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 flavors... 165 break; 166 167 default: 168 done = true; 169 break; 170 } 171 } 172 } 173 } 174 175 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 176 const char *alt_name, size_t reg_byte_size, 177 Stream &data) { 178 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 179 if (reg_info == NULL) 180 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 181 if (reg_info) { 182 lldb_private::RegisterValue reg_value; 183 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 184 if (reg_info->byte_size >= reg_byte_size) 185 data.Write(reg_value.GetBytes(), reg_byte_size); 186 else { 187 data.Write(reg_value.GetBytes(), reg_info->byte_size); 188 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 189 ++i) 190 data.PutChar(0); 191 } 192 return reg_byte_size; 193 } 194 } 195 // Just write zeros if all else fails 196 for (size_t i = 0; i < reg_byte_size; ++i) 197 data.PutChar(0); 198 return reg_byte_size; 199 } 200 201 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 202 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 203 if (reg_ctx_sp) { 204 RegisterContext *reg_ctx = reg_ctx_sp.get(); 205 206 data.PutHex32(GPRRegSet); // Flavor 207 data.PutHex32(GPRWordCount); 208 WriteRegister(reg_ctx, "rax", NULL, 8, data); 209 WriteRegister(reg_ctx, "rbx", NULL, 8, data); 210 WriteRegister(reg_ctx, "rcx", NULL, 8, data); 211 WriteRegister(reg_ctx, "rdx", NULL, 8, data); 212 WriteRegister(reg_ctx, "rdi", NULL, 8, data); 213 WriteRegister(reg_ctx, "rsi", NULL, 8, data); 214 WriteRegister(reg_ctx, "rbp", NULL, 8, data); 215 WriteRegister(reg_ctx, "rsp", NULL, 8, data); 216 WriteRegister(reg_ctx, "r8", NULL, 8, data); 217 WriteRegister(reg_ctx, "r9", NULL, 8, data); 218 WriteRegister(reg_ctx, "r10", NULL, 8, data); 219 WriteRegister(reg_ctx, "r11", NULL, 8, data); 220 WriteRegister(reg_ctx, "r12", NULL, 8, data); 221 WriteRegister(reg_ctx, "r13", NULL, 8, data); 222 WriteRegister(reg_ctx, "r14", NULL, 8, data); 223 WriteRegister(reg_ctx, "r15", NULL, 8, data); 224 WriteRegister(reg_ctx, "rip", NULL, 8, data); 225 WriteRegister(reg_ctx, "rflags", NULL, 8, data); 226 WriteRegister(reg_ctx, "cs", NULL, 8, data); 227 WriteRegister(reg_ctx, "fs", NULL, 8, data); 228 WriteRegister(reg_ctx, "gs", NULL, 8, data); 229 230 // // Write out the FPU registers 231 // const size_t fpu_byte_size = sizeof(FPU); 232 // size_t bytes_written = 0; 233 // data.PutHex32 (FPURegSet); 234 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 235 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 236 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 237 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 238 // data); // uint16_t fcw; // "fctrl" 239 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 240 // data); // uint16_t fsw; // "fstat" 241 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 242 // data); // uint8_t ftw; // "ftag" 243 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 244 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 245 // data); // uint16_t fop; // "fop" 246 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 247 // data); // uint32_t ip; // "fioff" 248 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 249 // data); // uint16_t cs; // "fiseg" 250 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 251 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 252 // data); // uint32_t dp; // "fooff" 253 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 254 // data); // uint16_t ds; // "foseg" 255 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 256 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 257 // data); // uint32_t mxcsr; 258 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 259 // 4, data);// uint32_t mxcsrmask; 260 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 261 // sizeof(MMSReg), data); 262 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 263 // sizeof(MMSReg), data); 264 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 265 // sizeof(MMSReg), data); 266 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 267 // sizeof(MMSReg), data); 268 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 269 // sizeof(MMSReg), data); 270 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 271 // sizeof(MMSReg), data); 272 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 273 // sizeof(MMSReg), data); 274 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 275 // sizeof(MMSReg), data); 276 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 277 // sizeof(XMMReg), data); 278 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 279 // sizeof(XMMReg), data); 280 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 281 // sizeof(XMMReg), data); 282 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 283 // sizeof(XMMReg), data); 284 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 285 // sizeof(XMMReg), data); 286 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 287 // sizeof(XMMReg), data); 288 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 289 // sizeof(XMMReg), data); 290 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 291 // sizeof(XMMReg), data); 292 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 293 // sizeof(XMMReg), data); 294 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 295 // sizeof(XMMReg), data); 296 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 297 // sizeof(XMMReg), data); 298 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 299 // sizeof(XMMReg), data); 300 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 301 // sizeof(XMMReg), data); 302 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 303 // sizeof(XMMReg), data); 304 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 305 // sizeof(XMMReg), data); 306 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 307 // sizeof(XMMReg), data); 308 // 309 // // Fill rest with zeros 310 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 311 // i) 312 // data.PutChar(0); 313 314 // Write out the EXC registers 315 data.PutHex32(EXCRegSet); 316 data.PutHex32(EXCWordCount); 317 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 318 WriteRegister(reg_ctx, "err", NULL, 4, data); 319 WriteRegister(reg_ctx, "faultvaddr", NULL, 8, data); 320 return true; 321 } 322 return false; 323 } 324 325 protected: 326 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 327 328 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 329 330 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 331 332 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 333 return 0; 334 } 335 336 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 337 return 0; 338 } 339 340 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 341 return 0; 342 } 343 }; 344 345 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 346 public: 347 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 348 const DataExtractor &data) 349 : RegisterContextDarwin_i386(thread, 0) { 350 SetRegisterDataFrom_LC_THREAD(data); 351 } 352 353 void InvalidateAllRegisters() override { 354 // Do nothing... registers are always valid... 355 } 356 357 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 358 lldb::offset_t offset = 0; 359 SetError(GPRRegSet, Read, -1); 360 SetError(FPURegSet, Read, -1); 361 SetError(EXCRegSet, Read, -1); 362 bool done = false; 363 364 while (!done) { 365 int flavor = data.GetU32(&offset); 366 if (flavor == 0) 367 done = true; 368 else { 369 uint32_t i; 370 uint32_t count = data.GetU32(&offset); 371 switch (flavor) { 372 case GPRRegSet: 373 for (i = 0; i < count; ++i) 374 (&gpr.eax)[i] = data.GetU32(&offset); 375 SetError(GPRRegSet, Read, 0); 376 done = true; 377 378 break; 379 case FPURegSet: 380 // TODO: fill in FPU regs.... 381 // SetError (FPURegSet, Read, -1); 382 done = true; 383 384 break; 385 case EXCRegSet: 386 exc.trapno = data.GetU32(&offset); 387 exc.err = data.GetU32(&offset); 388 exc.faultvaddr = data.GetU32(&offset); 389 SetError(EXCRegSet, Read, 0); 390 done = true; 391 break; 392 case 7: 393 case 8: 394 case 9: 395 // fancy flavors that encapsulate of the above flavors... 396 break; 397 398 default: 399 done = true; 400 break; 401 } 402 } 403 } 404 } 405 406 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 407 const char *alt_name, size_t reg_byte_size, 408 Stream &data) { 409 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 410 if (reg_info == NULL) 411 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 412 if (reg_info) { 413 lldb_private::RegisterValue reg_value; 414 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 415 if (reg_info->byte_size >= reg_byte_size) 416 data.Write(reg_value.GetBytes(), reg_byte_size); 417 else { 418 data.Write(reg_value.GetBytes(), reg_info->byte_size); 419 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 420 ++i) 421 data.PutChar(0); 422 } 423 return reg_byte_size; 424 } 425 } 426 // Just write zeros if all else fails 427 for (size_t i = 0; i < reg_byte_size; ++i) 428 data.PutChar(0); 429 return reg_byte_size; 430 } 431 432 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 433 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 434 if (reg_ctx_sp) { 435 RegisterContext *reg_ctx = reg_ctx_sp.get(); 436 437 data.PutHex32(GPRRegSet); // Flavor 438 data.PutHex32(GPRWordCount); 439 WriteRegister(reg_ctx, "eax", NULL, 4, data); 440 WriteRegister(reg_ctx, "ebx", NULL, 4, data); 441 WriteRegister(reg_ctx, "ecx", NULL, 4, data); 442 WriteRegister(reg_ctx, "edx", NULL, 4, data); 443 WriteRegister(reg_ctx, "edi", NULL, 4, data); 444 WriteRegister(reg_ctx, "esi", NULL, 4, data); 445 WriteRegister(reg_ctx, "ebp", NULL, 4, data); 446 WriteRegister(reg_ctx, "esp", NULL, 4, data); 447 WriteRegister(reg_ctx, "ss", NULL, 4, data); 448 WriteRegister(reg_ctx, "eflags", NULL, 4, data); 449 WriteRegister(reg_ctx, "eip", NULL, 4, data); 450 WriteRegister(reg_ctx, "cs", NULL, 4, data); 451 WriteRegister(reg_ctx, "ds", NULL, 4, data); 452 WriteRegister(reg_ctx, "es", NULL, 4, data); 453 WriteRegister(reg_ctx, "fs", NULL, 4, data); 454 WriteRegister(reg_ctx, "gs", NULL, 4, data); 455 456 // Write out the EXC registers 457 data.PutHex32(EXCRegSet); 458 data.PutHex32(EXCWordCount); 459 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 460 WriteRegister(reg_ctx, "err", NULL, 4, data); 461 WriteRegister(reg_ctx, "faultvaddr", NULL, 4, data); 462 return true; 463 } 464 return false; 465 } 466 467 protected: 468 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 469 470 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 471 472 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 473 474 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 475 return 0; 476 } 477 478 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 479 return 0; 480 } 481 482 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 483 return 0; 484 } 485 }; 486 487 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 488 public: 489 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 490 const DataExtractor &data) 491 : RegisterContextDarwin_arm(thread, 0) { 492 SetRegisterDataFrom_LC_THREAD(data); 493 } 494 495 void InvalidateAllRegisters() override { 496 // Do nothing... registers are always valid... 497 } 498 499 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 500 lldb::offset_t offset = 0; 501 SetError(GPRRegSet, Read, -1); 502 SetError(FPURegSet, Read, -1); 503 SetError(EXCRegSet, Read, -1); 504 bool done = false; 505 506 while (!done) { 507 int flavor = data.GetU32(&offset); 508 uint32_t count = data.GetU32(&offset); 509 lldb::offset_t next_thread_state = offset + (count * 4); 510 switch (flavor) { 511 case GPRAltRegSet: 512 case GPRRegSet: 513 for (uint32_t i = 0; i < count; ++i) { 514 gpr.r[i] = data.GetU32(&offset); 515 } 516 517 // Note that gpr.cpsr is also copied by the above loop; this loop 518 // technically extends one element past the end of the gpr.r[] array. 519 520 SetError(GPRRegSet, Read, 0); 521 offset = next_thread_state; 522 break; 523 524 case FPURegSet: { 525 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 526 const int fpu_reg_buf_size = sizeof(fpu.floats); 527 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 528 fpu_reg_buf) == fpu_reg_buf_size) { 529 offset += fpu_reg_buf_size; 530 fpu.fpscr = data.GetU32(&offset); 531 SetError(FPURegSet, Read, 0); 532 } else { 533 done = true; 534 } 535 } 536 offset = next_thread_state; 537 break; 538 539 case EXCRegSet: 540 if (count == 3) { 541 exc.exception = data.GetU32(&offset); 542 exc.fsr = data.GetU32(&offset); 543 exc.far = data.GetU32(&offset); 544 SetError(EXCRegSet, Read, 0); 545 } 546 done = true; 547 offset = next_thread_state; 548 break; 549 550 // Unknown register set flavor, stop trying to parse. 551 default: 552 done = true; 553 } 554 } 555 } 556 557 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 558 const char *alt_name, size_t reg_byte_size, 559 Stream &data) { 560 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 561 if (reg_info == NULL) 562 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 563 if (reg_info) { 564 lldb_private::RegisterValue reg_value; 565 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 566 if (reg_info->byte_size >= reg_byte_size) 567 data.Write(reg_value.GetBytes(), reg_byte_size); 568 else { 569 data.Write(reg_value.GetBytes(), reg_info->byte_size); 570 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 571 ++i) 572 data.PutChar(0); 573 } 574 return reg_byte_size; 575 } 576 } 577 // Just write zeros if all else fails 578 for (size_t i = 0; i < reg_byte_size; ++i) 579 data.PutChar(0); 580 return reg_byte_size; 581 } 582 583 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 584 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 585 if (reg_ctx_sp) { 586 RegisterContext *reg_ctx = reg_ctx_sp.get(); 587 588 data.PutHex32(GPRRegSet); // Flavor 589 data.PutHex32(GPRWordCount); 590 WriteRegister(reg_ctx, "r0", NULL, 4, data); 591 WriteRegister(reg_ctx, "r1", NULL, 4, data); 592 WriteRegister(reg_ctx, "r2", NULL, 4, data); 593 WriteRegister(reg_ctx, "r3", NULL, 4, data); 594 WriteRegister(reg_ctx, "r4", NULL, 4, data); 595 WriteRegister(reg_ctx, "r5", NULL, 4, data); 596 WriteRegister(reg_ctx, "r6", NULL, 4, data); 597 WriteRegister(reg_ctx, "r7", NULL, 4, data); 598 WriteRegister(reg_ctx, "r8", NULL, 4, data); 599 WriteRegister(reg_ctx, "r9", NULL, 4, data); 600 WriteRegister(reg_ctx, "r10", NULL, 4, data); 601 WriteRegister(reg_ctx, "r11", NULL, 4, data); 602 WriteRegister(reg_ctx, "r12", NULL, 4, data); 603 WriteRegister(reg_ctx, "sp", NULL, 4, data); 604 WriteRegister(reg_ctx, "lr", NULL, 4, data); 605 WriteRegister(reg_ctx, "pc", NULL, 4, data); 606 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 607 608 // Write out the EXC registers 609 // data.PutHex32 (EXCRegSet); 610 // data.PutHex32 (EXCWordCount); 611 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 612 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 613 // WriteRegister (reg_ctx, "far", NULL, 4, data); 614 return true; 615 } 616 return false; 617 } 618 619 protected: 620 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 621 622 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 623 624 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 625 626 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 627 628 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 629 return 0; 630 } 631 632 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 633 return 0; 634 } 635 636 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 637 return 0; 638 } 639 640 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 641 return -1; 642 } 643 }; 644 645 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 646 public: 647 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 648 const DataExtractor &data) 649 : RegisterContextDarwin_arm64(thread, 0) { 650 SetRegisterDataFrom_LC_THREAD(data); 651 } 652 653 void InvalidateAllRegisters() override { 654 // Do nothing... registers are always valid... 655 } 656 657 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 658 lldb::offset_t offset = 0; 659 SetError(GPRRegSet, Read, -1); 660 SetError(FPURegSet, Read, -1); 661 SetError(EXCRegSet, Read, -1); 662 bool done = false; 663 while (!done) { 664 int flavor = data.GetU32(&offset); 665 uint32_t count = data.GetU32(&offset); 666 lldb::offset_t next_thread_state = offset + (count * 4); 667 switch (flavor) { 668 case GPRRegSet: 669 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 670 // 32-bit register) 671 if (count >= (33 * 2) + 1) { 672 for (uint32_t i = 0; i < 29; ++i) 673 gpr.x[i] = data.GetU64(&offset); 674 gpr.fp = data.GetU64(&offset); 675 gpr.lr = data.GetU64(&offset); 676 gpr.sp = data.GetU64(&offset); 677 gpr.pc = data.GetU64(&offset); 678 gpr.cpsr = data.GetU32(&offset); 679 SetError(GPRRegSet, Read, 0); 680 } 681 offset = next_thread_state; 682 break; 683 case FPURegSet: { 684 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 685 const int fpu_reg_buf_size = sizeof(fpu); 686 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 687 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 688 fpu_reg_buf) == fpu_reg_buf_size) { 689 SetError(FPURegSet, Read, 0); 690 } else { 691 done = true; 692 } 693 } 694 offset = next_thread_state; 695 break; 696 case EXCRegSet: 697 if (count == 4) { 698 exc.far = data.GetU64(&offset); 699 exc.esr = data.GetU32(&offset); 700 exc.exception = data.GetU32(&offset); 701 SetError(EXCRegSet, Read, 0); 702 } 703 offset = next_thread_state; 704 break; 705 default: 706 done = true; 707 break; 708 } 709 } 710 } 711 712 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 713 const char *alt_name, size_t reg_byte_size, 714 Stream &data) { 715 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 716 if (reg_info == NULL) 717 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 718 if (reg_info) { 719 lldb_private::RegisterValue reg_value; 720 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 721 if (reg_info->byte_size >= reg_byte_size) 722 data.Write(reg_value.GetBytes(), reg_byte_size); 723 else { 724 data.Write(reg_value.GetBytes(), reg_info->byte_size); 725 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 726 ++i) 727 data.PutChar(0); 728 } 729 return reg_byte_size; 730 } 731 } 732 // Just write zeros if all else fails 733 for (size_t i = 0; i < reg_byte_size; ++i) 734 data.PutChar(0); 735 return reg_byte_size; 736 } 737 738 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 739 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 740 if (reg_ctx_sp) { 741 RegisterContext *reg_ctx = reg_ctx_sp.get(); 742 743 data.PutHex32(GPRRegSet); // Flavor 744 data.PutHex32(GPRWordCount); 745 WriteRegister(reg_ctx, "x0", NULL, 8, data); 746 WriteRegister(reg_ctx, "x1", NULL, 8, data); 747 WriteRegister(reg_ctx, "x2", NULL, 8, data); 748 WriteRegister(reg_ctx, "x3", NULL, 8, data); 749 WriteRegister(reg_ctx, "x4", NULL, 8, data); 750 WriteRegister(reg_ctx, "x5", NULL, 8, data); 751 WriteRegister(reg_ctx, "x6", NULL, 8, data); 752 WriteRegister(reg_ctx, "x7", NULL, 8, data); 753 WriteRegister(reg_ctx, "x8", NULL, 8, data); 754 WriteRegister(reg_ctx, "x9", NULL, 8, data); 755 WriteRegister(reg_ctx, "x10", NULL, 8, data); 756 WriteRegister(reg_ctx, "x11", NULL, 8, data); 757 WriteRegister(reg_ctx, "x12", NULL, 8, data); 758 WriteRegister(reg_ctx, "x13", NULL, 8, data); 759 WriteRegister(reg_ctx, "x14", NULL, 8, data); 760 WriteRegister(reg_ctx, "x15", NULL, 8, data); 761 WriteRegister(reg_ctx, "x16", NULL, 8, data); 762 WriteRegister(reg_ctx, "x17", NULL, 8, data); 763 WriteRegister(reg_ctx, "x18", NULL, 8, data); 764 WriteRegister(reg_ctx, "x19", NULL, 8, data); 765 WriteRegister(reg_ctx, "x20", NULL, 8, data); 766 WriteRegister(reg_ctx, "x21", NULL, 8, data); 767 WriteRegister(reg_ctx, "x22", NULL, 8, data); 768 WriteRegister(reg_ctx, "x23", NULL, 8, data); 769 WriteRegister(reg_ctx, "x24", NULL, 8, data); 770 WriteRegister(reg_ctx, "x25", NULL, 8, data); 771 WriteRegister(reg_ctx, "x26", NULL, 8, data); 772 WriteRegister(reg_ctx, "x27", NULL, 8, data); 773 WriteRegister(reg_ctx, "x28", NULL, 8, data); 774 WriteRegister(reg_ctx, "fp", NULL, 8, data); 775 WriteRegister(reg_ctx, "lr", NULL, 8, data); 776 WriteRegister(reg_ctx, "sp", NULL, 8, data); 777 WriteRegister(reg_ctx, "pc", NULL, 8, data); 778 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 779 780 // Write out the EXC registers 781 // data.PutHex32 (EXCRegSet); 782 // data.PutHex32 (EXCWordCount); 783 // WriteRegister (reg_ctx, "far", NULL, 8, data); 784 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 785 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 786 return true; 787 } 788 return false; 789 } 790 791 protected: 792 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 793 794 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 795 796 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 797 798 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 799 800 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 801 return 0; 802 } 803 804 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 805 return 0; 806 } 807 808 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 809 return 0; 810 } 811 812 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 813 return -1; 814 } 815 }; 816 817 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 818 switch (magic) { 819 case MH_MAGIC: 820 case MH_CIGAM: 821 return sizeof(struct mach_header); 822 823 case MH_MAGIC_64: 824 case MH_CIGAM_64: 825 return sizeof(struct mach_header_64); 826 break; 827 828 default: 829 break; 830 } 831 return 0; 832 } 833 834 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 835 836 void ObjectFileMachO::Initialize() { 837 PluginManager::RegisterPlugin( 838 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 839 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 840 } 841 842 void ObjectFileMachO::Terminate() { 843 PluginManager::UnregisterPlugin(CreateInstance); 844 } 845 846 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 847 static ConstString g_name("mach-o"); 848 return g_name; 849 } 850 851 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 852 return "Mach-o object file reader (32 and 64 bit)"; 853 } 854 855 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 856 DataBufferSP &data_sp, 857 lldb::offset_t data_offset, 858 const FileSpec *file, 859 lldb::offset_t file_offset, 860 lldb::offset_t length) { 861 if (!data_sp) { 862 data_sp = MapFileData(*file, length, file_offset); 863 if (!data_sp) 864 return nullptr; 865 data_offset = 0; 866 } 867 868 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 869 return nullptr; 870 871 // Update the data to contain the entire file if it doesn't already 872 if (data_sp->GetByteSize() < length) { 873 data_sp = MapFileData(*file, length, file_offset); 874 if (!data_sp) 875 return nullptr; 876 data_offset = 0; 877 } 878 auto objfile_ap = llvm::make_unique<ObjectFileMachO>( 879 module_sp, data_sp, data_offset, file, file_offset, length); 880 if (!objfile_ap || !objfile_ap->ParseHeader()) 881 return nullptr; 882 883 return objfile_ap.release(); 884 } 885 886 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 887 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 888 const ProcessSP &process_sp, lldb::addr_t header_addr) { 889 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 890 std::unique_ptr<ObjectFile> objfile_ap( 891 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 892 if (objfile_ap.get() && objfile_ap->ParseHeader()) 893 return objfile_ap.release(); 894 } 895 return NULL; 896 } 897 898 size_t ObjectFileMachO::GetModuleSpecifications( 899 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 900 lldb::offset_t data_offset, lldb::offset_t file_offset, 901 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 902 const size_t initial_count = specs.GetSize(); 903 904 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 905 DataExtractor data; 906 data.SetData(data_sp); 907 llvm::MachO::mach_header header; 908 if (ParseHeader(data, &data_offset, header)) { 909 size_t header_and_load_cmds = 910 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 911 if (header_and_load_cmds >= data_sp->GetByteSize()) { 912 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 913 data.SetData(data_sp); 914 data_offset = MachHeaderSizeFromMagic(header.magic); 915 } 916 if (data_sp) { 917 ModuleSpec spec; 918 spec.GetFileSpec() = file; 919 spec.SetObjectOffset(file_offset); 920 spec.SetObjectSize(length); 921 922 if (GetArchitecture(header, data, data_offset, 923 spec.GetArchitecture())) { 924 if (spec.GetArchitecture().IsValid()) { 925 GetUUID(header, data, data_offset, spec.GetUUID()); 926 specs.Append(spec); 927 } 928 } 929 } 930 } 931 } 932 return specs.GetSize() - initial_count; 933 } 934 935 const ConstString &ObjectFileMachO::GetSegmentNameTEXT() { 936 static ConstString g_segment_name_TEXT("__TEXT"); 937 return g_segment_name_TEXT; 938 } 939 940 const ConstString &ObjectFileMachO::GetSegmentNameDATA() { 941 static ConstString g_segment_name_DATA("__DATA"); 942 return g_segment_name_DATA; 943 } 944 945 const ConstString &ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 946 static ConstString g_segment_name("__DATA_DIRTY"); 947 return g_segment_name; 948 } 949 950 const ConstString &ObjectFileMachO::GetSegmentNameDATA_CONST() { 951 static ConstString g_segment_name("__DATA_CONST"); 952 return g_segment_name; 953 } 954 955 const ConstString &ObjectFileMachO::GetSegmentNameOBJC() { 956 static ConstString g_segment_name_OBJC("__OBJC"); 957 return g_segment_name_OBJC; 958 } 959 960 const ConstString &ObjectFileMachO::GetSegmentNameLINKEDIT() { 961 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 962 return g_section_name_LINKEDIT; 963 } 964 965 const ConstString &ObjectFileMachO::GetSectionNameEHFrame() { 966 static ConstString g_section_name_eh_frame("__eh_frame"); 967 return g_section_name_eh_frame; 968 } 969 970 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 971 lldb::addr_t data_offset, 972 lldb::addr_t data_length) { 973 DataExtractor data; 974 data.SetData(data_sp, data_offset, data_length); 975 lldb::offset_t offset = 0; 976 uint32_t magic = data.GetU32(&offset); 977 return MachHeaderSizeFromMagic(magic) != 0; 978 } 979 980 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 981 DataBufferSP &data_sp, 982 lldb::offset_t data_offset, 983 const FileSpec *file, 984 lldb::offset_t file_offset, 985 lldb::offset_t length) 986 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 987 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 988 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 989 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 990 ::memset(&m_header, 0, sizeof(m_header)); 991 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 992 } 993 994 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 995 lldb::DataBufferSP &header_data_sp, 996 const lldb::ProcessSP &process_sp, 997 lldb::addr_t header_addr) 998 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 999 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 1000 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 1001 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 1002 ::memset(&m_header, 0, sizeof(m_header)); 1003 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1004 } 1005 1006 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 1007 lldb::offset_t *data_offset_ptr, 1008 llvm::MachO::mach_header &header) { 1009 data.SetByteOrder(endian::InlHostByteOrder()); 1010 // Leave magic in the original byte order 1011 header.magic = data.GetU32(data_offset_ptr); 1012 bool can_parse = false; 1013 bool is_64_bit = false; 1014 switch (header.magic) { 1015 case MH_MAGIC: 1016 data.SetByteOrder(endian::InlHostByteOrder()); 1017 data.SetAddressByteSize(4); 1018 can_parse = true; 1019 break; 1020 1021 case MH_MAGIC_64: 1022 data.SetByteOrder(endian::InlHostByteOrder()); 1023 data.SetAddressByteSize(8); 1024 can_parse = true; 1025 is_64_bit = true; 1026 break; 1027 1028 case MH_CIGAM: 1029 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1030 ? eByteOrderLittle 1031 : eByteOrderBig); 1032 data.SetAddressByteSize(4); 1033 can_parse = true; 1034 break; 1035 1036 case MH_CIGAM_64: 1037 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1038 ? eByteOrderLittle 1039 : eByteOrderBig); 1040 data.SetAddressByteSize(8); 1041 is_64_bit = true; 1042 can_parse = true; 1043 break; 1044 1045 default: 1046 break; 1047 } 1048 1049 if (can_parse) { 1050 data.GetU32(data_offset_ptr, &header.cputype, 6); 1051 if (is_64_bit) 1052 *data_offset_ptr += 4; 1053 return true; 1054 } else { 1055 memset(&header, 0, sizeof(header)); 1056 } 1057 return false; 1058 } 1059 1060 bool ObjectFileMachO::ParseHeader() { 1061 ModuleSP module_sp(GetModule()); 1062 if (module_sp) { 1063 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1064 bool can_parse = false; 1065 lldb::offset_t offset = 0; 1066 m_data.SetByteOrder(endian::InlHostByteOrder()); 1067 // Leave magic in the original byte order 1068 m_header.magic = m_data.GetU32(&offset); 1069 switch (m_header.magic) { 1070 case MH_MAGIC: 1071 m_data.SetByteOrder(endian::InlHostByteOrder()); 1072 m_data.SetAddressByteSize(4); 1073 can_parse = true; 1074 break; 1075 1076 case MH_MAGIC_64: 1077 m_data.SetByteOrder(endian::InlHostByteOrder()); 1078 m_data.SetAddressByteSize(8); 1079 can_parse = true; 1080 break; 1081 1082 case MH_CIGAM: 1083 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1084 ? eByteOrderLittle 1085 : eByteOrderBig); 1086 m_data.SetAddressByteSize(4); 1087 can_parse = true; 1088 break; 1089 1090 case MH_CIGAM_64: 1091 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1092 ? eByteOrderLittle 1093 : eByteOrderBig); 1094 m_data.SetAddressByteSize(8); 1095 can_parse = true; 1096 break; 1097 1098 default: 1099 break; 1100 } 1101 1102 if (can_parse) { 1103 m_data.GetU32(&offset, &m_header.cputype, 6); 1104 1105 ArchSpec mach_arch; 1106 1107 if (GetArchitecture(mach_arch)) { 1108 // Check if the module has a required architecture 1109 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1110 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1111 return false; 1112 1113 if (SetModulesArchitecture(mach_arch)) { 1114 const size_t header_and_lc_size = 1115 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1116 if (m_data.GetByteSize() < header_and_lc_size) { 1117 DataBufferSP data_sp; 1118 ProcessSP process_sp(m_process_wp.lock()); 1119 if (process_sp) { 1120 data_sp = 1121 ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1122 } else { 1123 // Read in all only the load command data from the file on disk 1124 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset); 1125 if (data_sp->GetByteSize() != header_and_lc_size) 1126 return false; 1127 } 1128 if (data_sp) 1129 m_data.SetData(data_sp); 1130 } 1131 } 1132 return true; 1133 } 1134 } else { 1135 memset(&m_header, 0, sizeof(struct mach_header)); 1136 } 1137 } 1138 return false; 1139 } 1140 1141 ByteOrder ObjectFileMachO::GetByteOrder() const { 1142 return m_data.GetByteOrder(); 1143 } 1144 1145 bool ObjectFileMachO::IsExecutable() const { 1146 return m_header.filetype == MH_EXECUTE; 1147 } 1148 1149 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1150 return m_data.GetAddressByteSize(); 1151 } 1152 1153 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1154 Symtab *symtab = GetSymtab(); 1155 if (symtab) { 1156 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1157 if (symbol) { 1158 if (symbol->ValueIsAddress()) { 1159 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1160 if (section_sp) { 1161 const lldb::SectionType section_type = section_sp->GetType(); 1162 switch (section_type) { 1163 case eSectionTypeInvalid: 1164 return eAddressClassUnknown; 1165 1166 case eSectionTypeCode: 1167 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1168 // For ARM we have a bit in the n_desc field of the symbol that 1169 // tells us ARM/Thumb which is bit 0x0008. 1170 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1171 return eAddressClassCodeAlternateISA; 1172 } 1173 return eAddressClassCode; 1174 1175 case eSectionTypeContainer: 1176 return eAddressClassUnknown; 1177 1178 case eSectionTypeData: 1179 case eSectionTypeDataCString: 1180 case eSectionTypeDataCStringPointers: 1181 case eSectionTypeDataSymbolAddress: 1182 case eSectionTypeData4: 1183 case eSectionTypeData8: 1184 case eSectionTypeData16: 1185 case eSectionTypeDataPointers: 1186 case eSectionTypeZeroFill: 1187 case eSectionTypeDataObjCMessageRefs: 1188 case eSectionTypeDataObjCCFStrings: 1189 case eSectionTypeGoSymtab: 1190 return eAddressClassData; 1191 1192 case eSectionTypeDebug: 1193 case eSectionTypeDWARFDebugAbbrev: 1194 case eSectionTypeDWARFDebugAddr: 1195 case eSectionTypeDWARFDebugAranges: 1196 case eSectionTypeDWARFDebugCuIndex: 1197 case eSectionTypeDWARFDebugFrame: 1198 case eSectionTypeDWARFDebugInfo: 1199 case eSectionTypeDWARFDebugLine: 1200 case eSectionTypeDWARFDebugLoc: 1201 case eSectionTypeDWARFDebugMacInfo: 1202 case eSectionTypeDWARFDebugMacro: 1203 case eSectionTypeDWARFDebugNames: 1204 case eSectionTypeDWARFDebugPubNames: 1205 case eSectionTypeDWARFDebugPubTypes: 1206 case eSectionTypeDWARFDebugRanges: 1207 case eSectionTypeDWARFDebugStr: 1208 case eSectionTypeDWARFDebugStrOffsets: 1209 case eSectionTypeDWARFDebugTypes: 1210 case eSectionTypeDWARFAppleNames: 1211 case eSectionTypeDWARFAppleTypes: 1212 case eSectionTypeDWARFAppleNamespaces: 1213 case eSectionTypeDWARFAppleObjC: 1214 case eSectionTypeDWARFGNUDebugAltLink: 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 that tells 1246 // 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 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1353 EncryptedFileRanges result; 1354 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1355 1356 encryption_info_command encryption_cmd; 1357 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1358 const lldb::offset_t load_cmd_offset = offset; 1359 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1360 break; 1361 1362 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the 1363 // 3 fields we care about, so treat them the same. 1364 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1365 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1366 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1367 if (encryption_cmd.cryptid != 0) { 1368 EncryptedFileRanges::Entry entry; 1369 entry.SetRangeBase(encryption_cmd.cryptoff); 1370 entry.SetByteSize(encryption_cmd.cryptsize); 1371 result.Append(entry); 1372 } 1373 } 1374 } 1375 offset = load_cmd_offset + encryption_cmd.cmdsize; 1376 } 1377 1378 return result; 1379 } 1380 1381 void ObjectFileMachO::SanitizeSegmentCommand(segment_command_64 &seg_cmd, 1382 uint32_t cmd_idx) { 1383 if (m_length == 0 || seg_cmd.filesize == 0) 1384 return; 1385 1386 if (seg_cmd.fileoff > m_length) { 1387 // We have a load command that says it extends past the end of the file. 1388 // This is likely a corrupt file. We don't have any way to return an error 1389 // condition here (this method was likely invoked from something like 1390 // ObjectFile::GetSectionList()), so we just null out the section contents, 1391 // and dump a message to stdout. The most common case here is core file 1392 // debugging with a truncated file. 1393 const char *lc_segment_name = 1394 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1395 GetModule()->ReportWarning( 1396 "load command %u %s has a fileoff (0x%" PRIx64 1397 ") that extends beyond the end of the file (0x%" PRIx64 1398 "), ignoring this section", 1399 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1400 1401 seg_cmd.fileoff = 0; 1402 seg_cmd.filesize = 0; 1403 } 1404 1405 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1406 // We have a load command that says it extends past the end of the file. 1407 // This is likely a corrupt file. We don't have any way to return an error 1408 // condition here (this method was likely invoked from something like 1409 // ObjectFile::GetSectionList()), so we just null out the section contents, 1410 // and dump a message to stdout. The most common case here is core file 1411 // debugging with a truncated file. 1412 const char *lc_segment_name = 1413 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1414 GetModule()->ReportWarning( 1415 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1416 ") that extends beyond the end of the file (0x%" PRIx64 1417 "), the segment will be truncated to match", 1418 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1419 1420 // Truncate the length 1421 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1422 } 1423 } 1424 1425 static uint32_t GetSegmentPermissions(const segment_command_64 &seg_cmd) { 1426 uint32_t result = 0; 1427 if (seg_cmd.initprot & VM_PROT_READ) 1428 result |= ePermissionsReadable; 1429 if (seg_cmd.initprot & VM_PROT_WRITE) 1430 result |= ePermissionsWritable; 1431 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1432 result |= ePermissionsExecutable; 1433 return result; 1434 } 1435 1436 static lldb::SectionType GetSectionType(uint32_t flags, 1437 ConstString section_name) { 1438 1439 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1440 return eSectionTypeCode; 1441 1442 uint32_t mach_sect_type = flags & SECTION_TYPE; 1443 static ConstString g_sect_name_objc_data("__objc_data"); 1444 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1445 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1446 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1447 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1448 static ConstString g_sect_name_objc_const("__objc_const"); 1449 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1450 static ConstString g_sect_name_cfstring("__cfstring"); 1451 1452 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1453 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1454 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1455 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1456 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1457 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1458 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1459 static ConstString g_sect_name_dwarf_debug_names("__debug_names"); 1460 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1461 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1462 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1463 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1464 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1465 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1466 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1467 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1468 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1469 static ConstString g_sect_name_eh_frame("__eh_frame"); 1470 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1471 static ConstString g_sect_name_text("__text"); 1472 static ConstString g_sect_name_data("__data"); 1473 static ConstString g_sect_name_go_symtab("__gosymtab"); 1474 1475 if (section_name == g_sect_name_dwarf_debug_abbrev) 1476 return eSectionTypeDWARFDebugAbbrev; 1477 if (section_name == g_sect_name_dwarf_debug_aranges) 1478 return eSectionTypeDWARFDebugAranges; 1479 if (section_name == g_sect_name_dwarf_debug_frame) 1480 return eSectionTypeDWARFDebugFrame; 1481 if (section_name == g_sect_name_dwarf_debug_info) 1482 return eSectionTypeDWARFDebugInfo; 1483 if (section_name == g_sect_name_dwarf_debug_line) 1484 return eSectionTypeDWARFDebugLine; 1485 if (section_name == g_sect_name_dwarf_debug_loc) 1486 return eSectionTypeDWARFDebugLoc; 1487 if (section_name == g_sect_name_dwarf_debug_macinfo) 1488 return eSectionTypeDWARFDebugMacInfo; 1489 if (section_name == g_sect_name_dwarf_debug_names) 1490 return eSectionTypeDWARFDebugNames; 1491 if (section_name == g_sect_name_dwarf_debug_pubnames) 1492 return eSectionTypeDWARFDebugPubNames; 1493 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1494 return eSectionTypeDWARFDebugPubTypes; 1495 if (section_name == g_sect_name_dwarf_debug_ranges) 1496 return eSectionTypeDWARFDebugRanges; 1497 if (section_name == g_sect_name_dwarf_debug_str) 1498 return eSectionTypeDWARFDebugStr; 1499 if (section_name == g_sect_name_dwarf_debug_types) 1500 return eSectionTypeDWARFDebugTypes; 1501 if (section_name == g_sect_name_dwarf_apple_names) 1502 return eSectionTypeDWARFAppleNames; 1503 if (section_name == g_sect_name_dwarf_apple_types) 1504 return eSectionTypeDWARFAppleTypes; 1505 if (section_name == g_sect_name_dwarf_apple_namespaces) 1506 return eSectionTypeDWARFAppleNamespaces; 1507 if (section_name == g_sect_name_dwarf_apple_objc) 1508 return eSectionTypeDWARFAppleObjC; 1509 if (section_name == g_sect_name_objc_selrefs) 1510 return eSectionTypeDataCStringPointers; 1511 if (section_name == g_sect_name_objc_msgrefs) 1512 return eSectionTypeDataObjCMessageRefs; 1513 if (section_name == g_sect_name_eh_frame) 1514 return eSectionTypeEHFrame; 1515 if (section_name == g_sect_name_compact_unwind) 1516 return eSectionTypeCompactUnwind; 1517 if (section_name == g_sect_name_cfstring) 1518 return eSectionTypeDataObjCCFStrings; 1519 if (section_name == g_sect_name_go_symtab) 1520 return eSectionTypeGoSymtab; 1521 if (section_name == g_sect_name_objc_data || 1522 section_name == g_sect_name_objc_classrefs || 1523 section_name == g_sect_name_objc_superrefs || 1524 section_name == g_sect_name_objc_const || 1525 section_name == g_sect_name_objc_classlist) { 1526 return eSectionTypeDataPointers; 1527 } 1528 1529 switch (mach_sect_type) { 1530 // TODO: categorize sections by other flags for regular sections 1531 case S_REGULAR: 1532 if (section_name == g_sect_name_text) 1533 return eSectionTypeCode; 1534 if (section_name == g_sect_name_data) 1535 return eSectionTypeData; 1536 return eSectionTypeOther; 1537 case S_ZEROFILL: 1538 return eSectionTypeZeroFill; 1539 case S_CSTRING_LITERALS: // section with only literal C strings 1540 return eSectionTypeDataCString; 1541 case S_4BYTE_LITERALS: // section with only 4 byte literals 1542 return eSectionTypeData4; 1543 case S_8BYTE_LITERALS: // section with only 8 byte literals 1544 return eSectionTypeData8; 1545 case S_LITERAL_POINTERS: // section with only pointers to literals 1546 return eSectionTypeDataPointers; 1547 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1548 return eSectionTypeDataPointers; 1549 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1550 return eSectionTypeDataPointers; 1551 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1552 // the reserved2 field 1553 return eSectionTypeCode; 1554 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1555 // initialization 1556 return eSectionTypeDataPointers; 1557 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1558 // termination 1559 return eSectionTypeDataPointers; 1560 case S_COALESCED: 1561 return eSectionTypeOther; 1562 case S_GB_ZEROFILL: 1563 return eSectionTypeZeroFill; 1564 case S_INTERPOSING: // section with only pairs of function pointers for 1565 // interposing 1566 return eSectionTypeCode; 1567 case S_16BYTE_LITERALS: // section with only 16 byte literals 1568 return eSectionTypeData16; 1569 case S_DTRACE_DOF: 1570 return eSectionTypeDebug; 1571 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1572 return eSectionTypeDataPointers; 1573 default: 1574 return eSectionTypeOther; 1575 } 1576 } 1577 1578 struct ObjectFileMachO::SegmentParsingContext { 1579 const EncryptedFileRanges EncryptedRanges; 1580 lldb_private::SectionList &UnifiedList; 1581 uint32_t NextSegmentIdx = 0; 1582 uint32_t NextSectionIdx = 0; 1583 bool FileAddressesChanged = false; 1584 1585 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1586 lldb_private::SectionList &UnifiedList) 1587 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1588 }; 1589 1590 void ObjectFileMachO::ProcessSegmentCommand(const load_command &load_cmd_, 1591 lldb::offset_t offset, 1592 uint32_t cmd_idx, 1593 SegmentParsingContext &context) { 1594 segment_command_64 load_cmd; 1595 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1596 1597 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1598 return; 1599 1600 ModuleSP module_sp = GetModule(); 1601 const bool is_core = GetType() == eTypeCoreFile; 1602 const bool is_dsym = (m_header.filetype == MH_DSYM); 1603 bool add_section = true; 1604 bool add_to_unified = true; 1605 ConstString const_segname( 1606 load_cmd.segname, 1607 std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 1608 1609 SectionSP unified_section_sp( 1610 context.UnifiedList.FindSectionByName(const_segname)); 1611 if (is_dsym && unified_section_sp) { 1612 if (const_segname == GetSegmentNameLINKEDIT()) { 1613 // We need to keep the __LINKEDIT segment private to this object file 1614 // only 1615 add_to_unified = false; 1616 } else { 1617 // This is the dSYM file and this section has already been created by the 1618 // object file, no need to create it. 1619 add_section = false; 1620 } 1621 } 1622 load_cmd.vmaddr = m_data.GetAddress(&offset); 1623 load_cmd.vmsize = m_data.GetAddress(&offset); 1624 load_cmd.fileoff = m_data.GetAddress(&offset); 1625 load_cmd.filesize = m_data.GetAddress(&offset); 1626 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1627 return; 1628 1629 SanitizeSegmentCommand(load_cmd, cmd_idx); 1630 1631 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1632 const bool segment_is_encrypted = 1633 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1634 1635 // Keep a list of mach segments around in case we need to get at data that 1636 // isn't stored in the abstracted Sections. 1637 m_mach_segments.push_back(load_cmd); 1638 1639 // Use a segment ID of the segment index shifted left by 8 so they never 1640 // conflict with any of the sections. 1641 SectionSP segment_sp; 1642 if (add_section && (const_segname || is_core)) { 1643 segment_sp.reset(new Section( 1644 module_sp, // Module to which this section belongs 1645 this, // Object file to which this sections belongs 1646 ++context.NextSegmentIdx 1647 << 8, // Section ID is the 1 based segment index 1648 // shifted right by 8 bits as not to collide with any of the 256 1649 // section IDs that are possible 1650 const_segname, // Name of this section 1651 eSectionTypeContainer, // This section is a container of other 1652 // sections. 1653 load_cmd.vmaddr, // File VM address == addresses as they are 1654 // found in the object file 1655 load_cmd.vmsize, // VM size in bytes of this section 1656 load_cmd.fileoff, // Offset to the data for this section in 1657 // the file 1658 load_cmd.filesize, // Size in bytes of this section as found 1659 // in the file 1660 0, // Segments have no alignment information 1661 load_cmd.flags)); // Flags for this section 1662 1663 segment_sp->SetIsEncrypted(segment_is_encrypted); 1664 m_sections_ap->AddSection(segment_sp); 1665 segment_sp->SetPermissions(segment_permissions); 1666 if (add_to_unified) 1667 context.UnifiedList.AddSection(segment_sp); 1668 } else if (unified_section_sp) { 1669 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1670 // Check to see if the module was read from memory? 1671 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) { 1672 // We have a module that is in memory and needs to have its file 1673 // address adjusted. We need to do this because when we load a file 1674 // from memory, its addresses will be slid already, yet the addresses 1675 // in the new symbol file will still be unslid. Since everything is 1676 // stored as section offset, this shouldn't cause any problems. 1677 1678 // Make sure we've parsed the symbol table from the ObjectFile before 1679 // we go around changing its Sections. 1680 module_sp->GetObjectFile()->GetSymtab(); 1681 // eh_frame would present the same problems but we parse that on a per- 1682 // function basis as-needed so it's more difficult to remove its use of 1683 // the Sections. Realistically, the environments where this code path 1684 // will be taken will not have eh_frame sections. 1685 1686 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1687 1688 // Notify the module that the section addresses have been changed once 1689 // we're done so any file-address caches can be updated. 1690 context.FileAddressesChanged = true; 1691 } 1692 } 1693 m_sections_ap->AddSection(unified_section_sp); 1694 } 1695 1696 struct section_64 sect64; 1697 ::memset(§64, 0, sizeof(sect64)); 1698 // Push a section into our mach sections for the section at index zero 1699 // (NO_SECT) if we don't have any mach sections yet... 1700 if (m_mach_sections.empty()) 1701 m_mach_sections.push_back(sect64); 1702 uint32_t segment_sect_idx; 1703 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1704 1705 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1706 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1707 ++segment_sect_idx) { 1708 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1709 sizeof(sect64.sectname)) == NULL) 1710 break; 1711 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1712 sizeof(sect64.segname)) == NULL) 1713 break; 1714 sect64.addr = m_data.GetAddress(&offset); 1715 sect64.size = m_data.GetAddress(&offset); 1716 1717 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1718 break; 1719 1720 // Keep a list of mach sections around in case we need to get at data that 1721 // isn't stored in the abstracted Sections. 1722 m_mach_sections.push_back(sect64); 1723 1724 if (add_section) { 1725 ConstString section_name( 1726 sect64.sectname, 1727 std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1728 if (!const_segname) { 1729 // We have a segment with no name so we need to conjure up segments 1730 // that correspond to the section's segname if there isn't already such 1731 // a section. If there is such a section, we resize the section so that 1732 // it spans all sections. We also mark these sections as fake so 1733 // address matches don't hit if they land in the gaps between the child 1734 // sections. 1735 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1736 sizeof(sect64.segname)); 1737 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1738 if (segment_sp.get()) { 1739 Section *segment = segment_sp.get(); 1740 // Grow the section size as needed. 1741 const lldb::addr_t sect64_min_addr = sect64.addr; 1742 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1743 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1744 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1745 const lldb::addr_t curr_seg_max_addr = 1746 curr_seg_min_addr + curr_seg_byte_size; 1747 if (sect64_min_addr >= curr_seg_min_addr) { 1748 const lldb::addr_t new_seg_byte_size = 1749 sect64_max_addr - curr_seg_min_addr; 1750 // Only grow the section size if needed 1751 if (new_seg_byte_size > curr_seg_byte_size) 1752 segment->SetByteSize(new_seg_byte_size); 1753 } else { 1754 // We need to change the base address of the segment and adjust the 1755 // child section offsets for all existing children. 1756 const lldb::addr_t slide_amount = 1757 sect64_min_addr - curr_seg_min_addr; 1758 segment->Slide(slide_amount, false); 1759 segment->GetChildren().Slide(-slide_amount, false); 1760 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1761 } 1762 1763 // Grow the section size as needed. 1764 if (sect64.offset) { 1765 const lldb::addr_t segment_min_file_offset = 1766 segment->GetFileOffset(); 1767 const lldb::addr_t segment_max_file_offset = 1768 segment_min_file_offset + segment->GetFileSize(); 1769 1770 const lldb::addr_t section_min_file_offset = sect64.offset; 1771 const lldb::addr_t section_max_file_offset = 1772 section_min_file_offset + sect64.size; 1773 const lldb::addr_t new_file_offset = 1774 std::min(section_min_file_offset, segment_min_file_offset); 1775 const lldb::addr_t new_file_size = 1776 std::max(section_max_file_offset, segment_max_file_offset) - 1777 new_file_offset; 1778 segment->SetFileOffset(new_file_offset); 1779 segment->SetFileSize(new_file_size); 1780 } 1781 } else { 1782 // Create a fake section for the section's named segment 1783 segment_sp.reset(new Section( 1784 segment_sp, // Parent section 1785 module_sp, // Module to which this section belongs 1786 this, // Object file to which this section belongs 1787 ++context.NextSegmentIdx 1788 << 8, // Section ID is the 1 based segment index 1789 // shifted right by 8 bits as not to 1790 // collide with any of the 256 section IDs 1791 // that are possible 1792 const_segname, // Name of this section 1793 eSectionTypeContainer, // This section is a container of 1794 // other sections. 1795 sect64.addr, // File VM address == addresses as they are 1796 // found in the object file 1797 sect64.size, // VM size in bytes of this section 1798 sect64.offset, // Offset to the data for this section in 1799 // the file 1800 sect64.offset ? sect64.size : 0, // Size in bytes of 1801 // this section as 1802 // found in the file 1803 sect64.align, 1804 load_cmd.flags)); // Flags for this section 1805 segment_sp->SetIsFake(true); 1806 segment_sp->SetPermissions(segment_permissions); 1807 m_sections_ap->AddSection(segment_sp); 1808 if (add_to_unified) 1809 context.UnifiedList.AddSection(segment_sp); 1810 segment_sp->SetIsEncrypted(segment_is_encrypted); 1811 } 1812 } 1813 assert(segment_sp.get()); 1814 1815 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1816 1817 SectionSP section_sp(new Section( 1818 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1819 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1820 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1821 sect64.flags)); 1822 // Set the section to be encrypted to match the segment 1823 1824 bool section_is_encrypted = false; 1825 if (!segment_is_encrypted && load_cmd.filesize != 0) 1826 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1827 sect64.offset) != NULL; 1828 1829 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1830 section_sp->SetPermissions(segment_permissions); 1831 segment_sp->GetChildren().AddSection(section_sp); 1832 1833 if (segment_sp->IsFake()) { 1834 segment_sp.reset(); 1835 const_segname.Clear(); 1836 } 1837 } 1838 } 1839 if (segment_sp && is_dsym) { 1840 if (first_segment_sectID <= context.NextSectionIdx) { 1841 lldb::user_id_t sect_uid; 1842 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1843 ++sect_uid) { 1844 SectionSP curr_section_sp( 1845 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1846 SectionSP next_section_sp; 1847 if (sect_uid + 1 <= context.NextSectionIdx) 1848 next_section_sp = 1849 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1850 1851 if (curr_section_sp.get()) { 1852 if (curr_section_sp->GetByteSize() == 0) { 1853 if (next_section_sp.get() != NULL) 1854 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1855 curr_section_sp->GetFileAddress()); 1856 else 1857 curr_section_sp->SetByteSize(load_cmd.vmsize); 1858 } 1859 } 1860 } 1861 } 1862 } 1863 } 1864 1865 void ObjectFileMachO::ProcessDysymtabCommand(const load_command &load_cmd, 1866 lldb::offset_t offset) { 1867 m_dysymtab.cmd = load_cmd.cmd; 1868 m_dysymtab.cmdsize = load_cmd.cmdsize; 1869 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1870 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1871 } 1872 1873 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1874 if (m_sections_ap) 1875 return; 1876 1877 m_sections_ap.reset(new SectionList()); 1878 1879 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1880 // bool dump_sections = false; 1881 ModuleSP module_sp(GetModule()); 1882 1883 offset = MachHeaderSizeFromMagic(m_header.magic); 1884 1885 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1886 struct load_command load_cmd; 1887 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1888 const lldb::offset_t load_cmd_offset = offset; 1889 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1890 break; 1891 1892 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1893 ProcessSegmentCommand(load_cmd, offset, i, context); 1894 else if (load_cmd.cmd == LC_DYSYMTAB) 1895 ProcessDysymtabCommand(load_cmd, offset); 1896 1897 offset = load_cmd_offset + load_cmd.cmdsize; 1898 } 1899 1900 if (context.FileAddressesChanged && module_sp) 1901 module_sp->SectionFileAddressesChanged(); 1902 } 1903 1904 class MachSymtabSectionInfo { 1905 public: 1906 MachSymtabSectionInfo(SectionList *section_list) 1907 : m_section_list(section_list), m_section_infos() { 1908 // Get the number of sections down to a depth of 1 to include all segments 1909 // and their sections, but no other sections that may be added for debug 1910 // map or 1911 m_section_infos.resize(section_list->GetNumSections(1)); 1912 } 1913 1914 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1915 if (n_sect == 0) 1916 return SectionSP(); 1917 if (n_sect < m_section_infos.size()) { 1918 if (!m_section_infos[n_sect].section_sp) { 1919 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1920 m_section_infos[n_sect].section_sp = section_sp; 1921 if (section_sp) { 1922 m_section_infos[n_sect].vm_range.SetBaseAddress( 1923 section_sp->GetFileAddress()); 1924 m_section_infos[n_sect].vm_range.SetByteSize( 1925 section_sp->GetByteSize()); 1926 } else { 1927 Host::SystemLog(Host::eSystemLogError, 1928 "error: unable to find section for section %u\n", 1929 n_sect); 1930 } 1931 } 1932 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1933 // Symbol is in section. 1934 return m_section_infos[n_sect].section_sp; 1935 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1936 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1937 file_addr) { 1938 // Symbol is in section with zero size, but has the same start address 1939 // as the section. This can happen with linker symbols (symbols that 1940 // start with the letter 'l' or 'L'. 1941 return m_section_infos[n_sect].section_sp; 1942 } 1943 } 1944 return m_section_list->FindSectionContainingFileAddress(file_addr); 1945 } 1946 1947 protected: 1948 struct SectionInfo { 1949 SectionInfo() : vm_range(), section_sp() {} 1950 1951 VMRange vm_range; 1952 SectionSP section_sp; 1953 }; 1954 SectionList *m_section_list; 1955 std::vector<SectionInfo> m_section_infos; 1956 }; 1957 1958 struct TrieEntry { 1959 TrieEntry() 1960 : name(), address(LLDB_INVALID_ADDRESS), flags(0), other(0), 1961 import_name() {} 1962 1963 void Clear() { 1964 name.Clear(); 1965 address = LLDB_INVALID_ADDRESS; 1966 flags = 0; 1967 other = 0; 1968 import_name.Clear(); 1969 } 1970 1971 void Dump() const { 1972 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1973 static_cast<unsigned long long>(address), 1974 static_cast<unsigned long long>(flags), 1975 static_cast<unsigned long long>(other), name.GetCString()); 1976 if (import_name) 1977 printf(" -> \"%s\"\n", import_name.GetCString()); 1978 else 1979 printf("\n"); 1980 } 1981 ConstString name; 1982 uint64_t address; 1983 uint64_t flags; 1984 uint64_t other; 1985 ConstString import_name; 1986 }; 1987 1988 struct TrieEntryWithOffset { 1989 lldb::offset_t nodeOffset; 1990 TrieEntry entry; 1991 1992 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1993 1994 void Dump(uint32_t idx) const { 1995 printf("[%3u] 0x%16.16llx: ", idx, 1996 static_cast<unsigned long long>(nodeOffset)); 1997 entry.Dump(); 1998 } 1999 2000 bool operator<(const TrieEntryWithOffset &other) const { 2001 return (nodeOffset < other.nodeOffset); 2002 } 2003 }; 2004 2005 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2006 const bool is_arm, 2007 std::vector<llvm::StringRef> &nameSlices, 2008 std::set<lldb::addr_t> &resolver_addresses, 2009 std::vector<TrieEntryWithOffset> &output) { 2010 if (!data.ValidOffset(offset)) 2011 return true; 2012 2013 const uint64_t terminalSize = data.GetULEB128(&offset); 2014 lldb::offset_t children_offset = offset + terminalSize; 2015 if (terminalSize != 0) { 2016 TrieEntryWithOffset e(offset); 2017 e.entry.flags = data.GetULEB128(&offset); 2018 const char *import_name = NULL; 2019 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2020 e.entry.address = 0; 2021 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2022 import_name = data.GetCStr(&offset); 2023 } else { 2024 e.entry.address = data.GetULEB128(&offset); 2025 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2026 e.entry.other = data.GetULEB128(&offset); 2027 uint64_t resolver_addr = e.entry.other; 2028 if (is_arm) 2029 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2030 resolver_addresses.insert(resolver_addr); 2031 } else 2032 e.entry.other = 0; 2033 } 2034 // Only add symbols that are reexport symbols with a valid import name 2035 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && 2036 import_name[0]) { 2037 std::string name; 2038 if (!nameSlices.empty()) { 2039 for (auto name_slice : nameSlices) 2040 name.append(name_slice.data(), name_slice.size()); 2041 } 2042 if (name.size() > 1) { 2043 // Skip the leading '_' 2044 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2045 } 2046 if (import_name) { 2047 // Skip the leading '_' 2048 e.entry.import_name.SetCString(import_name + 1); 2049 } 2050 output.push_back(e); 2051 } 2052 } 2053 2054 const uint8_t childrenCount = data.GetU8(&children_offset); 2055 for (uint8_t i = 0; i < childrenCount; ++i) { 2056 const char *cstr = data.GetCStr(&children_offset); 2057 if (cstr) 2058 nameSlices.push_back(llvm::StringRef(cstr)); 2059 else 2060 return false; // Corrupt data 2061 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2062 if (childNodeOffset) { 2063 if (!ParseTrieEntries(data, childNodeOffset, is_arm, nameSlices, 2064 resolver_addresses, output)) { 2065 return false; 2066 } 2067 } 2068 nameSlices.pop_back(); 2069 } 2070 return true; 2071 } 2072 2073 // Read the UUID out of a dyld_shared_cache file on-disk. 2074 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2075 const ByteOrder byte_order, 2076 const uint32_t addr_byte_size) { 2077 UUID dsc_uuid; 2078 DataBufferSP DscData = MapFileData( 2079 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2080 if (!DscData) 2081 return dsc_uuid; 2082 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2083 2084 char version_str[7]; 2085 lldb::offset_t offset = 0; 2086 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2087 version_str[6] = '\0'; 2088 if (strcmp(version_str, "dyld_v") == 0) { 2089 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2090 uint8_t uuid_bytes[sizeof(uuid_t)]; 2091 memcpy(uuid_bytes, dsc_header_data.GetData(&offset, sizeof(uuid_t)), 2092 sizeof(uuid_t)); 2093 dsc_uuid.SetBytes(uuid_bytes); 2094 } 2095 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2096 if (log && dsc_uuid.IsValid()) { 2097 log->Printf("Shared cache %s has UUID %s", dyld_shared_cache.GetPath().c_str(), 2098 dsc_uuid.GetAsString().c_str()); 2099 } 2100 return dsc_uuid; 2101 } 2102 2103 size_t ObjectFileMachO::ParseSymtab() { 2104 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2105 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2106 m_file.GetFilename().AsCString("")); 2107 ModuleSP module_sp(GetModule()); 2108 if (!module_sp) 2109 return 0; 2110 2111 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2112 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2113 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2114 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2115 FunctionStarts function_starts; 2116 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2117 uint32_t i; 2118 FileSpecList dylib_files; 2119 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2120 static const llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2121 static const llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2122 static const llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2123 2124 for (i = 0; i < m_header.ncmds; ++i) { 2125 const lldb::offset_t cmd_offset = offset; 2126 // Read in the load command and load command size 2127 struct load_command lc; 2128 if (m_data.GetU32(&offset, &lc, 2) == NULL) 2129 break; 2130 // Watch for the symbol table load command 2131 switch (lc.cmd) { 2132 case LC_SYMTAB: 2133 symtab_load_command.cmd = lc.cmd; 2134 symtab_load_command.cmdsize = lc.cmdsize; 2135 // Read in the rest of the symtab load command 2136 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2137 0) // fill in symoff, nsyms, stroff, strsize fields 2138 return 0; 2139 if (symtab_load_command.symoff == 0) { 2140 if (log) 2141 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2142 return 0; 2143 } 2144 2145 if (symtab_load_command.stroff == 0) { 2146 if (log) 2147 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2148 return 0; 2149 } 2150 2151 if (symtab_load_command.nsyms == 0) { 2152 if (log) 2153 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2154 return 0; 2155 } 2156 2157 if (symtab_load_command.strsize == 0) { 2158 if (log) 2159 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2160 return 0; 2161 } 2162 break; 2163 2164 case LC_DYLD_INFO: 2165 case LC_DYLD_INFO_ONLY: 2166 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2167 dyld_info.cmd = lc.cmd; 2168 dyld_info.cmdsize = lc.cmdsize; 2169 } else { 2170 memset(&dyld_info, 0, sizeof(dyld_info)); 2171 } 2172 break; 2173 2174 case LC_LOAD_DYLIB: 2175 case LC_LOAD_WEAK_DYLIB: 2176 case LC_REEXPORT_DYLIB: 2177 case LC_LOADFVMLIB: 2178 case LC_LOAD_UPWARD_DYLIB: { 2179 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2180 const char *path = m_data.PeekCStr(name_offset); 2181 if (path) { 2182 FileSpec file_spec(path, false); 2183 // Strip the path if there is @rpath, @executable, etc so we just use 2184 // the basename 2185 if (path[0] == '@') 2186 file_spec.GetDirectory().Clear(); 2187 2188 if (lc.cmd == LC_REEXPORT_DYLIB) { 2189 m_reexported_dylibs.AppendIfUnique(file_spec); 2190 } 2191 2192 dylib_files.Append(file_spec); 2193 } 2194 } break; 2195 2196 case LC_FUNCTION_STARTS: 2197 function_starts_load_command.cmd = lc.cmd; 2198 function_starts_load_command.cmdsize = lc.cmdsize; 2199 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2200 NULL) // fill in symoff, nsyms, stroff, strsize fields 2201 memset(&function_starts_load_command, 0, 2202 sizeof(function_starts_load_command)); 2203 break; 2204 2205 default: 2206 break; 2207 } 2208 offset = cmd_offset + lc.cmdsize; 2209 } 2210 2211 if (symtab_load_command.cmd) { 2212 Symtab *symtab = m_symtab_ap.get(); 2213 SectionList *section_list = GetSectionList(); 2214 if (section_list == NULL) 2215 return 0; 2216 2217 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2218 const ByteOrder byte_order = m_data.GetByteOrder(); 2219 bool bit_width_32 = addr_byte_size == 4; 2220 const size_t nlist_byte_size = 2221 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2222 2223 DataExtractor nlist_data(NULL, 0, byte_order, addr_byte_size); 2224 DataExtractor strtab_data(NULL, 0, byte_order, addr_byte_size); 2225 DataExtractor function_starts_data(NULL, 0, byte_order, addr_byte_size); 2226 DataExtractor indirect_symbol_index_data(NULL, 0, byte_order, 2227 addr_byte_size); 2228 DataExtractor dyld_trie_data(NULL, 0, byte_order, addr_byte_size); 2229 2230 const addr_t nlist_data_byte_size = 2231 symtab_load_command.nsyms * nlist_byte_size; 2232 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2233 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2234 2235 ProcessSP process_sp(m_process_wp.lock()); 2236 Process *process = process_sp.get(); 2237 2238 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2239 2240 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2241 Target &target = process->GetTarget(); 2242 2243 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2244 2245 SectionSP linkedit_section_sp( 2246 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2247 // Reading mach file from memory in a process or core file... 2248 2249 if (linkedit_section_sp) { 2250 addr_t linkedit_load_addr = 2251 linkedit_section_sp->GetLoadBaseAddress(&target); 2252 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2253 // We might be trying to access the symbol table before the 2254 // __LINKEDIT's load address has been set in the target. We can't 2255 // fail to read the symbol table, so calculate the right address 2256 // manually 2257 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2258 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2259 } 2260 2261 const addr_t linkedit_file_offset = 2262 linkedit_section_sp->GetFileOffset(); 2263 const addr_t symoff_addr = linkedit_load_addr + 2264 symtab_load_command.symoff - 2265 linkedit_file_offset; 2266 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2267 linkedit_file_offset; 2268 2269 bool data_was_read = false; 2270 2271 #if defined(__APPLE__) && \ 2272 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2273 if (m_header.flags & 0x80000000u && 2274 process->GetAddressByteSize() == sizeof(void *)) { 2275 // This mach-o memory file is in the dyld shared cache. If this 2276 // program is not remote and this is iOS, then this process will 2277 // share the same shared cache as the process we are debugging and we 2278 // can read the entire __LINKEDIT from the address space in this 2279 // process. This is a needed optimization that is used for local iOS 2280 // debugging only since all shared libraries in the shared cache do 2281 // not have corresponding files that exist in the file system of the 2282 // device. They have been combined into a single file. This means we 2283 // always have to load these files from memory. All of the symbol and 2284 // string tables from all of the __LINKEDIT sections from the shared 2285 // libraries in the shared cache have been merged into a single large 2286 // symbol and string table. Reading all of this symbol and string 2287 // table data across can slow down debug launch times, so we optimize 2288 // this by reading the memory for the __LINKEDIT section from this 2289 // process. 2290 2291 UUID lldb_shared_cache; 2292 addr_t lldb_shared_cache_addr; 2293 GetLLDBSharedCacheUUID (lldb_shared_cache_addr, lldb_shared_cache); 2294 UUID process_shared_cache; 2295 addr_t process_shared_cache_addr; 2296 GetProcessSharedCacheUUID(process, process_shared_cache_addr, process_shared_cache); 2297 bool use_lldb_cache = true; 2298 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2299 (lldb_shared_cache != process_shared_cache 2300 || process_shared_cache_addr != lldb_shared_cache_addr)) { 2301 use_lldb_cache = false; 2302 } 2303 2304 PlatformSP platform_sp(target.GetPlatform()); 2305 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2306 data_was_read = true; 2307 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2308 eByteOrderLittle); 2309 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2310 eByteOrderLittle); 2311 if (function_starts_load_command.cmd) { 2312 const addr_t func_start_addr = 2313 linkedit_load_addr + function_starts_load_command.dataoff - 2314 linkedit_file_offset; 2315 function_starts_data.SetData( 2316 (void *)func_start_addr, 2317 function_starts_load_command.datasize, eByteOrderLittle); 2318 } 2319 } 2320 } 2321 #endif 2322 2323 if (!data_was_read) { 2324 // Always load dyld - the dynamic linker - from memory if we didn't 2325 // find a binary anywhere else. lldb will not register 2326 // dylib/framework/bundle loads/unloads if we don't have the dyld 2327 // symbols, we force dyld to load from memory despite the user's 2328 // target.memory-module-load-level setting. 2329 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2330 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2331 DataBufferSP nlist_data_sp( 2332 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2333 if (nlist_data_sp) 2334 nlist_data.SetData(nlist_data_sp, 0, 2335 nlist_data_sp->GetByteSize()); 2336 // Load strings individually from memory when loading from memory 2337 // since shared cache string tables contain strings for all symbols 2338 // from all shared cached libraries DataBufferSP strtab_data_sp 2339 // (ReadMemory (process_sp, strtab_addr, 2340 // strtab_data_byte_size)); 2341 // if (strtab_data_sp) 2342 // strtab_data.SetData (strtab_data_sp, 0, 2343 // strtab_data_sp->GetByteSize()); 2344 if (m_dysymtab.nindirectsyms != 0) { 2345 const addr_t indirect_syms_addr = linkedit_load_addr + 2346 m_dysymtab.indirectsymoff - 2347 linkedit_file_offset; 2348 DataBufferSP indirect_syms_data_sp( 2349 ReadMemory(process_sp, indirect_syms_addr, 2350 m_dysymtab.nindirectsyms * 4)); 2351 if (indirect_syms_data_sp) 2352 indirect_symbol_index_data.SetData( 2353 indirect_syms_data_sp, 0, 2354 indirect_syms_data_sp->GetByteSize()); 2355 } 2356 } else if (memory_module_load_level >= 2357 eMemoryModuleLoadLevelPartial) { 2358 if (function_starts_load_command.cmd) { 2359 const addr_t func_start_addr = 2360 linkedit_load_addr + function_starts_load_command.dataoff - 2361 linkedit_file_offset; 2362 DataBufferSP func_start_data_sp( 2363 ReadMemory(process_sp, func_start_addr, 2364 function_starts_load_command.datasize)); 2365 if (func_start_data_sp) 2366 function_starts_data.SetData(func_start_data_sp, 0, 2367 func_start_data_sp->GetByteSize()); 2368 } 2369 } 2370 } 2371 } 2372 } else { 2373 nlist_data.SetData(m_data, symtab_load_command.symoff, 2374 nlist_data_byte_size); 2375 strtab_data.SetData(m_data, symtab_load_command.stroff, 2376 strtab_data_byte_size); 2377 2378 if (dyld_info.export_size > 0) { 2379 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2380 dyld_info.export_size); 2381 } 2382 2383 if (m_dysymtab.nindirectsyms != 0) { 2384 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2385 m_dysymtab.nindirectsyms * 4); 2386 } 2387 if (function_starts_load_command.cmd) { 2388 function_starts_data.SetData(m_data, 2389 function_starts_load_command.dataoff, 2390 function_starts_load_command.datasize); 2391 } 2392 } 2393 2394 if (nlist_data.GetByteSize() == 0 && 2395 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2396 if (log) 2397 module_sp->LogMessage(log, "failed to read nlist data"); 2398 return 0; 2399 } 2400 2401 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2402 if (!have_strtab_data) { 2403 if (process) { 2404 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2405 if (log) 2406 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2407 return 0; 2408 } 2409 } else { 2410 if (log) 2411 module_sp->LogMessage(log, "failed to read strtab data"); 2412 return 0; 2413 } 2414 } 2415 2416 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2417 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2418 const ConstString &g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2419 const ConstString &g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2420 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2421 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2422 SectionSP text_section_sp( 2423 section_list->FindSectionByName(g_segment_name_TEXT)); 2424 SectionSP data_section_sp( 2425 section_list->FindSectionByName(g_segment_name_DATA)); 2426 SectionSP data_dirty_section_sp( 2427 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2428 SectionSP data_const_section_sp( 2429 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2430 SectionSP objc_section_sp( 2431 section_list->FindSectionByName(g_segment_name_OBJC)); 2432 SectionSP eh_frame_section_sp; 2433 if (text_section_sp.get()) 2434 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2435 g_section_name_eh_frame); 2436 else 2437 eh_frame_section_sp = 2438 section_list->FindSectionByName(g_section_name_eh_frame); 2439 2440 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2441 2442 // lldb works best if it knows the start address of all functions in a 2443 // module. Linker symbols or debug info are normally the best source of 2444 // information for start addr / size but they may be stripped in a released 2445 // binary. Two additional sources of information exist in Mach-O binaries: 2446 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2447 // function's start address in the 2448 // binary, relative to the text section. 2449 // eh_frame - the eh_frame FDEs have the start addr & size of 2450 // each function 2451 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2452 // all modern binaries. 2453 // Binaries built to run on older releases may need to use eh_frame 2454 // information. 2455 2456 if (text_section_sp && function_starts_data.GetByteSize()) { 2457 FunctionStarts::Entry function_start_entry; 2458 function_start_entry.data = false; 2459 lldb::offset_t function_start_offset = 0; 2460 function_start_entry.addr = text_section_sp->GetFileAddress(); 2461 uint64_t delta; 2462 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2463 0) { 2464 // Now append the current entry 2465 function_start_entry.addr += delta; 2466 function_starts.Append(function_start_entry); 2467 } 2468 } else { 2469 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2470 // load command claiming an eh_frame but it doesn't actually have the 2471 // eh_frame content. And if we have a dSYM, we don't need to do any of 2472 // this fill-in-the-missing-symbols works anyway - the debug info should 2473 // give us all the functions in the module. 2474 if (text_section_sp.get() && eh_frame_section_sp.get() && 2475 m_type != eTypeDebugInfo) { 2476 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2477 DWARFCallFrameInfo::EH); 2478 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2479 eh_frame.GetFunctionAddressAndSizeVector(functions); 2480 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2481 size_t count = functions.GetSize(); 2482 for (size_t i = 0; i < count; ++i) { 2483 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2484 functions.GetEntryAtIndex(i); 2485 if (func) { 2486 FunctionStarts::Entry function_start_entry; 2487 function_start_entry.addr = func->base - text_base_addr; 2488 function_starts.Append(function_start_entry); 2489 } 2490 } 2491 } 2492 } 2493 2494 const size_t function_starts_count = function_starts.GetSize(); 2495 2496 // For user process binaries (executables, dylibs, frameworks, bundles), if 2497 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2498 // going to assume the binary has been stripped. Don't allow assembly 2499 // language instruction emulation because we don't know proper function 2500 // start boundaries. 2501 // 2502 // For all other types of binaries (kernels, stand-alone bare board 2503 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2504 // sections - we should not make any assumptions about them based on that. 2505 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2506 m_allow_assembly_emulation_unwind_plans = false; 2507 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2508 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2509 2510 if (unwind_or_symbol_log) 2511 module_sp->LogMessage( 2512 unwind_or_symbol_log, 2513 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2514 } 2515 2516 const user_id_t TEXT_eh_frame_sectID = 2517 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2518 : static_cast<user_id_t>(NO_SECT); 2519 2520 lldb::offset_t nlist_data_offset = 0; 2521 2522 uint32_t N_SO_index = UINT32_MAX; 2523 2524 MachSymtabSectionInfo section_info(section_list); 2525 std::vector<uint32_t> N_FUN_indexes; 2526 std::vector<uint32_t> N_NSYM_indexes; 2527 std::vector<uint32_t> N_INCL_indexes; 2528 std::vector<uint32_t> N_BRAC_indexes; 2529 std::vector<uint32_t> N_COMM_indexes; 2530 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2531 typedef std::map<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2532 typedef std::map<const char *, uint32_t> ConstNameToSymbolIndexMap; 2533 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2534 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2535 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2536 // Any symbols that get merged into another will get an entry in this map 2537 // so we know 2538 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2539 uint32_t nlist_idx = 0; 2540 Symbol *symbol_ptr = NULL; 2541 2542 uint32_t sym_idx = 0; 2543 Symbol *sym = NULL; 2544 size_t num_syms = 0; 2545 std::string memory_symbol_name; 2546 uint32_t unmapped_local_symbols_found = 0; 2547 2548 std::vector<TrieEntryWithOffset> trie_entries; 2549 std::set<lldb::addr_t> resolver_addresses; 2550 2551 if (dyld_trie_data.GetByteSize() > 0) { 2552 std::vector<llvm::StringRef> nameSlices; 2553 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, 2554 resolver_addresses, trie_entries); 2555 2556 ConstString text_segment_name("__TEXT"); 2557 SectionSP text_segment_sp = 2558 GetSectionList()->FindSectionByName(text_segment_name); 2559 if (text_segment_sp) { 2560 const lldb::addr_t text_segment_file_addr = 2561 text_segment_sp->GetFileAddress(); 2562 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2563 for (auto &e : trie_entries) 2564 e.entry.address += text_segment_file_addr; 2565 } 2566 } 2567 } 2568 2569 typedef std::set<ConstString> IndirectSymbols; 2570 IndirectSymbols indirect_symbol_names; 2571 2572 #if defined(__APPLE__) && \ 2573 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2574 2575 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2576 // optimized by moving LOCAL symbols out of the memory mapped portion of 2577 // the DSC. The symbol information has all been retained, but it isn't 2578 // available in the normal nlist data. However, there *are* duplicate 2579 // entries of *some* 2580 // LOCAL symbols in the normal nlist data. To handle this situation 2581 // correctly, we must first attempt 2582 // to parse any DSC unmapped symbol information. If we find any, we set a 2583 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2584 2585 if (m_header.flags & 0x80000000u) { 2586 // Before we can start mapping the DSC, we need to make certain the 2587 // target process is actually using the cache we can find. 2588 2589 // Next we need to determine the correct path for the dyld shared cache. 2590 2591 ArchSpec header_arch; 2592 GetArchitecture(header_arch); 2593 char dsc_path[PATH_MAX]; 2594 char dsc_path_development[PATH_MAX]; 2595 2596 snprintf( 2597 dsc_path, sizeof(dsc_path), "%s%s%s", 2598 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2599 */ 2600 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2601 header_arch.GetArchitectureName()); 2602 2603 snprintf( 2604 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2605 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2606 */ 2607 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2608 header_arch.GetArchitectureName(), ".development"); 2609 2610 FileSpec dsc_nondevelopment_filespec(dsc_path, false); 2611 FileSpec dsc_development_filespec(dsc_path_development, false); 2612 FileSpec dsc_filespec; 2613 2614 UUID dsc_uuid; 2615 UUID process_shared_cache_uuid; 2616 addr_t process_shared_cache_base_addr; 2617 2618 if (process) { 2619 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, process_shared_cache_uuid); 2620 } 2621 2622 // First see if we can find an exact match for the inferior process 2623 // shared cache UUID in the development or non-development shared caches 2624 // on disk. 2625 if (process_shared_cache_uuid.IsValid()) { 2626 if (dsc_development_filespec.Exists()) { 2627 UUID dsc_development_uuid = GetSharedCacheUUID( 2628 dsc_development_filespec, byte_order, addr_byte_size); 2629 if (dsc_development_uuid.IsValid() && 2630 dsc_development_uuid == process_shared_cache_uuid) { 2631 dsc_filespec = dsc_development_filespec; 2632 dsc_uuid = dsc_development_uuid; 2633 } 2634 } 2635 if (!dsc_uuid.IsValid() && dsc_nondevelopment_filespec.Exists()) { 2636 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2637 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2638 if (dsc_nondevelopment_uuid.IsValid() && 2639 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2640 dsc_filespec = dsc_nondevelopment_filespec; 2641 dsc_uuid = dsc_nondevelopment_uuid; 2642 } 2643 } 2644 } 2645 2646 // Failing a UUID match, prefer the development dyld_shared cache if both 2647 // are present. 2648 if (!dsc_filespec.Exists()) { 2649 if (dsc_development_filespec.Exists()) { 2650 dsc_filespec = dsc_development_filespec; 2651 } else { 2652 dsc_filespec = dsc_nondevelopment_filespec; 2653 } 2654 } 2655 2656 /* The dyld_cache_header has a pointer to the 2657 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2658 The dyld_cache_local_symbols_info structure gives us three things: 2659 1. The start and count of the nlist records in the dyld_shared_cache 2660 file 2661 2. The start and size of the strings for these nlist records 2662 3. The start and count of dyld_cache_local_symbols_entry entries 2663 2664 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2665 dyld shared cache. 2666 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2667 the dyld shared cache. 2668 The dyld_cache_local_symbols_entry also lists the start of this 2669 dylib/framework's nlist records 2670 and the count of how many nlist records there are for this 2671 dylib/framework. 2672 */ 2673 2674 // Process the dyld shared cache header to find the unmapped symbols 2675 2676 DataBufferSP dsc_data_sp = MapFileData( 2677 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2678 if (!dsc_uuid.IsValid()) { 2679 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2680 } 2681 if (dsc_data_sp) { 2682 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2683 2684 bool uuid_match = true; 2685 if (dsc_uuid.IsValid() && process) { 2686 if (process_shared_cache_uuid.IsValid() && 2687 dsc_uuid != process_shared_cache_uuid) { 2688 // The on-disk dyld_shared_cache file is not the same as the one in 2689 // this process' memory, don't use it. 2690 uuid_match = false; 2691 ModuleSP module_sp(GetModule()); 2692 if (module_sp) 2693 module_sp->ReportWarning("process shared cache does not match " 2694 "on-disk dyld_shared_cache file, some " 2695 "symbol names will be missing."); 2696 } 2697 } 2698 2699 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2700 2701 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2702 2703 // If the mappingOffset points to a location inside the header, we've 2704 // opened an old dyld shared cache, and should not proceed further. 2705 if (uuid_match && 2706 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2707 2708 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2709 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2710 mappingOffset); 2711 2712 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2713 byte_order, addr_byte_size); 2714 offset = 0; 2715 2716 // The File addresses (from the in-memory Mach-O load commands) for 2717 // the shared libraries in the shared library cache need to be 2718 // adjusted by an offset to match up with the dylibOffset identifying 2719 // field in the dyld_cache_local_symbol_entry's. This offset is 2720 // recorded in mapping_offset_value. 2721 const uint64_t mapping_offset_value = 2722 dsc_mapping_info_data.GetU64(&offset); 2723 2724 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, 2725 localSymbolsOffset); 2726 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2727 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2728 2729 if (localSymbolsOffset && localSymbolsSize) { 2730 // Map the local symbols 2731 DataBufferSP dsc_local_symbols_data_sp = 2732 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2733 2734 if (dsc_local_symbols_data_sp) { 2735 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2736 byte_order, addr_byte_size); 2737 2738 offset = 0; 2739 2740 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2741 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2742 UndefinedNameToDescMap undefined_name_to_desc; 2743 SymbolIndexToName reexport_shlib_needs_fixup; 2744 2745 // Read the local_symbols_infos struct in one shot 2746 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2747 dsc_local_symbols_data.GetU32(&offset, 2748 &local_symbols_info.nlistOffset, 6); 2749 2750 SectionSP text_section_sp( 2751 section_list->FindSectionByName(GetSegmentNameTEXT())); 2752 2753 uint32_t header_file_offset = 2754 (text_section_sp->GetFileAddress() - mapping_offset_value); 2755 2756 offset = local_symbols_info.entriesOffset; 2757 for (uint32_t entry_index = 0; 2758 entry_index < local_symbols_info.entriesCount; 2759 entry_index++) { 2760 struct lldb_copy_dyld_cache_local_symbols_entry 2761 local_symbols_entry; 2762 local_symbols_entry.dylibOffset = 2763 dsc_local_symbols_data.GetU32(&offset); 2764 local_symbols_entry.nlistStartIndex = 2765 dsc_local_symbols_data.GetU32(&offset); 2766 local_symbols_entry.nlistCount = 2767 dsc_local_symbols_data.GetU32(&offset); 2768 2769 if (header_file_offset == local_symbols_entry.dylibOffset) { 2770 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2771 2772 // The normal nlist code cannot correctly size the Symbols 2773 // array, we need to allocate it here. 2774 sym = symtab->Resize( 2775 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2776 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2777 num_syms = symtab->GetNumSymbols(); 2778 2779 nlist_data_offset = 2780 local_symbols_info.nlistOffset + 2781 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2782 uint32_t string_table_offset = 2783 local_symbols_info.stringsOffset; 2784 2785 for (uint32_t nlist_index = 0; 2786 nlist_index < local_symbols_entry.nlistCount; 2787 nlist_index++) { 2788 ///////////////////////////// 2789 { 2790 struct nlist_64 nlist; 2791 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize( 2792 nlist_data_offset, nlist_byte_size)) 2793 break; 2794 2795 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked( 2796 &nlist_data_offset); 2797 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked( 2798 &nlist_data_offset); 2799 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked( 2800 &nlist_data_offset); 2801 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked( 2802 &nlist_data_offset); 2803 nlist.n_value = 2804 dsc_local_symbols_data.GetAddress_unchecked( 2805 &nlist_data_offset); 2806 2807 SymbolType type = eSymbolTypeInvalid; 2808 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2809 string_table_offset + nlist.n_strx); 2810 2811 if (symbol_name == NULL) { 2812 // No symbol should be NULL, even the symbols with no 2813 // string values should have an offset zero which 2814 // points to an empty C-string 2815 Host::SystemLog( 2816 Host::eSystemLogError, 2817 "error: DSC unmapped local symbol[%u] has invalid " 2818 "string table offset 0x%x in %s, ignoring symbol\n", 2819 entry_index, nlist.n_strx, 2820 module_sp->GetFileSpec().GetPath().c_str()); 2821 continue; 2822 } 2823 if (symbol_name[0] == '\0') 2824 symbol_name = NULL; 2825 2826 const char *symbol_name_non_abi_mangled = NULL; 2827 2828 SectionSP symbol_section; 2829 uint32_t symbol_byte_size = 0; 2830 bool add_nlist = true; 2831 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2832 bool demangled_is_synthesized = false; 2833 bool is_gsym = false; 2834 bool set_value = true; 2835 2836 assert(sym_idx < num_syms); 2837 2838 sym[sym_idx].SetDebug(is_debug); 2839 2840 if (is_debug) { 2841 switch (nlist.n_type) { 2842 case N_GSYM: 2843 // global symbol: name,,NO_SECT,type,0 2844 // Sometimes the N_GSYM value contains the address. 2845 2846 // FIXME: In the .o files, we have a GSYM and a debug 2847 // symbol for all the ObjC data. They 2848 // have the same address, but we want to ensure that 2849 // we always find only the real symbol, 'cause we 2850 // don't currently correctly attribute the 2851 // GSYM one to the ObjCClass/Ivar/MetaClass 2852 // symbol type. This is a temporary hack to make 2853 // sure the ObjectiveC symbols get treated correctly. 2854 // To do this right, we should coalesce all the GSYM 2855 // & global symbols that have the same address. 2856 2857 is_gsym = true; 2858 sym[sym_idx].SetExternal(true); 2859 2860 if (symbol_name && symbol_name[0] == '_' && 2861 symbol_name[1] == 'O') { 2862 llvm::StringRef symbol_name_ref(symbol_name); 2863 if (symbol_name_ref.startswith( 2864 g_objc_v2_prefix_class)) { 2865 symbol_name_non_abi_mangled = symbol_name + 1; 2866 symbol_name = 2867 symbol_name + g_objc_v2_prefix_class.size(); 2868 type = eSymbolTypeObjCClass; 2869 demangled_is_synthesized = true; 2870 2871 } else if (symbol_name_ref.startswith( 2872 g_objc_v2_prefix_metaclass)) { 2873 symbol_name_non_abi_mangled = symbol_name + 1; 2874 symbol_name = symbol_name + 2875 g_objc_v2_prefix_metaclass.size(); 2876 type = eSymbolTypeObjCMetaClass; 2877 demangled_is_synthesized = true; 2878 } else if (symbol_name_ref.startswith( 2879 g_objc_v2_prefix_ivar)) { 2880 symbol_name_non_abi_mangled = symbol_name + 1; 2881 symbol_name = 2882 symbol_name + g_objc_v2_prefix_ivar.size(); 2883 type = eSymbolTypeObjCIVar; 2884 demangled_is_synthesized = true; 2885 } 2886 } else { 2887 if (nlist.n_value != 0) 2888 symbol_section = section_info.GetSection( 2889 nlist.n_sect, nlist.n_value); 2890 type = eSymbolTypeData; 2891 } 2892 break; 2893 2894 case N_FNAME: 2895 // procedure name (f77 kludge): name,,NO_SECT,0,0 2896 type = eSymbolTypeCompiler; 2897 break; 2898 2899 case N_FUN: 2900 // procedure: name,,n_sect,linenumber,address 2901 if (symbol_name) { 2902 type = eSymbolTypeCode; 2903 symbol_section = section_info.GetSection( 2904 nlist.n_sect, nlist.n_value); 2905 2906 N_FUN_addr_to_sym_idx.insert( 2907 std::make_pair(nlist.n_value, sym_idx)); 2908 // We use the current number of symbols in the 2909 // symbol table in lieu of using nlist_idx in case 2910 // we ever start trimming entries out 2911 N_FUN_indexes.push_back(sym_idx); 2912 } else { 2913 type = eSymbolTypeCompiler; 2914 2915 if (!N_FUN_indexes.empty()) { 2916 // Copy the size of the function into the 2917 // original 2918 // STAB entry so we don't have 2919 // to hunt for it later 2920 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2921 ->SetByteSize(nlist.n_value); 2922 N_FUN_indexes.pop_back(); 2923 // We don't really need the end function STAB as 2924 // it contains the size which we already placed 2925 // with the original symbol, so don't add it if 2926 // we want a minimal symbol table 2927 add_nlist = false; 2928 } 2929 } 2930 break; 2931 2932 case N_STSYM: 2933 // static symbol: name,,n_sect,type,address 2934 N_STSYM_addr_to_sym_idx.insert( 2935 std::make_pair(nlist.n_value, sym_idx)); 2936 symbol_section = section_info.GetSection( 2937 nlist.n_sect, nlist.n_value); 2938 if (symbol_name && symbol_name[0]) { 2939 type = ObjectFile::GetSymbolTypeFromName( 2940 symbol_name + 1, eSymbolTypeData); 2941 } 2942 break; 2943 2944 case N_LCSYM: 2945 // .lcomm symbol: name,,n_sect,type,address 2946 symbol_section = section_info.GetSection( 2947 nlist.n_sect, nlist.n_value); 2948 type = eSymbolTypeCommonBlock; 2949 break; 2950 2951 case N_BNSYM: 2952 // We use the current number of symbols in the symbol 2953 // table in lieu of using nlist_idx in case we ever 2954 // start trimming entries out Skip these if we want 2955 // minimal symbol tables 2956 add_nlist = false; 2957 break; 2958 2959 case N_ENSYM: 2960 // Set the size of the N_BNSYM to the terminating 2961 // index of this N_ENSYM so that we can always skip 2962 // the entire symbol if we need to navigate more 2963 // quickly at the source level when parsing STABS 2964 // Skip these if we want minimal symbol tables 2965 add_nlist = false; 2966 break; 2967 2968 case N_OPT: 2969 // emitted with gcc2_compiled and in gcc source 2970 type = eSymbolTypeCompiler; 2971 break; 2972 2973 case N_RSYM: 2974 // register sym: name,,NO_SECT,type,register 2975 type = eSymbolTypeVariable; 2976 break; 2977 2978 case N_SLINE: 2979 // src line: 0,,n_sect,linenumber,address 2980 symbol_section = section_info.GetSection( 2981 nlist.n_sect, nlist.n_value); 2982 type = eSymbolTypeLineEntry; 2983 break; 2984 2985 case N_SSYM: 2986 // structure elt: name,,NO_SECT,type,struct_offset 2987 type = eSymbolTypeVariableType; 2988 break; 2989 2990 case N_SO: 2991 // source file name 2992 type = eSymbolTypeSourceFile; 2993 if (symbol_name == NULL) { 2994 add_nlist = false; 2995 if (N_SO_index != UINT32_MAX) { 2996 // Set the size of the N_SO to the terminating 2997 // index of this N_SO so that we can always skip 2998 // the entire N_SO if we need to navigate more 2999 // quickly at the source level when parsing STABS 3000 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3001 symbol_ptr->SetByteSize(sym_idx); 3002 symbol_ptr->SetSizeIsSibling(true); 3003 } 3004 N_NSYM_indexes.clear(); 3005 N_INCL_indexes.clear(); 3006 N_BRAC_indexes.clear(); 3007 N_COMM_indexes.clear(); 3008 N_FUN_indexes.clear(); 3009 N_SO_index = UINT32_MAX; 3010 } else { 3011 // We use the current number of symbols in the 3012 // symbol table in lieu of using nlist_idx in case 3013 // we ever start trimming entries out 3014 const bool N_SO_has_full_path = 3015 symbol_name[0] == '/'; 3016 if (N_SO_has_full_path) { 3017 if ((N_SO_index == sym_idx - 1) && 3018 ((sym_idx - 1) < num_syms)) { 3019 // We have two consecutive N_SO entries where 3020 // the first contains a directory and the 3021 // second contains a full path. 3022 sym[sym_idx - 1].GetMangled().SetValue( 3023 ConstString(symbol_name), false); 3024 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3025 add_nlist = false; 3026 } else { 3027 // This is the first entry in a N_SO that 3028 // contains a directory or 3029 // a full path to the source file 3030 N_SO_index = sym_idx; 3031 } 3032 } else if ((N_SO_index == sym_idx - 1) && 3033 ((sym_idx - 1) < num_syms)) { 3034 // This is usually the second N_SO entry that 3035 // contains just the filename, so here we combine 3036 // it with the first one if we are minimizing the 3037 // symbol table 3038 const char *so_path = 3039 sym[sym_idx - 1] 3040 .GetMangled() 3041 .GetDemangledName( 3042 lldb::eLanguageTypeUnknown) 3043 .AsCString(); 3044 if (so_path && so_path[0]) { 3045 std::string full_so_path(so_path); 3046 const size_t double_slash_pos = 3047 full_so_path.find("//"); 3048 if (double_slash_pos != std::string::npos) { 3049 // The linker has been generating bad N_SO 3050 // entries with doubled up paths 3051 // in the format "%s%s" where the first 3052 // string in the DW_AT_comp_dir, and the 3053 // second is the directory for the source 3054 // file so you end up with a path that looks 3055 // like "/tmp/src//tmp/src/" 3056 FileSpec so_dir(so_path, false); 3057 if (!so_dir.Exists()) { 3058 so_dir.SetFile( 3059 &full_so_path[double_slash_pos + 1], 3060 false); 3061 if (so_dir.Exists()) { 3062 // Trim off the incorrect path 3063 full_so_path.erase(0, 3064 double_slash_pos + 1); 3065 } 3066 } 3067 } 3068 if (*full_so_path.rbegin() != '/') 3069 full_so_path += '/'; 3070 full_so_path += symbol_name; 3071 sym[sym_idx - 1].GetMangled().SetValue( 3072 ConstString(full_so_path.c_str()), false); 3073 add_nlist = false; 3074 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3075 } 3076 } else { 3077 // This could be a relative path to a N_SO 3078 N_SO_index = sym_idx; 3079 } 3080 } 3081 break; 3082 3083 case N_OSO: 3084 // object file name: name,,0,0,st_mtime 3085 type = eSymbolTypeObjectFile; 3086 break; 3087 3088 case N_LSYM: 3089 // local sym: name,,NO_SECT,type,offset 3090 type = eSymbolTypeLocal; 3091 break; 3092 3093 //---------------------------------------------------------------------- 3094 // INCL scopes 3095 //---------------------------------------------------------------------- 3096 case N_BINCL: 3097 // include file beginning: name,,NO_SECT,0,sum We use 3098 // the current number of symbols in the symbol table 3099 // in lieu of using nlist_idx in case we ever start 3100 // trimming entries out 3101 N_INCL_indexes.push_back(sym_idx); 3102 type = eSymbolTypeScopeBegin; 3103 break; 3104 3105 case N_EINCL: 3106 // include file end: name,,NO_SECT,0,0 3107 // Set the size of the N_BINCL to the terminating 3108 // index of this N_EINCL so that we can always skip 3109 // the entire symbol if we need to navigate more 3110 // quickly at the source level when parsing STABS 3111 if (!N_INCL_indexes.empty()) { 3112 symbol_ptr = 3113 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3114 symbol_ptr->SetByteSize(sym_idx + 1); 3115 symbol_ptr->SetSizeIsSibling(true); 3116 N_INCL_indexes.pop_back(); 3117 } 3118 type = eSymbolTypeScopeEnd; 3119 break; 3120 3121 case N_SOL: 3122 // #included file name: name,,n_sect,0,address 3123 type = eSymbolTypeHeaderFile; 3124 3125 // We currently don't use the header files on darwin 3126 add_nlist = false; 3127 break; 3128 3129 case N_PARAMS: 3130 // compiler parameters: name,,NO_SECT,0,0 3131 type = eSymbolTypeCompiler; 3132 break; 3133 3134 case N_VERSION: 3135 // compiler version: name,,NO_SECT,0,0 3136 type = eSymbolTypeCompiler; 3137 break; 3138 3139 case N_OLEVEL: 3140 // compiler -O level: name,,NO_SECT,0,0 3141 type = eSymbolTypeCompiler; 3142 break; 3143 3144 case N_PSYM: 3145 // parameter: name,,NO_SECT,type,offset 3146 type = eSymbolTypeVariable; 3147 break; 3148 3149 case N_ENTRY: 3150 // alternate entry: name,,n_sect,linenumber,address 3151 symbol_section = section_info.GetSection( 3152 nlist.n_sect, nlist.n_value); 3153 type = eSymbolTypeLineEntry; 3154 break; 3155 3156 //---------------------------------------------------------------------- 3157 // Left and Right Braces 3158 //---------------------------------------------------------------------- 3159 case N_LBRAC: 3160 // left bracket: 0,,NO_SECT,nesting level,address We 3161 // use the current number of symbols in the symbol 3162 // table in lieu of using nlist_idx in case we ever 3163 // start trimming entries out 3164 symbol_section = section_info.GetSection( 3165 nlist.n_sect, nlist.n_value); 3166 N_BRAC_indexes.push_back(sym_idx); 3167 type = eSymbolTypeScopeBegin; 3168 break; 3169 3170 case N_RBRAC: 3171 // right bracket: 0,,NO_SECT,nesting level,address 3172 // Set the size of the N_LBRAC to the terminating 3173 // index of this N_RBRAC so that we can always skip 3174 // the entire symbol if we need to navigate more 3175 // quickly at the source level when parsing STABS 3176 symbol_section = section_info.GetSection( 3177 nlist.n_sect, nlist.n_value); 3178 if (!N_BRAC_indexes.empty()) { 3179 symbol_ptr = 3180 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3181 symbol_ptr->SetByteSize(sym_idx + 1); 3182 symbol_ptr->SetSizeIsSibling(true); 3183 N_BRAC_indexes.pop_back(); 3184 } 3185 type = eSymbolTypeScopeEnd; 3186 break; 3187 3188 case N_EXCL: 3189 // deleted include file: name,,NO_SECT,0,sum 3190 type = eSymbolTypeHeaderFile; 3191 break; 3192 3193 //---------------------------------------------------------------------- 3194 // COMM scopes 3195 //---------------------------------------------------------------------- 3196 case N_BCOMM: 3197 // begin common: name,,NO_SECT,0,0 3198 // We use the current number of symbols in the symbol 3199 // table in lieu of using nlist_idx in case we ever 3200 // start trimming entries out 3201 type = eSymbolTypeScopeBegin; 3202 N_COMM_indexes.push_back(sym_idx); 3203 break; 3204 3205 case N_ECOML: 3206 // end common (local name): 0,,n_sect,0,address 3207 symbol_section = section_info.GetSection( 3208 nlist.n_sect, nlist.n_value); 3209 // Fall through 3210 3211 case N_ECOMM: 3212 // end common: name,,n_sect,0,0 3213 // Set the size of the N_BCOMM to the terminating 3214 // index of this N_ECOMM/N_ECOML so that we can 3215 // always skip the entire symbol if we need to 3216 // navigate more quickly at the source level when 3217 // parsing STABS 3218 if (!N_COMM_indexes.empty()) { 3219 symbol_ptr = 3220 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3221 symbol_ptr->SetByteSize(sym_idx + 1); 3222 symbol_ptr->SetSizeIsSibling(true); 3223 N_COMM_indexes.pop_back(); 3224 } 3225 type = eSymbolTypeScopeEnd; 3226 break; 3227 3228 case N_LENG: 3229 // second stab entry with length information 3230 type = eSymbolTypeAdditional; 3231 break; 3232 3233 default: 3234 break; 3235 } 3236 } else { 3237 // uint8_t n_pext = N_PEXT & nlist.n_type; 3238 uint8_t n_type = N_TYPE & nlist.n_type; 3239 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3240 3241 switch (n_type) { 3242 case N_INDR: { 3243 const char *reexport_name_cstr = 3244 strtab_data.PeekCStr(nlist.n_value); 3245 if (reexport_name_cstr && reexport_name_cstr[0]) { 3246 type = eSymbolTypeReExported; 3247 ConstString reexport_name( 3248 reexport_name_cstr + 3249 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3250 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3251 set_value = false; 3252 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3253 indirect_symbol_names.insert( 3254 ConstString(symbol_name + 3255 ((symbol_name[0] == '_') ? 1 : 0))); 3256 } else 3257 type = eSymbolTypeUndefined; 3258 } break; 3259 3260 case N_UNDF: 3261 if (symbol_name && symbol_name[0]) { 3262 ConstString undefined_name( 3263 symbol_name + 3264 ((symbol_name[0] == '_') ? 1 : 0)); 3265 undefined_name_to_desc[undefined_name] = 3266 nlist.n_desc; 3267 } 3268 // Fall through 3269 case N_PBUD: 3270 type = eSymbolTypeUndefined; 3271 break; 3272 3273 case N_ABS: 3274 type = eSymbolTypeAbsolute; 3275 break; 3276 3277 case N_SECT: { 3278 symbol_section = section_info.GetSection( 3279 nlist.n_sect, nlist.n_value); 3280 3281 if (symbol_section == NULL) { 3282 // TODO: warn about this? 3283 add_nlist = false; 3284 break; 3285 } 3286 3287 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3288 type = eSymbolTypeException; 3289 } else { 3290 uint32_t section_type = 3291 symbol_section->Get() & SECTION_TYPE; 3292 3293 switch (section_type) { 3294 case S_CSTRING_LITERALS: 3295 type = eSymbolTypeData; 3296 break; // section with only literal C strings 3297 case S_4BYTE_LITERALS: 3298 type = eSymbolTypeData; 3299 break; // section with only 4 byte literals 3300 case S_8BYTE_LITERALS: 3301 type = eSymbolTypeData; 3302 break; // section with only 8 byte literals 3303 case S_LITERAL_POINTERS: 3304 type = eSymbolTypeTrampoline; 3305 break; // section with only pointers to literals 3306 case S_NON_LAZY_SYMBOL_POINTERS: 3307 type = eSymbolTypeTrampoline; 3308 break; // section with only non-lazy symbol 3309 // pointers 3310 case S_LAZY_SYMBOL_POINTERS: 3311 type = eSymbolTypeTrampoline; 3312 break; // section with only lazy symbol pointers 3313 case S_SYMBOL_STUBS: 3314 type = eSymbolTypeTrampoline; 3315 break; // section with only symbol stubs, byte 3316 // size of stub in the reserved2 field 3317 case S_MOD_INIT_FUNC_POINTERS: 3318 type = eSymbolTypeCode; 3319 break; // section with only function pointers for 3320 // initialization 3321 case S_MOD_TERM_FUNC_POINTERS: 3322 type = eSymbolTypeCode; 3323 break; // section with only function pointers for 3324 // termination 3325 case S_INTERPOSING: 3326 type = eSymbolTypeTrampoline; 3327 break; // section with only pairs of function 3328 // pointers for interposing 3329 case S_16BYTE_LITERALS: 3330 type = eSymbolTypeData; 3331 break; // section with only 16 byte literals 3332 case S_DTRACE_DOF: 3333 type = eSymbolTypeInstrumentation; 3334 break; 3335 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3336 type = eSymbolTypeTrampoline; 3337 break; 3338 default: 3339 switch (symbol_section->GetType()) { 3340 case lldb::eSectionTypeCode: 3341 type = eSymbolTypeCode; 3342 break; 3343 case eSectionTypeData: 3344 case eSectionTypeDataCString: // Inlined C string 3345 // data 3346 case eSectionTypeDataCStringPointers: // Pointers 3347 // to C 3348 // string 3349 // data 3350 case eSectionTypeDataSymbolAddress: // Address of 3351 // a symbol in 3352 // the symbol 3353 // table 3354 case eSectionTypeData4: 3355 case eSectionTypeData8: 3356 case eSectionTypeData16: 3357 type = eSymbolTypeData; 3358 break; 3359 default: 3360 break; 3361 } 3362 break; 3363 } 3364 3365 if (type == eSymbolTypeInvalid) { 3366 const char *symbol_sect_name = 3367 symbol_section->GetName().AsCString(); 3368 if (symbol_section->IsDescendant( 3369 text_section_sp.get())) { 3370 if (symbol_section->IsClear( 3371 S_ATTR_PURE_INSTRUCTIONS | 3372 S_ATTR_SELF_MODIFYING_CODE | 3373 S_ATTR_SOME_INSTRUCTIONS)) 3374 type = eSymbolTypeData; 3375 else 3376 type = eSymbolTypeCode; 3377 } else if (symbol_section->IsDescendant( 3378 data_section_sp.get()) || 3379 symbol_section->IsDescendant( 3380 data_dirty_section_sp.get()) || 3381 symbol_section->IsDescendant( 3382 data_const_section_sp.get())) { 3383 if (symbol_sect_name && 3384 ::strstr(symbol_sect_name, "__objc") == 3385 symbol_sect_name) { 3386 type = eSymbolTypeRuntime; 3387 3388 if (symbol_name) { 3389 llvm::StringRef symbol_name_ref( 3390 symbol_name); 3391 if (symbol_name_ref.startswith("_OBJC_")) { 3392 static const llvm::StringRef 3393 g_objc_v2_prefix_class( 3394 "_OBJC_CLASS_$_"); 3395 static const llvm::StringRef 3396 g_objc_v2_prefix_metaclass( 3397 "_OBJC_METACLASS_$_"); 3398 static const llvm::StringRef 3399 g_objc_v2_prefix_ivar( 3400 "_OBJC_IVAR_$_"); 3401 if (symbol_name_ref.startswith( 3402 g_objc_v2_prefix_class)) { 3403 symbol_name_non_abi_mangled = 3404 symbol_name + 1; 3405 symbol_name = 3406 symbol_name + 3407 g_objc_v2_prefix_class.size(); 3408 type = eSymbolTypeObjCClass; 3409 demangled_is_synthesized = true; 3410 } else if ( 3411 symbol_name_ref.startswith( 3412 g_objc_v2_prefix_metaclass)) { 3413 symbol_name_non_abi_mangled = 3414 symbol_name + 1; 3415 symbol_name = 3416 symbol_name + 3417 g_objc_v2_prefix_metaclass.size(); 3418 type = eSymbolTypeObjCMetaClass; 3419 demangled_is_synthesized = true; 3420 } else if (symbol_name_ref.startswith( 3421 g_objc_v2_prefix_ivar)) { 3422 symbol_name_non_abi_mangled = 3423 symbol_name + 1; 3424 symbol_name = 3425 symbol_name + 3426 g_objc_v2_prefix_ivar.size(); 3427 type = eSymbolTypeObjCIVar; 3428 demangled_is_synthesized = true; 3429 } 3430 } 3431 } 3432 } else if (symbol_sect_name && 3433 ::strstr(symbol_sect_name, 3434 "__gcc_except_tab") == 3435 symbol_sect_name) { 3436 type = eSymbolTypeException; 3437 } else { 3438 type = eSymbolTypeData; 3439 } 3440 } else if (symbol_sect_name && 3441 ::strstr(symbol_sect_name, 3442 "__IMPORT") == 3443 symbol_sect_name) { 3444 type = eSymbolTypeTrampoline; 3445 } else if (symbol_section->IsDescendant( 3446 objc_section_sp.get())) { 3447 type = eSymbolTypeRuntime; 3448 if (symbol_name && symbol_name[0] == '.') { 3449 llvm::StringRef symbol_name_ref(symbol_name); 3450 static const llvm::StringRef 3451 g_objc_v1_prefix_class( 3452 ".objc_class_name_"); 3453 if (symbol_name_ref.startswith( 3454 g_objc_v1_prefix_class)) { 3455 symbol_name_non_abi_mangled = symbol_name; 3456 symbol_name = symbol_name + 3457 g_objc_v1_prefix_class.size(); 3458 type = eSymbolTypeObjCClass; 3459 demangled_is_synthesized = true; 3460 } 3461 } 3462 } 3463 } 3464 } 3465 } break; 3466 } 3467 } 3468 3469 if (add_nlist) { 3470 uint64_t symbol_value = nlist.n_value; 3471 if (symbol_name_non_abi_mangled) { 3472 sym[sym_idx].GetMangled().SetMangledName( 3473 ConstString(symbol_name_non_abi_mangled)); 3474 sym[sym_idx].GetMangled().SetDemangledName( 3475 ConstString(symbol_name)); 3476 } else { 3477 bool symbol_name_is_mangled = false; 3478 3479 if (symbol_name && symbol_name[0] == '_') { 3480 symbol_name_is_mangled = symbol_name[1] == '_'; 3481 symbol_name++; // Skip the leading underscore 3482 } 3483 3484 if (symbol_name) { 3485 ConstString const_symbol_name(symbol_name); 3486 sym[sym_idx].GetMangled().SetValue( 3487 const_symbol_name, symbol_name_is_mangled); 3488 if (is_gsym && is_debug) { 3489 const char *gsym_name = 3490 sym[sym_idx] 3491 .GetMangled() 3492 .GetName(lldb::eLanguageTypeUnknown, 3493 Mangled::ePreferMangled) 3494 .GetCString(); 3495 if (gsym_name) 3496 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3497 } 3498 } 3499 } 3500 if (symbol_section) { 3501 const addr_t section_file_addr = 3502 symbol_section->GetFileAddress(); 3503 if (symbol_byte_size == 0 && 3504 function_starts_count > 0) { 3505 addr_t symbol_lookup_file_addr = nlist.n_value; 3506 // Do an exact address match for non-ARM addresses, 3507 // else get the closest since the symbol might be a 3508 // thumb symbol which has an address with bit zero 3509 // set 3510 FunctionStarts::Entry *func_start_entry = 3511 function_starts.FindEntry( 3512 symbol_lookup_file_addr, !is_arm); 3513 if (is_arm && func_start_entry) { 3514 // Verify that the function start address is the 3515 // symbol address (ARM) or the symbol address + 1 3516 // (thumb) 3517 if (func_start_entry->addr != 3518 symbol_lookup_file_addr && 3519 func_start_entry->addr != 3520 (symbol_lookup_file_addr + 1)) { 3521 // Not the right entry, NULL it out... 3522 func_start_entry = NULL; 3523 } 3524 } 3525 if (func_start_entry) { 3526 func_start_entry->data = true; 3527 3528 addr_t symbol_file_addr = func_start_entry->addr; 3529 uint32_t symbol_flags = 0; 3530 if (is_arm) { 3531 if (symbol_file_addr & 1) 3532 symbol_flags = 3533 MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3534 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3535 } 3536 3537 const FunctionStarts::Entry 3538 *next_func_start_entry = 3539 function_starts.FindNextEntry( 3540 func_start_entry); 3541 const addr_t section_end_file_addr = 3542 section_file_addr + 3543 symbol_section->GetByteSize(); 3544 if (next_func_start_entry) { 3545 addr_t next_symbol_file_addr = 3546 next_func_start_entry->addr; 3547 // Be sure the clear the Thumb address bit when 3548 // we calculate the size from the current and 3549 // next address 3550 if (is_arm) 3551 next_symbol_file_addr &= 3552 THUMB_ADDRESS_BIT_MASK; 3553 symbol_byte_size = std::min<lldb::addr_t>( 3554 next_symbol_file_addr - symbol_file_addr, 3555 section_end_file_addr - symbol_file_addr); 3556 } else { 3557 symbol_byte_size = 3558 section_end_file_addr - symbol_file_addr; 3559 } 3560 } 3561 } 3562 symbol_value -= section_file_addr; 3563 } 3564 3565 if (is_debug == false) { 3566 if (type == eSymbolTypeCode) { 3567 // See if we can find a N_FUN entry for any code 3568 // symbols. If we do find a match, and the name 3569 // matches, then we can merge the two into just the 3570 // function symbol to avoid duplicate entries in 3571 // the symbol table 3572 std::pair<ValueToSymbolIndexMap::const_iterator, 3573 ValueToSymbolIndexMap::const_iterator> 3574 range; 3575 range = N_FUN_addr_to_sym_idx.equal_range( 3576 nlist.n_value); 3577 if (range.first != range.second) { 3578 bool found_it = false; 3579 for (ValueToSymbolIndexMap::const_iterator pos = 3580 range.first; 3581 pos != range.second; ++pos) { 3582 if (sym[sym_idx].GetMangled().GetName( 3583 lldb::eLanguageTypeUnknown, 3584 Mangled::ePreferMangled) == 3585 sym[pos->second].GetMangled().GetName( 3586 lldb::eLanguageTypeUnknown, 3587 Mangled::ePreferMangled)) { 3588 m_nlist_idx_to_sym_idx[nlist_idx] = 3589 pos->second; 3590 // We just need the flags from the linker 3591 // symbol, so put these flags 3592 // into the N_FUN flags to avoid duplicate 3593 // symbols in the symbol table 3594 sym[pos->second].SetExternal( 3595 sym[sym_idx].IsExternal()); 3596 sym[pos->second].SetFlags(nlist.n_type << 16 | 3597 nlist.n_desc); 3598 if (resolver_addresses.find(nlist.n_value) != 3599 resolver_addresses.end()) 3600 sym[pos->second].SetType( 3601 eSymbolTypeResolver); 3602 sym[sym_idx].Clear(); 3603 found_it = true; 3604 break; 3605 } 3606 } 3607 if (found_it) 3608 continue; 3609 } else { 3610 if (resolver_addresses.find(nlist.n_value) != 3611 resolver_addresses.end()) 3612 type = eSymbolTypeResolver; 3613 } 3614 } else if (type == eSymbolTypeData || 3615 type == eSymbolTypeObjCClass || 3616 type == eSymbolTypeObjCMetaClass || 3617 type == eSymbolTypeObjCIVar) { 3618 // See if we can find a N_STSYM entry for any data 3619 // symbols. If we do find a match, and the name 3620 // matches, then we can merge the two into just the 3621 // Static symbol to avoid duplicate entries in the 3622 // symbol table 3623 std::pair<ValueToSymbolIndexMap::const_iterator, 3624 ValueToSymbolIndexMap::const_iterator> 3625 range; 3626 range = N_STSYM_addr_to_sym_idx.equal_range( 3627 nlist.n_value); 3628 if (range.first != range.second) { 3629 bool found_it = false; 3630 for (ValueToSymbolIndexMap::const_iterator pos = 3631 range.first; 3632 pos != range.second; ++pos) { 3633 if (sym[sym_idx].GetMangled().GetName( 3634 lldb::eLanguageTypeUnknown, 3635 Mangled::ePreferMangled) == 3636 sym[pos->second].GetMangled().GetName( 3637 lldb::eLanguageTypeUnknown, 3638 Mangled::ePreferMangled)) { 3639 m_nlist_idx_to_sym_idx[nlist_idx] = 3640 pos->second; 3641 // We just need the flags from the linker 3642 // symbol, so put these flags 3643 // into the N_STSYM flags to avoid duplicate 3644 // symbols in the symbol table 3645 sym[pos->second].SetExternal( 3646 sym[sym_idx].IsExternal()); 3647 sym[pos->second].SetFlags(nlist.n_type << 16 | 3648 nlist.n_desc); 3649 sym[sym_idx].Clear(); 3650 found_it = true; 3651 break; 3652 } 3653 } 3654 if (found_it) 3655 continue; 3656 } else { 3657 const char *gsym_name = 3658 sym[sym_idx] 3659 .GetMangled() 3660 .GetName(lldb::eLanguageTypeUnknown, 3661 Mangled::ePreferMangled) 3662 .GetCString(); 3663 if (gsym_name) { 3664 // Combine N_GSYM stab entries with the non 3665 // stab symbol 3666 ConstNameToSymbolIndexMap::const_iterator pos = 3667 N_GSYM_name_to_sym_idx.find(gsym_name); 3668 if (pos != N_GSYM_name_to_sym_idx.end()) { 3669 const uint32_t GSYM_sym_idx = pos->second; 3670 m_nlist_idx_to_sym_idx[nlist_idx] = 3671 GSYM_sym_idx; 3672 // Copy the address, because often the N_GSYM 3673 // address has an invalid address of zero 3674 // when the global is a common symbol 3675 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3676 symbol_section); 3677 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3678 symbol_value); 3679 // We just need the flags from the linker 3680 // symbol, so put these flags 3681 // into the N_GSYM flags to avoid duplicate 3682 // symbols in the symbol table 3683 sym[GSYM_sym_idx].SetFlags( 3684 nlist.n_type << 16 | nlist.n_desc); 3685 sym[sym_idx].Clear(); 3686 continue; 3687 } 3688 } 3689 } 3690 } 3691 } 3692 3693 sym[sym_idx].SetID(nlist_idx); 3694 sym[sym_idx].SetType(type); 3695 if (set_value) { 3696 sym[sym_idx].GetAddressRef().SetSection( 3697 symbol_section); 3698 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3699 } 3700 sym[sym_idx].SetFlags(nlist.n_type << 16 | 3701 nlist.n_desc); 3702 3703 if (symbol_byte_size > 0) 3704 sym[sym_idx].SetByteSize(symbol_byte_size); 3705 3706 if (demangled_is_synthesized) 3707 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3708 ++sym_idx; 3709 } else { 3710 sym[sym_idx].Clear(); 3711 } 3712 } 3713 ///////////////////////////// 3714 } 3715 break; // No more entries to consider 3716 } 3717 } 3718 3719 for (const auto &pos : reexport_shlib_needs_fixup) { 3720 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3721 if (undef_pos != undefined_name_to_desc.end()) { 3722 const uint8_t dylib_ordinal = 3723 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3724 if (dylib_ordinal > 0 && 3725 dylib_ordinal < dylib_files.GetSize()) 3726 sym[pos.first].SetReExportedSymbolSharedLibrary( 3727 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3728 } 3729 } 3730 } 3731 } 3732 } 3733 } 3734 } 3735 3736 // Must reset this in case it was mutated above! 3737 nlist_data_offset = 0; 3738 #endif 3739 3740 if (nlist_data.GetByteSize() > 0) { 3741 3742 // If the sym array was not created while parsing the DSC unmapped 3743 // symbols, create it now. 3744 if (sym == NULL) { 3745 sym = symtab->Resize(symtab_load_command.nsyms + 3746 m_dysymtab.nindirectsyms); 3747 num_syms = symtab->GetNumSymbols(); 3748 } 3749 3750 if (unmapped_local_symbols_found) { 3751 assert(m_dysymtab.ilocalsym == 0); 3752 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3753 nlist_idx = m_dysymtab.nlocalsym; 3754 } else { 3755 nlist_idx = 0; 3756 } 3757 3758 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3759 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3760 UndefinedNameToDescMap undefined_name_to_desc; 3761 SymbolIndexToName reexport_shlib_needs_fixup; 3762 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 3763 struct nlist_64 nlist; 3764 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, 3765 nlist_byte_size)) 3766 break; 3767 3768 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3769 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 3770 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 3771 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 3772 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 3773 3774 SymbolType type = eSymbolTypeInvalid; 3775 const char *symbol_name = NULL; 3776 3777 if (have_strtab_data) { 3778 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3779 3780 if (symbol_name == NULL) { 3781 // No symbol should be NULL, even the symbols with no string values 3782 // should have an offset zero which points to an empty C-string 3783 Host::SystemLog(Host::eSystemLogError, 3784 "error: symbol[%u] has invalid string table offset " 3785 "0x%x in %s, ignoring symbol\n", 3786 nlist_idx, nlist.n_strx, 3787 module_sp->GetFileSpec().GetPath().c_str()); 3788 continue; 3789 } 3790 if (symbol_name[0] == '\0') 3791 symbol_name = NULL; 3792 } else { 3793 const addr_t str_addr = strtab_addr + nlist.n_strx; 3794 Status str_error; 3795 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3796 str_error)) 3797 symbol_name = memory_symbol_name.c_str(); 3798 } 3799 const char *symbol_name_non_abi_mangled = NULL; 3800 3801 SectionSP symbol_section; 3802 lldb::addr_t symbol_byte_size = 0; 3803 bool add_nlist = true; 3804 bool is_gsym = false; 3805 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3806 bool demangled_is_synthesized = false; 3807 bool set_value = true; 3808 assert(sym_idx < num_syms); 3809 3810 sym[sym_idx].SetDebug(is_debug); 3811 3812 if (is_debug) { 3813 switch (nlist.n_type) { 3814 case N_GSYM: 3815 // global symbol: name,,NO_SECT,type,0 3816 // Sometimes the N_GSYM value contains the address. 3817 3818 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3819 // the ObjC data. They 3820 // have the same address, but we want to ensure that we always find 3821 // only the real symbol, 'cause we don't currently correctly 3822 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3823 // type. This is a temporary hack to make sure the ObjectiveC 3824 // symbols get treated correctly. To do this right, we should 3825 // coalesce all the GSYM & global symbols that have the same 3826 // address. 3827 is_gsym = true; 3828 sym[sym_idx].SetExternal(true); 3829 3830 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3831 llvm::StringRef symbol_name_ref(symbol_name); 3832 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3833 symbol_name_non_abi_mangled = symbol_name + 1; 3834 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3835 type = eSymbolTypeObjCClass; 3836 demangled_is_synthesized = true; 3837 3838 } else if (symbol_name_ref.startswith( 3839 g_objc_v2_prefix_metaclass)) { 3840 symbol_name_non_abi_mangled = symbol_name + 1; 3841 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3842 type = eSymbolTypeObjCMetaClass; 3843 demangled_is_synthesized = true; 3844 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3845 symbol_name_non_abi_mangled = symbol_name + 1; 3846 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3847 type = eSymbolTypeObjCIVar; 3848 demangled_is_synthesized = true; 3849 } 3850 } else { 3851 if (nlist.n_value != 0) 3852 symbol_section = 3853 section_info.GetSection(nlist.n_sect, nlist.n_value); 3854 type = eSymbolTypeData; 3855 } 3856 break; 3857 3858 case N_FNAME: 3859 // procedure name (f77 kludge): name,,NO_SECT,0,0 3860 type = eSymbolTypeCompiler; 3861 break; 3862 3863 case N_FUN: 3864 // procedure: name,,n_sect,linenumber,address 3865 if (symbol_name) { 3866 type = eSymbolTypeCode; 3867 symbol_section = 3868 section_info.GetSection(nlist.n_sect, nlist.n_value); 3869 3870 N_FUN_addr_to_sym_idx.insert( 3871 std::make_pair(nlist.n_value, sym_idx)); 3872 // We use the current number of symbols in the symbol table in 3873 // lieu of using nlist_idx in case we ever start trimming entries 3874 // out 3875 N_FUN_indexes.push_back(sym_idx); 3876 } else { 3877 type = eSymbolTypeCompiler; 3878 3879 if (!N_FUN_indexes.empty()) { 3880 // Copy the size of the function into the original STAB entry 3881 // so we don't have to hunt for it later 3882 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3883 ->SetByteSize(nlist.n_value); 3884 N_FUN_indexes.pop_back(); 3885 // We don't really need the end function STAB as it contains 3886 // the size which we already placed with the original symbol, 3887 // so don't add it if we want a minimal symbol table 3888 add_nlist = false; 3889 } 3890 } 3891 break; 3892 3893 case N_STSYM: 3894 // static symbol: name,,n_sect,type,address 3895 N_STSYM_addr_to_sym_idx.insert( 3896 std::make_pair(nlist.n_value, sym_idx)); 3897 symbol_section = 3898 section_info.GetSection(nlist.n_sect, nlist.n_value); 3899 if (symbol_name && symbol_name[0]) { 3900 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3901 eSymbolTypeData); 3902 } 3903 break; 3904 3905 case N_LCSYM: 3906 // .lcomm symbol: name,,n_sect,type,address 3907 symbol_section = 3908 section_info.GetSection(nlist.n_sect, nlist.n_value); 3909 type = eSymbolTypeCommonBlock; 3910 break; 3911 3912 case N_BNSYM: 3913 // We use the current number of symbols in the symbol table in lieu 3914 // of using nlist_idx in case we ever start trimming entries out 3915 // Skip these if we want minimal symbol tables 3916 add_nlist = false; 3917 break; 3918 3919 case N_ENSYM: 3920 // Set the size of the N_BNSYM to the terminating index of this 3921 // N_ENSYM so that we can always skip the entire symbol if we need 3922 // to navigate more quickly at the source level when parsing STABS 3923 // Skip these if we want minimal symbol tables 3924 add_nlist = false; 3925 break; 3926 3927 case N_OPT: 3928 // emitted with gcc2_compiled and in gcc source 3929 type = eSymbolTypeCompiler; 3930 break; 3931 3932 case N_RSYM: 3933 // register sym: name,,NO_SECT,type,register 3934 type = eSymbolTypeVariable; 3935 break; 3936 3937 case N_SLINE: 3938 // src line: 0,,n_sect,linenumber,address 3939 symbol_section = 3940 section_info.GetSection(nlist.n_sect, nlist.n_value); 3941 type = eSymbolTypeLineEntry; 3942 break; 3943 3944 case N_SSYM: 3945 // structure elt: name,,NO_SECT,type,struct_offset 3946 type = eSymbolTypeVariableType; 3947 break; 3948 3949 case N_SO: 3950 // source file name 3951 type = eSymbolTypeSourceFile; 3952 if (symbol_name == NULL) { 3953 add_nlist = false; 3954 if (N_SO_index != UINT32_MAX) { 3955 // Set the size of the N_SO to the terminating index of this 3956 // N_SO so that we can always skip the entire N_SO if we need 3957 // to navigate more quickly at the source level when parsing 3958 // STABS 3959 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3960 symbol_ptr->SetByteSize(sym_idx); 3961 symbol_ptr->SetSizeIsSibling(true); 3962 } 3963 N_NSYM_indexes.clear(); 3964 N_INCL_indexes.clear(); 3965 N_BRAC_indexes.clear(); 3966 N_COMM_indexes.clear(); 3967 N_FUN_indexes.clear(); 3968 N_SO_index = UINT32_MAX; 3969 } else { 3970 // We use the current number of symbols in the symbol table in 3971 // lieu of using nlist_idx in case we ever start trimming entries 3972 // out 3973 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3974 if (N_SO_has_full_path) { 3975 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3976 // We have two consecutive N_SO entries where the first 3977 // contains a directory and the second contains a full path. 3978 sym[sym_idx - 1].GetMangled().SetValue( 3979 ConstString(symbol_name), false); 3980 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3981 add_nlist = false; 3982 } else { 3983 // This is the first entry in a N_SO that contains a 3984 // directory or a full path to the source file 3985 N_SO_index = sym_idx; 3986 } 3987 } else if ((N_SO_index == sym_idx - 1) && 3988 ((sym_idx - 1) < num_syms)) { 3989 // This is usually the second N_SO entry that contains just the 3990 // filename, so here we combine it with the first one if we are 3991 // minimizing the symbol table 3992 const char *so_path = 3993 sym[sym_idx - 1] 3994 .GetMangled() 3995 .GetDemangledName(lldb::eLanguageTypeUnknown) 3996 .AsCString(); 3997 if (so_path && so_path[0]) { 3998 std::string full_so_path(so_path); 3999 const size_t double_slash_pos = full_so_path.find("//"); 4000 if (double_slash_pos != std::string::npos) { 4001 // The linker has been generating bad N_SO entries with 4002 // doubled up paths in the format "%s%s" where the first 4003 // string in the DW_AT_comp_dir, and the second is the 4004 // directory for the source file so you end up with a path 4005 // that looks like "/tmp/src//tmp/src/" 4006 FileSpec so_dir(so_path, false); 4007 if (!so_dir.Exists()) { 4008 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 4009 false); 4010 if (so_dir.Exists()) { 4011 // Trim off the incorrect path 4012 full_so_path.erase(0, double_slash_pos + 1); 4013 } 4014 } 4015 } 4016 if (*full_so_path.rbegin() != '/') 4017 full_so_path += '/'; 4018 full_so_path += symbol_name; 4019 sym[sym_idx - 1].GetMangled().SetValue( 4020 ConstString(full_so_path.c_str()), false); 4021 add_nlist = false; 4022 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4023 } 4024 } else { 4025 // This could be a relative path to a N_SO 4026 N_SO_index = sym_idx; 4027 } 4028 } 4029 break; 4030 4031 case N_OSO: 4032 // object file name: name,,0,0,st_mtime 4033 type = eSymbolTypeObjectFile; 4034 break; 4035 4036 case N_LSYM: 4037 // local sym: name,,NO_SECT,type,offset 4038 type = eSymbolTypeLocal; 4039 break; 4040 4041 //---------------------------------------------------------------------- 4042 // INCL scopes 4043 //---------------------------------------------------------------------- 4044 case N_BINCL: 4045 // include file beginning: name,,NO_SECT,0,sum We use the current 4046 // number of symbols in the symbol table in lieu of using nlist_idx 4047 // in case we ever start trimming entries out 4048 N_INCL_indexes.push_back(sym_idx); 4049 type = eSymbolTypeScopeBegin; 4050 break; 4051 4052 case N_EINCL: 4053 // include file end: name,,NO_SECT,0,0 4054 // Set the size of the N_BINCL to the terminating index of this 4055 // N_EINCL so that we can always skip the entire symbol if we need 4056 // to navigate more quickly at the source level when parsing STABS 4057 if (!N_INCL_indexes.empty()) { 4058 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4059 symbol_ptr->SetByteSize(sym_idx + 1); 4060 symbol_ptr->SetSizeIsSibling(true); 4061 N_INCL_indexes.pop_back(); 4062 } 4063 type = eSymbolTypeScopeEnd; 4064 break; 4065 4066 case N_SOL: 4067 // #included file name: name,,n_sect,0,address 4068 type = eSymbolTypeHeaderFile; 4069 4070 // We currently don't use the header files on darwin 4071 add_nlist = false; 4072 break; 4073 4074 case N_PARAMS: 4075 // compiler parameters: name,,NO_SECT,0,0 4076 type = eSymbolTypeCompiler; 4077 break; 4078 4079 case N_VERSION: 4080 // compiler version: name,,NO_SECT,0,0 4081 type = eSymbolTypeCompiler; 4082 break; 4083 4084 case N_OLEVEL: 4085 // compiler -O level: name,,NO_SECT,0,0 4086 type = eSymbolTypeCompiler; 4087 break; 4088 4089 case N_PSYM: 4090 // parameter: name,,NO_SECT,type,offset 4091 type = eSymbolTypeVariable; 4092 break; 4093 4094 case N_ENTRY: 4095 // alternate entry: name,,n_sect,linenumber,address 4096 symbol_section = 4097 section_info.GetSection(nlist.n_sect, nlist.n_value); 4098 type = eSymbolTypeLineEntry; 4099 break; 4100 4101 //---------------------------------------------------------------------- 4102 // Left and Right Braces 4103 //---------------------------------------------------------------------- 4104 case N_LBRAC: 4105 // left bracket: 0,,NO_SECT,nesting level,address We use the 4106 // current number of symbols in the symbol table in lieu of using 4107 // nlist_idx in case we ever start trimming entries out 4108 symbol_section = 4109 section_info.GetSection(nlist.n_sect, nlist.n_value); 4110 N_BRAC_indexes.push_back(sym_idx); 4111 type = eSymbolTypeScopeBegin; 4112 break; 4113 4114 case N_RBRAC: 4115 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4116 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4117 // can always skip the entire symbol if we need to navigate more 4118 // quickly at the source level when parsing STABS 4119 symbol_section = 4120 section_info.GetSection(nlist.n_sect, nlist.n_value); 4121 if (!N_BRAC_indexes.empty()) { 4122 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4123 symbol_ptr->SetByteSize(sym_idx + 1); 4124 symbol_ptr->SetSizeIsSibling(true); 4125 N_BRAC_indexes.pop_back(); 4126 } 4127 type = eSymbolTypeScopeEnd; 4128 break; 4129 4130 case N_EXCL: 4131 // deleted include file: name,,NO_SECT,0,sum 4132 type = eSymbolTypeHeaderFile; 4133 break; 4134 4135 //---------------------------------------------------------------------- 4136 // COMM scopes 4137 //---------------------------------------------------------------------- 4138 case N_BCOMM: 4139 // begin common: name,,NO_SECT,0,0 4140 // We use the current number of symbols in the symbol table in lieu 4141 // of using nlist_idx in case we ever start trimming entries out 4142 type = eSymbolTypeScopeBegin; 4143 N_COMM_indexes.push_back(sym_idx); 4144 break; 4145 4146 case N_ECOML: 4147 // end common (local name): 0,,n_sect,0,address 4148 symbol_section = 4149 section_info.GetSection(nlist.n_sect, nlist.n_value); 4150 LLVM_FALLTHROUGH; 4151 4152 case N_ECOMM: 4153 // end common: name,,n_sect,0,0 4154 // Set the size of the N_BCOMM to the terminating index of this 4155 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4156 // we need to navigate more quickly at the source level when 4157 // parsing STABS 4158 if (!N_COMM_indexes.empty()) { 4159 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4160 symbol_ptr->SetByteSize(sym_idx + 1); 4161 symbol_ptr->SetSizeIsSibling(true); 4162 N_COMM_indexes.pop_back(); 4163 } 4164 type = eSymbolTypeScopeEnd; 4165 break; 4166 4167 case N_LENG: 4168 // second stab entry with length information 4169 type = eSymbolTypeAdditional; 4170 break; 4171 4172 default: 4173 break; 4174 } 4175 } else { 4176 // uint8_t n_pext = N_PEXT & nlist.n_type; 4177 uint8_t n_type = N_TYPE & nlist.n_type; 4178 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4179 4180 switch (n_type) { 4181 case N_INDR: { 4182 const char *reexport_name_cstr = 4183 strtab_data.PeekCStr(nlist.n_value); 4184 if (reexport_name_cstr && reexport_name_cstr[0]) { 4185 type = eSymbolTypeReExported; 4186 ConstString reexport_name( 4187 reexport_name_cstr + 4188 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4189 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4190 set_value = false; 4191 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4192 indirect_symbol_names.insert( 4193 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4194 } else 4195 type = eSymbolTypeUndefined; 4196 } break; 4197 4198 case N_UNDF: 4199 if (symbol_name && symbol_name[0]) { 4200 ConstString undefined_name(symbol_name + 4201 ((symbol_name[0] == '_') ? 1 : 0)); 4202 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4203 } 4204 LLVM_FALLTHROUGH; 4205 4206 case N_PBUD: 4207 type = eSymbolTypeUndefined; 4208 break; 4209 4210 case N_ABS: 4211 type = eSymbolTypeAbsolute; 4212 break; 4213 4214 case N_SECT: { 4215 symbol_section = 4216 section_info.GetSection(nlist.n_sect, nlist.n_value); 4217 4218 if (!symbol_section) { 4219 // TODO: warn about this? 4220 add_nlist = false; 4221 break; 4222 } 4223 4224 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4225 type = eSymbolTypeException; 4226 } else { 4227 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4228 4229 switch (section_type) { 4230 case S_CSTRING_LITERALS: 4231 type = eSymbolTypeData; 4232 break; // section with only literal C strings 4233 case S_4BYTE_LITERALS: 4234 type = eSymbolTypeData; 4235 break; // section with only 4 byte literals 4236 case S_8BYTE_LITERALS: 4237 type = eSymbolTypeData; 4238 break; // section with only 8 byte literals 4239 case S_LITERAL_POINTERS: 4240 type = eSymbolTypeTrampoline; 4241 break; // section with only pointers to literals 4242 case S_NON_LAZY_SYMBOL_POINTERS: 4243 type = eSymbolTypeTrampoline; 4244 break; // section with only non-lazy symbol pointers 4245 case S_LAZY_SYMBOL_POINTERS: 4246 type = eSymbolTypeTrampoline; 4247 break; // section with only lazy symbol pointers 4248 case S_SYMBOL_STUBS: 4249 type = eSymbolTypeTrampoline; 4250 break; // section with only symbol stubs, byte size of stub in 4251 // the reserved2 field 4252 case S_MOD_INIT_FUNC_POINTERS: 4253 type = eSymbolTypeCode; 4254 break; // section with only function pointers for initialization 4255 case S_MOD_TERM_FUNC_POINTERS: 4256 type = eSymbolTypeCode; 4257 break; // section with only function pointers for termination 4258 case S_INTERPOSING: 4259 type = eSymbolTypeTrampoline; 4260 break; // section with only pairs of function pointers for 4261 // interposing 4262 case S_16BYTE_LITERALS: 4263 type = eSymbolTypeData; 4264 break; // section with only 16 byte literals 4265 case S_DTRACE_DOF: 4266 type = eSymbolTypeInstrumentation; 4267 break; 4268 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4269 type = eSymbolTypeTrampoline; 4270 break; 4271 default: 4272 switch (symbol_section->GetType()) { 4273 case lldb::eSectionTypeCode: 4274 type = eSymbolTypeCode; 4275 break; 4276 case eSectionTypeData: 4277 case eSectionTypeDataCString: // Inlined C string data 4278 case eSectionTypeDataCStringPointers: // Pointers to C string 4279 // data 4280 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4281 // the symbol table 4282 case eSectionTypeData4: 4283 case eSectionTypeData8: 4284 case eSectionTypeData16: 4285 type = eSymbolTypeData; 4286 break; 4287 default: 4288 break; 4289 } 4290 break; 4291 } 4292 4293 if (type == eSymbolTypeInvalid) { 4294 const char *symbol_sect_name = 4295 symbol_section->GetName().AsCString(); 4296 if (symbol_section->IsDescendant(text_section_sp.get())) { 4297 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4298 S_ATTR_SELF_MODIFYING_CODE | 4299 S_ATTR_SOME_INSTRUCTIONS)) 4300 type = eSymbolTypeData; 4301 else 4302 type = eSymbolTypeCode; 4303 } else if (symbol_section->IsDescendant( 4304 data_section_sp.get()) || 4305 symbol_section->IsDescendant( 4306 data_dirty_section_sp.get()) || 4307 symbol_section->IsDescendant( 4308 data_const_section_sp.get())) { 4309 if (symbol_sect_name && 4310 ::strstr(symbol_sect_name, "__objc") == 4311 symbol_sect_name) { 4312 type = eSymbolTypeRuntime; 4313 4314 if (symbol_name) { 4315 llvm::StringRef symbol_name_ref(symbol_name); 4316 if (symbol_name_ref.startswith("_OBJC_")) { 4317 static const llvm::StringRef g_objc_v2_prefix_class( 4318 "_OBJC_CLASS_$_"); 4319 static const llvm::StringRef g_objc_v2_prefix_metaclass( 4320 "_OBJC_METACLASS_$_"); 4321 static const llvm::StringRef g_objc_v2_prefix_ivar( 4322 "_OBJC_IVAR_$_"); 4323 if (symbol_name_ref.startswith( 4324 g_objc_v2_prefix_class)) { 4325 symbol_name_non_abi_mangled = symbol_name + 1; 4326 symbol_name = 4327 symbol_name + g_objc_v2_prefix_class.size(); 4328 type = eSymbolTypeObjCClass; 4329 demangled_is_synthesized = true; 4330 } else if (symbol_name_ref.startswith( 4331 g_objc_v2_prefix_metaclass)) { 4332 symbol_name_non_abi_mangled = symbol_name + 1; 4333 symbol_name = 4334 symbol_name + g_objc_v2_prefix_metaclass.size(); 4335 type = eSymbolTypeObjCMetaClass; 4336 demangled_is_synthesized = true; 4337 } else if (symbol_name_ref.startswith( 4338 g_objc_v2_prefix_ivar)) { 4339 symbol_name_non_abi_mangled = symbol_name + 1; 4340 symbol_name = 4341 symbol_name + g_objc_v2_prefix_ivar.size(); 4342 type = eSymbolTypeObjCIVar; 4343 demangled_is_synthesized = true; 4344 } 4345 } 4346 } 4347 } else if (symbol_sect_name && 4348 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4349 symbol_sect_name) { 4350 type = eSymbolTypeException; 4351 } else { 4352 type = eSymbolTypeData; 4353 } 4354 } else if (symbol_sect_name && 4355 ::strstr(symbol_sect_name, "__IMPORT") == 4356 symbol_sect_name) { 4357 type = eSymbolTypeTrampoline; 4358 } else if (symbol_section->IsDescendant( 4359 objc_section_sp.get())) { 4360 type = eSymbolTypeRuntime; 4361 if (symbol_name && symbol_name[0] == '.') { 4362 llvm::StringRef symbol_name_ref(symbol_name); 4363 static const llvm::StringRef g_objc_v1_prefix_class( 4364 ".objc_class_name_"); 4365 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4366 symbol_name_non_abi_mangled = symbol_name; 4367 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4368 type = eSymbolTypeObjCClass; 4369 demangled_is_synthesized = true; 4370 } 4371 } 4372 } 4373 } 4374 } 4375 } break; 4376 } 4377 } 4378 4379 if (add_nlist) { 4380 uint64_t symbol_value = nlist.n_value; 4381 4382 if (symbol_name_non_abi_mangled) { 4383 sym[sym_idx].GetMangled().SetMangledName( 4384 ConstString(symbol_name_non_abi_mangled)); 4385 sym[sym_idx].GetMangled().SetDemangledName( 4386 ConstString(symbol_name)); 4387 } else { 4388 bool symbol_name_is_mangled = false; 4389 4390 if (symbol_name && symbol_name[0] == '_') { 4391 symbol_name_is_mangled = symbol_name[1] == '_'; 4392 symbol_name++; // Skip the leading underscore 4393 } 4394 4395 if (symbol_name) { 4396 ConstString const_symbol_name(symbol_name); 4397 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4398 symbol_name_is_mangled); 4399 } 4400 } 4401 4402 if (is_gsym) { 4403 const char *gsym_name = sym[sym_idx] 4404 .GetMangled() 4405 .GetName(lldb::eLanguageTypeUnknown, 4406 Mangled::ePreferMangled) 4407 .GetCString(); 4408 if (gsym_name) 4409 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4410 } 4411 4412 if (symbol_section) { 4413 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4414 if (symbol_byte_size == 0 && function_starts_count > 0) { 4415 addr_t symbol_lookup_file_addr = nlist.n_value; 4416 // Do an exact address match for non-ARM addresses, else get the 4417 // closest since the symbol might be a thumb symbol which has an 4418 // address with bit zero set 4419 FunctionStarts::Entry *func_start_entry = 4420 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4421 if (is_arm && func_start_entry) { 4422 // Verify that the function start address is the symbol address 4423 // (ARM) or the symbol address + 1 (thumb) 4424 if (func_start_entry->addr != symbol_lookup_file_addr && 4425 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4426 // Not the right entry, NULL it out... 4427 func_start_entry = NULL; 4428 } 4429 } 4430 if (func_start_entry) { 4431 func_start_entry->data = true; 4432 4433 addr_t symbol_file_addr = func_start_entry->addr; 4434 if (is_arm) 4435 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4436 4437 const FunctionStarts::Entry *next_func_start_entry = 4438 function_starts.FindNextEntry(func_start_entry); 4439 const addr_t section_end_file_addr = 4440 section_file_addr + symbol_section->GetByteSize(); 4441 if (next_func_start_entry) { 4442 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4443 // Be sure the clear the Thumb address bit when we calculate 4444 // the size from the current and next address 4445 if (is_arm) 4446 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4447 symbol_byte_size = std::min<lldb::addr_t>( 4448 next_symbol_file_addr - symbol_file_addr, 4449 section_end_file_addr - symbol_file_addr); 4450 } else { 4451 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4452 } 4453 } 4454 } 4455 symbol_value -= section_file_addr; 4456 } 4457 4458 if (is_debug == false) { 4459 if (type == eSymbolTypeCode) { 4460 // See if we can find a N_FUN entry for any code symbols. If we 4461 // do find a match, and the name matches, then we can merge the 4462 // two into just the function symbol to avoid duplicate entries 4463 // in the symbol table 4464 std::pair<ValueToSymbolIndexMap::const_iterator, 4465 ValueToSymbolIndexMap::const_iterator> 4466 range; 4467 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4468 if (range.first != range.second) { 4469 bool found_it = false; 4470 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4471 pos != range.second; ++pos) { 4472 if (sym[sym_idx].GetMangled().GetName( 4473 lldb::eLanguageTypeUnknown, 4474 Mangled::ePreferMangled) == 4475 sym[pos->second].GetMangled().GetName( 4476 lldb::eLanguageTypeUnknown, 4477 Mangled::ePreferMangled)) { 4478 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4479 // We just need the flags from the linker symbol, so put 4480 // these flags into the N_FUN flags to avoid duplicate 4481 // symbols in the symbol table 4482 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4483 sym[pos->second].SetFlags(nlist.n_type << 16 | 4484 nlist.n_desc); 4485 if (resolver_addresses.find(nlist.n_value) != 4486 resolver_addresses.end()) 4487 sym[pos->second].SetType(eSymbolTypeResolver); 4488 sym[sym_idx].Clear(); 4489 found_it = true; 4490 break; 4491 } 4492 } 4493 if (found_it) 4494 continue; 4495 } else { 4496 if (resolver_addresses.find(nlist.n_value) != 4497 resolver_addresses.end()) 4498 type = eSymbolTypeResolver; 4499 } 4500 } else if (type == eSymbolTypeData || 4501 type == eSymbolTypeObjCClass || 4502 type == eSymbolTypeObjCMetaClass || 4503 type == eSymbolTypeObjCIVar) { 4504 // See if we can find a N_STSYM entry for any data symbols. If we 4505 // do find a match, and the name matches, then we can merge the 4506 // two into just the Static symbol to avoid duplicate entries in 4507 // the symbol table 4508 std::pair<ValueToSymbolIndexMap::const_iterator, 4509 ValueToSymbolIndexMap::const_iterator> 4510 range; 4511 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4512 if (range.first != range.second) { 4513 bool found_it = false; 4514 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4515 pos != range.second; ++pos) { 4516 if (sym[sym_idx].GetMangled().GetName( 4517 lldb::eLanguageTypeUnknown, 4518 Mangled::ePreferMangled) == 4519 sym[pos->second].GetMangled().GetName( 4520 lldb::eLanguageTypeUnknown, 4521 Mangled::ePreferMangled)) { 4522 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4523 // We just need the flags from the linker symbol, so put 4524 // these flags into the N_STSYM flags to avoid duplicate 4525 // symbols in the symbol table 4526 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4527 sym[pos->second].SetFlags(nlist.n_type << 16 | 4528 nlist.n_desc); 4529 sym[sym_idx].Clear(); 4530 found_it = true; 4531 break; 4532 } 4533 } 4534 if (found_it) 4535 continue; 4536 } else { 4537 // Combine N_GSYM stab entries with the non stab symbol 4538 const char *gsym_name = sym[sym_idx] 4539 .GetMangled() 4540 .GetName(lldb::eLanguageTypeUnknown, 4541 Mangled::ePreferMangled) 4542 .GetCString(); 4543 if (gsym_name) { 4544 ConstNameToSymbolIndexMap::const_iterator pos = 4545 N_GSYM_name_to_sym_idx.find(gsym_name); 4546 if (pos != N_GSYM_name_to_sym_idx.end()) { 4547 const uint32_t GSYM_sym_idx = pos->second; 4548 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4549 // Copy the address, because often the N_GSYM address has 4550 // an invalid address of zero when the global is a common 4551 // symbol 4552 sym[GSYM_sym_idx].GetAddressRef().SetSection( 4553 symbol_section); 4554 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4555 // We just need the flags from the linker symbol, so put 4556 // these flags into the N_GSYM flags to avoid duplicate 4557 // symbols in the symbol table 4558 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 4559 nlist.n_desc); 4560 sym[sym_idx].Clear(); 4561 continue; 4562 } 4563 } 4564 } 4565 } 4566 } 4567 4568 sym[sym_idx].SetID(nlist_idx); 4569 sym[sym_idx].SetType(type); 4570 if (set_value) { 4571 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4572 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4573 } 4574 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4575 4576 if (symbol_byte_size > 0) 4577 sym[sym_idx].SetByteSize(symbol_byte_size); 4578 4579 if (demangled_is_synthesized) 4580 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4581 4582 ++sym_idx; 4583 } else { 4584 sym[sym_idx].Clear(); 4585 } 4586 } 4587 4588 for (const auto &pos : reexport_shlib_needs_fixup) { 4589 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4590 if (undef_pos != undefined_name_to_desc.end()) { 4591 const uint8_t dylib_ordinal = 4592 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4593 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4594 sym[pos.first].SetReExportedSymbolSharedLibrary( 4595 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4596 } 4597 } 4598 } 4599 4600 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4601 4602 if (function_starts_count > 0) { 4603 uint32_t num_synthetic_function_symbols = 0; 4604 for (i = 0; i < function_starts_count; ++i) { 4605 if (function_starts.GetEntryRef(i).data == false) 4606 ++num_synthetic_function_symbols; 4607 } 4608 4609 if (num_synthetic_function_symbols > 0) { 4610 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4611 num_syms = sym_idx + num_synthetic_function_symbols; 4612 sym = symtab->Resize(num_syms); 4613 } 4614 for (i = 0; i < function_starts_count; ++i) { 4615 const FunctionStarts::Entry *func_start_entry = 4616 function_starts.GetEntryAtIndex(i); 4617 if (func_start_entry->data == false) { 4618 addr_t symbol_file_addr = func_start_entry->addr; 4619 uint32_t symbol_flags = 0; 4620 if (is_arm) { 4621 if (symbol_file_addr & 1) 4622 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4623 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4624 } 4625 Address symbol_addr; 4626 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4627 SectionSP symbol_section(symbol_addr.GetSection()); 4628 uint32_t symbol_byte_size = 0; 4629 if (symbol_section) { 4630 const addr_t section_file_addr = 4631 symbol_section->GetFileAddress(); 4632 const FunctionStarts::Entry *next_func_start_entry = 4633 function_starts.FindNextEntry(func_start_entry); 4634 const addr_t section_end_file_addr = 4635 section_file_addr + symbol_section->GetByteSize(); 4636 if (next_func_start_entry) { 4637 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4638 if (is_arm) 4639 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4640 symbol_byte_size = std::min<lldb::addr_t>( 4641 next_symbol_file_addr - symbol_file_addr, 4642 section_end_file_addr - symbol_file_addr); 4643 } else { 4644 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4645 } 4646 sym[sym_idx].SetID(synthetic_sym_id++); 4647 sym[sym_idx].GetMangled().SetDemangledName( 4648 GetNextSyntheticSymbolName()); 4649 sym[sym_idx].SetType(eSymbolTypeCode); 4650 sym[sym_idx].SetIsSynthetic(true); 4651 sym[sym_idx].GetAddressRef() = symbol_addr; 4652 if (symbol_flags) 4653 sym[sym_idx].SetFlags(symbol_flags); 4654 if (symbol_byte_size) 4655 sym[sym_idx].SetByteSize(symbol_byte_size); 4656 ++sym_idx; 4657 } 4658 } 4659 } 4660 } 4661 } 4662 } 4663 4664 // Trim our symbols down to just what we ended up with after removing any 4665 // symbols. 4666 if (sym_idx < num_syms) { 4667 num_syms = sym_idx; 4668 sym = symtab->Resize(num_syms); 4669 } 4670 4671 // Now synthesize indirect symbols 4672 if (m_dysymtab.nindirectsyms != 0) { 4673 if (indirect_symbol_index_data.GetByteSize()) { 4674 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4675 m_nlist_idx_to_sym_idx.end(); 4676 4677 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4678 ++sect_idx) { 4679 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4680 S_SYMBOL_STUBS) { 4681 uint32_t symbol_stub_byte_size = 4682 m_mach_sections[sect_idx].reserved2; 4683 if (symbol_stub_byte_size == 0) 4684 continue; 4685 4686 const uint32_t num_symbol_stubs = 4687 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4688 4689 if (num_symbol_stubs == 0) 4690 continue; 4691 4692 const uint32_t symbol_stub_index_offset = 4693 m_mach_sections[sect_idx].reserved1; 4694 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; 4695 ++stub_idx) { 4696 const uint32_t symbol_stub_index = 4697 symbol_stub_index_offset + stub_idx; 4698 const lldb::addr_t symbol_stub_addr = 4699 m_mach_sections[sect_idx].addr + 4700 (stub_idx * symbol_stub_byte_size); 4701 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4702 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4703 symbol_stub_offset, 4)) { 4704 const uint32_t stub_sym_id = 4705 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4706 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4707 continue; 4708 4709 NListIndexToSymbolIndexMap::const_iterator index_pos = 4710 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4711 Symbol *stub_symbol = NULL; 4712 if (index_pos != end_index_pos) { 4713 // We have a remapping from the original nlist index to a 4714 // current symbol index, so just look this up by index 4715 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4716 } else { 4717 // We need to lookup a symbol using the original nlist symbol 4718 // index since this index is coming from the S_SYMBOL_STUBS 4719 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4720 } 4721 4722 if (stub_symbol) { 4723 Address so_addr(symbol_stub_addr, section_list); 4724 4725 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4726 // Change the external symbol into a trampoline that makes 4727 // sense These symbols were N_UNDF N_EXT, and are useless 4728 // to us, so we can re-use them so we don't have to make up 4729 // a synthetic symbol for no good reason. 4730 if (resolver_addresses.find(symbol_stub_addr) == 4731 resolver_addresses.end()) 4732 stub_symbol->SetType(eSymbolTypeTrampoline); 4733 else 4734 stub_symbol->SetType(eSymbolTypeResolver); 4735 stub_symbol->SetExternal(false); 4736 stub_symbol->GetAddressRef() = so_addr; 4737 stub_symbol->SetByteSize(symbol_stub_byte_size); 4738 } else { 4739 // Make a synthetic symbol to describe the trampoline stub 4740 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4741 if (sym_idx >= num_syms) { 4742 sym = symtab->Resize(++num_syms); 4743 stub_symbol = NULL; // this pointer no longer valid 4744 } 4745 sym[sym_idx].SetID(synthetic_sym_id++); 4746 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4747 if (resolver_addresses.find(symbol_stub_addr) == 4748 resolver_addresses.end()) 4749 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4750 else 4751 sym[sym_idx].SetType(eSymbolTypeResolver); 4752 sym[sym_idx].SetIsSynthetic(true); 4753 sym[sym_idx].GetAddressRef() = so_addr; 4754 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4755 ++sym_idx; 4756 } 4757 } else { 4758 if (log) 4759 log->Warning("symbol stub referencing symbol table symbol " 4760 "%u that isn't in our minimal symbol table, " 4761 "fix this!!!", 4762 stub_sym_id); 4763 } 4764 } 4765 } 4766 } 4767 } 4768 } 4769 } 4770 4771 if (!trie_entries.empty()) { 4772 for (const auto &e : trie_entries) { 4773 if (e.entry.import_name) { 4774 // Only add indirect symbols from the Trie entries if we didn't have 4775 // a N_INDR nlist entry for this already 4776 if (indirect_symbol_names.find(e.entry.name) == 4777 indirect_symbol_names.end()) { 4778 // Make a synthetic symbol to describe re-exported symbol. 4779 if (sym_idx >= num_syms) 4780 sym = symtab->Resize(++num_syms); 4781 sym[sym_idx].SetID(synthetic_sym_id++); 4782 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4783 sym[sym_idx].SetType(eSymbolTypeReExported); 4784 sym[sym_idx].SetIsSynthetic(true); 4785 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4786 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4787 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4788 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4789 } 4790 ++sym_idx; 4791 } 4792 } 4793 } 4794 } 4795 4796 // StreamFile s(stdout, false); 4797 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4798 // symtab->Dump(&s, NULL, eSortOrderNone); 4799 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4800 // sizes 4801 symtab->CalculateSymbolSizes(); 4802 4803 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4804 // symtab->Dump(&s, NULL, eSortOrderNone); 4805 4806 return symtab->GetNumSymbols(); 4807 } 4808 return 0; 4809 } 4810 4811 void ObjectFileMachO::Dump(Stream *s) { 4812 ModuleSP module_sp(GetModule()); 4813 if (module_sp) { 4814 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4815 s->Printf("%p: ", static_cast<void *>(this)); 4816 s->Indent(); 4817 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4818 s->PutCString("ObjectFileMachO64"); 4819 else 4820 s->PutCString("ObjectFileMachO32"); 4821 4822 ArchSpec header_arch; 4823 GetArchitecture(header_arch); 4824 4825 *s << ", file = '" << m_file 4826 << "', triple = " << header_arch.GetTriple().getTriple() << "\n"; 4827 4828 SectionList *sections = GetSectionList(); 4829 if (sections) 4830 sections->Dump(s, NULL, true, UINT32_MAX); 4831 4832 if (m_symtab_ap.get()) 4833 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4834 } 4835 } 4836 4837 bool ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4838 const lldb_private::DataExtractor &data, 4839 lldb::offset_t lc_offset, 4840 lldb_private::UUID &uuid) { 4841 uint32_t i; 4842 struct uuid_command load_cmd; 4843 4844 lldb::offset_t offset = lc_offset; 4845 for (i = 0; i < header.ncmds; ++i) { 4846 const lldb::offset_t cmd_offset = offset; 4847 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4848 break; 4849 4850 if (load_cmd.cmd == LC_UUID) { 4851 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4852 4853 if (uuid_bytes) { 4854 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4855 // We pretend these object files have no UUID to prevent crashing. 4856 4857 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4858 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4859 0xbb, 0x14, 0xf0, 0x0d}; 4860 4861 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4862 return false; 4863 4864 uuid.SetBytes(uuid_bytes); 4865 return true; 4866 } 4867 return false; 4868 } 4869 offset = cmd_offset + load_cmd.cmdsize; 4870 } 4871 return false; 4872 } 4873 4874 static const char *GetOSName(uint32_t cmd) { 4875 switch (cmd) { 4876 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4877 return "ios"; 4878 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4879 return "macosx"; 4880 case llvm::MachO::LC_VERSION_MIN_TVOS: 4881 return "tvos"; 4882 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4883 return "watchos"; 4884 default: 4885 llvm_unreachable("unexpected LC_VERSION load command"); 4886 } 4887 } 4888 4889 bool ObjectFileMachO::GetArchitecture(const llvm::MachO::mach_header &header, 4890 const lldb_private::DataExtractor &data, 4891 lldb::offset_t lc_offset, 4892 ArchSpec &arch) { 4893 arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4894 4895 if (arch.IsValid()) { 4896 llvm::Triple &triple = arch.GetTriple(); 4897 4898 // Set OS to an unspecified unknown or a "*" so it can match any OS 4899 triple.setOS(llvm::Triple::UnknownOS); 4900 triple.setOSName(llvm::StringRef()); 4901 4902 if (header.filetype == MH_PRELOAD) { 4903 if (header.cputype == CPU_TYPE_ARM) { 4904 // If this is a 32-bit arm binary, and it's a standalone binary, force 4905 // the Vendor to Apple so we don't accidentally pick up the generic 4906 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4907 // frame pointer register; most other armv7 ABIs use a combination of 4908 // r7 and r11. 4909 triple.setVendor(llvm::Triple::Apple); 4910 } else { 4911 // Set vendor to an unspecified unknown or a "*" so it can match any 4912 // vendor This is required for correct behavior of EFI debugging on 4913 // x86_64 4914 triple.setVendor(llvm::Triple::UnknownVendor); 4915 triple.setVendorName(llvm::StringRef()); 4916 } 4917 return true; 4918 } else { 4919 struct load_command load_cmd; 4920 4921 lldb::offset_t offset = lc_offset; 4922 for (uint32_t i = 0; i < header.ncmds; ++i) { 4923 const lldb::offset_t cmd_offset = offset; 4924 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4925 break; 4926 4927 uint32_t major, minor, patch; 4928 struct version_min_command version_min; 4929 4930 llvm::SmallString<16> os_name; 4931 llvm::raw_svector_ostream os(os_name); 4932 4933 switch (load_cmd.cmd) { 4934 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4935 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4936 case llvm::MachO::LC_VERSION_MIN_TVOS: 4937 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4938 if (load_cmd.cmdsize != sizeof(version_min)) 4939 break; 4940 data.ExtractBytes(cmd_offset, 4941 sizeof(version_min), data.GetByteOrder(), 4942 &version_min); 4943 major = version_min.version >> 16; 4944 minor = (version_min.version >> 8) & 0xffu; 4945 patch = version_min.version & 0xffu; 4946 os << GetOSName(load_cmd.cmd) << major << '.' << minor << '.' 4947 << patch; 4948 triple.setOSName(os.str()); 4949 return true; 4950 default: 4951 break; 4952 } 4953 4954 offset = cmd_offset + load_cmd.cmdsize; 4955 } 4956 4957 if (header.filetype != MH_KEXT_BUNDLE) { 4958 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 4959 // so lets not say our Vendor is Apple, leave it as an unspecified 4960 // unknown 4961 triple.setVendor(llvm::Triple::UnknownVendor); 4962 triple.setVendorName(llvm::StringRef()); 4963 } 4964 } 4965 } 4966 return arch.IsValid(); 4967 } 4968 4969 bool ObjectFileMachO::GetUUID(lldb_private::UUID *uuid) { 4970 ModuleSP module_sp(GetModule()); 4971 if (module_sp) { 4972 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4973 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4974 return GetUUID(m_header, m_data, offset, *uuid); 4975 } 4976 return false; 4977 } 4978 4979 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 4980 uint32_t count = 0; 4981 ModuleSP module_sp(GetModule()); 4982 if (module_sp) { 4983 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4984 struct load_command load_cmd; 4985 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4986 std::vector<std::string> rpath_paths; 4987 std::vector<std::string> rpath_relative_paths; 4988 std::vector<std::string> at_exec_relative_paths; 4989 const bool resolve_path = false; // Don't resolve the dependent file paths 4990 // since they may not reside on this 4991 // system 4992 uint32_t i; 4993 for (i = 0; i < m_header.ncmds; ++i) { 4994 const uint32_t cmd_offset = offset; 4995 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4996 break; 4997 4998 switch (load_cmd.cmd) { 4999 case LC_RPATH: 5000 case LC_LOAD_DYLIB: 5001 case LC_LOAD_WEAK_DYLIB: 5002 case LC_REEXPORT_DYLIB: 5003 case LC_LOAD_DYLINKER: 5004 case LC_LOADFVMLIB: 5005 case LC_LOAD_UPWARD_DYLIB: { 5006 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5007 const char *path = m_data.PeekCStr(name_offset); 5008 if (path) { 5009 if (load_cmd.cmd == LC_RPATH) 5010 rpath_paths.push_back(path); 5011 else { 5012 if (path[0] == '@') { 5013 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5014 rpath_relative_paths.push_back(path + strlen("@rpath")); 5015 else if (strncmp(path, "@executable_path", 5016 strlen("@executable_path")) == 0) 5017 at_exec_relative_paths.push_back(path 5018 + strlen("@executable_path")); 5019 } else { 5020 FileSpec file_spec(path, resolve_path); 5021 if (files.AppendIfUnique(file_spec)) 5022 count++; 5023 } 5024 } 5025 } 5026 } break; 5027 5028 default: 5029 break; 5030 } 5031 offset = cmd_offset + load_cmd.cmdsize; 5032 } 5033 5034 FileSpec this_file_spec(m_file); 5035 this_file_spec.ResolvePath(); 5036 5037 if (!rpath_paths.empty()) { 5038 // Fixup all LC_RPATH values to be absolute paths 5039 std::string loader_path("@loader_path"); 5040 std::string executable_path("@executable_path"); 5041 for (auto &rpath : rpath_paths) { 5042 if (rpath.find(loader_path) == 0) { 5043 rpath.erase(0, loader_path.size()); 5044 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5045 } else if (rpath.find(executable_path) == 0) { 5046 rpath.erase(0, executable_path.size()); 5047 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5048 } 5049 } 5050 5051 for (const auto &rpath_relative_path : rpath_relative_paths) { 5052 for (const auto &rpath : rpath_paths) { 5053 std::string path = rpath; 5054 path += rpath_relative_path; 5055 // It is OK to resolve this path because we must find a file on disk 5056 // for us to accept it anyway if it is rpath relative. 5057 FileSpec file_spec(path, true); 5058 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) { 5059 count++; 5060 break; 5061 } 5062 } 5063 } 5064 } 5065 5066 // We may have @executable_paths but no RPATHS. Figure those out here. 5067 // Only do this if this object file is the executable. We have no way to 5068 // get back to the actual executable otherwise, so we won't get the right 5069 // path. 5070 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5071 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5072 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5073 FileSpec file_spec = 5074 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5075 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) 5076 count++; 5077 } 5078 } 5079 } 5080 return count; 5081 } 5082 5083 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5084 // If the object file is not an executable it can't hold the entry point. 5085 // m_entry_point_address is initialized to an invalid address, so we can just 5086 // return that. If m_entry_point_address is valid it means we've found it 5087 // already, so return the cached value. 5088 5089 if (!IsExecutable() || m_entry_point_address.IsValid()) 5090 return m_entry_point_address; 5091 5092 // Otherwise, look for the UnixThread or Thread command. The data for the 5093 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5094 // 5095 // uint32_t flavor - this is the flavor argument you would pass to 5096 // thread_get_state 5097 // uint32_t count - this is the count of longs in the thread state data 5098 // struct XXX_thread_state state - this is the structure from 5099 // <machine/thread_status.h> corresponding to the flavor. 5100 // <repeat this trio> 5101 // 5102 // So we just keep reading the various register flavors till we find the GPR 5103 // one, then read the PC out of there. 5104 // FIXME: We will need to have a "RegisterContext data provider" class at some 5105 // point that can get all the registers 5106 // out of data in this form & attach them to a given thread. That should 5107 // underlie the MacOS X User process plugin, and we'll also need it for the 5108 // MacOS X Core File process plugin. When we have that we can also use it 5109 // here. 5110 // 5111 // For now we hard-code the offsets and flavors we need: 5112 // 5113 // 5114 5115 ModuleSP module_sp(GetModule()); 5116 if (module_sp) { 5117 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5118 struct load_command load_cmd; 5119 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5120 uint32_t i; 5121 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5122 bool done = false; 5123 5124 for (i = 0; i < m_header.ncmds; ++i) { 5125 const lldb::offset_t cmd_offset = offset; 5126 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5127 break; 5128 5129 switch (load_cmd.cmd) { 5130 case LC_UNIXTHREAD: 5131 case LC_THREAD: { 5132 while (offset < cmd_offset + load_cmd.cmdsize) { 5133 uint32_t flavor = m_data.GetU32(&offset); 5134 uint32_t count = m_data.GetU32(&offset); 5135 if (count == 0) { 5136 // We've gotten off somehow, log and exit; 5137 return m_entry_point_address; 5138 } 5139 5140 switch (m_header.cputype) { 5141 case llvm::MachO::CPU_TYPE_ARM: 5142 if (flavor == 1 || 5143 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 from 5144 // mach/arm/thread_status.h 5145 { 5146 offset += 60; // This is the offset of pc in the GPR thread state 5147 // data structure. 5148 start_address = m_data.GetU32(&offset); 5149 done = true; 5150 } 5151 break; 5152 case llvm::MachO::CPU_TYPE_ARM64: 5153 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5154 { 5155 offset += 256; // This is the offset of pc in the GPR thread state 5156 // data structure. 5157 start_address = m_data.GetU64(&offset); 5158 done = true; 5159 } 5160 break; 5161 case llvm::MachO::CPU_TYPE_I386: 5162 if (flavor == 5163 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5164 { 5165 offset += 40; // This is the offset of eip in the GPR thread state 5166 // data structure. 5167 start_address = m_data.GetU32(&offset); 5168 done = true; 5169 } 5170 break; 5171 case llvm::MachO::CPU_TYPE_X86_64: 5172 if (flavor == 5173 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5174 { 5175 offset += 16 * 8; // This is the offset of rip in the GPR thread 5176 // state data structure. 5177 start_address = m_data.GetU64(&offset); 5178 done = true; 5179 } 5180 break; 5181 default: 5182 return m_entry_point_address; 5183 } 5184 // Haven't found the GPR flavor yet, skip over the data for this 5185 // flavor: 5186 if (done) 5187 break; 5188 offset += count * 4; 5189 } 5190 } break; 5191 case LC_MAIN: { 5192 ConstString text_segment_name("__TEXT"); 5193 uint64_t entryoffset = m_data.GetU64(&offset); 5194 SectionSP text_segment_sp = 5195 GetSectionList()->FindSectionByName(text_segment_name); 5196 if (text_segment_sp) { 5197 done = true; 5198 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5199 } 5200 } break; 5201 5202 default: 5203 break; 5204 } 5205 if (done) 5206 break; 5207 5208 // Go to the next load command: 5209 offset = cmd_offset + load_cmd.cmdsize; 5210 } 5211 5212 if (start_address != LLDB_INVALID_ADDRESS) { 5213 // We got the start address from the load commands, so now resolve that 5214 // address in the sections of this ObjectFile: 5215 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5216 start_address, GetSectionList())) { 5217 m_entry_point_address.Clear(); 5218 } 5219 } else { 5220 // We couldn't read the UnixThread load command - maybe it wasn't there. 5221 // As a fallback look for the "start" symbol in the main executable. 5222 5223 ModuleSP module_sp(GetModule()); 5224 5225 if (module_sp) { 5226 SymbolContextList contexts; 5227 SymbolContext context; 5228 if (module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5229 eSymbolTypeCode, contexts)) { 5230 if (contexts.GetContextAtIndex(0, context)) 5231 m_entry_point_address = context.symbol->GetAddress(); 5232 } 5233 } 5234 } 5235 } 5236 5237 return m_entry_point_address; 5238 } 5239 5240 lldb_private::Address ObjectFileMachO::GetHeaderAddress() { 5241 lldb_private::Address header_addr; 5242 SectionList *section_list = GetSectionList(); 5243 if (section_list) { 5244 SectionSP text_segment_sp( 5245 section_list->FindSectionByName(GetSegmentNameTEXT())); 5246 if (text_segment_sp) { 5247 header_addr.SetSection(text_segment_sp); 5248 header_addr.SetOffset(0); 5249 } 5250 } 5251 return header_addr; 5252 } 5253 5254 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5255 ModuleSP module_sp(GetModule()); 5256 if (module_sp) { 5257 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5258 if (!m_thread_context_offsets_valid) { 5259 m_thread_context_offsets_valid = true; 5260 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5261 FileRangeArray::Entry file_range; 5262 thread_command thread_cmd; 5263 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5264 const uint32_t cmd_offset = offset; 5265 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 5266 break; 5267 5268 if (thread_cmd.cmd == LC_THREAD) { 5269 file_range.SetRangeBase(offset); 5270 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5271 m_thread_context_offsets.Append(file_range); 5272 } 5273 offset = cmd_offset + thread_cmd.cmdsize; 5274 } 5275 } 5276 } 5277 return m_thread_context_offsets.GetSize(); 5278 } 5279 5280 std::string ObjectFileMachO::GetIdentifierString() { 5281 std::string result; 5282 ModuleSP module_sp(GetModule()); 5283 if (module_sp) { 5284 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5285 5286 // First, look over the load commands for an LC_NOTE load command with 5287 // data_owner string "kern ver str" & use that if found. 5288 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5289 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5290 const uint32_t cmd_offset = offset; 5291 load_command lc; 5292 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5293 break; 5294 if (lc.cmd == LC_NOTE) 5295 { 5296 char data_owner[17]; 5297 m_data.CopyData (offset, 16, data_owner); 5298 data_owner[16] = '\0'; 5299 offset += 16; 5300 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5301 uint64_t size = m_data.GetU64_unchecked (&offset); 5302 5303 // "kern ver str" has a uint32_t version and then a nul terminated 5304 // c-string. 5305 if (strcmp ("kern ver str", data_owner) == 0) 5306 { 5307 offset = fileoff; 5308 uint32_t version; 5309 if (m_data.GetU32 (&offset, &version, 1) != nullptr) 5310 { 5311 if (version == 1) 5312 { 5313 uint32_t strsize = size - sizeof (uint32_t); 5314 char *buf = (char*) malloc (strsize); 5315 if (buf) 5316 { 5317 m_data.CopyData (offset, strsize, buf); 5318 buf[strsize - 1] = '\0'; 5319 result = buf; 5320 if (buf) 5321 free (buf); 5322 return result; 5323 } 5324 } 5325 } 5326 } 5327 } 5328 offset = cmd_offset + lc.cmdsize; 5329 } 5330 5331 // Second, make a pass over the load commands looking for an obsolete 5332 // LC_IDENT load command. 5333 offset = MachHeaderSizeFromMagic(m_header.magic); 5334 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5335 const uint32_t cmd_offset = offset; 5336 struct ident_command ident_command; 5337 if (m_data.GetU32(&offset, &ident_command, 2) == NULL) 5338 break; 5339 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5340 char *buf = (char *) malloc (ident_command.cmdsize); 5341 if (buf != nullptr 5342 && m_data.CopyData (offset, ident_command.cmdsize, buf) == ident_command.cmdsize) { 5343 buf[ident_command.cmdsize - 1] = '\0'; 5344 result = buf; 5345 } 5346 if (buf) 5347 free (buf); 5348 } 5349 offset = cmd_offset + ident_command.cmdsize; 5350 } 5351 5352 } 5353 return result; 5354 } 5355 5356 bool ObjectFileMachO::GetCorefileMainBinaryInfo (addr_t &address, UUID &uuid) { 5357 address = LLDB_INVALID_ADDRESS; 5358 uuid.Clear(); 5359 ModuleSP module_sp(GetModule()); 5360 if (module_sp) { 5361 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5362 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5363 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5364 const uint32_t cmd_offset = offset; 5365 load_command lc; 5366 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5367 break; 5368 if (lc.cmd == LC_NOTE) 5369 { 5370 char data_owner[17]; 5371 memset (data_owner, 0, sizeof (data_owner)); 5372 m_data.CopyData (offset, 16, data_owner); 5373 offset += 16; 5374 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5375 uint64_t size = m_data.GetU64_unchecked (&offset); 5376 5377 // "main bin spec" (main binary specification) data payload is 5378 // formatted: 5379 // uint32_t version [currently 1] 5380 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user process] 5381 // uint64_t address [ UINT64_MAX if address not specified ] 5382 // uuid_t uuid [ all zero's if uuid not specified ] 5383 // uint32_t log2_pagesize [ process page size in log base 2, e.g. 4k pages are 12. 0 for unspecified ] 5384 5385 if (strcmp ("main bin spec", data_owner) == 0 && size >= 32) 5386 { 5387 offset = fileoff; 5388 uint32_t version; 5389 if (m_data.GetU32 (&offset, &version, 1) != nullptr && version == 1) 5390 { 5391 uint32_t type = 0; 5392 uuid_t raw_uuid; 5393 memset (raw_uuid, 0, sizeof (uuid_t)); 5394 5395 if (m_data.GetU32 (&offset, &type, 1) 5396 && m_data.GetU64 (&offset, &address, 1) 5397 && m_data.CopyData (offset, sizeof (uuid_t), raw_uuid) != 0 5398 && uuid.SetBytes (raw_uuid, sizeof (uuid_t))) 5399 { 5400 return true; 5401 } 5402 } 5403 } 5404 } 5405 offset = cmd_offset + lc.cmdsize; 5406 } 5407 } 5408 return false; 5409 } 5410 5411 lldb::RegisterContextSP 5412 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5413 lldb_private::Thread &thread) { 5414 lldb::RegisterContextSP reg_ctx_sp; 5415 5416 ModuleSP module_sp(GetModule()); 5417 if (module_sp) { 5418 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5419 if (!m_thread_context_offsets_valid) 5420 GetNumThreadContexts(); 5421 5422 const FileRangeArray::Entry *thread_context_file_range = 5423 m_thread_context_offsets.GetEntryAtIndex(idx); 5424 if (thread_context_file_range) { 5425 5426 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5427 thread_context_file_range->GetByteSize()); 5428 5429 switch (m_header.cputype) { 5430 case llvm::MachO::CPU_TYPE_ARM64: 5431 reg_ctx_sp.reset(new RegisterContextDarwin_arm64_Mach(thread, data)); 5432 break; 5433 5434 case llvm::MachO::CPU_TYPE_ARM: 5435 reg_ctx_sp.reset(new RegisterContextDarwin_arm_Mach(thread, data)); 5436 break; 5437 5438 case llvm::MachO::CPU_TYPE_I386: 5439 reg_ctx_sp.reset(new RegisterContextDarwin_i386_Mach(thread, data)); 5440 break; 5441 5442 case llvm::MachO::CPU_TYPE_X86_64: 5443 reg_ctx_sp.reset(new RegisterContextDarwin_x86_64_Mach(thread, data)); 5444 break; 5445 } 5446 } 5447 } 5448 return reg_ctx_sp; 5449 } 5450 5451 ObjectFile::Type ObjectFileMachO::CalculateType() { 5452 switch (m_header.filetype) { 5453 case MH_OBJECT: // 0x1u 5454 if (GetAddressByteSize() == 4) { 5455 // 32 bit kexts are just object files, but they do have a valid 5456 // UUID load command. 5457 UUID uuid; 5458 if (GetUUID(&uuid)) { 5459 // this checking for the UUID load command is not enough we could 5460 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5461 // this is required of kexts 5462 if (m_strata == eStrataInvalid) 5463 m_strata = eStrataKernel; 5464 return eTypeSharedLibrary; 5465 } 5466 } 5467 return eTypeObjectFile; 5468 5469 case MH_EXECUTE: 5470 return eTypeExecutable; // 0x2u 5471 case MH_FVMLIB: 5472 return eTypeSharedLibrary; // 0x3u 5473 case MH_CORE: 5474 return eTypeCoreFile; // 0x4u 5475 case MH_PRELOAD: 5476 return eTypeSharedLibrary; // 0x5u 5477 case MH_DYLIB: 5478 return eTypeSharedLibrary; // 0x6u 5479 case MH_DYLINKER: 5480 return eTypeDynamicLinker; // 0x7u 5481 case MH_BUNDLE: 5482 return eTypeSharedLibrary; // 0x8u 5483 case MH_DYLIB_STUB: 5484 return eTypeStubLibrary; // 0x9u 5485 case MH_DSYM: 5486 return eTypeDebugInfo; // 0xAu 5487 case MH_KEXT_BUNDLE: 5488 return eTypeSharedLibrary; // 0xBu 5489 default: 5490 break; 5491 } 5492 return eTypeUnknown; 5493 } 5494 5495 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5496 switch (m_header.filetype) { 5497 case MH_OBJECT: // 0x1u 5498 { 5499 // 32 bit kexts are just object files, but they do have a valid 5500 // UUID load command. 5501 UUID uuid; 5502 if (GetUUID(&uuid)) { 5503 // this checking for the UUID load command is not enough we could 5504 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5505 // this is required of kexts 5506 if (m_type == eTypeInvalid) 5507 m_type = eTypeSharedLibrary; 5508 5509 return eStrataKernel; 5510 } 5511 } 5512 return eStrataUnknown; 5513 5514 case MH_EXECUTE: // 0x2u 5515 // Check for the MH_DYLDLINK bit in the flags 5516 if (m_header.flags & MH_DYLDLINK) { 5517 return eStrataUser; 5518 } else { 5519 SectionList *section_list = GetSectionList(); 5520 if (section_list) { 5521 static ConstString g_kld_section_name("__KLD"); 5522 if (section_list->FindSectionByName(g_kld_section_name)) 5523 return eStrataKernel; 5524 } 5525 } 5526 return eStrataRawImage; 5527 5528 case MH_FVMLIB: 5529 return eStrataUser; // 0x3u 5530 case MH_CORE: 5531 return eStrataUnknown; // 0x4u 5532 case MH_PRELOAD: 5533 return eStrataRawImage; // 0x5u 5534 case MH_DYLIB: 5535 return eStrataUser; // 0x6u 5536 case MH_DYLINKER: 5537 return eStrataUser; // 0x7u 5538 case MH_BUNDLE: 5539 return eStrataUser; // 0x8u 5540 case MH_DYLIB_STUB: 5541 return eStrataUser; // 0x9u 5542 case MH_DSYM: 5543 return eStrataUnknown; // 0xAu 5544 case MH_KEXT_BUNDLE: 5545 return eStrataKernel; // 0xBu 5546 default: 5547 break; 5548 } 5549 return eStrataUnknown; 5550 } 5551 5552 uint32_t ObjectFileMachO::GetVersion(uint32_t *versions, 5553 uint32_t num_versions) { 5554 ModuleSP module_sp(GetModule()); 5555 if (module_sp) { 5556 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5557 struct dylib_command load_cmd; 5558 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5559 uint32_t version_cmd = 0; 5560 uint64_t version = 0; 5561 uint32_t i; 5562 for (i = 0; i < m_header.ncmds; ++i) { 5563 const lldb::offset_t cmd_offset = offset; 5564 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5565 break; 5566 5567 if (load_cmd.cmd == LC_ID_DYLIB) { 5568 if (version_cmd == 0) { 5569 version_cmd = load_cmd.cmd; 5570 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5571 break; 5572 version = load_cmd.dylib.current_version; 5573 } 5574 break; // Break for now unless there is another more complete version 5575 // number load command in the future. 5576 } 5577 offset = cmd_offset + load_cmd.cmdsize; 5578 } 5579 5580 if (version_cmd == LC_ID_DYLIB) { 5581 if (versions != NULL && num_versions > 0) { 5582 if (num_versions > 0) 5583 versions[0] = (version & 0xFFFF0000ull) >> 16; 5584 if (num_versions > 1) 5585 versions[1] = (version & 0x0000FF00ull) >> 8; 5586 if (num_versions > 2) 5587 versions[2] = (version & 0x000000FFull); 5588 // Fill in an remaining version numbers with invalid values 5589 for (i = 3; i < num_versions; ++i) 5590 versions[i] = UINT32_MAX; 5591 } 5592 // The LC_ID_DYLIB load command has a version with 3 version numbers in 5593 // it, so always return 3 5594 return 3; 5595 } 5596 } 5597 return false; 5598 } 5599 5600 bool ObjectFileMachO::GetArchitecture(ArchSpec &arch) { 5601 ModuleSP module_sp(GetModule()); 5602 if (module_sp) { 5603 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5604 return GetArchitecture(m_header, m_data, 5605 MachHeaderSizeFromMagic(m_header.magic), arch); 5606 } 5607 return false; 5608 } 5609 5610 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, addr_t &base_addr, UUID &uuid) { 5611 uuid.Clear(); 5612 base_addr = LLDB_INVALID_ADDRESS; 5613 if (process && process->GetDynamicLoader()) { 5614 DynamicLoader *dl = process->GetDynamicLoader(); 5615 LazyBool using_shared_cache; 5616 LazyBool private_shared_cache; 5617 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5618 private_shared_cache); 5619 } 5620 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5621 if (log) 5622 log->Printf("inferior process shared cache has a UUID of %s, base address 0x%" PRIx64 , uuid.GetAsString().c_str(), base_addr); 5623 } 5624 5625 // From dyld SPI header dyld_process_info.h 5626 typedef void *dyld_process_info; 5627 struct lldb_copy__dyld_process_cache_info { 5628 uuid_t cacheUUID; // UUID of cache used by process 5629 uint64_t cacheBaseAddress; // load address of dyld shared cache 5630 bool noCache; // process is running without a dyld cache 5631 bool privateCache; // process is using a private copy of its dyld cache 5632 }; 5633 5634 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with llvm 5635 // enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile errors. 5636 // So we need to use the actual underlying types of task_t and kern_return_t 5637 // below. 5638 extern "C" unsigned int /*task_t*/ mach_task_self(); 5639 5640 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5641 uuid.Clear(); 5642 base_addr = LLDB_INVALID_ADDRESS; 5643 5644 #if defined(__APPLE__) && \ 5645 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5646 uint8_t *(*dyld_get_all_image_infos)(void); 5647 dyld_get_all_image_infos = 5648 (uint8_t * (*)())dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5649 if (dyld_get_all_image_infos) { 5650 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5651 if (dyld_all_image_infos_address) { 5652 uint32_t *version = (uint32_t *) 5653 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5654 if (*version >= 13) { 5655 uuid_t *sharedCacheUUID_address = 0; 5656 int wordsize = sizeof(uint8_t *); 5657 if (wordsize == 8) { 5658 sharedCacheUUID_address = 5659 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5660 160); // sharedCacheUUID <mach-o/dyld_images.h> 5661 if (*version >= 15) 5662 base_addr = *(uint64_t *) ((uint8_t *) dyld_all_image_infos_address 5663 + 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5664 } else { 5665 sharedCacheUUID_address = 5666 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5667 84); // sharedCacheUUID <mach-o/dyld_images.h> 5668 if (*version >= 15) { 5669 base_addr = 0; 5670 base_addr = *(uint32_t *) ((uint8_t *) dyld_all_image_infos_address 5671 + 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5672 } 5673 } 5674 uuid.SetBytes(sharedCacheUUID_address); 5675 } 5676 } 5677 } else { 5678 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5679 dyld_process_info (*dyld_process_info_create)(unsigned int /* task_t */ task, uint64_t timestamp, unsigned int /*kern_return_t*/ *kernelError); 5680 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5681 void (*dyld_process_info_release)(dyld_process_info info); 5682 5683 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, unsigned int /*kern_return_t*/ *)) 5684 dlsym (RTLD_DEFAULT, "_dyld_process_info_create"); 5685 dyld_process_info_get_cache = (void (*)(void *, void *)) 5686 dlsym (RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5687 dyld_process_info_release = (void (*)(void *)) 5688 dlsym (RTLD_DEFAULT, "_dyld_process_info_release"); 5689 5690 if (dyld_process_info_create && dyld_process_info_get_cache) { 5691 unsigned int /*kern_return_t */ kern_ret; 5692 dyld_process_info process_info = dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5693 if (process_info) { 5694 struct lldb_copy__dyld_process_cache_info sc_info; 5695 memset (&sc_info, 0, sizeof (struct lldb_copy__dyld_process_cache_info)); 5696 dyld_process_info_get_cache (process_info, &sc_info); 5697 if (sc_info.cacheBaseAddress != 0) { 5698 base_addr = sc_info.cacheBaseAddress; 5699 uuid.SetBytes (sc_info.cacheUUID); 5700 } 5701 dyld_process_info_release (process_info); 5702 } 5703 } 5704 } 5705 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5706 if (log && uuid.IsValid()) 5707 log->Printf("lldb's in-memory shared cache has a UUID of %s base address of 0x%" PRIx64, uuid.GetAsString().c_str(), base_addr); 5708 #endif 5709 } 5710 5711 uint32_t ObjectFileMachO::GetMinimumOSVersion(uint32_t *versions, 5712 uint32_t num_versions) { 5713 if (m_min_os_versions.empty()) { 5714 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5715 bool success = false; 5716 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5717 const lldb::offset_t load_cmd_offset = offset; 5718 5719 version_min_command lc; 5720 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5721 break; 5722 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5723 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5724 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5725 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5726 if (m_data.GetU32(&offset, &lc.version, 5727 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5728 const uint32_t xxxx = lc.version >> 16; 5729 const uint32_t yy = (lc.version >> 8) & 0xffu; 5730 const uint32_t zz = lc.version & 0xffu; 5731 if (xxxx) { 5732 m_min_os_versions.push_back(xxxx); 5733 m_min_os_versions.push_back(yy); 5734 m_min_os_versions.push_back(zz); 5735 } 5736 success = true; 5737 } 5738 } 5739 offset = load_cmd_offset + lc.cmdsize; 5740 } 5741 5742 if (success == false) { 5743 // Push an invalid value so we don't keep trying to 5744 m_min_os_versions.push_back(UINT32_MAX); 5745 } 5746 } 5747 5748 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) { 5749 if (versions != NULL && num_versions > 0) { 5750 for (size_t i = 0; i < num_versions; ++i) { 5751 if (i < m_min_os_versions.size()) 5752 versions[i] = m_min_os_versions[i]; 5753 else 5754 versions[i] = 0; 5755 } 5756 } 5757 return m_min_os_versions.size(); 5758 } 5759 // Call the superclasses version that will empty out the data 5760 return ObjectFile::GetMinimumOSVersion(versions, num_versions); 5761 } 5762 5763 uint32_t ObjectFileMachO::GetSDKVersion(uint32_t *versions, 5764 uint32_t num_versions) { 5765 if (m_sdk_versions.empty()) { 5766 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5767 bool success = false; 5768 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5769 const lldb::offset_t load_cmd_offset = offset; 5770 5771 version_min_command lc; 5772 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5773 break; 5774 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5775 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5776 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5777 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5778 if (m_data.GetU32(&offset, &lc.version, 5779 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5780 const uint32_t xxxx = lc.sdk >> 16; 5781 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5782 const uint32_t zz = lc.sdk & 0xffu; 5783 if (xxxx) { 5784 m_sdk_versions.push_back(xxxx); 5785 m_sdk_versions.push_back(yy); 5786 m_sdk_versions.push_back(zz); 5787 } 5788 success = true; 5789 } 5790 } 5791 offset = load_cmd_offset + lc.cmdsize; 5792 } 5793 5794 if (success == false) { 5795 // Push an invalid value so we don't keep trying to 5796 m_sdk_versions.push_back(UINT32_MAX); 5797 } 5798 } 5799 5800 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) { 5801 if (versions != NULL && num_versions > 0) { 5802 for (size_t i = 0; i < num_versions; ++i) { 5803 if (i < m_sdk_versions.size()) 5804 versions[i] = m_sdk_versions[i]; 5805 else 5806 versions[i] = 0; 5807 } 5808 } 5809 return m_sdk_versions.size(); 5810 } 5811 // Call the superclasses version that will empty out the data 5812 return ObjectFile::GetSDKVersion(versions, num_versions); 5813 } 5814 5815 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5816 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5817 } 5818 5819 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5820 return m_allow_assembly_emulation_unwind_plans; 5821 } 5822 5823 //------------------------------------------------------------------ 5824 // PluginInterface protocol 5825 //------------------------------------------------------------------ 5826 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5827 return GetPluginNameStatic(); 5828 } 5829 5830 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5831 5832 Section *ObjectFileMachO::GetMachHeaderSection() { 5833 // Find the first address of the mach header which is the first non-zero file 5834 // sized section whose file offset is zero. This is the base file address of 5835 // the mach-o file which can be subtracted from the vmaddr of the other 5836 // segments found in memory and added to the load address 5837 ModuleSP module_sp = GetModule(); 5838 if (module_sp) { 5839 SectionList *section_list = GetSectionList(); 5840 if (section_list) { 5841 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5842 const size_t num_sections = section_list->GetSize(); 5843 5844 for (size_t sect_idx = 0; sect_idx < num_sections && 5845 mach_base_file_addr == LLDB_INVALID_ADDRESS; 5846 ++sect_idx) { 5847 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5848 if (section && section->GetFileSize() > 0 && 5849 section->GetFileOffset() == 0 && 5850 section->IsThreadSpecific() == false && 5851 module_sp.get() == section->GetModule().get()) { 5852 return section; 5853 } 5854 } 5855 } 5856 } 5857 return nullptr; 5858 } 5859 5860 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 5861 lldb::addr_t mach_header_load_address, const Section *mach_header_section, 5862 const Section *section) { 5863 ModuleSP module_sp = GetModule(); 5864 if (module_sp && mach_header_section && section && 5865 mach_header_load_address != LLDB_INVALID_ADDRESS) { 5866 lldb::addr_t mach_header_file_addr = mach_header_section->GetFileAddress(); 5867 if (mach_header_file_addr != LLDB_INVALID_ADDRESS) { 5868 if (section && section->GetFileSize() > 0 && 5869 section->IsThreadSpecific() == false && 5870 module_sp.get() == section->GetModule().get()) { 5871 // Ignore __LINKEDIT and __DWARF segments 5872 if (section->GetName() == GetSegmentNameLINKEDIT()) { 5873 // Only map __LINKEDIT if we have an in memory image and this isn't a 5874 // kernel binary like a kext or mach_kernel. 5875 const bool is_memory_image = (bool)m_process_wp.lock(); 5876 const Strata strata = GetStrata(); 5877 if (is_memory_image == false || strata == eStrataKernel) 5878 return LLDB_INVALID_ADDRESS; 5879 } 5880 return section->GetFileAddress() - mach_header_file_addr + 5881 mach_header_load_address; 5882 } 5883 } 5884 } 5885 return LLDB_INVALID_ADDRESS; 5886 } 5887 5888 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 5889 bool value_is_offset) { 5890 ModuleSP module_sp = GetModule(); 5891 if (module_sp) { 5892 size_t num_loaded_sections = 0; 5893 SectionList *section_list = GetSectionList(); 5894 if (section_list) { 5895 const size_t num_sections = section_list->GetSize(); 5896 5897 if (value_is_offset) { 5898 // "value" is an offset to apply to each top level segment 5899 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5900 // Iterate through the object file sections to find all of the 5901 // sections that size on disk (to avoid __PAGEZERO) and load them 5902 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5903 if (section_sp && section_sp->GetFileSize() > 0 && 5904 section_sp->IsThreadSpecific() == false && 5905 module_sp.get() == section_sp->GetModule().get()) { 5906 // Ignore __LINKEDIT and __DWARF segments 5907 if (section_sp->GetName() == GetSegmentNameLINKEDIT()) { 5908 // Only map __LINKEDIT if we have an in memory image and this 5909 // isn't a kernel binary like a kext or mach_kernel. 5910 const bool is_memory_image = (bool)m_process_wp.lock(); 5911 const Strata strata = GetStrata(); 5912 if (is_memory_image == false || strata == eStrataKernel) 5913 continue; 5914 } 5915 if (target.GetSectionLoadList().SetSectionLoadAddress( 5916 section_sp, section_sp->GetFileAddress() + value)) 5917 ++num_loaded_sections; 5918 } 5919 } 5920 } else { 5921 // "value" is the new base address of the mach_header, adjust each 5922 // section accordingly 5923 5924 Section *mach_header_section = GetMachHeaderSection(); 5925 if (mach_header_section) { 5926 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5927 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5928 5929 lldb::addr_t section_load_addr = 5930 CalculateSectionLoadAddressForMemoryImage( 5931 value, mach_header_section, section_sp.get()); 5932 if (section_load_addr != LLDB_INVALID_ADDRESS) { 5933 if (target.GetSectionLoadList().SetSectionLoadAddress( 5934 section_sp, section_load_addr)) 5935 ++num_loaded_sections; 5936 } 5937 } 5938 } 5939 } 5940 } 5941 return num_loaded_sections > 0; 5942 } 5943 return false; 5944 } 5945 5946 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 5947 const FileSpec &outfile, Status &error) { 5948 if (process_sp) { 5949 Target &target = process_sp->GetTarget(); 5950 const ArchSpec target_arch = target.GetArchitecture(); 5951 const llvm::Triple &target_triple = target_arch.GetTriple(); 5952 if (target_triple.getVendor() == llvm::Triple::Apple && 5953 (target_triple.getOS() == llvm::Triple::MacOSX || 5954 target_triple.getOS() == llvm::Triple::IOS || 5955 target_triple.getOS() == llvm::Triple::WatchOS || 5956 target_triple.getOS() == llvm::Triple::TvOS)) { 5957 bool make_core = false; 5958 switch (target_arch.GetMachine()) { 5959 case llvm::Triple::aarch64: 5960 case llvm::Triple::arm: 5961 case llvm::Triple::thumb: 5962 case llvm::Triple::x86: 5963 case llvm::Triple::x86_64: 5964 make_core = true; 5965 break; 5966 default: 5967 error.SetErrorStringWithFormat("unsupported core architecture: %s", 5968 target_triple.str().c_str()); 5969 break; 5970 } 5971 5972 if (make_core) { 5973 std::vector<segment_command_64> segment_load_commands; 5974 // uint32_t range_info_idx = 0; 5975 MemoryRegionInfo range_info; 5976 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 5977 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 5978 const ByteOrder byte_order = target_arch.GetByteOrder(); 5979 if (range_error.Success()) { 5980 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 5981 const addr_t addr = range_info.GetRange().GetRangeBase(); 5982 const addr_t size = range_info.GetRange().GetByteSize(); 5983 5984 if (size == 0) 5985 break; 5986 5987 // Calculate correct protections 5988 uint32_t prot = 0; 5989 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 5990 prot |= VM_PROT_READ; 5991 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 5992 prot |= VM_PROT_WRITE; 5993 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 5994 prot |= VM_PROT_EXECUTE; 5995 5996 // printf ("[%3u] [0x%16.16" PRIx64 " - 5997 // 0x%16.16" PRIx64 ") %c%c%c\n", 5998 // range_info_idx, 5999 // addr, 6000 // size, 6001 // (prot & VM_PROT_READ ) ? 'r' : 6002 // '-', 6003 // (prot & VM_PROT_WRITE ) ? 'w' : 6004 // '-', 6005 // (prot & VM_PROT_EXECUTE) ? 'x' : 6006 // '-'); 6007 6008 if (prot != 0) { 6009 uint32_t cmd_type = LC_SEGMENT_64; 6010 uint32_t segment_size = sizeof(segment_command_64); 6011 if (addr_byte_size == 4) { 6012 cmd_type = LC_SEGMENT; 6013 segment_size = sizeof(segment_command); 6014 } 6015 segment_command_64 segment = { 6016 cmd_type, // uint32_t cmd; 6017 segment_size, // uint32_t cmdsize; 6018 {0}, // char segname[16]; 6019 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6020 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6021 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6022 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6023 prot, // uint32_t maxprot; 6024 prot, // uint32_t initprot; 6025 0, // uint32_t nsects; 6026 0}; // uint32_t flags; 6027 segment_load_commands.push_back(segment); 6028 } else { 6029 // No protections and a size of 1 used to be returned from old 6030 // debugservers when we asked about a region that was past the 6031 // last memory region and it indicates the end... 6032 if (size == 1) 6033 break; 6034 } 6035 6036 range_error = process_sp->GetMemoryRegionInfo( 6037 range_info.GetRange().GetRangeEnd(), range_info); 6038 if (range_error.Fail()) 6039 break; 6040 } 6041 6042 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6043 6044 mach_header_64 mach_header; 6045 if (addr_byte_size == 8) { 6046 mach_header.magic = MH_MAGIC_64; 6047 } else { 6048 mach_header.magic = MH_MAGIC; 6049 } 6050 mach_header.cputype = target_arch.GetMachOCPUType(); 6051 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6052 mach_header.filetype = MH_CORE; 6053 mach_header.ncmds = segment_load_commands.size(); 6054 mach_header.flags = 0; 6055 mach_header.reserved = 0; 6056 ThreadList &thread_list = process_sp->GetThreadList(); 6057 const uint32_t num_threads = thread_list.GetSize(); 6058 6059 // Make an array of LC_THREAD data items. Each one contains the 6060 // contents of the LC_THREAD load command. The data doesn't contain 6061 // the load command + load command size, we will add the load command 6062 // and load command size as we emit the data. 6063 std::vector<StreamString> LC_THREAD_datas(num_threads); 6064 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6065 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6066 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6067 LC_THREAD_data.SetByteOrder(byte_order); 6068 } 6069 for (uint32_t thread_idx = 0; thread_idx < num_threads; 6070 ++thread_idx) { 6071 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6072 if (thread_sp) { 6073 switch (mach_header.cputype) { 6074 case llvm::MachO::CPU_TYPE_ARM64: 6075 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6076 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6077 break; 6078 6079 case llvm::MachO::CPU_TYPE_ARM: 6080 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6081 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6082 break; 6083 6084 case llvm::MachO::CPU_TYPE_I386: 6085 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6086 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6087 break; 6088 6089 case llvm::MachO::CPU_TYPE_X86_64: 6090 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6091 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6092 break; 6093 } 6094 } 6095 } 6096 6097 // The size of the load command is the size of the segments... 6098 if (addr_byte_size == 8) { 6099 mach_header.sizeofcmds = segment_load_commands.size() * 6100 sizeof(struct segment_command_64); 6101 } else { 6102 mach_header.sizeofcmds = 6103 segment_load_commands.size() * sizeof(struct segment_command); 6104 } 6105 6106 // and the size of all LC_THREAD load command 6107 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6108 ++mach_header.ncmds; 6109 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6110 } 6111 6112 printf("mach_header: 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x " 6113 "0x%8.8x 0x%8.8x\n", 6114 mach_header.magic, mach_header.cputype, mach_header.cpusubtype, 6115 mach_header.filetype, mach_header.ncmds, 6116 mach_header.sizeofcmds, mach_header.flags, 6117 mach_header.reserved); 6118 6119 // Write the mach header 6120 buffer.PutHex32(mach_header.magic); 6121 buffer.PutHex32(mach_header.cputype); 6122 buffer.PutHex32(mach_header.cpusubtype); 6123 buffer.PutHex32(mach_header.filetype); 6124 buffer.PutHex32(mach_header.ncmds); 6125 buffer.PutHex32(mach_header.sizeofcmds); 6126 buffer.PutHex32(mach_header.flags); 6127 if (addr_byte_size == 8) { 6128 buffer.PutHex32(mach_header.reserved); 6129 } 6130 6131 // Skip the mach header and all load commands and align to the next 6132 // 0x1000 byte boundary 6133 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6134 if (file_offset & 0x00000fff) { 6135 file_offset += 0x00001000ull; 6136 file_offset &= (~0x00001000ull + 1); 6137 } 6138 6139 for (auto &segment : segment_load_commands) { 6140 segment.fileoff = file_offset; 6141 file_offset += segment.filesize; 6142 } 6143 6144 // Write out all of the LC_THREAD load commands 6145 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6146 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6147 buffer.PutHex32(LC_THREAD); 6148 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6149 buffer.Write(LC_THREAD_data.GetString().data(), 6150 LC_THREAD_data_size); 6151 } 6152 6153 // Write out all of the segment load commands 6154 for (const auto &segment : segment_load_commands) { 6155 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6156 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6157 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6158 segment.cmd, segment.cmdsize, segment.vmaddr, 6159 segment.vmaddr + segment.vmsize, segment.fileoff, 6160 segment.filesize, segment.maxprot, segment.initprot, 6161 segment.nsects, segment.flags); 6162 6163 buffer.PutHex32(segment.cmd); 6164 buffer.PutHex32(segment.cmdsize); 6165 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6166 if (addr_byte_size == 8) { 6167 buffer.PutHex64(segment.vmaddr); 6168 buffer.PutHex64(segment.vmsize); 6169 buffer.PutHex64(segment.fileoff); 6170 buffer.PutHex64(segment.filesize); 6171 } else { 6172 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6173 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6174 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6175 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6176 } 6177 buffer.PutHex32(segment.maxprot); 6178 buffer.PutHex32(segment.initprot); 6179 buffer.PutHex32(segment.nsects); 6180 buffer.PutHex32(segment.flags); 6181 } 6182 6183 File core_file; 6184 std::string core_file_path(outfile.GetPath()); 6185 error = core_file.Open(core_file_path.c_str(), 6186 File::eOpenOptionWrite | 6187 File::eOpenOptionTruncate | 6188 File::eOpenOptionCanCreate); 6189 if (error.Success()) { 6190 // Read 1 page at a time 6191 uint8_t bytes[0x1000]; 6192 // Write the mach header and load commands out to the core file 6193 size_t bytes_written = buffer.GetString().size(); 6194 error = core_file.Write(buffer.GetString().data(), bytes_written); 6195 if (error.Success()) { 6196 // Now write the file data for all memory segments in the process 6197 for (const auto &segment : segment_load_commands) { 6198 if (core_file.SeekFromStart(segment.fileoff) == -1) { 6199 error.SetErrorStringWithFormat( 6200 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6201 segment.fileoff, core_file_path.c_str()); 6202 break; 6203 } 6204 6205 printf("Saving %" PRId64 6206 " bytes of data for memory region at 0x%" PRIx64 "\n", 6207 segment.vmsize, segment.vmaddr); 6208 addr_t bytes_left = segment.vmsize; 6209 addr_t addr = segment.vmaddr; 6210 Status memory_read_error; 6211 while (bytes_left > 0 && error.Success()) { 6212 const size_t bytes_to_read = 6213 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6214 const size_t bytes_read = process_sp->ReadMemory( 6215 addr, bytes, bytes_to_read, memory_read_error); 6216 if (bytes_read == bytes_to_read) { 6217 size_t bytes_written = bytes_read; 6218 error = core_file.Write(bytes, bytes_written); 6219 bytes_left -= bytes_read; 6220 addr += bytes_read; 6221 } else { 6222 // Some pages within regions are not readable, those should 6223 // be zero filled 6224 memset(bytes, 0, bytes_to_read); 6225 size_t bytes_written = bytes_to_read; 6226 error = core_file.Write(bytes, bytes_written); 6227 bytes_left -= bytes_to_read; 6228 addr += bytes_to_read; 6229 } 6230 } 6231 } 6232 } 6233 } 6234 } else { 6235 error.SetErrorString( 6236 "process doesn't support getting memory region info"); 6237 } 6238 } 6239 return true; // This is the right plug to handle saving core files for 6240 // this process 6241 } 6242 } 6243 return false; 6244 } 6245