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 "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 17 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 18 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 19 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 20 #include "lldb/Core/Debugger.h" 21 #include "lldb/Core/FileSpecList.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/RegisterValue.h" 27 #include "lldb/Core/Section.h" 28 #include "lldb/Core/StreamFile.h" 29 #include "lldb/Host/Host.h" 30 #include "lldb/Symbol/DWARFCallFrameInfo.h" 31 #include "lldb/Symbol/ObjectFile.h" 32 #include "lldb/Target/DynamicLoader.h" 33 #include "lldb/Target/MemoryRegionInfo.h" 34 #include "lldb/Target/Platform.h" 35 #include "lldb/Target/Process.h" 36 #include "lldb/Target/SectionLoadList.h" 37 #include "lldb/Target/Target.h" 38 #include "lldb/Target/Thread.h" 39 #include "lldb/Target/ThreadList.h" 40 #include "lldb/Utility/ArchSpec.h" 41 #include "lldb/Utility/DataBuffer.h" 42 #include "lldb/Utility/FileSpec.h" 43 #include "lldb/Utility/Log.h" 44 #include "lldb/Utility/Status.h" 45 #include "lldb/Utility/StreamString.h" 46 #include "lldb/Utility/Timer.h" 47 #include "lldb/Utility/UUID.h" 48 49 #include "lldb/Utility/SafeMachO.h" 50 51 #include "llvm/Support/MemoryBuffer.h" 52 53 #include "ObjectFileMachO.h" 54 55 #if defined(__APPLE__) && \ 56 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 57 // GetLLDBSharedCacheUUID() needs to call dlsym() 58 #include <dlfcn.h> 59 #endif 60 61 #ifndef __APPLE__ 62 #include "Utility/UuidCompatibility.h" 63 #else 64 #include <uuid/uuid.h> 65 #endif 66 67 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 68 using namespace lldb; 69 using namespace lldb_private; 70 using namespace llvm::MachO; 71 72 // Some structure definitions needed for parsing the dyld shared cache files 73 // found on iOS devices. 74 75 struct lldb_copy_dyld_cache_header_v1 { 76 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 77 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 78 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 79 uint32_t imagesOffset; 80 uint32_t imagesCount; 81 uint64_t dyldBaseAddress; 82 uint64_t codeSignatureOffset; 83 uint64_t codeSignatureSize; 84 uint64_t slideInfoOffset; 85 uint64_t slideInfoSize; 86 uint64_t localSymbolsOffset; 87 uint64_t localSymbolsSize; 88 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 89 // and later 90 }; 91 92 struct lldb_copy_dyld_cache_mapping_info { 93 uint64_t address; 94 uint64_t size; 95 uint64_t fileOffset; 96 uint32_t maxProt; 97 uint32_t initProt; 98 }; 99 100 struct lldb_copy_dyld_cache_local_symbols_info { 101 uint32_t nlistOffset; 102 uint32_t nlistCount; 103 uint32_t stringsOffset; 104 uint32_t stringsSize; 105 uint32_t entriesOffset; 106 uint32_t entriesCount; 107 }; 108 struct lldb_copy_dyld_cache_local_symbols_entry { 109 uint32_t dylibOffset; 110 uint32_t nlistStartIndex; 111 uint32_t nlistCount; 112 }; 113 114 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 115 public: 116 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 117 const DataExtractor &data) 118 : RegisterContextDarwin_x86_64(thread, 0) { 119 SetRegisterDataFrom_LC_THREAD(data); 120 } 121 122 void InvalidateAllRegisters() override { 123 // Do nothing... registers are always valid... 124 } 125 126 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 127 lldb::offset_t offset = 0; 128 SetError(GPRRegSet, Read, -1); 129 SetError(FPURegSet, Read, -1); 130 SetError(EXCRegSet, Read, -1); 131 bool done = false; 132 133 while (!done) { 134 int flavor = data.GetU32(&offset); 135 if (flavor == 0) 136 done = true; 137 else { 138 uint32_t i; 139 uint32_t count = data.GetU32(&offset); 140 switch (flavor) { 141 case GPRRegSet: 142 for (i = 0; i < count; ++i) 143 (&gpr.rax)[i] = data.GetU64(&offset); 144 SetError(GPRRegSet, Read, 0); 145 done = true; 146 147 break; 148 case FPURegSet: 149 // TODO: fill in FPU regs.... 150 // SetError (FPURegSet, Read, -1); 151 done = true; 152 153 break; 154 case EXCRegSet: 155 exc.trapno = data.GetU32(&offset); 156 exc.err = data.GetU32(&offset); 157 exc.faultvaddr = data.GetU64(&offset); 158 SetError(EXCRegSet, Read, 0); 159 done = true; 160 break; 161 case 7: 162 case 8: 163 case 9: 164 // fancy flavors that encapsulate of the above 165 // flavors... 166 break; 167 168 default: 169 done = true; 170 break; 171 } 172 } 173 } 174 } 175 176 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 177 const char *alt_name, size_t reg_byte_size, 178 Stream &data) { 179 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 180 if (reg_info == NULL) 181 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 182 if (reg_info) { 183 lldb_private::RegisterValue reg_value; 184 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 185 if (reg_info->byte_size >= reg_byte_size) 186 data.Write(reg_value.GetBytes(), reg_byte_size); 187 else { 188 data.Write(reg_value.GetBytes(), reg_info->byte_size); 189 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 190 ++i) 191 data.PutChar(0); 192 } 193 return reg_byte_size; 194 } 195 } 196 // Just write zeros if all else fails 197 for (size_t i = 0; i < reg_byte_size; ++i) 198 data.PutChar(0); 199 return reg_byte_size; 200 } 201 202 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 203 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 204 if (reg_ctx_sp) { 205 RegisterContext *reg_ctx = reg_ctx_sp.get(); 206 207 data.PutHex32(GPRRegSet); // Flavor 208 data.PutHex32(GPRWordCount); 209 WriteRegister(reg_ctx, "rax", NULL, 8, data); 210 WriteRegister(reg_ctx, "rbx", NULL, 8, data); 211 WriteRegister(reg_ctx, "rcx", NULL, 8, data); 212 WriteRegister(reg_ctx, "rdx", NULL, 8, data); 213 WriteRegister(reg_ctx, "rdi", NULL, 8, data); 214 WriteRegister(reg_ctx, "rsi", NULL, 8, data); 215 WriteRegister(reg_ctx, "rbp", NULL, 8, data); 216 WriteRegister(reg_ctx, "rsp", NULL, 8, data); 217 WriteRegister(reg_ctx, "r8", NULL, 8, data); 218 WriteRegister(reg_ctx, "r9", NULL, 8, data); 219 WriteRegister(reg_ctx, "r10", NULL, 8, data); 220 WriteRegister(reg_ctx, "r11", NULL, 8, data); 221 WriteRegister(reg_ctx, "r12", NULL, 8, data); 222 WriteRegister(reg_ctx, "r13", NULL, 8, data); 223 WriteRegister(reg_ctx, "r14", NULL, 8, data); 224 WriteRegister(reg_ctx, "r15", NULL, 8, data); 225 WriteRegister(reg_ctx, "rip", NULL, 8, data); 226 WriteRegister(reg_ctx, "rflags", NULL, 8, data); 227 WriteRegister(reg_ctx, "cs", NULL, 8, data); 228 WriteRegister(reg_ctx, "fs", NULL, 8, data); 229 WriteRegister(reg_ctx, "gs", NULL, 8, data); 230 231 // // Write out the FPU registers 232 // const size_t fpu_byte_size = sizeof(FPU); 233 // size_t bytes_written = 0; 234 // data.PutHex32 (FPURegSet); 235 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 236 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 237 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 238 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 239 // data); // uint16_t fcw; // "fctrl" 240 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 241 // data); // uint16_t fsw; // "fstat" 242 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 243 // data); // uint8_t ftw; // "ftag" 244 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 245 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 246 // data); // uint16_t fop; // "fop" 247 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 248 // data); // uint32_t ip; // "fioff" 249 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 250 // data); // uint16_t cs; // "fiseg" 251 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 252 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 253 // data); // uint32_t dp; // "fooff" 254 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 255 // data); // uint16_t ds; // "foseg" 256 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 257 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 258 // data); // uint32_t mxcsr; 259 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 260 // 4, data);// uint32_t mxcsrmask; 261 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 262 // sizeof(MMSReg), data); 263 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 264 // sizeof(MMSReg), data); 265 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 266 // sizeof(MMSReg), data); 267 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 268 // sizeof(MMSReg), data); 269 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 270 // sizeof(MMSReg), data); 271 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 272 // sizeof(MMSReg), data); 273 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 274 // sizeof(MMSReg), data); 275 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 276 // sizeof(MMSReg), data); 277 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 278 // sizeof(XMMReg), data); 279 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 280 // sizeof(XMMReg), data); 281 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 282 // sizeof(XMMReg), data); 283 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 284 // sizeof(XMMReg), data); 285 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 286 // sizeof(XMMReg), data); 287 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 288 // sizeof(XMMReg), data); 289 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 290 // sizeof(XMMReg), data); 291 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 292 // sizeof(XMMReg), data); 293 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 294 // sizeof(XMMReg), data); 295 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 296 // sizeof(XMMReg), data); 297 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 298 // sizeof(XMMReg), data); 299 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 300 // sizeof(XMMReg), data); 301 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 302 // sizeof(XMMReg), data); 303 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 304 // sizeof(XMMReg), data); 305 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 306 // sizeof(XMMReg), data); 307 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 308 // sizeof(XMMReg), data); 309 // 310 // // Fill rest with zeros 311 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 312 // i) 313 // data.PutChar(0); 314 315 // Write out the EXC registers 316 data.PutHex32(EXCRegSet); 317 data.PutHex32(EXCWordCount); 318 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 319 WriteRegister(reg_ctx, "err", NULL, 4, data); 320 WriteRegister(reg_ctx, "faultvaddr", NULL, 8, data); 321 return true; 322 } 323 return false; 324 } 325 326 protected: 327 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 328 329 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 330 331 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 332 333 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 334 return 0; 335 } 336 337 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 338 return 0; 339 } 340 341 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 342 return 0; 343 } 344 }; 345 346 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 347 public: 348 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 349 const DataExtractor &data) 350 : RegisterContextDarwin_i386(thread, 0) { 351 SetRegisterDataFrom_LC_THREAD(data); 352 } 353 354 void InvalidateAllRegisters() override { 355 // Do nothing... registers are always valid... 356 } 357 358 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 359 lldb::offset_t offset = 0; 360 SetError(GPRRegSet, Read, -1); 361 SetError(FPURegSet, Read, -1); 362 SetError(EXCRegSet, Read, -1); 363 bool done = false; 364 365 while (!done) { 366 int flavor = data.GetU32(&offset); 367 if (flavor == 0) 368 done = true; 369 else { 370 uint32_t i; 371 uint32_t count = data.GetU32(&offset); 372 switch (flavor) { 373 case GPRRegSet: 374 for (i = 0; i < count; ++i) 375 (&gpr.eax)[i] = data.GetU32(&offset); 376 SetError(GPRRegSet, Read, 0); 377 done = true; 378 379 break; 380 case FPURegSet: 381 // TODO: fill in FPU regs.... 382 // SetError (FPURegSet, Read, -1); 383 done = true; 384 385 break; 386 case EXCRegSet: 387 exc.trapno = data.GetU32(&offset); 388 exc.err = data.GetU32(&offset); 389 exc.faultvaddr = data.GetU32(&offset); 390 SetError(EXCRegSet, Read, 0); 391 done = true; 392 break; 393 case 7: 394 case 8: 395 case 9: 396 // fancy flavors that encapsulate of the above 397 // flavors... 398 break; 399 400 default: 401 done = true; 402 break; 403 } 404 } 405 } 406 } 407 408 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 409 const char *alt_name, size_t reg_byte_size, 410 Stream &data) { 411 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 412 if (reg_info == NULL) 413 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 414 if (reg_info) { 415 lldb_private::RegisterValue reg_value; 416 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 417 if (reg_info->byte_size >= reg_byte_size) 418 data.Write(reg_value.GetBytes(), reg_byte_size); 419 else { 420 data.Write(reg_value.GetBytes(), reg_info->byte_size); 421 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 422 ++i) 423 data.PutChar(0); 424 } 425 return reg_byte_size; 426 } 427 } 428 // Just write zeros if all else fails 429 for (size_t i = 0; i < reg_byte_size; ++i) 430 data.PutChar(0); 431 return reg_byte_size; 432 } 433 434 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 435 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 436 if (reg_ctx_sp) { 437 RegisterContext *reg_ctx = reg_ctx_sp.get(); 438 439 data.PutHex32(GPRRegSet); // Flavor 440 data.PutHex32(GPRWordCount); 441 WriteRegister(reg_ctx, "eax", NULL, 4, data); 442 WriteRegister(reg_ctx, "ebx", NULL, 4, data); 443 WriteRegister(reg_ctx, "ecx", NULL, 4, data); 444 WriteRegister(reg_ctx, "edx", NULL, 4, data); 445 WriteRegister(reg_ctx, "edi", NULL, 4, data); 446 WriteRegister(reg_ctx, "esi", NULL, 4, data); 447 WriteRegister(reg_ctx, "ebp", NULL, 4, data); 448 WriteRegister(reg_ctx, "esp", NULL, 4, data); 449 WriteRegister(reg_ctx, "ss", NULL, 4, data); 450 WriteRegister(reg_ctx, "eflags", NULL, 4, data); 451 WriteRegister(reg_ctx, "eip", NULL, 4, data); 452 WriteRegister(reg_ctx, "cs", NULL, 4, data); 453 WriteRegister(reg_ctx, "ds", NULL, 4, data); 454 WriteRegister(reg_ctx, "es", NULL, 4, data); 455 WriteRegister(reg_ctx, "fs", NULL, 4, data); 456 WriteRegister(reg_ctx, "gs", NULL, 4, data); 457 458 // Write out the EXC registers 459 data.PutHex32(EXCRegSet); 460 data.PutHex32(EXCWordCount); 461 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 462 WriteRegister(reg_ctx, "err", NULL, 4, data); 463 WriteRegister(reg_ctx, "faultvaddr", NULL, 4, data); 464 return true; 465 } 466 return false; 467 } 468 469 protected: 470 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 471 472 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 473 474 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 475 476 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 477 return 0; 478 } 479 480 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 481 return 0; 482 } 483 484 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 485 return 0; 486 } 487 }; 488 489 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 490 public: 491 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 492 const DataExtractor &data) 493 : RegisterContextDarwin_arm(thread, 0) { 494 SetRegisterDataFrom_LC_THREAD(data); 495 } 496 497 void InvalidateAllRegisters() override { 498 // Do nothing... registers are always valid... 499 } 500 501 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 502 lldb::offset_t offset = 0; 503 SetError(GPRRegSet, Read, -1); 504 SetError(FPURegSet, Read, -1); 505 SetError(EXCRegSet, Read, -1); 506 bool done = false; 507 508 while (!done) { 509 int flavor = data.GetU32(&offset); 510 uint32_t count = data.GetU32(&offset); 511 lldb::offset_t next_thread_state = offset + (count * 4); 512 switch (flavor) { 513 case GPRAltRegSet: 514 case GPRRegSet: 515 for (uint32_t i = 0; i < count; ++i) { 516 gpr.r[i] = data.GetU32(&offset); 517 } 518 519 // Note that gpr.cpsr is also copied by the above loop; this loop 520 // technically extends 521 // one element past the end of the gpr.r[] array. 522 523 SetError(GPRRegSet, Read, 0); 524 offset = next_thread_state; 525 break; 526 527 case FPURegSet: { 528 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 529 const int fpu_reg_buf_size = sizeof(fpu.floats); 530 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 531 fpu_reg_buf) == fpu_reg_buf_size) { 532 offset += fpu_reg_buf_size; 533 fpu.fpscr = data.GetU32(&offset); 534 SetError(FPURegSet, Read, 0); 535 } else { 536 done = true; 537 } 538 } 539 offset = next_thread_state; 540 break; 541 542 case EXCRegSet: 543 if (count == 3) { 544 exc.exception = data.GetU32(&offset); 545 exc.fsr = data.GetU32(&offset); 546 exc.far = data.GetU32(&offset); 547 SetError(EXCRegSet, Read, 0); 548 } 549 done = true; 550 offset = next_thread_state; 551 break; 552 553 // Unknown register set flavor, stop trying to parse. 554 default: 555 done = true; 556 } 557 } 558 } 559 560 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 561 const char *alt_name, size_t reg_byte_size, 562 Stream &data) { 563 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 564 if (reg_info == NULL) 565 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 566 if (reg_info) { 567 lldb_private::RegisterValue reg_value; 568 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 569 if (reg_info->byte_size >= reg_byte_size) 570 data.Write(reg_value.GetBytes(), reg_byte_size); 571 else { 572 data.Write(reg_value.GetBytes(), reg_info->byte_size); 573 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 574 ++i) 575 data.PutChar(0); 576 } 577 return reg_byte_size; 578 } 579 } 580 // Just write zeros if all else fails 581 for (size_t i = 0; i < reg_byte_size; ++i) 582 data.PutChar(0); 583 return reg_byte_size; 584 } 585 586 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 587 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 588 if (reg_ctx_sp) { 589 RegisterContext *reg_ctx = reg_ctx_sp.get(); 590 591 data.PutHex32(GPRRegSet); // Flavor 592 data.PutHex32(GPRWordCount); 593 WriteRegister(reg_ctx, "r0", NULL, 4, data); 594 WriteRegister(reg_ctx, "r1", NULL, 4, data); 595 WriteRegister(reg_ctx, "r2", NULL, 4, data); 596 WriteRegister(reg_ctx, "r3", NULL, 4, data); 597 WriteRegister(reg_ctx, "r4", NULL, 4, data); 598 WriteRegister(reg_ctx, "r5", NULL, 4, data); 599 WriteRegister(reg_ctx, "r6", NULL, 4, data); 600 WriteRegister(reg_ctx, "r7", NULL, 4, data); 601 WriteRegister(reg_ctx, "r8", NULL, 4, data); 602 WriteRegister(reg_ctx, "r9", NULL, 4, data); 603 WriteRegister(reg_ctx, "r10", NULL, 4, data); 604 WriteRegister(reg_ctx, "r11", NULL, 4, data); 605 WriteRegister(reg_ctx, "r12", NULL, 4, data); 606 WriteRegister(reg_ctx, "sp", NULL, 4, data); 607 WriteRegister(reg_ctx, "lr", NULL, 4, data); 608 WriteRegister(reg_ctx, "pc", NULL, 4, data); 609 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 610 611 // Write out the EXC registers 612 // data.PutHex32 (EXCRegSet); 613 // data.PutHex32 (EXCWordCount); 614 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 615 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 616 // WriteRegister (reg_ctx, "far", NULL, 4, data); 617 return true; 618 } 619 return false; 620 } 621 622 protected: 623 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 624 625 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 626 627 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 628 629 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 630 631 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 632 return 0; 633 } 634 635 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 636 return 0; 637 } 638 639 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 640 return 0; 641 } 642 643 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 644 return -1; 645 } 646 }; 647 648 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 649 public: 650 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 651 const DataExtractor &data) 652 : RegisterContextDarwin_arm64(thread, 0) { 653 SetRegisterDataFrom_LC_THREAD(data); 654 } 655 656 void InvalidateAllRegisters() override { 657 // Do nothing... registers are always valid... 658 } 659 660 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 661 lldb::offset_t offset = 0; 662 SetError(GPRRegSet, Read, -1); 663 SetError(FPURegSet, Read, -1); 664 SetError(EXCRegSet, Read, -1); 665 bool done = false; 666 while (!done) { 667 int flavor = data.GetU32(&offset); 668 uint32_t count = data.GetU32(&offset); 669 lldb::offset_t next_thread_state = offset + (count * 4); 670 switch (flavor) { 671 case GPRRegSet: 672 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 673 // 32-bit register) 674 if (count >= (33 * 2) + 1) { 675 for (uint32_t i = 0; i < 29; ++i) 676 gpr.x[i] = data.GetU64(&offset); 677 gpr.fp = data.GetU64(&offset); 678 gpr.lr = data.GetU64(&offset); 679 gpr.sp = data.GetU64(&offset); 680 gpr.pc = data.GetU64(&offset); 681 gpr.cpsr = data.GetU32(&offset); 682 SetError(GPRRegSet, Read, 0); 683 } 684 offset = next_thread_state; 685 break; 686 case FPURegSet: { 687 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 688 const int fpu_reg_buf_size = sizeof(fpu); 689 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 690 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 691 fpu_reg_buf) == fpu_reg_buf_size) { 692 SetError(FPURegSet, Read, 0); 693 } else { 694 done = true; 695 } 696 } 697 offset = next_thread_state; 698 break; 699 case EXCRegSet: 700 if (count == 4) { 701 exc.far = data.GetU64(&offset); 702 exc.esr = data.GetU32(&offset); 703 exc.exception = data.GetU32(&offset); 704 SetError(EXCRegSet, Read, 0); 705 } 706 offset = next_thread_state; 707 break; 708 default: 709 done = true; 710 break; 711 } 712 } 713 } 714 715 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 716 const char *alt_name, size_t reg_byte_size, 717 Stream &data) { 718 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 719 if (reg_info == NULL) 720 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 721 if (reg_info) { 722 lldb_private::RegisterValue reg_value; 723 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 724 if (reg_info->byte_size >= reg_byte_size) 725 data.Write(reg_value.GetBytes(), reg_byte_size); 726 else { 727 data.Write(reg_value.GetBytes(), reg_info->byte_size); 728 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 729 ++i) 730 data.PutChar(0); 731 } 732 return reg_byte_size; 733 } 734 } 735 // Just write zeros if all else fails 736 for (size_t i = 0; i < reg_byte_size; ++i) 737 data.PutChar(0); 738 return reg_byte_size; 739 } 740 741 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 742 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 743 if (reg_ctx_sp) { 744 RegisterContext *reg_ctx = reg_ctx_sp.get(); 745 746 data.PutHex32(GPRRegSet); // Flavor 747 data.PutHex32(GPRWordCount); 748 WriteRegister(reg_ctx, "x0", NULL, 8, data); 749 WriteRegister(reg_ctx, "x1", NULL, 8, data); 750 WriteRegister(reg_ctx, "x2", NULL, 8, data); 751 WriteRegister(reg_ctx, "x3", NULL, 8, data); 752 WriteRegister(reg_ctx, "x4", NULL, 8, data); 753 WriteRegister(reg_ctx, "x5", NULL, 8, data); 754 WriteRegister(reg_ctx, "x6", NULL, 8, data); 755 WriteRegister(reg_ctx, "x7", NULL, 8, data); 756 WriteRegister(reg_ctx, "x8", NULL, 8, data); 757 WriteRegister(reg_ctx, "x9", NULL, 8, data); 758 WriteRegister(reg_ctx, "x10", NULL, 8, data); 759 WriteRegister(reg_ctx, "x11", NULL, 8, data); 760 WriteRegister(reg_ctx, "x12", NULL, 8, data); 761 WriteRegister(reg_ctx, "x13", NULL, 8, data); 762 WriteRegister(reg_ctx, "x14", NULL, 8, data); 763 WriteRegister(reg_ctx, "x15", NULL, 8, data); 764 WriteRegister(reg_ctx, "x16", NULL, 8, data); 765 WriteRegister(reg_ctx, "x17", NULL, 8, data); 766 WriteRegister(reg_ctx, "x18", NULL, 8, data); 767 WriteRegister(reg_ctx, "x19", NULL, 8, data); 768 WriteRegister(reg_ctx, "x20", NULL, 8, data); 769 WriteRegister(reg_ctx, "x21", NULL, 8, data); 770 WriteRegister(reg_ctx, "x22", NULL, 8, data); 771 WriteRegister(reg_ctx, "x23", NULL, 8, data); 772 WriteRegister(reg_ctx, "x24", NULL, 8, data); 773 WriteRegister(reg_ctx, "x25", NULL, 8, data); 774 WriteRegister(reg_ctx, "x26", NULL, 8, data); 775 WriteRegister(reg_ctx, "x27", NULL, 8, data); 776 WriteRegister(reg_ctx, "x28", NULL, 8, data); 777 WriteRegister(reg_ctx, "fp", NULL, 8, data); 778 WriteRegister(reg_ctx, "lr", NULL, 8, data); 779 WriteRegister(reg_ctx, "sp", NULL, 8, data); 780 WriteRegister(reg_ctx, "pc", NULL, 8, data); 781 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 782 783 // Write out the EXC registers 784 // data.PutHex32 (EXCRegSet); 785 // data.PutHex32 (EXCWordCount); 786 // WriteRegister (reg_ctx, "far", NULL, 8, data); 787 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 788 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 789 return true; 790 } 791 return false; 792 } 793 794 protected: 795 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 796 797 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 798 799 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 800 801 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 802 803 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 804 return 0; 805 } 806 807 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 808 return 0; 809 } 810 811 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 812 return 0; 813 } 814 815 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 816 return -1; 817 } 818 }; 819 820 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 821 switch (magic) { 822 case MH_MAGIC: 823 case MH_CIGAM: 824 return sizeof(struct mach_header); 825 826 case MH_MAGIC_64: 827 case MH_CIGAM_64: 828 return sizeof(struct mach_header_64); 829 break; 830 831 default: 832 break; 833 } 834 return 0; 835 } 836 837 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 838 839 void ObjectFileMachO::Initialize() { 840 PluginManager::RegisterPlugin( 841 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 842 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 843 } 844 845 void ObjectFileMachO::Terminate() { 846 PluginManager::UnregisterPlugin(CreateInstance); 847 } 848 849 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 850 static ConstString g_name("mach-o"); 851 return g_name; 852 } 853 854 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 855 return "Mach-o object file reader (32 and 64 bit)"; 856 } 857 858 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 859 DataBufferSP &data_sp, 860 lldb::offset_t data_offset, 861 const FileSpec *file, 862 lldb::offset_t file_offset, 863 lldb::offset_t length) { 864 if (!data_sp) { 865 data_sp = MapFileData(*file, length, file_offset); 866 if (!data_sp) 867 return nullptr; 868 data_offset = 0; 869 } 870 871 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 872 return nullptr; 873 874 // Update the data to contain the entire file if it doesn't already 875 if (data_sp->GetByteSize() < length) { 876 data_sp = MapFileData(*file, length, file_offset); 877 if (!data_sp) 878 return nullptr; 879 data_offset = 0; 880 } 881 auto objfile_ap = llvm::make_unique<ObjectFileMachO>( 882 module_sp, data_sp, data_offset, file, file_offset, length); 883 if (!objfile_ap || !objfile_ap->ParseHeader()) 884 return nullptr; 885 886 return objfile_ap.release(); 887 } 888 889 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 890 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 891 const ProcessSP &process_sp, lldb::addr_t header_addr) { 892 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 893 std::unique_ptr<ObjectFile> objfile_ap( 894 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 895 if (objfile_ap.get() && objfile_ap->ParseHeader()) 896 return objfile_ap.release(); 897 } 898 return NULL; 899 } 900 901 size_t ObjectFileMachO::GetModuleSpecifications( 902 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 903 lldb::offset_t data_offset, lldb::offset_t file_offset, 904 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 905 const size_t initial_count = specs.GetSize(); 906 907 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 908 DataExtractor data; 909 data.SetData(data_sp); 910 llvm::MachO::mach_header header; 911 if (ParseHeader(data, &data_offset, header)) { 912 size_t header_and_load_cmds = 913 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 914 if (header_and_load_cmds >= data_sp->GetByteSize()) { 915 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 916 data.SetData(data_sp); 917 data_offset = MachHeaderSizeFromMagic(header.magic); 918 } 919 if (data_sp) { 920 ModuleSpec spec; 921 spec.GetFileSpec() = file; 922 spec.SetObjectOffset(file_offset); 923 spec.SetObjectSize(length); 924 925 if (GetArchitecture(header, data, data_offset, 926 spec.GetArchitecture())) { 927 if (spec.GetArchitecture().IsValid()) { 928 GetUUID(header, data, data_offset, spec.GetUUID()); 929 specs.Append(spec); 930 } 931 } 932 } 933 } 934 } 935 return specs.GetSize() - initial_count; 936 } 937 938 const ConstString &ObjectFileMachO::GetSegmentNameTEXT() { 939 static ConstString g_segment_name_TEXT("__TEXT"); 940 return g_segment_name_TEXT; 941 } 942 943 const ConstString &ObjectFileMachO::GetSegmentNameDATA() { 944 static ConstString g_segment_name_DATA("__DATA"); 945 return g_segment_name_DATA; 946 } 947 948 const ConstString &ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 949 static ConstString g_segment_name("__DATA_DIRTY"); 950 return g_segment_name; 951 } 952 953 const ConstString &ObjectFileMachO::GetSegmentNameDATA_CONST() { 954 static ConstString g_segment_name("__DATA_CONST"); 955 return g_segment_name; 956 } 957 958 const ConstString &ObjectFileMachO::GetSegmentNameOBJC() { 959 static ConstString g_segment_name_OBJC("__OBJC"); 960 return g_segment_name_OBJC; 961 } 962 963 const ConstString &ObjectFileMachO::GetSegmentNameLINKEDIT() { 964 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 965 return g_section_name_LINKEDIT; 966 } 967 968 const ConstString &ObjectFileMachO::GetSectionNameEHFrame() { 969 static ConstString g_section_name_eh_frame("__eh_frame"); 970 return g_section_name_eh_frame; 971 } 972 973 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 974 lldb::addr_t data_offset, 975 lldb::addr_t data_length) { 976 DataExtractor data; 977 data.SetData(data_sp, data_offset, data_length); 978 lldb::offset_t offset = 0; 979 uint32_t magic = data.GetU32(&offset); 980 return MachHeaderSizeFromMagic(magic) != 0; 981 } 982 983 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 984 DataBufferSP &data_sp, 985 lldb::offset_t data_offset, 986 const FileSpec *file, 987 lldb::offset_t file_offset, 988 lldb::offset_t length) 989 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 990 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 991 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 992 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 993 ::memset(&m_header, 0, sizeof(m_header)); 994 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 995 } 996 997 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 998 lldb::DataBufferSP &header_data_sp, 999 const lldb::ProcessSP &process_sp, 1000 lldb::addr_t header_addr) 1001 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 1002 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 1003 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 1004 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 1005 ::memset(&m_header, 0, sizeof(m_header)); 1006 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1007 } 1008 1009 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 1010 lldb::offset_t *data_offset_ptr, 1011 llvm::MachO::mach_header &header) { 1012 data.SetByteOrder(endian::InlHostByteOrder()); 1013 // Leave magic in the original byte order 1014 header.magic = data.GetU32(data_offset_ptr); 1015 bool can_parse = false; 1016 bool is_64_bit = false; 1017 switch (header.magic) { 1018 case MH_MAGIC: 1019 data.SetByteOrder(endian::InlHostByteOrder()); 1020 data.SetAddressByteSize(4); 1021 can_parse = true; 1022 break; 1023 1024 case MH_MAGIC_64: 1025 data.SetByteOrder(endian::InlHostByteOrder()); 1026 data.SetAddressByteSize(8); 1027 can_parse = true; 1028 is_64_bit = true; 1029 break; 1030 1031 case MH_CIGAM: 1032 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1033 ? eByteOrderLittle 1034 : eByteOrderBig); 1035 data.SetAddressByteSize(4); 1036 can_parse = true; 1037 break; 1038 1039 case MH_CIGAM_64: 1040 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1041 ? eByteOrderLittle 1042 : eByteOrderBig); 1043 data.SetAddressByteSize(8); 1044 is_64_bit = true; 1045 can_parse = true; 1046 break; 1047 1048 default: 1049 break; 1050 } 1051 1052 if (can_parse) { 1053 data.GetU32(data_offset_ptr, &header.cputype, 6); 1054 if (is_64_bit) 1055 *data_offset_ptr += 4; 1056 return true; 1057 } else { 1058 memset(&header, 0, sizeof(header)); 1059 } 1060 return false; 1061 } 1062 1063 bool ObjectFileMachO::ParseHeader() { 1064 ModuleSP module_sp(GetModule()); 1065 if (module_sp) { 1066 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1067 bool can_parse = false; 1068 lldb::offset_t offset = 0; 1069 m_data.SetByteOrder(endian::InlHostByteOrder()); 1070 // Leave magic in the original byte order 1071 m_header.magic = m_data.GetU32(&offset); 1072 switch (m_header.magic) { 1073 case MH_MAGIC: 1074 m_data.SetByteOrder(endian::InlHostByteOrder()); 1075 m_data.SetAddressByteSize(4); 1076 can_parse = true; 1077 break; 1078 1079 case MH_MAGIC_64: 1080 m_data.SetByteOrder(endian::InlHostByteOrder()); 1081 m_data.SetAddressByteSize(8); 1082 can_parse = true; 1083 break; 1084 1085 case MH_CIGAM: 1086 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1087 ? eByteOrderLittle 1088 : eByteOrderBig); 1089 m_data.SetAddressByteSize(4); 1090 can_parse = true; 1091 break; 1092 1093 case MH_CIGAM_64: 1094 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1095 ? eByteOrderLittle 1096 : eByteOrderBig); 1097 m_data.SetAddressByteSize(8); 1098 can_parse = true; 1099 break; 1100 1101 default: 1102 break; 1103 } 1104 1105 if (can_parse) { 1106 m_data.GetU32(&offset, &m_header.cputype, 6); 1107 1108 ArchSpec mach_arch; 1109 1110 if (GetArchitecture(mach_arch)) { 1111 // Check if the module has a required architecture 1112 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1113 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1114 return false; 1115 1116 if (SetModulesArchitecture(mach_arch)) { 1117 const size_t header_and_lc_size = 1118 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1119 if (m_data.GetByteSize() < header_and_lc_size) { 1120 DataBufferSP data_sp; 1121 ProcessSP process_sp(m_process_wp.lock()); 1122 if (process_sp) { 1123 data_sp = 1124 ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1125 } else { 1126 // Read in all only the load command data from the file on disk 1127 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset); 1128 if (data_sp->GetByteSize() != header_and_lc_size) 1129 return false; 1130 } 1131 if (data_sp) 1132 m_data.SetData(data_sp); 1133 } 1134 } 1135 return true; 1136 } 1137 } else { 1138 memset(&m_header, 0, sizeof(struct mach_header)); 1139 } 1140 } 1141 return false; 1142 } 1143 1144 ByteOrder ObjectFileMachO::GetByteOrder() const { 1145 return m_data.GetByteOrder(); 1146 } 1147 1148 bool ObjectFileMachO::IsExecutable() const { 1149 return m_header.filetype == MH_EXECUTE; 1150 } 1151 1152 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1153 return m_data.GetAddressByteSize(); 1154 } 1155 1156 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1157 Symtab *symtab = GetSymtab(); 1158 if (symtab) { 1159 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1160 if (symbol) { 1161 if (symbol->ValueIsAddress()) { 1162 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1163 if (section_sp) { 1164 const lldb::SectionType section_type = section_sp->GetType(); 1165 switch (section_type) { 1166 case eSectionTypeInvalid: 1167 return eAddressClassUnknown; 1168 1169 case eSectionTypeCode: 1170 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1171 // For ARM we have a bit in the n_desc field of the symbol 1172 // that tells us ARM/Thumb which is bit 0x0008. 1173 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1174 return eAddressClassCodeAlternateISA; 1175 } 1176 return eAddressClassCode; 1177 1178 case eSectionTypeContainer: 1179 return eAddressClassUnknown; 1180 1181 case eSectionTypeData: 1182 case eSectionTypeDataCString: 1183 case eSectionTypeDataCStringPointers: 1184 case eSectionTypeDataSymbolAddress: 1185 case eSectionTypeData4: 1186 case eSectionTypeData8: 1187 case eSectionTypeData16: 1188 case eSectionTypeDataPointers: 1189 case eSectionTypeZeroFill: 1190 case eSectionTypeDataObjCMessageRefs: 1191 case eSectionTypeDataObjCCFStrings: 1192 case eSectionTypeGoSymtab: 1193 return eAddressClassData; 1194 1195 case eSectionTypeDebug: 1196 case eSectionTypeDWARFDebugAbbrev: 1197 case eSectionTypeDWARFDebugAddr: 1198 case eSectionTypeDWARFDebugAranges: 1199 case eSectionTypeDWARFDebugCuIndex: 1200 case eSectionTypeDWARFDebugFrame: 1201 case eSectionTypeDWARFDebugInfo: 1202 case eSectionTypeDWARFDebugLine: 1203 case eSectionTypeDWARFDebugLoc: 1204 case eSectionTypeDWARFDebugMacInfo: 1205 case eSectionTypeDWARFDebugMacro: 1206 case eSectionTypeDWARFDebugPubNames: 1207 case eSectionTypeDWARFDebugPubTypes: 1208 case eSectionTypeDWARFDebugRanges: 1209 case eSectionTypeDWARFDebugStr: 1210 case eSectionTypeDWARFDebugStrOffsets: 1211 case eSectionTypeDWARFAppleNames: 1212 case eSectionTypeDWARFAppleTypes: 1213 case eSectionTypeDWARFAppleNamespaces: 1214 case eSectionTypeDWARFAppleObjC: 1215 return eAddressClassDebug; 1216 1217 case eSectionTypeEHFrame: 1218 case eSectionTypeARMexidx: 1219 case eSectionTypeARMextab: 1220 case eSectionTypeCompactUnwind: 1221 return eAddressClassRuntime; 1222 1223 case eSectionTypeAbsoluteAddress: 1224 case eSectionTypeELFSymbolTable: 1225 case eSectionTypeELFDynamicSymbols: 1226 case eSectionTypeELFRelocationEntries: 1227 case eSectionTypeELFDynamicLinkInfo: 1228 case eSectionTypeOther: 1229 return eAddressClassUnknown; 1230 } 1231 } 1232 } 1233 1234 const SymbolType symbol_type = symbol->GetType(); 1235 switch (symbol_type) { 1236 case eSymbolTypeAny: 1237 return eAddressClassUnknown; 1238 case eSymbolTypeAbsolute: 1239 return eAddressClassUnknown; 1240 1241 case eSymbolTypeCode: 1242 case eSymbolTypeTrampoline: 1243 case eSymbolTypeResolver: 1244 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1245 // For ARM we have a bit in the n_desc field of the symbol 1246 // that tells us ARM/Thumb which is bit 0x0008. 1247 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1248 return eAddressClassCodeAlternateISA; 1249 } 1250 return eAddressClassCode; 1251 1252 case eSymbolTypeData: 1253 return eAddressClassData; 1254 case eSymbolTypeRuntime: 1255 return eAddressClassRuntime; 1256 case eSymbolTypeException: 1257 return eAddressClassRuntime; 1258 case eSymbolTypeSourceFile: 1259 return eAddressClassDebug; 1260 case eSymbolTypeHeaderFile: 1261 return eAddressClassDebug; 1262 case eSymbolTypeObjectFile: 1263 return eAddressClassDebug; 1264 case eSymbolTypeCommonBlock: 1265 return eAddressClassDebug; 1266 case eSymbolTypeBlock: 1267 return eAddressClassDebug; 1268 case eSymbolTypeLocal: 1269 return eAddressClassData; 1270 case eSymbolTypeParam: 1271 return eAddressClassData; 1272 case eSymbolTypeVariable: 1273 return eAddressClassData; 1274 case eSymbolTypeVariableType: 1275 return eAddressClassDebug; 1276 case eSymbolTypeLineEntry: 1277 return eAddressClassDebug; 1278 case eSymbolTypeLineHeader: 1279 return eAddressClassDebug; 1280 case eSymbolTypeScopeBegin: 1281 return eAddressClassDebug; 1282 case eSymbolTypeScopeEnd: 1283 return eAddressClassDebug; 1284 case eSymbolTypeAdditional: 1285 return eAddressClassUnknown; 1286 case eSymbolTypeCompiler: 1287 return eAddressClassDebug; 1288 case eSymbolTypeInstrumentation: 1289 return eAddressClassDebug; 1290 case eSymbolTypeUndefined: 1291 return eAddressClassUnknown; 1292 case eSymbolTypeObjCClass: 1293 return eAddressClassRuntime; 1294 case eSymbolTypeObjCMetaClass: 1295 return eAddressClassRuntime; 1296 case eSymbolTypeObjCIVar: 1297 return eAddressClassRuntime; 1298 case eSymbolTypeReExported: 1299 return eAddressClassRuntime; 1300 } 1301 } 1302 } 1303 return eAddressClassUnknown; 1304 } 1305 1306 Symtab *ObjectFileMachO::GetSymtab() { 1307 ModuleSP module_sp(GetModule()); 1308 if (module_sp) { 1309 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1310 if (m_symtab_ap.get() == NULL) { 1311 m_symtab_ap.reset(new Symtab(this)); 1312 std::lock_guard<std::recursive_mutex> symtab_guard( 1313 m_symtab_ap->GetMutex()); 1314 ParseSymtab(); 1315 m_symtab_ap->Finalize(); 1316 } 1317 } 1318 return m_symtab_ap.get(); 1319 } 1320 1321 bool ObjectFileMachO::IsStripped() { 1322 if (m_dysymtab.cmd == 0) { 1323 ModuleSP module_sp(GetModule()); 1324 if (module_sp) { 1325 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1326 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1327 const lldb::offset_t load_cmd_offset = offset; 1328 1329 load_command lc; 1330 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1331 break; 1332 if (lc.cmd == LC_DYSYMTAB) { 1333 m_dysymtab.cmd = lc.cmd; 1334 m_dysymtab.cmdsize = lc.cmdsize; 1335 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1336 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1337 NULL) { 1338 // Clear m_dysymtab if we were unable to read all items from the 1339 // load command 1340 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1341 } 1342 } 1343 offset = load_cmd_offset + lc.cmdsize; 1344 } 1345 } 1346 } 1347 if (m_dysymtab.cmd) 1348 return m_dysymtab.nlocalsym <= 1; 1349 return false; 1350 } 1351 1352 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1353 if (!m_sections_ap.get()) { 1354 m_sections_ap.reset(new SectionList()); 1355 1356 const bool is_dsym = (m_header.filetype == MH_DSYM); 1357 lldb::user_id_t segID = 0; 1358 lldb::user_id_t sectID = 0; 1359 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1360 uint32_t i; 1361 const bool is_core = GetType() == eTypeCoreFile; 1362 // bool dump_sections = false; 1363 ModuleSP module_sp(GetModule()); 1364 // First look up any LC_ENCRYPTION_INFO load commands 1365 typedef RangeArray<uint32_t, uint32_t, 8> EncryptedFileRanges; 1366 EncryptedFileRanges encrypted_file_ranges; 1367 encryption_info_command encryption_cmd; 1368 for (i = 0; i < m_header.ncmds; ++i) { 1369 const lldb::offset_t load_cmd_offset = offset; 1370 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1371 break; 1372 1373 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for 1374 // the 3 fields we care about, so treat them the same. 1375 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1376 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1377 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1378 if (encryption_cmd.cryptid != 0) { 1379 EncryptedFileRanges::Entry entry; 1380 entry.SetRangeBase(encryption_cmd.cryptoff); 1381 entry.SetByteSize(encryption_cmd.cryptsize); 1382 encrypted_file_ranges.Append(entry); 1383 } 1384 } 1385 } 1386 offset = load_cmd_offset + encryption_cmd.cmdsize; 1387 } 1388 1389 bool section_file_addresses_changed = false; 1390 1391 offset = MachHeaderSizeFromMagic(m_header.magic); 1392 1393 struct segment_command_64 load_cmd; 1394 for (i = 0; i < m_header.ncmds; ++i) { 1395 const lldb::offset_t load_cmd_offset = offset; 1396 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1397 break; 1398 1399 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) { 1400 if (m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) { 1401 bool add_section = true; 1402 bool add_to_unified = true; 1403 ConstString const_segname(load_cmd.segname, 1404 std::min<size_t>(strlen(load_cmd.segname), 1405 sizeof(load_cmd.segname))); 1406 1407 SectionSP unified_section_sp( 1408 unified_section_list.FindSectionByName(const_segname)); 1409 if (is_dsym && unified_section_sp) { 1410 if (const_segname == GetSegmentNameLINKEDIT()) { 1411 // We need to keep the __LINKEDIT segment private to this object 1412 // file only 1413 add_to_unified = false; 1414 } else { 1415 // This is the dSYM file and this section has already been created 1416 // by 1417 // the object file, no need to create it. 1418 add_section = false; 1419 } 1420 } 1421 load_cmd.vmaddr = m_data.GetAddress(&offset); 1422 load_cmd.vmsize = m_data.GetAddress(&offset); 1423 load_cmd.fileoff = m_data.GetAddress(&offset); 1424 load_cmd.filesize = m_data.GetAddress(&offset); 1425 if (m_length != 0 && load_cmd.filesize != 0) { 1426 if (load_cmd.fileoff > m_length) { 1427 // We have a load command that says it extends past the end of the 1428 // file. This is likely 1429 // a corrupt file. We don't have any way to return an error 1430 // condition here (this method 1431 // was likely invoked from something like 1432 // ObjectFile::GetSectionList()) -- all we can do 1433 // is null out the SectionList vector and if a process has been 1434 // set up, dump a message 1435 // to stdout. The most common case here is core file debugging 1436 // with a truncated file. 1437 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 1438 ? "LC_SEGMENT_64" 1439 : "LC_SEGMENT"; 1440 module_sp->ReportWarning( 1441 "load command %u %s has a fileoff (0x%" PRIx64 1442 ") that extends beyond the end of the file (0x%" PRIx64 1443 "), ignoring this section", 1444 i, lc_segment_name, load_cmd.fileoff, m_length); 1445 1446 load_cmd.fileoff = 0; 1447 load_cmd.filesize = 0; 1448 } 1449 1450 if (load_cmd.fileoff + load_cmd.filesize > m_length) { 1451 // We have a load command that says it extends past the end of the 1452 // file. This is likely 1453 // a corrupt file. We don't have any way to return an error 1454 // condition here (this method 1455 // was likely invoked from something like 1456 // ObjectFile::GetSectionList()) -- all we can do 1457 // is null out the SectionList vector and if a process has been 1458 // set up, dump a message 1459 // to stdout. The most common case here is core file debugging 1460 // with a truncated file. 1461 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 1462 ? "LC_SEGMENT_64" 1463 : "LC_SEGMENT"; 1464 GetModule()->ReportWarning( 1465 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1466 ") that extends beyond the end of the file (0x%" PRIx64 1467 "), the segment will be truncated to match", 1468 i, lc_segment_name, load_cmd.fileoff + load_cmd.filesize, 1469 m_length); 1470 1471 // Tuncase the length 1472 load_cmd.filesize = m_length - load_cmd.fileoff; 1473 } 1474 } 1475 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) { 1476 uint32_t segment_permissions = 0; 1477 if (load_cmd.initprot & VM_PROT_READ) 1478 segment_permissions |= ePermissionsReadable; 1479 if (load_cmd.initprot & VM_PROT_WRITE) 1480 segment_permissions |= ePermissionsWritable; 1481 if (load_cmd.initprot & VM_PROT_EXECUTE) 1482 segment_permissions |= ePermissionsExecutable; 1483 1484 const bool segment_is_encrypted = 1485 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1486 1487 // Keep a list of mach segments around in case we need to 1488 // get at data that isn't stored in the abstracted Sections. 1489 m_mach_segments.push_back(load_cmd); 1490 1491 // Use a segment ID of the segment index shifted left by 8 so they 1492 // never conflict with any of the sections. 1493 SectionSP segment_sp; 1494 if (add_section && (const_segname || is_core)) { 1495 segment_sp.reset(new Section( 1496 module_sp, // Module to which this section belongs 1497 this, // Object file to which this sections belongs 1498 ++segID << 8, // Section ID is the 1 based segment index 1499 // shifted right by 8 bits as not to collide 1500 // with any of the 256 section IDs that are 1501 // possible 1502 const_segname, // Name of this section 1503 eSectionTypeContainer, // This section is a container of other 1504 // sections. 1505 load_cmd.vmaddr, // File VM address == addresses as they are 1506 // found in the object file 1507 load_cmd.vmsize, // VM size in bytes of this section 1508 load_cmd.fileoff, // Offset to the data for this section in 1509 // the file 1510 load_cmd.filesize, // Size in bytes of this section as found 1511 // in the file 1512 0, // Segments have no alignment information 1513 load_cmd.flags)); // Flags for this section 1514 1515 segment_sp->SetIsEncrypted(segment_is_encrypted); 1516 m_sections_ap->AddSection(segment_sp); 1517 segment_sp->SetPermissions(segment_permissions); 1518 if (add_to_unified) 1519 unified_section_list.AddSection(segment_sp); 1520 } else if (unified_section_sp) { 1521 if (is_dsym && 1522 unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1523 // Check to see if the module was read from memory? 1524 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) { 1525 // We have a module that is in memory and needs to have its 1526 // file address adjusted. We need to do this because when we 1527 // load a file from memory, its addresses will be slid 1528 // already, 1529 // yet the addresses in the new symbol file will still be 1530 // unslid. 1531 // Since everything is stored as section offset, this 1532 // shouldn't 1533 // cause any problems. 1534 1535 // Make sure we've parsed the symbol table from the 1536 // ObjectFile before we go around changing its Sections. 1537 module_sp->GetObjectFile()->GetSymtab(); 1538 // eh_frame would present the same problems but we parse that 1539 // on 1540 // a per-function basis as-needed so it's more difficult to 1541 // remove its use of the Sections. Realistically, the 1542 // environments 1543 // where this code path will be taken will not have eh_frame 1544 // sections. 1545 1546 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1547 1548 // Notify the module that the section addresses have been 1549 // changed once 1550 // we're done so any file-address caches can be updated. 1551 section_file_addresses_changed = true; 1552 } 1553 } 1554 m_sections_ap->AddSection(unified_section_sp); 1555 } 1556 1557 struct section_64 sect64; 1558 ::memset(§64, 0, sizeof(sect64)); 1559 // Push a section into our mach sections for the section at 1560 // index zero (NO_SECT) if we don't have any mach sections yet... 1561 if (m_mach_sections.empty()) 1562 m_mach_sections.push_back(sect64); 1563 uint32_t segment_sect_idx; 1564 const lldb::user_id_t first_segment_sectID = sectID + 1; 1565 1566 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1567 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1568 ++segment_sect_idx) { 1569 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1570 sizeof(sect64.sectname)) == NULL) 1571 break; 1572 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1573 sizeof(sect64.segname)) == NULL) 1574 break; 1575 sect64.addr = m_data.GetAddress(&offset); 1576 sect64.size = m_data.GetAddress(&offset); 1577 1578 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1579 break; 1580 1581 // Keep a list of mach sections around in case we need to 1582 // get at data that isn't stored in the abstracted Sections. 1583 m_mach_sections.push_back(sect64); 1584 1585 if (add_section) { 1586 ConstString section_name( 1587 sect64.sectname, std::min<size_t>(strlen(sect64.sectname), 1588 sizeof(sect64.sectname))); 1589 if (!const_segname) { 1590 // We have a segment with no name so we need to conjure up 1591 // segments that correspond to the section's segname if there 1592 // isn't already such a section. If there is such a section, 1593 // we resize the section so that it spans all sections. 1594 // We also mark these sections as fake so address matches 1595 // don't 1596 // hit if they land in the gaps between the child sections. 1597 const_segname.SetTrimmedCStringWithLength( 1598 sect64.segname, sizeof(sect64.segname)); 1599 segment_sp = 1600 unified_section_list.FindSectionByName(const_segname); 1601 if (segment_sp.get()) { 1602 Section *segment = segment_sp.get(); 1603 // Grow the section size as needed. 1604 const lldb::addr_t sect64_min_addr = sect64.addr; 1605 const lldb::addr_t sect64_max_addr = 1606 sect64_min_addr + sect64.size; 1607 const lldb::addr_t curr_seg_byte_size = 1608 segment->GetByteSize(); 1609 const lldb::addr_t curr_seg_min_addr = 1610 segment->GetFileAddress(); 1611 const lldb::addr_t curr_seg_max_addr = 1612 curr_seg_min_addr + curr_seg_byte_size; 1613 if (sect64_min_addr >= curr_seg_min_addr) { 1614 const lldb::addr_t new_seg_byte_size = 1615 sect64_max_addr - curr_seg_min_addr; 1616 // Only grow the section size if needed 1617 if (new_seg_byte_size > curr_seg_byte_size) 1618 segment->SetByteSize(new_seg_byte_size); 1619 } else { 1620 // We need to change the base address of the segment and 1621 // adjust the child section offsets for all existing 1622 // children. 1623 const lldb::addr_t slide_amount = 1624 sect64_min_addr - curr_seg_min_addr; 1625 segment->Slide(slide_amount, false); 1626 segment->GetChildren().Slide(-slide_amount, false); 1627 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1628 } 1629 1630 // Grow the section size as needed. 1631 if (sect64.offset) { 1632 const lldb::addr_t segment_min_file_offset = 1633 segment->GetFileOffset(); 1634 const lldb::addr_t segment_max_file_offset = 1635 segment_min_file_offset + segment->GetFileSize(); 1636 1637 const lldb::addr_t section_min_file_offset = 1638 sect64.offset; 1639 const lldb::addr_t section_max_file_offset = 1640 section_min_file_offset + sect64.size; 1641 const lldb::addr_t new_file_offset = std::min( 1642 section_min_file_offset, segment_min_file_offset); 1643 const lldb::addr_t new_file_size = 1644 std::max(section_max_file_offset, 1645 segment_max_file_offset) - 1646 new_file_offset; 1647 segment->SetFileOffset(new_file_offset); 1648 segment->SetFileSize(new_file_size); 1649 } 1650 } else { 1651 // Create a fake section for the section's named segment 1652 segment_sp.reset(new Section( 1653 segment_sp, // Parent section 1654 module_sp, // Module to which this section belongs 1655 this, // Object file to which this section belongs 1656 ++segID << 8, // Section ID is the 1 based segment index 1657 // shifted right by 8 bits as not to 1658 // collide with any of the 256 section IDs 1659 // that are possible 1660 const_segname, // Name of this section 1661 eSectionTypeContainer, // This section is a container of 1662 // other sections. 1663 sect64.addr, // File VM address == addresses as they are 1664 // found in the object file 1665 sect64.size, // VM size in bytes of this section 1666 sect64.offset, // Offset to the data for this section in 1667 // the file 1668 sect64.offset ? sect64.size : 0, // Size in bytes of 1669 // this section as 1670 // found in the file 1671 sect64.align, 1672 load_cmd.flags)); // Flags for this section 1673 segment_sp->SetIsFake(true); 1674 segment_sp->SetPermissions(segment_permissions); 1675 m_sections_ap->AddSection(segment_sp); 1676 if (add_to_unified) 1677 unified_section_list.AddSection(segment_sp); 1678 segment_sp->SetIsEncrypted(segment_is_encrypted); 1679 } 1680 } 1681 assert(segment_sp.get()); 1682 1683 lldb::SectionType sect_type = eSectionTypeOther; 1684 1685 if (sect64.flags & 1686 (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1687 sect_type = eSectionTypeCode; 1688 else { 1689 uint32_t mach_sect_type = sect64.flags & SECTION_TYPE; 1690 static ConstString g_sect_name_objc_data("__objc_data"); 1691 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1692 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1693 static ConstString g_sect_name_objc_classrefs( 1694 "__objc_classrefs"); 1695 static ConstString g_sect_name_objc_superrefs( 1696 "__objc_superrefs"); 1697 static ConstString g_sect_name_objc_const("__objc_const"); 1698 static ConstString g_sect_name_objc_classlist( 1699 "__objc_classlist"); 1700 static ConstString g_sect_name_cfstring("__cfstring"); 1701 1702 static ConstString g_sect_name_dwarf_debug_abbrev( 1703 "__debug_abbrev"); 1704 static ConstString g_sect_name_dwarf_debug_aranges( 1705 "__debug_aranges"); 1706 static ConstString g_sect_name_dwarf_debug_frame( 1707 "__debug_frame"); 1708 static ConstString g_sect_name_dwarf_debug_info( 1709 "__debug_info"); 1710 static ConstString g_sect_name_dwarf_debug_line( 1711 "__debug_line"); 1712 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1713 static ConstString g_sect_name_dwarf_debug_macinfo( 1714 "__debug_macinfo"); 1715 static ConstString g_sect_name_dwarf_debug_pubnames( 1716 "__debug_pubnames"); 1717 static ConstString g_sect_name_dwarf_debug_pubtypes( 1718 "__debug_pubtypes"); 1719 static ConstString g_sect_name_dwarf_debug_ranges( 1720 "__debug_ranges"); 1721 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1722 static ConstString g_sect_name_dwarf_apple_names( 1723 "__apple_names"); 1724 static ConstString g_sect_name_dwarf_apple_types( 1725 "__apple_types"); 1726 static ConstString g_sect_name_dwarf_apple_namespaces( 1727 "__apple_namespac"); 1728 static ConstString g_sect_name_dwarf_apple_objc( 1729 "__apple_objc"); 1730 static ConstString g_sect_name_eh_frame("__eh_frame"); 1731 static ConstString g_sect_name_compact_unwind( 1732 "__unwind_info"); 1733 static ConstString g_sect_name_text("__text"); 1734 static ConstString g_sect_name_data("__data"); 1735 static ConstString g_sect_name_go_symtab("__gosymtab"); 1736 1737 if (section_name == g_sect_name_dwarf_debug_abbrev) 1738 sect_type = eSectionTypeDWARFDebugAbbrev; 1739 else if (section_name == g_sect_name_dwarf_debug_aranges) 1740 sect_type = eSectionTypeDWARFDebugAranges; 1741 else if (section_name == g_sect_name_dwarf_debug_frame) 1742 sect_type = eSectionTypeDWARFDebugFrame; 1743 else if (section_name == g_sect_name_dwarf_debug_info) 1744 sect_type = eSectionTypeDWARFDebugInfo; 1745 else if (section_name == g_sect_name_dwarf_debug_line) 1746 sect_type = eSectionTypeDWARFDebugLine; 1747 else if (section_name == g_sect_name_dwarf_debug_loc) 1748 sect_type = eSectionTypeDWARFDebugLoc; 1749 else if (section_name == g_sect_name_dwarf_debug_macinfo) 1750 sect_type = eSectionTypeDWARFDebugMacInfo; 1751 else if (section_name == g_sect_name_dwarf_debug_pubnames) 1752 sect_type = eSectionTypeDWARFDebugPubNames; 1753 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 1754 sect_type = eSectionTypeDWARFDebugPubTypes; 1755 else if (section_name == g_sect_name_dwarf_debug_ranges) 1756 sect_type = eSectionTypeDWARFDebugRanges; 1757 else if (section_name == g_sect_name_dwarf_debug_str) 1758 sect_type = eSectionTypeDWARFDebugStr; 1759 else if (section_name == g_sect_name_dwarf_apple_names) 1760 sect_type = eSectionTypeDWARFAppleNames; 1761 else if (section_name == g_sect_name_dwarf_apple_types) 1762 sect_type = eSectionTypeDWARFAppleTypes; 1763 else if (section_name == g_sect_name_dwarf_apple_namespaces) 1764 sect_type = eSectionTypeDWARFAppleNamespaces; 1765 else if (section_name == g_sect_name_dwarf_apple_objc) 1766 sect_type = eSectionTypeDWARFAppleObjC; 1767 else if (section_name == g_sect_name_objc_selrefs) 1768 sect_type = eSectionTypeDataCStringPointers; 1769 else if (section_name == g_sect_name_objc_msgrefs) 1770 sect_type = eSectionTypeDataObjCMessageRefs; 1771 else if (section_name == g_sect_name_eh_frame) 1772 sect_type = eSectionTypeEHFrame; 1773 else if (section_name == g_sect_name_compact_unwind) 1774 sect_type = eSectionTypeCompactUnwind; 1775 else if (section_name == g_sect_name_cfstring) 1776 sect_type = eSectionTypeDataObjCCFStrings; 1777 else if (section_name == g_sect_name_go_symtab) 1778 sect_type = eSectionTypeGoSymtab; 1779 else if (section_name == g_sect_name_objc_data || 1780 section_name == g_sect_name_objc_classrefs || 1781 section_name == g_sect_name_objc_superrefs || 1782 section_name == g_sect_name_objc_const || 1783 section_name == g_sect_name_objc_classlist) { 1784 sect_type = eSectionTypeDataPointers; 1785 } 1786 1787 if (sect_type == eSectionTypeOther) { 1788 switch (mach_sect_type) { 1789 // TODO: categorize sections by other flags for regular 1790 // sections 1791 case S_REGULAR: 1792 if (section_name == g_sect_name_text) 1793 sect_type = eSectionTypeCode; 1794 else if (section_name == g_sect_name_data) 1795 sect_type = eSectionTypeData; 1796 else 1797 sect_type = eSectionTypeOther; 1798 break; 1799 case S_ZEROFILL: 1800 sect_type = eSectionTypeZeroFill; 1801 break; 1802 case S_CSTRING_LITERALS: 1803 sect_type = eSectionTypeDataCString; 1804 break; // section with only literal C strings 1805 case S_4BYTE_LITERALS: 1806 sect_type = eSectionTypeData4; 1807 break; // section with only 4 byte literals 1808 case S_8BYTE_LITERALS: 1809 sect_type = eSectionTypeData8; 1810 break; // section with only 8 byte literals 1811 case S_LITERAL_POINTERS: 1812 sect_type = eSectionTypeDataPointers; 1813 break; // section with only pointers to literals 1814 case S_NON_LAZY_SYMBOL_POINTERS: 1815 sect_type = eSectionTypeDataPointers; 1816 break; // section with only non-lazy symbol pointers 1817 case S_LAZY_SYMBOL_POINTERS: 1818 sect_type = eSectionTypeDataPointers; 1819 break; // section with only lazy symbol pointers 1820 case S_SYMBOL_STUBS: 1821 sect_type = eSectionTypeCode; 1822 break; // section with only symbol stubs, byte size of 1823 // stub in the reserved2 field 1824 case S_MOD_INIT_FUNC_POINTERS: 1825 sect_type = eSectionTypeDataPointers; 1826 break; // section with only function pointers for 1827 // initialization 1828 case S_MOD_TERM_FUNC_POINTERS: 1829 sect_type = eSectionTypeDataPointers; 1830 break; // section with only function pointers for 1831 // termination 1832 case S_COALESCED: 1833 sect_type = eSectionTypeOther; 1834 break; 1835 case S_GB_ZEROFILL: 1836 sect_type = eSectionTypeZeroFill; 1837 break; 1838 case S_INTERPOSING: 1839 sect_type = eSectionTypeCode; 1840 break; // section with only pairs of function pointers for 1841 // interposing 1842 case S_16BYTE_LITERALS: 1843 sect_type = eSectionTypeData16; 1844 break; // section with only 16 byte literals 1845 case S_DTRACE_DOF: 1846 sect_type = eSectionTypeDebug; 1847 break; 1848 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1849 sect_type = eSectionTypeDataPointers; 1850 break; 1851 default: 1852 break; 1853 } 1854 } 1855 } 1856 1857 SectionSP section_sp(new Section( 1858 segment_sp, module_sp, this, ++sectID, section_name, 1859 sect_type, sect64.addr - segment_sp->GetFileAddress(), 1860 sect64.size, sect64.offset, 1861 sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1862 sect64.flags)); 1863 // Set the section to be encrypted to match the segment 1864 1865 bool section_is_encrypted = false; 1866 if (!segment_is_encrypted && load_cmd.filesize != 0) 1867 section_is_encrypted = 1868 encrypted_file_ranges.FindEntryThatContains( 1869 sect64.offset) != NULL; 1870 1871 section_sp->SetIsEncrypted(segment_is_encrypted || 1872 section_is_encrypted); 1873 section_sp->SetPermissions(segment_permissions); 1874 segment_sp->GetChildren().AddSection(section_sp); 1875 1876 if (segment_sp->IsFake()) { 1877 segment_sp.reset(); 1878 const_segname.Clear(); 1879 } 1880 } 1881 } 1882 if (segment_sp && is_dsym) { 1883 if (first_segment_sectID <= sectID) { 1884 lldb::user_id_t sect_uid; 1885 for (sect_uid = first_segment_sectID; sect_uid <= sectID; 1886 ++sect_uid) { 1887 SectionSP curr_section_sp( 1888 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1889 SectionSP next_section_sp; 1890 if (sect_uid + 1 <= sectID) 1891 next_section_sp = 1892 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1893 1894 if (curr_section_sp.get()) { 1895 if (curr_section_sp->GetByteSize() == 0) { 1896 if (next_section_sp.get() != NULL) 1897 curr_section_sp->SetByteSize( 1898 next_section_sp->GetFileAddress() - 1899 curr_section_sp->GetFileAddress()); 1900 else 1901 curr_section_sp->SetByteSize(load_cmd.vmsize); 1902 } 1903 } 1904 } 1905 } 1906 } 1907 } 1908 } 1909 } else if (load_cmd.cmd == LC_DYSYMTAB) { 1910 m_dysymtab.cmd = load_cmd.cmd; 1911 m_dysymtab.cmdsize = load_cmd.cmdsize; 1912 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1913 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1914 } 1915 1916 offset = load_cmd_offset + load_cmd.cmdsize; 1917 } 1918 1919 if (section_file_addresses_changed && module_sp.get()) { 1920 module_sp->SectionFileAddressesChanged(); 1921 } 1922 } 1923 } 1924 1925 class MachSymtabSectionInfo { 1926 public: 1927 MachSymtabSectionInfo(SectionList *section_list) 1928 : m_section_list(section_list), m_section_infos() { 1929 // Get the number of sections down to a depth of 1 to include 1930 // all segments and their sections, but no other sections that 1931 // may be added for debug map or 1932 m_section_infos.resize(section_list->GetNumSections(1)); 1933 } 1934 1935 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1936 if (n_sect == 0) 1937 return SectionSP(); 1938 if (n_sect < m_section_infos.size()) { 1939 if (!m_section_infos[n_sect].section_sp) { 1940 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1941 m_section_infos[n_sect].section_sp = section_sp; 1942 if (section_sp) { 1943 m_section_infos[n_sect].vm_range.SetBaseAddress( 1944 section_sp->GetFileAddress()); 1945 m_section_infos[n_sect].vm_range.SetByteSize( 1946 section_sp->GetByteSize()); 1947 } else { 1948 Host::SystemLog(Host::eSystemLogError, 1949 "error: unable to find section for section %u\n", 1950 n_sect); 1951 } 1952 } 1953 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1954 // Symbol is in section. 1955 return m_section_infos[n_sect].section_sp; 1956 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1957 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1958 file_addr) { 1959 // Symbol is in section with zero size, but has the same start 1960 // address as the section. This can happen with linker symbols 1961 // (symbols that start with the letter 'l' or 'L'. 1962 return m_section_infos[n_sect].section_sp; 1963 } 1964 } 1965 return m_section_list->FindSectionContainingFileAddress(file_addr); 1966 } 1967 1968 protected: 1969 struct SectionInfo { 1970 SectionInfo() : vm_range(), section_sp() {} 1971 1972 VMRange vm_range; 1973 SectionSP section_sp; 1974 }; 1975 SectionList *m_section_list; 1976 std::vector<SectionInfo> m_section_infos; 1977 }; 1978 1979 struct TrieEntry { 1980 TrieEntry() 1981 : name(), address(LLDB_INVALID_ADDRESS), flags(0), other(0), 1982 import_name() {} 1983 1984 void Clear() { 1985 name.Clear(); 1986 address = LLDB_INVALID_ADDRESS; 1987 flags = 0; 1988 other = 0; 1989 import_name.Clear(); 1990 } 1991 1992 void Dump() const { 1993 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1994 static_cast<unsigned long long>(address), 1995 static_cast<unsigned long long>(flags), 1996 static_cast<unsigned long long>(other), name.GetCString()); 1997 if (import_name) 1998 printf(" -> \"%s\"\n", import_name.GetCString()); 1999 else 2000 printf("\n"); 2001 } 2002 ConstString name; 2003 uint64_t address; 2004 uint64_t flags; 2005 uint64_t other; 2006 ConstString import_name; 2007 }; 2008 2009 struct TrieEntryWithOffset { 2010 lldb::offset_t nodeOffset; 2011 TrieEntry entry; 2012 2013 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 2014 2015 void Dump(uint32_t idx) const { 2016 printf("[%3u] 0x%16.16llx: ", idx, 2017 static_cast<unsigned long long>(nodeOffset)); 2018 entry.Dump(); 2019 } 2020 2021 bool operator<(const TrieEntryWithOffset &other) const { 2022 return (nodeOffset < other.nodeOffset); 2023 } 2024 }; 2025 2026 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2027 const bool is_arm, 2028 std::vector<llvm::StringRef> &nameSlices, 2029 std::set<lldb::addr_t> &resolver_addresses, 2030 std::vector<TrieEntryWithOffset> &output) { 2031 if (!data.ValidOffset(offset)) 2032 return true; 2033 2034 const uint64_t terminalSize = data.GetULEB128(&offset); 2035 lldb::offset_t children_offset = offset + terminalSize; 2036 if (terminalSize != 0) { 2037 TrieEntryWithOffset e(offset); 2038 e.entry.flags = data.GetULEB128(&offset); 2039 const char *import_name = NULL; 2040 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2041 e.entry.address = 0; 2042 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2043 import_name = data.GetCStr(&offset); 2044 } else { 2045 e.entry.address = data.GetULEB128(&offset); 2046 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2047 e.entry.other = data.GetULEB128(&offset); 2048 uint64_t resolver_addr = e.entry.other; 2049 if (is_arm) 2050 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2051 resolver_addresses.insert(resolver_addr); 2052 } else 2053 e.entry.other = 0; 2054 } 2055 // Only add symbols that are reexport symbols with a valid import name 2056 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && 2057 import_name[0]) { 2058 std::string name; 2059 if (!nameSlices.empty()) { 2060 for (auto name_slice : nameSlices) 2061 name.append(name_slice.data(), name_slice.size()); 2062 } 2063 if (name.size() > 1) { 2064 // Skip the leading '_' 2065 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2066 } 2067 if (import_name) { 2068 // Skip the leading '_' 2069 e.entry.import_name.SetCString(import_name + 1); 2070 } 2071 output.push_back(e); 2072 } 2073 } 2074 2075 const uint8_t childrenCount = data.GetU8(&children_offset); 2076 for (uint8_t i = 0; i < childrenCount; ++i) { 2077 const char *cstr = data.GetCStr(&children_offset); 2078 if (cstr) 2079 nameSlices.push_back(llvm::StringRef(cstr)); 2080 else 2081 return false; // Corrupt data 2082 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2083 if (childNodeOffset) { 2084 if (!ParseTrieEntries(data, childNodeOffset, is_arm, nameSlices, 2085 resolver_addresses, output)) { 2086 return false; 2087 } 2088 } 2089 nameSlices.pop_back(); 2090 } 2091 return true; 2092 } 2093 2094 // Read the UUID out of a dyld_shared_cache file on-disk. 2095 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2096 const ByteOrder byte_order, 2097 const uint32_t addr_byte_size) { 2098 UUID dsc_uuid; 2099 DataBufferSP DscData = MapFileData( 2100 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2101 if (!DscData) 2102 return dsc_uuid; 2103 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2104 2105 char version_str[7]; 2106 lldb::offset_t offset = 0; 2107 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2108 version_str[6] = '\0'; 2109 if (strcmp(version_str, "dyld_v") == 0) { 2110 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2111 uint8_t uuid_bytes[sizeof(uuid_t)]; 2112 memcpy(uuid_bytes, dsc_header_data.GetData(&offset, sizeof(uuid_t)), 2113 sizeof(uuid_t)); 2114 dsc_uuid.SetBytes(uuid_bytes); 2115 } 2116 return dsc_uuid; 2117 } 2118 2119 size_t ObjectFileMachO::ParseSymtab() { 2120 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2121 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2122 m_file.GetFilename().AsCString("")); 2123 ModuleSP module_sp(GetModule()); 2124 if (!module_sp) 2125 return 0; 2126 2127 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2128 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2129 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2130 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2131 FunctionStarts function_starts; 2132 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2133 uint32_t i; 2134 FileSpecList dylib_files; 2135 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2136 static const llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2137 static const llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2138 static const llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2139 2140 for (i = 0; i < m_header.ncmds; ++i) { 2141 const lldb::offset_t cmd_offset = offset; 2142 // Read in the load command and load command size 2143 struct load_command lc; 2144 if (m_data.GetU32(&offset, &lc, 2) == NULL) 2145 break; 2146 // Watch for the symbol table load command 2147 switch (lc.cmd) { 2148 case LC_SYMTAB: 2149 symtab_load_command.cmd = lc.cmd; 2150 symtab_load_command.cmdsize = lc.cmdsize; 2151 // Read in the rest of the symtab load command 2152 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2153 0) // fill in symoff, nsyms, stroff, strsize fields 2154 return 0; 2155 if (symtab_load_command.symoff == 0) { 2156 if (log) 2157 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2158 return 0; 2159 } 2160 2161 if (symtab_load_command.stroff == 0) { 2162 if (log) 2163 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2164 return 0; 2165 } 2166 2167 if (symtab_load_command.nsyms == 0) { 2168 if (log) 2169 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2170 return 0; 2171 } 2172 2173 if (symtab_load_command.strsize == 0) { 2174 if (log) 2175 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2176 return 0; 2177 } 2178 break; 2179 2180 case LC_DYLD_INFO: 2181 case LC_DYLD_INFO_ONLY: 2182 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2183 dyld_info.cmd = lc.cmd; 2184 dyld_info.cmdsize = lc.cmdsize; 2185 } else { 2186 memset(&dyld_info, 0, sizeof(dyld_info)); 2187 } 2188 break; 2189 2190 case LC_LOAD_DYLIB: 2191 case LC_LOAD_WEAK_DYLIB: 2192 case LC_REEXPORT_DYLIB: 2193 case LC_LOADFVMLIB: 2194 case LC_LOAD_UPWARD_DYLIB: { 2195 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2196 const char *path = m_data.PeekCStr(name_offset); 2197 if (path) { 2198 FileSpec file_spec(path, false); 2199 // Strip the path if there is @rpath, @executable, etc so we just use 2200 // the basename 2201 if (path[0] == '@') 2202 file_spec.GetDirectory().Clear(); 2203 2204 if (lc.cmd == LC_REEXPORT_DYLIB) { 2205 m_reexported_dylibs.AppendIfUnique(file_spec); 2206 } 2207 2208 dylib_files.Append(file_spec); 2209 } 2210 } break; 2211 2212 case LC_FUNCTION_STARTS: 2213 function_starts_load_command.cmd = lc.cmd; 2214 function_starts_load_command.cmdsize = lc.cmdsize; 2215 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2216 NULL) // fill in symoff, nsyms, stroff, strsize fields 2217 memset(&function_starts_load_command, 0, 2218 sizeof(function_starts_load_command)); 2219 break; 2220 2221 default: 2222 break; 2223 } 2224 offset = cmd_offset + lc.cmdsize; 2225 } 2226 2227 if (symtab_load_command.cmd) { 2228 Symtab *symtab = m_symtab_ap.get(); 2229 SectionList *section_list = GetSectionList(); 2230 if (section_list == NULL) 2231 return 0; 2232 2233 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2234 const ByteOrder byte_order = m_data.GetByteOrder(); 2235 bool bit_width_32 = addr_byte_size == 4; 2236 const size_t nlist_byte_size = 2237 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2238 2239 DataExtractor nlist_data(NULL, 0, byte_order, addr_byte_size); 2240 DataExtractor strtab_data(NULL, 0, byte_order, addr_byte_size); 2241 DataExtractor function_starts_data(NULL, 0, byte_order, addr_byte_size); 2242 DataExtractor indirect_symbol_index_data(NULL, 0, byte_order, 2243 addr_byte_size); 2244 DataExtractor dyld_trie_data(NULL, 0, byte_order, addr_byte_size); 2245 2246 const addr_t nlist_data_byte_size = 2247 symtab_load_command.nsyms * nlist_byte_size; 2248 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2249 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2250 2251 ProcessSP process_sp(m_process_wp.lock()); 2252 Process *process = process_sp.get(); 2253 2254 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2255 2256 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2257 Target &target = process->GetTarget(); 2258 2259 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2260 2261 SectionSP linkedit_section_sp( 2262 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2263 // Reading mach file from memory in a process or core file... 2264 2265 if (linkedit_section_sp) { 2266 addr_t linkedit_load_addr = 2267 linkedit_section_sp->GetLoadBaseAddress(&target); 2268 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2269 // We might be trying to access the symbol table before the 2270 // __LINKEDIT's load 2271 // address has been set in the target. We can't fail to read the 2272 // symbol table, 2273 // so calculate the right address manually 2274 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2275 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2276 } 2277 2278 const addr_t linkedit_file_offset = 2279 linkedit_section_sp->GetFileOffset(); 2280 const addr_t symoff_addr = linkedit_load_addr + 2281 symtab_load_command.symoff - 2282 linkedit_file_offset; 2283 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2284 linkedit_file_offset; 2285 2286 bool data_was_read = false; 2287 2288 #if defined(__APPLE__) && \ 2289 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2290 if (m_header.flags & 0x80000000u && 2291 process->GetAddressByteSize() == sizeof(void *)) { 2292 // This mach-o memory file is in the dyld shared cache. If this 2293 // program is not remote and this is iOS, then this process will 2294 // share the same shared cache as the process we are debugging and 2295 // we can read the entire __LINKEDIT from the address space in this 2296 // process. This is a needed optimization that is used for local iOS 2297 // debugging only since all shared libraries in the shared cache do 2298 // not have corresponding files that exist in the file system of the 2299 // device. They have been combined into a single file. This means we 2300 // always have to load these files from memory. All of the symbol and 2301 // string tables from all of the __LINKEDIT sections from the shared 2302 // libraries in the shared cache have been merged into a single large 2303 // symbol and string table. Reading all of this symbol and string 2304 // table 2305 // data across can slow down debug launch times, so we optimize this 2306 // by 2307 // reading the memory for the __LINKEDIT section from this process. 2308 2309 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 2310 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 2311 bool use_lldb_cache = true; 2312 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2313 lldb_shared_cache != process_shared_cache) { 2314 use_lldb_cache = false; 2315 ModuleSP module_sp(GetModule()); 2316 if (module_sp) 2317 module_sp->ReportWarning("shared cache in process does not match " 2318 "lldb's own shared cache, startup will " 2319 "be slow."); 2320 } 2321 2322 PlatformSP platform_sp(target.GetPlatform()); 2323 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2324 data_was_read = true; 2325 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2326 eByteOrderLittle); 2327 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2328 eByteOrderLittle); 2329 if (function_starts_load_command.cmd) { 2330 const addr_t func_start_addr = 2331 linkedit_load_addr + function_starts_load_command.dataoff - 2332 linkedit_file_offset; 2333 function_starts_data.SetData( 2334 (void *)func_start_addr, 2335 function_starts_load_command.datasize, eByteOrderLittle); 2336 } 2337 } 2338 } 2339 #endif 2340 2341 if (!data_was_read) { 2342 // Always load dyld - the dynamic linker - from memory if we didn't 2343 // find a binary anywhere else. 2344 // lldb will not register dylib/framework/bundle loads/unloads if we 2345 // don't have the dyld symbols, 2346 // we force dyld to load from memory despite the user's 2347 // target.memory-module-load-level setting. 2348 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2349 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2350 DataBufferSP nlist_data_sp( 2351 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2352 if (nlist_data_sp) 2353 nlist_data.SetData(nlist_data_sp, 0, 2354 nlist_data_sp->GetByteSize()); 2355 // Load strings individually from memory when loading from memory 2356 // since shared cache 2357 // string tables contain strings for all symbols from all shared 2358 // cached libraries 2359 // DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, 2360 // strtab_data_byte_size)); 2361 // if (strtab_data_sp) 2362 // strtab_data.SetData (strtab_data_sp, 0, 2363 // strtab_data_sp->GetByteSize()); 2364 if (m_dysymtab.nindirectsyms != 0) { 2365 const addr_t indirect_syms_addr = linkedit_load_addr + 2366 m_dysymtab.indirectsymoff - 2367 linkedit_file_offset; 2368 DataBufferSP indirect_syms_data_sp( 2369 ReadMemory(process_sp, indirect_syms_addr, 2370 m_dysymtab.nindirectsyms * 4)); 2371 if (indirect_syms_data_sp) 2372 indirect_symbol_index_data.SetData( 2373 indirect_syms_data_sp, 0, 2374 indirect_syms_data_sp->GetByteSize()); 2375 } 2376 } else if (memory_module_load_level >= 2377 eMemoryModuleLoadLevelPartial) { 2378 if (function_starts_load_command.cmd) { 2379 const addr_t func_start_addr = 2380 linkedit_load_addr + function_starts_load_command.dataoff - 2381 linkedit_file_offset; 2382 DataBufferSP func_start_data_sp( 2383 ReadMemory(process_sp, func_start_addr, 2384 function_starts_load_command.datasize)); 2385 if (func_start_data_sp) 2386 function_starts_data.SetData(func_start_data_sp, 0, 2387 func_start_data_sp->GetByteSize()); 2388 } 2389 } 2390 } 2391 } 2392 } else { 2393 nlist_data.SetData(m_data, symtab_load_command.symoff, 2394 nlist_data_byte_size); 2395 strtab_data.SetData(m_data, symtab_load_command.stroff, 2396 strtab_data_byte_size); 2397 2398 if (dyld_info.export_size > 0) { 2399 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2400 dyld_info.export_size); 2401 } 2402 2403 if (m_dysymtab.nindirectsyms != 0) { 2404 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2405 m_dysymtab.nindirectsyms * 4); 2406 } 2407 if (function_starts_load_command.cmd) { 2408 function_starts_data.SetData(m_data, 2409 function_starts_load_command.dataoff, 2410 function_starts_load_command.datasize); 2411 } 2412 } 2413 2414 if (nlist_data.GetByteSize() == 0 && 2415 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2416 if (log) 2417 module_sp->LogMessage(log, "failed to read nlist data"); 2418 return 0; 2419 } 2420 2421 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2422 if (!have_strtab_data) { 2423 if (process) { 2424 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2425 if (log) 2426 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2427 return 0; 2428 } 2429 } else { 2430 if (log) 2431 module_sp->LogMessage(log, "failed to read strtab data"); 2432 return 0; 2433 } 2434 } 2435 2436 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2437 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2438 const ConstString &g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2439 const ConstString &g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2440 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2441 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2442 SectionSP text_section_sp( 2443 section_list->FindSectionByName(g_segment_name_TEXT)); 2444 SectionSP data_section_sp( 2445 section_list->FindSectionByName(g_segment_name_DATA)); 2446 SectionSP data_dirty_section_sp( 2447 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2448 SectionSP data_const_section_sp( 2449 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2450 SectionSP objc_section_sp( 2451 section_list->FindSectionByName(g_segment_name_OBJC)); 2452 SectionSP eh_frame_section_sp; 2453 if (text_section_sp.get()) 2454 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2455 g_section_name_eh_frame); 2456 else 2457 eh_frame_section_sp = 2458 section_list->FindSectionByName(g_section_name_eh_frame); 2459 2460 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2461 2462 // lldb works best if it knows the start address of all functions in a 2463 // module. 2464 // Linker symbols or debug info are normally the best source of information 2465 // for start addr / size but 2466 // they may be stripped in a released binary. 2467 // Two additional sources of information exist in Mach-O binaries: 2468 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2469 // function's start address in the 2470 // binary, relative to the text section. 2471 // eh_frame - the eh_frame FDEs have the start addr & size of 2472 // each function 2473 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2474 // all modern binaries. 2475 // Binaries built to run on older releases may need to use eh_frame 2476 // information. 2477 2478 if (text_section_sp && function_starts_data.GetByteSize()) { 2479 FunctionStarts::Entry function_start_entry; 2480 function_start_entry.data = false; 2481 lldb::offset_t function_start_offset = 0; 2482 function_start_entry.addr = text_section_sp->GetFileAddress(); 2483 uint64_t delta; 2484 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2485 0) { 2486 // Now append the current entry 2487 function_start_entry.addr += delta; 2488 function_starts.Append(function_start_entry); 2489 } 2490 } else { 2491 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2492 // load command claiming an eh_frame 2493 // but it doesn't actually have the eh_frame content. And if we have a 2494 // dSYM, we don't need to do any 2495 // of this fill-in-the-missing-symbols works anyway - the debug info 2496 // should give us all the functions in 2497 // the module. 2498 if (text_section_sp.get() && eh_frame_section_sp.get() && 2499 m_type != eTypeDebugInfo) { 2500 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2501 DWARFCallFrameInfo::EH); 2502 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2503 eh_frame.GetFunctionAddressAndSizeVector(functions); 2504 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2505 size_t count = functions.GetSize(); 2506 for (size_t i = 0; i < count; ++i) { 2507 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2508 functions.GetEntryAtIndex(i); 2509 if (func) { 2510 FunctionStarts::Entry function_start_entry; 2511 function_start_entry.addr = func->base - text_base_addr; 2512 function_starts.Append(function_start_entry); 2513 } 2514 } 2515 } 2516 } 2517 2518 const size_t function_starts_count = function_starts.GetSize(); 2519 2520 // For user process binaries (executables, dylibs, frameworks, bundles), if 2521 // we don't have 2522 // LC_FUNCTION_STARTS/eh_frame section in this binary, we're going to assume 2523 // the binary 2524 // has been stripped. Don't allow assembly language instruction emulation 2525 // because we don't 2526 // know proper function start boundaries. 2527 // 2528 // For all other types of binaries (kernels, stand-alone bare board 2529 // binaries, kexts), they 2530 // may not have LC_FUNCTION_STARTS / eh_frame sections - we should not make 2531 // any assumptions 2532 // about them based on that. 2533 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2534 m_allow_assembly_emulation_unwind_plans = false; 2535 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2536 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2537 2538 if (unwind_or_symbol_log) 2539 module_sp->LogMessage( 2540 unwind_or_symbol_log, 2541 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2542 } 2543 2544 const user_id_t TEXT_eh_frame_sectID = 2545 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2546 : static_cast<user_id_t>(NO_SECT); 2547 2548 lldb::offset_t nlist_data_offset = 0; 2549 2550 uint32_t N_SO_index = UINT32_MAX; 2551 2552 MachSymtabSectionInfo section_info(section_list); 2553 std::vector<uint32_t> N_FUN_indexes; 2554 std::vector<uint32_t> N_NSYM_indexes; 2555 std::vector<uint32_t> N_INCL_indexes; 2556 std::vector<uint32_t> N_BRAC_indexes; 2557 std::vector<uint32_t> N_COMM_indexes; 2558 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2559 typedef std::map<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2560 typedef std::map<const char *, uint32_t> ConstNameToSymbolIndexMap; 2561 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2562 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2563 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2564 // Any symbols that get merged into another will get an entry 2565 // in this map so we know 2566 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2567 uint32_t nlist_idx = 0; 2568 Symbol *symbol_ptr = NULL; 2569 2570 uint32_t sym_idx = 0; 2571 Symbol *sym = NULL; 2572 size_t num_syms = 0; 2573 std::string memory_symbol_name; 2574 uint32_t unmapped_local_symbols_found = 0; 2575 2576 std::vector<TrieEntryWithOffset> trie_entries; 2577 std::set<lldb::addr_t> resolver_addresses; 2578 2579 if (dyld_trie_data.GetByteSize() > 0) { 2580 std::vector<llvm::StringRef> nameSlices; 2581 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, 2582 resolver_addresses, trie_entries); 2583 2584 ConstString text_segment_name("__TEXT"); 2585 SectionSP text_segment_sp = 2586 GetSectionList()->FindSectionByName(text_segment_name); 2587 if (text_segment_sp) { 2588 const lldb::addr_t text_segment_file_addr = 2589 text_segment_sp->GetFileAddress(); 2590 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2591 for (auto &e : trie_entries) 2592 e.entry.address += text_segment_file_addr; 2593 } 2594 } 2595 } 2596 2597 typedef std::set<ConstString> IndirectSymbols; 2598 IndirectSymbols indirect_symbol_names; 2599 2600 #if defined(__APPLE__) && \ 2601 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2602 2603 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2604 // optimized by moving LOCAL 2605 // symbols out of the memory mapped portion of the DSC. The symbol 2606 // information has all been retained, 2607 // but it isn't available in the normal nlist data. However, there *are* 2608 // duplicate entries of *some* 2609 // LOCAL symbols in the normal nlist data. To handle this situation 2610 // correctly, we must first attempt 2611 // to parse any DSC unmapped symbol information. If we find any, we set a 2612 // flag that tells the normal 2613 // nlist parser to ignore all LOCAL symbols. 2614 2615 if (m_header.flags & 0x80000000u) { 2616 // Before we can start mapping the DSC, we need to make certain the target 2617 // process is actually 2618 // using the cache we can find. 2619 2620 // Next we need to determine the correct path for the dyld shared cache. 2621 2622 ArchSpec header_arch; 2623 GetArchitecture(header_arch); 2624 char dsc_path[PATH_MAX]; 2625 char dsc_path_development[PATH_MAX]; 2626 2627 snprintf( 2628 dsc_path, sizeof(dsc_path), "%s%s%s", 2629 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2630 */ 2631 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2632 header_arch.GetArchitectureName()); 2633 2634 snprintf( 2635 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2636 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2637 */ 2638 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2639 header_arch.GetArchitectureName(), ".development"); 2640 2641 FileSpec dsc_nondevelopment_filespec(dsc_path, false); 2642 FileSpec dsc_development_filespec(dsc_path_development, false); 2643 FileSpec dsc_filespec; 2644 2645 UUID dsc_uuid; 2646 UUID process_shared_cache_uuid; 2647 2648 if (process) { 2649 process_shared_cache_uuid = GetProcessSharedCacheUUID(process); 2650 } 2651 2652 // First see if we can find an exact match for the inferior process shared 2653 // cache UUID in 2654 // the development or non-development shared caches on disk. 2655 if (process_shared_cache_uuid.IsValid()) { 2656 if (dsc_development_filespec.Exists()) { 2657 UUID dsc_development_uuid = GetSharedCacheUUID( 2658 dsc_development_filespec, byte_order, addr_byte_size); 2659 if (dsc_development_uuid.IsValid() && 2660 dsc_development_uuid == process_shared_cache_uuid) { 2661 dsc_filespec = dsc_development_filespec; 2662 dsc_uuid = dsc_development_uuid; 2663 } 2664 } 2665 if (!dsc_uuid.IsValid() && dsc_nondevelopment_filespec.Exists()) { 2666 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2667 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2668 if (dsc_nondevelopment_uuid.IsValid() && 2669 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2670 dsc_filespec = dsc_nondevelopment_filespec; 2671 dsc_uuid = dsc_nondevelopment_uuid; 2672 } 2673 } 2674 } 2675 2676 // Failing a UUID match, prefer the development dyld_shared cache if both 2677 // are present. 2678 if (!dsc_filespec.Exists()) { 2679 if (dsc_development_filespec.Exists()) { 2680 dsc_filespec = dsc_development_filespec; 2681 } else { 2682 dsc_filespec = dsc_nondevelopment_filespec; 2683 } 2684 } 2685 2686 /* The dyld_cache_header has a pointer to the 2687 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2688 The dyld_cache_local_symbols_info structure gives us three things: 2689 1. The start and count of the nlist records in the dyld_shared_cache 2690 file 2691 2. The start and size of the strings for these nlist records 2692 3. The start and count of dyld_cache_local_symbols_entry entries 2693 2694 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2695 dyld shared cache. 2696 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2697 the dyld shared cache. 2698 The dyld_cache_local_symbols_entry also lists the start of this 2699 dylib/framework's nlist records 2700 and the count of how many nlist records there are for this 2701 dylib/framework. 2702 */ 2703 2704 // Process the dyld shared cache header to find the unmapped symbols 2705 2706 DataBufferSP dsc_data_sp = MapFileData( 2707 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2708 if (!dsc_uuid.IsValid()) { 2709 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2710 } 2711 if (dsc_data_sp) { 2712 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2713 2714 bool uuid_match = true; 2715 if (dsc_uuid.IsValid() && process) { 2716 if (process_shared_cache_uuid.IsValid() && 2717 dsc_uuid != process_shared_cache_uuid) { 2718 // The on-disk dyld_shared_cache file is not the same as the one in 2719 // this 2720 // process' memory, don't use it. 2721 uuid_match = false; 2722 ModuleSP module_sp(GetModule()); 2723 if (module_sp) 2724 module_sp->ReportWarning("process shared cache does not match " 2725 "on-disk dyld_shared_cache file, some " 2726 "symbol names will be missing."); 2727 } 2728 } 2729 2730 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2731 2732 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2733 2734 // If the mappingOffset points to a location inside the header, we've 2735 // opened an old dyld shared cache, and should not proceed further. 2736 if (uuid_match && 2737 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2738 2739 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2740 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2741 mappingOffset); 2742 2743 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2744 byte_order, addr_byte_size); 2745 offset = 0; 2746 2747 // The File addresses (from the in-memory Mach-O load commands) for 2748 // the shared libraries 2749 // in the shared library cache need to be adjusted by an offset to 2750 // match up with the 2751 // dylibOffset identifying field in the 2752 // dyld_cache_local_symbol_entry's. This offset is 2753 // recorded in mapping_offset_value. 2754 const uint64_t mapping_offset_value = 2755 dsc_mapping_info_data.GetU64(&offset); 2756 2757 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, 2758 localSymbolsOffset); 2759 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2760 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2761 2762 if (localSymbolsOffset && localSymbolsSize) { 2763 // Map the local symbols 2764 DataBufferSP dsc_local_symbols_data_sp = 2765 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2766 2767 if (dsc_local_symbols_data_sp) { 2768 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2769 byte_order, addr_byte_size); 2770 2771 offset = 0; 2772 2773 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2774 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2775 UndefinedNameToDescMap undefined_name_to_desc; 2776 SymbolIndexToName reexport_shlib_needs_fixup; 2777 2778 // Read the local_symbols_infos struct in one shot 2779 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2780 dsc_local_symbols_data.GetU32(&offset, 2781 &local_symbols_info.nlistOffset, 6); 2782 2783 SectionSP text_section_sp( 2784 section_list->FindSectionByName(GetSegmentNameTEXT())); 2785 2786 uint32_t header_file_offset = 2787 (text_section_sp->GetFileAddress() - mapping_offset_value); 2788 2789 offset = local_symbols_info.entriesOffset; 2790 for (uint32_t entry_index = 0; 2791 entry_index < local_symbols_info.entriesCount; 2792 entry_index++) { 2793 struct lldb_copy_dyld_cache_local_symbols_entry 2794 local_symbols_entry; 2795 local_symbols_entry.dylibOffset = 2796 dsc_local_symbols_data.GetU32(&offset); 2797 local_symbols_entry.nlistStartIndex = 2798 dsc_local_symbols_data.GetU32(&offset); 2799 local_symbols_entry.nlistCount = 2800 dsc_local_symbols_data.GetU32(&offset); 2801 2802 if (header_file_offset == local_symbols_entry.dylibOffset) { 2803 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2804 2805 // The normal nlist code cannot correctly size the Symbols 2806 // array, we need to allocate it here. 2807 sym = symtab->Resize( 2808 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2809 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2810 num_syms = symtab->GetNumSymbols(); 2811 2812 nlist_data_offset = 2813 local_symbols_info.nlistOffset + 2814 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2815 uint32_t string_table_offset = 2816 local_symbols_info.stringsOffset; 2817 2818 for (uint32_t nlist_index = 0; 2819 nlist_index < local_symbols_entry.nlistCount; 2820 nlist_index++) { 2821 ///////////////////////////// 2822 { 2823 struct nlist_64 nlist; 2824 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize( 2825 nlist_data_offset, nlist_byte_size)) 2826 break; 2827 2828 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked( 2829 &nlist_data_offset); 2830 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked( 2831 &nlist_data_offset); 2832 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked( 2833 &nlist_data_offset); 2834 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked( 2835 &nlist_data_offset); 2836 nlist.n_value = 2837 dsc_local_symbols_data.GetAddress_unchecked( 2838 &nlist_data_offset); 2839 2840 SymbolType type = eSymbolTypeInvalid; 2841 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2842 string_table_offset + nlist.n_strx); 2843 2844 if (symbol_name == NULL) { 2845 // No symbol should be NULL, even the symbols with no 2846 // string values should have an offset zero which points 2847 // to an empty C-string 2848 Host::SystemLog( 2849 Host::eSystemLogError, 2850 "error: DSC unmapped local symbol[%u] has invalid " 2851 "string table offset 0x%x in %s, ignoring symbol\n", 2852 entry_index, nlist.n_strx, 2853 module_sp->GetFileSpec().GetPath().c_str()); 2854 continue; 2855 } 2856 if (symbol_name[0] == '\0') 2857 symbol_name = NULL; 2858 2859 const char *symbol_name_non_abi_mangled = NULL; 2860 2861 SectionSP symbol_section; 2862 uint32_t symbol_byte_size = 0; 2863 bool add_nlist = true; 2864 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2865 bool demangled_is_synthesized = false; 2866 bool is_gsym = false; 2867 bool set_value = true; 2868 2869 assert(sym_idx < num_syms); 2870 2871 sym[sym_idx].SetDebug(is_debug); 2872 2873 if (is_debug) { 2874 switch (nlist.n_type) { 2875 case N_GSYM: 2876 // global symbol: name,,NO_SECT,type,0 2877 // Sometimes the N_GSYM value contains the address. 2878 2879 // FIXME: In the .o files, we have a GSYM and a debug 2880 // symbol for all the ObjC data. They 2881 // have the same address, but we want to ensure that 2882 // we always find only the real symbol, 2883 // 'cause we don't currently correctly attribute the 2884 // GSYM one to the ObjCClass/Ivar/MetaClass 2885 // symbol type. This is a temporary hack to make sure 2886 // the ObjectiveC symbols get treated 2887 // correctly. To do this right, we should coalesce 2888 // all the GSYM & global symbols that have the 2889 // same address. 2890 2891 is_gsym = true; 2892 sym[sym_idx].SetExternal(true); 2893 2894 if (symbol_name && symbol_name[0] == '_' && 2895 symbol_name[1] == 'O') { 2896 llvm::StringRef symbol_name_ref(symbol_name); 2897 if (symbol_name_ref.startswith( 2898 g_objc_v2_prefix_class)) { 2899 symbol_name_non_abi_mangled = symbol_name + 1; 2900 symbol_name = 2901 symbol_name + g_objc_v2_prefix_class.size(); 2902 type = eSymbolTypeObjCClass; 2903 demangled_is_synthesized = true; 2904 2905 } else if (symbol_name_ref.startswith( 2906 g_objc_v2_prefix_metaclass)) { 2907 symbol_name_non_abi_mangled = symbol_name + 1; 2908 symbol_name = symbol_name + 2909 g_objc_v2_prefix_metaclass.size(); 2910 type = eSymbolTypeObjCMetaClass; 2911 demangled_is_synthesized = true; 2912 } else if (symbol_name_ref.startswith( 2913 g_objc_v2_prefix_ivar)) { 2914 symbol_name_non_abi_mangled = symbol_name + 1; 2915 symbol_name = 2916 symbol_name + g_objc_v2_prefix_ivar.size(); 2917 type = eSymbolTypeObjCIVar; 2918 demangled_is_synthesized = true; 2919 } 2920 } else { 2921 if (nlist.n_value != 0) 2922 symbol_section = section_info.GetSection( 2923 nlist.n_sect, nlist.n_value); 2924 type = eSymbolTypeData; 2925 } 2926 break; 2927 2928 case N_FNAME: 2929 // procedure name (f77 kludge): name,,NO_SECT,0,0 2930 type = eSymbolTypeCompiler; 2931 break; 2932 2933 case N_FUN: 2934 // procedure: name,,n_sect,linenumber,address 2935 if (symbol_name) { 2936 type = eSymbolTypeCode; 2937 symbol_section = section_info.GetSection( 2938 nlist.n_sect, nlist.n_value); 2939 2940 N_FUN_addr_to_sym_idx.insert( 2941 std::make_pair(nlist.n_value, sym_idx)); 2942 // We use the current number of symbols in the 2943 // symbol table in lieu of 2944 // using nlist_idx in case we ever start trimming 2945 // entries out 2946 N_FUN_indexes.push_back(sym_idx); 2947 } else { 2948 type = eSymbolTypeCompiler; 2949 2950 if (!N_FUN_indexes.empty()) { 2951 // Copy the size of the function into the original 2952 // STAB entry so we don't have 2953 // to hunt for it later 2954 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2955 ->SetByteSize(nlist.n_value); 2956 N_FUN_indexes.pop_back(); 2957 // We don't really need the end function STAB as 2958 // it contains the size which 2959 // we already placed with the original symbol, so 2960 // don't add it if we want a 2961 // minimal symbol table 2962 add_nlist = false; 2963 } 2964 } 2965 break; 2966 2967 case N_STSYM: 2968 // static symbol: name,,n_sect,type,address 2969 N_STSYM_addr_to_sym_idx.insert( 2970 std::make_pair(nlist.n_value, sym_idx)); 2971 symbol_section = section_info.GetSection( 2972 nlist.n_sect, nlist.n_value); 2973 if (symbol_name && symbol_name[0]) { 2974 type = ObjectFile::GetSymbolTypeFromName( 2975 symbol_name + 1, eSymbolTypeData); 2976 } 2977 break; 2978 2979 case N_LCSYM: 2980 // .lcomm symbol: name,,n_sect,type,address 2981 symbol_section = section_info.GetSection( 2982 nlist.n_sect, nlist.n_value); 2983 type = eSymbolTypeCommonBlock; 2984 break; 2985 2986 case N_BNSYM: 2987 // We use the current number of symbols in the symbol 2988 // table in lieu of 2989 // using nlist_idx in case we ever start trimming 2990 // entries out 2991 // Skip these if we want minimal symbol tables 2992 add_nlist = false; 2993 break; 2994 2995 case N_ENSYM: 2996 // Set the size of the N_BNSYM to the terminating 2997 // index of this N_ENSYM 2998 // so that we can always skip the entire symbol if we 2999 // need to navigate 3000 // more quickly at the source level when parsing STABS 3001 // Skip these if we want minimal symbol tables 3002 add_nlist = false; 3003 break; 3004 3005 case N_OPT: 3006 // emitted with gcc2_compiled and in gcc source 3007 type = eSymbolTypeCompiler; 3008 break; 3009 3010 case N_RSYM: 3011 // register sym: name,,NO_SECT,type,register 3012 type = eSymbolTypeVariable; 3013 break; 3014 3015 case N_SLINE: 3016 // src line: 0,,n_sect,linenumber,address 3017 symbol_section = section_info.GetSection( 3018 nlist.n_sect, nlist.n_value); 3019 type = eSymbolTypeLineEntry; 3020 break; 3021 3022 case N_SSYM: 3023 // structure elt: name,,NO_SECT,type,struct_offset 3024 type = eSymbolTypeVariableType; 3025 break; 3026 3027 case N_SO: 3028 // source file name 3029 type = eSymbolTypeSourceFile; 3030 if (symbol_name == NULL) { 3031 add_nlist = false; 3032 if (N_SO_index != UINT32_MAX) { 3033 // Set the size of the N_SO to the terminating 3034 // index of this N_SO 3035 // so that we can always skip the entire N_SO if 3036 // we need to navigate 3037 // more quickly at the source level when parsing 3038 // STABS 3039 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3040 symbol_ptr->SetByteSize(sym_idx); 3041 symbol_ptr->SetSizeIsSibling(true); 3042 } 3043 N_NSYM_indexes.clear(); 3044 N_INCL_indexes.clear(); 3045 N_BRAC_indexes.clear(); 3046 N_COMM_indexes.clear(); 3047 N_FUN_indexes.clear(); 3048 N_SO_index = UINT32_MAX; 3049 } else { 3050 // We use the current number of symbols in the 3051 // symbol table in lieu of 3052 // using nlist_idx in case we ever start trimming 3053 // entries out 3054 const bool N_SO_has_full_path = 3055 symbol_name[0] == '/'; 3056 if (N_SO_has_full_path) { 3057 if ((N_SO_index == sym_idx - 1) && 3058 ((sym_idx - 1) < num_syms)) { 3059 // We have two consecutive N_SO entries where 3060 // the first contains a directory 3061 // and the second contains a full path. 3062 sym[sym_idx - 1].GetMangled().SetValue( 3063 ConstString(symbol_name), false); 3064 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3065 add_nlist = false; 3066 } else { 3067 // This is the first entry in a N_SO that 3068 // contains a directory or 3069 // a full path to the source file 3070 N_SO_index = sym_idx; 3071 } 3072 } else if ((N_SO_index == sym_idx - 1) && 3073 ((sym_idx - 1) < num_syms)) { 3074 // This is usually the second N_SO entry that 3075 // contains just the filename, 3076 // so here we combine it with the first one if we 3077 // are minimizing the symbol table 3078 const char *so_path = 3079 sym[sym_idx - 1] 3080 .GetMangled() 3081 .GetDemangledName( 3082 lldb::eLanguageTypeUnknown) 3083 .AsCString(); 3084 if (so_path && so_path[0]) { 3085 std::string full_so_path(so_path); 3086 const size_t double_slash_pos = 3087 full_so_path.find("//"); 3088 if (double_slash_pos != std::string::npos) { 3089 // The linker has been generating bad N_SO 3090 // entries with doubled up paths 3091 // in the format "%s%s" where the first string 3092 // in the DW_AT_comp_dir, 3093 // and the second is the directory for the 3094 // source file so you end up with 3095 // a path that looks like "/tmp/src//tmp/src/" 3096 FileSpec so_dir(so_path, false); 3097 if (!so_dir.Exists()) { 3098 so_dir.SetFile( 3099 &full_so_path[double_slash_pos + 1], 3100 false); 3101 if (so_dir.Exists()) { 3102 // Trim off the incorrect path 3103 full_so_path.erase(0, 3104 double_slash_pos + 1); 3105 } 3106 } 3107 } 3108 if (*full_so_path.rbegin() != '/') 3109 full_so_path += '/'; 3110 full_so_path += symbol_name; 3111 sym[sym_idx - 1].GetMangled().SetValue( 3112 ConstString(full_so_path.c_str()), false); 3113 add_nlist = false; 3114 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3115 } 3116 } else { 3117 // This could be a relative path to a N_SO 3118 N_SO_index = sym_idx; 3119 } 3120 } 3121 break; 3122 3123 case N_OSO: 3124 // object file name: name,,0,0,st_mtime 3125 type = eSymbolTypeObjectFile; 3126 break; 3127 3128 case N_LSYM: 3129 // local sym: name,,NO_SECT,type,offset 3130 type = eSymbolTypeLocal; 3131 break; 3132 3133 //---------------------------------------------------------------------- 3134 // INCL scopes 3135 //---------------------------------------------------------------------- 3136 case N_BINCL: 3137 // include file beginning: name,,NO_SECT,0,sum 3138 // We use the current number of symbols in the symbol 3139 // table in lieu of 3140 // using nlist_idx in case we ever start trimming 3141 // entries out 3142 N_INCL_indexes.push_back(sym_idx); 3143 type = eSymbolTypeScopeBegin; 3144 break; 3145 3146 case N_EINCL: 3147 // include file end: name,,NO_SECT,0,0 3148 // Set the size of the N_BINCL to the terminating 3149 // index of this N_EINCL 3150 // so that we can always skip the entire symbol if we 3151 // need to navigate 3152 // more quickly at the source level when parsing STABS 3153 if (!N_INCL_indexes.empty()) { 3154 symbol_ptr = 3155 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3156 symbol_ptr->SetByteSize(sym_idx + 1); 3157 symbol_ptr->SetSizeIsSibling(true); 3158 N_INCL_indexes.pop_back(); 3159 } 3160 type = eSymbolTypeScopeEnd; 3161 break; 3162 3163 case N_SOL: 3164 // #included file name: name,,n_sect,0,address 3165 type = eSymbolTypeHeaderFile; 3166 3167 // We currently don't use the header files on darwin 3168 add_nlist = false; 3169 break; 3170 3171 case N_PARAMS: 3172 // compiler parameters: name,,NO_SECT,0,0 3173 type = eSymbolTypeCompiler; 3174 break; 3175 3176 case N_VERSION: 3177 // compiler version: name,,NO_SECT,0,0 3178 type = eSymbolTypeCompiler; 3179 break; 3180 3181 case N_OLEVEL: 3182 // compiler -O level: name,,NO_SECT,0,0 3183 type = eSymbolTypeCompiler; 3184 break; 3185 3186 case N_PSYM: 3187 // parameter: name,,NO_SECT,type,offset 3188 type = eSymbolTypeVariable; 3189 break; 3190 3191 case N_ENTRY: 3192 // alternate entry: name,,n_sect,linenumber,address 3193 symbol_section = section_info.GetSection( 3194 nlist.n_sect, nlist.n_value); 3195 type = eSymbolTypeLineEntry; 3196 break; 3197 3198 //---------------------------------------------------------------------- 3199 // Left and Right Braces 3200 //---------------------------------------------------------------------- 3201 case N_LBRAC: 3202 // left bracket: 0,,NO_SECT,nesting level,address 3203 // We use the current number of symbols in the symbol 3204 // table in lieu of 3205 // using nlist_idx in case we ever start trimming 3206 // entries out 3207 symbol_section = section_info.GetSection( 3208 nlist.n_sect, nlist.n_value); 3209 N_BRAC_indexes.push_back(sym_idx); 3210 type = eSymbolTypeScopeBegin; 3211 break; 3212 3213 case N_RBRAC: 3214 // right bracket: 0,,NO_SECT,nesting level,address 3215 // Set the size of the N_LBRAC to the terminating 3216 // index of this N_RBRAC 3217 // so that we can always skip the entire symbol if we 3218 // need to navigate 3219 // more quickly at the source level when parsing STABS 3220 symbol_section = section_info.GetSection( 3221 nlist.n_sect, nlist.n_value); 3222 if (!N_BRAC_indexes.empty()) { 3223 symbol_ptr = 3224 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3225 symbol_ptr->SetByteSize(sym_idx + 1); 3226 symbol_ptr->SetSizeIsSibling(true); 3227 N_BRAC_indexes.pop_back(); 3228 } 3229 type = eSymbolTypeScopeEnd; 3230 break; 3231 3232 case N_EXCL: 3233 // deleted include file: name,,NO_SECT,0,sum 3234 type = eSymbolTypeHeaderFile; 3235 break; 3236 3237 //---------------------------------------------------------------------- 3238 // COMM scopes 3239 //---------------------------------------------------------------------- 3240 case N_BCOMM: 3241 // begin common: name,,NO_SECT,0,0 3242 // We use the current number of symbols in the symbol 3243 // table in lieu of 3244 // using nlist_idx in case we ever start trimming 3245 // entries out 3246 type = eSymbolTypeScopeBegin; 3247 N_COMM_indexes.push_back(sym_idx); 3248 break; 3249 3250 case N_ECOML: 3251 // end common (local name): 0,,n_sect,0,address 3252 symbol_section = section_info.GetSection( 3253 nlist.n_sect, nlist.n_value); 3254 // Fall through 3255 3256 case N_ECOMM: 3257 // end common: name,,n_sect,0,0 3258 // Set the size of the N_BCOMM to the terminating 3259 // index of this N_ECOMM/N_ECOML 3260 // so that we can always skip the entire symbol if we 3261 // need to navigate 3262 // more quickly at the source level when parsing STABS 3263 if (!N_COMM_indexes.empty()) { 3264 symbol_ptr = 3265 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3266 symbol_ptr->SetByteSize(sym_idx + 1); 3267 symbol_ptr->SetSizeIsSibling(true); 3268 N_COMM_indexes.pop_back(); 3269 } 3270 type = eSymbolTypeScopeEnd; 3271 break; 3272 3273 case N_LENG: 3274 // second stab entry with length information 3275 type = eSymbolTypeAdditional; 3276 break; 3277 3278 default: 3279 break; 3280 } 3281 } else { 3282 // uint8_t n_pext = N_PEXT & nlist.n_type; 3283 uint8_t n_type = N_TYPE & nlist.n_type; 3284 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3285 3286 switch (n_type) { 3287 case N_INDR: { 3288 const char *reexport_name_cstr = 3289 strtab_data.PeekCStr(nlist.n_value); 3290 if (reexport_name_cstr && reexport_name_cstr[0]) { 3291 type = eSymbolTypeReExported; 3292 ConstString reexport_name( 3293 reexport_name_cstr + 3294 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3295 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3296 set_value = false; 3297 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3298 indirect_symbol_names.insert( 3299 ConstString(symbol_name + 3300 ((symbol_name[0] == '_') ? 1 : 0))); 3301 } else 3302 type = eSymbolTypeUndefined; 3303 } break; 3304 3305 case N_UNDF: 3306 if (symbol_name && symbol_name[0]) { 3307 ConstString undefined_name( 3308 symbol_name + 3309 ((symbol_name[0] == '_') ? 1 : 0)); 3310 undefined_name_to_desc[undefined_name] = 3311 nlist.n_desc; 3312 } 3313 // Fall through 3314 case N_PBUD: 3315 type = eSymbolTypeUndefined; 3316 break; 3317 3318 case N_ABS: 3319 type = eSymbolTypeAbsolute; 3320 break; 3321 3322 case N_SECT: { 3323 symbol_section = section_info.GetSection( 3324 nlist.n_sect, nlist.n_value); 3325 3326 if (symbol_section == NULL) { 3327 // TODO: warn about this? 3328 add_nlist = false; 3329 break; 3330 } 3331 3332 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3333 type = eSymbolTypeException; 3334 } else { 3335 uint32_t section_type = 3336 symbol_section->Get() & SECTION_TYPE; 3337 3338 switch (section_type) { 3339 case S_CSTRING_LITERALS: 3340 type = eSymbolTypeData; 3341 break; // section with only literal C strings 3342 case S_4BYTE_LITERALS: 3343 type = eSymbolTypeData; 3344 break; // section with only 4 byte literals 3345 case S_8BYTE_LITERALS: 3346 type = eSymbolTypeData; 3347 break; // section with only 8 byte literals 3348 case S_LITERAL_POINTERS: 3349 type = eSymbolTypeTrampoline; 3350 break; // section with only pointers to literals 3351 case S_NON_LAZY_SYMBOL_POINTERS: 3352 type = eSymbolTypeTrampoline; 3353 break; // section with only non-lazy symbol 3354 // pointers 3355 case S_LAZY_SYMBOL_POINTERS: 3356 type = eSymbolTypeTrampoline; 3357 break; // section with only lazy symbol pointers 3358 case S_SYMBOL_STUBS: 3359 type = eSymbolTypeTrampoline; 3360 break; // section with only symbol stubs, byte 3361 // size of stub in the reserved2 field 3362 case S_MOD_INIT_FUNC_POINTERS: 3363 type = eSymbolTypeCode; 3364 break; // section with only function pointers for 3365 // initialization 3366 case S_MOD_TERM_FUNC_POINTERS: 3367 type = eSymbolTypeCode; 3368 break; // section with only function pointers for 3369 // termination 3370 case S_INTERPOSING: 3371 type = eSymbolTypeTrampoline; 3372 break; // section with only pairs of function 3373 // pointers for interposing 3374 case S_16BYTE_LITERALS: 3375 type = eSymbolTypeData; 3376 break; // section with only 16 byte literals 3377 case S_DTRACE_DOF: 3378 type = eSymbolTypeInstrumentation; 3379 break; 3380 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3381 type = eSymbolTypeTrampoline; 3382 break; 3383 default: 3384 switch (symbol_section->GetType()) { 3385 case lldb::eSectionTypeCode: 3386 type = eSymbolTypeCode; 3387 break; 3388 case eSectionTypeData: 3389 case eSectionTypeDataCString: // Inlined C string 3390 // data 3391 case eSectionTypeDataCStringPointers: // Pointers 3392 // to C 3393 // string 3394 // data 3395 case eSectionTypeDataSymbolAddress: // Address of 3396 // a symbol in 3397 // the symbol 3398 // table 3399 case eSectionTypeData4: 3400 case eSectionTypeData8: 3401 case eSectionTypeData16: 3402 type = eSymbolTypeData; 3403 break; 3404 default: 3405 break; 3406 } 3407 break; 3408 } 3409 3410 if (type == eSymbolTypeInvalid) { 3411 const char *symbol_sect_name = 3412 symbol_section->GetName().AsCString(); 3413 if (symbol_section->IsDescendant( 3414 text_section_sp.get())) { 3415 if (symbol_section->IsClear( 3416 S_ATTR_PURE_INSTRUCTIONS | 3417 S_ATTR_SELF_MODIFYING_CODE | 3418 S_ATTR_SOME_INSTRUCTIONS)) 3419 type = eSymbolTypeData; 3420 else 3421 type = eSymbolTypeCode; 3422 } else if (symbol_section->IsDescendant( 3423 data_section_sp.get()) || 3424 symbol_section->IsDescendant( 3425 data_dirty_section_sp.get()) || 3426 symbol_section->IsDescendant( 3427 data_const_section_sp.get())) { 3428 if (symbol_sect_name && 3429 ::strstr(symbol_sect_name, "__objc") == 3430 symbol_sect_name) { 3431 type = eSymbolTypeRuntime; 3432 3433 if (symbol_name) { 3434 llvm::StringRef symbol_name_ref( 3435 symbol_name); 3436 if (symbol_name_ref.startswith("_OBJC_")) { 3437 static const llvm::StringRef 3438 g_objc_v2_prefix_class( 3439 "_OBJC_CLASS_$_"); 3440 static const llvm::StringRef 3441 g_objc_v2_prefix_metaclass( 3442 "_OBJC_METACLASS_$_"); 3443 static const llvm::StringRef 3444 g_objc_v2_prefix_ivar( 3445 "_OBJC_IVAR_$_"); 3446 if (symbol_name_ref.startswith( 3447 g_objc_v2_prefix_class)) { 3448 symbol_name_non_abi_mangled = 3449 symbol_name + 1; 3450 symbol_name = 3451 symbol_name + 3452 g_objc_v2_prefix_class.size(); 3453 type = eSymbolTypeObjCClass; 3454 demangled_is_synthesized = true; 3455 } else if ( 3456 symbol_name_ref.startswith( 3457 g_objc_v2_prefix_metaclass)) { 3458 symbol_name_non_abi_mangled = 3459 symbol_name + 1; 3460 symbol_name = 3461 symbol_name + 3462 g_objc_v2_prefix_metaclass.size(); 3463 type = eSymbolTypeObjCMetaClass; 3464 demangled_is_synthesized = true; 3465 } else if (symbol_name_ref.startswith( 3466 g_objc_v2_prefix_ivar)) { 3467 symbol_name_non_abi_mangled = 3468 symbol_name + 1; 3469 symbol_name = 3470 symbol_name + 3471 g_objc_v2_prefix_ivar.size(); 3472 type = eSymbolTypeObjCIVar; 3473 demangled_is_synthesized = true; 3474 } 3475 } 3476 } 3477 } else if (symbol_sect_name && 3478 ::strstr(symbol_sect_name, 3479 "__gcc_except_tab") == 3480 symbol_sect_name) { 3481 type = eSymbolTypeException; 3482 } else { 3483 type = eSymbolTypeData; 3484 } 3485 } else if (symbol_sect_name && 3486 ::strstr(symbol_sect_name, 3487 "__IMPORT") == 3488 symbol_sect_name) { 3489 type = eSymbolTypeTrampoline; 3490 } else if (symbol_section->IsDescendant( 3491 objc_section_sp.get())) { 3492 type = eSymbolTypeRuntime; 3493 if (symbol_name && symbol_name[0] == '.') { 3494 llvm::StringRef symbol_name_ref(symbol_name); 3495 static const llvm::StringRef 3496 g_objc_v1_prefix_class( 3497 ".objc_class_name_"); 3498 if (symbol_name_ref.startswith( 3499 g_objc_v1_prefix_class)) { 3500 symbol_name_non_abi_mangled = symbol_name; 3501 symbol_name = symbol_name + 3502 g_objc_v1_prefix_class.size(); 3503 type = eSymbolTypeObjCClass; 3504 demangled_is_synthesized = true; 3505 } 3506 } 3507 } 3508 } 3509 } 3510 } break; 3511 } 3512 } 3513 3514 if (add_nlist) { 3515 uint64_t symbol_value = nlist.n_value; 3516 if (symbol_name_non_abi_mangled) { 3517 sym[sym_idx].GetMangled().SetMangledName( 3518 ConstString(symbol_name_non_abi_mangled)); 3519 sym[sym_idx].GetMangled().SetDemangledName( 3520 ConstString(symbol_name)); 3521 } else { 3522 bool symbol_name_is_mangled = false; 3523 3524 if (symbol_name && symbol_name[0] == '_') { 3525 symbol_name_is_mangled = symbol_name[1] == '_'; 3526 symbol_name++; // Skip the leading underscore 3527 } 3528 3529 if (symbol_name) { 3530 ConstString const_symbol_name(symbol_name); 3531 sym[sym_idx].GetMangled().SetValue( 3532 const_symbol_name, symbol_name_is_mangled); 3533 if (is_gsym && is_debug) { 3534 const char *gsym_name = 3535 sym[sym_idx] 3536 .GetMangled() 3537 .GetName(lldb::eLanguageTypeUnknown, 3538 Mangled::ePreferMangled) 3539 .GetCString(); 3540 if (gsym_name) 3541 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3542 } 3543 } 3544 } 3545 if (symbol_section) { 3546 const addr_t section_file_addr = 3547 symbol_section->GetFileAddress(); 3548 if (symbol_byte_size == 0 && 3549 function_starts_count > 0) { 3550 addr_t symbol_lookup_file_addr = nlist.n_value; 3551 // Do an exact address match for non-ARM addresses, 3552 // else get the closest since 3553 // the symbol might be a thumb symbol which has an 3554 // address with bit zero set 3555 FunctionStarts::Entry *func_start_entry = 3556 function_starts.FindEntry( 3557 symbol_lookup_file_addr, !is_arm); 3558 if (is_arm && func_start_entry) { 3559 // Verify that the function start address is the 3560 // symbol address (ARM) 3561 // or the symbol address + 1 (thumb) 3562 if (func_start_entry->addr != 3563 symbol_lookup_file_addr && 3564 func_start_entry->addr != 3565 (symbol_lookup_file_addr + 1)) { 3566 // Not the right entry, NULL it out... 3567 func_start_entry = NULL; 3568 } 3569 } 3570 if (func_start_entry) { 3571 func_start_entry->data = true; 3572 3573 addr_t symbol_file_addr = func_start_entry->addr; 3574 uint32_t symbol_flags = 0; 3575 if (is_arm) { 3576 if (symbol_file_addr & 1) 3577 symbol_flags = 3578 MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3579 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3580 } 3581 3582 const FunctionStarts::Entry 3583 *next_func_start_entry = 3584 function_starts.FindNextEntry( 3585 func_start_entry); 3586 const addr_t section_end_file_addr = 3587 section_file_addr + 3588 symbol_section->GetByteSize(); 3589 if (next_func_start_entry) { 3590 addr_t next_symbol_file_addr = 3591 next_func_start_entry->addr; 3592 // Be sure the clear the Thumb address bit when 3593 // we calculate the size 3594 // from the current and next address 3595 if (is_arm) 3596 next_symbol_file_addr &= 3597 THUMB_ADDRESS_BIT_MASK; 3598 symbol_byte_size = std::min<lldb::addr_t>( 3599 next_symbol_file_addr - symbol_file_addr, 3600 section_end_file_addr - symbol_file_addr); 3601 } else { 3602 symbol_byte_size = 3603 section_end_file_addr - symbol_file_addr; 3604 } 3605 } 3606 } 3607 symbol_value -= section_file_addr; 3608 } 3609 3610 if (is_debug == false) { 3611 if (type == eSymbolTypeCode) { 3612 // See if we can find a N_FUN entry for any code 3613 // symbols. 3614 // If we do find a match, and the name matches, then 3615 // we 3616 // can merge the two into just the function symbol 3617 // to avoid 3618 // duplicate entries in the symbol table 3619 std::pair<ValueToSymbolIndexMap::const_iterator, 3620 ValueToSymbolIndexMap::const_iterator> 3621 range; 3622 range = N_FUN_addr_to_sym_idx.equal_range( 3623 nlist.n_value); 3624 if (range.first != range.second) { 3625 bool found_it = false; 3626 for (ValueToSymbolIndexMap::const_iterator pos = 3627 range.first; 3628 pos != range.second; ++pos) { 3629 if (sym[sym_idx].GetMangled().GetName( 3630 lldb::eLanguageTypeUnknown, 3631 Mangled::ePreferMangled) == 3632 sym[pos->second].GetMangled().GetName( 3633 lldb::eLanguageTypeUnknown, 3634 Mangled::ePreferMangled)) { 3635 m_nlist_idx_to_sym_idx[nlist_idx] = 3636 pos->second; 3637 // We just need the flags from the linker 3638 // symbol, so put these flags 3639 // into the N_FUN flags to avoid duplicate 3640 // symbols in the symbol table 3641 sym[pos->second].SetExternal( 3642 sym[sym_idx].IsExternal()); 3643 sym[pos->second].SetFlags(nlist.n_type << 16 | 3644 nlist.n_desc); 3645 if (resolver_addresses.find(nlist.n_value) != 3646 resolver_addresses.end()) 3647 sym[pos->second].SetType( 3648 eSymbolTypeResolver); 3649 sym[sym_idx].Clear(); 3650 found_it = true; 3651 break; 3652 } 3653 } 3654 if (found_it) 3655 continue; 3656 } else { 3657 if (resolver_addresses.find(nlist.n_value) != 3658 resolver_addresses.end()) 3659 type = eSymbolTypeResolver; 3660 } 3661 } else if (type == eSymbolTypeData || 3662 type == eSymbolTypeObjCClass || 3663 type == eSymbolTypeObjCMetaClass || 3664 type == eSymbolTypeObjCIVar) { 3665 // See if we can find a N_STSYM entry for any data 3666 // symbols. 3667 // If we do find a match, and the name matches, then 3668 // we 3669 // can merge the two into just the Static symbol to 3670 // avoid 3671 // duplicate entries in the symbol table 3672 std::pair<ValueToSymbolIndexMap::const_iterator, 3673 ValueToSymbolIndexMap::const_iterator> 3674 range; 3675 range = N_STSYM_addr_to_sym_idx.equal_range( 3676 nlist.n_value); 3677 if (range.first != range.second) { 3678 bool found_it = false; 3679 for (ValueToSymbolIndexMap::const_iterator pos = 3680 range.first; 3681 pos != range.second; ++pos) { 3682 if (sym[sym_idx].GetMangled().GetName( 3683 lldb::eLanguageTypeUnknown, 3684 Mangled::ePreferMangled) == 3685 sym[pos->second].GetMangled().GetName( 3686 lldb::eLanguageTypeUnknown, 3687 Mangled::ePreferMangled)) { 3688 m_nlist_idx_to_sym_idx[nlist_idx] = 3689 pos->second; 3690 // We just need the flags from the linker 3691 // symbol, so put these flags 3692 // into the N_STSYM flags to avoid duplicate 3693 // symbols in the symbol table 3694 sym[pos->second].SetExternal( 3695 sym[sym_idx].IsExternal()); 3696 sym[pos->second].SetFlags(nlist.n_type << 16 | 3697 nlist.n_desc); 3698 sym[sym_idx].Clear(); 3699 found_it = true; 3700 break; 3701 } 3702 } 3703 if (found_it) 3704 continue; 3705 } else { 3706 const char *gsym_name = 3707 sym[sym_idx] 3708 .GetMangled() 3709 .GetName(lldb::eLanguageTypeUnknown, 3710 Mangled::ePreferMangled) 3711 .GetCString(); 3712 if (gsym_name) { 3713 // Combine N_GSYM stab entries with the non stab 3714 // symbol 3715 ConstNameToSymbolIndexMap::const_iterator pos = 3716 N_GSYM_name_to_sym_idx.find(gsym_name); 3717 if (pos != N_GSYM_name_to_sym_idx.end()) { 3718 const uint32_t GSYM_sym_idx = pos->second; 3719 m_nlist_idx_to_sym_idx[nlist_idx] = 3720 GSYM_sym_idx; 3721 // Copy the address, because often the N_GSYM 3722 // address has an invalid address of zero 3723 // when the global is a common symbol 3724 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3725 symbol_section); 3726 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3727 symbol_value); 3728 // We just need the flags from the linker 3729 // symbol, so put these flags 3730 // into the N_GSYM flags to avoid duplicate 3731 // symbols in the symbol table 3732 sym[GSYM_sym_idx].SetFlags( 3733 nlist.n_type << 16 | nlist.n_desc); 3734 sym[sym_idx].Clear(); 3735 continue; 3736 } 3737 } 3738 } 3739 } 3740 } 3741 3742 sym[sym_idx].SetID(nlist_idx); 3743 sym[sym_idx].SetType(type); 3744 if (set_value) { 3745 sym[sym_idx].GetAddressRef().SetSection( 3746 symbol_section); 3747 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3748 } 3749 sym[sym_idx].SetFlags(nlist.n_type << 16 | 3750 nlist.n_desc); 3751 3752 if (symbol_byte_size > 0) 3753 sym[sym_idx].SetByteSize(symbol_byte_size); 3754 3755 if (demangled_is_synthesized) 3756 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3757 ++sym_idx; 3758 } else { 3759 sym[sym_idx].Clear(); 3760 } 3761 } 3762 ///////////////////////////// 3763 } 3764 break; // No more entries to consider 3765 } 3766 } 3767 3768 for (const auto &pos : reexport_shlib_needs_fixup) { 3769 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3770 if (undef_pos != undefined_name_to_desc.end()) { 3771 const uint8_t dylib_ordinal = 3772 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3773 if (dylib_ordinal > 0 && 3774 dylib_ordinal < dylib_files.GetSize()) 3775 sym[pos.first].SetReExportedSymbolSharedLibrary( 3776 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3777 } 3778 } 3779 } 3780 } 3781 } 3782 } 3783 } 3784 3785 // Must reset this in case it was mutated above! 3786 nlist_data_offset = 0; 3787 #endif 3788 3789 if (nlist_data.GetByteSize() > 0) { 3790 3791 // If the sym array was not created while parsing the DSC unmapped 3792 // symbols, create it now. 3793 if (sym == NULL) { 3794 sym = symtab->Resize(symtab_load_command.nsyms + 3795 m_dysymtab.nindirectsyms); 3796 num_syms = symtab->GetNumSymbols(); 3797 } 3798 3799 if (unmapped_local_symbols_found) { 3800 assert(m_dysymtab.ilocalsym == 0); 3801 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3802 nlist_idx = m_dysymtab.nlocalsym; 3803 } else { 3804 nlist_idx = 0; 3805 } 3806 3807 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3808 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3809 UndefinedNameToDescMap undefined_name_to_desc; 3810 SymbolIndexToName reexport_shlib_needs_fixup; 3811 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 3812 struct nlist_64 nlist; 3813 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, 3814 nlist_byte_size)) 3815 break; 3816 3817 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3818 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 3819 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 3820 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 3821 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 3822 3823 SymbolType type = eSymbolTypeInvalid; 3824 const char *symbol_name = NULL; 3825 3826 if (have_strtab_data) { 3827 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3828 3829 if (symbol_name == NULL) { 3830 // No symbol should be NULL, even the symbols with no 3831 // string values should have an offset zero which points 3832 // to an empty C-string 3833 Host::SystemLog(Host::eSystemLogError, 3834 "error: symbol[%u] has invalid string table offset " 3835 "0x%x in %s, ignoring symbol\n", 3836 nlist_idx, nlist.n_strx, 3837 module_sp->GetFileSpec().GetPath().c_str()); 3838 continue; 3839 } 3840 if (symbol_name[0] == '\0') 3841 symbol_name = NULL; 3842 } else { 3843 const addr_t str_addr = strtab_addr + nlist.n_strx; 3844 Status str_error; 3845 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3846 str_error)) 3847 symbol_name = memory_symbol_name.c_str(); 3848 } 3849 const char *symbol_name_non_abi_mangled = NULL; 3850 3851 SectionSP symbol_section; 3852 lldb::addr_t symbol_byte_size = 0; 3853 bool add_nlist = true; 3854 bool is_gsym = false; 3855 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3856 bool demangled_is_synthesized = false; 3857 bool set_value = true; 3858 assert(sym_idx < num_syms); 3859 3860 sym[sym_idx].SetDebug(is_debug); 3861 3862 if (is_debug) { 3863 switch (nlist.n_type) { 3864 case N_GSYM: 3865 // global symbol: name,,NO_SECT,type,0 3866 // Sometimes the N_GSYM value contains the address. 3867 3868 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3869 // the ObjC data. They 3870 // have the same address, but we want to ensure that we always find 3871 // only the real symbol, 3872 // 'cause we don't currently correctly attribute the GSYM one to the 3873 // ObjCClass/Ivar/MetaClass 3874 // symbol type. This is a temporary hack to make sure the 3875 // ObjectiveC symbols get treated 3876 // correctly. To do this right, we should coalesce all the GSYM & 3877 // global symbols that have the 3878 // same address. 3879 is_gsym = true; 3880 sym[sym_idx].SetExternal(true); 3881 3882 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3883 llvm::StringRef symbol_name_ref(symbol_name); 3884 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3885 symbol_name_non_abi_mangled = symbol_name + 1; 3886 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3887 type = eSymbolTypeObjCClass; 3888 demangled_is_synthesized = true; 3889 3890 } else if (symbol_name_ref.startswith( 3891 g_objc_v2_prefix_metaclass)) { 3892 symbol_name_non_abi_mangled = symbol_name + 1; 3893 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3894 type = eSymbolTypeObjCMetaClass; 3895 demangled_is_synthesized = true; 3896 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3897 symbol_name_non_abi_mangled = symbol_name + 1; 3898 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3899 type = eSymbolTypeObjCIVar; 3900 demangled_is_synthesized = true; 3901 } 3902 } else { 3903 if (nlist.n_value != 0) 3904 symbol_section = 3905 section_info.GetSection(nlist.n_sect, nlist.n_value); 3906 type = eSymbolTypeData; 3907 } 3908 break; 3909 3910 case N_FNAME: 3911 // procedure name (f77 kludge): name,,NO_SECT,0,0 3912 type = eSymbolTypeCompiler; 3913 break; 3914 3915 case N_FUN: 3916 // procedure: name,,n_sect,linenumber,address 3917 if (symbol_name) { 3918 type = eSymbolTypeCode; 3919 symbol_section = 3920 section_info.GetSection(nlist.n_sect, nlist.n_value); 3921 3922 N_FUN_addr_to_sym_idx.insert( 3923 std::make_pair(nlist.n_value, sym_idx)); 3924 // We use the current number of symbols in the symbol table in 3925 // lieu of 3926 // using nlist_idx in case we ever start trimming entries out 3927 N_FUN_indexes.push_back(sym_idx); 3928 } else { 3929 type = eSymbolTypeCompiler; 3930 3931 if (!N_FUN_indexes.empty()) { 3932 // Copy the size of the function into the original STAB entry so 3933 // we don't have 3934 // to hunt for it later 3935 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3936 ->SetByteSize(nlist.n_value); 3937 N_FUN_indexes.pop_back(); 3938 // We don't really need the end function STAB as it contains the 3939 // size which 3940 // we already placed with the original symbol, so don't add it 3941 // if we want a 3942 // minimal symbol table 3943 add_nlist = false; 3944 } 3945 } 3946 break; 3947 3948 case N_STSYM: 3949 // static symbol: name,,n_sect,type,address 3950 N_STSYM_addr_to_sym_idx.insert( 3951 std::make_pair(nlist.n_value, sym_idx)); 3952 symbol_section = 3953 section_info.GetSection(nlist.n_sect, nlist.n_value); 3954 if (symbol_name && symbol_name[0]) { 3955 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3956 eSymbolTypeData); 3957 } 3958 break; 3959 3960 case N_LCSYM: 3961 // .lcomm symbol: name,,n_sect,type,address 3962 symbol_section = 3963 section_info.GetSection(nlist.n_sect, nlist.n_value); 3964 type = eSymbolTypeCommonBlock; 3965 break; 3966 3967 case N_BNSYM: 3968 // We use the current number of symbols in the symbol table in lieu 3969 // of 3970 // using nlist_idx in case we ever start trimming entries out 3971 // Skip these if we want minimal symbol tables 3972 add_nlist = false; 3973 break; 3974 3975 case N_ENSYM: 3976 // Set the size of the N_BNSYM to the terminating index of this 3977 // N_ENSYM 3978 // so that we can always skip the entire symbol if we need to 3979 // navigate 3980 // more quickly at the source level when parsing STABS 3981 // Skip these if we want minimal symbol tables 3982 add_nlist = false; 3983 break; 3984 3985 case N_OPT: 3986 // emitted with gcc2_compiled and in gcc source 3987 type = eSymbolTypeCompiler; 3988 break; 3989 3990 case N_RSYM: 3991 // register sym: name,,NO_SECT,type,register 3992 type = eSymbolTypeVariable; 3993 break; 3994 3995 case N_SLINE: 3996 // src line: 0,,n_sect,linenumber,address 3997 symbol_section = 3998 section_info.GetSection(nlist.n_sect, nlist.n_value); 3999 type = eSymbolTypeLineEntry; 4000 break; 4001 4002 case N_SSYM: 4003 // structure elt: name,,NO_SECT,type,struct_offset 4004 type = eSymbolTypeVariableType; 4005 break; 4006 4007 case N_SO: 4008 // source file name 4009 type = eSymbolTypeSourceFile; 4010 if (symbol_name == NULL) { 4011 add_nlist = false; 4012 if (N_SO_index != UINT32_MAX) { 4013 // Set the size of the N_SO to the terminating index of this 4014 // N_SO 4015 // so that we can always skip the entire N_SO if we need to 4016 // navigate 4017 // more quickly at the source level when parsing STABS 4018 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 4019 symbol_ptr->SetByteSize(sym_idx); 4020 symbol_ptr->SetSizeIsSibling(true); 4021 } 4022 N_NSYM_indexes.clear(); 4023 N_INCL_indexes.clear(); 4024 N_BRAC_indexes.clear(); 4025 N_COMM_indexes.clear(); 4026 N_FUN_indexes.clear(); 4027 N_SO_index = UINT32_MAX; 4028 } else { 4029 // We use the current number of symbols in the symbol table in 4030 // lieu of 4031 // using nlist_idx in case we ever start trimming entries out 4032 const bool N_SO_has_full_path = symbol_name[0] == '/'; 4033 if (N_SO_has_full_path) { 4034 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 4035 // We have two consecutive N_SO entries where the first 4036 // contains a directory 4037 // and the second contains a full path. 4038 sym[sym_idx - 1].GetMangled().SetValue( 4039 ConstString(symbol_name), false); 4040 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4041 add_nlist = false; 4042 } else { 4043 // This is the first entry in a N_SO that contains a directory 4044 // or 4045 // a full path to the source file 4046 N_SO_index = sym_idx; 4047 } 4048 } else if ((N_SO_index == sym_idx - 1) && 4049 ((sym_idx - 1) < num_syms)) { 4050 // This is usually the second N_SO entry that contains just the 4051 // filename, 4052 // so here we combine it with the first one if we are minimizing 4053 // the symbol table 4054 const char *so_path = 4055 sym[sym_idx - 1] 4056 .GetMangled() 4057 .GetDemangledName(lldb::eLanguageTypeUnknown) 4058 .AsCString(); 4059 if (so_path && so_path[0]) { 4060 std::string full_so_path(so_path); 4061 const size_t double_slash_pos = full_so_path.find("//"); 4062 if (double_slash_pos != std::string::npos) { 4063 // The linker has been generating bad N_SO entries with 4064 // doubled up paths 4065 // in the format "%s%s" where the first string in the 4066 // DW_AT_comp_dir, 4067 // and the second is the directory for the source file so 4068 // you end up with 4069 // a path that looks like "/tmp/src//tmp/src/" 4070 FileSpec so_dir(so_path, false); 4071 if (!so_dir.Exists()) { 4072 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 4073 false); 4074 if (so_dir.Exists()) { 4075 // Trim off the incorrect path 4076 full_so_path.erase(0, double_slash_pos + 1); 4077 } 4078 } 4079 } 4080 if (*full_so_path.rbegin() != '/') 4081 full_so_path += '/'; 4082 full_so_path += symbol_name; 4083 sym[sym_idx - 1].GetMangled().SetValue( 4084 ConstString(full_so_path.c_str()), false); 4085 add_nlist = false; 4086 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4087 } 4088 } else { 4089 // This could be a relative path to a N_SO 4090 N_SO_index = sym_idx; 4091 } 4092 } 4093 break; 4094 4095 case N_OSO: 4096 // object file name: name,,0,0,st_mtime 4097 type = eSymbolTypeObjectFile; 4098 break; 4099 4100 case N_LSYM: 4101 // local sym: name,,NO_SECT,type,offset 4102 type = eSymbolTypeLocal; 4103 break; 4104 4105 //---------------------------------------------------------------------- 4106 // INCL scopes 4107 //---------------------------------------------------------------------- 4108 case N_BINCL: 4109 // include file beginning: name,,NO_SECT,0,sum 4110 // We use the current number of symbols in the symbol table in lieu 4111 // of 4112 // using nlist_idx in case we ever start trimming entries out 4113 N_INCL_indexes.push_back(sym_idx); 4114 type = eSymbolTypeScopeBegin; 4115 break; 4116 4117 case N_EINCL: 4118 // include file end: name,,NO_SECT,0,0 4119 // Set the size of the N_BINCL to the terminating index of this 4120 // N_EINCL 4121 // so that we can always skip the entire symbol if we need to 4122 // navigate 4123 // more quickly at the source level when parsing STABS 4124 if (!N_INCL_indexes.empty()) { 4125 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4126 symbol_ptr->SetByteSize(sym_idx + 1); 4127 symbol_ptr->SetSizeIsSibling(true); 4128 N_INCL_indexes.pop_back(); 4129 } 4130 type = eSymbolTypeScopeEnd; 4131 break; 4132 4133 case N_SOL: 4134 // #included file name: name,,n_sect,0,address 4135 type = eSymbolTypeHeaderFile; 4136 4137 // We currently don't use the header files on darwin 4138 add_nlist = false; 4139 break; 4140 4141 case N_PARAMS: 4142 // compiler parameters: name,,NO_SECT,0,0 4143 type = eSymbolTypeCompiler; 4144 break; 4145 4146 case N_VERSION: 4147 // compiler version: name,,NO_SECT,0,0 4148 type = eSymbolTypeCompiler; 4149 break; 4150 4151 case N_OLEVEL: 4152 // compiler -O level: name,,NO_SECT,0,0 4153 type = eSymbolTypeCompiler; 4154 break; 4155 4156 case N_PSYM: 4157 // parameter: name,,NO_SECT,type,offset 4158 type = eSymbolTypeVariable; 4159 break; 4160 4161 case N_ENTRY: 4162 // alternate entry: name,,n_sect,linenumber,address 4163 symbol_section = 4164 section_info.GetSection(nlist.n_sect, nlist.n_value); 4165 type = eSymbolTypeLineEntry; 4166 break; 4167 4168 //---------------------------------------------------------------------- 4169 // Left and Right Braces 4170 //---------------------------------------------------------------------- 4171 case N_LBRAC: 4172 // left bracket: 0,,NO_SECT,nesting level,address 4173 // We use the current number of symbols in the symbol table in lieu 4174 // of 4175 // using nlist_idx in case we ever start trimming entries out 4176 symbol_section = 4177 section_info.GetSection(nlist.n_sect, nlist.n_value); 4178 N_BRAC_indexes.push_back(sym_idx); 4179 type = eSymbolTypeScopeBegin; 4180 break; 4181 4182 case N_RBRAC: 4183 // right bracket: 0,,NO_SECT,nesting level,address 4184 // Set the size of the N_LBRAC to the terminating index of this 4185 // N_RBRAC 4186 // so that we can always skip the entire symbol if we need to 4187 // navigate 4188 // more quickly at the source level when parsing STABS 4189 symbol_section = 4190 section_info.GetSection(nlist.n_sect, nlist.n_value); 4191 if (!N_BRAC_indexes.empty()) { 4192 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4193 symbol_ptr->SetByteSize(sym_idx + 1); 4194 symbol_ptr->SetSizeIsSibling(true); 4195 N_BRAC_indexes.pop_back(); 4196 } 4197 type = eSymbolTypeScopeEnd; 4198 break; 4199 4200 case N_EXCL: 4201 // deleted include file: name,,NO_SECT,0,sum 4202 type = eSymbolTypeHeaderFile; 4203 break; 4204 4205 //---------------------------------------------------------------------- 4206 // COMM scopes 4207 //---------------------------------------------------------------------- 4208 case N_BCOMM: 4209 // begin common: name,,NO_SECT,0,0 4210 // We use the current number of symbols in the symbol table in lieu 4211 // of 4212 // using nlist_idx in case we ever start trimming entries out 4213 type = eSymbolTypeScopeBegin; 4214 N_COMM_indexes.push_back(sym_idx); 4215 break; 4216 4217 case N_ECOML: 4218 // end common (local name): 0,,n_sect,0,address 4219 symbol_section = 4220 section_info.GetSection(nlist.n_sect, nlist.n_value); 4221 LLVM_FALLTHROUGH; 4222 4223 case N_ECOMM: 4224 // end common: name,,n_sect,0,0 4225 // Set the size of the N_BCOMM to the terminating index of this 4226 // N_ECOMM/N_ECOML 4227 // so that we can always skip the entire symbol if we need to 4228 // navigate 4229 // more quickly at the source level when parsing STABS 4230 if (!N_COMM_indexes.empty()) { 4231 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4232 symbol_ptr->SetByteSize(sym_idx + 1); 4233 symbol_ptr->SetSizeIsSibling(true); 4234 N_COMM_indexes.pop_back(); 4235 } 4236 type = eSymbolTypeScopeEnd; 4237 break; 4238 4239 case N_LENG: 4240 // second stab entry with length information 4241 type = eSymbolTypeAdditional; 4242 break; 4243 4244 default: 4245 break; 4246 } 4247 } else { 4248 // uint8_t n_pext = N_PEXT & nlist.n_type; 4249 uint8_t n_type = N_TYPE & nlist.n_type; 4250 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4251 4252 switch (n_type) { 4253 case N_INDR: { 4254 const char *reexport_name_cstr = 4255 strtab_data.PeekCStr(nlist.n_value); 4256 if (reexport_name_cstr && reexport_name_cstr[0]) { 4257 type = eSymbolTypeReExported; 4258 ConstString reexport_name( 4259 reexport_name_cstr + 4260 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4261 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4262 set_value = false; 4263 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4264 indirect_symbol_names.insert( 4265 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4266 } else 4267 type = eSymbolTypeUndefined; 4268 } break; 4269 4270 case N_UNDF: 4271 if (symbol_name && symbol_name[0]) { 4272 ConstString undefined_name(symbol_name + 4273 ((symbol_name[0] == '_') ? 1 : 0)); 4274 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4275 } 4276 LLVM_FALLTHROUGH; 4277 4278 case N_PBUD: 4279 type = eSymbolTypeUndefined; 4280 break; 4281 4282 case N_ABS: 4283 type = eSymbolTypeAbsolute; 4284 break; 4285 4286 case N_SECT: { 4287 symbol_section = 4288 section_info.GetSection(nlist.n_sect, nlist.n_value); 4289 4290 if (!symbol_section) { 4291 // TODO: warn about this? 4292 add_nlist = false; 4293 break; 4294 } 4295 4296 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4297 type = eSymbolTypeException; 4298 } else { 4299 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4300 4301 switch (section_type) { 4302 case S_CSTRING_LITERALS: 4303 type = eSymbolTypeData; 4304 break; // section with only literal C strings 4305 case S_4BYTE_LITERALS: 4306 type = eSymbolTypeData; 4307 break; // section with only 4 byte literals 4308 case S_8BYTE_LITERALS: 4309 type = eSymbolTypeData; 4310 break; // section with only 8 byte literals 4311 case S_LITERAL_POINTERS: 4312 type = eSymbolTypeTrampoline; 4313 break; // section with only pointers to literals 4314 case S_NON_LAZY_SYMBOL_POINTERS: 4315 type = eSymbolTypeTrampoline; 4316 break; // section with only non-lazy symbol pointers 4317 case S_LAZY_SYMBOL_POINTERS: 4318 type = eSymbolTypeTrampoline; 4319 break; // section with only lazy symbol pointers 4320 case S_SYMBOL_STUBS: 4321 type = eSymbolTypeTrampoline; 4322 break; // section with only symbol stubs, byte size of stub in 4323 // the reserved2 field 4324 case S_MOD_INIT_FUNC_POINTERS: 4325 type = eSymbolTypeCode; 4326 break; // section with only function pointers for initialization 4327 case S_MOD_TERM_FUNC_POINTERS: 4328 type = eSymbolTypeCode; 4329 break; // section with only function pointers for termination 4330 case S_INTERPOSING: 4331 type = eSymbolTypeTrampoline; 4332 break; // section with only pairs of function pointers for 4333 // interposing 4334 case S_16BYTE_LITERALS: 4335 type = eSymbolTypeData; 4336 break; // section with only 16 byte literals 4337 case S_DTRACE_DOF: 4338 type = eSymbolTypeInstrumentation; 4339 break; 4340 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4341 type = eSymbolTypeTrampoline; 4342 break; 4343 default: 4344 switch (symbol_section->GetType()) { 4345 case lldb::eSectionTypeCode: 4346 type = eSymbolTypeCode; 4347 break; 4348 case eSectionTypeData: 4349 case eSectionTypeDataCString: // Inlined C string data 4350 case eSectionTypeDataCStringPointers: // Pointers to C string 4351 // data 4352 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4353 // the symbol table 4354 case eSectionTypeData4: 4355 case eSectionTypeData8: 4356 case eSectionTypeData16: 4357 type = eSymbolTypeData; 4358 break; 4359 default: 4360 break; 4361 } 4362 break; 4363 } 4364 4365 if (type == eSymbolTypeInvalid) { 4366 const char *symbol_sect_name = 4367 symbol_section->GetName().AsCString(); 4368 if (symbol_section->IsDescendant(text_section_sp.get())) { 4369 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4370 S_ATTR_SELF_MODIFYING_CODE | 4371 S_ATTR_SOME_INSTRUCTIONS)) 4372 type = eSymbolTypeData; 4373 else 4374 type = eSymbolTypeCode; 4375 } else if (symbol_section->IsDescendant( 4376 data_section_sp.get()) || 4377 symbol_section->IsDescendant( 4378 data_dirty_section_sp.get()) || 4379 symbol_section->IsDescendant( 4380 data_const_section_sp.get())) { 4381 if (symbol_sect_name && 4382 ::strstr(symbol_sect_name, "__objc") == 4383 symbol_sect_name) { 4384 type = eSymbolTypeRuntime; 4385 4386 if (symbol_name) { 4387 llvm::StringRef symbol_name_ref(symbol_name); 4388 if (symbol_name_ref.startswith("_OBJC_")) { 4389 static const llvm::StringRef g_objc_v2_prefix_class( 4390 "_OBJC_CLASS_$_"); 4391 static const llvm::StringRef g_objc_v2_prefix_metaclass( 4392 "_OBJC_METACLASS_$_"); 4393 static const llvm::StringRef g_objc_v2_prefix_ivar( 4394 "_OBJC_IVAR_$_"); 4395 if (symbol_name_ref.startswith( 4396 g_objc_v2_prefix_class)) { 4397 symbol_name_non_abi_mangled = symbol_name + 1; 4398 symbol_name = 4399 symbol_name + g_objc_v2_prefix_class.size(); 4400 type = eSymbolTypeObjCClass; 4401 demangled_is_synthesized = true; 4402 } else if (symbol_name_ref.startswith( 4403 g_objc_v2_prefix_metaclass)) { 4404 symbol_name_non_abi_mangled = symbol_name + 1; 4405 symbol_name = 4406 symbol_name + g_objc_v2_prefix_metaclass.size(); 4407 type = eSymbolTypeObjCMetaClass; 4408 demangled_is_synthesized = true; 4409 } else if (symbol_name_ref.startswith( 4410 g_objc_v2_prefix_ivar)) { 4411 symbol_name_non_abi_mangled = symbol_name + 1; 4412 symbol_name = 4413 symbol_name + g_objc_v2_prefix_ivar.size(); 4414 type = eSymbolTypeObjCIVar; 4415 demangled_is_synthesized = true; 4416 } 4417 } 4418 } 4419 } else if (symbol_sect_name && 4420 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4421 symbol_sect_name) { 4422 type = eSymbolTypeException; 4423 } else { 4424 type = eSymbolTypeData; 4425 } 4426 } else if (symbol_sect_name && 4427 ::strstr(symbol_sect_name, "__IMPORT") == 4428 symbol_sect_name) { 4429 type = eSymbolTypeTrampoline; 4430 } else if (symbol_section->IsDescendant( 4431 objc_section_sp.get())) { 4432 type = eSymbolTypeRuntime; 4433 if (symbol_name && symbol_name[0] == '.') { 4434 llvm::StringRef symbol_name_ref(symbol_name); 4435 static const llvm::StringRef g_objc_v1_prefix_class( 4436 ".objc_class_name_"); 4437 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4438 symbol_name_non_abi_mangled = symbol_name; 4439 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4440 type = eSymbolTypeObjCClass; 4441 demangled_is_synthesized = true; 4442 } 4443 } 4444 } 4445 } 4446 } 4447 } break; 4448 } 4449 } 4450 4451 if (add_nlist) { 4452 uint64_t symbol_value = nlist.n_value; 4453 4454 if (symbol_name_non_abi_mangled) { 4455 sym[sym_idx].GetMangled().SetMangledName( 4456 ConstString(symbol_name_non_abi_mangled)); 4457 sym[sym_idx].GetMangled().SetDemangledName( 4458 ConstString(symbol_name)); 4459 } else { 4460 bool symbol_name_is_mangled = false; 4461 4462 if (symbol_name && symbol_name[0] == '_') { 4463 symbol_name_is_mangled = symbol_name[1] == '_'; 4464 symbol_name++; // Skip the leading underscore 4465 } 4466 4467 if (symbol_name) { 4468 ConstString const_symbol_name(symbol_name); 4469 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4470 symbol_name_is_mangled); 4471 } 4472 } 4473 4474 if (is_gsym) { 4475 const char *gsym_name = sym[sym_idx] 4476 .GetMangled() 4477 .GetName(lldb::eLanguageTypeUnknown, 4478 Mangled::ePreferMangled) 4479 .GetCString(); 4480 if (gsym_name) 4481 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4482 } 4483 4484 if (symbol_section) { 4485 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4486 if (symbol_byte_size == 0 && function_starts_count > 0) { 4487 addr_t symbol_lookup_file_addr = nlist.n_value; 4488 // Do an exact address match for non-ARM addresses, else get the 4489 // closest since 4490 // the symbol might be a thumb symbol which has an address with 4491 // bit zero set 4492 FunctionStarts::Entry *func_start_entry = 4493 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4494 if (is_arm && func_start_entry) { 4495 // Verify that the function start address is the symbol address 4496 // (ARM) 4497 // or the symbol address + 1 (thumb) 4498 if (func_start_entry->addr != symbol_lookup_file_addr && 4499 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4500 // Not the right entry, NULL it out... 4501 func_start_entry = NULL; 4502 } 4503 } 4504 if (func_start_entry) { 4505 func_start_entry->data = true; 4506 4507 addr_t symbol_file_addr = func_start_entry->addr; 4508 if (is_arm) 4509 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4510 4511 const FunctionStarts::Entry *next_func_start_entry = 4512 function_starts.FindNextEntry(func_start_entry); 4513 const addr_t section_end_file_addr = 4514 section_file_addr + symbol_section->GetByteSize(); 4515 if (next_func_start_entry) { 4516 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4517 // Be sure the clear the Thumb address bit when we calculate 4518 // the size 4519 // from the current and next address 4520 if (is_arm) 4521 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4522 symbol_byte_size = std::min<lldb::addr_t>( 4523 next_symbol_file_addr - symbol_file_addr, 4524 section_end_file_addr - symbol_file_addr); 4525 } else { 4526 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4527 } 4528 } 4529 } 4530 symbol_value -= section_file_addr; 4531 } 4532 4533 if (is_debug == false) { 4534 if (type == eSymbolTypeCode) { 4535 // See if we can find a N_FUN entry for any code symbols. 4536 // If we do find a match, and the name matches, then we 4537 // can merge the two into just the function symbol to avoid 4538 // duplicate entries in the symbol table 4539 std::pair<ValueToSymbolIndexMap::const_iterator, 4540 ValueToSymbolIndexMap::const_iterator> 4541 range; 4542 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4543 if (range.first != range.second) { 4544 bool found_it = false; 4545 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4546 pos != range.second; ++pos) { 4547 if (sym[sym_idx].GetMangled().GetName( 4548 lldb::eLanguageTypeUnknown, 4549 Mangled::ePreferMangled) == 4550 sym[pos->second].GetMangled().GetName( 4551 lldb::eLanguageTypeUnknown, 4552 Mangled::ePreferMangled)) { 4553 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4554 // We just need the flags from the linker symbol, so put 4555 // these flags 4556 // into the N_FUN flags to avoid duplicate symbols in the 4557 // symbol table 4558 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4559 sym[pos->second].SetFlags(nlist.n_type << 16 | 4560 nlist.n_desc); 4561 if (resolver_addresses.find(nlist.n_value) != 4562 resolver_addresses.end()) 4563 sym[pos->second].SetType(eSymbolTypeResolver); 4564 sym[sym_idx].Clear(); 4565 found_it = true; 4566 break; 4567 } 4568 } 4569 if (found_it) 4570 continue; 4571 } else { 4572 if (resolver_addresses.find(nlist.n_value) != 4573 resolver_addresses.end()) 4574 type = eSymbolTypeResolver; 4575 } 4576 } else if (type == eSymbolTypeData || 4577 type == eSymbolTypeObjCClass || 4578 type == eSymbolTypeObjCMetaClass || 4579 type == eSymbolTypeObjCIVar) { 4580 // See if we can find a N_STSYM entry for any data symbols. 4581 // If we do find a match, and the name matches, then we 4582 // can merge the two into just the Static symbol to avoid 4583 // duplicate entries in the symbol table 4584 std::pair<ValueToSymbolIndexMap::const_iterator, 4585 ValueToSymbolIndexMap::const_iterator> 4586 range; 4587 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4588 if (range.first != range.second) { 4589 bool found_it = false; 4590 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4591 pos != range.second; ++pos) { 4592 if (sym[sym_idx].GetMangled().GetName( 4593 lldb::eLanguageTypeUnknown, 4594 Mangled::ePreferMangled) == 4595 sym[pos->second].GetMangled().GetName( 4596 lldb::eLanguageTypeUnknown, 4597 Mangled::ePreferMangled)) { 4598 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4599 // We just need the flags from the linker symbol, so put 4600 // these flags 4601 // into the N_STSYM flags to avoid duplicate symbols in the 4602 // symbol table 4603 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4604 sym[pos->second].SetFlags(nlist.n_type << 16 | 4605 nlist.n_desc); 4606 sym[sym_idx].Clear(); 4607 found_it = true; 4608 break; 4609 } 4610 } 4611 if (found_it) 4612 continue; 4613 } else { 4614 // Combine N_GSYM stab entries with the non stab symbol 4615 const char *gsym_name = sym[sym_idx] 4616 .GetMangled() 4617 .GetName(lldb::eLanguageTypeUnknown, 4618 Mangled::ePreferMangled) 4619 .GetCString(); 4620 if (gsym_name) { 4621 ConstNameToSymbolIndexMap::const_iterator pos = 4622 N_GSYM_name_to_sym_idx.find(gsym_name); 4623 if (pos != N_GSYM_name_to_sym_idx.end()) { 4624 const uint32_t GSYM_sym_idx = pos->second; 4625 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4626 // Copy the address, because often the N_GSYM address has an 4627 // invalid address of zero 4628 // when the global is a common symbol 4629 sym[GSYM_sym_idx].GetAddressRef().SetSection( 4630 symbol_section); 4631 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4632 // We just need the flags from the linker symbol, so put 4633 // these flags 4634 // into the N_GSYM flags to avoid duplicate symbols in the 4635 // symbol table 4636 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 4637 nlist.n_desc); 4638 sym[sym_idx].Clear(); 4639 continue; 4640 } 4641 } 4642 } 4643 } 4644 } 4645 4646 sym[sym_idx].SetID(nlist_idx); 4647 sym[sym_idx].SetType(type); 4648 if (set_value) { 4649 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4650 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4651 } 4652 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4653 4654 if (symbol_byte_size > 0) 4655 sym[sym_idx].SetByteSize(symbol_byte_size); 4656 4657 if (demangled_is_synthesized) 4658 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4659 4660 ++sym_idx; 4661 } else { 4662 sym[sym_idx].Clear(); 4663 } 4664 } 4665 4666 for (const auto &pos : reexport_shlib_needs_fixup) { 4667 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4668 if (undef_pos != undefined_name_to_desc.end()) { 4669 const uint8_t dylib_ordinal = 4670 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4671 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4672 sym[pos.first].SetReExportedSymbolSharedLibrary( 4673 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4674 } 4675 } 4676 } 4677 4678 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4679 4680 if (function_starts_count > 0) { 4681 uint32_t num_synthetic_function_symbols = 0; 4682 for (i = 0; i < function_starts_count; ++i) { 4683 if (function_starts.GetEntryRef(i).data == false) 4684 ++num_synthetic_function_symbols; 4685 } 4686 4687 if (num_synthetic_function_symbols > 0) { 4688 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4689 num_syms = sym_idx + num_synthetic_function_symbols; 4690 sym = symtab->Resize(num_syms); 4691 } 4692 for (i = 0; i < function_starts_count; ++i) { 4693 const FunctionStarts::Entry *func_start_entry = 4694 function_starts.GetEntryAtIndex(i); 4695 if (func_start_entry->data == false) { 4696 addr_t symbol_file_addr = func_start_entry->addr; 4697 uint32_t symbol_flags = 0; 4698 if (is_arm) { 4699 if (symbol_file_addr & 1) 4700 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4701 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4702 } 4703 Address symbol_addr; 4704 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4705 SectionSP symbol_section(symbol_addr.GetSection()); 4706 uint32_t symbol_byte_size = 0; 4707 if (symbol_section) { 4708 const addr_t section_file_addr = 4709 symbol_section->GetFileAddress(); 4710 const FunctionStarts::Entry *next_func_start_entry = 4711 function_starts.FindNextEntry(func_start_entry); 4712 const addr_t section_end_file_addr = 4713 section_file_addr + symbol_section->GetByteSize(); 4714 if (next_func_start_entry) { 4715 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4716 if (is_arm) 4717 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4718 symbol_byte_size = std::min<lldb::addr_t>( 4719 next_symbol_file_addr - symbol_file_addr, 4720 section_end_file_addr - symbol_file_addr); 4721 } else { 4722 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4723 } 4724 sym[sym_idx].SetID(synthetic_sym_id++); 4725 sym[sym_idx].GetMangled().SetDemangledName( 4726 GetNextSyntheticSymbolName()); 4727 sym[sym_idx].SetType(eSymbolTypeCode); 4728 sym[sym_idx].SetIsSynthetic(true); 4729 sym[sym_idx].GetAddressRef() = symbol_addr; 4730 if (symbol_flags) 4731 sym[sym_idx].SetFlags(symbol_flags); 4732 if (symbol_byte_size) 4733 sym[sym_idx].SetByteSize(symbol_byte_size); 4734 ++sym_idx; 4735 } 4736 } 4737 } 4738 } 4739 } 4740 } 4741 4742 // Trim our symbols down to just what we ended up with after 4743 // removing any symbols. 4744 if (sym_idx < num_syms) { 4745 num_syms = sym_idx; 4746 sym = symtab->Resize(num_syms); 4747 } 4748 4749 // Now synthesize indirect symbols 4750 if (m_dysymtab.nindirectsyms != 0) { 4751 if (indirect_symbol_index_data.GetByteSize()) { 4752 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4753 m_nlist_idx_to_sym_idx.end(); 4754 4755 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4756 ++sect_idx) { 4757 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4758 S_SYMBOL_STUBS) { 4759 uint32_t symbol_stub_byte_size = 4760 m_mach_sections[sect_idx].reserved2; 4761 if (symbol_stub_byte_size == 0) 4762 continue; 4763 4764 const uint32_t num_symbol_stubs = 4765 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4766 4767 if (num_symbol_stubs == 0) 4768 continue; 4769 4770 const uint32_t symbol_stub_index_offset = 4771 m_mach_sections[sect_idx].reserved1; 4772 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; 4773 ++stub_idx) { 4774 const uint32_t symbol_stub_index = 4775 symbol_stub_index_offset + stub_idx; 4776 const lldb::addr_t symbol_stub_addr = 4777 m_mach_sections[sect_idx].addr + 4778 (stub_idx * symbol_stub_byte_size); 4779 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4780 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4781 symbol_stub_offset, 4)) { 4782 const uint32_t stub_sym_id = 4783 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4784 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4785 continue; 4786 4787 NListIndexToSymbolIndexMap::const_iterator index_pos = 4788 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4789 Symbol *stub_symbol = NULL; 4790 if (index_pos != end_index_pos) { 4791 // We have a remapping from the original nlist index to 4792 // a current symbol index, so just look this up by index 4793 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4794 } else { 4795 // We need to lookup a symbol using the original nlist 4796 // symbol index since this index is coming from the 4797 // S_SYMBOL_STUBS 4798 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4799 } 4800 4801 if (stub_symbol) { 4802 Address so_addr(symbol_stub_addr, section_list); 4803 4804 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4805 // Change the external symbol into a trampoline that makes 4806 // sense 4807 // These symbols were N_UNDF N_EXT, and are useless to us, 4808 // so we 4809 // can re-use them so we don't have to make up a synthetic 4810 // symbol 4811 // for no good reason. 4812 if (resolver_addresses.find(symbol_stub_addr) == 4813 resolver_addresses.end()) 4814 stub_symbol->SetType(eSymbolTypeTrampoline); 4815 else 4816 stub_symbol->SetType(eSymbolTypeResolver); 4817 stub_symbol->SetExternal(false); 4818 stub_symbol->GetAddressRef() = so_addr; 4819 stub_symbol->SetByteSize(symbol_stub_byte_size); 4820 } else { 4821 // Make a synthetic symbol to describe the trampoline stub 4822 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4823 if (sym_idx >= num_syms) { 4824 sym = symtab->Resize(++num_syms); 4825 stub_symbol = NULL; // this pointer no longer valid 4826 } 4827 sym[sym_idx].SetID(synthetic_sym_id++); 4828 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4829 if (resolver_addresses.find(symbol_stub_addr) == 4830 resolver_addresses.end()) 4831 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4832 else 4833 sym[sym_idx].SetType(eSymbolTypeResolver); 4834 sym[sym_idx].SetIsSynthetic(true); 4835 sym[sym_idx].GetAddressRef() = so_addr; 4836 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4837 ++sym_idx; 4838 } 4839 } else { 4840 if (log) 4841 log->Warning("symbol stub referencing symbol table symbol " 4842 "%u that isn't in our minimal symbol table, " 4843 "fix this!!!", 4844 stub_sym_id); 4845 } 4846 } 4847 } 4848 } 4849 } 4850 } 4851 } 4852 4853 if (!trie_entries.empty()) { 4854 for (const auto &e : trie_entries) { 4855 if (e.entry.import_name) { 4856 // Only add indirect symbols from the Trie entries if we 4857 // didn't have a N_INDR nlist entry for this already 4858 if (indirect_symbol_names.find(e.entry.name) == 4859 indirect_symbol_names.end()) { 4860 // Make a synthetic symbol to describe re-exported symbol. 4861 if (sym_idx >= num_syms) 4862 sym = symtab->Resize(++num_syms); 4863 sym[sym_idx].SetID(synthetic_sym_id++); 4864 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4865 sym[sym_idx].SetType(eSymbolTypeReExported); 4866 sym[sym_idx].SetIsSynthetic(true); 4867 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4868 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4869 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4870 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4871 } 4872 ++sym_idx; 4873 } 4874 } 4875 } 4876 } 4877 4878 // StreamFile s(stdout, false); 4879 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4880 // symtab->Dump(&s, NULL, eSortOrderNone); 4881 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4882 // sizes 4883 symtab->CalculateSymbolSizes(); 4884 4885 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4886 // symtab->Dump(&s, NULL, eSortOrderNone); 4887 4888 return symtab->GetNumSymbols(); 4889 } 4890 return 0; 4891 } 4892 4893 void ObjectFileMachO::Dump(Stream *s) { 4894 ModuleSP module_sp(GetModule()); 4895 if (module_sp) { 4896 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4897 s->Printf("%p: ", static_cast<void *>(this)); 4898 s->Indent(); 4899 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4900 s->PutCString("ObjectFileMachO64"); 4901 else 4902 s->PutCString("ObjectFileMachO32"); 4903 4904 ArchSpec header_arch; 4905 GetArchitecture(header_arch); 4906 4907 *s << ", file = '" << m_file 4908 << "', arch = " << header_arch.GetArchitectureName() << "\n"; 4909 4910 SectionList *sections = GetSectionList(); 4911 if (sections) 4912 sections->Dump(s, NULL, true, UINT32_MAX); 4913 4914 if (m_symtab_ap.get()) 4915 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4916 } 4917 } 4918 4919 bool ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4920 const lldb_private::DataExtractor &data, 4921 lldb::offset_t lc_offset, 4922 lldb_private::UUID &uuid) { 4923 uint32_t i; 4924 struct uuid_command load_cmd; 4925 4926 lldb::offset_t offset = lc_offset; 4927 for (i = 0; i < header.ncmds; ++i) { 4928 const lldb::offset_t cmd_offset = offset; 4929 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4930 break; 4931 4932 if (load_cmd.cmd == LC_UUID) { 4933 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4934 4935 if (uuid_bytes) { 4936 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4937 // We pretend these object files have no UUID to prevent crashing. 4938 4939 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4940 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4941 0xbb, 0x14, 0xf0, 0x0d}; 4942 4943 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4944 return false; 4945 4946 uuid.SetBytes(uuid_bytes); 4947 return true; 4948 } 4949 return false; 4950 } 4951 offset = cmd_offset + load_cmd.cmdsize; 4952 } 4953 return false; 4954 } 4955 4956 bool ObjectFileMachO::GetArchitecture(const llvm::MachO::mach_header &header, 4957 const lldb_private::DataExtractor &data, 4958 lldb::offset_t lc_offset, 4959 ArchSpec &arch) { 4960 arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4961 4962 if (arch.IsValid()) { 4963 llvm::Triple &triple = arch.GetTriple(); 4964 4965 // Set OS to an unspecified unknown or a "*" so it can match any OS 4966 triple.setOS(llvm::Triple::UnknownOS); 4967 triple.setOSName(llvm::StringRef()); 4968 4969 if (header.filetype == MH_PRELOAD) { 4970 if (header.cputype == CPU_TYPE_ARM) { 4971 // If this is a 32-bit arm binary, and it's a standalone binary, 4972 // force the Vendor to Apple so we don't accidentally pick up 4973 // the generic armv7 ABI at runtime. Apple's armv7 ABI always uses 4974 // r7 for the frame pointer register; most other armv7 ABIs use a 4975 // combination of r7 and r11. 4976 triple.setVendor(llvm::Triple::Apple); 4977 } else { 4978 // Set vendor to an unspecified unknown or a "*" so it can match any 4979 // vendor 4980 // This is required for correct behavior of EFI debugging on x86_64 4981 triple.setVendor(llvm::Triple::UnknownVendor); 4982 triple.setVendorName(llvm::StringRef()); 4983 } 4984 return true; 4985 } else { 4986 struct load_command load_cmd; 4987 4988 lldb::offset_t offset = lc_offset; 4989 for (uint32_t i = 0; i < header.ncmds; ++i) { 4990 const lldb::offset_t cmd_offset = offset; 4991 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4992 break; 4993 4994 switch (load_cmd.cmd) { 4995 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4996 triple.setOS(llvm::Triple::IOS); 4997 return true; 4998 4999 case llvm::MachO::LC_VERSION_MIN_MACOSX: 5000 triple.setOS(llvm::Triple::MacOSX); 5001 return true; 5002 5003 case llvm::MachO::LC_VERSION_MIN_TVOS: 5004 triple.setOS(llvm::Triple::TvOS); 5005 return true; 5006 5007 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 5008 triple.setOS(llvm::Triple::WatchOS); 5009 return true; 5010 5011 default: 5012 break; 5013 } 5014 5015 offset = cmd_offset + load_cmd.cmdsize; 5016 } 5017 5018 if (header.filetype != MH_KEXT_BUNDLE) { 5019 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 5020 // so lets not say our Vendor is Apple, leave it as an unspecified 5021 // unknown 5022 triple.setVendor(llvm::Triple::UnknownVendor); 5023 triple.setVendorName(llvm::StringRef()); 5024 } 5025 } 5026 } 5027 return arch.IsValid(); 5028 } 5029 5030 bool ObjectFileMachO::GetUUID(lldb_private::UUID *uuid) { 5031 ModuleSP module_sp(GetModule()); 5032 if (module_sp) { 5033 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5034 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5035 return GetUUID(m_header, m_data, offset, *uuid); 5036 } 5037 return false; 5038 } 5039 5040 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5041 uint32_t count = 0; 5042 ModuleSP module_sp(GetModule()); 5043 if (module_sp) { 5044 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5045 struct load_command load_cmd; 5046 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5047 std::vector<std::string> rpath_paths; 5048 std::vector<std::string> rpath_relative_paths; 5049 std::vector<std::string> at_exec_relative_paths; 5050 const bool resolve_path = false; // Don't resolve the dependent file paths 5051 // since they may not reside on this system 5052 uint32_t i; 5053 for (i = 0; i < m_header.ncmds; ++i) { 5054 const uint32_t cmd_offset = offset; 5055 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5056 break; 5057 5058 switch (load_cmd.cmd) { 5059 case LC_RPATH: 5060 case LC_LOAD_DYLIB: 5061 case LC_LOAD_WEAK_DYLIB: 5062 case LC_REEXPORT_DYLIB: 5063 case LC_LOAD_DYLINKER: 5064 case LC_LOADFVMLIB: 5065 case LC_LOAD_UPWARD_DYLIB: { 5066 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5067 const char *path = m_data.PeekCStr(name_offset); 5068 if (path) { 5069 if (load_cmd.cmd == LC_RPATH) 5070 rpath_paths.push_back(path); 5071 else { 5072 if (path[0] == '@') { 5073 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5074 rpath_relative_paths.push_back(path + strlen("@rpath")); 5075 else if (strncmp(path, "@executable_path", 5076 strlen("@executable_path")) == 0) 5077 at_exec_relative_paths.push_back(path 5078 + strlen("@executable_path")); 5079 } else { 5080 FileSpec file_spec(path, resolve_path); 5081 if (files.AppendIfUnique(file_spec)) 5082 count++; 5083 } 5084 } 5085 } 5086 } break; 5087 5088 default: 5089 break; 5090 } 5091 offset = cmd_offset + load_cmd.cmdsize; 5092 } 5093 5094 FileSpec this_file_spec(m_file); 5095 this_file_spec.ResolvePath(); 5096 5097 if (!rpath_paths.empty()) { 5098 // Fixup all LC_RPATH values to be absolute paths 5099 std::string loader_path("@loader_path"); 5100 std::string executable_path("@executable_path"); 5101 for (auto &rpath : rpath_paths) { 5102 if (rpath.find(loader_path) == 0) { 5103 rpath.erase(0, loader_path.size()); 5104 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5105 } else if (rpath.find(executable_path) == 0) { 5106 rpath.erase(0, executable_path.size()); 5107 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5108 } 5109 } 5110 5111 for (const auto &rpath_relative_path : rpath_relative_paths) { 5112 for (const auto &rpath : rpath_paths) { 5113 std::string path = rpath; 5114 path += rpath_relative_path; 5115 // It is OK to resolve this path because we must find a file on 5116 // disk for us to accept it anyway if it is rpath relative. 5117 FileSpec file_spec(path, true); 5118 // Remove any redundant parts of the path (like "../foo") since 5119 // LC_RPATH values often contain "..". 5120 file_spec = file_spec.GetNormalizedPath(); 5121 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) { 5122 count++; 5123 break; 5124 } 5125 } 5126 } 5127 } 5128 5129 // We may have @executable_paths but no RPATHS. Figure those out here. 5130 // Only do this if this object file is the executable. We have no way to 5131 // get back to the actual executable otherwise, so we won't get the right 5132 // path. 5133 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5134 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5135 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5136 FileSpec file_spec = 5137 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5138 file_spec = file_spec.GetNormalizedPath(); 5139 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) { 5140 count++; 5141 break; 5142 } 5143 } 5144 } 5145 } 5146 return count; 5147 } 5148 5149 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5150 // If the object file is not an executable it can't hold the entry point. 5151 // m_entry_point_address 5152 // is initialized to an invalid address, so we can just return that. 5153 // If m_entry_point_address is valid it means we've found it already, so 5154 // return the cached value. 5155 5156 if (!IsExecutable() || m_entry_point_address.IsValid()) 5157 return m_entry_point_address; 5158 5159 // Otherwise, look for the UnixThread or Thread command. The data for the 5160 // Thread command is given in 5161 // /usr/include/mach-o.h, but it is basically: 5162 // 5163 // uint32_t flavor - this is the flavor argument you would pass to 5164 // thread_get_state 5165 // uint32_t count - this is the count of longs in the thread state data 5166 // struct XXX_thread_state state - this is the structure from 5167 // <machine/thread_status.h> corresponding to the flavor. 5168 // <repeat this trio> 5169 // 5170 // So we just keep reading the various register flavors till we find the GPR 5171 // one, then read the PC out of there. 5172 // FIXME: We will need to have a "RegisterContext data provider" class at some 5173 // point that can get all the registers 5174 // out of data in this form & attach them to a given thread. That should 5175 // underlie the MacOS X User process plugin, 5176 // and we'll also need it for the MacOS X Core File process plugin. When we 5177 // have that we can also use it here. 5178 // 5179 // For now we hard-code the offsets and flavors we need: 5180 // 5181 // 5182 5183 ModuleSP module_sp(GetModule()); 5184 if (module_sp) { 5185 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5186 struct load_command load_cmd; 5187 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5188 uint32_t i; 5189 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5190 bool done = false; 5191 5192 for (i = 0; i < m_header.ncmds; ++i) { 5193 const lldb::offset_t cmd_offset = offset; 5194 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5195 break; 5196 5197 switch (load_cmd.cmd) { 5198 case LC_UNIXTHREAD: 5199 case LC_THREAD: { 5200 while (offset < cmd_offset + load_cmd.cmdsize) { 5201 uint32_t flavor = m_data.GetU32(&offset); 5202 uint32_t count = m_data.GetU32(&offset); 5203 if (count == 0) { 5204 // We've gotten off somehow, log and exit; 5205 return m_entry_point_address; 5206 } 5207 5208 switch (m_header.cputype) { 5209 case llvm::MachO::CPU_TYPE_ARM: 5210 if (flavor == 1 || 5211 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 from 5212 // mach/arm/thread_status.h 5213 { 5214 offset += 60; // This is the offset of pc in the GPR thread state 5215 // data structure. 5216 start_address = m_data.GetU32(&offset); 5217 done = true; 5218 } 5219 break; 5220 case llvm::MachO::CPU_TYPE_ARM64: 5221 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5222 { 5223 offset += 256; // This is the offset of pc in the GPR thread state 5224 // data structure. 5225 start_address = m_data.GetU64(&offset); 5226 done = true; 5227 } 5228 break; 5229 case llvm::MachO::CPU_TYPE_I386: 5230 if (flavor == 5231 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5232 { 5233 offset += 40; // This is the offset of eip in the GPR thread state 5234 // data structure. 5235 start_address = m_data.GetU32(&offset); 5236 done = true; 5237 } 5238 break; 5239 case llvm::MachO::CPU_TYPE_X86_64: 5240 if (flavor == 5241 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5242 { 5243 offset += 16 * 8; // This is the offset of rip in the GPR thread 5244 // state data structure. 5245 start_address = m_data.GetU64(&offset); 5246 done = true; 5247 } 5248 break; 5249 default: 5250 return m_entry_point_address; 5251 } 5252 // Haven't found the GPR flavor yet, skip over the data for this 5253 // flavor: 5254 if (done) 5255 break; 5256 offset += count * 4; 5257 } 5258 } break; 5259 case LC_MAIN: { 5260 ConstString text_segment_name("__TEXT"); 5261 uint64_t entryoffset = m_data.GetU64(&offset); 5262 SectionSP text_segment_sp = 5263 GetSectionList()->FindSectionByName(text_segment_name); 5264 if (text_segment_sp) { 5265 done = true; 5266 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5267 } 5268 } break; 5269 5270 default: 5271 break; 5272 } 5273 if (done) 5274 break; 5275 5276 // Go to the next load command: 5277 offset = cmd_offset + load_cmd.cmdsize; 5278 } 5279 5280 if (start_address != LLDB_INVALID_ADDRESS) { 5281 // We got the start address from the load commands, so now resolve that 5282 // address in the sections 5283 // of this ObjectFile: 5284 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5285 start_address, GetSectionList())) { 5286 m_entry_point_address.Clear(); 5287 } 5288 } else { 5289 // We couldn't read the UnixThread load command - maybe it wasn't there. 5290 // As a fallback look for the 5291 // "start" symbol in the main executable. 5292 5293 ModuleSP module_sp(GetModule()); 5294 5295 if (module_sp) { 5296 SymbolContextList contexts; 5297 SymbolContext context; 5298 if (module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5299 eSymbolTypeCode, contexts)) { 5300 if (contexts.GetContextAtIndex(0, context)) 5301 m_entry_point_address = context.symbol->GetAddress(); 5302 } 5303 } 5304 } 5305 } 5306 5307 return m_entry_point_address; 5308 } 5309 5310 lldb_private::Address ObjectFileMachO::GetHeaderAddress() { 5311 lldb_private::Address header_addr; 5312 SectionList *section_list = GetSectionList(); 5313 if (section_list) { 5314 SectionSP text_segment_sp( 5315 section_list->FindSectionByName(GetSegmentNameTEXT())); 5316 if (text_segment_sp) { 5317 header_addr.SetSection(text_segment_sp); 5318 header_addr.SetOffset(0); 5319 } 5320 } 5321 return header_addr; 5322 } 5323 5324 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5325 ModuleSP module_sp(GetModule()); 5326 if (module_sp) { 5327 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5328 if (!m_thread_context_offsets_valid) { 5329 m_thread_context_offsets_valid = true; 5330 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5331 FileRangeArray::Entry file_range; 5332 thread_command thread_cmd; 5333 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5334 const uint32_t cmd_offset = offset; 5335 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 5336 break; 5337 5338 if (thread_cmd.cmd == LC_THREAD) { 5339 file_range.SetRangeBase(offset); 5340 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5341 m_thread_context_offsets.Append(file_range); 5342 } 5343 offset = cmd_offset + thread_cmd.cmdsize; 5344 } 5345 } 5346 } 5347 return m_thread_context_offsets.GetSize(); 5348 } 5349 5350 std::string ObjectFileMachO::GetIdentifierString() { 5351 std::string result; 5352 ModuleSP module_sp(GetModule()); 5353 if (module_sp) { 5354 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5355 5356 // First, look over the load commands for an LC_NOTE load command 5357 // with data_owner string "kern ver str" & use that if found. 5358 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5359 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5360 const uint32_t cmd_offset = offset; 5361 load_command lc; 5362 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5363 break; 5364 if (lc.cmd == LC_NOTE) 5365 { 5366 char data_owner[17]; 5367 m_data.CopyData (offset, 16, data_owner); 5368 data_owner[16] = '\0'; 5369 offset += 16; 5370 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5371 uint64_t size = m_data.GetU64_unchecked (&offset); 5372 5373 // "kern ver str" has a uint32_t version and then a 5374 // nul terminated c-string. 5375 if (strcmp ("kern ver str", data_owner) == 0) 5376 { 5377 offset = fileoff; 5378 uint32_t version; 5379 if (m_data.GetU32 (&offset, &version, 1) != nullptr) 5380 { 5381 if (version == 1) 5382 { 5383 uint32_t strsize = size - sizeof (uint32_t); 5384 char *buf = (char*) malloc (strsize); 5385 if (buf) 5386 { 5387 m_data.CopyData (offset, strsize, buf); 5388 buf[strsize - 1] = '\0'; 5389 result = buf; 5390 if (buf) 5391 free (buf); 5392 return result; 5393 } 5394 } 5395 } 5396 } 5397 } 5398 offset = cmd_offset + lc.cmdsize; 5399 } 5400 5401 // Second, make a pass over the load commands looking for an 5402 // obsolete LC_IDENT load command. 5403 offset = MachHeaderSizeFromMagic(m_header.magic); 5404 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5405 const uint32_t cmd_offset = offset; 5406 struct ident_command ident_command; 5407 if (m_data.GetU32(&offset, &ident_command, 2) == NULL) 5408 break; 5409 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5410 char *buf = (char *) malloc (ident_command.cmdsize); 5411 if (buf != nullptr 5412 && m_data.CopyData (offset, ident_command.cmdsize, buf) == ident_command.cmdsize) { 5413 buf[ident_command.cmdsize - 1] = '\0'; 5414 result = buf; 5415 } 5416 if (buf) 5417 free (buf); 5418 } 5419 offset = cmd_offset + ident_command.cmdsize; 5420 } 5421 5422 } 5423 return result; 5424 } 5425 5426 bool ObjectFileMachO::GetCorefileMainBinaryInfo (addr_t &address, UUID &uuid) { 5427 address = LLDB_INVALID_ADDRESS; 5428 uuid.Clear(); 5429 ModuleSP module_sp(GetModule()); 5430 if (module_sp) { 5431 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5432 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5433 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5434 const uint32_t cmd_offset = offset; 5435 load_command lc; 5436 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5437 break; 5438 if (lc.cmd == LC_NOTE) 5439 { 5440 char data_owner[17]; 5441 memset (data_owner, 0, sizeof (data_owner)); 5442 m_data.CopyData (offset, 16, data_owner); 5443 offset += 16; 5444 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5445 uint64_t size = m_data.GetU64_unchecked (&offset); 5446 5447 // "main bin spec" (main binary specification) data payload is formatted: 5448 // uint32_t version [currently 1] 5449 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user process] 5450 // uint64_t address [ UINT64_MAX if address not specified ] 5451 // uuid_t uuid [ all zero's if uuid not specified ] 5452 // uint32_t log2_pagesize [ process page size in log base 2, e.g. 4k pages are 12. 0 for unspecified ] 5453 5454 if (strcmp ("main bin spec", data_owner) == 0 && size >= 32) 5455 { 5456 offset = fileoff; 5457 uint32_t version; 5458 if (m_data.GetU32 (&offset, &version, 1) != nullptr && version == 1) 5459 { 5460 uint32_t type = 0; 5461 uuid_t raw_uuid; 5462 memset (raw_uuid, 0, sizeof (uuid_t)); 5463 5464 if (m_data.GetU32 (&offset, &type, 1) 5465 && m_data.GetU64 (&offset, &address, 1) 5466 && m_data.CopyData (offset, sizeof (uuid_t), raw_uuid) != 0 5467 && uuid.SetBytes (raw_uuid, sizeof (uuid_t))) 5468 { 5469 return true; 5470 } 5471 } 5472 } 5473 } 5474 offset = cmd_offset + lc.cmdsize; 5475 } 5476 } 5477 return false; 5478 } 5479 5480 lldb::RegisterContextSP 5481 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5482 lldb_private::Thread &thread) { 5483 lldb::RegisterContextSP reg_ctx_sp; 5484 5485 ModuleSP module_sp(GetModule()); 5486 if (module_sp) { 5487 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5488 if (!m_thread_context_offsets_valid) 5489 GetNumThreadContexts(); 5490 5491 const FileRangeArray::Entry *thread_context_file_range = 5492 m_thread_context_offsets.GetEntryAtIndex(idx); 5493 if (thread_context_file_range) { 5494 5495 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5496 thread_context_file_range->GetByteSize()); 5497 5498 switch (m_header.cputype) { 5499 case llvm::MachO::CPU_TYPE_ARM64: 5500 reg_ctx_sp.reset(new RegisterContextDarwin_arm64_Mach(thread, data)); 5501 break; 5502 5503 case llvm::MachO::CPU_TYPE_ARM: 5504 reg_ctx_sp.reset(new RegisterContextDarwin_arm_Mach(thread, data)); 5505 break; 5506 5507 case llvm::MachO::CPU_TYPE_I386: 5508 reg_ctx_sp.reset(new RegisterContextDarwin_i386_Mach(thread, data)); 5509 break; 5510 5511 case llvm::MachO::CPU_TYPE_X86_64: 5512 reg_ctx_sp.reset(new RegisterContextDarwin_x86_64_Mach(thread, data)); 5513 break; 5514 } 5515 } 5516 } 5517 return reg_ctx_sp; 5518 } 5519 5520 ObjectFile::Type ObjectFileMachO::CalculateType() { 5521 switch (m_header.filetype) { 5522 case MH_OBJECT: // 0x1u 5523 if (GetAddressByteSize() == 4) { 5524 // 32 bit kexts are just object files, but they do have a valid 5525 // UUID load command. 5526 UUID uuid; 5527 if (GetUUID(&uuid)) { 5528 // this checking for the UUID load command is not enough 5529 // we could eventually look for the symbol named 5530 // "OSKextGetCurrentIdentifier" as this is required of kexts 5531 if (m_strata == eStrataInvalid) 5532 m_strata = eStrataKernel; 5533 return eTypeSharedLibrary; 5534 } 5535 } 5536 return eTypeObjectFile; 5537 5538 case MH_EXECUTE: 5539 return eTypeExecutable; // 0x2u 5540 case MH_FVMLIB: 5541 return eTypeSharedLibrary; // 0x3u 5542 case MH_CORE: 5543 return eTypeCoreFile; // 0x4u 5544 case MH_PRELOAD: 5545 return eTypeSharedLibrary; // 0x5u 5546 case MH_DYLIB: 5547 return eTypeSharedLibrary; // 0x6u 5548 case MH_DYLINKER: 5549 return eTypeDynamicLinker; // 0x7u 5550 case MH_BUNDLE: 5551 return eTypeSharedLibrary; // 0x8u 5552 case MH_DYLIB_STUB: 5553 return eTypeStubLibrary; // 0x9u 5554 case MH_DSYM: 5555 return eTypeDebugInfo; // 0xAu 5556 case MH_KEXT_BUNDLE: 5557 return eTypeSharedLibrary; // 0xBu 5558 default: 5559 break; 5560 } 5561 return eTypeUnknown; 5562 } 5563 5564 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5565 switch (m_header.filetype) { 5566 case MH_OBJECT: // 0x1u 5567 { 5568 // 32 bit kexts are just object files, but they do have a valid 5569 // UUID load command. 5570 UUID uuid; 5571 if (GetUUID(&uuid)) { 5572 // this checking for the UUID load command is not enough 5573 // we could eventually look for the symbol named 5574 // "OSKextGetCurrentIdentifier" as this is required of kexts 5575 if (m_type == eTypeInvalid) 5576 m_type = eTypeSharedLibrary; 5577 5578 return eStrataKernel; 5579 } 5580 } 5581 return eStrataUnknown; 5582 5583 case MH_EXECUTE: // 0x2u 5584 // Check for the MH_DYLDLINK bit in the flags 5585 if (m_header.flags & MH_DYLDLINK) { 5586 return eStrataUser; 5587 } else { 5588 SectionList *section_list = GetSectionList(); 5589 if (section_list) { 5590 static ConstString g_kld_section_name("__KLD"); 5591 if (section_list->FindSectionByName(g_kld_section_name)) 5592 return eStrataKernel; 5593 } 5594 } 5595 return eStrataRawImage; 5596 5597 case MH_FVMLIB: 5598 return eStrataUser; // 0x3u 5599 case MH_CORE: 5600 return eStrataUnknown; // 0x4u 5601 case MH_PRELOAD: 5602 return eStrataRawImage; // 0x5u 5603 case MH_DYLIB: 5604 return eStrataUser; // 0x6u 5605 case MH_DYLINKER: 5606 return eStrataUser; // 0x7u 5607 case MH_BUNDLE: 5608 return eStrataUser; // 0x8u 5609 case MH_DYLIB_STUB: 5610 return eStrataUser; // 0x9u 5611 case MH_DSYM: 5612 return eStrataUnknown; // 0xAu 5613 case MH_KEXT_BUNDLE: 5614 return eStrataKernel; // 0xBu 5615 default: 5616 break; 5617 } 5618 return eStrataUnknown; 5619 } 5620 5621 uint32_t ObjectFileMachO::GetVersion(uint32_t *versions, 5622 uint32_t num_versions) { 5623 ModuleSP module_sp(GetModule()); 5624 if (module_sp) { 5625 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5626 struct dylib_command load_cmd; 5627 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5628 uint32_t version_cmd = 0; 5629 uint64_t version = 0; 5630 uint32_t i; 5631 for (i = 0; i < m_header.ncmds; ++i) { 5632 const lldb::offset_t cmd_offset = offset; 5633 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5634 break; 5635 5636 if (load_cmd.cmd == LC_ID_DYLIB) { 5637 if (version_cmd == 0) { 5638 version_cmd = load_cmd.cmd; 5639 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5640 break; 5641 version = load_cmd.dylib.current_version; 5642 } 5643 break; // Break for now unless there is another more complete version 5644 // number load command in the future. 5645 } 5646 offset = cmd_offset + load_cmd.cmdsize; 5647 } 5648 5649 if (version_cmd == LC_ID_DYLIB) { 5650 if (versions != NULL && num_versions > 0) { 5651 if (num_versions > 0) 5652 versions[0] = (version & 0xFFFF0000ull) >> 16; 5653 if (num_versions > 1) 5654 versions[1] = (version & 0x0000FF00ull) >> 8; 5655 if (num_versions > 2) 5656 versions[2] = (version & 0x000000FFull); 5657 // Fill in an remaining version numbers with invalid values 5658 for (i = 3; i < num_versions; ++i) 5659 versions[i] = UINT32_MAX; 5660 } 5661 // The LC_ID_DYLIB load command has a version with 3 version numbers 5662 // in it, so always return 3 5663 return 3; 5664 } 5665 } 5666 return false; 5667 } 5668 5669 bool ObjectFileMachO::GetArchitecture(ArchSpec &arch) { 5670 ModuleSP module_sp(GetModule()); 5671 if (module_sp) { 5672 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5673 return GetArchitecture(m_header, m_data, 5674 MachHeaderSizeFromMagic(m_header.magic), arch); 5675 } 5676 return false; 5677 } 5678 5679 UUID ObjectFileMachO::GetProcessSharedCacheUUID(Process *process) { 5680 UUID uuid; 5681 if (process && process->GetDynamicLoader()) { 5682 DynamicLoader *dl = process->GetDynamicLoader(); 5683 addr_t load_address; 5684 LazyBool using_shared_cache; 5685 LazyBool private_shared_cache; 5686 dl->GetSharedCacheInformation(load_address, uuid, using_shared_cache, 5687 private_shared_cache); 5688 } 5689 return uuid; 5690 } 5691 5692 UUID ObjectFileMachO::GetLLDBSharedCacheUUID() { 5693 UUID uuid; 5694 #if defined(__APPLE__) && \ 5695 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5696 uint8_t *(*dyld_get_all_image_infos)(void); 5697 dyld_get_all_image_infos = 5698 (uint8_t * (*)())dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5699 if (dyld_get_all_image_infos) { 5700 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5701 if (dyld_all_image_infos_address) { 5702 uint32_t *version = (uint32_t *) 5703 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5704 if (*version >= 13) { 5705 uuid_t *sharedCacheUUID_address = 0; 5706 int wordsize = sizeof(uint8_t *); 5707 if (wordsize == 8) { 5708 sharedCacheUUID_address = 5709 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5710 160); // sharedCacheUUID <mach-o/dyld_images.h> 5711 } else { 5712 sharedCacheUUID_address = 5713 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5714 84); // sharedCacheUUID <mach-o/dyld_images.h> 5715 } 5716 uuid.SetBytes(sharedCacheUUID_address); 5717 } 5718 } 5719 } 5720 #endif 5721 return uuid; 5722 } 5723 5724 uint32_t ObjectFileMachO::GetMinimumOSVersion(uint32_t *versions, 5725 uint32_t num_versions) { 5726 if (m_min_os_versions.empty()) { 5727 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5728 bool success = false; 5729 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5730 const lldb::offset_t load_cmd_offset = offset; 5731 5732 version_min_command lc; 5733 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5734 break; 5735 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5736 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5737 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5738 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5739 if (m_data.GetU32(&offset, &lc.version, 5740 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5741 const uint32_t xxxx = lc.version >> 16; 5742 const uint32_t yy = (lc.version >> 8) & 0xffu; 5743 const uint32_t zz = lc.version & 0xffu; 5744 if (xxxx) { 5745 m_min_os_versions.push_back(xxxx); 5746 m_min_os_versions.push_back(yy); 5747 m_min_os_versions.push_back(zz); 5748 } 5749 success = true; 5750 } 5751 } 5752 offset = load_cmd_offset + lc.cmdsize; 5753 } 5754 5755 if (success == false) { 5756 // Push an invalid value so we don't keep trying to 5757 m_min_os_versions.push_back(UINT32_MAX); 5758 } 5759 } 5760 5761 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) { 5762 if (versions != NULL && num_versions > 0) { 5763 for (size_t i = 0; i < num_versions; ++i) { 5764 if (i < m_min_os_versions.size()) 5765 versions[i] = m_min_os_versions[i]; 5766 else 5767 versions[i] = 0; 5768 } 5769 } 5770 return m_min_os_versions.size(); 5771 } 5772 // Call the superclasses version that will empty out the data 5773 return ObjectFile::GetMinimumOSVersion(versions, num_versions); 5774 } 5775 5776 uint32_t ObjectFileMachO::GetSDKVersion(uint32_t *versions, 5777 uint32_t num_versions) { 5778 if (m_sdk_versions.empty()) { 5779 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5780 bool success = false; 5781 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5782 const lldb::offset_t load_cmd_offset = offset; 5783 5784 version_min_command lc; 5785 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5786 break; 5787 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5788 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5789 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5790 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5791 if (m_data.GetU32(&offset, &lc.version, 5792 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5793 const uint32_t xxxx = lc.sdk >> 16; 5794 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5795 const uint32_t zz = lc.sdk & 0xffu; 5796 if (xxxx) { 5797 m_sdk_versions.push_back(xxxx); 5798 m_sdk_versions.push_back(yy); 5799 m_sdk_versions.push_back(zz); 5800 } 5801 success = true; 5802 } 5803 } 5804 offset = load_cmd_offset + lc.cmdsize; 5805 } 5806 5807 if (success == false) { 5808 // Push an invalid value so we don't keep trying to 5809 m_sdk_versions.push_back(UINT32_MAX); 5810 } 5811 } 5812 5813 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) { 5814 if (versions != NULL && num_versions > 0) { 5815 for (size_t i = 0; i < num_versions; ++i) { 5816 if (i < m_sdk_versions.size()) 5817 versions[i] = m_sdk_versions[i]; 5818 else 5819 versions[i] = 0; 5820 } 5821 } 5822 return m_sdk_versions.size(); 5823 } 5824 // Call the superclasses version that will empty out the data 5825 return ObjectFile::GetSDKVersion(versions, num_versions); 5826 } 5827 5828 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5829 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5830 } 5831 5832 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5833 return m_allow_assembly_emulation_unwind_plans; 5834 } 5835 5836 //------------------------------------------------------------------ 5837 // PluginInterface protocol 5838 //------------------------------------------------------------------ 5839 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5840 return GetPluginNameStatic(); 5841 } 5842 5843 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5844 5845 Section *ObjectFileMachO::GetMachHeaderSection() { 5846 // Find the first address of the mach header which is the first non-zero 5847 // file sized section whose file offset is zero. This is the base file address 5848 // of the mach-o file which can be subtracted from the vmaddr of the other 5849 // segments found in memory and added to the load address 5850 ModuleSP module_sp = GetModule(); 5851 if (module_sp) { 5852 SectionList *section_list = GetSectionList(); 5853 if (section_list) { 5854 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5855 const size_t num_sections = section_list->GetSize(); 5856 5857 for (size_t sect_idx = 0; sect_idx < num_sections && 5858 mach_base_file_addr == LLDB_INVALID_ADDRESS; 5859 ++sect_idx) { 5860 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5861 if (section && section->GetFileSize() > 0 && 5862 section->GetFileOffset() == 0 && 5863 section->IsThreadSpecific() == false && 5864 module_sp.get() == section->GetModule().get()) { 5865 return section; 5866 } 5867 } 5868 } 5869 } 5870 return nullptr; 5871 } 5872 5873 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 5874 lldb::addr_t mach_header_load_address, const Section *mach_header_section, 5875 const Section *section) { 5876 ModuleSP module_sp = GetModule(); 5877 if (module_sp && mach_header_section && section && 5878 mach_header_load_address != LLDB_INVALID_ADDRESS) { 5879 lldb::addr_t mach_header_file_addr = mach_header_section->GetFileAddress(); 5880 if (mach_header_file_addr != LLDB_INVALID_ADDRESS) { 5881 if (section && section->GetFileSize() > 0 && 5882 section->IsThreadSpecific() == false && 5883 module_sp.get() == section->GetModule().get()) { 5884 // Ignore __LINKEDIT and __DWARF segments 5885 if (section->GetName() == GetSegmentNameLINKEDIT()) { 5886 // Only map __LINKEDIT if we have an in memory image and this isn't 5887 // a kernel binary like a kext or mach_kernel. 5888 const bool is_memory_image = (bool)m_process_wp.lock(); 5889 const Strata strata = GetStrata(); 5890 if (is_memory_image == false || strata == eStrataKernel) 5891 return LLDB_INVALID_ADDRESS; 5892 } 5893 return section->GetFileAddress() - mach_header_file_addr + 5894 mach_header_load_address; 5895 } 5896 } 5897 } 5898 return LLDB_INVALID_ADDRESS; 5899 } 5900 5901 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 5902 bool value_is_offset) { 5903 ModuleSP module_sp = GetModule(); 5904 if (module_sp) { 5905 size_t num_loaded_sections = 0; 5906 SectionList *section_list = GetSectionList(); 5907 if (section_list) { 5908 const size_t num_sections = section_list->GetSize(); 5909 5910 if (value_is_offset) { 5911 // "value" is an offset to apply to each top level segment 5912 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5913 // Iterate through the object file sections to find all 5914 // of the sections that size on disk (to avoid __PAGEZERO) 5915 // and load them 5916 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5917 if (section_sp && section_sp->GetFileSize() > 0 && 5918 section_sp->IsThreadSpecific() == false && 5919 module_sp.get() == section_sp->GetModule().get()) { 5920 // Ignore __LINKEDIT and __DWARF segments 5921 if (section_sp->GetName() == GetSegmentNameLINKEDIT()) { 5922 // Only map __LINKEDIT if we have an in memory image and this 5923 // isn't 5924 // a kernel binary like a kext or mach_kernel. 5925 const bool is_memory_image = (bool)m_process_wp.lock(); 5926 const Strata strata = GetStrata(); 5927 if (is_memory_image == false || strata == eStrataKernel) 5928 continue; 5929 } 5930 if (target.GetSectionLoadList().SetSectionLoadAddress( 5931 section_sp, section_sp->GetFileAddress() + value)) 5932 ++num_loaded_sections; 5933 } 5934 } 5935 } else { 5936 // "value" is the new base address of the mach_header, adjust each 5937 // section accordingly 5938 5939 Section *mach_header_section = GetMachHeaderSection(); 5940 if (mach_header_section) { 5941 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5942 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5943 5944 lldb::addr_t section_load_addr = 5945 CalculateSectionLoadAddressForMemoryImage( 5946 value, mach_header_section, section_sp.get()); 5947 if (section_load_addr != LLDB_INVALID_ADDRESS) { 5948 if (target.GetSectionLoadList().SetSectionLoadAddress( 5949 section_sp, section_load_addr)) 5950 ++num_loaded_sections; 5951 } 5952 } 5953 } 5954 } 5955 } 5956 return num_loaded_sections > 0; 5957 } 5958 return false; 5959 } 5960 5961 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 5962 const FileSpec &outfile, Status &error) { 5963 if (process_sp) { 5964 Target &target = process_sp->GetTarget(); 5965 const ArchSpec target_arch = target.GetArchitecture(); 5966 const llvm::Triple &target_triple = target_arch.GetTriple(); 5967 if (target_triple.getVendor() == llvm::Triple::Apple && 5968 (target_triple.getOS() == llvm::Triple::MacOSX || 5969 target_triple.getOS() == llvm::Triple::IOS || 5970 target_triple.getOS() == llvm::Triple::WatchOS || 5971 target_triple.getOS() == llvm::Triple::TvOS)) { 5972 bool make_core = false; 5973 switch (target_arch.GetMachine()) { 5974 case llvm::Triple::aarch64: 5975 case llvm::Triple::arm: 5976 case llvm::Triple::thumb: 5977 case llvm::Triple::x86: 5978 case llvm::Triple::x86_64: 5979 make_core = true; 5980 break; 5981 default: 5982 error.SetErrorStringWithFormat("unsupported core architecture: %s", 5983 target_triple.str().c_str()); 5984 break; 5985 } 5986 5987 if (make_core) { 5988 std::vector<segment_command_64> segment_load_commands; 5989 // uint32_t range_info_idx = 0; 5990 MemoryRegionInfo range_info; 5991 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 5992 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 5993 const ByteOrder byte_order = target_arch.GetByteOrder(); 5994 if (range_error.Success()) { 5995 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 5996 const addr_t addr = range_info.GetRange().GetRangeBase(); 5997 const addr_t size = range_info.GetRange().GetByteSize(); 5998 5999 if (size == 0) 6000 break; 6001 6002 // Calculate correct protections 6003 uint32_t prot = 0; 6004 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6005 prot |= VM_PROT_READ; 6006 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6007 prot |= VM_PROT_WRITE; 6008 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6009 prot |= VM_PROT_EXECUTE; 6010 6011 // printf ("[%3u] [0x%16.16" PRIx64 " - 6012 // 0x%16.16" PRIx64 ") %c%c%c\n", 6013 // range_info_idx, 6014 // addr, 6015 // size, 6016 // (prot & VM_PROT_READ ) ? 'r' : 6017 // '-', 6018 // (prot & VM_PROT_WRITE ) ? 'w' : 6019 // '-', 6020 // (prot & VM_PROT_EXECUTE) ? 'x' : 6021 // '-'); 6022 6023 if (prot != 0) { 6024 uint32_t cmd_type = LC_SEGMENT_64; 6025 uint32_t segment_size = sizeof(segment_command_64); 6026 if (addr_byte_size == 4) { 6027 cmd_type = LC_SEGMENT; 6028 segment_size = sizeof(segment_command); 6029 } 6030 segment_command_64 segment = { 6031 cmd_type, // uint32_t cmd; 6032 segment_size, // uint32_t cmdsize; 6033 {0}, // char segname[16]; 6034 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6035 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6036 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6037 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6038 prot, // uint32_t maxprot; 6039 prot, // uint32_t initprot; 6040 0, // uint32_t nsects; 6041 0}; // uint32_t flags; 6042 segment_load_commands.push_back(segment); 6043 } else { 6044 // No protections and a size of 1 used to be returned from old 6045 // debugservers when we asked about a region that was past the 6046 // last memory region and it indicates the end... 6047 if (size == 1) 6048 break; 6049 } 6050 6051 range_error = process_sp->GetMemoryRegionInfo( 6052 range_info.GetRange().GetRangeEnd(), range_info); 6053 if (range_error.Fail()) 6054 break; 6055 } 6056 6057 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6058 6059 mach_header_64 mach_header; 6060 if (addr_byte_size == 8) { 6061 mach_header.magic = MH_MAGIC_64; 6062 } else { 6063 mach_header.magic = MH_MAGIC; 6064 } 6065 mach_header.cputype = target_arch.GetMachOCPUType(); 6066 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6067 mach_header.filetype = MH_CORE; 6068 mach_header.ncmds = segment_load_commands.size(); 6069 mach_header.flags = 0; 6070 mach_header.reserved = 0; 6071 ThreadList &thread_list = process_sp->GetThreadList(); 6072 const uint32_t num_threads = thread_list.GetSize(); 6073 6074 // Make an array of LC_THREAD data items. Each one contains 6075 // the contents of the LC_THREAD load command. The data doesn't 6076 // contain the load command + load command size, we will 6077 // add the load command and load command size as we emit the data. 6078 std::vector<StreamString> LC_THREAD_datas(num_threads); 6079 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6080 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6081 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6082 LC_THREAD_data.SetByteOrder(byte_order); 6083 } 6084 for (uint32_t thread_idx = 0; thread_idx < num_threads; 6085 ++thread_idx) { 6086 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6087 if (thread_sp) { 6088 switch (mach_header.cputype) { 6089 case llvm::MachO::CPU_TYPE_ARM64: 6090 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6091 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6092 break; 6093 6094 case llvm::MachO::CPU_TYPE_ARM: 6095 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6096 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6097 break; 6098 6099 case llvm::MachO::CPU_TYPE_I386: 6100 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6101 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6102 break; 6103 6104 case llvm::MachO::CPU_TYPE_X86_64: 6105 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6106 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6107 break; 6108 } 6109 } 6110 } 6111 6112 // The size of the load command is the size of the segments... 6113 if (addr_byte_size == 8) { 6114 mach_header.sizeofcmds = segment_load_commands.size() * 6115 sizeof(struct segment_command_64); 6116 } else { 6117 mach_header.sizeofcmds = 6118 segment_load_commands.size() * sizeof(struct segment_command); 6119 } 6120 6121 // and the size of all LC_THREAD load command 6122 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6123 ++mach_header.ncmds; 6124 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6125 } 6126 6127 printf("mach_header: 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x " 6128 "0x%8.8x 0x%8.8x\n", 6129 mach_header.magic, mach_header.cputype, mach_header.cpusubtype, 6130 mach_header.filetype, mach_header.ncmds, 6131 mach_header.sizeofcmds, mach_header.flags, 6132 mach_header.reserved); 6133 6134 // Write the mach header 6135 buffer.PutHex32(mach_header.magic); 6136 buffer.PutHex32(mach_header.cputype); 6137 buffer.PutHex32(mach_header.cpusubtype); 6138 buffer.PutHex32(mach_header.filetype); 6139 buffer.PutHex32(mach_header.ncmds); 6140 buffer.PutHex32(mach_header.sizeofcmds); 6141 buffer.PutHex32(mach_header.flags); 6142 if (addr_byte_size == 8) { 6143 buffer.PutHex32(mach_header.reserved); 6144 } 6145 6146 // Skip the mach header and all load commands and align to the next 6147 // 0x1000 byte boundary 6148 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6149 if (file_offset & 0x00000fff) { 6150 file_offset += 0x00001000ull; 6151 file_offset &= (~0x00001000ull + 1); 6152 } 6153 6154 for (auto &segment : segment_load_commands) { 6155 segment.fileoff = file_offset; 6156 file_offset += segment.filesize; 6157 } 6158 6159 // Write out all of the LC_THREAD load commands 6160 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6161 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6162 buffer.PutHex32(LC_THREAD); 6163 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6164 buffer.Write(LC_THREAD_data.GetString().data(), 6165 LC_THREAD_data_size); 6166 } 6167 6168 // Write out all of the segment load commands 6169 for (const auto &segment : segment_load_commands) { 6170 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6171 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6172 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6173 segment.cmd, segment.cmdsize, segment.vmaddr, 6174 segment.vmaddr + segment.vmsize, segment.fileoff, 6175 segment.filesize, segment.maxprot, segment.initprot, 6176 segment.nsects, segment.flags); 6177 6178 buffer.PutHex32(segment.cmd); 6179 buffer.PutHex32(segment.cmdsize); 6180 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6181 if (addr_byte_size == 8) { 6182 buffer.PutHex64(segment.vmaddr); 6183 buffer.PutHex64(segment.vmsize); 6184 buffer.PutHex64(segment.fileoff); 6185 buffer.PutHex64(segment.filesize); 6186 } else { 6187 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6188 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6189 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6190 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6191 } 6192 buffer.PutHex32(segment.maxprot); 6193 buffer.PutHex32(segment.initprot); 6194 buffer.PutHex32(segment.nsects); 6195 buffer.PutHex32(segment.flags); 6196 } 6197 6198 File core_file; 6199 std::string core_file_path(outfile.GetPath()); 6200 error = core_file.Open(core_file_path.c_str(), 6201 File::eOpenOptionWrite | 6202 File::eOpenOptionTruncate | 6203 File::eOpenOptionCanCreate); 6204 if (error.Success()) { 6205 // Read 1 page at a time 6206 uint8_t bytes[0x1000]; 6207 // Write the mach header and load commands out to the core file 6208 size_t bytes_written = buffer.GetString().size(); 6209 error = core_file.Write(buffer.GetString().data(), bytes_written); 6210 if (error.Success()) { 6211 // Now write the file data for all memory segments in the process 6212 for (const auto &segment : segment_load_commands) { 6213 if (core_file.SeekFromStart(segment.fileoff) == -1) { 6214 error.SetErrorStringWithFormat( 6215 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6216 segment.fileoff, core_file_path.c_str()); 6217 break; 6218 } 6219 6220 printf("Saving %" PRId64 6221 " bytes of data for memory region at 0x%" PRIx64 "\n", 6222 segment.vmsize, segment.vmaddr); 6223 addr_t bytes_left = segment.vmsize; 6224 addr_t addr = segment.vmaddr; 6225 Status memory_read_error; 6226 while (bytes_left > 0 && error.Success()) { 6227 const size_t bytes_to_read = 6228 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6229 const size_t bytes_read = process_sp->ReadMemory( 6230 addr, bytes, bytes_to_read, memory_read_error); 6231 if (bytes_read == bytes_to_read) { 6232 size_t bytes_written = bytes_read; 6233 error = core_file.Write(bytes, bytes_written); 6234 bytes_left -= bytes_read; 6235 addr += bytes_read; 6236 } else { 6237 // Some pages within regions are not readable, those 6238 // should be zero filled 6239 memset(bytes, 0, bytes_to_read); 6240 size_t bytes_written = bytes_to_read; 6241 error = core_file.Write(bytes, bytes_written); 6242 bytes_left -= bytes_to_read; 6243 addr += bytes_to_read; 6244 } 6245 } 6246 } 6247 } 6248 } 6249 } else { 6250 error.SetErrorString( 6251 "process doesn't support getting memory region info"); 6252 } 6253 } 6254 return true; // This is the right plug to handle saving core files for 6255 // this process 6256 } 6257 } 6258 return false; 6259 } 6260