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