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