1 //===-- ObjectFileMachO.cpp -------------------------------------*- C++ -*-===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 10 // C Includes 11 // C++ Includes 12 // Other libraries and framework includes 13 #include "llvm/ADT/StringRef.h" 14 15 // Project includes 16 #include "lldb/Core/ArchSpec.h" 17 #include "lldb/Core/DataBuffer.h" 18 #include "lldb/Core/Debugger.h" 19 #include "lldb/Core/Error.h" 20 #include "lldb/Core/FileSpecList.h" 21 #include "lldb/Core/Log.h" 22 #include "lldb/Core/Module.h" 23 #include "lldb/Core/ModuleSpec.h" 24 #include "lldb/Core/PluginManager.h" 25 #include "lldb/Core/RangeMap.h" 26 #include "lldb/Core/Section.h" 27 #include "lldb/Core/StreamFile.h" 28 #include "lldb/Core/StreamString.h" 29 #include "lldb/Core/Timer.h" 30 #include "lldb/Core/UUID.h" 31 #include "lldb/Host/Host.h" 32 #include "lldb/Host/FileSpec.h" 33 #include "lldb/Symbol/DWARFCallFrameInfo.h" 34 #include "lldb/Symbol/ObjectFile.h" 35 #include "lldb/Target/MemoryRegionInfo.h" 36 #include "lldb/Target/Platform.h" 37 #include "lldb/Target/Process.h" 38 #include "lldb/Target/SectionLoadList.h" 39 #include "lldb/Target/Target.h" 40 #include "lldb/Target/Thread.h" 41 #include "lldb/Target/ThreadList.h" 42 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 43 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 44 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 45 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 46 47 #include "lldb/Utility/SafeMachO.h" 48 49 #include "ObjectFileMachO.h" 50 51 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 52 // GetLLDBSharedCacheUUID() needs to call dlsym() 53 #include <dlfcn.h> 54 #endif 55 56 #ifndef __APPLE__ 57 #include "Utility/UuidCompatibility.h" 58 #endif 59 60 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 61 using namespace lldb; 62 using namespace lldb_private; 63 using namespace llvm::MachO; 64 65 // Some structure definitions needed for parsing the dyld shared cache files 66 // found on iOS devices. 67 68 struct lldb_copy_dyld_cache_header_v1 69 { 70 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 71 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 72 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 73 uint32_t imagesOffset; 74 uint32_t imagesCount; 75 uint64_t dyldBaseAddress; 76 uint64_t codeSignatureOffset; 77 uint64_t codeSignatureSize; 78 uint64_t slideInfoOffset; 79 uint64_t slideInfoSize; 80 uint64_t localSymbolsOffset; 81 uint64_t localSymbolsSize; 82 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 and later 83 }; 84 85 struct lldb_copy_dyld_cache_mapping_info 86 { 87 uint64_t address; 88 uint64_t size; 89 uint64_t fileOffset; 90 uint32_t maxProt; 91 uint32_t initProt; 92 }; 93 94 struct lldb_copy_dyld_cache_local_symbols_info 95 { 96 uint32_t nlistOffset; 97 uint32_t nlistCount; 98 uint32_t stringsOffset; 99 uint32_t stringsSize; 100 uint32_t entriesOffset; 101 uint32_t entriesCount; 102 }; 103 struct lldb_copy_dyld_cache_local_symbols_entry 104 { 105 uint32_t dylibOffset; 106 uint32_t nlistStartIndex; 107 uint32_t nlistCount; 108 }; 109 110 111 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 112 { 113 public: 114 RegisterContextDarwin_x86_64_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 115 RegisterContextDarwin_x86_64 (thread, 0) 116 { 117 SetRegisterDataFrom_LC_THREAD (data); 118 } 119 120 void 121 InvalidateAllRegisters() override 122 { 123 // Do nothing... registers are always valid... 124 } 125 126 void 127 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 128 { 129 lldb::offset_t offset = 0; 130 SetError (GPRRegSet, Read, -1); 131 SetError (FPURegSet, Read, -1); 132 SetError (EXCRegSet, Read, -1); 133 bool done = false; 134 135 while (!done) 136 { 137 int flavor = data.GetU32 (&offset); 138 if (flavor == 0) 139 done = true; 140 else 141 { 142 uint32_t i; 143 uint32_t count = data.GetU32 (&offset); 144 switch (flavor) 145 { 146 case GPRRegSet: 147 for (i=0; i<count; ++i) 148 (&gpr.rax)[i] = data.GetU64(&offset); 149 SetError (GPRRegSet, Read, 0); 150 done = true; 151 152 break; 153 case FPURegSet: 154 // TODO: fill in FPU regs.... 155 //SetError (FPURegSet, Read, -1); 156 done = true; 157 158 break; 159 case EXCRegSet: 160 exc.trapno = data.GetU32(&offset); 161 exc.err = data.GetU32(&offset); 162 exc.faultvaddr = data.GetU64(&offset); 163 SetError (EXCRegSet, Read, 0); 164 done = true; 165 break; 166 case 7: 167 case 8: 168 case 9: 169 // fancy flavors that encapsulate of the above 170 // flavors... 171 break; 172 173 default: 174 done = true; 175 break; 176 } 177 } 178 } 179 } 180 181 static size_t 182 WriteRegister (RegisterContext *reg_ctx, const char *name, const char *alt_name, size_t reg_byte_size, Stream &data) 183 { 184 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 185 if (reg_info == NULL) 186 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 187 if (reg_info) 188 { 189 lldb_private::RegisterValue reg_value; 190 if (reg_ctx->ReadRegister(reg_info, reg_value)) 191 { 192 if (reg_info->byte_size >= reg_byte_size) 193 data.Write(reg_value.GetBytes(), reg_byte_size); 194 else 195 { 196 data.Write(reg_value.GetBytes(), reg_info->byte_size); 197 for (size_t i=0, n = reg_byte_size - reg_info->byte_size; i<n; ++ i) 198 data.PutChar(0); 199 } 200 return reg_byte_size; 201 } 202 } 203 // Just write zeros if all else fails 204 for (size_t i=0; i<reg_byte_size; ++ i) 205 data.PutChar(0); 206 return reg_byte_size; 207 } 208 209 static bool 210 Create_LC_THREAD (Thread *thread, Stream &data) 211 { 212 RegisterContextSP reg_ctx_sp (thread->GetRegisterContext()); 213 if (reg_ctx_sp) 214 { 215 RegisterContext *reg_ctx = reg_ctx_sp.get(); 216 217 data.PutHex32 (GPRRegSet); // Flavor 218 data.PutHex32 (GPRWordCount); 219 WriteRegister (reg_ctx, "rax", NULL, 8, data); 220 WriteRegister (reg_ctx, "rbx", NULL, 8, data); 221 WriteRegister (reg_ctx, "rcx", NULL, 8, data); 222 WriteRegister (reg_ctx, "rdx", NULL, 8, data); 223 WriteRegister (reg_ctx, "rdi", NULL, 8, data); 224 WriteRegister (reg_ctx, "rsi", NULL, 8, data); 225 WriteRegister (reg_ctx, "rbp", NULL, 8, data); 226 WriteRegister (reg_ctx, "rsp", NULL, 8, data); 227 WriteRegister (reg_ctx, "r8", NULL, 8, data); 228 WriteRegister (reg_ctx, "r9", NULL, 8, data); 229 WriteRegister (reg_ctx, "r10", NULL, 8, data); 230 WriteRegister (reg_ctx, "r11", NULL, 8, data); 231 WriteRegister (reg_ctx, "r12", NULL, 8, data); 232 WriteRegister (reg_ctx, "r13", NULL, 8, data); 233 WriteRegister (reg_ctx, "r14", NULL, 8, data); 234 WriteRegister (reg_ctx, "r15", NULL, 8, data); 235 WriteRegister (reg_ctx, "rip", NULL, 8, data); 236 WriteRegister (reg_ctx, "rflags", NULL, 8, data); 237 WriteRegister (reg_ctx, "cs", NULL, 8, data); 238 WriteRegister (reg_ctx, "fs", NULL, 8, data); 239 WriteRegister (reg_ctx, "gs", NULL, 8, data); 240 241 // // Write out the FPU registers 242 // const size_t fpu_byte_size = sizeof(FPU); 243 // size_t bytes_written = 0; 244 // data.PutHex32 (FPURegSet); 245 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 246 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 247 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 248 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, data); // uint16_t fcw; // "fctrl" 249 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, data); // uint16_t fsw; // "fstat" 250 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, data); // uint8_t ftw; // "ftag" 251 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 252 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, data); // uint16_t fop; // "fop" 253 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, data); // uint32_t ip; // "fioff" 254 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, data); // uint16_t cs; // "fiseg" 255 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 256 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, data); // uint32_t dp; // "fooff" 257 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, data); // uint16_t ds; // "foseg" 258 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 259 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, data); // uint32_t mxcsr; 260 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 4, data);// uint32_t mxcsrmask; 261 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, sizeof(MMSReg), data); 262 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, sizeof(MMSReg), data); 263 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, sizeof(MMSReg), data); 264 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, sizeof(MMSReg), data); 265 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, sizeof(MMSReg), data); 266 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, sizeof(MMSReg), data); 267 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, sizeof(MMSReg), data); 268 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, sizeof(MMSReg), data); 269 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, sizeof(XMMReg), data); 270 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, sizeof(XMMReg), data); 271 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, sizeof(XMMReg), data); 272 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, sizeof(XMMReg), data); 273 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, sizeof(XMMReg), data); 274 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, sizeof(XMMReg), data); 275 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, sizeof(XMMReg), data); 276 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, sizeof(XMMReg), data); 277 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, sizeof(XMMReg), data); 278 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, sizeof(XMMReg), data); 279 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, sizeof(XMMReg), data); 280 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, sizeof(XMMReg), data); 281 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, sizeof(XMMReg), data); 282 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, sizeof(XMMReg), data); 283 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, sizeof(XMMReg), data); 284 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, sizeof(XMMReg), data); 285 // 286 // // Fill rest with zeros 287 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ i) 288 // data.PutChar(0); 289 290 // Write out the EXC registers 291 data.PutHex32 (EXCRegSet); 292 data.PutHex32 (EXCWordCount); 293 WriteRegister (reg_ctx, "trapno", NULL, 4, data); 294 WriteRegister (reg_ctx, "err", NULL, 4, data); 295 WriteRegister (reg_ctx, "faultvaddr", NULL, 8, data); 296 return true; 297 } 298 return false; 299 } 300 301 protected: 302 int 303 DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override 304 { 305 return 0; 306 } 307 308 int 309 DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override 310 { 311 return 0; 312 } 313 314 int 315 DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override 316 { 317 return 0; 318 } 319 320 int 321 DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override 322 { 323 return 0; 324 } 325 326 int 327 DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override 328 { 329 return 0; 330 } 331 332 int 333 DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override 334 { 335 return 0; 336 } 337 }; 338 339 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 340 { 341 public: 342 RegisterContextDarwin_i386_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 343 RegisterContextDarwin_i386 (thread, 0) 344 { 345 SetRegisterDataFrom_LC_THREAD (data); 346 } 347 348 void 349 InvalidateAllRegisters() override 350 { 351 // Do nothing... registers are always valid... 352 } 353 354 void 355 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 356 { 357 lldb::offset_t offset = 0; 358 SetError (GPRRegSet, Read, -1); 359 SetError (FPURegSet, Read, -1); 360 SetError (EXCRegSet, Read, -1); 361 bool done = false; 362 363 while (!done) 364 { 365 int flavor = data.GetU32 (&offset); 366 if (flavor == 0) 367 done = true; 368 else 369 { 370 uint32_t i; 371 uint32_t count = data.GetU32 (&offset); 372 switch (flavor) 373 { 374 case GPRRegSet: 375 for (i=0; i<count; ++i) 376 (&gpr.eax)[i] = data.GetU32(&offset); 377 SetError (GPRRegSet, Read, 0); 378 done = true; 379 380 break; 381 case FPURegSet: 382 // TODO: fill in FPU regs.... 383 //SetError (FPURegSet, Read, -1); 384 done = true; 385 386 break; 387 case EXCRegSet: 388 exc.trapno = data.GetU32(&offset); 389 exc.err = data.GetU32(&offset); 390 exc.faultvaddr = data.GetU32(&offset); 391 SetError (EXCRegSet, Read, 0); 392 done = true; 393 break; 394 case 7: 395 case 8: 396 case 9: 397 // fancy flavors that encapsulate of the above 398 // flavors... 399 break; 400 401 default: 402 done = true; 403 break; 404 } 405 } 406 } 407 } 408 409 static size_t 410 WriteRegister (RegisterContext *reg_ctx, const char *name, const char *alt_name, size_t reg_byte_size, Stream &data) 411 { 412 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 413 if (reg_info == NULL) 414 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 415 if (reg_info) 416 { 417 lldb_private::RegisterValue reg_value; 418 if (reg_ctx->ReadRegister(reg_info, reg_value)) 419 { 420 if (reg_info->byte_size >= reg_byte_size) 421 data.Write(reg_value.GetBytes(), reg_byte_size); 422 else 423 { 424 data.Write(reg_value.GetBytes(), reg_info->byte_size); 425 for (size_t i=0, n = reg_byte_size - reg_info->byte_size; i<n; ++ i) 426 data.PutChar(0); 427 } 428 return reg_byte_size; 429 } 430 } 431 // Just write zeros if all else fails 432 for (size_t i=0; i<reg_byte_size; ++ i) 433 data.PutChar(0); 434 return reg_byte_size; 435 } 436 437 static bool 438 Create_LC_THREAD (Thread *thread, Stream &data) 439 { 440 RegisterContextSP reg_ctx_sp (thread->GetRegisterContext()); 441 if (reg_ctx_sp) 442 { 443 RegisterContext *reg_ctx = reg_ctx_sp.get(); 444 445 data.PutHex32 (GPRRegSet); // Flavor 446 data.PutHex32 (GPRWordCount); 447 WriteRegister (reg_ctx, "eax", NULL, 4, data); 448 WriteRegister (reg_ctx, "ebx", NULL, 4, data); 449 WriteRegister (reg_ctx, "ecx", NULL, 4, data); 450 WriteRegister (reg_ctx, "edx", NULL, 4, data); 451 WriteRegister (reg_ctx, "edi", NULL, 4, data); 452 WriteRegister (reg_ctx, "esi", NULL, 4, data); 453 WriteRegister (reg_ctx, "ebp", NULL, 4, data); 454 WriteRegister (reg_ctx, "esp", NULL, 4, data); 455 WriteRegister (reg_ctx, "ss", NULL, 4, data); 456 WriteRegister (reg_ctx, "eflags", NULL, 4, data); 457 WriteRegister (reg_ctx, "eip", NULL, 4, data); 458 WriteRegister (reg_ctx, "cs", NULL, 4, data); 459 WriteRegister (reg_ctx, "ds", NULL, 4, data); 460 WriteRegister (reg_ctx, "es", NULL, 4, data); 461 WriteRegister (reg_ctx, "fs", NULL, 4, data); 462 WriteRegister (reg_ctx, "gs", NULL, 4, data); 463 464 // Write out the EXC registers 465 data.PutHex32 (EXCRegSet); 466 data.PutHex32 (EXCWordCount); 467 WriteRegister (reg_ctx, "trapno", NULL, 4, data); 468 WriteRegister (reg_ctx, "err", NULL, 4, data); 469 WriteRegister (reg_ctx, "faultvaddr", NULL, 4, data); 470 return true; 471 } 472 return false; 473 } 474 475 protected: 476 int 477 DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override 478 { 479 return 0; 480 } 481 482 int 483 DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override 484 { 485 return 0; 486 } 487 488 int 489 DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override 490 { 491 return 0; 492 } 493 494 int 495 DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override 496 { 497 return 0; 498 } 499 500 int 501 DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override 502 { 503 return 0; 504 } 505 506 int 507 DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override 508 { 509 return 0; 510 } 511 }; 512 513 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm 514 { 515 public: 516 RegisterContextDarwin_arm_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 517 RegisterContextDarwin_arm (thread, 0) 518 { 519 SetRegisterDataFrom_LC_THREAD (data); 520 } 521 522 void 523 InvalidateAllRegisters() override 524 { 525 // Do nothing... registers are always valid... 526 } 527 528 void 529 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 530 { 531 lldb::offset_t offset = 0; 532 SetError (GPRRegSet, Read, -1); 533 SetError (FPURegSet, Read, -1); 534 SetError (EXCRegSet, Read, -1); 535 bool done = false; 536 537 while (!done) 538 { 539 int flavor = data.GetU32 (&offset); 540 uint32_t count = data.GetU32 (&offset); 541 lldb::offset_t next_thread_state = offset + (count * 4); 542 switch (flavor) 543 { 544 case GPRAltRegSet: 545 case GPRRegSet: 546 for (uint32_t i=0; i<count; ++i) 547 { 548 gpr.r[i] = data.GetU32(&offset); 549 } 550 551 // Note that gpr.cpsr is also copied by the above loop; this loop technically extends 552 // one element past the end of the gpr.r[] array. 553 554 SetError (GPRRegSet, Read, 0); 555 offset = next_thread_state; 556 break; 557 558 case FPURegSet: 559 { 560 uint8_t *fpu_reg_buf = (uint8_t*) &fpu.floats.s[0]; 561 const int fpu_reg_buf_size = sizeof (fpu.floats); 562 if (data.ExtractBytes (offset, fpu_reg_buf_size, eByteOrderLittle, fpu_reg_buf) == fpu_reg_buf_size) 563 { 564 offset += fpu_reg_buf_size; 565 fpu.fpscr = data.GetU32(&offset); 566 SetError (FPURegSet, Read, 0); 567 } 568 else 569 { 570 done = true; 571 } 572 } 573 offset = next_thread_state; 574 break; 575 576 case EXCRegSet: 577 if (count == 3) 578 { 579 exc.exception = data.GetU32(&offset); 580 exc.fsr = data.GetU32(&offset); 581 exc.far = data.GetU32(&offset); 582 SetError (EXCRegSet, Read, 0); 583 } 584 done = true; 585 offset = next_thread_state; 586 break; 587 588 // Unknown register set flavor, stop trying to parse. 589 default: 590 done = true; 591 } 592 } 593 } 594 595 static size_t 596 WriteRegister (RegisterContext *reg_ctx, const char *name, const char *alt_name, size_t reg_byte_size, Stream &data) 597 { 598 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 599 if (reg_info == NULL) 600 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 601 if (reg_info) 602 { 603 lldb_private::RegisterValue reg_value; 604 if (reg_ctx->ReadRegister(reg_info, reg_value)) 605 { 606 if (reg_info->byte_size >= reg_byte_size) 607 data.Write(reg_value.GetBytes(), reg_byte_size); 608 else 609 { 610 data.Write(reg_value.GetBytes(), reg_info->byte_size); 611 for (size_t i=0, n = reg_byte_size - reg_info->byte_size; i<n; ++ i) 612 data.PutChar(0); 613 } 614 return reg_byte_size; 615 } 616 } 617 // Just write zeros if all else fails 618 for (size_t i=0; i<reg_byte_size; ++ i) 619 data.PutChar(0); 620 return reg_byte_size; 621 } 622 623 static bool 624 Create_LC_THREAD (Thread *thread, Stream &data) 625 { 626 RegisterContextSP reg_ctx_sp (thread->GetRegisterContext()); 627 if (reg_ctx_sp) 628 { 629 RegisterContext *reg_ctx = reg_ctx_sp.get(); 630 631 data.PutHex32 (GPRRegSet); // Flavor 632 data.PutHex32 (GPRWordCount); 633 WriteRegister (reg_ctx, "r0", NULL, 4, data); 634 WriteRegister (reg_ctx, "r1", NULL, 4, data); 635 WriteRegister (reg_ctx, "r2", NULL, 4, data); 636 WriteRegister (reg_ctx, "r3", NULL, 4, data); 637 WriteRegister (reg_ctx, "r4", NULL, 4, data); 638 WriteRegister (reg_ctx, "r5", NULL, 4, data); 639 WriteRegister (reg_ctx, "r6", NULL, 4, data); 640 WriteRegister (reg_ctx, "r7", NULL, 4, data); 641 WriteRegister (reg_ctx, "r8", NULL, 4, data); 642 WriteRegister (reg_ctx, "r9", NULL, 4, data); 643 WriteRegister (reg_ctx, "r10", NULL, 4, data); 644 WriteRegister (reg_ctx, "r11", NULL, 4, data); 645 WriteRegister (reg_ctx, "r12", NULL, 4, data); 646 WriteRegister (reg_ctx, "sp", NULL, 4, data); 647 WriteRegister (reg_ctx, "lr", NULL, 4, data); 648 WriteRegister (reg_ctx, "pc", NULL, 4, data); 649 WriteRegister (reg_ctx, "cpsr", NULL, 4, data); 650 651 // Write out the EXC registers 652 // data.PutHex32 (EXCRegSet); 653 // data.PutHex32 (EXCWordCount); 654 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 655 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 656 // WriteRegister (reg_ctx, "far", NULL, 4, data); 657 return true; 658 } 659 return false; 660 } 661 662 protected: 663 int 664 DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override 665 { 666 return -1; 667 } 668 669 int 670 DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override 671 { 672 return -1; 673 } 674 675 int 676 DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override 677 { 678 return -1; 679 } 680 681 int 682 DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override 683 { 684 return -1; 685 } 686 687 int 688 DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override 689 { 690 return 0; 691 } 692 693 int 694 DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override 695 { 696 return 0; 697 } 698 699 int 700 DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override 701 { 702 return 0; 703 } 704 705 int 706 DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override 707 { 708 return -1; 709 } 710 }; 711 712 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 713 { 714 public: 715 RegisterContextDarwin_arm64_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 716 RegisterContextDarwin_arm64 (thread, 0) 717 { 718 SetRegisterDataFrom_LC_THREAD (data); 719 } 720 721 void 722 InvalidateAllRegisters() override 723 { 724 // Do nothing... registers are always valid... 725 } 726 727 void 728 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 729 { 730 lldb::offset_t offset = 0; 731 SetError (GPRRegSet, Read, -1); 732 SetError (FPURegSet, Read, -1); 733 SetError (EXCRegSet, Read, -1); 734 bool done = false; 735 while (!done) 736 { 737 int flavor = data.GetU32 (&offset); 738 uint32_t count = data.GetU32 (&offset); 739 lldb::offset_t next_thread_state = offset + (count * 4); 740 switch (flavor) 741 { 742 case GPRRegSet: 743 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 32-bit register) 744 if (count >= (33 * 2) + 1) 745 { 746 for (uint32_t i=0; i<33; ++i) 747 gpr.x[i] = data.GetU64(&offset); 748 gpr.cpsr = data.GetU32(&offset); 749 SetError (GPRRegSet, Read, 0); 750 } 751 offset = next_thread_state; 752 break; 753 case FPURegSet: 754 { 755 uint8_t *fpu_reg_buf = (uint8_t*) &fpu.v[0]; 756 const int fpu_reg_buf_size = sizeof (fpu); 757 if (fpu_reg_buf_size == count 758 && data.ExtractBytes (offset, fpu_reg_buf_size, eByteOrderLittle, fpu_reg_buf) == fpu_reg_buf_size) 759 { 760 SetError (FPURegSet, Read, 0); 761 } 762 else 763 { 764 done = true; 765 } 766 } 767 offset = next_thread_state; 768 break; 769 case EXCRegSet: 770 if (count == 4) 771 { 772 exc.far = data.GetU64(&offset); 773 exc.esr = data.GetU32(&offset); 774 exc.exception = data.GetU32(&offset); 775 SetError (EXCRegSet, Read, 0); 776 } 777 offset = next_thread_state; 778 break; 779 default: 780 done = true; 781 break; 782 } 783 } 784 } 785 786 static size_t 787 WriteRegister (RegisterContext *reg_ctx, const char *name, const char *alt_name, size_t reg_byte_size, Stream &data) 788 { 789 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 790 if (reg_info == NULL) 791 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 792 if (reg_info) 793 { 794 lldb_private::RegisterValue reg_value; 795 if (reg_ctx->ReadRegister(reg_info, reg_value)) 796 { 797 if (reg_info->byte_size >= reg_byte_size) 798 data.Write(reg_value.GetBytes(), reg_byte_size); 799 else 800 { 801 data.Write(reg_value.GetBytes(), reg_info->byte_size); 802 for (size_t i=0, n = reg_byte_size - reg_info->byte_size; i<n; ++ i) 803 data.PutChar(0); 804 } 805 return reg_byte_size; 806 } 807 } 808 // Just write zeros if all else fails 809 for (size_t i=0; i<reg_byte_size; ++ i) 810 data.PutChar(0); 811 return reg_byte_size; 812 } 813 814 static bool 815 Create_LC_THREAD (Thread *thread, Stream &data) 816 { 817 RegisterContextSP reg_ctx_sp (thread->GetRegisterContext()); 818 if (reg_ctx_sp) 819 { 820 RegisterContext *reg_ctx = reg_ctx_sp.get(); 821 822 data.PutHex32 (GPRRegSet); // Flavor 823 data.PutHex32 (GPRWordCount); 824 WriteRegister (reg_ctx, "x0", NULL, 8, data); 825 WriteRegister (reg_ctx, "x1", NULL, 8, data); 826 WriteRegister (reg_ctx, "x2", NULL, 8, data); 827 WriteRegister (reg_ctx, "x3", NULL, 8, data); 828 WriteRegister (reg_ctx, "x4", NULL, 8, data); 829 WriteRegister (reg_ctx, "x5", NULL, 8, data); 830 WriteRegister (reg_ctx, "x6", NULL, 8, data); 831 WriteRegister (reg_ctx, "x7", NULL, 8, data); 832 WriteRegister (reg_ctx, "x8", NULL, 8, data); 833 WriteRegister (reg_ctx, "x9", NULL, 8, data); 834 WriteRegister (reg_ctx, "x10", NULL, 8, data); 835 WriteRegister (reg_ctx, "x11", NULL, 8, data); 836 WriteRegister (reg_ctx, "x12", NULL, 8, data); 837 WriteRegister (reg_ctx, "x13", NULL, 8, data); 838 WriteRegister (reg_ctx, "x14", NULL, 8, data); 839 WriteRegister (reg_ctx, "x15", NULL, 8, data); 840 WriteRegister (reg_ctx, "x16", NULL, 8, data); 841 WriteRegister (reg_ctx, "x17", NULL, 8, data); 842 WriteRegister (reg_ctx, "x18", NULL, 8, data); 843 WriteRegister (reg_ctx, "x19", NULL, 8, data); 844 WriteRegister (reg_ctx, "x20", NULL, 8, data); 845 WriteRegister (reg_ctx, "x21", NULL, 8, data); 846 WriteRegister (reg_ctx, "x22", NULL, 8, data); 847 WriteRegister (reg_ctx, "x23", NULL, 8, data); 848 WriteRegister (reg_ctx, "x24", NULL, 8, data); 849 WriteRegister (reg_ctx, "x25", NULL, 8, data); 850 WriteRegister (reg_ctx, "x26", NULL, 8, data); 851 WriteRegister (reg_ctx, "x27", NULL, 8, data); 852 WriteRegister (reg_ctx, "x28", NULL, 8, data); 853 WriteRegister (reg_ctx, "fp", NULL, 8, data); 854 WriteRegister (reg_ctx, "lr", NULL, 8, data); 855 WriteRegister (reg_ctx, "sp", NULL, 8, data); 856 WriteRegister (reg_ctx, "pc", NULL, 8, data); 857 WriteRegister (reg_ctx, "cpsr", NULL, 4, data); 858 859 // Write out the EXC registers 860 // data.PutHex32 (EXCRegSet); 861 // data.PutHex32 (EXCWordCount); 862 // WriteRegister (reg_ctx, "far", NULL, 8, data); 863 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 864 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 865 return true; 866 } 867 return false; 868 } 869 870 protected: 871 int 872 DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override 873 { 874 return -1; 875 } 876 877 int 878 DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override 879 { 880 return -1; 881 } 882 883 int 884 DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override 885 { 886 return -1; 887 } 888 889 int 890 DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override 891 { 892 return -1; 893 } 894 895 int 896 DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override 897 { 898 return 0; 899 } 900 901 int 902 DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override 903 { 904 return 0; 905 } 906 907 int 908 DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override 909 { 910 return 0; 911 } 912 913 int 914 DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override 915 { 916 return -1; 917 } 918 }; 919 920 static uint32_t 921 MachHeaderSizeFromMagic(uint32_t magic) 922 { 923 switch (magic) 924 { 925 case MH_MAGIC: 926 case MH_CIGAM: 927 return sizeof(struct mach_header); 928 929 case MH_MAGIC_64: 930 case MH_CIGAM_64: 931 return sizeof(struct mach_header_64); 932 break; 933 934 default: 935 break; 936 } 937 return 0; 938 } 939 940 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 941 942 void 943 ObjectFileMachO::Initialize() 944 { 945 PluginManager::RegisterPlugin (GetPluginNameStatic(), 946 GetPluginDescriptionStatic(), 947 CreateInstance, 948 CreateMemoryInstance, 949 GetModuleSpecifications, 950 SaveCore); 951 } 952 953 void 954 ObjectFileMachO::Terminate() 955 { 956 PluginManager::UnregisterPlugin (CreateInstance); 957 } 958 959 lldb_private::ConstString 960 ObjectFileMachO::GetPluginNameStatic() 961 { 962 static ConstString g_name("mach-o"); 963 return g_name; 964 } 965 966 const char * 967 ObjectFileMachO::GetPluginDescriptionStatic() 968 { 969 return "Mach-o object file reader (32 and 64 bit)"; 970 } 971 972 ObjectFile * 973 ObjectFileMachO::CreateInstance (const lldb::ModuleSP &module_sp, 974 DataBufferSP& data_sp, 975 lldb::offset_t data_offset, 976 const FileSpec* file, 977 lldb::offset_t file_offset, 978 lldb::offset_t length) 979 { 980 if (!data_sp) 981 { 982 data_sp = file->MemoryMapFileContentsIfLocal(file_offset, length); 983 data_offset = 0; 984 } 985 986 if (ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 987 { 988 // Update the data to contain the entire file if it doesn't already 989 if (data_sp->GetByteSize() < length) 990 { 991 data_sp = file->MemoryMapFileContentsIfLocal(file_offset, length); 992 data_offset = 0; 993 } 994 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, data_offset, file, file_offset, length)); 995 if (objfile_ap.get() && objfile_ap->ParseHeader()) 996 return objfile_ap.release(); 997 } 998 return NULL; 999 } 1000 1001 ObjectFile * 1002 ObjectFileMachO::CreateMemoryInstance (const lldb::ModuleSP &module_sp, 1003 DataBufferSP& data_sp, 1004 const ProcessSP &process_sp, 1005 lldb::addr_t header_addr) 1006 { 1007 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) 1008 { 1009 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, process_sp, header_addr)); 1010 if (objfile_ap.get() && objfile_ap->ParseHeader()) 1011 return objfile_ap.release(); 1012 } 1013 return NULL; 1014 } 1015 1016 size_t 1017 ObjectFileMachO::GetModuleSpecifications (const lldb_private::FileSpec& file, 1018 lldb::DataBufferSP& data_sp, 1019 lldb::offset_t data_offset, 1020 lldb::offset_t file_offset, 1021 lldb::offset_t length, 1022 lldb_private::ModuleSpecList &specs) 1023 { 1024 const size_t initial_count = specs.GetSize(); 1025 1026 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) 1027 { 1028 DataExtractor data; 1029 data.SetData(data_sp); 1030 llvm::MachO::mach_header header; 1031 if (ParseHeader (data, &data_offset, header)) 1032 { 1033 size_t header_and_load_cmds = header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 1034 if (header_and_load_cmds >= data_sp->GetByteSize()) 1035 { 1036 data_sp = file.ReadFileContents(file_offset, header_and_load_cmds); 1037 data.SetData(data_sp); 1038 data_offset = MachHeaderSizeFromMagic(header.magic); 1039 } 1040 if (data_sp) 1041 { 1042 ModuleSpec spec; 1043 spec.GetFileSpec() = file; 1044 spec.SetObjectOffset(file_offset); 1045 spec.SetObjectSize(length); 1046 1047 if (GetArchitecture (header, data, data_offset, spec.GetArchitecture())) 1048 { 1049 if (spec.GetArchitecture().IsValid()) 1050 { 1051 GetUUID (header, data, data_offset, spec.GetUUID()); 1052 specs.Append(spec); 1053 } 1054 } 1055 } 1056 } 1057 } 1058 return specs.GetSize() - initial_count; 1059 } 1060 1061 const ConstString & 1062 ObjectFileMachO::GetSegmentNameTEXT() 1063 { 1064 static ConstString g_segment_name_TEXT ("__TEXT"); 1065 return g_segment_name_TEXT; 1066 } 1067 1068 const ConstString & 1069 ObjectFileMachO::GetSegmentNameDATA() 1070 { 1071 static ConstString g_segment_name_DATA ("__DATA"); 1072 return g_segment_name_DATA; 1073 } 1074 1075 const ConstString & 1076 ObjectFileMachO::GetSegmentNameDATA_DIRTY() 1077 { 1078 static ConstString g_segment_name ("__DATA_DIRTY"); 1079 return g_segment_name; 1080 } 1081 1082 const ConstString & 1083 ObjectFileMachO::GetSegmentNameDATA_CONST() 1084 { 1085 static ConstString g_segment_name ("__DATA_CONST"); 1086 return g_segment_name; 1087 } 1088 1089 const ConstString & 1090 ObjectFileMachO::GetSegmentNameOBJC() 1091 { 1092 static ConstString g_segment_name_OBJC ("__OBJC"); 1093 return g_segment_name_OBJC; 1094 } 1095 1096 const ConstString & 1097 ObjectFileMachO::GetSegmentNameLINKEDIT() 1098 { 1099 static ConstString g_section_name_LINKEDIT ("__LINKEDIT"); 1100 return g_section_name_LINKEDIT; 1101 } 1102 1103 const ConstString & 1104 ObjectFileMachO::GetSectionNameEHFrame() 1105 { 1106 static ConstString g_section_name_eh_frame ("__eh_frame"); 1107 return g_section_name_eh_frame; 1108 } 1109 1110 bool 1111 ObjectFileMachO::MagicBytesMatch (DataBufferSP& data_sp, 1112 lldb::addr_t data_offset, 1113 lldb::addr_t data_length) 1114 { 1115 DataExtractor data; 1116 data.SetData (data_sp, data_offset, data_length); 1117 lldb::offset_t offset = 0; 1118 uint32_t magic = data.GetU32(&offset); 1119 return MachHeaderSizeFromMagic(magic) != 0; 1120 } 1121 1122 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 1123 DataBufferSP& data_sp, 1124 lldb::offset_t data_offset, 1125 const FileSpec* file, 1126 lldb::offset_t file_offset, 1127 lldb::offset_t length) : 1128 ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 1129 m_mach_segments(), 1130 m_mach_sections(), 1131 m_entry_point_address(), 1132 m_thread_context_offsets(), 1133 m_thread_context_offsets_valid(false), 1134 m_reexported_dylibs (), 1135 m_allow_assembly_emulation_unwind_plans (true) 1136 { 1137 ::memset (&m_header, 0, sizeof(m_header)); 1138 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 1139 } 1140 1141 ObjectFileMachO::ObjectFileMachO (const lldb::ModuleSP &module_sp, 1142 lldb::DataBufferSP& header_data_sp, 1143 const lldb::ProcessSP &process_sp, 1144 lldb::addr_t header_addr) : 1145 ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 1146 m_mach_segments(), 1147 m_mach_sections(), 1148 m_entry_point_address(), 1149 m_thread_context_offsets(), 1150 m_thread_context_offsets_valid(false), 1151 m_reexported_dylibs (), 1152 m_allow_assembly_emulation_unwind_plans (true) 1153 { 1154 ::memset (&m_header, 0, sizeof(m_header)); 1155 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 1156 } 1157 1158 bool 1159 ObjectFileMachO::ParseHeader (DataExtractor &data, 1160 lldb::offset_t *data_offset_ptr, 1161 llvm::MachO::mach_header &header) 1162 { 1163 data.SetByteOrder (endian::InlHostByteOrder()); 1164 // Leave magic in the original byte order 1165 header.magic = data.GetU32(data_offset_ptr); 1166 bool can_parse = false; 1167 bool is_64_bit = false; 1168 switch (header.magic) 1169 { 1170 case MH_MAGIC: 1171 data.SetByteOrder (endian::InlHostByteOrder()); 1172 data.SetAddressByteSize(4); 1173 can_parse = true; 1174 break; 1175 1176 case MH_MAGIC_64: 1177 data.SetByteOrder (endian::InlHostByteOrder()); 1178 data.SetAddressByteSize(8); 1179 can_parse = true; 1180 is_64_bit = true; 1181 break; 1182 1183 case MH_CIGAM: 1184 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 1185 data.SetAddressByteSize(4); 1186 can_parse = true; 1187 break; 1188 1189 case MH_CIGAM_64: 1190 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 1191 data.SetAddressByteSize(8); 1192 is_64_bit = true; 1193 can_parse = true; 1194 break; 1195 1196 default: 1197 break; 1198 } 1199 1200 if (can_parse) 1201 { 1202 data.GetU32(data_offset_ptr, &header.cputype, 6); 1203 if (is_64_bit) 1204 *data_offset_ptr += 4; 1205 return true; 1206 } 1207 else 1208 { 1209 memset(&header, 0, sizeof(header)); 1210 } 1211 return false; 1212 } 1213 1214 bool 1215 ObjectFileMachO::ParseHeader () 1216 { 1217 ModuleSP module_sp(GetModule()); 1218 if (module_sp) 1219 { 1220 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1221 bool can_parse = false; 1222 lldb::offset_t offset = 0; 1223 m_data.SetByteOrder (endian::InlHostByteOrder()); 1224 // Leave magic in the original byte order 1225 m_header.magic = m_data.GetU32(&offset); 1226 switch (m_header.magic) 1227 { 1228 case MH_MAGIC: 1229 m_data.SetByteOrder (endian::InlHostByteOrder()); 1230 m_data.SetAddressByteSize(4); 1231 can_parse = true; 1232 break; 1233 1234 case MH_MAGIC_64: 1235 m_data.SetByteOrder (endian::InlHostByteOrder()); 1236 m_data.SetAddressByteSize(8); 1237 can_parse = true; 1238 break; 1239 1240 case MH_CIGAM: 1241 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 1242 m_data.SetAddressByteSize(4); 1243 can_parse = true; 1244 break; 1245 1246 case MH_CIGAM_64: 1247 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 1248 m_data.SetAddressByteSize(8); 1249 can_parse = true; 1250 break; 1251 1252 default: 1253 break; 1254 } 1255 1256 if (can_parse) 1257 { 1258 m_data.GetU32(&offset, &m_header.cputype, 6); 1259 1260 1261 ArchSpec mach_arch; 1262 1263 if (GetArchitecture (mach_arch)) 1264 { 1265 // Check if the module has a required architecture 1266 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1267 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1268 return false; 1269 1270 if (SetModulesArchitecture (mach_arch)) 1271 { 1272 const size_t header_and_lc_size = m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1273 if (m_data.GetByteSize() < header_and_lc_size) 1274 { 1275 DataBufferSP data_sp; 1276 ProcessSP process_sp (m_process_wp.lock()); 1277 if (process_sp) 1278 { 1279 data_sp = ReadMemory (process_sp, m_memory_addr, header_and_lc_size); 1280 } 1281 else 1282 { 1283 // Read in all only the load command data from the file on disk 1284 data_sp = m_file.ReadFileContents(m_file_offset, header_and_lc_size); 1285 if (data_sp->GetByteSize() != header_and_lc_size) 1286 return false; 1287 } 1288 if (data_sp) 1289 m_data.SetData (data_sp); 1290 } 1291 } 1292 return true; 1293 } 1294 } 1295 else 1296 { 1297 memset(&m_header, 0, sizeof(struct mach_header)); 1298 } 1299 } 1300 return false; 1301 } 1302 1303 ByteOrder 1304 ObjectFileMachO::GetByteOrder () const 1305 { 1306 return m_data.GetByteOrder (); 1307 } 1308 1309 bool 1310 ObjectFileMachO::IsExecutable() const 1311 { 1312 return m_header.filetype == MH_EXECUTE; 1313 } 1314 1315 uint32_t 1316 ObjectFileMachO::GetAddressByteSize () const 1317 { 1318 return m_data.GetAddressByteSize (); 1319 } 1320 1321 AddressClass 1322 ObjectFileMachO::GetAddressClass (lldb::addr_t file_addr) 1323 { 1324 Symtab *symtab = GetSymtab(); 1325 if (symtab) 1326 { 1327 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1328 if (symbol) 1329 { 1330 if (symbol->ValueIsAddress()) 1331 { 1332 SectionSP section_sp (symbol->GetAddressRef().GetSection()); 1333 if (section_sp) 1334 { 1335 const lldb::SectionType section_type = section_sp->GetType(); 1336 switch (section_type) 1337 { 1338 case eSectionTypeInvalid: 1339 return eAddressClassUnknown; 1340 1341 case eSectionTypeCode: 1342 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) 1343 { 1344 // For ARM we have a bit in the n_desc field of the symbol 1345 // that tells us ARM/Thumb which is bit 0x0008. 1346 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1347 return eAddressClassCodeAlternateISA; 1348 } 1349 return eAddressClassCode; 1350 1351 case eSectionTypeContainer: 1352 return eAddressClassUnknown; 1353 1354 case eSectionTypeData: 1355 case eSectionTypeDataCString: 1356 case eSectionTypeDataCStringPointers: 1357 case eSectionTypeDataSymbolAddress: 1358 case eSectionTypeData4: 1359 case eSectionTypeData8: 1360 case eSectionTypeData16: 1361 case eSectionTypeDataPointers: 1362 case eSectionTypeZeroFill: 1363 case eSectionTypeDataObjCMessageRefs: 1364 case eSectionTypeDataObjCCFStrings: 1365 case eSectionTypeGoSymtab: 1366 return eAddressClassData; 1367 1368 case eSectionTypeDebug: 1369 case eSectionTypeDWARFDebugAbbrev: 1370 case eSectionTypeDWARFDebugAddr: 1371 case eSectionTypeDWARFDebugAranges: 1372 case eSectionTypeDWARFDebugFrame: 1373 case eSectionTypeDWARFDebugInfo: 1374 case eSectionTypeDWARFDebugLine: 1375 case eSectionTypeDWARFDebugLoc: 1376 case eSectionTypeDWARFDebugMacInfo: 1377 case eSectionTypeDWARFDebugMacro: 1378 case eSectionTypeDWARFDebugPubNames: 1379 case eSectionTypeDWARFDebugPubTypes: 1380 case eSectionTypeDWARFDebugRanges: 1381 case eSectionTypeDWARFDebugStr: 1382 case eSectionTypeDWARFDebugStrOffsets: 1383 case eSectionTypeDWARFAppleNames: 1384 case eSectionTypeDWARFAppleTypes: 1385 case eSectionTypeDWARFAppleNamespaces: 1386 case eSectionTypeDWARFAppleObjC: 1387 return eAddressClassDebug; 1388 1389 case eSectionTypeEHFrame: 1390 case eSectionTypeARMexidx: 1391 case eSectionTypeARMextab: 1392 case eSectionTypeCompactUnwind: 1393 return eAddressClassRuntime; 1394 1395 case eSectionTypeAbsoluteAddress: 1396 case eSectionTypeELFSymbolTable: 1397 case eSectionTypeELFDynamicSymbols: 1398 case eSectionTypeELFRelocationEntries: 1399 case eSectionTypeELFDynamicLinkInfo: 1400 case eSectionTypeOther: 1401 return eAddressClassUnknown; 1402 } 1403 } 1404 } 1405 1406 const SymbolType symbol_type = symbol->GetType(); 1407 switch (symbol_type) 1408 { 1409 case eSymbolTypeAny: return eAddressClassUnknown; 1410 case eSymbolTypeAbsolute: return eAddressClassUnknown; 1411 1412 case eSymbolTypeCode: 1413 case eSymbolTypeTrampoline: 1414 case eSymbolTypeResolver: 1415 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) 1416 { 1417 // For ARM we have a bit in the n_desc field of the symbol 1418 // that tells us ARM/Thumb which is bit 0x0008. 1419 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1420 return eAddressClassCodeAlternateISA; 1421 } 1422 return eAddressClassCode; 1423 1424 case eSymbolTypeData: return eAddressClassData; 1425 case eSymbolTypeRuntime: return eAddressClassRuntime; 1426 case eSymbolTypeException: return eAddressClassRuntime; 1427 case eSymbolTypeSourceFile: return eAddressClassDebug; 1428 case eSymbolTypeHeaderFile: return eAddressClassDebug; 1429 case eSymbolTypeObjectFile: return eAddressClassDebug; 1430 case eSymbolTypeCommonBlock: return eAddressClassDebug; 1431 case eSymbolTypeBlock: return eAddressClassDebug; 1432 case eSymbolTypeLocal: return eAddressClassData; 1433 case eSymbolTypeParam: return eAddressClassData; 1434 case eSymbolTypeVariable: return eAddressClassData; 1435 case eSymbolTypeVariableType: return eAddressClassDebug; 1436 case eSymbolTypeLineEntry: return eAddressClassDebug; 1437 case eSymbolTypeLineHeader: return eAddressClassDebug; 1438 case eSymbolTypeScopeBegin: return eAddressClassDebug; 1439 case eSymbolTypeScopeEnd: return eAddressClassDebug; 1440 case eSymbolTypeAdditional: return eAddressClassUnknown; 1441 case eSymbolTypeCompiler: return eAddressClassDebug; 1442 case eSymbolTypeInstrumentation:return eAddressClassDebug; 1443 case eSymbolTypeUndefined: return eAddressClassUnknown; 1444 case eSymbolTypeObjCClass: return eAddressClassRuntime; 1445 case eSymbolTypeObjCMetaClass: return eAddressClassRuntime; 1446 case eSymbolTypeObjCIVar: return eAddressClassRuntime; 1447 case eSymbolTypeReExported: return eAddressClassRuntime; 1448 } 1449 } 1450 } 1451 return eAddressClassUnknown; 1452 } 1453 1454 Symtab * 1455 ObjectFileMachO::GetSymtab() 1456 { 1457 ModuleSP module_sp(GetModule()); 1458 if (module_sp) 1459 { 1460 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1461 if (m_symtab_ap.get() == NULL) 1462 { 1463 m_symtab_ap.reset(new Symtab(this)); 1464 std::lock_guard<std::recursive_mutex> symtab_guard(m_symtab_ap->GetMutex()); 1465 ParseSymtab (); 1466 m_symtab_ap->Finalize (); 1467 } 1468 } 1469 return m_symtab_ap.get(); 1470 } 1471 1472 bool 1473 ObjectFileMachO::IsStripped () 1474 { 1475 if (m_dysymtab.cmd == 0) 1476 { 1477 ModuleSP module_sp(GetModule()); 1478 if (module_sp) 1479 { 1480 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1481 for (uint32_t i=0; i<m_header.ncmds; ++i) 1482 { 1483 const lldb::offset_t load_cmd_offset = offset; 1484 1485 load_command lc; 1486 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1487 break; 1488 if (lc.cmd == LC_DYSYMTAB) 1489 { 1490 m_dysymtab.cmd = lc.cmd; 1491 m_dysymtab.cmdsize = lc.cmdsize; 1492 if (m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == NULL) 1493 { 1494 // Clear m_dysymtab if we were unable to read all items from the load command 1495 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 1496 } 1497 } 1498 offset = load_cmd_offset + lc.cmdsize; 1499 } 1500 } 1501 } 1502 if (m_dysymtab.cmd) 1503 return m_dysymtab.nlocalsym <= 1; 1504 return false; 1505 } 1506 1507 void 1508 ObjectFileMachO::CreateSections (SectionList &unified_section_list) 1509 { 1510 if (!m_sections_ap.get()) 1511 { 1512 m_sections_ap.reset(new SectionList()); 1513 1514 const bool is_dsym = (m_header.filetype == MH_DSYM); 1515 lldb::user_id_t segID = 0; 1516 lldb::user_id_t sectID = 0; 1517 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1518 uint32_t i; 1519 const bool is_core = GetType() == eTypeCoreFile; 1520 //bool dump_sections = false; 1521 ModuleSP module_sp (GetModule()); 1522 // First look up any LC_ENCRYPTION_INFO load commands 1523 typedef RangeArray<uint32_t, uint32_t, 8> EncryptedFileRanges; 1524 EncryptedFileRanges encrypted_file_ranges; 1525 encryption_info_command encryption_cmd; 1526 for (i=0; i<m_header.ncmds; ++i) 1527 { 1528 const lldb::offset_t load_cmd_offset = offset; 1529 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1530 break; 1531 1532 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for 1533 // the 3 fields we care about, so treat them the same. 1534 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) 1535 { 1536 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) 1537 { 1538 if (encryption_cmd.cryptid != 0) 1539 { 1540 EncryptedFileRanges::Entry entry; 1541 entry.SetRangeBase(encryption_cmd.cryptoff); 1542 entry.SetByteSize(encryption_cmd.cryptsize); 1543 encrypted_file_ranges.Append(entry); 1544 } 1545 } 1546 } 1547 offset = load_cmd_offset + encryption_cmd.cmdsize; 1548 } 1549 1550 bool section_file_addresses_changed = false; 1551 1552 offset = MachHeaderSizeFromMagic(m_header.magic); 1553 1554 struct segment_command_64 load_cmd; 1555 for (i=0; i<m_header.ncmds; ++i) 1556 { 1557 const lldb::offset_t load_cmd_offset = offset; 1558 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1559 break; 1560 1561 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1562 { 1563 if (m_data.GetU8(&offset, (uint8_t*)load_cmd.segname, 16)) 1564 { 1565 bool add_section = true; 1566 bool add_to_unified = true; 1567 ConstString const_segname (load_cmd.segname, std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 1568 1569 SectionSP unified_section_sp(unified_section_list.FindSectionByName(const_segname)); 1570 if (is_dsym && unified_section_sp) 1571 { 1572 if (const_segname == GetSegmentNameLINKEDIT()) 1573 { 1574 // We need to keep the __LINKEDIT segment private to this object file only 1575 add_to_unified = false; 1576 } 1577 else 1578 { 1579 // This is the dSYM file and this section has already been created by 1580 // the object file, no need to create it. 1581 add_section = false; 1582 } 1583 } 1584 load_cmd.vmaddr = m_data.GetAddress(&offset); 1585 load_cmd.vmsize = m_data.GetAddress(&offset); 1586 load_cmd.fileoff = m_data.GetAddress(&offset); 1587 load_cmd.filesize = m_data.GetAddress(&offset); 1588 if (m_length != 0 && load_cmd.filesize != 0) 1589 { 1590 if (load_cmd.fileoff > m_length) 1591 { 1592 // We have a load command that says it extends past the end of the file. This is likely 1593 // a corrupt file. We don't have any way to return an error condition here (this method 1594 // was likely invoked from something like ObjectFile::GetSectionList()) -- all we can do 1595 // is null out the SectionList vector and if a process has been set up, dump a message 1596 // to stdout. The most common case here is core file debugging with a truncated file. 1597 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1598 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", 1599 i, 1600 lc_segment_name, 1601 load_cmd.fileoff, 1602 m_length); 1603 1604 load_cmd.fileoff = 0; 1605 load_cmd.filesize = 0; 1606 } 1607 1608 if (load_cmd.fileoff + load_cmd.filesize > m_length) 1609 { 1610 // We have a load command that says it extends past the end of the file. This is likely 1611 // a corrupt file. We don't have any way to return an error condition here (this method 1612 // was likely invoked from something like ObjectFile::GetSectionList()) -- all we can do 1613 // is null out the SectionList vector and if a process has been set up, dump a message 1614 // to stdout. The most common case here is core file debugging with a truncated file. 1615 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1616 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", 1617 i, 1618 lc_segment_name, 1619 load_cmd.fileoff + load_cmd.filesize, 1620 m_length); 1621 1622 // Tuncase the length 1623 load_cmd.filesize = m_length - load_cmd.fileoff; 1624 } 1625 } 1626 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1627 { 1628 1629 const bool segment_is_encrypted = (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1630 1631 // Keep a list of mach segments around in case we need to 1632 // get at data that isn't stored in the abstracted Sections. 1633 m_mach_segments.push_back (load_cmd); 1634 1635 // Use a segment ID of the segment index shifted left by 8 so they 1636 // never conflict with any of the sections. 1637 SectionSP segment_sp; 1638 if (add_section && (const_segname || is_core)) 1639 { 1640 segment_sp.reset(new Section (module_sp, // Module to which this section belongs 1641 this, // Object file to which this sections belongs 1642 ++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 1643 const_segname, // Name of this section 1644 eSectionTypeContainer, // This section is a container of other sections. 1645 load_cmd.vmaddr, // File VM address == addresses as they are found in the object file 1646 load_cmd.vmsize, // VM size in bytes of this section 1647 load_cmd.fileoff, // Offset to the data for this section in the file 1648 load_cmd.filesize, // Size in bytes of this section as found in the file 1649 0, // Segments have no alignment information 1650 load_cmd.flags)); // Flags for this section 1651 1652 segment_sp->SetIsEncrypted (segment_is_encrypted); 1653 m_sections_ap->AddSection(segment_sp); 1654 if (add_to_unified) 1655 unified_section_list.AddSection(segment_sp); 1656 } 1657 else if (unified_section_sp) 1658 { 1659 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) 1660 { 1661 // Check to see if the module was read from memory? 1662 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) 1663 { 1664 // We have a module that is in memory and needs to have its 1665 // file address adjusted. We need to do this because when we 1666 // load a file from memory, its addresses will be slid already, 1667 // yet the addresses in the new symbol file will still be unslid. 1668 // Since everything is stored as section offset, this shouldn't 1669 // cause any problems. 1670 1671 // Make sure we've parsed the symbol table from the 1672 // ObjectFile before we go around changing its Sections. 1673 module_sp->GetObjectFile()->GetSymtab(); 1674 // eh_frame would present the same problems but we parse that on 1675 // a per-function basis as-needed so it's more difficult to 1676 // remove its use of the Sections. Realistically, the environments 1677 // where this code path will be taken will not have eh_frame sections. 1678 1679 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1680 1681 // Notify the module that the section addresses have been changed once 1682 // we're done so any file-address caches can be updated. 1683 section_file_addresses_changed = true; 1684 } 1685 } 1686 m_sections_ap->AddSection(unified_section_sp); 1687 } 1688 1689 struct section_64 sect64; 1690 ::memset (§64, 0, sizeof(sect64)); 1691 // Push a section into our mach sections for the section at 1692 // index zero (NO_SECT) if we don't have any mach sections yet... 1693 if (m_mach_sections.empty()) 1694 m_mach_sections.push_back(sect64); 1695 uint32_t segment_sect_idx; 1696 const lldb::user_id_t first_segment_sectID = sectID + 1; 1697 1698 1699 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1700 for (segment_sect_idx=0; segment_sect_idx<load_cmd.nsects; ++segment_sect_idx) 1701 { 1702 if (m_data.GetU8(&offset, (uint8_t*)sect64.sectname, sizeof(sect64.sectname)) == NULL) 1703 break; 1704 if (m_data.GetU8(&offset, (uint8_t*)sect64.segname, sizeof(sect64.segname)) == NULL) 1705 break; 1706 sect64.addr = m_data.GetAddress(&offset); 1707 sect64.size = m_data.GetAddress(&offset); 1708 1709 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1710 break; 1711 1712 // Keep a list of mach sections around in case we need to 1713 // get at data that isn't stored in the abstracted Sections. 1714 m_mach_sections.push_back (sect64); 1715 1716 if (add_section) 1717 { 1718 ConstString section_name (sect64.sectname, std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1719 if (!const_segname) 1720 { 1721 // We have a segment with no name so we need to conjure up 1722 // segments that correspond to the section's segname if there 1723 // isn't already such a section. If there is such a section, 1724 // we resize the section so that it spans all sections. 1725 // We also mark these sections as fake so address matches don't 1726 // hit if they land in the gaps between the child sections. 1727 const_segname.SetTrimmedCStringWithLength(sect64.segname, sizeof(sect64.segname)); 1728 segment_sp = unified_section_list.FindSectionByName (const_segname); 1729 if (segment_sp.get()) 1730 { 1731 Section *segment = segment_sp.get(); 1732 // Grow the section size as needed. 1733 const lldb::addr_t sect64_min_addr = sect64.addr; 1734 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1735 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1736 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1737 const lldb::addr_t curr_seg_max_addr = curr_seg_min_addr + curr_seg_byte_size; 1738 if (sect64_min_addr >= curr_seg_min_addr) 1739 { 1740 const lldb::addr_t new_seg_byte_size = sect64_max_addr - curr_seg_min_addr; 1741 // Only grow the section size if needed 1742 if (new_seg_byte_size > curr_seg_byte_size) 1743 segment->SetByteSize (new_seg_byte_size); 1744 } 1745 else 1746 { 1747 // We need to change the base address of the segment and 1748 // adjust the child section offsets for all existing children. 1749 const lldb::addr_t slide_amount = sect64_min_addr - curr_seg_min_addr; 1750 segment->Slide(slide_amount, false); 1751 segment->GetChildren().Slide(-slide_amount, false); 1752 segment->SetByteSize (curr_seg_max_addr - sect64_min_addr); 1753 } 1754 1755 // Grow the section size as needed. 1756 if (sect64.offset) 1757 { 1758 const lldb::addr_t segment_min_file_offset = segment->GetFileOffset(); 1759 const lldb::addr_t segment_max_file_offset = segment_min_file_offset + segment->GetFileSize(); 1760 1761 const lldb::addr_t section_min_file_offset = sect64.offset; 1762 const lldb::addr_t section_max_file_offset = section_min_file_offset + sect64.size; 1763 const lldb::addr_t new_file_offset = std::min (section_min_file_offset, segment_min_file_offset); 1764 const lldb::addr_t new_file_size = std::max (section_max_file_offset, segment_max_file_offset) - new_file_offset; 1765 segment->SetFileOffset (new_file_offset); 1766 segment->SetFileSize (new_file_size); 1767 } 1768 } 1769 else 1770 { 1771 // Create a fake section for the section's named segment 1772 segment_sp.reset(new Section (segment_sp, // Parent section 1773 module_sp, // Module to which this section belongs 1774 this, // Object file to which this section belongs 1775 ++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 1776 const_segname, // Name of this section 1777 eSectionTypeContainer, // This section is a container of other sections. 1778 sect64.addr, // File VM address == addresses as they are found in the object file 1779 sect64.size, // VM size in bytes of this section 1780 sect64.offset, // Offset to the data for this section in the file 1781 sect64.offset ? sect64.size : 0, // Size in bytes of this section as found in the file 1782 sect64.align, 1783 load_cmd.flags)); // Flags for this section 1784 segment_sp->SetIsFake(true); 1785 1786 m_sections_ap->AddSection(segment_sp); 1787 if (add_to_unified) 1788 unified_section_list.AddSection(segment_sp); 1789 segment_sp->SetIsEncrypted (segment_is_encrypted); 1790 } 1791 } 1792 assert (segment_sp.get()); 1793 1794 lldb::SectionType sect_type = eSectionTypeOther; 1795 1796 if (sect64.flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1797 sect_type = eSectionTypeCode; 1798 else 1799 { 1800 uint32_t mach_sect_type = sect64.flags & SECTION_TYPE; 1801 static ConstString g_sect_name_objc_data ("__objc_data"); 1802 static ConstString g_sect_name_objc_msgrefs ("__objc_msgrefs"); 1803 static ConstString g_sect_name_objc_selrefs ("__objc_selrefs"); 1804 static ConstString g_sect_name_objc_classrefs ("__objc_classrefs"); 1805 static ConstString g_sect_name_objc_superrefs ("__objc_superrefs"); 1806 static ConstString g_sect_name_objc_const ("__objc_const"); 1807 static ConstString g_sect_name_objc_classlist ("__objc_classlist"); 1808 static ConstString g_sect_name_cfstring ("__cfstring"); 1809 1810 static ConstString g_sect_name_dwarf_debug_abbrev ("__debug_abbrev"); 1811 static ConstString g_sect_name_dwarf_debug_aranges ("__debug_aranges"); 1812 static ConstString g_sect_name_dwarf_debug_frame ("__debug_frame"); 1813 static ConstString g_sect_name_dwarf_debug_info ("__debug_info"); 1814 static ConstString g_sect_name_dwarf_debug_line ("__debug_line"); 1815 static ConstString g_sect_name_dwarf_debug_loc ("__debug_loc"); 1816 static ConstString g_sect_name_dwarf_debug_macinfo ("__debug_macinfo"); 1817 static ConstString g_sect_name_dwarf_debug_pubnames ("__debug_pubnames"); 1818 static ConstString g_sect_name_dwarf_debug_pubtypes ("__debug_pubtypes"); 1819 static ConstString g_sect_name_dwarf_debug_ranges ("__debug_ranges"); 1820 static ConstString g_sect_name_dwarf_debug_str ("__debug_str"); 1821 static ConstString g_sect_name_dwarf_apple_names ("__apple_names"); 1822 static ConstString g_sect_name_dwarf_apple_types ("__apple_types"); 1823 static ConstString g_sect_name_dwarf_apple_namespaces ("__apple_namespac"); 1824 static ConstString g_sect_name_dwarf_apple_objc ("__apple_objc"); 1825 static ConstString g_sect_name_eh_frame ("__eh_frame"); 1826 static ConstString g_sect_name_compact_unwind ("__unwind_info"); 1827 static ConstString g_sect_name_text ("__text"); 1828 static ConstString g_sect_name_data ("__data"); 1829 static ConstString g_sect_name_go_symtab ("__gosymtab"); 1830 1831 if (section_name == g_sect_name_dwarf_debug_abbrev) 1832 sect_type = eSectionTypeDWARFDebugAbbrev; 1833 else if (section_name == g_sect_name_dwarf_debug_aranges) 1834 sect_type = eSectionTypeDWARFDebugAranges; 1835 else if (section_name == g_sect_name_dwarf_debug_frame) 1836 sect_type = eSectionTypeDWARFDebugFrame; 1837 else if (section_name == g_sect_name_dwarf_debug_info) 1838 sect_type = eSectionTypeDWARFDebugInfo; 1839 else if (section_name == g_sect_name_dwarf_debug_line) 1840 sect_type = eSectionTypeDWARFDebugLine; 1841 else if (section_name == g_sect_name_dwarf_debug_loc) 1842 sect_type = eSectionTypeDWARFDebugLoc; 1843 else if (section_name == g_sect_name_dwarf_debug_macinfo) 1844 sect_type = eSectionTypeDWARFDebugMacInfo; 1845 else if (section_name == g_sect_name_dwarf_debug_pubnames) 1846 sect_type = eSectionTypeDWARFDebugPubNames; 1847 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 1848 sect_type = eSectionTypeDWARFDebugPubTypes; 1849 else if (section_name == g_sect_name_dwarf_debug_ranges) 1850 sect_type = eSectionTypeDWARFDebugRanges; 1851 else if (section_name == g_sect_name_dwarf_debug_str) 1852 sect_type = eSectionTypeDWARFDebugStr; 1853 else if (section_name == g_sect_name_dwarf_apple_names) 1854 sect_type = eSectionTypeDWARFAppleNames; 1855 else if (section_name == g_sect_name_dwarf_apple_types) 1856 sect_type = eSectionTypeDWARFAppleTypes; 1857 else if (section_name == g_sect_name_dwarf_apple_namespaces) 1858 sect_type = eSectionTypeDWARFAppleNamespaces; 1859 else if (section_name == g_sect_name_dwarf_apple_objc) 1860 sect_type = eSectionTypeDWARFAppleObjC; 1861 else if (section_name == g_sect_name_objc_selrefs) 1862 sect_type = eSectionTypeDataCStringPointers; 1863 else if (section_name == g_sect_name_objc_msgrefs) 1864 sect_type = eSectionTypeDataObjCMessageRefs; 1865 else if (section_name == g_sect_name_eh_frame) 1866 sect_type = eSectionTypeEHFrame; 1867 else if (section_name == g_sect_name_compact_unwind) 1868 sect_type = eSectionTypeCompactUnwind; 1869 else if (section_name == g_sect_name_cfstring) 1870 sect_type = eSectionTypeDataObjCCFStrings; 1871 else if (section_name == g_sect_name_go_symtab) 1872 sect_type = eSectionTypeGoSymtab; 1873 else if (section_name == g_sect_name_objc_data || 1874 section_name == g_sect_name_objc_classrefs || 1875 section_name == g_sect_name_objc_superrefs || 1876 section_name == g_sect_name_objc_const || 1877 section_name == g_sect_name_objc_classlist) 1878 { 1879 sect_type = eSectionTypeDataPointers; 1880 } 1881 1882 if (sect_type == eSectionTypeOther) 1883 { 1884 switch (mach_sect_type) 1885 { 1886 // TODO: categorize sections by other flags for regular sections 1887 case S_REGULAR: 1888 if (section_name == g_sect_name_text) 1889 sect_type = eSectionTypeCode; 1890 else if (section_name == g_sect_name_data) 1891 sect_type = eSectionTypeData; 1892 else 1893 sect_type = eSectionTypeOther; 1894 break; 1895 case S_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1896 case S_CSTRING_LITERALS: sect_type = eSectionTypeDataCString; break; // section with only literal C strings 1897 case S_4BYTE_LITERALS: sect_type = eSectionTypeData4; break; // section with only 4 byte literals 1898 case S_8BYTE_LITERALS: sect_type = eSectionTypeData8; break; // section with only 8 byte literals 1899 case S_LITERAL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only pointers to literals 1900 case S_NON_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only non-lazy symbol pointers 1901 case S_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only lazy symbol pointers 1902 case S_SYMBOL_STUBS: sect_type = eSectionTypeCode; break; // section with only symbol stubs, byte size of stub in the reserved2 field 1903 case S_MOD_INIT_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for initialization 1904 case S_MOD_TERM_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for termination 1905 case S_COALESCED: sect_type = eSectionTypeOther; break; 1906 case S_GB_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1907 case S_INTERPOSING: sect_type = eSectionTypeCode; break; // section with only pairs of function pointers for interposing 1908 case S_16BYTE_LITERALS: sect_type = eSectionTypeData16; break; // section with only 16 byte literals 1909 case S_DTRACE_DOF: sect_type = eSectionTypeDebug; break; 1910 case S_LAZY_DYLIB_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; 1911 default: break; 1912 } 1913 } 1914 } 1915 1916 SectionSP section_sp(new Section (segment_sp, 1917 module_sp, 1918 this, 1919 ++sectID, 1920 section_name, 1921 sect_type, 1922 sect64.addr - segment_sp->GetFileAddress(), 1923 sect64.size, 1924 sect64.offset, 1925 sect64.offset == 0 ? 0 : sect64.size, 1926 sect64.align, 1927 sect64.flags)); 1928 // Set the section to be encrypted to match the segment 1929 1930 bool section_is_encrypted = false; 1931 if (!segment_is_encrypted && load_cmd.filesize != 0) 1932 section_is_encrypted = encrypted_file_ranges.FindEntryThatContains(sect64.offset) != NULL; 1933 1934 section_sp->SetIsEncrypted (segment_is_encrypted || section_is_encrypted); 1935 segment_sp->GetChildren().AddSection(section_sp); 1936 1937 if (segment_sp->IsFake()) 1938 { 1939 segment_sp.reset(); 1940 const_segname.Clear(); 1941 } 1942 } 1943 } 1944 if (segment_sp && is_dsym) 1945 { 1946 if (first_segment_sectID <= sectID) 1947 { 1948 lldb::user_id_t sect_uid; 1949 for (sect_uid = first_segment_sectID; sect_uid <= sectID; ++sect_uid) 1950 { 1951 SectionSP curr_section_sp(segment_sp->GetChildren().FindSectionByID (sect_uid)); 1952 SectionSP next_section_sp; 1953 if (sect_uid + 1 <= sectID) 1954 next_section_sp = segment_sp->GetChildren().FindSectionByID (sect_uid+1); 1955 1956 if (curr_section_sp.get()) 1957 { 1958 if (curr_section_sp->GetByteSize() == 0) 1959 { 1960 if (next_section_sp.get() != NULL) 1961 curr_section_sp->SetByteSize ( next_section_sp->GetFileAddress() - curr_section_sp->GetFileAddress() ); 1962 else 1963 curr_section_sp->SetByteSize ( load_cmd.vmsize ); 1964 } 1965 } 1966 } 1967 } 1968 } 1969 } 1970 } 1971 } 1972 else if (load_cmd.cmd == LC_DYSYMTAB) 1973 { 1974 m_dysymtab.cmd = load_cmd.cmd; 1975 m_dysymtab.cmdsize = load_cmd.cmdsize; 1976 m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1977 } 1978 1979 offset = load_cmd_offset + load_cmd.cmdsize; 1980 } 1981 1982 1983 if (section_file_addresses_changed && module_sp.get()) 1984 { 1985 module_sp->SectionFileAddressesChanged(); 1986 } 1987 } 1988 } 1989 1990 class MachSymtabSectionInfo 1991 { 1992 public: 1993 MachSymtabSectionInfo (SectionList *section_list) : 1994 m_section_list (section_list), 1995 m_section_infos() 1996 { 1997 // Get the number of sections down to a depth of 1 to include 1998 // all segments and their sections, but no other sections that 1999 // may be added for debug map or 2000 m_section_infos.resize(section_list->GetNumSections(1)); 2001 } 2002 2003 SectionSP 2004 GetSection (uint8_t n_sect, addr_t file_addr) 2005 { 2006 if (n_sect == 0) 2007 return SectionSP(); 2008 if (n_sect < m_section_infos.size()) 2009 { 2010 if (!m_section_infos[n_sect].section_sp) 2011 { 2012 SectionSP section_sp (m_section_list->FindSectionByID (n_sect)); 2013 m_section_infos[n_sect].section_sp = section_sp; 2014 if (section_sp) 2015 { 2016 m_section_infos[n_sect].vm_range.SetBaseAddress (section_sp->GetFileAddress()); 2017 m_section_infos[n_sect].vm_range.SetByteSize (section_sp->GetByteSize()); 2018 } 2019 else 2020 { 2021 Host::SystemLog (Host::eSystemLogError, "error: unable to find section for section %u\n", n_sect); 2022 } 2023 } 2024 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) 2025 { 2026 // Symbol is in section. 2027 return m_section_infos[n_sect].section_sp; 2028 } 2029 else if (m_section_infos[n_sect].vm_range.GetByteSize () == 0 && 2030 m_section_infos[n_sect].vm_range.GetBaseAddress() == file_addr) 2031 { 2032 // Symbol is in section with zero size, but has the same start 2033 // address as the section. This can happen with linker symbols 2034 // (symbols that start with the letter 'l' or 'L'. 2035 return m_section_infos[n_sect].section_sp; 2036 } 2037 } 2038 return m_section_list->FindSectionContainingFileAddress(file_addr); 2039 } 2040 2041 protected: 2042 struct SectionInfo 2043 { 2044 SectionInfo () : 2045 vm_range(), 2046 section_sp () 2047 { 2048 } 2049 2050 VMRange vm_range; 2051 SectionSP section_sp; 2052 }; 2053 SectionList *m_section_list; 2054 std::vector<SectionInfo> m_section_infos; 2055 }; 2056 2057 struct TrieEntry 2058 { 2059 TrieEntry () : 2060 name(), 2061 address(LLDB_INVALID_ADDRESS), 2062 flags (0), 2063 other(0), 2064 import_name() 2065 { 2066 } 2067 2068 void 2069 Clear () 2070 { 2071 name.Clear(); 2072 address = LLDB_INVALID_ADDRESS; 2073 flags = 0; 2074 other = 0; 2075 import_name.Clear(); 2076 } 2077 2078 void 2079 Dump () const 2080 { 2081 printf ("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 2082 static_cast<unsigned long long>(address), 2083 static_cast<unsigned long long>(flags), 2084 static_cast<unsigned long long>(other), name.GetCString()); 2085 if (import_name) 2086 printf (" -> \"%s\"\n", import_name.GetCString()); 2087 else 2088 printf ("\n"); 2089 } 2090 ConstString name; 2091 uint64_t address; 2092 uint64_t flags; 2093 uint64_t other; 2094 ConstString import_name; 2095 }; 2096 2097 struct TrieEntryWithOffset 2098 { 2099 lldb::offset_t nodeOffset; 2100 TrieEntry entry; 2101 2102 TrieEntryWithOffset (lldb::offset_t offset) : 2103 nodeOffset (offset), 2104 entry() 2105 { 2106 } 2107 2108 void 2109 Dump (uint32_t idx) const 2110 { 2111 printf ("[%3u] 0x%16.16llx: ", idx, 2112 static_cast<unsigned long long>(nodeOffset)); 2113 entry.Dump(); 2114 } 2115 2116 bool 2117 operator<(const TrieEntryWithOffset& other) const 2118 { 2119 return ( nodeOffset < other.nodeOffset ); 2120 } 2121 }; 2122 2123 static bool 2124 ParseTrieEntries (DataExtractor &data, 2125 lldb::offset_t offset, 2126 const bool is_arm, 2127 std::vector<llvm::StringRef> &nameSlices, 2128 std::set<lldb::addr_t> &resolver_addresses, 2129 std::vector<TrieEntryWithOffset>& output) 2130 { 2131 if (!data.ValidOffset(offset)) 2132 return true; 2133 2134 const uint64_t terminalSize = data.GetULEB128(&offset); 2135 lldb::offset_t children_offset = offset + terminalSize; 2136 if ( terminalSize != 0 ) { 2137 TrieEntryWithOffset e (offset); 2138 e.entry.flags = data.GetULEB128(&offset); 2139 const char *import_name = NULL; 2140 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT ) { 2141 e.entry.address = 0; 2142 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2143 import_name = data.GetCStr(&offset); 2144 } 2145 else { 2146 e.entry.address = data.GetULEB128(&offset); 2147 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER ) 2148 { 2149 e.entry.other = data.GetULEB128(&offset); 2150 uint64_t resolver_addr = e.entry.other; 2151 if (is_arm) 2152 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2153 resolver_addresses.insert(resolver_addr); 2154 } 2155 else 2156 e.entry.other = 0; 2157 } 2158 // Only add symbols that are reexport symbols with a valid import name 2159 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && import_name[0]) 2160 { 2161 std::string name; 2162 if (!nameSlices.empty()) 2163 { 2164 for (auto name_slice: nameSlices) 2165 name.append(name_slice.data(), name_slice.size()); 2166 } 2167 if (name.size() > 1) 2168 { 2169 // Skip the leading '_' 2170 e.entry.name.SetCStringWithLength(name.c_str() + 1,name.size() - 1); 2171 } 2172 if (import_name) 2173 { 2174 // Skip the leading '_' 2175 e.entry.import_name.SetCString(import_name+1); 2176 } 2177 output.push_back(e); 2178 } 2179 } 2180 2181 const uint8_t childrenCount = data.GetU8(&children_offset); 2182 for (uint8_t i=0; i < childrenCount; ++i) { 2183 const char *cstr = data.GetCStr(&children_offset); 2184 if (cstr) 2185 nameSlices.push_back(llvm::StringRef(cstr)); 2186 else 2187 return false; // Corrupt data 2188 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2189 if (childNodeOffset) 2190 { 2191 if (!ParseTrieEntries(data, 2192 childNodeOffset, 2193 is_arm, 2194 nameSlices, 2195 resolver_addresses, 2196 output)) 2197 { 2198 return false; 2199 } 2200 } 2201 nameSlices.pop_back(); 2202 } 2203 return true; 2204 } 2205 2206 // Read the UUID out of a dyld_shared_cache file on-disk. 2207 UUID 2208 ObjectFileMachO::GetSharedCacheUUID (FileSpec dyld_shared_cache, const ByteOrder byte_order, const uint32_t addr_byte_size) 2209 { 2210 UUID dsc_uuid; 2211 DataBufferSP dsc_data_sp = dyld_shared_cache.MemoryMapFileContentsIfLocal(0, sizeof(struct lldb_copy_dyld_cache_header_v1)); 2212 if (dsc_data_sp) 2213 { 2214 DataExtractor dsc_header_data (dsc_data_sp, byte_order, addr_byte_size); 2215 2216 char version_str[7]; 2217 lldb::offset_t offset = 0; 2218 memcpy (version_str, dsc_header_data.GetData (&offset, 6), 6); 2219 version_str[6] = '\0'; 2220 if (strcmp (version_str, "dyld_v") == 0) 2221 { 2222 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, uuid); 2223 uint8_t uuid_bytes[sizeof (uuid_t)]; 2224 memcpy (uuid_bytes, dsc_header_data.GetData (&offset, sizeof (uuid_t)), sizeof (uuid_t)); 2225 dsc_uuid.SetBytes (uuid_bytes); 2226 } 2227 } 2228 return dsc_uuid; 2229 } 2230 2231 size_t 2232 ObjectFileMachO::ParseSymtab () 2233 { 2234 Timer scoped_timer(__PRETTY_FUNCTION__, 2235 "ObjectFileMachO::ParseSymtab () module = %s", 2236 m_file.GetFilename().AsCString("")); 2237 ModuleSP module_sp (GetModule()); 2238 if (!module_sp) 2239 return 0; 2240 2241 struct symtab_command symtab_load_command = { 0, 0, 0, 0, 0, 0 }; 2242 struct linkedit_data_command function_starts_load_command = { 0, 0, 0, 0 }; 2243 struct dyld_info_command dyld_info = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; 2244 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2245 FunctionStarts function_starts; 2246 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2247 uint32_t i; 2248 FileSpecList dylib_files; 2249 Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_SYMBOLS)); 2250 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 2251 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 2252 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 2253 2254 for (i=0; i<m_header.ncmds; ++i) 2255 { 2256 const lldb::offset_t cmd_offset = offset; 2257 // Read in the load command and load command size 2258 struct load_command lc; 2259 if (m_data.GetU32(&offset, &lc, 2) == NULL) 2260 break; 2261 // Watch for the symbol table load command 2262 switch (lc.cmd) 2263 { 2264 case LC_SYMTAB: 2265 symtab_load_command.cmd = lc.cmd; 2266 symtab_load_command.cmdsize = lc.cmdsize; 2267 // Read in the rest of the symtab load command 2268 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 0) // fill in symoff, nsyms, stroff, strsize fields 2269 return 0; 2270 if (symtab_load_command.symoff == 0) 2271 { 2272 if (log) 2273 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2274 return 0; 2275 } 2276 2277 if (symtab_load_command.stroff == 0) 2278 { 2279 if (log) 2280 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2281 return 0; 2282 } 2283 2284 if (symtab_load_command.nsyms == 0) 2285 { 2286 if (log) 2287 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2288 return 0; 2289 } 2290 2291 if (symtab_load_command.strsize == 0) 2292 { 2293 if (log) 2294 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2295 return 0; 2296 } 2297 break; 2298 2299 case LC_DYLD_INFO: 2300 case LC_DYLD_INFO_ONLY: 2301 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) 2302 { 2303 dyld_info.cmd = lc.cmd; 2304 dyld_info.cmdsize = lc.cmdsize; 2305 } 2306 else 2307 { 2308 memset (&dyld_info, 0, sizeof(dyld_info)); 2309 } 2310 break; 2311 2312 case LC_LOAD_DYLIB: 2313 case LC_LOAD_WEAK_DYLIB: 2314 case LC_REEXPORT_DYLIB: 2315 case LC_LOADFVMLIB: 2316 case LC_LOAD_UPWARD_DYLIB: 2317 { 2318 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2319 const char *path = m_data.PeekCStr(name_offset); 2320 if (path) 2321 { 2322 FileSpec file_spec(path, false); 2323 // Strip the path if there is @rpath, @executable, etc so we just use the basename 2324 if (path[0] == '@') 2325 file_spec.GetDirectory().Clear(); 2326 2327 if (lc.cmd == LC_REEXPORT_DYLIB) 2328 { 2329 m_reexported_dylibs.AppendIfUnique(file_spec); 2330 } 2331 2332 dylib_files.Append(file_spec); 2333 } 2334 } 2335 break; 2336 2337 case LC_FUNCTION_STARTS: 2338 function_starts_load_command.cmd = lc.cmd; 2339 function_starts_load_command.cmdsize = lc.cmdsize; 2340 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == NULL) // fill in symoff, nsyms, stroff, strsize fields 2341 memset (&function_starts_load_command, 0, sizeof(function_starts_load_command)); 2342 break; 2343 2344 default: 2345 break; 2346 } 2347 offset = cmd_offset + lc.cmdsize; 2348 } 2349 2350 if (symtab_load_command.cmd) 2351 { 2352 Symtab *symtab = m_symtab_ap.get(); 2353 SectionList *section_list = GetSectionList(); 2354 if (section_list == NULL) 2355 return 0; 2356 2357 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2358 const ByteOrder byte_order = m_data.GetByteOrder(); 2359 bool bit_width_32 = addr_byte_size == 4; 2360 const size_t nlist_byte_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2361 2362 DataExtractor nlist_data (NULL, 0, byte_order, addr_byte_size); 2363 DataExtractor strtab_data (NULL, 0, byte_order, addr_byte_size); 2364 DataExtractor function_starts_data (NULL, 0, byte_order, addr_byte_size); 2365 DataExtractor indirect_symbol_index_data (NULL, 0, byte_order, addr_byte_size); 2366 DataExtractor dyld_trie_data (NULL, 0, byte_order, addr_byte_size); 2367 2368 const addr_t nlist_data_byte_size = symtab_load_command.nsyms * nlist_byte_size; 2369 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2370 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2371 2372 ProcessSP process_sp (m_process_wp.lock()); 2373 Process *process = process_sp.get(); 2374 2375 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2376 2377 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) 2378 { 2379 Target &target = process->GetTarget(); 2380 2381 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2382 2383 SectionSP linkedit_section_sp(section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2384 // Reading mach file from memory in a process or core file... 2385 2386 if (linkedit_section_sp) 2387 { 2388 addr_t linkedit_load_addr = linkedit_section_sp->GetLoadBaseAddress(&target); 2389 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) 2390 { 2391 // We might be trying to access the symbol table before the __LINKEDIT's load 2392 // address has been set in the target. We can't fail to read the symbol table, 2393 // so calculate the right address manually 2394 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage(m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2395 } 2396 2397 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2398 const addr_t symoff_addr = linkedit_load_addr + symtab_load_command.symoff - linkedit_file_offset; 2399 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - linkedit_file_offset; 2400 2401 bool data_was_read = false; 2402 2403 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 2404 if (m_header.flags & 0x80000000u && process->GetAddressByteSize() == sizeof (void*)) 2405 { 2406 // This mach-o memory file is in the dyld shared cache. If this 2407 // program is not remote and this is iOS, then this process will 2408 // share the same shared cache as the process we are debugging and 2409 // we can read the entire __LINKEDIT from the address space in this 2410 // process. This is a needed optimization that is used for local iOS 2411 // debugging only since all shared libraries in the shared cache do 2412 // not have corresponding files that exist in the file system of the 2413 // device. They have been combined into a single file. This means we 2414 // always have to load these files from memory. All of the symbol and 2415 // string tables from all of the __LINKEDIT sections from the shared 2416 // libraries in the shared cache have been merged into a single large 2417 // symbol and string table. Reading all of this symbol and string table 2418 // data across can slow down debug launch times, so we optimize this by 2419 // reading the memory for the __LINKEDIT section from this process. 2420 2421 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 2422 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 2423 bool use_lldb_cache = true; 2424 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && lldb_shared_cache != process_shared_cache) 2425 { 2426 use_lldb_cache = false; 2427 ModuleSP module_sp (GetModule()); 2428 if (module_sp) 2429 module_sp->ReportWarning ("shared cache in process does not match lldb's own shared cache, startup will be slow."); 2430 2431 } 2432 2433 PlatformSP platform_sp (target.GetPlatform()); 2434 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) 2435 { 2436 data_was_read = true; 2437 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, eByteOrderLittle); 2438 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, eByteOrderLittle); 2439 if (function_starts_load_command.cmd) 2440 { 2441 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 2442 function_starts_data.SetData ((void *)func_start_addr, function_starts_load_command.datasize, eByteOrderLittle); 2443 } 2444 } 2445 } 2446 #endif 2447 2448 if (!data_was_read) 2449 { 2450 if (memory_module_load_level == eMemoryModuleLoadLevelComplete) 2451 { 2452 DataBufferSP nlist_data_sp (ReadMemory (process_sp, symoff_addr, nlist_data_byte_size)); 2453 if (nlist_data_sp) 2454 nlist_data.SetData (nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2455 // Load strings individually from memory when loading from memory since shared cache 2456 // string tables contain strings for all symbols from all shared cached libraries 2457 //DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, strtab_data_byte_size)); 2458 //if (strtab_data_sp) 2459 // strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 2460 if (m_dysymtab.nindirectsyms != 0) 2461 { 2462 const addr_t indirect_syms_addr = linkedit_load_addr + m_dysymtab.indirectsymoff - linkedit_file_offset; 2463 DataBufferSP indirect_syms_data_sp (ReadMemory (process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 2464 if (indirect_syms_data_sp) 2465 indirect_symbol_index_data.SetData (indirect_syms_data_sp, 0, indirect_syms_data_sp->GetByteSize()); 2466 } 2467 } 2468 2469 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) 2470 { 2471 if (function_starts_load_command.cmd) 2472 { 2473 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 2474 DataBufferSP func_start_data_sp (ReadMemory (process_sp, func_start_addr, function_starts_load_command.datasize)); 2475 if (func_start_data_sp) 2476 function_starts_data.SetData (func_start_data_sp, 0, func_start_data_sp->GetByteSize()); 2477 } 2478 } 2479 } 2480 } 2481 } 2482 else 2483 { 2484 nlist_data.SetData (m_data, 2485 symtab_load_command.symoff, 2486 nlist_data_byte_size); 2487 strtab_data.SetData (m_data, 2488 symtab_load_command.stroff, 2489 strtab_data_byte_size); 2490 2491 if (dyld_info.export_size > 0) 2492 { 2493 dyld_trie_data.SetData (m_data, 2494 dyld_info.export_off, 2495 dyld_info.export_size); 2496 } 2497 2498 if (m_dysymtab.nindirectsyms != 0) 2499 { 2500 indirect_symbol_index_data.SetData (m_data, 2501 m_dysymtab.indirectsymoff, 2502 m_dysymtab.nindirectsyms * 4); 2503 } 2504 if (function_starts_load_command.cmd) 2505 { 2506 function_starts_data.SetData (m_data, 2507 function_starts_load_command.dataoff, 2508 function_starts_load_command.datasize); 2509 } 2510 } 2511 2512 if (nlist_data.GetByteSize() == 0 && memory_module_load_level == eMemoryModuleLoadLevelComplete) 2513 { 2514 if (log) 2515 module_sp->LogMessage(log, "failed to read nlist data"); 2516 return 0; 2517 } 2518 2519 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2520 if (!have_strtab_data) 2521 { 2522 if (process) 2523 { 2524 if (strtab_addr == LLDB_INVALID_ADDRESS) 2525 { 2526 if (log) 2527 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2528 return 0; 2529 } 2530 } 2531 else 2532 { 2533 if (log) 2534 module_sp->LogMessage(log, "failed to read strtab data"); 2535 return 0; 2536 } 2537 } 2538 2539 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2540 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2541 const ConstString &g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2542 const ConstString &g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2543 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2544 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2545 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 2546 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 2547 SectionSP data_dirty_section_sp(section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2548 SectionSP data_const_section_sp(section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2549 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 2550 SectionSP eh_frame_section_sp; 2551 if (text_section_sp.get()) 2552 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 2553 else 2554 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 2555 2556 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2557 2558 // lldb works best if it knows the start address of all functions in a module. 2559 // Linker symbols or debug info are normally the best source of information for start addr / size but 2560 // they may be stripped in a released binary. 2561 // Two additional sources of information exist in Mach-O binaries: 2562 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each function's start address in the 2563 // binary, relative to the text section. 2564 // eh_frame - the eh_frame FDEs have the start addr & size of each function 2565 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on all modern binaries. 2566 // Binaries built to run on older releases may need to use eh_frame information. 2567 2568 if (text_section_sp && function_starts_data.GetByteSize()) 2569 { 2570 FunctionStarts::Entry function_start_entry; 2571 function_start_entry.data = false; 2572 lldb::offset_t function_start_offset = 0; 2573 function_start_entry.addr = text_section_sp->GetFileAddress(); 2574 uint64_t delta; 2575 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 0) 2576 { 2577 // Now append the current entry 2578 function_start_entry.addr += delta; 2579 function_starts.Append(function_start_entry); 2580 } 2581 } 2582 else 2583 { 2584 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the load command claiming an eh_frame 2585 // but it doesn't actually have the eh_frame content. And if we have a dSYM, we don't need to do any 2586 // of this fill-in-the-missing-symbols works anyway - the debug info should give us all the functions in 2587 // the module. 2588 if (text_section_sp.get() && eh_frame_section_sp.get() && m_type != eTypeDebugInfo) 2589 { 2590 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, eRegisterKindEHFrame, true); 2591 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2592 eh_frame.GetFunctionAddressAndSizeVector (functions); 2593 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2594 size_t count = functions.GetSize(); 2595 for (size_t i = 0; i < count; ++i) 2596 { 2597 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = functions.GetEntryAtIndex (i); 2598 if (func) 2599 { 2600 FunctionStarts::Entry function_start_entry; 2601 function_start_entry.addr = func->base - text_base_addr; 2602 function_starts.Append(function_start_entry); 2603 } 2604 } 2605 } 2606 } 2607 2608 const size_t function_starts_count = function_starts.GetSize(); 2609 2610 // For user process binaries (executables, dylibs, frameworks, bundles), if we don't have 2611 // LC_FUNCTION_STARTS/eh_frame section in this binary, we're going to assume the binary 2612 // has been stripped. Don't allow assembly language instruction emulation because we don't 2613 // know proper function start boundaries. 2614 // 2615 // For all other types of binaries (kernels, stand-alone bare board binaries, kexts), they 2616 // may not have LC_FUNCTION_STARTS / eh_frame sections - we should not make any assumptions 2617 // about them based on that. 2618 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) 2619 { 2620 m_allow_assembly_emulation_unwind_plans = false; 2621 Log *unwind_or_symbol_log (lldb_private::GetLogIfAnyCategoriesSet (LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2622 2623 if (unwind_or_symbol_log) 2624 module_sp->LogMessage(unwind_or_symbol_log, "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2625 } 2626 2627 const user_id_t TEXT_eh_frame_sectID = 2628 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2629 : static_cast<user_id_t>(NO_SECT); 2630 2631 lldb::offset_t nlist_data_offset = 0; 2632 2633 uint32_t N_SO_index = UINT32_MAX; 2634 2635 MachSymtabSectionInfo section_info (section_list); 2636 std::vector<uint32_t> N_FUN_indexes; 2637 std::vector<uint32_t> N_NSYM_indexes; 2638 std::vector<uint32_t> N_INCL_indexes; 2639 std::vector<uint32_t> N_BRAC_indexes; 2640 std::vector<uint32_t> N_COMM_indexes; 2641 typedef std::multimap <uint64_t, uint32_t> ValueToSymbolIndexMap; 2642 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2643 typedef std::map <const char *, uint32_t> ConstNameToSymbolIndexMap; 2644 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2645 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2646 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2647 // Any symbols that get merged into another will get an entry 2648 // in this map so we know 2649 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2650 uint32_t nlist_idx = 0; 2651 Symbol *symbol_ptr = NULL; 2652 2653 uint32_t sym_idx = 0; 2654 Symbol *sym = NULL; 2655 size_t num_syms = 0; 2656 std::string memory_symbol_name; 2657 uint32_t unmapped_local_symbols_found = 0; 2658 2659 std::vector<TrieEntryWithOffset> trie_entries; 2660 std::set<lldb::addr_t> resolver_addresses; 2661 2662 if (dyld_trie_data.GetByteSize() > 0) 2663 { 2664 std::vector<llvm::StringRef> nameSlices; 2665 ParseTrieEntries (dyld_trie_data, 2666 0, 2667 is_arm, 2668 nameSlices, 2669 resolver_addresses, 2670 trie_entries); 2671 2672 ConstString text_segment_name ("__TEXT"); 2673 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 2674 if (text_segment_sp) 2675 { 2676 const lldb::addr_t text_segment_file_addr = text_segment_sp->GetFileAddress(); 2677 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) 2678 { 2679 for (auto &e : trie_entries) 2680 e.entry.address += text_segment_file_addr; 2681 } 2682 } 2683 } 2684 2685 typedef std::set<ConstString> IndirectSymbols; 2686 IndirectSymbols indirect_symbol_names; 2687 2688 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 2689 2690 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been optimized by moving LOCAL 2691 // symbols out of the memory mapped portion of the DSC. The symbol information has all been retained, 2692 // but it isn't available in the normal nlist data. However, there *are* duplicate entries of *some* 2693 // LOCAL symbols in the normal nlist data. To handle this situation correctly, we must first attempt 2694 // to parse any DSC unmapped symbol information. If we find any, we set a flag that tells the normal 2695 // nlist parser to ignore all LOCAL symbols. 2696 2697 if (m_header.flags & 0x80000000u) 2698 { 2699 // Before we can start mapping the DSC, we need to make certain the target process is actually 2700 // using the cache we can find. 2701 2702 // Next we need to determine the correct path for the dyld shared cache. 2703 2704 ArchSpec header_arch; 2705 GetArchitecture(header_arch); 2706 char dsc_path[PATH_MAX]; 2707 char dsc_path_development[PATH_MAX]; 2708 2709 snprintf(dsc_path, sizeof(dsc_path), "%s%s%s", 2710 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR */ 2711 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2712 header_arch.GetArchitectureName()); 2713 2714 snprintf(dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2715 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR */ 2716 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2717 header_arch.GetArchitectureName(), 2718 ".development"); 2719 2720 FileSpec dsc_nondevelopment_filespec(dsc_path, false); 2721 FileSpec dsc_development_filespec(dsc_path_development, false); 2722 FileSpec dsc_filespec; 2723 2724 UUID dsc_uuid; 2725 UUID process_shared_cache_uuid; 2726 2727 if (process) 2728 { 2729 process_shared_cache_uuid = GetProcessSharedCacheUUID(process); 2730 } 2731 2732 // First see if we can find an exact match for the inferior process shared cache UUID in 2733 // the development or non-development shared caches on disk. 2734 if (process_shared_cache_uuid.IsValid()) 2735 { 2736 if (dsc_development_filespec.Exists()) 2737 { 2738 UUID dsc_development_uuid = GetSharedCacheUUID (dsc_development_filespec, byte_order, addr_byte_size); 2739 if (dsc_development_uuid.IsValid() && dsc_development_uuid == process_shared_cache_uuid) 2740 { 2741 dsc_filespec = dsc_development_filespec; 2742 dsc_uuid = dsc_development_uuid; 2743 } 2744 } 2745 if (!dsc_uuid.IsValid() && dsc_nondevelopment_filespec.Exists()) 2746 { 2747 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID (dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2748 if (dsc_nondevelopment_uuid.IsValid() && dsc_nondevelopment_uuid == process_shared_cache_uuid) 2749 { 2750 dsc_filespec = dsc_nondevelopment_filespec; 2751 dsc_uuid = dsc_nondevelopment_uuid; 2752 } 2753 } 2754 } 2755 2756 // Failing a UUID match, prefer the development dyld_shared cache if both are present. 2757 if (!dsc_filespec.Exists()) 2758 { 2759 if (dsc_development_filespec.Exists()) 2760 { 2761 dsc_filespec = dsc_development_filespec; 2762 } 2763 else 2764 { 2765 dsc_filespec = dsc_nondevelopment_filespec; 2766 } 2767 } 2768 2769 /* The dyld_cache_header has a pointer to the dyld_cache_local_symbols_info structure (localSymbolsOffset). 2770 The dyld_cache_local_symbols_info structure gives us three things: 2771 1. The start and count of the nlist records in the dyld_shared_cache file 2772 2. The start and size of the strings for these nlist records 2773 3. The start and count of dyld_cache_local_symbols_entry entries 2774 2775 There is one dyld_cache_local_symbols_entry per dylib/framework in the dyld shared cache. 2776 The "dylibOffset" field is the Mach-O header of this dylib/framework in the dyld shared cache. 2777 The dyld_cache_local_symbols_entry also lists the start of this dylib/framework's nlist records 2778 and the count of how many nlist records there are for this dylib/framework. 2779 */ 2780 2781 // Process the dyld shared cache header to find the unmapped symbols 2782 2783 DataBufferSP dsc_data_sp = dsc_filespec.MemoryMapFileContentsIfLocal(0, sizeof(struct lldb_copy_dyld_cache_header_v1)); 2784 if (!dsc_uuid.IsValid()) 2785 { 2786 dsc_uuid = GetSharedCacheUUID (dsc_filespec, byte_order, addr_byte_size); 2787 } 2788 if (dsc_data_sp) 2789 { 2790 DataExtractor dsc_header_data (dsc_data_sp, byte_order, addr_byte_size); 2791 2792 bool uuid_match = true; 2793 if (dsc_uuid.IsValid() && process) 2794 { 2795 if (process_shared_cache_uuid.IsValid() && dsc_uuid != process_shared_cache_uuid) 2796 { 2797 // The on-disk dyld_shared_cache file is not the same as the one in this 2798 // process' memory, don't use it. 2799 uuid_match = false; 2800 ModuleSP module_sp (GetModule()); 2801 if (module_sp) 2802 module_sp->ReportWarning ("process shared cache does not match on-disk dyld_shared_cache file, some symbol names will be missing."); 2803 } 2804 } 2805 2806 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2807 2808 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2809 2810 // If the mappingOffset points to a location inside the header, we've 2811 // opened an old dyld shared cache, and should not proceed further. 2812 if (uuid_match && mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) 2813 { 2814 2815 DataBufferSP dsc_mapping_info_data_sp = dsc_filespec.MemoryMapFileContentsIfLocal(mappingOffset, sizeof (struct lldb_copy_dyld_cache_mapping_info)); 2816 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, byte_order, addr_byte_size); 2817 offset = 0; 2818 2819 // The File addresses (from the in-memory Mach-O load commands) for the shared libraries 2820 // in the shared library cache need to be adjusted by an offset to match up with the 2821 // dylibOffset identifying field in the dyld_cache_local_symbol_entry's. This offset is 2822 // recorded in mapping_offset_value. 2823 const uint64_t mapping_offset_value = dsc_mapping_info_data.GetU64(&offset); 2824 2825 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2826 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2827 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2828 2829 if (localSymbolsOffset && localSymbolsSize) 2830 { 2831 // Map the local symbols 2832 if (DataBufferSP dsc_local_symbols_data_sp = dsc_filespec.MemoryMapFileContentsIfLocal(localSymbolsOffset, localSymbolsSize)) 2833 { 2834 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, byte_order, addr_byte_size); 2835 2836 offset = 0; 2837 2838 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2839 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2840 UndefinedNameToDescMap undefined_name_to_desc; 2841 SymbolIndexToName reexport_shlib_needs_fixup; 2842 2843 2844 // Read the local_symbols_infos struct in one shot 2845 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2846 dsc_local_symbols_data.GetU32(&offset, &local_symbols_info.nlistOffset, 6); 2847 2848 SectionSP text_section_sp(section_list->FindSectionByName(GetSegmentNameTEXT())); 2849 2850 uint32_t header_file_offset = (text_section_sp->GetFileAddress() - mapping_offset_value); 2851 2852 offset = local_symbols_info.entriesOffset; 2853 for (uint32_t entry_index = 0; entry_index < local_symbols_info.entriesCount; entry_index++) 2854 { 2855 struct lldb_copy_dyld_cache_local_symbols_entry local_symbols_entry; 2856 local_symbols_entry.dylibOffset = dsc_local_symbols_data.GetU32(&offset); 2857 local_symbols_entry.nlistStartIndex = dsc_local_symbols_data.GetU32(&offset); 2858 local_symbols_entry.nlistCount = dsc_local_symbols_data.GetU32(&offset); 2859 2860 if (header_file_offset == local_symbols_entry.dylibOffset) 2861 { 2862 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2863 2864 // The normal nlist code cannot correctly size the Symbols array, we need to allocate it here. 2865 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms + unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2866 num_syms = symtab->GetNumSymbols(); 2867 2868 nlist_data_offset = local_symbols_info.nlistOffset + (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2869 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2870 2871 for (uint32_t nlist_index = 0; nlist_index < local_symbols_entry.nlistCount; nlist_index++) 2872 { 2873 ///////////////////////////// 2874 { 2875 struct nlist_64 nlist; 2876 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2877 break; 2878 2879 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked(&nlist_data_offset); 2880 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2881 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2882 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked (&nlist_data_offset); 2883 nlist.n_value = dsc_local_symbols_data.GetAddress_unchecked (&nlist_data_offset); 2884 2885 SymbolType type = eSymbolTypeInvalid; 2886 const char *symbol_name = dsc_local_symbols_data.PeekCStr(string_table_offset + nlist.n_strx); 2887 2888 if (symbol_name == NULL) 2889 { 2890 // No symbol should be NULL, even the symbols with no 2891 // string values should have an offset zero which points 2892 // to an empty C-string 2893 Host::SystemLog (Host::eSystemLogError, 2894 "error: DSC unmapped local symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 2895 entry_index, 2896 nlist.n_strx, 2897 module_sp->GetFileSpec().GetPath().c_str()); 2898 continue; 2899 } 2900 if (symbol_name[0] == '\0') 2901 symbol_name = NULL; 2902 2903 const char *symbol_name_non_abi_mangled = NULL; 2904 2905 SectionSP symbol_section; 2906 uint32_t symbol_byte_size = 0; 2907 bool add_nlist = true; 2908 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2909 bool demangled_is_synthesized = false; 2910 bool is_gsym = false; 2911 bool set_value = true; 2912 2913 assert (sym_idx < num_syms); 2914 2915 sym[sym_idx].SetDebug (is_debug); 2916 2917 if (is_debug) 2918 { 2919 switch (nlist.n_type) 2920 { 2921 case N_GSYM: 2922 // global symbol: name,,NO_SECT,type,0 2923 // Sometimes the N_GSYM value contains the address. 2924 2925 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 2926 // have the same address, but we want to ensure that we always find only the real symbol, 2927 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 2928 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 2929 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 2930 // same address. 2931 2932 is_gsym = true; 2933 sym[sym_idx].SetExternal(true); 2934 2935 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') 2936 { 2937 llvm::StringRef symbol_name_ref(symbol_name); 2938 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 2939 { 2940 symbol_name_non_abi_mangled = symbol_name + 1; 2941 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2942 type = eSymbolTypeObjCClass; 2943 demangled_is_synthesized = true; 2944 2945 } 2946 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 2947 { 2948 symbol_name_non_abi_mangled = symbol_name + 1; 2949 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2950 type = eSymbolTypeObjCMetaClass; 2951 demangled_is_synthesized = true; 2952 } 2953 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 2954 { 2955 symbol_name_non_abi_mangled = symbol_name + 1; 2956 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2957 type = eSymbolTypeObjCIVar; 2958 demangled_is_synthesized = true; 2959 } 2960 } 2961 else 2962 { 2963 if (nlist.n_value != 0) 2964 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2965 type = eSymbolTypeData; 2966 } 2967 break; 2968 2969 case N_FNAME: 2970 // procedure name (f77 kludge): name,,NO_SECT,0,0 2971 type = eSymbolTypeCompiler; 2972 break; 2973 2974 case N_FUN: 2975 // procedure: name,,n_sect,linenumber,address 2976 if (symbol_name) 2977 { 2978 type = eSymbolTypeCode; 2979 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2980 2981 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 2982 // We use the current number of symbols in the symbol table in lieu of 2983 // using nlist_idx in case we ever start trimming entries out 2984 N_FUN_indexes.push_back(sym_idx); 2985 } 2986 else 2987 { 2988 type = eSymbolTypeCompiler; 2989 2990 if ( !N_FUN_indexes.empty() ) 2991 { 2992 // Copy the size of the function into the original STAB entry so we don't have 2993 // to hunt for it later 2994 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 2995 N_FUN_indexes.pop_back(); 2996 // We don't really need the end function STAB as it contains the size which 2997 // we already placed with the original symbol, so don't add it if we want a 2998 // minimal symbol table 2999 add_nlist = false; 3000 } 3001 } 3002 break; 3003 3004 case N_STSYM: 3005 // static symbol: name,,n_sect,type,address 3006 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3007 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3008 if (symbol_name && symbol_name[0]) 3009 { 3010 type = ObjectFile::GetSymbolTypeFromName(symbol_name+1, eSymbolTypeData); 3011 } 3012 break; 3013 3014 case N_LCSYM: 3015 // .lcomm symbol: name,,n_sect,type,address 3016 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3017 type = eSymbolTypeCommonBlock; 3018 break; 3019 3020 case N_BNSYM: 3021 // We use the current number of symbols in the symbol table in lieu of 3022 // using nlist_idx in case we ever start trimming entries out 3023 // Skip these if we want minimal symbol tables 3024 add_nlist = false; 3025 break; 3026 3027 case N_ENSYM: 3028 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 3029 // so that we can always skip the entire symbol if we need to navigate 3030 // more quickly at the source level when parsing STABS 3031 // Skip these if we want minimal symbol tables 3032 add_nlist = false; 3033 break; 3034 3035 case N_OPT: 3036 // emitted with gcc2_compiled and in gcc source 3037 type = eSymbolTypeCompiler; 3038 break; 3039 3040 case N_RSYM: 3041 // register sym: name,,NO_SECT,type,register 3042 type = eSymbolTypeVariable; 3043 break; 3044 3045 case N_SLINE: 3046 // src line: 0,,n_sect,linenumber,address 3047 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3048 type = eSymbolTypeLineEntry; 3049 break; 3050 3051 case N_SSYM: 3052 // structure elt: name,,NO_SECT,type,struct_offset 3053 type = eSymbolTypeVariableType; 3054 break; 3055 3056 case N_SO: 3057 // source file name 3058 type = eSymbolTypeSourceFile; 3059 if (symbol_name == NULL) 3060 { 3061 add_nlist = false; 3062 if (N_SO_index != UINT32_MAX) 3063 { 3064 // Set the size of the N_SO to the terminating index of this N_SO 3065 // so that we can always skip the entire N_SO if we need to navigate 3066 // more quickly at the source level when parsing STABS 3067 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3068 symbol_ptr->SetByteSize(sym_idx); 3069 symbol_ptr->SetSizeIsSibling(true); 3070 } 3071 N_NSYM_indexes.clear(); 3072 N_INCL_indexes.clear(); 3073 N_BRAC_indexes.clear(); 3074 N_COMM_indexes.clear(); 3075 N_FUN_indexes.clear(); 3076 N_SO_index = UINT32_MAX; 3077 } 3078 else 3079 { 3080 // We use the current number of symbols in the symbol table in lieu of 3081 // using nlist_idx in case we ever start trimming entries out 3082 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3083 if (N_SO_has_full_path) 3084 { 3085 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3086 { 3087 // We have two consecutive N_SO entries where the first contains a directory 3088 // and the second contains a full path. 3089 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 3090 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3091 add_nlist = false; 3092 } 3093 else 3094 { 3095 // This is the first entry in a N_SO that contains a directory or 3096 // a full path to the source file 3097 N_SO_index = sym_idx; 3098 } 3099 } 3100 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3101 { 3102 // This is usually the second N_SO entry that contains just the filename, 3103 // so here we combine it with the first one if we are minimizing the symbol table 3104 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName(lldb::eLanguageTypeUnknown).AsCString(); 3105 if (so_path && so_path[0]) 3106 { 3107 std::string full_so_path (so_path); 3108 const size_t double_slash_pos = full_so_path.find("//"); 3109 if (double_slash_pos != std::string::npos) 3110 { 3111 // The linker has been generating bad N_SO entries with doubled up paths 3112 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 3113 // and the second is the directory for the source file so you end up with 3114 // a path that looks like "/tmp/src//tmp/src/" 3115 FileSpec so_dir(so_path, false); 3116 if (!so_dir.Exists()) 3117 { 3118 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 3119 if (so_dir.Exists()) 3120 { 3121 // Trim off the incorrect path 3122 full_so_path.erase(0, double_slash_pos + 1); 3123 } 3124 } 3125 } 3126 if (*full_so_path.rbegin() != '/') 3127 full_so_path += '/'; 3128 full_so_path += symbol_name; 3129 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 3130 add_nlist = false; 3131 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3132 } 3133 } 3134 else 3135 { 3136 // This could be a relative path to a N_SO 3137 N_SO_index = sym_idx; 3138 } 3139 } 3140 break; 3141 3142 case N_OSO: 3143 // object file name: name,,0,0,st_mtime 3144 type = eSymbolTypeObjectFile; 3145 break; 3146 3147 case N_LSYM: 3148 // local sym: name,,NO_SECT,type,offset 3149 type = eSymbolTypeLocal; 3150 break; 3151 3152 //---------------------------------------------------------------------- 3153 // INCL scopes 3154 //---------------------------------------------------------------------- 3155 case N_BINCL: 3156 // include file beginning: name,,NO_SECT,0,sum 3157 // We use the current number of symbols in the symbol table in lieu of 3158 // using nlist_idx in case we ever start trimming entries out 3159 N_INCL_indexes.push_back(sym_idx); 3160 type = eSymbolTypeScopeBegin; 3161 break; 3162 3163 case N_EINCL: 3164 // include file end: name,,NO_SECT,0,0 3165 // Set the size of the N_BINCL to the terminating index of this N_EINCL 3166 // so that we can always skip the entire symbol if we need to navigate 3167 // more quickly at the source level when parsing STABS 3168 if ( !N_INCL_indexes.empty() ) 3169 { 3170 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3171 symbol_ptr->SetByteSize(sym_idx + 1); 3172 symbol_ptr->SetSizeIsSibling(true); 3173 N_INCL_indexes.pop_back(); 3174 } 3175 type = eSymbolTypeScopeEnd; 3176 break; 3177 3178 case N_SOL: 3179 // #included file name: name,,n_sect,0,address 3180 type = eSymbolTypeHeaderFile; 3181 3182 // We currently don't use the header files on darwin 3183 add_nlist = false; 3184 break; 3185 3186 case N_PARAMS: 3187 // compiler parameters: name,,NO_SECT,0,0 3188 type = eSymbolTypeCompiler; 3189 break; 3190 3191 case N_VERSION: 3192 // compiler version: name,,NO_SECT,0,0 3193 type = eSymbolTypeCompiler; 3194 break; 3195 3196 case N_OLEVEL: 3197 // compiler -O level: name,,NO_SECT,0,0 3198 type = eSymbolTypeCompiler; 3199 break; 3200 3201 case N_PSYM: 3202 // parameter: name,,NO_SECT,type,offset 3203 type = eSymbolTypeVariable; 3204 break; 3205 3206 case N_ENTRY: 3207 // alternate entry: name,,n_sect,linenumber,address 3208 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3209 type = eSymbolTypeLineEntry; 3210 break; 3211 3212 //---------------------------------------------------------------------- 3213 // Left and Right Braces 3214 //---------------------------------------------------------------------- 3215 case N_LBRAC: 3216 // left bracket: 0,,NO_SECT,nesting level,address 3217 // We use the current number of symbols in the symbol table in lieu of 3218 // using nlist_idx in case we ever start trimming entries out 3219 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3220 N_BRAC_indexes.push_back(sym_idx); 3221 type = eSymbolTypeScopeBegin; 3222 break; 3223 3224 case N_RBRAC: 3225 // right bracket: 0,,NO_SECT,nesting level,address 3226 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 3227 // so that we can always skip the entire symbol if we need to navigate 3228 // more quickly at the source level when parsing STABS 3229 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3230 if ( !N_BRAC_indexes.empty() ) 3231 { 3232 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3233 symbol_ptr->SetByteSize(sym_idx + 1); 3234 symbol_ptr->SetSizeIsSibling(true); 3235 N_BRAC_indexes.pop_back(); 3236 } 3237 type = eSymbolTypeScopeEnd; 3238 break; 3239 3240 case N_EXCL: 3241 // deleted include file: name,,NO_SECT,0,sum 3242 type = eSymbolTypeHeaderFile; 3243 break; 3244 3245 //---------------------------------------------------------------------- 3246 // COMM scopes 3247 //---------------------------------------------------------------------- 3248 case N_BCOMM: 3249 // begin common: name,,NO_SECT,0,0 3250 // We use the current number of symbols in the symbol table in lieu of 3251 // using nlist_idx in case we ever start trimming entries out 3252 type = eSymbolTypeScopeBegin; 3253 N_COMM_indexes.push_back(sym_idx); 3254 break; 3255 3256 case N_ECOML: 3257 // end common (local name): 0,,n_sect,0,address 3258 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3259 // Fall through 3260 3261 case N_ECOMM: 3262 // end common: name,,n_sect,0,0 3263 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 3264 // so that we can always skip the entire symbol if we need to navigate 3265 // more quickly at the source level when parsing STABS 3266 if ( !N_COMM_indexes.empty() ) 3267 { 3268 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 3269 symbol_ptr->SetByteSize(sym_idx + 1); 3270 symbol_ptr->SetSizeIsSibling(true); 3271 N_COMM_indexes.pop_back(); 3272 } 3273 type = eSymbolTypeScopeEnd; 3274 break; 3275 3276 case N_LENG: 3277 // second stab entry with length information 3278 type = eSymbolTypeAdditional; 3279 break; 3280 3281 default: break; 3282 } 3283 } 3284 else 3285 { 3286 //uint8_t n_pext = N_PEXT & nlist.n_type; 3287 uint8_t n_type = N_TYPE & nlist.n_type; 3288 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3289 3290 switch (n_type) 3291 { 3292 case N_INDR: 3293 { 3294 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 3295 if (reexport_name_cstr && reexport_name_cstr[0]) 3296 { 3297 type = eSymbolTypeReExported; 3298 ConstString reexport_name(reexport_name_cstr + ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3299 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3300 set_value = false; 3301 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3302 indirect_symbol_names.insert(ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3303 } 3304 else 3305 type = eSymbolTypeUndefined; 3306 } 3307 break; 3308 3309 case N_UNDF: 3310 if (symbol_name && symbol_name[0]) 3311 { 3312 ConstString undefined_name(symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3313 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3314 } 3315 // Fall through 3316 case N_PBUD: 3317 type = eSymbolTypeUndefined; 3318 break; 3319 3320 case N_ABS: 3321 type = eSymbolTypeAbsolute; 3322 break; 3323 3324 case N_SECT: 3325 { 3326 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3327 3328 if (symbol_section == NULL) 3329 { 3330 // TODO: warn about this? 3331 add_nlist = false; 3332 break; 3333 } 3334 3335 if (TEXT_eh_frame_sectID == nlist.n_sect) 3336 { 3337 type = eSymbolTypeException; 3338 } 3339 else 3340 { 3341 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 3342 3343 switch (section_type) 3344 { 3345 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 3346 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 3347 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 3348 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 3349 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 3350 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 3351 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 3352 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 3353 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 3354 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 3355 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 3356 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 3357 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 3358 default: 3359 switch (symbol_section->GetType()) 3360 { 3361 case lldb::eSectionTypeCode: 3362 type = eSymbolTypeCode; 3363 break; 3364 case eSectionTypeData: 3365 case eSectionTypeDataCString: // Inlined C string data 3366 case eSectionTypeDataCStringPointers: // Pointers to C string data 3367 case eSectionTypeDataSymbolAddress: // Address of a symbol in the symbol table 3368 case eSectionTypeData4: 3369 case eSectionTypeData8: 3370 case eSectionTypeData16: 3371 type = eSymbolTypeData; 3372 break; 3373 default: 3374 break; 3375 } 3376 break; 3377 } 3378 3379 if (type == eSymbolTypeInvalid) 3380 { 3381 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 3382 if (symbol_section->IsDescendant (text_section_sp.get())) 3383 { 3384 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 3385 S_ATTR_SELF_MODIFYING_CODE | 3386 S_ATTR_SOME_INSTRUCTIONS)) 3387 type = eSymbolTypeData; 3388 else 3389 type = eSymbolTypeCode; 3390 } 3391 else if (symbol_section->IsDescendant(data_section_sp.get()) || 3392 symbol_section->IsDescendant(data_dirty_section_sp.get()) || 3393 symbol_section->IsDescendant(data_const_section_sp.get())) 3394 { 3395 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 3396 { 3397 type = eSymbolTypeRuntime; 3398 3399 if (symbol_name) 3400 { 3401 llvm::StringRef symbol_name_ref(symbol_name); 3402 if (symbol_name_ref.startswith("_OBJC_")) 3403 { 3404 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 3405 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 3406 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 3407 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 3408 { 3409 symbol_name_non_abi_mangled = symbol_name + 1; 3410 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3411 type = eSymbolTypeObjCClass; 3412 demangled_is_synthesized = true; 3413 } 3414 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 3415 { 3416 symbol_name_non_abi_mangled = symbol_name + 1; 3417 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3418 type = eSymbolTypeObjCMetaClass; 3419 demangled_is_synthesized = true; 3420 } 3421 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 3422 { 3423 symbol_name_non_abi_mangled = symbol_name + 1; 3424 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3425 type = eSymbolTypeObjCIVar; 3426 demangled_is_synthesized = true; 3427 } 3428 } 3429 } 3430 } 3431 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 3432 { 3433 type = eSymbolTypeException; 3434 } 3435 else 3436 { 3437 type = eSymbolTypeData; 3438 } 3439 } 3440 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 3441 { 3442 type = eSymbolTypeTrampoline; 3443 } 3444 else if (symbol_section->IsDescendant(objc_section_sp.get())) 3445 { 3446 type = eSymbolTypeRuntime; 3447 if (symbol_name && symbol_name[0] == '.') 3448 { 3449 llvm::StringRef symbol_name_ref(symbol_name); 3450 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 3451 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 3452 { 3453 symbol_name_non_abi_mangled = symbol_name; 3454 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 3455 type = eSymbolTypeObjCClass; 3456 demangled_is_synthesized = true; 3457 } 3458 } 3459 } 3460 } 3461 } 3462 } 3463 break; 3464 } 3465 } 3466 3467 if (add_nlist) 3468 { 3469 uint64_t symbol_value = nlist.n_value; 3470 if (symbol_name_non_abi_mangled) 3471 { 3472 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 3473 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 3474 } 3475 else 3476 { 3477 bool symbol_name_is_mangled = false; 3478 3479 if (symbol_name && symbol_name[0] == '_') 3480 { 3481 symbol_name_is_mangled = symbol_name[1] == '_'; 3482 symbol_name++; // Skip the leading underscore 3483 } 3484 3485 if (symbol_name) 3486 { 3487 ConstString const_symbol_name(symbol_name); 3488 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 3489 if (is_gsym && is_debug) 3490 { 3491 const char *gsym_name = sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled).GetCString(); 3492 if (gsym_name) 3493 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3494 } 3495 } 3496 } 3497 if (symbol_section) 3498 { 3499 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3500 if (symbol_byte_size == 0 && function_starts_count > 0) 3501 { 3502 addr_t symbol_lookup_file_addr = nlist.n_value; 3503 // Do an exact address match for non-ARM addresses, else get the closest since 3504 // the symbol might be a thumb symbol which has an address with bit zero set 3505 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 3506 if (is_arm && func_start_entry) 3507 { 3508 // Verify that the function start address is the symbol address (ARM) 3509 // or the symbol address + 1 (thumb) 3510 if (func_start_entry->addr != symbol_lookup_file_addr && 3511 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 3512 { 3513 // Not the right entry, NULL it out... 3514 func_start_entry = NULL; 3515 } 3516 } 3517 if (func_start_entry) 3518 { 3519 func_start_entry->data = true; 3520 3521 addr_t symbol_file_addr = func_start_entry->addr; 3522 uint32_t symbol_flags = 0; 3523 if (is_arm) 3524 { 3525 if (symbol_file_addr & 1) 3526 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3527 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3528 } 3529 3530 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3531 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3532 if (next_func_start_entry) 3533 { 3534 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3535 // Be sure the clear the Thumb address bit when we calculate the size 3536 // from the current and next address 3537 if (is_arm) 3538 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3539 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3540 } 3541 else 3542 { 3543 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3544 } 3545 } 3546 } 3547 symbol_value -= section_file_addr; 3548 } 3549 3550 if (is_debug == false) 3551 { 3552 if (type == eSymbolTypeCode) 3553 { 3554 // See if we can find a N_FUN entry for any code symbols. 3555 // If we do find a match, and the name matches, then we 3556 // can merge the two into just the function symbol to avoid 3557 // duplicate entries in the symbol table 3558 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3559 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3560 if (range.first != range.second) 3561 { 3562 bool found_it = false; 3563 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3564 { 3565 if (sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled)) 3566 { 3567 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3568 // We just need the flags from the linker symbol, so put these flags 3569 // into the N_FUN flags to avoid duplicate symbols in the symbol table 3570 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3571 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3572 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3573 sym[pos->second].SetType (eSymbolTypeResolver); 3574 sym[sym_idx].Clear(); 3575 found_it = true; 3576 break; 3577 } 3578 } 3579 if (found_it) 3580 continue; 3581 } 3582 else 3583 { 3584 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3585 type = eSymbolTypeResolver; 3586 } 3587 } 3588 else if (type == eSymbolTypeData || 3589 type == eSymbolTypeObjCClass || 3590 type == eSymbolTypeObjCMetaClass || 3591 type == eSymbolTypeObjCIVar ) 3592 { 3593 // See if we can find a N_STSYM entry for any data symbols. 3594 // If we do find a match, and the name matches, then we 3595 // can merge the two into just the Static symbol to avoid 3596 // duplicate entries in the symbol table 3597 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3598 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 3599 if (range.first != range.second) 3600 { 3601 bool found_it = false; 3602 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3603 { 3604 if (sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled)) 3605 { 3606 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3607 // We just need the flags from the linker symbol, so put these flags 3608 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3609 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3610 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3611 sym[sym_idx].Clear(); 3612 found_it = true; 3613 break; 3614 } 3615 } 3616 if (found_it) 3617 continue; 3618 } 3619 else 3620 { 3621 const char *gsym_name = sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled).GetCString(); 3622 if (gsym_name) 3623 { 3624 // Combine N_GSYM stab entries with the non stab symbol 3625 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(gsym_name); 3626 if (pos != N_GSYM_name_to_sym_idx.end()) 3627 { 3628 const uint32_t GSYM_sym_idx = pos->second; 3629 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 3630 // Copy the address, because often the N_GSYM address has an invalid address of zero 3631 // when the global is a common symbol 3632 sym[GSYM_sym_idx].GetAddressRef().SetSection (symbol_section); 3633 sym[GSYM_sym_idx].GetAddressRef().SetOffset (symbol_value); 3634 // We just need the flags from the linker symbol, so put these flags 3635 // into the N_GSYM flags to avoid duplicate symbols in the symbol table 3636 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3637 sym[sym_idx].Clear(); 3638 continue; 3639 } 3640 } 3641 } 3642 } 3643 } 3644 3645 sym[sym_idx].SetID (nlist_idx); 3646 sym[sym_idx].SetType (type); 3647 if (set_value) 3648 { 3649 sym[sym_idx].GetAddressRef().SetSection (symbol_section); 3650 sym[sym_idx].GetAddressRef().SetOffset (symbol_value); 3651 } 3652 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3653 3654 if (symbol_byte_size > 0) 3655 sym[sym_idx].SetByteSize(symbol_byte_size); 3656 3657 if (demangled_is_synthesized) 3658 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3659 ++sym_idx; 3660 } 3661 else 3662 { 3663 sym[sym_idx].Clear(); 3664 } 3665 3666 } 3667 ///////////////////////////// 3668 } 3669 break; // No more entries to consider 3670 } 3671 } 3672 3673 for (const auto &pos :reexport_shlib_needs_fixup) 3674 { 3675 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3676 if (undef_pos != undefined_name_to_desc.end()) 3677 { 3678 const uint8_t dylib_ordinal = llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3679 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3680 sym[pos.first].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(dylib_ordinal-1)); 3681 } 3682 } 3683 } 3684 } 3685 } 3686 } 3687 } 3688 3689 // Must reset this in case it was mutated above! 3690 nlist_data_offset = 0; 3691 #endif 3692 3693 if (nlist_data.GetByteSize() > 0) 3694 { 3695 3696 // If the sym array was not created while parsing the DSC unmapped 3697 // symbols, create it now. 3698 if (sym == NULL) 3699 { 3700 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3701 num_syms = symtab->GetNumSymbols(); 3702 } 3703 3704 if (unmapped_local_symbols_found) 3705 { 3706 assert(m_dysymtab.ilocalsym == 0); 3707 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3708 nlist_idx = m_dysymtab.nlocalsym; 3709 } 3710 else 3711 { 3712 nlist_idx = 0; 3713 } 3714 3715 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3716 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3717 UndefinedNameToDescMap undefined_name_to_desc; 3718 SymbolIndexToName reexport_shlib_needs_fixup; 3719 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 3720 { 3721 struct nlist_64 nlist; 3722 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 3723 break; 3724 3725 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3726 nlist.n_type = nlist_data.GetU8_unchecked (&nlist_data_offset); 3727 nlist.n_sect = nlist_data.GetU8_unchecked (&nlist_data_offset); 3728 nlist.n_desc = nlist_data.GetU16_unchecked (&nlist_data_offset); 3729 nlist.n_value = nlist_data.GetAddress_unchecked (&nlist_data_offset); 3730 3731 SymbolType type = eSymbolTypeInvalid; 3732 const char *symbol_name = NULL; 3733 3734 if (have_strtab_data) 3735 { 3736 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3737 3738 if (symbol_name == NULL) 3739 { 3740 // No symbol should be NULL, even the symbols with no 3741 // string values should have an offset zero which points 3742 // to an empty C-string 3743 Host::SystemLog (Host::eSystemLogError, 3744 "error: symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 3745 nlist_idx, 3746 nlist.n_strx, 3747 module_sp->GetFileSpec().GetPath().c_str()); 3748 continue; 3749 } 3750 if (symbol_name[0] == '\0') 3751 symbol_name = NULL; 3752 } 3753 else 3754 { 3755 const addr_t str_addr = strtab_addr + nlist.n_strx; 3756 Error str_error; 3757 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, str_error)) 3758 symbol_name = memory_symbol_name.c_str(); 3759 } 3760 const char *symbol_name_non_abi_mangled = NULL; 3761 3762 SectionSP symbol_section; 3763 lldb::addr_t symbol_byte_size = 0; 3764 bool add_nlist = true; 3765 bool is_gsym = false; 3766 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3767 bool demangled_is_synthesized = false; 3768 bool set_value = true; 3769 assert (sym_idx < num_syms); 3770 3771 sym[sym_idx].SetDebug (is_debug); 3772 3773 if (is_debug) 3774 { 3775 switch (nlist.n_type) 3776 { 3777 case N_GSYM: 3778 // global symbol: name,,NO_SECT,type,0 3779 // Sometimes the N_GSYM value contains the address. 3780 3781 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 3782 // have the same address, but we want to ensure that we always find only the real symbol, 3783 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 3784 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 3785 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 3786 // same address. 3787 is_gsym = true; 3788 sym[sym_idx].SetExternal(true); 3789 3790 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') 3791 { 3792 llvm::StringRef symbol_name_ref(symbol_name); 3793 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 3794 { 3795 symbol_name_non_abi_mangled = symbol_name + 1; 3796 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3797 type = eSymbolTypeObjCClass; 3798 demangled_is_synthesized = true; 3799 3800 } 3801 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 3802 { 3803 symbol_name_non_abi_mangled = symbol_name + 1; 3804 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3805 type = eSymbolTypeObjCMetaClass; 3806 demangled_is_synthesized = true; 3807 } 3808 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 3809 { 3810 symbol_name_non_abi_mangled = symbol_name + 1; 3811 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3812 type = eSymbolTypeObjCIVar; 3813 demangled_is_synthesized = true; 3814 } 3815 } 3816 else 3817 { 3818 if (nlist.n_value != 0) 3819 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3820 type = eSymbolTypeData; 3821 } 3822 break; 3823 3824 case N_FNAME: 3825 // procedure name (f77 kludge): name,,NO_SECT,0,0 3826 type = eSymbolTypeCompiler; 3827 break; 3828 3829 case N_FUN: 3830 // procedure: name,,n_sect,linenumber,address 3831 if (symbol_name) 3832 { 3833 type = eSymbolTypeCode; 3834 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3835 3836 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3837 // We use the current number of symbols in the symbol table in lieu of 3838 // using nlist_idx in case we ever start trimming entries out 3839 N_FUN_indexes.push_back(sym_idx); 3840 } 3841 else 3842 { 3843 type = eSymbolTypeCompiler; 3844 3845 if ( !N_FUN_indexes.empty() ) 3846 { 3847 // Copy the size of the function into the original STAB entry so we don't have 3848 // to hunt for it later 3849 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 3850 N_FUN_indexes.pop_back(); 3851 // We don't really need the end function STAB as it contains the size which 3852 // we already placed with the original symbol, so don't add it if we want a 3853 // minimal symbol table 3854 add_nlist = false; 3855 } 3856 } 3857 break; 3858 3859 case N_STSYM: 3860 // static symbol: name,,n_sect,type,address 3861 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3862 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3863 if (symbol_name && symbol_name[0]) 3864 { 3865 type = ObjectFile::GetSymbolTypeFromName(symbol_name+1, eSymbolTypeData); 3866 } 3867 break; 3868 3869 case N_LCSYM: 3870 // .lcomm symbol: name,,n_sect,type,address 3871 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3872 type = eSymbolTypeCommonBlock; 3873 break; 3874 3875 case N_BNSYM: 3876 // We use the current number of symbols in the symbol table in lieu of 3877 // using nlist_idx in case we ever start trimming entries out 3878 // Skip these if we want minimal symbol tables 3879 add_nlist = false; 3880 break; 3881 3882 case N_ENSYM: 3883 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 3884 // so that we can always skip the entire symbol if we need to navigate 3885 // more quickly at the source level when parsing STABS 3886 // Skip these if we want minimal symbol tables 3887 add_nlist = false; 3888 break; 3889 3890 3891 case N_OPT: 3892 // emitted with gcc2_compiled and in gcc source 3893 type = eSymbolTypeCompiler; 3894 break; 3895 3896 case N_RSYM: 3897 // register sym: name,,NO_SECT,type,register 3898 type = eSymbolTypeVariable; 3899 break; 3900 3901 case N_SLINE: 3902 // src line: 0,,n_sect,linenumber,address 3903 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3904 type = eSymbolTypeLineEntry; 3905 break; 3906 3907 case N_SSYM: 3908 // structure elt: name,,NO_SECT,type,struct_offset 3909 type = eSymbolTypeVariableType; 3910 break; 3911 3912 case N_SO: 3913 // source file name 3914 type = eSymbolTypeSourceFile; 3915 if (symbol_name == NULL) 3916 { 3917 add_nlist = false; 3918 if (N_SO_index != UINT32_MAX) 3919 { 3920 // Set the size of the N_SO to the terminating index of this N_SO 3921 // so that we can always skip the entire N_SO if we need to navigate 3922 // more quickly at the source level when parsing STABS 3923 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3924 symbol_ptr->SetByteSize(sym_idx); 3925 symbol_ptr->SetSizeIsSibling(true); 3926 } 3927 N_NSYM_indexes.clear(); 3928 N_INCL_indexes.clear(); 3929 N_BRAC_indexes.clear(); 3930 N_COMM_indexes.clear(); 3931 N_FUN_indexes.clear(); 3932 N_SO_index = UINT32_MAX; 3933 } 3934 else 3935 { 3936 // We use the current number of symbols in the symbol table in lieu of 3937 // using nlist_idx in case we ever start trimming entries out 3938 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3939 if (N_SO_has_full_path) 3940 { 3941 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3942 { 3943 // We have two consecutive N_SO entries where the first contains a directory 3944 // and the second contains a full path. 3945 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 3946 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3947 add_nlist = false; 3948 } 3949 else 3950 { 3951 // This is the first entry in a N_SO that contains a directory or 3952 // a full path to the source file 3953 N_SO_index = sym_idx; 3954 } 3955 } 3956 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3957 { 3958 // This is usually the second N_SO entry that contains just the filename, 3959 // so here we combine it with the first one if we are minimizing the symbol table 3960 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName(lldb::eLanguageTypeUnknown).AsCString(); 3961 if (so_path && so_path[0]) 3962 { 3963 std::string full_so_path (so_path); 3964 const size_t double_slash_pos = full_so_path.find("//"); 3965 if (double_slash_pos != std::string::npos) 3966 { 3967 // The linker has been generating bad N_SO entries with doubled up paths 3968 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 3969 // and the second is the directory for the source file so you end up with 3970 // a path that looks like "/tmp/src//tmp/src/" 3971 FileSpec so_dir(so_path, false); 3972 if (!so_dir.Exists()) 3973 { 3974 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 3975 if (so_dir.Exists()) 3976 { 3977 // Trim off the incorrect path 3978 full_so_path.erase(0, double_slash_pos + 1); 3979 } 3980 } 3981 } 3982 if (*full_so_path.rbegin() != '/') 3983 full_so_path += '/'; 3984 full_so_path += symbol_name; 3985 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 3986 add_nlist = false; 3987 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3988 } 3989 } 3990 else 3991 { 3992 // This could be a relative path to a N_SO 3993 N_SO_index = sym_idx; 3994 } 3995 } 3996 break; 3997 3998 case N_OSO: 3999 // object file name: name,,0,0,st_mtime 4000 type = eSymbolTypeObjectFile; 4001 break; 4002 4003 case N_LSYM: 4004 // local sym: name,,NO_SECT,type,offset 4005 type = eSymbolTypeLocal; 4006 break; 4007 4008 //---------------------------------------------------------------------- 4009 // INCL scopes 4010 //---------------------------------------------------------------------- 4011 case N_BINCL: 4012 // include file beginning: name,,NO_SECT,0,sum 4013 // We use the current number of symbols in the symbol table in lieu of 4014 // using nlist_idx in case we ever start trimming entries out 4015 N_INCL_indexes.push_back(sym_idx); 4016 type = eSymbolTypeScopeBegin; 4017 break; 4018 4019 case N_EINCL: 4020 // include file end: name,,NO_SECT,0,0 4021 // Set the size of the N_BINCL to the terminating index of this N_EINCL 4022 // so that we can always skip the entire symbol if we need to navigate 4023 // more quickly at the source level when parsing STABS 4024 if ( !N_INCL_indexes.empty() ) 4025 { 4026 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4027 symbol_ptr->SetByteSize(sym_idx + 1); 4028 symbol_ptr->SetSizeIsSibling(true); 4029 N_INCL_indexes.pop_back(); 4030 } 4031 type = eSymbolTypeScopeEnd; 4032 break; 4033 4034 case N_SOL: 4035 // #included file name: name,,n_sect,0,address 4036 type = eSymbolTypeHeaderFile; 4037 4038 // We currently don't use the header files on darwin 4039 add_nlist = false; 4040 break; 4041 4042 case N_PARAMS: 4043 // compiler parameters: name,,NO_SECT,0,0 4044 type = eSymbolTypeCompiler; 4045 break; 4046 4047 case N_VERSION: 4048 // compiler version: name,,NO_SECT,0,0 4049 type = eSymbolTypeCompiler; 4050 break; 4051 4052 case N_OLEVEL: 4053 // compiler -O level: name,,NO_SECT,0,0 4054 type = eSymbolTypeCompiler; 4055 break; 4056 4057 case N_PSYM: 4058 // parameter: name,,NO_SECT,type,offset 4059 type = eSymbolTypeVariable; 4060 break; 4061 4062 case N_ENTRY: 4063 // alternate entry: name,,n_sect,linenumber,address 4064 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 4065 type = eSymbolTypeLineEntry; 4066 break; 4067 4068 //---------------------------------------------------------------------- 4069 // Left and Right Braces 4070 //---------------------------------------------------------------------- 4071 case N_LBRAC: 4072 // left bracket: 0,,NO_SECT,nesting level,address 4073 // We use the current number of symbols in the symbol table in lieu of 4074 // using nlist_idx in case we ever start trimming entries out 4075 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 4076 N_BRAC_indexes.push_back(sym_idx); 4077 type = eSymbolTypeScopeBegin; 4078 break; 4079 4080 case N_RBRAC: 4081 // right bracket: 0,,NO_SECT,nesting level,address 4082 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 4083 // so that we can always skip the entire symbol if we need to navigate 4084 // more quickly at the source level when parsing STABS 4085 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 4086 if ( !N_BRAC_indexes.empty() ) 4087 { 4088 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4089 symbol_ptr->SetByteSize(sym_idx + 1); 4090 symbol_ptr->SetSizeIsSibling(true); 4091 N_BRAC_indexes.pop_back(); 4092 } 4093 type = eSymbolTypeScopeEnd; 4094 break; 4095 4096 case N_EXCL: 4097 // deleted include file: name,,NO_SECT,0,sum 4098 type = eSymbolTypeHeaderFile; 4099 break; 4100 4101 //---------------------------------------------------------------------- 4102 // COMM scopes 4103 //---------------------------------------------------------------------- 4104 case N_BCOMM: 4105 // begin common: name,,NO_SECT,0,0 4106 // We use the current number of symbols in the symbol table in lieu of 4107 // using nlist_idx in case we ever start trimming entries out 4108 type = eSymbolTypeScopeBegin; 4109 N_COMM_indexes.push_back(sym_idx); 4110 break; 4111 4112 case N_ECOML: 4113 // end common (local name): 0,,n_sect,0,address 4114 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 4115 LLVM_FALLTHROUGH; 4116 4117 case N_ECOMM: 4118 // end common: name,,n_sect,0,0 4119 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 4120 // so that we can always skip the entire symbol if we need to navigate 4121 // more quickly at the source level when parsing STABS 4122 if ( !N_COMM_indexes.empty() ) 4123 { 4124 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4125 symbol_ptr->SetByteSize(sym_idx + 1); 4126 symbol_ptr->SetSizeIsSibling(true); 4127 N_COMM_indexes.pop_back(); 4128 } 4129 type = eSymbolTypeScopeEnd; 4130 break; 4131 4132 case N_LENG: 4133 // second stab entry with length information 4134 type = eSymbolTypeAdditional; 4135 break; 4136 4137 default: break; 4138 } 4139 } 4140 else 4141 { 4142 //uint8_t n_pext = N_PEXT & nlist.n_type; 4143 uint8_t n_type = N_TYPE & nlist.n_type; 4144 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4145 4146 switch (n_type) 4147 { 4148 case N_INDR: 4149 { 4150 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4151 if (reexport_name_cstr && reexport_name_cstr[0]) 4152 { 4153 type = eSymbolTypeReExported; 4154 ConstString reexport_name(reexport_name_cstr + ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4155 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4156 set_value = false; 4157 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4158 indirect_symbol_names.insert(ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4159 } 4160 else 4161 type = eSymbolTypeUndefined; 4162 } 4163 break; 4164 4165 case N_UNDF: 4166 if (symbol_name && symbol_name[0]) 4167 { 4168 ConstString undefined_name(symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 4169 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4170 } 4171 LLVM_FALLTHROUGH; 4172 4173 case N_PBUD: 4174 type = eSymbolTypeUndefined; 4175 break; 4176 4177 case N_ABS: 4178 type = eSymbolTypeAbsolute; 4179 break; 4180 4181 case N_SECT: 4182 { 4183 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 4184 4185 if (!symbol_section) 4186 { 4187 // TODO: warn about this? 4188 add_nlist = false; 4189 break; 4190 } 4191 4192 if (TEXT_eh_frame_sectID == nlist.n_sect) 4193 { 4194 type = eSymbolTypeException; 4195 } 4196 else 4197 { 4198 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4199 4200 switch (section_type) 4201 { 4202 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 4203 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 4204 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 4205 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 4206 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 4207 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 4208 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 4209 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 4210 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 4211 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 4212 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 4213 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 4214 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 4215 default: 4216 switch (symbol_section->GetType()) 4217 { 4218 case lldb::eSectionTypeCode: 4219 type = eSymbolTypeCode; 4220 break; 4221 case eSectionTypeData: 4222 case eSectionTypeDataCString: // Inlined C string data 4223 case eSectionTypeDataCStringPointers: // Pointers to C string data 4224 case eSectionTypeDataSymbolAddress: // Address of a symbol in the symbol table 4225 case eSectionTypeData4: 4226 case eSectionTypeData8: 4227 case eSectionTypeData16: 4228 type = eSymbolTypeData; 4229 break; 4230 default: 4231 break; 4232 } 4233 break; 4234 } 4235 4236 if (type == eSymbolTypeInvalid) 4237 { 4238 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 4239 if (symbol_section->IsDescendant (text_section_sp.get())) 4240 { 4241 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4242 S_ATTR_SELF_MODIFYING_CODE | 4243 S_ATTR_SOME_INSTRUCTIONS)) 4244 type = eSymbolTypeData; 4245 else 4246 type = eSymbolTypeCode; 4247 } 4248 else 4249 if (symbol_section->IsDescendant(data_section_sp.get()) || 4250 symbol_section->IsDescendant(data_dirty_section_sp.get()) || 4251 symbol_section->IsDescendant(data_const_section_sp.get())) 4252 { 4253 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 4254 { 4255 type = eSymbolTypeRuntime; 4256 4257 if (symbol_name) 4258 { 4259 llvm::StringRef symbol_name_ref(symbol_name); 4260 if (symbol_name_ref.startswith("_OBJC_")) 4261 { 4262 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 4263 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 4264 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 4265 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 4266 { 4267 symbol_name_non_abi_mangled = symbol_name + 1; 4268 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 4269 type = eSymbolTypeObjCClass; 4270 demangled_is_synthesized = true; 4271 } 4272 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 4273 { 4274 symbol_name_non_abi_mangled = symbol_name + 1; 4275 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 4276 type = eSymbolTypeObjCMetaClass; 4277 demangled_is_synthesized = true; 4278 } 4279 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 4280 { 4281 symbol_name_non_abi_mangled = symbol_name + 1; 4282 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 4283 type = eSymbolTypeObjCIVar; 4284 demangled_is_synthesized = true; 4285 } 4286 } 4287 } 4288 } 4289 else 4290 if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 4291 { 4292 type = eSymbolTypeException; 4293 } 4294 else 4295 { 4296 type = eSymbolTypeData; 4297 } 4298 } 4299 else 4300 if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 4301 { 4302 type = eSymbolTypeTrampoline; 4303 } 4304 else 4305 if (symbol_section->IsDescendant(objc_section_sp.get())) 4306 { 4307 type = eSymbolTypeRuntime; 4308 if (symbol_name && symbol_name[0] == '.') 4309 { 4310 llvm::StringRef symbol_name_ref(symbol_name); 4311 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 4312 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 4313 { 4314 symbol_name_non_abi_mangled = symbol_name; 4315 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4316 type = eSymbolTypeObjCClass; 4317 demangled_is_synthesized = true; 4318 } 4319 } 4320 } 4321 } 4322 } 4323 } 4324 break; 4325 } 4326 } 4327 4328 if (add_nlist) 4329 { 4330 uint64_t symbol_value = nlist.n_value; 4331 4332 if (symbol_name_non_abi_mangled) 4333 { 4334 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 4335 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 4336 } 4337 else 4338 { 4339 bool symbol_name_is_mangled = false; 4340 4341 if (symbol_name && symbol_name[0] == '_') 4342 { 4343 symbol_name_is_mangled = symbol_name[1] == '_'; 4344 symbol_name++; // Skip the leading underscore 4345 } 4346 4347 if (symbol_name) 4348 { 4349 ConstString const_symbol_name(symbol_name); 4350 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 4351 } 4352 } 4353 4354 if (is_gsym) 4355 { 4356 const char *gsym_name = sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled).GetCString(); 4357 if (gsym_name) 4358 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4359 } 4360 4361 if (symbol_section) 4362 { 4363 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4364 if (symbol_byte_size == 0 && function_starts_count > 0) 4365 { 4366 addr_t symbol_lookup_file_addr = nlist.n_value; 4367 // Do an exact address match for non-ARM addresses, else get the closest since 4368 // the symbol might be a thumb symbol which has an address with bit zero set 4369 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 4370 if (is_arm && func_start_entry) 4371 { 4372 // Verify that the function start address is the symbol address (ARM) 4373 // or the symbol address + 1 (thumb) 4374 if (func_start_entry->addr != symbol_lookup_file_addr && 4375 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 4376 { 4377 // Not the right entry, NULL it out... 4378 func_start_entry = NULL; 4379 } 4380 } 4381 if (func_start_entry) 4382 { 4383 func_start_entry->data = true; 4384 4385 addr_t symbol_file_addr = func_start_entry->addr; 4386 if (is_arm) 4387 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4388 4389 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 4390 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 4391 if (next_func_start_entry) 4392 { 4393 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4394 // Be sure the clear the Thumb address bit when we calculate the size 4395 // from the current and next address 4396 if (is_arm) 4397 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4398 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 4399 } 4400 else 4401 { 4402 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4403 } 4404 } 4405 } 4406 symbol_value -= section_file_addr; 4407 } 4408 4409 if (is_debug == false) 4410 { 4411 if (type == eSymbolTypeCode) 4412 { 4413 // See if we can find a N_FUN entry for any code symbols. 4414 // If we do find a match, and the name matches, then we 4415 // can merge the two into just the function symbol to avoid 4416 // duplicate entries in the symbol table 4417 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 4418 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4419 if (range.first != range.second) 4420 { 4421 bool found_it = false; 4422 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 4423 { 4424 if (sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled)) 4425 { 4426 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4427 // We just need the flags from the linker symbol, so put these flags 4428 // into the N_FUN flags to avoid duplicate symbols in the symbol table 4429 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4430 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4431 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 4432 sym[pos->second].SetType (eSymbolTypeResolver); 4433 sym[sym_idx].Clear(); 4434 found_it = true; 4435 break; 4436 } 4437 } 4438 if (found_it) 4439 continue; 4440 } 4441 else 4442 { 4443 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 4444 type = eSymbolTypeResolver; 4445 } 4446 } 4447 else if (type == eSymbolTypeData || 4448 type == eSymbolTypeObjCClass || 4449 type == eSymbolTypeObjCMetaClass || 4450 type == eSymbolTypeObjCIVar ) 4451 { 4452 // See if we can find a N_STSYM entry for any data symbols. 4453 // If we do find a match, and the name matches, then we 4454 // can merge the two into just the Static symbol to avoid 4455 // duplicate entries in the symbol table 4456 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 4457 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4458 if (range.first != range.second) 4459 { 4460 bool found_it = false; 4461 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 4462 { 4463 if (sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled)) 4464 { 4465 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4466 // We just need the flags from the linker symbol, so put these flags 4467 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 4468 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4469 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4470 sym[sym_idx].Clear(); 4471 found_it = true; 4472 break; 4473 } 4474 } 4475 if (found_it) 4476 continue; 4477 } 4478 else 4479 { 4480 // Combine N_GSYM stab entries with the non stab symbol 4481 const char *gsym_name = sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled).GetCString(); 4482 if (gsym_name) 4483 { 4484 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(gsym_name); 4485 if (pos != N_GSYM_name_to_sym_idx.end()) 4486 { 4487 const uint32_t GSYM_sym_idx = pos->second; 4488 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4489 // Copy the address, because often the N_GSYM address has an invalid address of zero 4490 // when the global is a common symbol 4491 sym[GSYM_sym_idx].GetAddressRef().SetSection (symbol_section); 4492 sym[GSYM_sym_idx].GetAddressRef().SetOffset (symbol_value); 4493 // We just need the flags from the linker symbol, so put these flags 4494 // into the N_GSYM flags to avoid duplicate symbols in the symbol table 4495 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4496 sym[sym_idx].Clear(); 4497 continue; 4498 } 4499 } 4500 } 4501 } 4502 } 4503 4504 sym[sym_idx].SetID (nlist_idx); 4505 sym[sym_idx].SetType (type); 4506 if (set_value) 4507 { 4508 sym[sym_idx].GetAddressRef().SetSection (symbol_section); 4509 sym[sym_idx].GetAddressRef().SetOffset (symbol_value); 4510 } 4511 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4512 4513 if (symbol_byte_size > 0) 4514 sym[sym_idx].SetByteSize(symbol_byte_size); 4515 4516 if (demangled_is_synthesized) 4517 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4518 4519 ++sym_idx; 4520 } 4521 else 4522 { 4523 sym[sym_idx].Clear(); 4524 } 4525 } 4526 4527 for (const auto &pos :reexport_shlib_needs_fixup) 4528 { 4529 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4530 if (undef_pos != undefined_name_to_desc.end()) 4531 { 4532 const uint8_t dylib_ordinal = llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4533 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4534 sym[pos.first].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(dylib_ordinal-1)); 4535 } 4536 } 4537 } 4538 4539 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4540 4541 if (function_starts_count > 0) 4542 { 4543 uint32_t num_synthetic_function_symbols = 0; 4544 for (i=0; i<function_starts_count; ++i) 4545 { 4546 if (function_starts.GetEntryRef (i).data == false) 4547 ++num_synthetic_function_symbols; 4548 } 4549 4550 if (num_synthetic_function_symbols > 0) 4551 { 4552 if (num_syms < sym_idx + num_synthetic_function_symbols) 4553 { 4554 num_syms = sym_idx + num_synthetic_function_symbols; 4555 sym = symtab->Resize (num_syms); 4556 } 4557 for (i=0; i<function_starts_count; ++i) 4558 { 4559 const FunctionStarts::Entry *func_start_entry = function_starts.GetEntryAtIndex (i); 4560 if (func_start_entry->data == false) 4561 { 4562 addr_t symbol_file_addr = func_start_entry->addr; 4563 uint32_t symbol_flags = 0; 4564 if (is_arm) 4565 { 4566 if (symbol_file_addr & 1) 4567 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4568 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4569 } 4570 Address symbol_addr; 4571 if (module_sp->ResolveFileAddress (symbol_file_addr, symbol_addr)) 4572 { 4573 SectionSP symbol_section (symbol_addr.GetSection()); 4574 uint32_t symbol_byte_size = 0; 4575 if (symbol_section) 4576 { 4577 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4578 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 4579 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 4580 if (next_func_start_entry) 4581 { 4582 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4583 if (is_arm) 4584 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4585 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 4586 } 4587 else 4588 { 4589 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4590 } 4591 sym[sym_idx].SetID (synthetic_sym_id++); 4592 sym[sym_idx].GetMangled().SetDemangledName(GetNextSyntheticSymbolName()); 4593 sym[sym_idx].SetType (eSymbolTypeCode); 4594 sym[sym_idx].SetIsSynthetic (true); 4595 sym[sym_idx].GetAddressRef() = symbol_addr; 4596 if (symbol_flags) 4597 sym[sym_idx].SetFlags (symbol_flags); 4598 if (symbol_byte_size) 4599 sym[sym_idx].SetByteSize (symbol_byte_size); 4600 ++sym_idx; 4601 } 4602 } 4603 } 4604 } 4605 } 4606 } 4607 4608 // Trim our symbols down to just what we ended up with after 4609 // removing any symbols. 4610 if (sym_idx < num_syms) 4611 { 4612 num_syms = sym_idx; 4613 sym = symtab->Resize (num_syms); 4614 } 4615 4616 // Now synthesize indirect symbols 4617 if (m_dysymtab.nindirectsyms != 0) 4618 { 4619 if (indirect_symbol_index_data.GetByteSize()) 4620 { 4621 NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end(); 4622 4623 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx) 4624 { 4625 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == S_SYMBOL_STUBS) 4626 { 4627 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4628 if (symbol_stub_byte_size == 0) 4629 continue; 4630 4631 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4632 4633 if (num_symbol_stubs == 0) 4634 continue; 4635 4636 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1; 4637 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) 4638 { 4639 const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx; 4640 const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size); 4641 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4642 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4)) 4643 { 4644 const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset); 4645 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4646 continue; 4647 4648 NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id); 4649 Symbol *stub_symbol = NULL; 4650 if (index_pos != end_index_pos) 4651 { 4652 // We have a remapping from the original nlist index to 4653 // a current symbol index, so just look this up by index 4654 stub_symbol = symtab->SymbolAtIndex (index_pos->second); 4655 } 4656 else 4657 { 4658 // We need to lookup a symbol using the original nlist 4659 // symbol index since this index is coming from the 4660 // S_SYMBOL_STUBS 4661 stub_symbol = symtab->FindSymbolByID (stub_sym_id); 4662 } 4663 4664 if (stub_symbol) 4665 { 4666 Address so_addr(symbol_stub_addr, section_list); 4667 4668 if (stub_symbol->GetType() == eSymbolTypeUndefined) 4669 { 4670 // Change the external symbol into a trampoline that makes sense 4671 // These symbols were N_UNDF N_EXT, and are useless to us, so we 4672 // can re-use them so we don't have to make up a synthetic symbol 4673 // for no good reason. 4674 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 4675 stub_symbol->SetType (eSymbolTypeTrampoline); 4676 else 4677 stub_symbol->SetType (eSymbolTypeResolver); 4678 stub_symbol->SetExternal (false); 4679 stub_symbol->GetAddressRef() = so_addr; 4680 stub_symbol->SetByteSize (symbol_stub_byte_size); 4681 } 4682 else 4683 { 4684 // Make a synthetic symbol to describe the trampoline stub 4685 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4686 if (sym_idx >= num_syms) 4687 { 4688 sym = symtab->Resize (++num_syms); 4689 stub_symbol = NULL; // this pointer no longer valid 4690 } 4691 sym[sym_idx].SetID (synthetic_sym_id++); 4692 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4693 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 4694 sym[sym_idx].SetType (eSymbolTypeTrampoline); 4695 else 4696 sym[sym_idx].SetType (eSymbolTypeResolver); 4697 sym[sym_idx].SetIsSynthetic (true); 4698 sym[sym_idx].GetAddressRef() = so_addr; 4699 sym[sym_idx].SetByteSize (symbol_stub_byte_size); 4700 ++sym_idx; 4701 } 4702 } 4703 else 4704 { 4705 if (log) 4706 log->Warning ("symbol stub referencing symbol table symbol %u that isn't in our minimal symbol table, fix this!!!", stub_sym_id); 4707 } 4708 } 4709 } 4710 } 4711 } 4712 } 4713 } 4714 4715 if (!trie_entries.empty()) 4716 { 4717 for (const auto &e : trie_entries) 4718 { 4719 if (e.entry.import_name) 4720 { 4721 // Only add indirect symbols from the Trie entries if we 4722 // didn't have a N_INDR nlist entry for this already 4723 if (indirect_symbol_names.find(e.entry.name) == indirect_symbol_names.end()) 4724 { 4725 // Make a synthetic symbol to describe re-exported symbol. 4726 if (sym_idx >= num_syms) 4727 sym = symtab->Resize (++num_syms); 4728 sym[sym_idx].SetID (synthetic_sym_id++); 4729 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4730 sym[sym_idx].SetType (eSymbolTypeReExported); 4731 sym[sym_idx].SetIsSynthetic (true); 4732 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4733 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) 4734 { 4735 sym[sym_idx].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(e.entry.other-1)); 4736 } 4737 ++sym_idx; 4738 } 4739 } 4740 } 4741 } 4742 4743 // StreamFile s(stdout, false); 4744 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4745 // symtab->Dump(&s, NULL, eSortOrderNone); 4746 // Set symbol byte sizes correctly since mach-o nlist entries don't have sizes 4747 symtab->CalculateSymbolSizes(); 4748 4749 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4750 // symtab->Dump(&s, NULL, eSortOrderNone); 4751 4752 return symtab->GetNumSymbols(); 4753 } 4754 return 0; 4755 } 4756 4757 void 4758 ObjectFileMachO::Dump (Stream *s) 4759 { 4760 ModuleSP module_sp(GetModule()); 4761 if (module_sp) 4762 { 4763 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4764 s->Printf("%p: ", static_cast<void*>(this)); 4765 s->Indent(); 4766 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4767 s->PutCString("ObjectFileMachO64"); 4768 else 4769 s->PutCString("ObjectFileMachO32"); 4770 4771 ArchSpec header_arch; 4772 GetArchitecture(header_arch); 4773 4774 *s << ", file = '" << m_file << "', arch = " << header_arch.GetArchitectureName() << "\n"; 4775 4776 SectionList *sections = GetSectionList(); 4777 if (sections) 4778 sections->Dump(s, NULL, true, UINT32_MAX); 4779 4780 if (m_symtab_ap.get()) 4781 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4782 } 4783 } 4784 4785 bool 4786 ObjectFileMachO::GetUUID (const llvm::MachO::mach_header &header, 4787 const lldb_private::DataExtractor &data, 4788 lldb::offset_t lc_offset, 4789 lldb_private::UUID& uuid) 4790 { 4791 uint32_t i; 4792 struct uuid_command load_cmd; 4793 4794 lldb::offset_t offset = lc_offset; 4795 for (i=0; i<header.ncmds; ++i) 4796 { 4797 const lldb::offset_t cmd_offset = offset; 4798 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4799 break; 4800 4801 if (load_cmd.cmd == LC_UUID) 4802 { 4803 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4804 4805 if (uuid_bytes) 4806 { 4807 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4808 // We pretend these object files have no UUID to prevent crashing. 4809 4810 const uint8_t opencl_uuid[] = { 0x8c, 0x8e, 0xb3, 0x9b, 4811 0x3b, 0xa8, 4812 0x4b, 0x16, 4813 0xb6, 0xa4, 4814 0x27, 0x63, 0xbb, 0x14, 0xf0, 0x0d }; 4815 4816 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4817 return false; 4818 4819 uuid.SetBytes (uuid_bytes); 4820 return true; 4821 } 4822 return false; 4823 } 4824 offset = cmd_offset + load_cmd.cmdsize; 4825 } 4826 return false; 4827 } 4828 4829 bool 4830 ObjectFileMachO::GetArchitecture (const llvm::MachO::mach_header &header, 4831 const lldb_private::DataExtractor &data, 4832 lldb::offset_t lc_offset, 4833 ArchSpec &arch) 4834 { 4835 arch.SetArchitecture (eArchTypeMachO, header.cputype, header.cpusubtype); 4836 4837 if (arch.IsValid()) 4838 { 4839 llvm::Triple &triple = arch.GetTriple(); 4840 4841 // Set OS to an unspecified unknown or a "*" so it can match any OS 4842 triple.setOS(llvm::Triple::UnknownOS); 4843 triple.setOSName(llvm::StringRef()); 4844 4845 if (header.filetype == MH_PRELOAD) 4846 { 4847 if (header.cputype == CPU_TYPE_ARM) 4848 { 4849 // If this is a 32-bit arm binary, and it's a standalone binary, 4850 // force the Vendor to Apple so we don't accidentally pick up 4851 // the generic armv7 ABI at runtime. Apple's armv7 ABI always uses 4852 // r7 for the frame pointer register; most other armv7 ABIs use a 4853 // combination of r7 and r11. 4854 triple.setVendor(llvm::Triple::Apple); 4855 } 4856 else 4857 { 4858 // Set vendor to an unspecified unknown or a "*" so it can match any vendor 4859 // This is required for correct behavior of EFI debugging on x86_64 4860 triple.setVendor(llvm::Triple::UnknownVendor); 4861 triple.setVendorName(llvm::StringRef()); 4862 } 4863 return true; 4864 } 4865 else 4866 { 4867 struct load_command load_cmd; 4868 4869 lldb::offset_t offset = lc_offset; 4870 for (uint32_t i=0; i<header.ncmds; ++i) 4871 { 4872 const lldb::offset_t cmd_offset = offset; 4873 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4874 break; 4875 4876 switch (load_cmd.cmd) 4877 { 4878 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4879 triple.setOS (llvm::Triple::IOS); 4880 return true; 4881 4882 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4883 triple.setOS (llvm::Triple::MacOSX); 4884 return true; 4885 4886 case llvm::MachO::LC_VERSION_MIN_TVOS: 4887 triple.setOS (llvm::Triple::TvOS); 4888 return true; 4889 4890 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4891 triple.setOS (llvm::Triple::WatchOS); 4892 return true; 4893 4894 default: 4895 break; 4896 } 4897 4898 offset = cmd_offset + load_cmd.cmdsize; 4899 } 4900 4901 if (header.filetype != MH_KEXT_BUNDLE) 4902 { 4903 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 4904 // so lets not say our Vendor is Apple, leave it as an unspecified unknown 4905 triple.setVendor(llvm::Triple::UnknownVendor); 4906 triple.setVendorName(llvm::StringRef()); 4907 } 4908 } 4909 } 4910 return arch.IsValid(); 4911 } 4912 4913 bool 4914 ObjectFileMachO::GetUUID (lldb_private::UUID* uuid) 4915 { 4916 ModuleSP module_sp(GetModule()); 4917 if (module_sp) 4918 { 4919 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4920 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4921 return GetUUID (m_header, m_data, offset, *uuid); 4922 } 4923 return false; 4924 } 4925 4926 uint32_t 4927 ObjectFileMachO::GetDependentModules (FileSpecList& files) 4928 { 4929 uint32_t count = 0; 4930 ModuleSP module_sp(GetModule()); 4931 if (module_sp) 4932 { 4933 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4934 struct load_command load_cmd; 4935 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4936 std::vector<std::string> rpath_paths; 4937 std::vector<std::string> rpath_relative_paths; 4938 const bool resolve_path = false; // Don't resolve the dependent file paths since they may not reside on this system 4939 uint32_t i; 4940 for (i=0; i<m_header.ncmds; ++i) 4941 { 4942 const uint32_t cmd_offset = offset; 4943 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4944 break; 4945 4946 switch (load_cmd.cmd) 4947 { 4948 case LC_RPATH: 4949 case LC_LOAD_DYLIB: 4950 case LC_LOAD_WEAK_DYLIB: 4951 case LC_REEXPORT_DYLIB: 4952 case LC_LOAD_DYLINKER: 4953 case LC_LOADFVMLIB: 4954 case LC_LOAD_UPWARD_DYLIB: 4955 { 4956 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 4957 const char *path = m_data.PeekCStr(name_offset); 4958 if (path) 4959 { 4960 if (load_cmd.cmd == LC_RPATH) 4961 rpath_paths.push_back(path); 4962 else 4963 { 4964 if (path[0] == '@') 4965 { 4966 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 4967 rpath_relative_paths.push_back(path + strlen("@rpath")); 4968 } 4969 else 4970 { 4971 FileSpec file_spec(path, resolve_path); 4972 if (files.AppendIfUnique(file_spec)) 4973 count++; 4974 } 4975 } 4976 } 4977 } 4978 break; 4979 4980 default: 4981 break; 4982 } 4983 offset = cmd_offset + load_cmd.cmdsize; 4984 } 4985 4986 if (!rpath_paths.empty()) 4987 { 4988 // Fixup all LC_RPATH values to be absolute paths 4989 FileSpec this_file_spec(m_file); 4990 this_file_spec.ResolvePath(); 4991 std::string loader_path("@loader_path"); 4992 std::string executable_path("@executable_path"); 4993 for (auto &rpath : rpath_paths) 4994 { 4995 if (rpath.find(loader_path) == 0) 4996 { 4997 rpath.erase(0, loader_path.size()); 4998 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 4999 } 5000 else if (rpath.find(executable_path) == 0) 5001 { 5002 rpath.erase(0, executable_path.size()); 5003 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5004 } 5005 } 5006 5007 for (const auto &rpath_relative_path : rpath_relative_paths) 5008 { 5009 for (const auto &rpath : rpath_paths) 5010 { 5011 std::string path = rpath; 5012 path += rpath_relative_path; 5013 // It is OK to resolve this path because we must find a file on 5014 // disk for us to accept it anyway if it is rpath relative. 5015 FileSpec file_spec(path, true); 5016 // Remove any redundant parts of the path (like "../foo") since 5017 // LC_RPATH values often contain "..". 5018 file_spec.NormalizePath (); 5019 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) 5020 { 5021 count++; 5022 break; 5023 } 5024 } 5025 } 5026 } 5027 } 5028 return count; 5029 } 5030 5031 lldb_private::Address 5032 ObjectFileMachO::GetEntryPointAddress () 5033 { 5034 // If the object file is not an executable it can't hold the entry point. m_entry_point_address 5035 // is initialized to an invalid address, so we can just return that. 5036 // If m_entry_point_address is valid it means we've found it already, so return the cached value. 5037 5038 if (!IsExecutable() || m_entry_point_address.IsValid()) 5039 return m_entry_point_address; 5040 5041 // Otherwise, look for the UnixThread or Thread command. The data for the Thread command is given in 5042 // /usr/include/mach-o.h, but it is basically: 5043 // 5044 // uint32_t flavor - this is the flavor argument you would pass to thread_get_state 5045 // uint32_t count - this is the count of longs in the thread state data 5046 // struct XXX_thread_state state - this is the structure from <machine/thread_status.h> corresponding to the flavor. 5047 // <repeat this trio> 5048 // 5049 // So we just keep reading the various register flavors till we find the GPR one, then read the PC out of there. 5050 // FIXME: We will need to have a "RegisterContext data provider" class at some point that can get all the registers 5051 // out of data in this form & attach them to a given thread. That should underlie the MacOS X User process plugin, 5052 // and we'll also need it for the MacOS X Core File process plugin. When we have that we can also use it here. 5053 // 5054 // For now we hard-code the offsets and flavors we need: 5055 // 5056 // 5057 5058 ModuleSP module_sp(GetModule()); 5059 if (module_sp) 5060 { 5061 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5062 struct load_command load_cmd; 5063 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5064 uint32_t i; 5065 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5066 bool done = false; 5067 5068 for (i=0; i<m_header.ncmds; ++i) 5069 { 5070 const lldb::offset_t cmd_offset = offset; 5071 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5072 break; 5073 5074 switch (load_cmd.cmd) 5075 { 5076 case LC_UNIXTHREAD: 5077 case LC_THREAD: 5078 { 5079 while (offset < cmd_offset + load_cmd.cmdsize) 5080 { 5081 uint32_t flavor = m_data.GetU32(&offset); 5082 uint32_t count = m_data.GetU32(&offset); 5083 if (count == 0) 5084 { 5085 // We've gotten off somehow, log and exit; 5086 return m_entry_point_address; 5087 } 5088 5089 switch (m_header.cputype) 5090 { 5091 case llvm::MachO::CPU_TYPE_ARM: 5092 if (flavor == 1 || flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 from mach/arm/thread_status.h 5093 { 5094 offset += 60; // This is the offset of pc in the GPR thread state data structure. 5095 start_address = m_data.GetU32(&offset); 5096 done = true; 5097 } 5098 break; 5099 case llvm::MachO::CPU_TYPE_ARM64: 5100 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5101 { 5102 offset += 256; // This is the offset of pc in the GPR thread state data structure. 5103 start_address = m_data.GetU64(&offset); 5104 done = true; 5105 } 5106 break; 5107 case llvm::MachO::CPU_TYPE_I386: 5108 if (flavor == 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5109 { 5110 offset += 40; // This is the offset of eip in the GPR thread state data structure. 5111 start_address = m_data.GetU32(&offset); 5112 done = true; 5113 } 5114 break; 5115 case llvm::MachO::CPU_TYPE_X86_64: 5116 if (flavor == 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5117 { 5118 offset += 16 * 8; // This is the offset of rip in the GPR thread state data structure. 5119 start_address = m_data.GetU64(&offset); 5120 done = true; 5121 } 5122 break; 5123 default: 5124 return m_entry_point_address; 5125 } 5126 // Haven't found the GPR flavor yet, skip over the data for this flavor: 5127 if (done) 5128 break; 5129 offset += count * 4; 5130 } 5131 } 5132 break; 5133 case LC_MAIN: 5134 { 5135 ConstString text_segment_name ("__TEXT"); 5136 uint64_t entryoffset = m_data.GetU64(&offset); 5137 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 5138 if (text_segment_sp) 5139 { 5140 done = true; 5141 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5142 } 5143 } 5144 break; 5145 5146 default: 5147 break; 5148 } 5149 if (done) 5150 break; 5151 5152 // Go to the next load command: 5153 offset = cmd_offset + load_cmd.cmdsize; 5154 } 5155 5156 if (start_address != LLDB_INVALID_ADDRESS) 5157 { 5158 // We got the start address from the load commands, so now resolve that address in the sections 5159 // of this ObjectFile: 5160 if (!m_entry_point_address.ResolveAddressUsingFileSections (start_address, GetSectionList())) 5161 { 5162 m_entry_point_address.Clear(); 5163 } 5164 } 5165 else 5166 { 5167 // We couldn't read the UnixThread load command - maybe it wasn't there. As a fallback look for the 5168 // "start" symbol in the main executable. 5169 5170 ModuleSP module_sp (GetModule()); 5171 5172 if (module_sp) 5173 { 5174 SymbolContextList contexts; 5175 SymbolContext context; 5176 if (module_sp->FindSymbolsWithNameAndType(ConstString ("start"), eSymbolTypeCode, contexts)) 5177 { 5178 if (contexts.GetContextAtIndex(0, context)) 5179 m_entry_point_address = context.symbol->GetAddress(); 5180 } 5181 } 5182 } 5183 } 5184 5185 return m_entry_point_address; 5186 } 5187 5188 lldb_private::Address 5189 ObjectFileMachO::GetHeaderAddress () 5190 { 5191 lldb_private::Address header_addr; 5192 SectionList *section_list = GetSectionList(); 5193 if (section_list) 5194 { 5195 SectionSP text_segment_sp (section_list->FindSectionByName (GetSegmentNameTEXT())); 5196 if (text_segment_sp) 5197 { 5198 header_addr.SetSection (text_segment_sp); 5199 header_addr.SetOffset (0); 5200 } 5201 } 5202 return header_addr; 5203 } 5204 5205 uint32_t 5206 ObjectFileMachO::GetNumThreadContexts () 5207 { 5208 ModuleSP module_sp(GetModule()); 5209 if (module_sp) 5210 { 5211 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5212 if (!m_thread_context_offsets_valid) 5213 { 5214 m_thread_context_offsets_valid = true; 5215 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5216 FileRangeArray::Entry file_range; 5217 thread_command thread_cmd; 5218 for (uint32_t i=0; i<m_header.ncmds; ++i) 5219 { 5220 const uint32_t cmd_offset = offset; 5221 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 5222 break; 5223 5224 if (thread_cmd.cmd == LC_THREAD) 5225 { 5226 file_range.SetRangeBase (offset); 5227 file_range.SetByteSize (thread_cmd.cmdsize - 8); 5228 m_thread_context_offsets.Append (file_range); 5229 } 5230 offset = cmd_offset + thread_cmd.cmdsize; 5231 } 5232 } 5233 } 5234 return m_thread_context_offsets.GetSize(); 5235 } 5236 5237 lldb::RegisterContextSP 5238 ObjectFileMachO::GetThreadContextAtIndex (uint32_t idx, lldb_private::Thread &thread) 5239 { 5240 lldb::RegisterContextSP reg_ctx_sp; 5241 5242 ModuleSP module_sp(GetModule()); 5243 if (module_sp) 5244 { 5245 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5246 if (!m_thread_context_offsets_valid) 5247 GetNumThreadContexts (); 5248 5249 const FileRangeArray::Entry *thread_context_file_range = m_thread_context_offsets.GetEntryAtIndex (idx); 5250 if (thread_context_file_range) 5251 { 5252 5253 DataExtractor data (m_data, 5254 thread_context_file_range->GetRangeBase(), 5255 thread_context_file_range->GetByteSize()); 5256 5257 switch (m_header.cputype) 5258 { 5259 case llvm::MachO::CPU_TYPE_ARM64: 5260 reg_ctx_sp.reset (new RegisterContextDarwin_arm64_Mach (thread, data)); 5261 break; 5262 5263 case llvm::MachO::CPU_TYPE_ARM: 5264 reg_ctx_sp.reset (new RegisterContextDarwin_arm_Mach (thread, data)); 5265 break; 5266 5267 case llvm::MachO::CPU_TYPE_I386: 5268 reg_ctx_sp.reset (new RegisterContextDarwin_i386_Mach (thread, data)); 5269 break; 5270 5271 case llvm::MachO::CPU_TYPE_X86_64: 5272 reg_ctx_sp.reset (new RegisterContextDarwin_x86_64_Mach (thread, data)); 5273 break; 5274 } 5275 } 5276 } 5277 return reg_ctx_sp; 5278 } 5279 5280 ObjectFile::Type 5281 ObjectFileMachO::CalculateType() 5282 { 5283 switch (m_header.filetype) 5284 { 5285 case MH_OBJECT: // 0x1u 5286 if (GetAddressByteSize () == 4) 5287 { 5288 // 32 bit kexts are just object files, but they do have a valid 5289 // UUID load command. 5290 UUID uuid; 5291 if (GetUUID(&uuid)) 5292 { 5293 // this checking for the UUID load command is not enough 5294 // we could eventually look for the symbol named 5295 // "OSKextGetCurrentIdentifier" as this is required of kexts 5296 if (m_strata == eStrataInvalid) 5297 m_strata = eStrataKernel; 5298 return eTypeSharedLibrary; 5299 } 5300 } 5301 return eTypeObjectFile; 5302 5303 case MH_EXECUTE: return eTypeExecutable; // 0x2u 5304 case MH_FVMLIB: return eTypeSharedLibrary; // 0x3u 5305 case MH_CORE: return eTypeCoreFile; // 0x4u 5306 case MH_PRELOAD: return eTypeSharedLibrary; // 0x5u 5307 case MH_DYLIB: return eTypeSharedLibrary; // 0x6u 5308 case MH_DYLINKER: return eTypeDynamicLinker; // 0x7u 5309 case MH_BUNDLE: return eTypeSharedLibrary; // 0x8u 5310 case MH_DYLIB_STUB: return eTypeStubLibrary; // 0x9u 5311 case MH_DSYM: return eTypeDebugInfo; // 0xAu 5312 case MH_KEXT_BUNDLE: return eTypeSharedLibrary; // 0xBu 5313 default: 5314 break; 5315 } 5316 return eTypeUnknown; 5317 } 5318 5319 ObjectFile::Strata 5320 ObjectFileMachO::CalculateStrata() 5321 { 5322 switch (m_header.filetype) 5323 { 5324 case MH_OBJECT: // 0x1u 5325 { 5326 // 32 bit kexts are just object files, but they do have a valid 5327 // UUID load command. 5328 UUID uuid; 5329 if (GetUUID(&uuid)) 5330 { 5331 // this checking for the UUID load command is not enough 5332 // we could eventually look for the symbol named 5333 // "OSKextGetCurrentIdentifier" as this is required of kexts 5334 if (m_type == eTypeInvalid) 5335 m_type = eTypeSharedLibrary; 5336 5337 return eStrataKernel; 5338 } 5339 } 5340 return eStrataUnknown; 5341 5342 case MH_EXECUTE: // 0x2u 5343 // Check for the MH_DYLDLINK bit in the flags 5344 if (m_header.flags & MH_DYLDLINK) 5345 { 5346 return eStrataUser; 5347 } 5348 else 5349 { 5350 SectionList *section_list = GetSectionList(); 5351 if (section_list) 5352 { 5353 static ConstString g_kld_section_name ("__KLD"); 5354 if (section_list->FindSectionByName(g_kld_section_name)) 5355 return eStrataKernel; 5356 } 5357 } 5358 return eStrataRawImage; 5359 5360 case MH_FVMLIB: return eStrataUser; // 0x3u 5361 case MH_CORE: return eStrataUnknown; // 0x4u 5362 case MH_PRELOAD: return eStrataRawImage; // 0x5u 5363 case MH_DYLIB: return eStrataUser; // 0x6u 5364 case MH_DYLINKER: return eStrataUser; // 0x7u 5365 case MH_BUNDLE: return eStrataUser; // 0x8u 5366 case MH_DYLIB_STUB: return eStrataUser; // 0x9u 5367 case MH_DSYM: return eStrataUnknown; // 0xAu 5368 case MH_KEXT_BUNDLE: return eStrataKernel; // 0xBu 5369 default: 5370 break; 5371 } 5372 return eStrataUnknown; 5373 } 5374 5375 uint32_t 5376 ObjectFileMachO::GetVersion (uint32_t *versions, uint32_t num_versions) 5377 { 5378 ModuleSP module_sp(GetModule()); 5379 if (module_sp) 5380 { 5381 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5382 struct dylib_command load_cmd; 5383 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5384 uint32_t version_cmd = 0; 5385 uint64_t version = 0; 5386 uint32_t i; 5387 for (i=0; i<m_header.ncmds; ++i) 5388 { 5389 const lldb::offset_t cmd_offset = offset; 5390 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5391 break; 5392 5393 if (load_cmd.cmd == LC_ID_DYLIB) 5394 { 5395 if (version_cmd == 0) 5396 { 5397 version_cmd = load_cmd.cmd; 5398 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5399 break; 5400 version = load_cmd.dylib.current_version; 5401 } 5402 break; // Break for now unless there is another more complete version 5403 // number load command in the future. 5404 } 5405 offset = cmd_offset + load_cmd.cmdsize; 5406 } 5407 5408 if (version_cmd == LC_ID_DYLIB) 5409 { 5410 if (versions != NULL && num_versions > 0) 5411 { 5412 if (num_versions > 0) 5413 versions[0] = (version & 0xFFFF0000ull) >> 16; 5414 if (num_versions > 1) 5415 versions[1] = (version & 0x0000FF00ull) >> 8; 5416 if (num_versions > 2) 5417 versions[2] = (version & 0x000000FFull); 5418 // Fill in an remaining version numbers with invalid values 5419 for (i=3; i<num_versions; ++i) 5420 versions[i] = UINT32_MAX; 5421 } 5422 // The LC_ID_DYLIB load command has a version with 3 version numbers 5423 // in it, so always return 3 5424 return 3; 5425 } 5426 } 5427 return false; 5428 } 5429 5430 bool 5431 ObjectFileMachO::GetArchitecture (ArchSpec &arch) 5432 { 5433 ModuleSP module_sp(GetModule()); 5434 if (module_sp) 5435 { 5436 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5437 return GetArchitecture (m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), arch); 5438 } 5439 return false; 5440 } 5441 5442 UUID 5443 ObjectFileMachO::GetProcessSharedCacheUUID (Process *process) 5444 { 5445 UUID uuid; 5446 if (process) 5447 { 5448 addr_t all_image_infos = process->GetImageInfoAddress(); 5449 5450 // The address returned by GetImageInfoAddress may be the address of dyld (don't want) 5451 // or it may be the address of the dyld_all_image_infos structure (want). The first four 5452 // bytes will be either the version field (all_image_infos) or a Mach-O file magic constant. 5453 // Version 13 and higher of dyld_all_image_infos is required to get the sharedCacheUUID field. 5454 5455 Error err; 5456 uint32_t version_or_magic = process->ReadUnsignedIntegerFromMemory (all_image_infos, 4, -1, err); 5457 if (version_or_magic != static_cast<uint32_t>(-1) 5458 && version_or_magic != MH_MAGIC 5459 && version_or_magic != MH_CIGAM 5460 && version_or_magic != MH_MAGIC_64 5461 && version_or_magic != MH_CIGAM_64 5462 && version_or_magic >= 13) 5463 { 5464 addr_t sharedCacheUUID_address = LLDB_INVALID_ADDRESS; 5465 int wordsize = process->GetAddressByteSize(); 5466 if (wordsize == 8) 5467 { 5468 sharedCacheUUID_address = all_image_infos + 160; // sharedCacheUUID <mach-o/dyld_images.h> 5469 } 5470 if (wordsize == 4) 5471 { 5472 sharedCacheUUID_address = all_image_infos + 84; // sharedCacheUUID <mach-o/dyld_images.h> 5473 } 5474 if (sharedCacheUUID_address != LLDB_INVALID_ADDRESS) 5475 { 5476 uuid_t shared_cache_uuid; 5477 if (process->ReadMemory (sharedCacheUUID_address, shared_cache_uuid, sizeof (uuid_t), err) == sizeof (uuid_t)) 5478 { 5479 uuid.SetBytes (shared_cache_uuid); 5480 } 5481 } 5482 } 5483 } 5484 return uuid; 5485 } 5486 5487 UUID 5488 ObjectFileMachO::GetLLDBSharedCacheUUID () 5489 { 5490 UUID uuid; 5491 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 5492 uint8_t *(*dyld_get_all_image_infos)(void); 5493 dyld_get_all_image_infos = (uint8_t*(*)()) dlsym (RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5494 if (dyld_get_all_image_infos) 5495 { 5496 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5497 if (dyld_all_image_infos_address) 5498 { 5499 uint32_t *version = (uint32_t*) dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5500 if (*version >= 13) 5501 { 5502 uuid_t *sharedCacheUUID_address = 0; 5503 int wordsize = sizeof (uint8_t *); 5504 if (wordsize == 8) 5505 { 5506 sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 160); // sharedCacheUUID <mach-o/dyld_images.h> 5507 } 5508 else 5509 { 5510 sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 84); // sharedCacheUUID <mach-o/dyld_images.h> 5511 } 5512 uuid.SetBytes (sharedCacheUUID_address); 5513 } 5514 } 5515 } 5516 #endif 5517 return uuid; 5518 } 5519 5520 uint32_t 5521 ObjectFileMachO::GetMinimumOSVersion (uint32_t *versions, uint32_t num_versions) 5522 { 5523 if (m_min_os_versions.empty()) 5524 { 5525 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5526 bool success = false; 5527 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 5528 { 5529 const lldb::offset_t load_cmd_offset = offset; 5530 5531 version_min_command lc; 5532 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5533 break; 5534 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX 5535 || lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS 5536 || lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS 5537 || lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) 5538 { 5539 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 5540 { 5541 const uint32_t xxxx = lc.version >> 16; 5542 const uint32_t yy = (lc.version >> 8) & 0xffu; 5543 const uint32_t zz = lc.version & 0xffu; 5544 if (xxxx) 5545 { 5546 m_min_os_versions.push_back(xxxx); 5547 m_min_os_versions.push_back(yy); 5548 m_min_os_versions.push_back(zz); 5549 } 5550 success = true; 5551 } 5552 } 5553 offset = load_cmd_offset + lc.cmdsize; 5554 } 5555 5556 if (success == false) 5557 { 5558 // Push an invalid value so we don't keep trying to 5559 m_min_os_versions.push_back(UINT32_MAX); 5560 } 5561 } 5562 5563 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) 5564 { 5565 if (versions != NULL && num_versions > 0) 5566 { 5567 for (size_t i=0; i<num_versions; ++i) 5568 { 5569 if (i < m_min_os_versions.size()) 5570 versions[i] = m_min_os_versions[i]; 5571 else 5572 versions[i] = 0; 5573 } 5574 } 5575 return m_min_os_versions.size(); 5576 } 5577 // Call the superclasses version that will empty out the data 5578 return ObjectFile::GetMinimumOSVersion (versions, num_versions); 5579 } 5580 5581 uint32_t 5582 ObjectFileMachO::GetSDKVersion(uint32_t *versions, uint32_t num_versions) 5583 { 5584 if (m_sdk_versions.empty()) 5585 { 5586 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5587 bool success = false; 5588 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 5589 { 5590 const lldb::offset_t load_cmd_offset = offset; 5591 5592 version_min_command lc; 5593 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5594 break; 5595 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX 5596 || lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS 5597 || lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS 5598 || lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) 5599 { 5600 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 5601 { 5602 const uint32_t xxxx = lc.sdk >> 16; 5603 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5604 const uint32_t zz = lc.sdk & 0xffu; 5605 if (xxxx) 5606 { 5607 m_sdk_versions.push_back(xxxx); 5608 m_sdk_versions.push_back(yy); 5609 m_sdk_versions.push_back(zz); 5610 } 5611 success = true; 5612 } 5613 } 5614 offset = load_cmd_offset + lc.cmdsize; 5615 } 5616 5617 if (success == false) 5618 { 5619 // Push an invalid value so we don't keep trying to 5620 m_sdk_versions.push_back(UINT32_MAX); 5621 } 5622 } 5623 5624 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) 5625 { 5626 if (versions != NULL && num_versions > 0) 5627 { 5628 for (size_t i=0; i<num_versions; ++i) 5629 { 5630 if (i < m_sdk_versions.size()) 5631 versions[i] = m_sdk_versions[i]; 5632 else 5633 versions[i] = 0; 5634 } 5635 } 5636 return m_sdk_versions.size(); 5637 } 5638 // Call the superclasses version that will empty out the data 5639 return ObjectFile::GetSDKVersion (versions, num_versions); 5640 } 5641 5642 bool 5643 ObjectFileMachO::GetIsDynamicLinkEditor() 5644 { 5645 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5646 } 5647 5648 bool 5649 ObjectFileMachO::AllowAssemblyEmulationUnwindPlans () 5650 { 5651 return m_allow_assembly_emulation_unwind_plans; 5652 } 5653 5654 //------------------------------------------------------------------ 5655 // PluginInterface protocol 5656 //------------------------------------------------------------------ 5657 lldb_private::ConstString 5658 ObjectFileMachO::GetPluginName() 5659 { 5660 return GetPluginNameStatic(); 5661 } 5662 5663 uint32_t 5664 ObjectFileMachO::GetPluginVersion() 5665 { 5666 return 1; 5667 } 5668 5669 Section * 5670 ObjectFileMachO::GetMachHeaderSection() 5671 { 5672 // Find the first address of the mach header which is the first non-zero 5673 // file sized section whose file offset is zero. This is the base file address 5674 // of the mach-o file which can be subtracted from the vmaddr of the other 5675 // segments found in memory and added to the load address 5676 ModuleSP module_sp = GetModule(); 5677 if (module_sp) 5678 { 5679 SectionList *section_list = GetSectionList (); 5680 if (section_list) 5681 { 5682 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5683 const size_t num_sections = section_list->GetSize(); 5684 5685 for (size_t sect_idx = 0; 5686 sect_idx < num_sections && mach_base_file_addr == LLDB_INVALID_ADDRESS; 5687 ++sect_idx) 5688 { 5689 Section *section = section_list->GetSectionAtIndex (sect_idx).get(); 5690 if (section && 5691 section->GetFileSize() > 0 && 5692 section->GetFileOffset() == 0 && 5693 section->IsThreadSpecific() == false && 5694 module_sp.get() == section->GetModule().get()) 5695 { 5696 return section; 5697 } 5698 } 5699 } 5700 } 5701 return nullptr; 5702 } 5703 5704 lldb::addr_t 5705 ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage(lldb::addr_t mach_header_load_address, const Section *mach_header_section, const Section *section) 5706 { 5707 ModuleSP module_sp = GetModule(); 5708 if (module_sp && mach_header_section && section && mach_header_load_address != LLDB_INVALID_ADDRESS) 5709 { 5710 lldb::addr_t mach_header_file_addr = mach_header_section->GetFileAddress(); 5711 if (mach_header_file_addr != LLDB_INVALID_ADDRESS) 5712 { 5713 if (section && 5714 section->GetFileSize() > 0 && 5715 section->IsThreadSpecific() == false && 5716 module_sp.get() == section->GetModule().get()) 5717 { 5718 // Ignore __LINKEDIT and __DWARF segments 5719 if (section->GetName() == GetSegmentNameLINKEDIT()) 5720 { 5721 // Only map __LINKEDIT if we have an in memory image and this isn't 5722 // a kernel binary like a kext or mach_kernel. 5723 const bool is_memory_image = (bool)m_process_wp.lock(); 5724 const Strata strata = GetStrata(); 5725 if (is_memory_image == false || strata == eStrataKernel) 5726 return LLDB_INVALID_ADDRESS; 5727 } 5728 return section->GetFileAddress() - mach_header_file_addr + mach_header_load_address; 5729 } 5730 } 5731 } 5732 return LLDB_INVALID_ADDRESS; 5733 } 5734 5735 bool 5736 ObjectFileMachO::SetLoadAddress (Target &target, 5737 lldb::addr_t value, 5738 bool value_is_offset) 5739 { 5740 ModuleSP module_sp = GetModule(); 5741 if (module_sp) 5742 { 5743 size_t num_loaded_sections = 0; 5744 SectionList *section_list = GetSectionList (); 5745 if (section_list) 5746 { 5747 const size_t num_sections = section_list->GetSize(); 5748 5749 if (value_is_offset) 5750 { 5751 // "value" is an offset to apply to each top level segment 5752 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) 5753 { 5754 // Iterate through the object file sections to find all 5755 // of the sections that size on disk (to avoid __PAGEZERO) 5756 // and load them 5757 SectionSP section_sp (section_list->GetSectionAtIndex (sect_idx)); 5758 if (section_sp && 5759 section_sp->GetFileSize() > 0 && 5760 section_sp->IsThreadSpecific() == false && 5761 module_sp.get() == section_sp->GetModule().get()) 5762 { 5763 // Ignore __LINKEDIT and __DWARF segments 5764 if (section_sp->GetName() == GetSegmentNameLINKEDIT()) 5765 { 5766 // Only map __LINKEDIT if we have an in memory image and this isn't 5767 // a kernel binary like a kext or mach_kernel. 5768 const bool is_memory_image = (bool)m_process_wp.lock(); 5769 const Strata strata = GetStrata(); 5770 if (is_memory_image == false || strata == eStrataKernel) 5771 continue; 5772 } 5773 if (target.GetSectionLoadList().SetSectionLoadAddress (section_sp, section_sp->GetFileAddress() + value)) 5774 ++num_loaded_sections; 5775 } 5776 } 5777 } 5778 else 5779 { 5780 // "value" is the new base address of the mach_header, adjust each 5781 // section accordingly 5782 5783 Section *mach_header_section = GetMachHeaderSection(); 5784 if (mach_header_section) 5785 { 5786 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) 5787 { 5788 SectionSP section_sp (section_list->GetSectionAtIndex (sect_idx)); 5789 5790 lldb::addr_t section_load_addr = CalculateSectionLoadAddressForMemoryImage(value, mach_header_section, section_sp.get()); 5791 if (section_load_addr != LLDB_INVALID_ADDRESS) 5792 { 5793 if (target.GetSectionLoadList().SetSectionLoadAddress (section_sp, section_load_addr)) 5794 ++num_loaded_sections; 5795 } 5796 } 5797 } 5798 } 5799 } 5800 return num_loaded_sections > 0; 5801 } 5802 return false; 5803 } 5804 5805 bool 5806 ObjectFileMachO::SaveCore (const lldb::ProcessSP &process_sp, 5807 const FileSpec &outfile, 5808 Error &error) 5809 { 5810 if (process_sp) 5811 { 5812 Target &target = process_sp->GetTarget(); 5813 const ArchSpec target_arch = target.GetArchitecture(); 5814 const llvm::Triple &target_triple = target_arch.GetTriple(); 5815 if (target_triple.getVendor() == llvm::Triple::Apple && 5816 (target_triple.getOS() == llvm::Triple::MacOSX 5817 || target_triple.getOS() == llvm::Triple::IOS 5818 || target_triple.getOS() == llvm::Triple::WatchOS 5819 || target_triple.getOS() == llvm::Triple::TvOS)) 5820 { 5821 bool make_core = false; 5822 switch (target_arch.GetMachine()) 5823 { 5824 case llvm::Triple::aarch64: 5825 case llvm::Triple::arm: 5826 case llvm::Triple::thumb: 5827 case llvm::Triple::x86: 5828 case llvm::Triple::x86_64: 5829 make_core = true; 5830 break; 5831 default: 5832 error.SetErrorStringWithFormat ("unsupported core architecture: %s", target_triple.str().c_str()); 5833 break; 5834 } 5835 5836 if (make_core) 5837 { 5838 std::vector<segment_command_64> segment_load_commands; 5839 // uint32_t range_info_idx = 0; 5840 MemoryRegionInfo range_info; 5841 Error range_error = process_sp->GetMemoryRegionInfo(0, range_info); 5842 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 5843 const ByteOrder byte_order = target_arch.GetByteOrder(); 5844 if (range_error.Success()) 5845 { 5846 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) 5847 { 5848 const addr_t addr = range_info.GetRange().GetRangeBase(); 5849 const addr_t size = range_info.GetRange().GetByteSize(); 5850 5851 if (size == 0) 5852 break; 5853 5854 // Calculate correct protections 5855 uint32_t prot = 0; 5856 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 5857 prot |= VM_PROT_READ; 5858 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 5859 prot |= VM_PROT_WRITE; 5860 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 5861 prot |= VM_PROT_EXECUTE; 5862 5863 // printf ("[%3u] [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 ") %c%c%c\n", 5864 // range_info_idx, 5865 // addr, 5866 // size, 5867 // (prot & VM_PROT_READ ) ? 'r' : '-', 5868 // (prot & VM_PROT_WRITE ) ? 'w' : '-', 5869 // (prot & VM_PROT_EXECUTE) ? 'x' : '-'); 5870 5871 if (prot != 0) 5872 { 5873 uint32_t cmd_type = LC_SEGMENT_64; 5874 uint32_t segment_size = sizeof (segment_command_64); 5875 if (addr_byte_size == 4) 5876 { 5877 cmd_type = LC_SEGMENT; 5878 segment_size = sizeof (segment_command); 5879 } 5880 segment_command_64 segment = { 5881 cmd_type, // uint32_t cmd; 5882 segment_size, // uint32_t cmdsize; 5883 {0}, // char segname[16]; 5884 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 5885 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 5886 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 5887 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 5888 prot, // uint32_t maxprot; 5889 prot, // uint32_t initprot; 5890 0, // uint32_t nsects; 5891 0 }; // uint32_t flags; 5892 segment_load_commands.push_back(segment); 5893 } 5894 else 5895 { 5896 // No protections and a size of 1 used to be returned from old 5897 // debugservers when we asked about a region that was past the 5898 // last memory region and it indicates the end... 5899 if (size == 1) 5900 break; 5901 } 5902 5903 range_error = process_sp->GetMemoryRegionInfo(range_info.GetRange().GetRangeEnd(), range_info); 5904 if (range_error.Fail()) 5905 break; 5906 } 5907 5908 StreamString buffer (Stream::eBinary, 5909 addr_byte_size, 5910 byte_order); 5911 5912 mach_header_64 mach_header; 5913 if (addr_byte_size == 8) 5914 { 5915 mach_header.magic = MH_MAGIC_64; 5916 } 5917 else 5918 { 5919 mach_header.magic = MH_MAGIC; 5920 } 5921 mach_header.cputype = target_arch.GetMachOCPUType(); 5922 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 5923 mach_header.filetype = MH_CORE; 5924 mach_header.ncmds = segment_load_commands.size(); 5925 mach_header.flags = 0; 5926 mach_header.reserved = 0; 5927 ThreadList &thread_list = process_sp->GetThreadList(); 5928 const uint32_t num_threads = thread_list.GetSize(); 5929 5930 // Make an array of LC_THREAD data items. Each one contains 5931 // the contents of the LC_THREAD load command. The data doesn't 5932 // contain the load command + load command size, we will 5933 // add the load command and load command size as we emit the data. 5934 std::vector<StreamString> LC_THREAD_datas(num_threads); 5935 for (auto &LC_THREAD_data : LC_THREAD_datas) 5936 { 5937 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 5938 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 5939 LC_THREAD_data.SetByteOrder(byte_order); 5940 } 5941 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) 5942 { 5943 ThreadSP thread_sp (thread_list.GetThreadAtIndex(thread_idx)); 5944 if (thread_sp) 5945 { 5946 switch (mach_header.cputype) 5947 { 5948 case llvm::MachO::CPU_TYPE_ARM64: 5949 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5950 break; 5951 5952 case llvm::MachO::CPU_TYPE_ARM: 5953 RegisterContextDarwin_arm_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5954 break; 5955 5956 case llvm::MachO::CPU_TYPE_I386: 5957 RegisterContextDarwin_i386_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5958 break; 5959 5960 case llvm::MachO::CPU_TYPE_X86_64: 5961 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5962 break; 5963 } 5964 5965 } 5966 } 5967 5968 // The size of the load command is the size of the segments... 5969 if (addr_byte_size == 8) 5970 { 5971 mach_header.sizeofcmds = segment_load_commands.size() * sizeof (struct segment_command_64); 5972 } 5973 else 5974 { 5975 mach_header.sizeofcmds = segment_load_commands.size() * sizeof (struct segment_command); 5976 } 5977 5978 // and the size of all LC_THREAD load command 5979 for (const auto &LC_THREAD_data : LC_THREAD_datas) 5980 { 5981 ++mach_header.ncmds; 5982 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 5983 } 5984 5985 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", 5986 mach_header.magic, 5987 mach_header.cputype, 5988 mach_header.cpusubtype, 5989 mach_header.filetype, 5990 mach_header.ncmds, 5991 mach_header.sizeofcmds, 5992 mach_header.flags, 5993 mach_header.reserved); 5994 5995 // Write the mach header 5996 buffer.PutHex32(mach_header.magic); 5997 buffer.PutHex32(mach_header.cputype); 5998 buffer.PutHex32(mach_header.cpusubtype); 5999 buffer.PutHex32(mach_header.filetype); 6000 buffer.PutHex32(mach_header.ncmds); 6001 buffer.PutHex32(mach_header.sizeofcmds); 6002 buffer.PutHex32(mach_header.flags); 6003 if (addr_byte_size == 8) 6004 { 6005 buffer.PutHex32(mach_header.reserved); 6006 } 6007 6008 // Skip the mach header and all load commands and align to the next 6009 // 0x1000 byte boundary 6010 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6011 if (file_offset & 0x00000fff) 6012 { 6013 file_offset += 0x00001000ull; 6014 file_offset &= (~0x00001000ull + 1); 6015 } 6016 6017 for (auto &segment : segment_load_commands) 6018 { 6019 segment.fileoff = file_offset; 6020 file_offset += segment.filesize; 6021 } 6022 6023 // Write out all of the LC_THREAD load commands 6024 for (const auto &LC_THREAD_data : LC_THREAD_datas) 6025 { 6026 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6027 buffer.PutHex32(LC_THREAD); 6028 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6029 buffer.Write(LC_THREAD_data.GetData(), LC_THREAD_data_size); 6030 } 6031 6032 // Write out all of the segment load commands 6033 for (const auto &segment : segment_load_commands) 6034 { 6035 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", 6036 segment.cmd, 6037 segment.cmdsize, 6038 segment.vmaddr, 6039 segment.vmaddr + segment.vmsize, 6040 segment.fileoff, 6041 segment.filesize, 6042 segment.maxprot, 6043 segment.initprot, 6044 segment.nsects, 6045 segment.flags); 6046 6047 buffer.PutHex32(segment.cmd); 6048 buffer.PutHex32(segment.cmdsize); 6049 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6050 if (addr_byte_size == 8) 6051 { 6052 buffer.PutHex64(segment.vmaddr); 6053 buffer.PutHex64(segment.vmsize); 6054 buffer.PutHex64(segment.fileoff); 6055 buffer.PutHex64(segment.filesize); 6056 } 6057 else 6058 { 6059 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6060 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6061 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6062 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6063 } 6064 buffer.PutHex32(segment.maxprot); 6065 buffer.PutHex32(segment.initprot); 6066 buffer.PutHex32(segment.nsects); 6067 buffer.PutHex32(segment.flags); 6068 } 6069 6070 File core_file; 6071 std::string core_file_path(outfile.GetPath()); 6072 error = core_file.Open(core_file_path.c_str(), 6073 File::eOpenOptionWrite | 6074 File::eOpenOptionTruncate | 6075 File::eOpenOptionCanCreate); 6076 if (error.Success()) 6077 { 6078 // Read 1 page at a time 6079 uint8_t bytes[0x1000]; 6080 // Write the mach header and load commands out to the core file 6081 size_t bytes_written = buffer.GetString().size(); 6082 error = core_file.Write(buffer.GetString().data(), bytes_written); 6083 if (error.Success()) 6084 { 6085 // Now write the file data for all memory segments in the process 6086 for (const auto &segment : segment_load_commands) 6087 { 6088 if (core_file.SeekFromStart(segment.fileoff) == -1) 6089 { 6090 error.SetErrorStringWithFormat("unable to seek to offset 0x%" PRIx64 " in '%s'", segment.fileoff, core_file_path.c_str()); 6091 break; 6092 } 6093 6094 printf ("Saving %" PRId64 " bytes of data for memory region at 0x%" PRIx64 "\n", segment.vmsize, segment.vmaddr); 6095 addr_t bytes_left = segment.vmsize; 6096 addr_t addr = segment.vmaddr; 6097 Error memory_read_error; 6098 while (bytes_left > 0 && error.Success()) 6099 { 6100 const size_t bytes_to_read = bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6101 const size_t bytes_read = process_sp->ReadMemory(addr, bytes, bytes_to_read, memory_read_error); 6102 if (bytes_read == bytes_to_read) 6103 { 6104 size_t bytes_written = bytes_read; 6105 error = core_file.Write(bytes, bytes_written); 6106 bytes_left -= bytes_read; 6107 addr += bytes_read; 6108 } 6109 else 6110 { 6111 // Some pages within regions are not readable, those 6112 // should be zero filled 6113 memset (bytes, 0, bytes_to_read); 6114 size_t bytes_written = bytes_to_read; 6115 error = core_file.Write(bytes, bytes_written); 6116 bytes_left -= bytes_to_read; 6117 addr += bytes_to_read; 6118 } 6119 } 6120 } 6121 } 6122 } 6123 } 6124 else 6125 { 6126 error.SetErrorString("process doesn't support getting memory region info"); 6127 } 6128 } 6129 return true; // This is the right plug to handle saving core files for this process 6130 } 6131 } 6132 return false; 6133 } 6134