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 eSectionTypeDWARFDebugPubNames: 1204 case eSectionTypeDWARFDebugPubTypes: 1205 case eSectionTypeDWARFDebugRanges: 1206 case eSectionTypeDWARFDebugStr: 1207 case eSectionTypeDWARFDebugStrOffsets: 1208 case eSectionTypeDWARFAppleNames: 1209 case eSectionTypeDWARFAppleTypes: 1210 case eSectionTypeDWARFAppleNamespaces: 1211 case eSectionTypeDWARFAppleObjC: 1212 case eSectionTypeDWARFGNUDebugAltLink: 1213 return eAddressClassDebug; 1214 1215 case eSectionTypeEHFrame: 1216 case eSectionTypeARMexidx: 1217 case eSectionTypeARMextab: 1218 case eSectionTypeCompactUnwind: 1219 return eAddressClassRuntime; 1220 1221 case eSectionTypeAbsoluteAddress: 1222 case eSectionTypeELFSymbolTable: 1223 case eSectionTypeELFDynamicSymbols: 1224 case eSectionTypeELFRelocationEntries: 1225 case eSectionTypeELFDynamicLinkInfo: 1226 case eSectionTypeOther: 1227 return eAddressClassUnknown; 1228 } 1229 } 1230 } 1231 1232 const SymbolType symbol_type = symbol->GetType(); 1233 switch (symbol_type) { 1234 case eSymbolTypeAny: 1235 return eAddressClassUnknown; 1236 case eSymbolTypeAbsolute: 1237 return eAddressClassUnknown; 1238 1239 case eSymbolTypeCode: 1240 case eSymbolTypeTrampoline: 1241 case eSymbolTypeResolver: 1242 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1243 // For ARM we have a bit in the n_desc field of the symbol that tells 1244 // us ARM/Thumb which is bit 0x0008. 1245 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1246 return eAddressClassCodeAlternateISA; 1247 } 1248 return eAddressClassCode; 1249 1250 case eSymbolTypeData: 1251 return eAddressClassData; 1252 case eSymbolTypeRuntime: 1253 return eAddressClassRuntime; 1254 case eSymbolTypeException: 1255 return eAddressClassRuntime; 1256 case eSymbolTypeSourceFile: 1257 return eAddressClassDebug; 1258 case eSymbolTypeHeaderFile: 1259 return eAddressClassDebug; 1260 case eSymbolTypeObjectFile: 1261 return eAddressClassDebug; 1262 case eSymbolTypeCommonBlock: 1263 return eAddressClassDebug; 1264 case eSymbolTypeBlock: 1265 return eAddressClassDebug; 1266 case eSymbolTypeLocal: 1267 return eAddressClassData; 1268 case eSymbolTypeParam: 1269 return eAddressClassData; 1270 case eSymbolTypeVariable: 1271 return eAddressClassData; 1272 case eSymbolTypeVariableType: 1273 return eAddressClassDebug; 1274 case eSymbolTypeLineEntry: 1275 return eAddressClassDebug; 1276 case eSymbolTypeLineHeader: 1277 return eAddressClassDebug; 1278 case eSymbolTypeScopeBegin: 1279 return eAddressClassDebug; 1280 case eSymbolTypeScopeEnd: 1281 return eAddressClassDebug; 1282 case eSymbolTypeAdditional: 1283 return eAddressClassUnknown; 1284 case eSymbolTypeCompiler: 1285 return eAddressClassDebug; 1286 case eSymbolTypeInstrumentation: 1287 return eAddressClassDebug; 1288 case eSymbolTypeUndefined: 1289 return eAddressClassUnknown; 1290 case eSymbolTypeObjCClass: 1291 return eAddressClassRuntime; 1292 case eSymbolTypeObjCMetaClass: 1293 return eAddressClassRuntime; 1294 case eSymbolTypeObjCIVar: 1295 return eAddressClassRuntime; 1296 case eSymbolTypeReExported: 1297 return eAddressClassRuntime; 1298 } 1299 } 1300 } 1301 return eAddressClassUnknown; 1302 } 1303 1304 Symtab *ObjectFileMachO::GetSymtab() { 1305 ModuleSP module_sp(GetModule()); 1306 if (module_sp) { 1307 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1308 if (m_symtab_ap.get() == NULL) { 1309 m_symtab_ap.reset(new Symtab(this)); 1310 std::lock_guard<std::recursive_mutex> symtab_guard( 1311 m_symtab_ap->GetMutex()); 1312 ParseSymtab(); 1313 m_symtab_ap->Finalize(); 1314 } 1315 } 1316 return m_symtab_ap.get(); 1317 } 1318 1319 bool ObjectFileMachO::IsStripped() { 1320 if (m_dysymtab.cmd == 0) { 1321 ModuleSP module_sp(GetModule()); 1322 if (module_sp) { 1323 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1324 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1325 const lldb::offset_t load_cmd_offset = offset; 1326 1327 load_command lc; 1328 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1329 break; 1330 if (lc.cmd == LC_DYSYMTAB) { 1331 m_dysymtab.cmd = lc.cmd; 1332 m_dysymtab.cmdsize = lc.cmdsize; 1333 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1334 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1335 NULL) { 1336 // Clear m_dysymtab if we were unable to read all items from the 1337 // load command 1338 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1339 } 1340 } 1341 offset = load_cmd_offset + lc.cmdsize; 1342 } 1343 } 1344 } 1345 if (m_dysymtab.cmd) 1346 return m_dysymtab.nlocalsym <= 1; 1347 return false; 1348 } 1349 1350 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1351 EncryptedFileRanges result; 1352 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1353 1354 encryption_info_command encryption_cmd; 1355 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1356 const lldb::offset_t load_cmd_offset = offset; 1357 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1358 break; 1359 1360 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the 1361 // 3 fields we care about, so treat them the same. 1362 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1363 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1364 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1365 if (encryption_cmd.cryptid != 0) { 1366 EncryptedFileRanges::Entry entry; 1367 entry.SetRangeBase(encryption_cmd.cryptoff); 1368 entry.SetByteSize(encryption_cmd.cryptsize); 1369 result.Append(entry); 1370 } 1371 } 1372 } 1373 offset = load_cmd_offset + encryption_cmd.cmdsize; 1374 } 1375 1376 return result; 1377 } 1378 1379 void ObjectFileMachO::SanitizeSegmentCommand(segment_command_64 &seg_cmd, 1380 uint32_t cmd_idx) { 1381 if (m_length == 0 || seg_cmd.filesize == 0) 1382 return; 1383 1384 if (seg_cmd.fileoff > m_length) { 1385 // We have a load command that says it extends past the end of the file. 1386 // This is likely a corrupt file. We don't have any way to return an error 1387 // condition here (this method was likely invoked from something like 1388 // ObjectFile::GetSectionList()), so we just null out the section contents, 1389 // and dump a message to stdout. The most common case here is core file 1390 // debugging with a truncated file. 1391 const char *lc_segment_name = 1392 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1393 GetModule()->ReportWarning( 1394 "load command %u %s has a fileoff (0x%" PRIx64 1395 ") that extends beyond the end of the file (0x%" PRIx64 1396 "), ignoring this section", 1397 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1398 1399 seg_cmd.fileoff = 0; 1400 seg_cmd.filesize = 0; 1401 } 1402 1403 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1404 // We have a load command that says it extends past the end of the file. 1405 // This is likely a corrupt file. We don't have any way to return an error 1406 // condition here (this method was likely invoked from something like 1407 // ObjectFile::GetSectionList()), so we just null out the section contents, 1408 // and dump a message to stdout. The most common case here is core file 1409 // debugging with a truncated file. 1410 const char *lc_segment_name = 1411 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1412 GetModule()->ReportWarning( 1413 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1414 ") that extends beyond the end of the file (0x%" PRIx64 1415 "), the segment will be truncated to match", 1416 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1417 1418 // Truncate the length 1419 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1420 } 1421 } 1422 1423 static uint32_t GetSegmentPermissions(const segment_command_64 &seg_cmd) { 1424 uint32_t result = 0; 1425 if (seg_cmd.initprot & VM_PROT_READ) 1426 result |= ePermissionsReadable; 1427 if (seg_cmd.initprot & VM_PROT_WRITE) 1428 result |= ePermissionsWritable; 1429 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1430 result |= ePermissionsExecutable; 1431 return result; 1432 } 1433 1434 static lldb::SectionType GetSectionType(uint32_t flags, 1435 ConstString section_name) { 1436 1437 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1438 return eSectionTypeCode; 1439 1440 uint32_t mach_sect_type = flags & SECTION_TYPE; 1441 static ConstString g_sect_name_objc_data("__objc_data"); 1442 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1443 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1444 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1445 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1446 static ConstString g_sect_name_objc_const("__objc_const"); 1447 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1448 static ConstString g_sect_name_cfstring("__cfstring"); 1449 1450 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1451 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1452 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1453 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1454 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1455 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1456 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1457 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1458 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1459 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1460 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1461 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1462 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1463 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1464 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1465 static ConstString g_sect_name_eh_frame("__eh_frame"); 1466 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1467 static ConstString g_sect_name_text("__text"); 1468 static ConstString g_sect_name_data("__data"); 1469 static ConstString g_sect_name_go_symtab("__gosymtab"); 1470 1471 if (section_name == g_sect_name_dwarf_debug_abbrev) 1472 return eSectionTypeDWARFDebugAbbrev; 1473 if (section_name == g_sect_name_dwarf_debug_aranges) 1474 return eSectionTypeDWARFDebugAranges; 1475 if (section_name == g_sect_name_dwarf_debug_frame) 1476 return eSectionTypeDWARFDebugFrame; 1477 if (section_name == g_sect_name_dwarf_debug_info) 1478 return eSectionTypeDWARFDebugInfo; 1479 if (section_name == g_sect_name_dwarf_debug_line) 1480 return eSectionTypeDWARFDebugLine; 1481 if (section_name == g_sect_name_dwarf_debug_loc) 1482 return eSectionTypeDWARFDebugLoc; 1483 if (section_name == g_sect_name_dwarf_debug_macinfo) 1484 return eSectionTypeDWARFDebugMacInfo; 1485 if (section_name == g_sect_name_dwarf_debug_pubnames) 1486 return eSectionTypeDWARFDebugPubNames; 1487 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1488 return eSectionTypeDWARFDebugPubTypes; 1489 if (section_name == g_sect_name_dwarf_debug_ranges) 1490 return eSectionTypeDWARFDebugRanges; 1491 if (section_name == g_sect_name_dwarf_debug_str) 1492 return eSectionTypeDWARFDebugStr; 1493 if (section_name == g_sect_name_dwarf_apple_names) 1494 return eSectionTypeDWARFAppleNames; 1495 if (section_name == g_sect_name_dwarf_apple_types) 1496 return eSectionTypeDWARFAppleTypes; 1497 if (section_name == g_sect_name_dwarf_apple_namespaces) 1498 return eSectionTypeDWARFAppleNamespaces; 1499 if (section_name == g_sect_name_dwarf_apple_objc) 1500 return eSectionTypeDWARFAppleObjC; 1501 if (section_name == g_sect_name_objc_selrefs) 1502 return eSectionTypeDataCStringPointers; 1503 if (section_name == g_sect_name_objc_msgrefs) 1504 return eSectionTypeDataObjCMessageRefs; 1505 if (section_name == g_sect_name_eh_frame) 1506 return eSectionTypeEHFrame; 1507 if (section_name == g_sect_name_compact_unwind) 1508 return eSectionTypeCompactUnwind; 1509 if (section_name == g_sect_name_cfstring) 1510 return eSectionTypeDataObjCCFStrings; 1511 if (section_name == g_sect_name_go_symtab) 1512 return eSectionTypeGoSymtab; 1513 if (section_name == g_sect_name_objc_data || 1514 section_name == g_sect_name_objc_classrefs || 1515 section_name == g_sect_name_objc_superrefs || 1516 section_name == g_sect_name_objc_const || 1517 section_name == g_sect_name_objc_classlist) { 1518 return eSectionTypeDataPointers; 1519 } 1520 1521 switch (mach_sect_type) { 1522 // TODO: categorize sections by other flags for regular sections 1523 case S_REGULAR: 1524 if (section_name == g_sect_name_text) 1525 return eSectionTypeCode; 1526 if (section_name == g_sect_name_data) 1527 return eSectionTypeData; 1528 return eSectionTypeOther; 1529 case S_ZEROFILL: 1530 return eSectionTypeZeroFill; 1531 case S_CSTRING_LITERALS: // section with only literal C strings 1532 return eSectionTypeDataCString; 1533 case S_4BYTE_LITERALS: // section with only 4 byte literals 1534 return eSectionTypeData4; 1535 case S_8BYTE_LITERALS: // section with only 8 byte literals 1536 return eSectionTypeData8; 1537 case S_LITERAL_POINTERS: // section with only pointers to literals 1538 return eSectionTypeDataPointers; 1539 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1540 return eSectionTypeDataPointers; 1541 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1542 return eSectionTypeDataPointers; 1543 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1544 // the reserved2 field 1545 return eSectionTypeCode; 1546 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1547 // initialization 1548 return eSectionTypeDataPointers; 1549 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1550 // termination 1551 return eSectionTypeDataPointers; 1552 case S_COALESCED: 1553 return eSectionTypeOther; 1554 case S_GB_ZEROFILL: 1555 return eSectionTypeZeroFill; 1556 case S_INTERPOSING: // section with only pairs of function pointers for 1557 // interposing 1558 return eSectionTypeCode; 1559 case S_16BYTE_LITERALS: // section with only 16 byte literals 1560 return eSectionTypeData16; 1561 case S_DTRACE_DOF: 1562 return eSectionTypeDebug; 1563 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1564 return eSectionTypeDataPointers; 1565 default: 1566 return eSectionTypeOther; 1567 } 1568 } 1569 1570 struct ObjectFileMachO::SegmentParsingContext { 1571 const EncryptedFileRanges EncryptedRanges; 1572 lldb_private::SectionList &UnifiedList; 1573 uint32_t NextSegmentIdx = 0; 1574 uint32_t NextSectionIdx = 0; 1575 bool FileAddressesChanged = false; 1576 1577 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1578 lldb_private::SectionList &UnifiedList) 1579 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1580 }; 1581 1582 void ObjectFileMachO::ProcessSegmentCommand(const load_command &load_cmd_, 1583 lldb::offset_t offset, 1584 uint32_t cmd_idx, 1585 SegmentParsingContext &context) { 1586 segment_command_64 load_cmd; 1587 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1588 1589 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1590 return; 1591 1592 ModuleSP module_sp = GetModule(); 1593 const bool is_core = GetType() == eTypeCoreFile; 1594 const bool is_dsym = (m_header.filetype == MH_DSYM); 1595 bool add_section = true; 1596 bool add_to_unified = true; 1597 ConstString const_segname( 1598 load_cmd.segname, 1599 std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 1600 1601 SectionSP unified_section_sp( 1602 context.UnifiedList.FindSectionByName(const_segname)); 1603 if (is_dsym && unified_section_sp) { 1604 if (const_segname == GetSegmentNameLINKEDIT()) { 1605 // We need to keep the __LINKEDIT segment private to this object file 1606 // only 1607 add_to_unified = false; 1608 } else { 1609 // This is the dSYM file and this section has already been created by the 1610 // object file, no need to create it. 1611 add_section = false; 1612 } 1613 } 1614 load_cmd.vmaddr = m_data.GetAddress(&offset); 1615 load_cmd.vmsize = m_data.GetAddress(&offset); 1616 load_cmd.fileoff = m_data.GetAddress(&offset); 1617 load_cmd.filesize = m_data.GetAddress(&offset); 1618 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1619 return; 1620 1621 SanitizeSegmentCommand(load_cmd, cmd_idx); 1622 1623 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1624 const bool segment_is_encrypted = 1625 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1626 1627 // Keep a list of mach segments around in case we need to get at data that 1628 // isn't stored in the abstracted Sections. 1629 m_mach_segments.push_back(load_cmd); 1630 1631 // Use a segment ID of the segment index shifted left by 8 so they never 1632 // conflict with any of the sections. 1633 SectionSP segment_sp; 1634 if (add_section && (const_segname || is_core)) { 1635 segment_sp.reset(new Section( 1636 module_sp, // Module to which this section belongs 1637 this, // Object file to which this sections belongs 1638 ++context.NextSegmentIdx 1639 << 8, // Section ID is the 1 based segment index 1640 // shifted right by 8 bits as not to collide with any of the 256 1641 // section IDs that are possible 1642 const_segname, // Name of this section 1643 eSectionTypeContainer, // This section is a container of other 1644 // sections. 1645 load_cmd.vmaddr, // File VM address == addresses as they are 1646 // found in the object file 1647 load_cmd.vmsize, // VM size in bytes of this section 1648 load_cmd.fileoff, // Offset to the data for this section in 1649 // the file 1650 load_cmd.filesize, // Size in bytes of this section as found 1651 // in the file 1652 0, // Segments have no alignment information 1653 load_cmd.flags)); // Flags for this section 1654 1655 segment_sp->SetIsEncrypted(segment_is_encrypted); 1656 m_sections_ap->AddSection(segment_sp); 1657 segment_sp->SetPermissions(segment_permissions); 1658 if (add_to_unified) 1659 context.UnifiedList.AddSection(segment_sp); 1660 } else if (unified_section_sp) { 1661 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1662 // Check to see if the module was read from memory? 1663 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) { 1664 // We have a module that is in memory and needs to have its file 1665 // address adjusted. We need to do this because when we load a file 1666 // from memory, its addresses will be slid already, yet the addresses 1667 // in the new symbol file will still be unslid. Since everything is 1668 // stored as section offset, this shouldn't cause any problems. 1669 1670 // Make sure we've parsed the symbol table from the ObjectFile before 1671 // we go around changing its Sections. 1672 module_sp->GetObjectFile()->GetSymtab(); 1673 // eh_frame would present the same problems but we parse that on a per- 1674 // function basis as-needed so it's more difficult to remove its use of 1675 // the Sections. Realistically, the environments where this code path 1676 // will be taken will not have eh_frame sections. 1677 1678 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1679 1680 // Notify the module that the section addresses have been changed once 1681 // we're done so any file-address caches can be updated. 1682 context.FileAddressesChanged = true; 1683 } 1684 } 1685 m_sections_ap->AddSection(unified_section_sp); 1686 } 1687 1688 struct section_64 sect64; 1689 ::memset(§64, 0, sizeof(sect64)); 1690 // Push a section into our mach sections for the section at index zero 1691 // (NO_SECT) if we don't have any mach sections yet... 1692 if (m_mach_sections.empty()) 1693 m_mach_sections.push_back(sect64); 1694 uint32_t segment_sect_idx; 1695 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1696 1697 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1698 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1699 ++segment_sect_idx) { 1700 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1701 sizeof(sect64.sectname)) == NULL) 1702 break; 1703 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1704 sizeof(sect64.segname)) == NULL) 1705 break; 1706 sect64.addr = m_data.GetAddress(&offset); 1707 sect64.size = m_data.GetAddress(&offset); 1708 1709 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1710 break; 1711 1712 // Keep a list of mach sections around in case we need to get at data that 1713 // isn't stored in the abstracted Sections. 1714 m_mach_sections.push_back(sect64); 1715 1716 if (add_section) { 1717 ConstString section_name( 1718 sect64.sectname, 1719 std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1720 if (!const_segname) { 1721 // We have a segment with no name so we need to conjure up segments 1722 // that correspond to the section's segname if there isn't already such 1723 // a section. If there is such a section, we resize the section so that 1724 // it spans all sections. We also mark these sections as fake so 1725 // address matches don't hit if they land in the gaps between the child 1726 // sections. 1727 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1728 sizeof(sect64.segname)); 1729 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1730 if (segment_sp.get()) { 1731 Section *segment = segment_sp.get(); 1732 // Grow the section size as needed. 1733 const lldb::addr_t sect64_min_addr = sect64.addr; 1734 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1735 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1736 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1737 const lldb::addr_t curr_seg_max_addr = 1738 curr_seg_min_addr + curr_seg_byte_size; 1739 if (sect64_min_addr >= curr_seg_min_addr) { 1740 const lldb::addr_t new_seg_byte_size = 1741 sect64_max_addr - curr_seg_min_addr; 1742 // Only grow the section size if needed 1743 if (new_seg_byte_size > curr_seg_byte_size) 1744 segment->SetByteSize(new_seg_byte_size); 1745 } else { 1746 // We need to change the base address of the segment and adjust the 1747 // child section offsets for all existing children. 1748 const lldb::addr_t slide_amount = 1749 sect64_min_addr - curr_seg_min_addr; 1750 segment->Slide(slide_amount, false); 1751 segment->GetChildren().Slide(-slide_amount, false); 1752 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1753 } 1754 1755 // Grow the section size as needed. 1756 if (sect64.offset) { 1757 const lldb::addr_t segment_min_file_offset = 1758 segment->GetFileOffset(); 1759 const lldb::addr_t segment_max_file_offset = 1760 segment_min_file_offset + segment->GetFileSize(); 1761 1762 const lldb::addr_t section_min_file_offset = sect64.offset; 1763 const lldb::addr_t section_max_file_offset = 1764 section_min_file_offset + sect64.size; 1765 const lldb::addr_t new_file_offset = 1766 std::min(section_min_file_offset, segment_min_file_offset); 1767 const lldb::addr_t new_file_size = 1768 std::max(section_max_file_offset, segment_max_file_offset) - 1769 new_file_offset; 1770 segment->SetFileOffset(new_file_offset); 1771 segment->SetFileSize(new_file_size); 1772 } 1773 } else { 1774 // Create a fake section for the section's named segment 1775 segment_sp.reset(new Section( 1776 segment_sp, // Parent section 1777 module_sp, // Module to which this section belongs 1778 this, // Object file to which this section belongs 1779 ++context.NextSegmentIdx 1780 << 8, // Section ID is the 1 based segment index 1781 // shifted right by 8 bits as not to 1782 // collide with any of the 256 section IDs 1783 // that are possible 1784 const_segname, // Name of this section 1785 eSectionTypeContainer, // This section is a container of 1786 // other sections. 1787 sect64.addr, // File VM address == addresses as they are 1788 // found in the object file 1789 sect64.size, // VM size in bytes of this section 1790 sect64.offset, // Offset to the data for this section in 1791 // the file 1792 sect64.offset ? sect64.size : 0, // Size in bytes of 1793 // this section as 1794 // found in the file 1795 sect64.align, 1796 load_cmd.flags)); // Flags for this section 1797 segment_sp->SetIsFake(true); 1798 segment_sp->SetPermissions(segment_permissions); 1799 m_sections_ap->AddSection(segment_sp); 1800 if (add_to_unified) 1801 context.UnifiedList.AddSection(segment_sp); 1802 segment_sp->SetIsEncrypted(segment_is_encrypted); 1803 } 1804 } 1805 assert(segment_sp.get()); 1806 1807 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1808 1809 SectionSP section_sp(new Section( 1810 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1811 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1812 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1813 sect64.flags)); 1814 // Set the section to be encrypted to match the segment 1815 1816 bool section_is_encrypted = false; 1817 if (!segment_is_encrypted && load_cmd.filesize != 0) 1818 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1819 sect64.offset) != NULL; 1820 1821 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1822 section_sp->SetPermissions(segment_permissions); 1823 segment_sp->GetChildren().AddSection(section_sp); 1824 1825 if (segment_sp->IsFake()) { 1826 segment_sp.reset(); 1827 const_segname.Clear(); 1828 } 1829 } 1830 } 1831 if (segment_sp && is_dsym) { 1832 if (first_segment_sectID <= context.NextSectionIdx) { 1833 lldb::user_id_t sect_uid; 1834 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1835 ++sect_uid) { 1836 SectionSP curr_section_sp( 1837 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1838 SectionSP next_section_sp; 1839 if (sect_uid + 1 <= context.NextSectionIdx) 1840 next_section_sp = 1841 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1842 1843 if (curr_section_sp.get()) { 1844 if (curr_section_sp->GetByteSize() == 0) { 1845 if (next_section_sp.get() != NULL) 1846 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1847 curr_section_sp->GetFileAddress()); 1848 else 1849 curr_section_sp->SetByteSize(load_cmd.vmsize); 1850 } 1851 } 1852 } 1853 } 1854 } 1855 } 1856 1857 void ObjectFileMachO::ProcessDysymtabCommand(const load_command &load_cmd, 1858 lldb::offset_t offset) { 1859 m_dysymtab.cmd = load_cmd.cmd; 1860 m_dysymtab.cmdsize = load_cmd.cmdsize; 1861 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1862 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1863 } 1864 1865 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1866 if (m_sections_ap) 1867 return; 1868 1869 m_sections_ap.reset(new SectionList()); 1870 1871 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1872 // bool dump_sections = false; 1873 ModuleSP module_sp(GetModule()); 1874 1875 offset = MachHeaderSizeFromMagic(m_header.magic); 1876 1877 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1878 struct load_command load_cmd; 1879 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1880 const lldb::offset_t load_cmd_offset = offset; 1881 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1882 break; 1883 1884 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1885 ProcessSegmentCommand(load_cmd, offset, i, context); 1886 else if (load_cmd.cmd == LC_DYSYMTAB) 1887 ProcessDysymtabCommand(load_cmd, offset); 1888 1889 offset = load_cmd_offset + load_cmd.cmdsize; 1890 } 1891 1892 if (context.FileAddressesChanged && module_sp) 1893 module_sp->SectionFileAddressesChanged(); 1894 } 1895 1896 class MachSymtabSectionInfo { 1897 public: 1898 MachSymtabSectionInfo(SectionList *section_list) 1899 : m_section_list(section_list), m_section_infos() { 1900 // Get the number of sections down to a depth of 1 to include all segments 1901 // and their sections, but no other sections that may be added for debug 1902 // map or 1903 m_section_infos.resize(section_list->GetNumSections(1)); 1904 } 1905 1906 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1907 if (n_sect == 0) 1908 return SectionSP(); 1909 if (n_sect < m_section_infos.size()) { 1910 if (!m_section_infos[n_sect].section_sp) { 1911 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1912 m_section_infos[n_sect].section_sp = section_sp; 1913 if (section_sp) { 1914 m_section_infos[n_sect].vm_range.SetBaseAddress( 1915 section_sp->GetFileAddress()); 1916 m_section_infos[n_sect].vm_range.SetByteSize( 1917 section_sp->GetByteSize()); 1918 } else { 1919 Host::SystemLog(Host::eSystemLogError, 1920 "error: unable to find section for section %u\n", 1921 n_sect); 1922 } 1923 } 1924 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1925 // Symbol is in section. 1926 return m_section_infos[n_sect].section_sp; 1927 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1928 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1929 file_addr) { 1930 // Symbol is in section with zero size, but has the same start address 1931 // as the section. This can happen with linker symbols (symbols that 1932 // start with the letter 'l' or 'L'. 1933 return m_section_infos[n_sect].section_sp; 1934 } 1935 } 1936 return m_section_list->FindSectionContainingFileAddress(file_addr); 1937 } 1938 1939 protected: 1940 struct SectionInfo { 1941 SectionInfo() : vm_range(), section_sp() {} 1942 1943 VMRange vm_range; 1944 SectionSP section_sp; 1945 }; 1946 SectionList *m_section_list; 1947 std::vector<SectionInfo> m_section_infos; 1948 }; 1949 1950 struct TrieEntry { 1951 TrieEntry() 1952 : name(), address(LLDB_INVALID_ADDRESS), flags(0), other(0), 1953 import_name() {} 1954 1955 void Clear() { 1956 name.Clear(); 1957 address = LLDB_INVALID_ADDRESS; 1958 flags = 0; 1959 other = 0; 1960 import_name.Clear(); 1961 } 1962 1963 void Dump() const { 1964 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1965 static_cast<unsigned long long>(address), 1966 static_cast<unsigned long long>(flags), 1967 static_cast<unsigned long long>(other), name.GetCString()); 1968 if (import_name) 1969 printf(" -> \"%s\"\n", import_name.GetCString()); 1970 else 1971 printf("\n"); 1972 } 1973 ConstString name; 1974 uint64_t address; 1975 uint64_t flags; 1976 uint64_t other; 1977 ConstString import_name; 1978 }; 1979 1980 struct TrieEntryWithOffset { 1981 lldb::offset_t nodeOffset; 1982 TrieEntry entry; 1983 1984 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1985 1986 void Dump(uint32_t idx) const { 1987 printf("[%3u] 0x%16.16llx: ", idx, 1988 static_cast<unsigned long long>(nodeOffset)); 1989 entry.Dump(); 1990 } 1991 1992 bool operator<(const TrieEntryWithOffset &other) const { 1993 return (nodeOffset < other.nodeOffset); 1994 } 1995 }; 1996 1997 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 1998 const bool is_arm, 1999 std::vector<llvm::StringRef> &nameSlices, 2000 std::set<lldb::addr_t> &resolver_addresses, 2001 std::vector<TrieEntryWithOffset> &output) { 2002 if (!data.ValidOffset(offset)) 2003 return true; 2004 2005 const uint64_t terminalSize = data.GetULEB128(&offset); 2006 lldb::offset_t children_offset = offset + terminalSize; 2007 if (terminalSize != 0) { 2008 TrieEntryWithOffset e(offset); 2009 e.entry.flags = data.GetULEB128(&offset); 2010 const char *import_name = NULL; 2011 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2012 e.entry.address = 0; 2013 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2014 import_name = data.GetCStr(&offset); 2015 } else { 2016 e.entry.address = data.GetULEB128(&offset); 2017 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2018 e.entry.other = data.GetULEB128(&offset); 2019 uint64_t resolver_addr = e.entry.other; 2020 if (is_arm) 2021 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2022 resolver_addresses.insert(resolver_addr); 2023 } else 2024 e.entry.other = 0; 2025 } 2026 // Only add symbols that are reexport symbols with a valid import name 2027 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && 2028 import_name[0]) { 2029 std::string name; 2030 if (!nameSlices.empty()) { 2031 for (auto name_slice : nameSlices) 2032 name.append(name_slice.data(), name_slice.size()); 2033 } 2034 if (name.size() > 1) { 2035 // Skip the leading '_' 2036 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2037 } 2038 if (import_name) { 2039 // Skip the leading '_' 2040 e.entry.import_name.SetCString(import_name + 1); 2041 } 2042 output.push_back(e); 2043 } 2044 } 2045 2046 const uint8_t childrenCount = data.GetU8(&children_offset); 2047 for (uint8_t i = 0; i < childrenCount; ++i) { 2048 const char *cstr = data.GetCStr(&children_offset); 2049 if (cstr) 2050 nameSlices.push_back(llvm::StringRef(cstr)); 2051 else 2052 return false; // Corrupt data 2053 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2054 if (childNodeOffset) { 2055 if (!ParseTrieEntries(data, childNodeOffset, is_arm, nameSlices, 2056 resolver_addresses, output)) { 2057 return false; 2058 } 2059 } 2060 nameSlices.pop_back(); 2061 } 2062 return true; 2063 } 2064 2065 // Read the UUID out of a dyld_shared_cache file on-disk. 2066 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2067 const ByteOrder byte_order, 2068 const uint32_t addr_byte_size) { 2069 UUID dsc_uuid; 2070 DataBufferSP DscData = MapFileData( 2071 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2072 if (!DscData) 2073 return dsc_uuid; 2074 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2075 2076 char version_str[7]; 2077 lldb::offset_t offset = 0; 2078 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2079 version_str[6] = '\0'; 2080 if (strcmp(version_str, "dyld_v") == 0) { 2081 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2082 uint8_t uuid_bytes[sizeof(uuid_t)]; 2083 memcpy(uuid_bytes, dsc_header_data.GetData(&offset, sizeof(uuid_t)), 2084 sizeof(uuid_t)); 2085 dsc_uuid.SetBytes(uuid_bytes); 2086 } 2087 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2088 if (log && dsc_uuid.IsValid()) { 2089 log->Printf("Shared cache %s has UUID %s", dyld_shared_cache.GetPath().c_str(), 2090 dsc_uuid.GetAsString().c_str()); 2091 } 2092 return dsc_uuid; 2093 } 2094 2095 size_t ObjectFileMachO::ParseSymtab() { 2096 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2097 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2098 m_file.GetFilename().AsCString("")); 2099 ModuleSP module_sp(GetModule()); 2100 if (!module_sp) 2101 return 0; 2102 2103 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2104 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2105 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2106 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2107 FunctionStarts function_starts; 2108 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2109 uint32_t i; 2110 FileSpecList dylib_files; 2111 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2112 static const llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2113 static const llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2114 static const llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2115 2116 for (i = 0; i < m_header.ncmds; ++i) { 2117 const lldb::offset_t cmd_offset = offset; 2118 // Read in the load command and load command size 2119 struct load_command lc; 2120 if (m_data.GetU32(&offset, &lc, 2) == NULL) 2121 break; 2122 // Watch for the symbol table load command 2123 switch (lc.cmd) { 2124 case LC_SYMTAB: 2125 symtab_load_command.cmd = lc.cmd; 2126 symtab_load_command.cmdsize = lc.cmdsize; 2127 // Read in the rest of the symtab load command 2128 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2129 0) // fill in symoff, nsyms, stroff, strsize fields 2130 return 0; 2131 if (symtab_load_command.symoff == 0) { 2132 if (log) 2133 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2134 return 0; 2135 } 2136 2137 if (symtab_load_command.stroff == 0) { 2138 if (log) 2139 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2140 return 0; 2141 } 2142 2143 if (symtab_load_command.nsyms == 0) { 2144 if (log) 2145 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2146 return 0; 2147 } 2148 2149 if (symtab_load_command.strsize == 0) { 2150 if (log) 2151 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2152 return 0; 2153 } 2154 break; 2155 2156 case LC_DYLD_INFO: 2157 case LC_DYLD_INFO_ONLY: 2158 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2159 dyld_info.cmd = lc.cmd; 2160 dyld_info.cmdsize = lc.cmdsize; 2161 } else { 2162 memset(&dyld_info, 0, sizeof(dyld_info)); 2163 } 2164 break; 2165 2166 case LC_LOAD_DYLIB: 2167 case LC_LOAD_WEAK_DYLIB: 2168 case LC_REEXPORT_DYLIB: 2169 case LC_LOADFVMLIB: 2170 case LC_LOAD_UPWARD_DYLIB: { 2171 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2172 const char *path = m_data.PeekCStr(name_offset); 2173 if (path) { 2174 FileSpec file_spec(path, false); 2175 // Strip the path if there is @rpath, @executable, etc so we just use 2176 // the basename 2177 if (path[0] == '@') 2178 file_spec.GetDirectory().Clear(); 2179 2180 if (lc.cmd == LC_REEXPORT_DYLIB) { 2181 m_reexported_dylibs.AppendIfUnique(file_spec); 2182 } 2183 2184 dylib_files.Append(file_spec); 2185 } 2186 } break; 2187 2188 case LC_FUNCTION_STARTS: 2189 function_starts_load_command.cmd = lc.cmd; 2190 function_starts_load_command.cmdsize = lc.cmdsize; 2191 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2192 NULL) // fill in symoff, nsyms, stroff, strsize fields 2193 memset(&function_starts_load_command, 0, 2194 sizeof(function_starts_load_command)); 2195 break; 2196 2197 default: 2198 break; 2199 } 2200 offset = cmd_offset + lc.cmdsize; 2201 } 2202 2203 if (symtab_load_command.cmd) { 2204 Symtab *symtab = m_symtab_ap.get(); 2205 SectionList *section_list = GetSectionList(); 2206 if (section_list == NULL) 2207 return 0; 2208 2209 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2210 const ByteOrder byte_order = m_data.GetByteOrder(); 2211 bool bit_width_32 = addr_byte_size == 4; 2212 const size_t nlist_byte_size = 2213 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2214 2215 DataExtractor nlist_data(NULL, 0, byte_order, addr_byte_size); 2216 DataExtractor strtab_data(NULL, 0, byte_order, addr_byte_size); 2217 DataExtractor function_starts_data(NULL, 0, byte_order, addr_byte_size); 2218 DataExtractor indirect_symbol_index_data(NULL, 0, byte_order, 2219 addr_byte_size); 2220 DataExtractor dyld_trie_data(NULL, 0, byte_order, addr_byte_size); 2221 2222 const addr_t nlist_data_byte_size = 2223 symtab_load_command.nsyms * nlist_byte_size; 2224 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2225 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2226 2227 ProcessSP process_sp(m_process_wp.lock()); 2228 Process *process = process_sp.get(); 2229 2230 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2231 2232 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2233 Target &target = process->GetTarget(); 2234 2235 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2236 2237 SectionSP linkedit_section_sp( 2238 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2239 // Reading mach file from memory in a process or core file... 2240 2241 if (linkedit_section_sp) { 2242 addr_t linkedit_load_addr = 2243 linkedit_section_sp->GetLoadBaseAddress(&target); 2244 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2245 // We might be trying to access the symbol table before the 2246 // __LINKEDIT's load address has been set in the target. We can't 2247 // fail to read the symbol table, so calculate the right address 2248 // manually 2249 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2250 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2251 } 2252 2253 const addr_t linkedit_file_offset = 2254 linkedit_section_sp->GetFileOffset(); 2255 const addr_t symoff_addr = linkedit_load_addr + 2256 symtab_load_command.symoff - 2257 linkedit_file_offset; 2258 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2259 linkedit_file_offset; 2260 2261 bool data_was_read = false; 2262 2263 #if defined(__APPLE__) && \ 2264 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2265 if (m_header.flags & 0x80000000u && 2266 process->GetAddressByteSize() == sizeof(void *)) { 2267 // This mach-o memory file is in the dyld shared cache. If this 2268 // program is not remote and this is iOS, then this process will 2269 // share the same shared cache as the process we are debugging and we 2270 // can read the entire __LINKEDIT from the address space in this 2271 // process. This is a needed optimization that is used for local iOS 2272 // debugging only since all shared libraries in the shared cache do 2273 // not have corresponding files that exist in the file system of the 2274 // device. They have been combined into a single file. This means we 2275 // always have to load these files from memory. All of the symbol and 2276 // string tables from all of the __LINKEDIT sections from the shared 2277 // libraries in the shared cache have been merged into a single large 2278 // symbol and string table. Reading all of this symbol and string 2279 // table data across can slow down debug launch times, so we optimize 2280 // this by reading the memory for the __LINKEDIT section from this 2281 // process. 2282 2283 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 2284 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 2285 bool use_lldb_cache = true; 2286 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2287 lldb_shared_cache != process_shared_cache) { 2288 use_lldb_cache = false; 2289 ModuleSP module_sp(GetModule()); 2290 if (module_sp) 2291 module_sp->ReportWarning("shared cache in process does not match " 2292 "lldb's own shared cache, startup will " 2293 "be slow."); 2294 } 2295 2296 PlatformSP platform_sp(target.GetPlatform()); 2297 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2298 data_was_read = true; 2299 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2300 eByteOrderLittle); 2301 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2302 eByteOrderLittle); 2303 if (function_starts_load_command.cmd) { 2304 const addr_t func_start_addr = 2305 linkedit_load_addr + function_starts_load_command.dataoff - 2306 linkedit_file_offset; 2307 function_starts_data.SetData( 2308 (void *)func_start_addr, 2309 function_starts_load_command.datasize, eByteOrderLittle); 2310 } 2311 } 2312 } 2313 #endif 2314 2315 if (!data_was_read) { 2316 // Always load dyld - the dynamic linker - from memory if we didn't 2317 // find a binary anywhere else. lldb will not register 2318 // dylib/framework/bundle loads/unloads if we don't have the dyld 2319 // symbols, we force dyld to load from memory despite the user's 2320 // target.memory-module-load-level setting. 2321 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2322 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2323 DataBufferSP nlist_data_sp( 2324 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2325 if (nlist_data_sp) 2326 nlist_data.SetData(nlist_data_sp, 0, 2327 nlist_data_sp->GetByteSize()); 2328 // Load strings individually from memory when loading from memory 2329 // since shared cache string tables contain strings for all symbols 2330 // from all shared cached libraries DataBufferSP strtab_data_sp 2331 // (ReadMemory (process_sp, strtab_addr, 2332 // strtab_data_byte_size)); 2333 // if (strtab_data_sp) 2334 // strtab_data.SetData (strtab_data_sp, 0, 2335 // strtab_data_sp->GetByteSize()); 2336 if (m_dysymtab.nindirectsyms != 0) { 2337 const addr_t indirect_syms_addr = linkedit_load_addr + 2338 m_dysymtab.indirectsymoff - 2339 linkedit_file_offset; 2340 DataBufferSP indirect_syms_data_sp( 2341 ReadMemory(process_sp, indirect_syms_addr, 2342 m_dysymtab.nindirectsyms * 4)); 2343 if (indirect_syms_data_sp) 2344 indirect_symbol_index_data.SetData( 2345 indirect_syms_data_sp, 0, 2346 indirect_syms_data_sp->GetByteSize()); 2347 } 2348 } else if (memory_module_load_level >= 2349 eMemoryModuleLoadLevelPartial) { 2350 if (function_starts_load_command.cmd) { 2351 const addr_t func_start_addr = 2352 linkedit_load_addr + function_starts_load_command.dataoff - 2353 linkedit_file_offset; 2354 DataBufferSP func_start_data_sp( 2355 ReadMemory(process_sp, func_start_addr, 2356 function_starts_load_command.datasize)); 2357 if (func_start_data_sp) 2358 function_starts_data.SetData(func_start_data_sp, 0, 2359 func_start_data_sp->GetByteSize()); 2360 } 2361 } 2362 } 2363 } 2364 } else { 2365 nlist_data.SetData(m_data, symtab_load_command.symoff, 2366 nlist_data_byte_size); 2367 strtab_data.SetData(m_data, symtab_load_command.stroff, 2368 strtab_data_byte_size); 2369 2370 if (dyld_info.export_size > 0) { 2371 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2372 dyld_info.export_size); 2373 } 2374 2375 if (m_dysymtab.nindirectsyms != 0) { 2376 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2377 m_dysymtab.nindirectsyms * 4); 2378 } 2379 if (function_starts_load_command.cmd) { 2380 function_starts_data.SetData(m_data, 2381 function_starts_load_command.dataoff, 2382 function_starts_load_command.datasize); 2383 } 2384 } 2385 2386 if (nlist_data.GetByteSize() == 0 && 2387 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2388 if (log) 2389 module_sp->LogMessage(log, "failed to read nlist data"); 2390 return 0; 2391 } 2392 2393 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2394 if (!have_strtab_data) { 2395 if (process) { 2396 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2397 if (log) 2398 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2399 return 0; 2400 } 2401 } else { 2402 if (log) 2403 module_sp->LogMessage(log, "failed to read strtab data"); 2404 return 0; 2405 } 2406 } 2407 2408 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2409 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2410 const ConstString &g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2411 const ConstString &g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2412 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2413 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2414 SectionSP text_section_sp( 2415 section_list->FindSectionByName(g_segment_name_TEXT)); 2416 SectionSP data_section_sp( 2417 section_list->FindSectionByName(g_segment_name_DATA)); 2418 SectionSP data_dirty_section_sp( 2419 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2420 SectionSP data_const_section_sp( 2421 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2422 SectionSP objc_section_sp( 2423 section_list->FindSectionByName(g_segment_name_OBJC)); 2424 SectionSP eh_frame_section_sp; 2425 if (text_section_sp.get()) 2426 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2427 g_section_name_eh_frame); 2428 else 2429 eh_frame_section_sp = 2430 section_list->FindSectionByName(g_section_name_eh_frame); 2431 2432 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2433 2434 // lldb works best if it knows the start address of all functions in a 2435 // module. Linker symbols or debug info are normally the best source of 2436 // information for start addr / size but they may be stripped in a released 2437 // binary. Two additional sources of information exist in Mach-O binaries: 2438 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2439 // function's start address in the 2440 // binary, relative to the text section. 2441 // eh_frame - the eh_frame FDEs have the start addr & size of 2442 // each function 2443 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2444 // all modern binaries. 2445 // Binaries built to run on older releases may need to use eh_frame 2446 // information. 2447 2448 if (text_section_sp && function_starts_data.GetByteSize()) { 2449 FunctionStarts::Entry function_start_entry; 2450 function_start_entry.data = false; 2451 lldb::offset_t function_start_offset = 0; 2452 function_start_entry.addr = text_section_sp->GetFileAddress(); 2453 uint64_t delta; 2454 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2455 0) { 2456 // Now append the current entry 2457 function_start_entry.addr += delta; 2458 function_starts.Append(function_start_entry); 2459 } 2460 } else { 2461 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2462 // load command claiming an eh_frame but it doesn't actually have the 2463 // eh_frame content. And if we have a dSYM, we don't need to do any of 2464 // this fill-in-the-missing-symbols works anyway - the debug info should 2465 // give us all the functions in the module. 2466 if (text_section_sp.get() && eh_frame_section_sp.get() && 2467 m_type != eTypeDebugInfo) { 2468 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2469 DWARFCallFrameInfo::EH); 2470 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2471 eh_frame.GetFunctionAddressAndSizeVector(functions); 2472 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2473 size_t count = functions.GetSize(); 2474 for (size_t i = 0; i < count; ++i) { 2475 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2476 functions.GetEntryAtIndex(i); 2477 if (func) { 2478 FunctionStarts::Entry function_start_entry; 2479 function_start_entry.addr = func->base - text_base_addr; 2480 function_starts.Append(function_start_entry); 2481 } 2482 } 2483 } 2484 } 2485 2486 const size_t function_starts_count = function_starts.GetSize(); 2487 2488 // For user process binaries (executables, dylibs, frameworks, bundles), if 2489 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2490 // going to assume the binary has been stripped. Don't allow assembly 2491 // language instruction emulation because we don't know proper function 2492 // start boundaries. 2493 // 2494 // For all other types of binaries (kernels, stand-alone bare board 2495 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2496 // sections - we should not make any assumptions about them based on that. 2497 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2498 m_allow_assembly_emulation_unwind_plans = false; 2499 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2500 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2501 2502 if (unwind_or_symbol_log) 2503 module_sp->LogMessage( 2504 unwind_or_symbol_log, 2505 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2506 } 2507 2508 const user_id_t TEXT_eh_frame_sectID = 2509 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2510 : static_cast<user_id_t>(NO_SECT); 2511 2512 lldb::offset_t nlist_data_offset = 0; 2513 2514 uint32_t N_SO_index = UINT32_MAX; 2515 2516 MachSymtabSectionInfo section_info(section_list); 2517 std::vector<uint32_t> N_FUN_indexes; 2518 std::vector<uint32_t> N_NSYM_indexes; 2519 std::vector<uint32_t> N_INCL_indexes; 2520 std::vector<uint32_t> N_BRAC_indexes; 2521 std::vector<uint32_t> N_COMM_indexes; 2522 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2523 typedef std::map<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2524 typedef std::map<const char *, uint32_t> ConstNameToSymbolIndexMap; 2525 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2526 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2527 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2528 // Any symbols that get merged into another will get an entry in this map 2529 // so we know 2530 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2531 uint32_t nlist_idx = 0; 2532 Symbol *symbol_ptr = NULL; 2533 2534 uint32_t sym_idx = 0; 2535 Symbol *sym = NULL; 2536 size_t num_syms = 0; 2537 std::string memory_symbol_name; 2538 uint32_t unmapped_local_symbols_found = 0; 2539 2540 std::vector<TrieEntryWithOffset> trie_entries; 2541 std::set<lldb::addr_t> resolver_addresses; 2542 2543 if (dyld_trie_data.GetByteSize() > 0) { 2544 std::vector<llvm::StringRef> nameSlices; 2545 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, 2546 resolver_addresses, trie_entries); 2547 2548 ConstString text_segment_name("__TEXT"); 2549 SectionSP text_segment_sp = 2550 GetSectionList()->FindSectionByName(text_segment_name); 2551 if (text_segment_sp) { 2552 const lldb::addr_t text_segment_file_addr = 2553 text_segment_sp->GetFileAddress(); 2554 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2555 for (auto &e : trie_entries) 2556 e.entry.address += text_segment_file_addr; 2557 } 2558 } 2559 } 2560 2561 typedef std::set<ConstString> IndirectSymbols; 2562 IndirectSymbols indirect_symbol_names; 2563 2564 #if defined(__APPLE__) && \ 2565 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2566 2567 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2568 // optimized by moving LOCAL symbols out of the memory mapped portion of 2569 // the DSC. The symbol information has all been retained, but it isn't 2570 // available in the normal nlist data. However, there *are* duplicate 2571 // entries of *some* 2572 // LOCAL symbols in the normal nlist data. To handle this situation 2573 // correctly, we must first attempt 2574 // to parse any DSC unmapped symbol information. If we find any, we set a 2575 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2576 2577 if (m_header.flags & 0x80000000u) { 2578 // Before we can start mapping the DSC, we need to make certain the 2579 // target process is actually using the cache we can find. 2580 2581 // Next we need to determine the correct path for the dyld shared cache. 2582 2583 ArchSpec header_arch; 2584 GetArchitecture(header_arch); 2585 char dsc_path[PATH_MAX]; 2586 char dsc_path_development[PATH_MAX]; 2587 2588 snprintf( 2589 dsc_path, sizeof(dsc_path), "%s%s%s", 2590 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2591 */ 2592 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2593 header_arch.GetArchitectureName()); 2594 2595 snprintf( 2596 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2597 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2598 */ 2599 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2600 header_arch.GetArchitectureName(), ".development"); 2601 2602 FileSpec dsc_nondevelopment_filespec(dsc_path, false); 2603 FileSpec dsc_development_filespec(dsc_path_development, false); 2604 FileSpec dsc_filespec; 2605 2606 UUID dsc_uuid; 2607 UUID process_shared_cache_uuid; 2608 2609 if (process) { 2610 process_shared_cache_uuid = GetProcessSharedCacheUUID(process); 2611 } 2612 2613 // First see if we can find an exact match for the inferior process 2614 // shared cache UUID in the development or non-development shared caches 2615 // on disk. 2616 if (process_shared_cache_uuid.IsValid()) { 2617 if (dsc_development_filespec.Exists()) { 2618 UUID dsc_development_uuid = GetSharedCacheUUID( 2619 dsc_development_filespec, byte_order, addr_byte_size); 2620 if (dsc_development_uuid.IsValid() && 2621 dsc_development_uuid == process_shared_cache_uuid) { 2622 dsc_filespec = dsc_development_filespec; 2623 dsc_uuid = dsc_development_uuid; 2624 } 2625 } 2626 if (!dsc_uuid.IsValid() && dsc_nondevelopment_filespec.Exists()) { 2627 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2628 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2629 if (dsc_nondevelopment_uuid.IsValid() && 2630 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2631 dsc_filespec = dsc_nondevelopment_filespec; 2632 dsc_uuid = dsc_nondevelopment_uuid; 2633 } 2634 } 2635 } 2636 2637 // Failing a UUID match, prefer the development dyld_shared cache if both 2638 // are present. 2639 if (!dsc_filespec.Exists()) { 2640 if (dsc_development_filespec.Exists()) { 2641 dsc_filespec = dsc_development_filespec; 2642 } else { 2643 dsc_filespec = dsc_nondevelopment_filespec; 2644 } 2645 } 2646 2647 /* The dyld_cache_header has a pointer to the 2648 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2649 The dyld_cache_local_symbols_info structure gives us three things: 2650 1. The start and count of the nlist records in the dyld_shared_cache 2651 file 2652 2. The start and size of the strings for these nlist records 2653 3. The start and count of dyld_cache_local_symbols_entry entries 2654 2655 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2656 dyld shared cache. 2657 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2658 the dyld shared cache. 2659 The dyld_cache_local_symbols_entry also lists the start of this 2660 dylib/framework's nlist records 2661 and the count of how many nlist records there are for this 2662 dylib/framework. 2663 */ 2664 2665 // Process the dyld shared cache header to find the unmapped symbols 2666 2667 DataBufferSP dsc_data_sp = MapFileData( 2668 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2669 if (!dsc_uuid.IsValid()) { 2670 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2671 } 2672 if (dsc_data_sp) { 2673 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2674 2675 bool uuid_match = true; 2676 if (dsc_uuid.IsValid() && process) { 2677 if (process_shared_cache_uuid.IsValid() && 2678 dsc_uuid != process_shared_cache_uuid) { 2679 // The on-disk dyld_shared_cache file is not the same as the one in 2680 // this process' memory, don't use it. 2681 uuid_match = false; 2682 ModuleSP module_sp(GetModule()); 2683 if (module_sp) 2684 module_sp->ReportWarning("process shared cache does not match " 2685 "on-disk dyld_shared_cache file, some " 2686 "symbol names will be missing."); 2687 } 2688 } 2689 2690 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2691 2692 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2693 2694 // If the mappingOffset points to a location inside the header, we've 2695 // opened an old dyld shared cache, and should not proceed further. 2696 if (uuid_match && 2697 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2698 2699 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2700 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2701 mappingOffset); 2702 2703 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2704 byte_order, addr_byte_size); 2705 offset = 0; 2706 2707 // The File addresses (from the in-memory Mach-O load commands) for 2708 // the shared libraries in the shared library cache need to be 2709 // adjusted by an offset to match up with the dylibOffset identifying 2710 // field in the dyld_cache_local_symbol_entry's. This offset is 2711 // recorded in mapping_offset_value. 2712 const uint64_t mapping_offset_value = 2713 dsc_mapping_info_data.GetU64(&offset); 2714 2715 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, 2716 localSymbolsOffset); 2717 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2718 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2719 2720 if (localSymbolsOffset && localSymbolsSize) { 2721 // Map the local symbols 2722 DataBufferSP dsc_local_symbols_data_sp = 2723 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2724 2725 if (dsc_local_symbols_data_sp) { 2726 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2727 byte_order, addr_byte_size); 2728 2729 offset = 0; 2730 2731 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2732 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2733 UndefinedNameToDescMap undefined_name_to_desc; 2734 SymbolIndexToName reexport_shlib_needs_fixup; 2735 2736 // Read the local_symbols_infos struct in one shot 2737 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2738 dsc_local_symbols_data.GetU32(&offset, 2739 &local_symbols_info.nlistOffset, 6); 2740 2741 SectionSP text_section_sp( 2742 section_list->FindSectionByName(GetSegmentNameTEXT())); 2743 2744 uint32_t header_file_offset = 2745 (text_section_sp->GetFileAddress() - mapping_offset_value); 2746 2747 offset = local_symbols_info.entriesOffset; 2748 for (uint32_t entry_index = 0; 2749 entry_index < local_symbols_info.entriesCount; 2750 entry_index++) { 2751 struct lldb_copy_dyld_cache_local_symbols_entry 2752 local_symbols_entry; 2753 local_symbols_entry.dylibOffset = 2754 dsc_local_symbols_data.GetU32(&offset); 2755 local_symbols_entry.nlistStartIndex = 2756 dsc_local_symbols_data.GetU32(&offset); 2757 local_symbols_entry.nlistCount = 2758 dsc_local_symbols_data.GetU32(&offset); 2759 2760 if (header_file_offset == local_symbols_entry.dylibOffset) { 2761 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2762 2763 // The normal nlist code cannot correctly size the Symbols 2764 // array, we need to allocate it here. 2765 sym = symtab->Resize( 2766 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2767 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2768 num_syms = symtab->GetNumSymbols(); 2769 2770 nlist_data_offset = 2771 local_symbols_info.nlistOffset + 2772 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2773 uint32_t string_table_offset = 2774 local_symbols_info.stringsOffset; 2775 2776 for (uint32_t nlist_index = 0; 2777 nlist_index < local_symbols_entry.nlistCount; 2778 nlist_index++) { 2779 ///////////////////////////// 2780 { 2781 struct nlist_64 nlist; 2782 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize( 2783 nlist_data_offset, nlist_byte_size)) 2784 break; 2785 2786 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked( 2787 &nlist_data_offset); 2788 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked( 2789 &nlist_data_offset); 2790 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked( 2791 &nlist_data_offset); 2792 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked( 2793 &nlist_data_offset); 2794 nlist.n_value = 2795 dsc_local_symbols_data.GetAddress_unchecked( 2796 &nlist_data_offset); 2797 2798 SymbolType type = eSymbolTypeInvalid; 2799 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2800 string_table_offset + nlist.n_strx); 2801 2802 if (symbol_name == NULL) { 2803 // No symbol should be NULL, even the symbols with no 2804 // string values should have an offset zero which 2805 // points to an empty C-string 2806 Host::SystemLog( 2807 Host::eSystemLogError, 2808 "error: DSC unmapped local symbol[%u] has invalid " 2809 "string table offset 0x%x in %s, ignoring symbol\n", 2810 entry_index, nlist.n_strx, 2811 module_sp->GetFileSpec().GetPath().c_str()); 2812 continue; 2813 } 2814 if (symbol_name[0] == '\0') 2815 symbol_name = NULL; 2816 2817 const char *symbol_name_non_abi_mangled = NULL; 2818 2819 SectionSP symbol_section; 2820 uint32_t symbol_byte_size = 0; 2821 bool add_nlist = true; 2822 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2823 bool demangled_is_synthesized = false; 2824 bool is_gsym = false; 2825 bool set_value = true; 2826 2827 assert(sym_idx < num_syms); 2828 2829 sym[sym_idx].SetDebug(is_debug); 2830 2831 if (is_debug) { 2832 switch (nlist.n_type) { 2833 case N_GSYM: 2834 // global symbol: name,,NO_SECT,type,0 2835 // Sometimes the N_GSYM value contains the address. 2836 2837 // FIXME: In the .o files, we have a GSYM and a debug 2838 // symbol for all the ObjC data. They 2839 // have the same address, but we want to ensure that 2840 // we always find only the real symbol, 'cause we 2841 // don't currently correctly attribute the 2842 // GSYM one to the ObjCClass/Ivar/MetaClass 2843 // symbol type. This is a temporary hack to make 2844 // sure the ObjectiveC symbols get treated correctly. 2845 // To do this right, we should coalesce all the GSYM 2846 // & global symbols that have the same address. 2847 2848 is_gsym = true; 2849 sym[sym_idx].SetExternal(true); 2850 2851 if (symbol_name && symbol_name[0] == '_' && 2852 symbol_name[1] == 'O') { 2853 llvm::StringRef symbol_name_ref(symbol_name); 2854 if (symbol_name_ref.startswith( 2855 g_objc_v2_prefix_class)) { 2856 symbol_name_non_abi_mangled = symbol_name + 1; 2857 symbol_name = 2858 symbol_name + g_objc_v2_prefix_class.size(); 2859 type = eSymbolTypeObjCClass; 2860 demangled_is_synthesized = true; 2861 2862 } else if (symbol_name_ref.startswith( 2863 g_objc_v2_prefix_metaclass)) { 2864 symbol_name_non_abi_mangled = symbol_name + 1; 2865 symbol_name = symbol_name + 2866 g_objc_v2_prefix_metaclass.size(); 2867 type = eSymbolTypeObjCMetaClass; 2868 demangled_is_synthesized = true; 2869 } else if (symbol_name_ref.startswith( 2870 g_objc_v2_prefix_ivar)) { 2871 symbol_name_non_abi_mangled = symbol_name + 1; 2872 symbol_name = 2873 symbol_name + g_objc_v2_prefix_ivar.size(); 2874 type = eSymbolTypeObjCIVar; 2875 demangled_is_synthesized = true; 2876 } 2877 } else { 2878 if (nlist.n_value != 0) 2879 symbol_section = section_info.GetSection( 2880 nlist.n_sect, nlist.n_value); 2881 type = eSymbolTypeData; 2882 } 2883 break; 2884 2885 case N_FNAME: 2886 // procedure name (f77 kludge): name,,NO_SECT,0,0 2887 type = eSymbolTypeCompiler; 2888 break; 2889 2890 case N_FUN: 2891 // procedure: name,,n_sect,linenumber,address 2892 if (symbol_name) { 2893 type = eSymbolTypeCode; 2894 symbol_section = section_info.GetSection( 2895 nlist.n_sect, nlist.n_value); 2896 2897 N_FUN_addr_to_sym_idx.insert( 2898 std::make_pair(nlist.n_value, sym_idx)); 2899 // We use the current number of symbols in the 2900 // symbol table in lieu of using nlist_idx in case 2901 // we ever start trimming entries out 2902 N_FUN_indexes.push_back(sym_idx); 2903 } else { 2904 type = eSymbolTypeCompiler; 2905 2906 if (!N_FUN_indexes.empty()) { 2907 // Copy the size of the function into the 2908 // original 2909 // STAB entry so we don't have 2910 // to hunt for it later 2911 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2912 ->SetByteSize(nlist.n_value); 2913 N_FUN_indexes.pop_back(); 2914 // We don't really need the end function STAB as 2915 // it contains the size which we already placed 2916 // with the original symbol, so don't add it if 2917 // we want a minimal symbol table 2918 add_nlist = false; 2919 } 2920 } 2921 break; 2922 2923 case N_STSYM: 2924 // static symbol: name,,n_sect,type,address 2925 N_STSYM_addr_to_sym_idx.insert( 2926 std::make_pair(nlist.n_value, sym_idx)); 2927 symbol_section = section_info.GetSection( 2928 nlist.n_sect, nlist.n_value); 2929 if (symbol_name && symbol_name[0]) { 2930 type = ObjectFile::GetSymbolTypeFromName( 2931 symbol_name + 1, eSymbolTypeData); 2932 } 2933 break; 2934 2935 case N_LCSYM: 2936 // .lcomm symbol: name,,n_sect,type,address 2937 symbol_section = section_info.GetSection( 2938 nlist.n_sect, nlist.n_value); 2939 type = eSymbolTypeCommonBlock; 2940 break; 2941 2942 case N_BNSYM: 2943 // We use the current number of symbols in the symbol 2944 // table in lieu of using nlist_idx in case we ever 2945 // start trimming entries out Skip these if we want 2946 // minimal symbol tables 2947 add_nlist = false; 2948 break; 2949 2950 case N_ENSYM: 2951 // Set the size of the N_BNSYM to the terminating 2952 // index of this N_ENSYM so that we can always skip 2953 // the entire symbol if we need to navigate more 2954 // quickly at the source level when parsing STABS 2955 // Skip these if we want minimal symbol tables 2956 add_nlist = false; 2957 break; 2958 2959 case N_OPT: 2960 // emitted with gcc2_compiled and in gcc source 2961 type = eSymbolTypeCompiler; 2962 break; 2963 2964 case N_RSYM: 2965 // register sym: name,,NO_SECT,type,register 2966 type = eSymbolTypeVariable; 2967 break; 2968 2969 case N_SLINE: 2970 // src line: 0,,n_sect,linenumber,address 2971 symbol_section = section_info.GetSection( 2972 nlist.n_sect, nlist.n_value); 2973 type = eSymbolTypeLineEntry; 2974 break; 2975 2976 case N_SSYM: 2977 // structure elt: name,,NO_SECT,type,struct_offset 2978 type = eSymbolTypeVariableType; 2979 break; 2980 2981 case N_SO: 2982 // source file name 2983 type = eSymbolTypeSourceFile; 2984 if (symbol_name == NULL) { 2985 add_nlist = false; 2986 if (N_SO_index != UINT32_MAX) { 2987 // Set the size of the N_SO to the terminating 2988 // index of this N_SO so that we can always skip 2989 // the entire N_SO if we need to navigate more 2990 // quickly at the source level when parsing STABS 2991 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2992 symbol_ptr->SetByteSize(sym_idx); 2993 symbol_ptr->SetSizeIsSibling(true); 2994 } 2995 N_NSYM_indexes.clear(); 2996 N_INCL_indexes.clear(); 2997 N_BRAC_indexes.clear(); 2998 N_COMM_indexes.clear(); 2999 N_FUN_indexes.clear(); 3000 N_SO_index = UINT32_MAX; 3001 } else { 3002 // We use the current number of symbols in the 3003 // symbol table in lieu of using nlist_idx in case 3004 // we ever start trimming entries out 3005 const bool N_SO_has_full_path = 3006 symbol_name[0] == '/'; 3007 if (N_SO_has_full_path) { 3008 if ((N_SO_index == sym_idx - 1) && 3009 ((sym_idx - 1) < num_syms)) { 3010 // We have two consecutive N_SO entries where 3011 // the first contains a directory and the 3012 // second contains a full path. 3013 sym[sym_idx - 1].GetMangled().SetValue( 3014 ConstString(symbol_name), false); 3015 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3016 add_nlist = false; 3017 } else { 3018 // This is the first entry in a N_SO that 3019 // contains a directory or 3020 // a full path to the source file 3021 N_SO_index = sym_idx; 3022 } 3023 } else if ((N_SO_index == sym_idx - 1) && 3024 ((sym_idx - 1) < num_syms)) { 3025 // This is usually the second N_SO entry that 3026 // contains just the filename, so here we combine 3027 // it with the first one if we are minimizing the 3028 // symbol table 3029 const char *so_path = 3030 sym[sym_idx - 1] 3031 .GetMangled() 3032 .GetDemangledName( 3033 lldb::eLanguageTypeUnknown) 3034 .AsCString(); 3035 if (so_path && so_path[0]) { 3036 std::string full_so_path(so_path); 3037 const size_t double_slash_pos = 3038 full_so_path.find("//"); 3039 if (double_slash_pos != std::string::npos) { 3040 // The linker has been generating bad N_SO 3041 // entries with doubled up paths 3042 // in the format "%s%s" where the first 3043 // string in the DW_AT_comp_dir, and the 3044 // second is the directory for the source 3045 // file so you end up with a path that looks 3046 // like "/tmp/src//tmp/src/" 3047 FileSpec so_dir(so_path, false); 3048 if (!so_dir.Exists()) { 3049 so_dir.SetFile( 3050 &full_so_path[double_slash_pos + 1], 3051 false); 3052 if (so_dir.Exists()) { 3053 // Trim off the incorrect path 3054 full_so_path.erase(0, 3055 double_slash_pos + 1); 3056 } 3057 } 3058 } 3059 if (*full_so_path.rbegin() != '/') 3060 full_so_path += '/'; 3061 full_so_path += symbol_name; 3062 sym[sym_idx - 1].GetMangled().SetValue( 3063 ConstString(full_so_path.c_str()), false); 3064 add_nlist = false; 3065 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3066 } 3067 } else { 3068 // This could be a relative path to a N_SO 3069 N_SO_index = sym_idx; 3070 } 3071 } 3072 break; 3073 3074 case N_OSO: 3075 // object file name: name,,0,0,st_mtime 3076 type = eSymbolTypeObjectFile; 3077 break; 3078 3079 case N_LSYM: 3080 // local sym: name,,NO_SECT,type,offset 3081 type = eSymbolTypeLocal; 3082 break; 3083 3084 //---------------------------------------------------------------------- 3085 // INCL scopes 3086 //---------------------------------------------------------------------- 3087 case N_BINCL: 3088 // include file beginning: name,,NO_SECT,0,sum We use 3089 // the current number of symbols in the symbol table 3090 // in lieu of using nlist_idx in case we ever start 3091 // trimming entries out 3092 N_INCL_indexes.push_back(sym_idx); 3093 type = eSymbolTypeScopeBegin; 3094 break; 3095 3096 case N_EINCL: 3097 // include file end: name,,NO_SECT,0,0 3098 // Set the size of the N_BINCL to the terminating 3099 // index of this N_EINCL so that we can always skip 3100 // the entire symbol if we need to navigate more 3101 // quickly at the source level when parsing STABS 3102 if (!N_INCL_indexes.empty()) { 3103 symbol_ptr = 3104 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3105 symbol_ptr->SetByteSize(sym_idx + 1); 3106 symbol_ptr->SetSizeIsSibling(true); 3107 N_INCL_indexes.pop_back(); 3108 } 3109 type = eSymbolTypeScopeEnd; 3110 break; 3111 3112 case N_SOL: 3113 // #included file name: name,,n_sect,0,address 3114 type = eSymbolTypeHeaderFile; 3115 3116 // We currently don't use the header files on darwin 3117 add_nlist = false; 3118 break; 3119 3120 case N_PARAMS: 3121 // compiler parameters: name,,NO_SECT,0,0 3122 type = eSymbolTypeCompiler; 3123 break; 3124 3125 case N_VERSION: 3126 // compiler version: name,,NO_SECT,0,0 3127 type = eSymbolTypeCompiler; 3128 break; 3129 3130 case N_OLEVEL: 3131 // compiler -O level: name,,NO_SECT,0,0 3132 type = eSymbolTypeCompiler; 3133 break; 3134 3135 case N_PSYM: 3136 // parameter: name,,NO_SECT,type,offset 3137 type = eSymbolTypeVariable; 3138 break; 3139 3140 case N_ENTRY: 3141 // alternate entry: name,,n_sect,linenumber,address 3142 symbol_section = section_info.GetSection( 3143 nlist.n_sect, nlist.n_value); 3144 type = eSymbolTypeLineEntry; 3145 break; 3146 3147 //---------------------------------------------------------------------- 3148 // Left and Right Braces 3149 //---------------------------------------------------------------------- 3150 case N_LBRAC: 3151 // left bracket: 0,,NO_SECT,nesting level,address We 3152 // use the current number of symbols in the symbol 3153 // table in lieu of using nlist_idx in case we ever 3154 // start trimming entries out 3155 symbol_section = section_info.GetSection( 3156 nlist.n_sect, nlist.n_value); 3157 N_BRAC_indexes.push_back(sym_idx); 3158 type = eSymbolTypeScopeBegin; 3159 break; 3160 3161 case N_RBRAC: 3162 // right bracket: 0,,NO_SECT,nesting level,address 3163 // Set the size of the N_LBRAC to the terminating 3164 // index of this N_RBRAC so that we can always skip 3165 // the entire symbol if we need to navigate more 3166 // quickly at the source level when parsing STABS 3167 symbol_section = section_info.GetSection( 3168 nlist.n_sect, nlist.n_value); 3169 if (!N_BRAC_indexes.empty()) { 3170 symbol_ptr = 3171 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3172 symbol_ptr->SetByteSize(sym_idx + 1); 3173 symbol_ptr->SetSizeIsSibling(true); 3174 N_BRAC_indexes.pop_back(); 3175 } 3176 type = eSymbolTypeScopeEnd; 3177 break; 3178 3179 case N_EXCL: 3180 // deleted include file: name,,NO_SECT,0,sum 3181 type = eSymbolTypeHeaderFile; 3182 break; 3183 3184 //---------------------------------------------------------------------- 3185 // COMM scopes 3186 //---------------------------------------------------------------------- 3187 case N_BCOMM: 3188 // begin common: name,,NO_SECT,0,0 3189 // We use the current number of symbols in the symbol 3190 // table in lieu of using nlist_idx in case we ever 3191 // start trimming entries out 3192 type = eSymbolTypeScopeBegin; 3193 N_COMM_indexes.push_back(sym_idx); 3194 break; 3195 3196 case N_ECOML: 3197 // end common (local name): 0,,n_sect,0,address 3198 symbol_section = section_info.GetSection( 3199 nlist.n_sect, nlist.n_value); 3200 // Fall through 3201 3202 case N_ECOMM: 3203 // end common: name,,n_sect,0,0 3204 // Set the size of the N_BCOMM to the terminating 3205 // index of this N_ECOMM/N_ECOML so that we can 3206 // always skip the entire symbol if we need to 3207 // navigate more quickly at the source level when 3208 // parsing STABS 3209 if (!N_COMM_indexes.empty()) { 3210 symbol_ptr = 3211 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3212 symbol_ptr->SetByteSize(sym_idx + 1); 3213 symbol_ptr->SetSizeIsSibling(true); 3214 N_COMM_indexes.pop_back(); 3215 } 3216 type = eSymbolTypeScopeEnd; 3217 break; 3218 3219 case N_LENG: 3220 // second stab entry with length information 3221 type = eSymbolTypeAdditional; 3222 break; 3223 3224 default: 3225 break; 3226 } 3227 } else { 3228 // uint8_t n_pext = N_PEXT & nlist.n_type; 3229 uint8_t n_type = N_TYPE & nlist.n_type; 3230 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3231 3232 switch (n_type) { 3233 case N_INDR: { 3234 const char *reexport_name_cstr = 3235 strtab_data.PeekCStr(nlist.n_value); 3236 if (reexport_name_cstr && reexport_name_cstr[0]) { 3237 type = eSymbolTypeReExported; 3238 ConstString reexport_name( 3239 reexport_name_cstr + 3240 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3241 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3242 set_value = false; 3243 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3244 indirect_symbol_names.insert( 3245 ConstString(symbol_name + 3246 ((symbol_name[0] == '_') ? 1 : 0))); 3247 } else 3248 type = eSymbolTypeUndefined; 3249 } break; 3250 3251 case N_UNDF: 3252 if (symbol_name && symbol_name[0]) { 3253 ConstString undefined_name( 3254 symbol_name + 3255 ((symbol_name[0] == '_') ? 1 : 0)); 3256 undefined_name_to_desc[undefined_name] = 3257 nlist.n_desc; 3258 } 3259 // Fall through 3260 case N_PBUD: 3261 type = eSymbolTypeUndefined; 3262 break; 3263 3264 case N_ABS: 3265 type = eSymbolTypeAbsolute; 3266 break; 3267 3268 case N_SECT: { 3269 symbol_section = section_info.GetSection( 3270 nlist.n_sect, nlist.n_value); 3271 3272 if (symbol_section == NULL) { 3273 // TODO: warn about this? 3274 add_nlist = false; 3275 break; 3276 } 3277 3278 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3279 type = eSymbolTypeException; 3280 } else { 3281 uint32_t section_type = 3282 symbol_section->Get() & SECTION_TYPE; 3283 3284 switch (section_type) { 3285 case S_CSTRING_LITERALS: 3286 type = eSymbolTypeData; 3287 break; // section with only literal C strings 3288 case S_4BYTE_LITERALS: 3289 type = eSymbolTypeData; 3290 break; // section with only 4 byte literals 3291 case S_8BYTE_LITERALS: 3292 type = eSymbolTypeData; 3293 break; // section with only 8 byte literals 3294 case S_LITERAL_POINTERS: 3295 type = eSymbolTypeTrampoline; 3296 break; // section with only pointers to literals 3297 case S_NON_LAZY_SYMBOL_POINTERS: 3298 type = eSymbolTypeTrampoline; 3299 break; // section with only non-lazy symbol 3300 // pointers 3301 case S_LAZY_SYMBOL_POINTERS: 3302 type = eSymbolTypeTrampoline; 3303 break; // section with only lazy symbol pointers 3304 case S_SYMBOL_STUBS: 3305 type = eSymbolTypeTrampoline; 3306 break; // section with only symbol stubs, byte 3307 // size of stub in the reserved2 field 3308 case S_MOD_INIT_FUNC_POINTERS: 3309 type = eSymbolTypeCode; 3310 break; // section with only function pointers for 3311 // initialization 3312 case S_MOD_TERM_FUNC_POINTERS: 3313 type = eSymbolTypeCode; 3314 break; // section with only function pointers for 3315 // termination 3316 case S_INTERPOSING: 3317 type = eSymbolTypeTrampoline; 3318 break; // section with only pairs of function 3319 // pointers for interposing 3320 case S_16BYTE_LITERALS: 3321 type = eSymbolTypeData; 3322 break; // section with only 16 byte literals 3323 case S_DTRACE_DOF: 3324 type = eSymbolTypeInstrumentation; 3325 break; 3326 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3327 type = eSymbolTypeTrampoline; 3328 break; 3329 default: 3330 switch (symbol_section->GetType()) { 3331 case lldb::eSectionTypeCode: 3332 type = eSymbolTypeCode; 3333 break; 3334 case eSectionTypeData: 3335 case eSectionTypeDataCString: // Inlined C string 3336 // data 3337 case eSectionTypeDataCStringPointers: // Pointers 3338 // to C 3339 // string 3340 // data 3341 case eSectionTypeDataSymbolAddress: // Address of 3342 // a symbol in 3343 // the symbol 3344 // table 3345 case eSectionTypeData4: 3346 case eSectionTypeData8: 3347 case eSectionTypeData16: 3348 type = eSymbolTypeData; 3349 break; 3350 default: 3351 break; 3352 } 3353 break; 3354 } 3355 3356 if (type == eSymbolTypeInvalid) { 3357 const char *symbol_sect_name = 3358 symbol_section->GetName().AsCString(); 3359 if (symbol_section->IsDescendant( 3360 text_section_sp.get())) { 3361 if (symbol_section->IsClear( 3362 S_ATTR_PURE_INSTRUCTIONS | 3363 S_ATTR_SELF_MODIFYING_CODE | 3364 S_ATTR_SOME_INSTRUCTIONS)) 3365 type = eSymbolTypeData; 3366 else 3367 type = eSymbolTypeCode; 3368 } else if (symbol_section->IsDescendant( 3369 data_section_sp.get()) || 3370 symbol_section->IsDescendant( 3371 data_dirty_section_sp.get()) || 3372 symbol_section->IsDescendant( 3373 data_const_section_sp.get())) { 3374 if (symbol_sect_name && 3375 ::strstr(symbol_sect_name, "__objc") == 3376 symbol_sect_name) { 3377 type = eSymbolTypeRuntime; 3378 3379 if (symbol_name) { 3380 llvm::StringRef symbol_name_ref( 3381 symbol_name); 3382 if (symbol_name_ref.startswith("_OBJC_")) { 3383 static const llvm::StringRef 3384 g_objc_v2_prefix_class( 3385 "_OBJC_CLASS_$_"); 3386 static const llvm::StringRef 3387 g_objc_v2_prefix_metaclass( 3388 "_OBJC_METACLASS_$_"); 3389 static const llvm::StringRef 3390 g_objc_v2_prefix_ivar( 3391 "_OBJC_IVAR_$_"); 3392 if (symbol_name_ref.startswith( 3393 g_objc_v2_prefix_class)) { 3394 symbol_name_non_abi_mangled = 3395 symbol_name + 1; 3396 symbol_name = 3397 symbol_name + 3398 g_objc_v2_prefix_class.size(); 3399 type = eSymbolTypeObjCClass; 3400 demangled_is_synthesized = true; 3401 } else if ( 3402 symbol_name_ref.startswith( 3403 g_objc_v2_prefix_metaclass)) { 3404 symbol_name_non_abi_mangled = 3405 symbol_name + 1; 3406 symbol_name = 3407 symbol_name + 3408 g_objc_v2_prefix_metaclass.size(); 3409 type = eSymbolTypeObjCMetaClass; 3410 demangled_is_synthesized = true; 3411 } else if (symbol_name_ref.startswith( 3412 g_objc_v2_prefix_ivar)) { 3413 symbol_name_non_abi_mangled = 3414 symbol_name + 1; 3415 symbol_name = 3416 symbol_name + 3417 g_objc_v2_prefix_ivar.size(); 3418 type = eSymbolTypeObjCIVar; 3419 demangled_is_synthesized = true; 3420 } 3421 } 3422 } 3423 } else if (symbol_sect_name && 3424 ::strstr(symbol_sect_name, 3425 "__gcc_except_tab") == 3426 symbol_sect_name) { 3427 type = eSymbolTypeException; 3428 } else { 3429 type = eSymbolTypeData; 3430 } 3431 } else if (symbol_sect_name && 3432 ::strstr(symbol_sect_name, 3433 "__IMPORT") == 3434 symbol_sect_name) { 3435 type = eSymbolTypeTrampoline; 3436 } else if (symbol_section->IsDescendant( 3437 objc_section_sp.get())) { 3438 type = eSymbolTypeRuntime; 3439 if (symbol_name && symbol_name[0] == '.') { 3440 llvm::StringRef symbol_name_ref(symbol_name); 3441 static const llvm::StringRef 3442 g_objc_v1_prefix_class( 3443 ".objc_class_name_"); 3444 if (symbol_name_ref.startswith( 3445 g_objc_v1_prefix_class)) { 3446 symbol_name_non_abi_mangled = symbol_name; 3447 symbol_name = symbol_name + 3448 g_objc_v1_prefix_class.size(); 3449 type = eSymbolTypeObjCClass; 3450 demangled_is_synthesized = true; 3451 } 3452 } 3453 } 3454 } 3455 } 3456 } break; 3457 } 3458 } 3459 3460 if (add_nlist) { 3461 uint64_t symbol_value = nlist.n_value; 3462 if (symbol_name_non_abi_mangled) { 3463 sym[sym_idx].GetMangled().SetMangledName( 3464 ConstString(symbol_name_non_abi_mangled)); 3465 sym[sym_idx].GetMangled().SetDemangledName( 3466 ConstString(symbol_name)); 3467 } else { 3468 bool symbol_name_is_mangled = false; 3469 3470 if (symbol_name && symbol_name[0] == '_') { 3471 symbol_name_is_mangled = symbol_name[1] == '_'; 3472 symbol_name++; // Skip the leading underscore 3473 } 3474 3475 if (symbol_name) { 3476 ConstString const_symbol_name(symbol_name); 3477 sym[sym_idx].GetMangled().SetValue( 3478 const_symbol_name, symbol_name_is_mangled); 3479 if (is_gsym && is_debug) { 3480 const char *gsym_name = 3481 sym[sym_idx] 3482 .GetMangled() 3483 .GetName(lldb::eLanguageTypeUnknown, 3484 Mangled::ePreferMangled) 3485 .GetCString(); 3486 if (gsym_name) 3487 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3488 } 3489 } 3490 } 3491 if (symbol_section) { 3492 const addr_t section_file_addr = 3493 symbol_section->GetFileAddress(); 3494 if (symbol_byte_size == 0 && 3495 function_starts_count > 0) { 3496 addr_t symbol_lookup_file_addr = nlist.n_value; 3497 // Do an exact address match for non-ARM addresses, 3498 // else get the closest since the symbol might be a 3499 // thumb symbol which has an address with bit zero 3500 // set 3501 FunctionStarts::Entry *func_start_entry = 3502 function_starts.FindEntry( 3503 symbol_lookup_file_addr, !is_arm); 3504 if (is_arm && func_start_entry) { 3505 // Verify that the function start address is the 3506 // symbol address (ARM) or the symbol address + 1 3507 // (thumb) 3508 if (func_start_entry->addr != 3509 symbol_lookup_file_addr && 3510 func_start_entry->addr != 3511 (symbol_lookup_file_addr + 1)) { 3512 // Not the right entry, NULL it out... 3513 func_start_entry = NULL; 3514 } 3515 } 3516 if (func_start_entry) { 3517 func_start_entry->data = true; 3518 3519 addr_t symbol_file_addr = func_start_entry->addr; 3520 uint32_t symbol_flags = 0; 3521 if (is_arm) { 3522 if (symbol_file_addr & 1) 3523 symbol_flags = 3524 MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3525 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3526 } 3527 3528 const FunctionStarts::Entry 3529 *next_func_start_entry = 3530 function_starts.FindNextEntry( 3531 func_start_entry); 3532 const addr_t section_end_file_addr = 3533 section_file_addr + 3534 symbol_section->GetByteSize(); 3535 if (next_func_start_entry) { 3536 addr_t next_symbol_file_addr = 3537 next_func_start_entry->addr; 3538 // Be sure the clear the Thumb address bit when 3539 // we calculate the size from the current and 3540 // next address 3541 if (is_arm) 3542 next_symbol_file_addr &= 3543 THUMB_ADDRESS_BIT_MASK; 3544 symbol_byte_size = std::min<lldb::addr_t>( 3545 next_symbol_file_addr - symbol_file_addr, 3546 section_end_file_addr - symbol_file_addr); 3547 } else { 3548 symbol_byte_size = 3549 section_end_file_addr - symbol_file_addr; 3550 } 3551 } 3552 } 3553 symbol_value -= section_file_addr; 3554 } 3555 3556 if (is_debug == false) { 3557 if (type == eSymbolTypeCode) { 3558 // See if we can find a N_FUN entry for any code 3559 // symbols. If we do find a match, and the name 3560 // matches, then we can merge the two into just the 3561 // function symbol to avoid duplicate entries in 3562 // the symbol table 3563 std::pair<ValueToSymbolIndexMap::const_iterator, 3564 ValueToSymbolIndexMap::const_iterator> 3565 range; 3566 range = N_FUN_addr_to_sym_idx.equal_range( 3567 nlist.n_value); 3568 if (range.first != range.second) { 3569 bool found_it = false; 3570 for (ValueToSymbolIndexMap::const_iterator pos = 3571 range.first; 3572 pos != range.second; ++pos) { 3573 if (sym[sym_idx].GetMangled().GetName( 3574 lldb::eLanguageTypeUnknown, 3575 Mangled::ePreferMangled) == 3576 sym[pos->second].GetMangled().GetName( 3577 lldb::eLanguageTypeUnknown, 3578 Mangled::ePreferMangled)) { 3579 m_nlist_idx_to_sym_idx[nlist_idx] = 3580 pos->second; 3581 // We just need the flags from the linker 3582 // symbol, so put these flags 3583 // into the N_FUN flags to avoid duplicate 3584 // symbols in the symbol table 3585 sym[pos->second].SetExternal( 3586 sym[sym_idx].IsExternal()); 3587 sym[pos->second].SetFlags(nlist.n_type << 16 | 3588 nlist.n_desc); 3589 if (resolver_addresses.find(nlist.n_value) != 3590 resolver_addresses.end()) 3591 sym[pos->second].SetType( 3592 eSymbolTypeResolver); 3593 sym[sym_idx].Clear(); 3594 found_it = true; 3595 break; 3596 } 3597 } 3598 if (found_it) 3599 continue; 3600 } else { 3601 if (resolver_addresses.find(nlist.n_value) != 3602 resolver_addresses.end()) 3603 type = eSymbolTypeResolver; 3604 } 3605 } else if (type == eSymbolTypeData || 3606 type == eSymbolTypeObjCClass || 3607 type == eSymbolTypeObjCMetaClass || 3608 type == eSymbolTypeObjCIVar) { 3609 // See if we can find a N_STSYM entry for any data 3610 // symbols. If we do find a match, and the name 3611 // matches, then we can merge the two into just the 3612 // Static symbol to avoid duplicate entries in the 3613 // symbol table 3614 std::pair<ValueToSymbolIndexMap::const_iterator, 3615 ValueToSymbolIndexMap::const_iterator> 3616 range; 3617 range = N_STSYM_addr_to_sym_idx.equal_range( 3618 nlist.n_value); 3619 if (range.first != range.second) { 3620 bool found_it = false; 3621 for (ValueToSymbolIndexMap::const_iterator pos = 3622 range.first; 3623 pos != range.second; ++pos) { 3624 if (sym[sym_idx].GetMangled().GetName( 3625 lldb::eLanguageTypeUnknown, 3626 Mangled::ePreferMangled) == 3627 sym[pos->second].GetMangled().GetName( 3628 lldb::eLanguageTypeUnknown, 3629 Mangled::ePreferMangled)) { 3630 m_nlist_idx_to_sym_idx[nlist_idx] = 3631 pos->second; 3632 // We just need the flags from the linker 3633 // symbol, so put these flags 3634 // into the N_STSYM flags to avoid duplicate 3635 // symbols in the symbol table 3636 sym[pos->second].SetExternal( 3637 sym[sym_idx].IsExternal()); 3638 sym[pos->second].SetFlags(nlist.n_type << 16 | 3639 nlist.n_desc); 3640 sym[sym_idx].Clear(); 3641 found_it = true; 3642 break; 3643 } 3644 } 3645 if (found_it) 3646 continue; 3647 } else { 3648 const char *gsym_name = 3649 sym[sym_idx] 3650 .GetMangled() 3651 .GetName(lldb::eLanguageTypeUnknown, 3652 Mangled::ePreferMangled) 3653 .GetCString(); 3654 if (gsym_name) { 3655 // Combine N_GSYM stab entries with the non 3656 // stab symbol 3657 ConstNameToSymbolIndexMap::const_iterator pos = 3658 N_GSYM_name_to_sym_idx.find(gsym_name); 3659 if (pos != N_GSYM_name_to_sym_idx.end()) { 3660 const uint32_t GSYM_sym_idx = pos->second; 3661 m_nlist_idx_to_sym_idx[nlist_idx] = 3662 GSYM_sym_idx; 3663 // Copy the address, because often the N_GSYM 3664 // address has an invalid address of zero 3665 // when the global is a common symbol 3666 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3667 symbol_section); 3668 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3669 symbol_value); 3670 // We just need the flags from the linker 3671 // symbol, so put these flags 3672 // into the N_GSYM flags to avoid duplicate 3673 // symbols in the symbol table 3674 sym[GSYM_sym_idx].SetFlags( 3675 nlist.n_type << 16 | nlist.n_desc); 3676 sym[sym_idx].Clear(); 3677 continue; 3678 } 3679 } 3680 } 3681 } 3682 } 3683 3684 sym[sym_idx].SetID(nlist_idx); 3685 sym[sym_idx].SetType(type); 3686 if (set_value) { 3687 sym[sym_idx].GetAddressRef().SetSection( 3688 symbol_section); 3689 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3690 } 3691 sym[sym_idx].SetFlags(nlist.n_type << 16 | 3692 nlist.n_desc); 3693 3694 if (symbol_byte_size > 0) 3695 sym[sym_idx].SetByteSize(symbol_byte_size); 3696 3697 if (demangled_is_synthesized) 3698 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3699 ++sym_idx; 3700 } else { 3701 sym[sym_idx].Clear(); 3702 } 3703 } 3704 ///////////////////////////// 3705 } 3706 break; // No more entries to consider 3707 } 3708 } 3709 3710 for (const auto &pos : reexport_shlib_needs_fixup) { 3711 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3712 if (undef_pos != undefined_name_to_desc.end()) { 3713 const uint8_t dylib_ordinal = 3714 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3715 if (dylib_ordinal > 0 && 3716 dylib_ordinal < dylib_files.GetSize()) 3717 sym[pos.first].SetReExportedSymbolSharedLibrary( 3718 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3719 } 3720 } 3721 } 3722 } 3723 } 3724 } 3725 } 3726 3727 // Must reset this in case it was mutated above! 3728 nlist_data_offset = 0; 3729 #endif 3730 3731 if (nlist_data.GetByteSize() > 0) { 3732 3733 // If the sym array was not created while parsing the DSC unmapped 3734 // symbols, create it now. 3735 if (sym == NULL) { 3736 sym = symtab->Resize(symtab_load_command.nsyms + 3737 m_dysymtab.nindirectsyms); 3738 num_syms = symtab->GetNumSymbols(); 3739 } 3740 3741 if (unmapped_local_symbols_found) { 3742 assert(m_dysymtab.ilocalsym == 0); 3743 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3744 nlist_idx = m_dysymtab.nlocalsym; 3745 } else { 3746 nlist_idx = 0; 3747 } 3748 3749 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3750 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3751 UndefinedNameToDescMap undefined_name_to_desc; 3752 SymbolIndexToName reexport_shlib_needs_fixup; 3753 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 3754 struct nlist_64 nlist; 3755 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, 3756 nlist_byte_size)) 3757 break; 3758 3759 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3760 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 3761 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 3762 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 3763 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 3764 3765 SymbolType type = eSymbolTypeInvalid; 3766 const char *symbol_name = NULL; 3767 3768 if (have_strtab_data) { 3769 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3770 3771 if (symbol_name == NULL) { 3772 // No symbol should be NULL, even the symbols with no string values 3773 // should have an offset zero which points to an empty C-string 3774 Host::SystemLog(Host::eSystemLogError, 3775 "error: symbol[%u] has invalid string table offset " 3776 "0x%x in %s, ignoring symbol\n", 3777 nlist_idx, nlist.n_strx, 3778 module_sp->GetFileSpec().GetPath().c_str()); 3779 continue; 3780 } 3781 if (symbol_name[0] == '\0') 3782 symbol_name = NULL; 3783 } else { 3784 const addr_t str_addr = strtab_addr + nlist.n_strx; 3785 Status str_error; 3786 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3787 str_error)) 3788 symbol_name = memory_symbol_name.c_str(); 3789 } 3790 const char *symbol_name_non_abi_mangled = NULL; 3791 3792 SectionSP symbol_section; 3793 lldb::addr_t symbol_byte_size = 0; 3794 bool add_nlist = true; 3795 bool is_gsym = false; 3796 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3797 bool demangled_is_synthesized = false; 3798 bool set_value = true; 3799 assert(sym_idx < num_syms); 3800 3801 sym[sym_idx].SetDebug(is_debug); 3802 3803 if (is_debug) { 3804 switch (nlist.n_type) { 3805 case N_GSYM: 3806 // global symbol: name,,NO_SECT,type,0 3807 // Sometimes the N_GSYM value contains the address. 3808 3809 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3810 // the ObjC data. They 3811 // have the same address, but we want to ensure that we always find 3812 // only the real symbol, 'cause we don't currently correctly 3813 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3814 // type. This is a temporary hack to make sure the ObjectiveC 3815 // symbols get treated correctly. To do this right, we should 3816 // coalesce all the GSYM & global symbols that have the same 3817 // address. 3818 is_gsym = true; 3819 sym[sym_idx].SetExternal(true); 3820 3821 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3822 llvm::StringRef symbol_name_ref(symbol_name); 3823 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3824 symbol_name_non_abi_mangled = symbol_name + 1; 3825 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3826 type = eSymbolTypeObjCClass; 3827 demangled_is_synthesized = true; 3828 3829 } else if (symbol_name_ref.startswith( 3830 g_objc_v2_prefix_metaclass)) { 3831 symbol_name_non_abi_mangled = symbol_name + 1; 3832 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3833 type = eSymbolTypeObjCMetaClass; 3834 demangled_is_synthesized = true; 3835 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3836 symbol_name_non_abi_mangled = symbol_name + 1; 3837 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3838 type = eSymbolTypeObjCIVar; 3839 demangled_is_synthesized = true; 3840 } 3841 } else { 3842 if (nlist.n_value != 0) 3843 symbol_section = 3844 section_info.GetSection(nlist.n_sect, nlist.n_value); 3845 type = eSymbolTypeData; 3846 } 3847 break; 3848 3849 case N_FNAME: 3850 // procedure name (f77 kludge): name,,NO_SECT,0,0 3851 type = eSymbolTypeCompiler; 3852 break; 3853 3854 case N_FUN: 3855 // procedure: name,,n_sect,linenumber,address 3856 if (symbol_name) { 3857 type = eSymbolTypeCode; 3858 symbol_section = 3859 section_info.GetSection(nlist.n_sect, nlist.n_value); 3860 3861 N_FUN_addr_to_sym_idx.insert( 3862 std::make_pair(nlist.n_value, sym_idx)); 3863 // We use the current number of symbols in the symbol table in 3864 // lieu of using nlist_idx in case we ever start trimming entries 3865 // out 3866 N_FUN_indexes.push_back(sym_idx); 3867 } else { 3868 type = eSymbolTypeCompiler; 3869 3870 if (!N_FUN_indexes.empty()) { 3871 // Copy the size of the function into the original STAB entry 3872 // so we don't have to hunt for it later 3873 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3874 ->SetByteSize(nlist.n_value); 3875 N_FUN_indexes.pop_back(); 3876 // We don't really need the end function STAB as it contains 3877 // the size which we already placed with the original symbol, 3878 // so don't add it if we want a minimal symbol table 3879 add_nlist = false; 3880 } 3881 } 3882 break; 3883 3884 case N_STSYM: 3885 // static symbol: name,,n_sect,type,address 3886 N_STSYM_addr_to_sym_idx.insert( 3887 std::make_pair(nlist.n_value, sym_idx)); 3888 symbol_section = 3889 section_info.GetSection(nlist.n_sect, nlist.n_value); 3890 if (symbol_name && symbol_name[0]) { 3891 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3892 eSymbolTypeData); 3893 } 3894 break; 3895 3896 case N_LCSYM: 3897 // .lcomm symbol: name,,n_sect,type,address 3898 symbol_section = 3899 section_info.GetSection(nlist.n_sect, nlist.n_value); 3900 type = eSymbolTypeCommonBlock; 3901 break; 3902 3903 case N_BNSYM: 3904 // We use the current number of symbols in the symbol table in lieu 3905 // of using nlist_idx in case we ever start trimming entries out 3906 // Skip these if we want minimal symbol tables 3907 add_nlist = false; 3908 break; 3909 3910 case N_ENSYM: 3911 // Set the size of the N_BNSYM to the terminating index of this 3912 // N_ENSYM so that we can always skip the entire symbol if we need 3913 // to navigate more quickly at the source level when parsing STABS 3914 // Skip these if we want minimal symbol tables 3915 add_nlist = false; 3916 break; 3917 3918 case N_OPT: 3919 // emitted with gcc2_compiled and in gcc source 3920 type = eSymbolTypeCompiler; 3921 break; 3922 3923 case N_RSYM: 3924 // register sym: name,,NO_SECT,type,register 3925 type = eSymbolTypeVariable; 3926 break; 3927 3928 case N_SLINE: 3929 // src line: 0,,n_sect,linenumber,address 3930 symbol_section = 3931 section_info.GetSection(nlist.n_sect, nlist.n_value); 3932 type = eSymbolTypeLineEntry; 3933 break; 3934 3935 case N_SSYM: 3936 // structure elt: name,,NO_SECT,type,struct_offset 3937 type = eSymbolTypeVariableType; 3938 break; 3939 3940 case N_SO: 3941 // source file name 3942 type = eSymbolTypeSourceFile; 3943 if (symbol_name == NULL) { 3944 add_nlist = false; 3945 if (N_SO_index != UINT32_MAX) { 3946 // Set the size of the N_SO to the terminating index of this 3947 // N_SO so that we can always skip the entire N_SO if we need 3948 // to navigate more quickly at the source level when parsing 3949 // STABS 3950 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3951 symbol_ptr->SetByteSize(sym_idx); 3952 symbol_ptr->SetSizeIsSibling(true); 3953 } 3954 N_NSYM_indexes.clear(); 3955 N_INCL_indexes.clear(); 3956 N_BRAC_indexes.clear(); 3957 N_COMM_indexes.clear(); 3958 N_FUN_indexes.clear(); 3959 N_SO_index = UINT32_MAX; 3960 } else { 3961 // We use the current number of symbols in the symbol table in 3962 // lieu of using nlist_idx in case we ever start trimming entries 3963 // out 3964 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3965 if (N_SO_has_full_path) { 3966 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3967 // We have two consecutive N_SO entries where the first 3968 // contains a directory and the second contains a full path. 3969 sym[sym_idx - 1].GetMangled().SetValue( 3970 ConstString(symbol_name), false); 3971 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3972 add_nlist = false; 3973 } else { 3974 // This is the first entry in a N_SO that contains a 3975 // directory or a full path to the source file 3976 N_SO_index = sym_idx; 3977 } 3978 } else if ((N_SO_index == sym_idx - 1) && 3979 ((sym_idx - 1) < num_syms)) { 3980 // This is usually the second N_SO entry that contains just the 3981 // filename, so here we combine it with the first one if we are 3982 // minimizing the symbol table 3983 const char *so_path = 3984 sym[sym_idx - 1] 3985 .GetMangled() 3986 .GetDemangledName(lldb::eLanguageTypeUnknown) 3987 .AsCString(); 3988 if (so_path && so_path[0]) { 3989 std::string full_so_path(so_path); 3990 const size_t double_slash_pos = full_so_path.find("//"); 3991 if (double_slash_pos != std::string::npos) { 3992 // The linker has been generating bad N_SO entries with 3993 // doubled up paths in the format "%s%s" where the first 3994 // string in the DW_AT_comp_dir, and the second is the 3995 // directory for the source file so you end up with a path 3996 // that looks like "/tmp/src//tmp/src/" 3997 FileSpec so_dir(so_path, false); 3998 if (!so_dir.Exists()) { 3999 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 4000 false); 4001 if (so_dir.Exists()) { 4002 // Trim off the incorrect path 4003 full_so_path.erase(0, double_slash_pos + 1); 4004 } 4005 } 4006 } 4007 if (*full_so_path.rbegin() != '/') 4008 full_so_path += '/'; 4009 full_so_path += symbol_name; 4010 sym[sym_idx - 1].GetMangled().SetValue( 4011 ConstString(full_so_path.c_str()), false); 4012 add_nlist = false; 4013 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4014 } 4015 } else { 4016 // This could be a relative path to a N_SO 4017 N_SO_index = sym_idx; 4018 } 4019 } 4020 break; 4021 4022 case N_OSO: 4023 // object file name: name,,0,0,st_mtime 4024 type = eSymbolTypeObjectFile; 4025 break; 4026 4027 case N_LSYM: 4028 // local sym: name,,NO_SECT,type,offset 4029 type = eSymbolTypeLocal; 4030 break; 4031 4032 //---------------------------------------------------------------------- 4033 // INCL scopes 4034 //---------------------------------------------------------------------- 4035 case N_BINCL: 4036 // include file beginning: name,,NO_SECT,0,sum We use the current 4037 // number of symbols in the symbol table in lieu of using nlist_idx 4038 // in case we ever start trimming entries out 4039 N_INCL_indexes.push_back(sym_idx); 4040 type = eSymbolTypeScopeBegin; 4041 break; 4042 4043 case N_EINCL: 4044 // include file end: name,,NO_SECT,0,0 4045 // Set the size of the N_BINCL to the terminating index of this 4046 // N_EINCL so that we can always skip the entire symbol if we need 4047 // to navigate more quickly at the source level when parsing STABS 4048 if (!N_INCL_indexes.empty()) { 4049 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4050 symbol_ptr->SetByteSize(sym_idx + 1); 4051 symbol_ptr->SetSizeIsSibling(true); 4052 N_INCL_indexes.pop_back(); 4053 } 4054 type = eSymbolTypeScopeEnd; 4055 break; 4056 4057 case N_SOL: 4058 // #included file name: name,,n_sect,0,address 4059 type = eSymbolTypeHeaderFile; 4060 4061 // We currently don't use the header files on darwin 4062 add_nlist = false; 4063 break; 4064 4065 case N_PARAMS: 4066 // compiler parameters: name,,NO_SECT,0,0 4067 type = eSymbolTypeCompiler; 4068 break; 4069 4070 case N_VERSION: 4071 // compiler version: name,,NO_SECT,0,0 4072 type = eSymbolTypeCompiler; 4073 break; 4074 4075 case N_OLEVEL: 4076 // compiler -O level: name,,NO_SECT,0,0 4077 type = eSymbolTypeCompiler; 4078 break; 4079 4080 case N_PSYM: 4081 // parameter: name,,NO_SECT,type,offset 4082 type = eSymbolTypeVariable; 4083 break; 4084 4085 case N_ENTRY: 4086 // alternate entry: name,,n_sect,linenumber,address 4087 symbol_section = 4088 section_info.GetSection(nlist.n_sect, nlist.n_value); 4089 type = eSymbolTypeLineEntry; 4090 break; 4091 4092 //---------------------------------------------------------------------- 4093 // Left and Right Braces 4094 //---------------------------------------------------------------------- 4095 case N_LBRAC: 4096 // left bracket: 0,,NO_SECT,nesting level,address We use the 4097 // current number of symbols in the symbol table in lieu of using 4098 // nlist_idx in case we ever start trimming entries out 4099 symbol_section = 4100 section_info.GetSection(nlist.n_sect, nlist.n_value); 4101 N_BRAC_indexes.push_back(sym_idx); 4102 type = eSymbolTypeScopeBegin; 4103 break; 4104 4105 case N_RBRAC: 4106 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4107 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4108 // can always skip the entire symbol if we need to navigate more 4109 // quickly at the source level when parsing STABS 4110 symbol_section = 4111 section_info.GetSection(nlist.n_sect, nlist.n_value); 4112 if (!N_BRAC_indexes.empty()) { 4113 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4114 symbol_ptr->SetByteSize(sym_idx + 1); 4115 symbol_ptr->SetSizeIsSibling(true); 4116 N_BRAC_indexes.pop_back(); 4117 } 4118 type = eSymbolTypeScopeEnd; 4119 break; 4120 4121 case N_EXCL: 4122 // deleted include file: name,,NO_SECT,0,sum 4123 type = eSymbolTypeHeaderFile; 4124 break; 4125 4126 //---------------------------------------------------------------------- 4127 // COMM scopes 4128 //---------------------------------------------------------------------- 4129 case N_BCOMM: 4130 // begin common: name,,NO_SECT,0,0 4131 // We use the current number of symbols in the symbol table in lieu 4132 // of using nlist_idx in case we ever start trimming entries out 4133 type = eSymbolTypeScopeBegin; 4134 N_COMM_indexes.push_back(sym_idx); 4135 break; 4136 4137 case N_ECOML: 4138 // end common (local name): 0,,n_sect,0,address 4139 symbol_section = 4140 section_info.GetSection(nlist.n_sect, nlist.n_value); 4141 LLVM_FALLTHROUGH; 4142 4143 case N_ECOMM: 4144 // end common: name,,n_sect,0,0 4145 // Set the size of the N_BCOMM to the terminating index of this 4146 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4147 // we need to navigate more quickly at the source level when 4148 // parsing STABS 4149 if (!N_COMM_indexes.empty()) { 4150 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4151 symbol_ptr->SetByteSize(sym_idx + 1); 4152 symbol_ptr->SetSizeIsSibling(true); 4153 N_COMM_indexes.pop_back(); 4154 } 4155 type = eSymbolTypeScopeEnd; 4156 break; 4157 4158 case N_LENG: 4159 // second stab entry with length information 4160 type = eSymbolTypeAdditional; 4161 break; 4162 4163 default: 4164 break; 4165 } 4166 } else { 4167 // uint8_t n_pext = N_PEXT & nlist.n_type; 4168 uint8_t n_type = N_TYPE & nlist.n_type; 4169 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4170 4171 switch (n_type) { 4172 case N_INDR: { 4173 const char *reexport_name_cstr = 4174 strtab_data.PeekCStr(nlist.n_value); 4175 if (reexport_name_cstr && reexport_name_cstr[0]) { 4176 type = eSymbolTypeReExported; 4177 ConstString reexport_name( 4178 reexport_name_cstr + 4179 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4180 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4181 set_value = false; 4182 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4183 indirect_symbol_names.insert( 4184 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4185 } else 4186 type = eSymbolTypeUndefined; 4187 } break; 4188 4189 case N_UNDF: 4190 if (symbol_name && symbol_name[0]) { 4191 ConstString undefined_name(symbol_name + 4192 ((symbol_name[0] == '_') ? 1 : 0)); 4193 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4194 } 4195 LLVM_FALLTHROUGH; 4196 4197 case N_PBUD: 4198 type = eSymbolTypeUndefined; 4199 break; 4200 4201 case N_ABS: 4202 type = eSymbolTypeAbsolute; 4203 break; 4204 4205 case N_SECT: { 4206 symbol_section = 4207 section_info.GetSection(nlist.n_sect, nlist.n_value); 4208 4209 if (!symbol_section) { 4210 // TODO: warn about this? 4211 add_nlist = false; 4212 break; 4213 } 4214 4215 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4216 type = eSymbolTypeException; 4217 } else { 4218 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4219 4220 switch (section_type) { 4221 case S_CSTRING_LITERALS: 4222 type = eSymbolTypeData; 4223 break; // section with only literal C strings 4224 case S_4BYTE_LITERALS: 4225 type = eSymbolTypeData; 4226 break; // section with only 4 byte literals 4227 case S_8BYTE_LITERALS: 4228 type = eSymbolTypeData; 4229 break; // section with only 8 byte literals 4230 case S_LITERAL_POINTERS: 4231 type = eSymbolTypeTrampoline; 4232 break; // section with only pointers to literals 4233 case S_NON_LAZY_SYMBOL_POINTERS: 4234 type = eSymbolTypeTrampoline; 4235 break; // section with only non-lazy symbol pointers 4236 case S_LAZY_SYMBOL_POINTERS: 4237 type = eSymbolTypeTrampoline; 4238 break; // section with only lazy symbol pointers 4239 case S_SYMBOL_STUBS: 4240 type = eSymbolTypeTrampoline; 4241 break; // section with only symbol stubs, byte size of stub in 4242 // the reserved2 field 4243 case S_MOD_INIT_FUNC_POINTERS: 4244 type = eSymbolTypeCode; 4245 break; // section with only function pointers for initialization 4246 case S_MOD_TERM_FUNC_POINTERS: 4247 type = eSymbolTypeCode; 4248 break; // section with only function pointers for termination 4249 case S_INTERPOSING: 4250 type = eSymbolTypeTrampoline; 4251 break; // section with only pairs of function pointers for 4252 // interposing 4253 case S_16BYTE_LITERALS: 4254 type = eSymbolTypeData; 4255 break; // section with only 16 byte literals 4256 case S_DTRACE_DOF: 4257 type = eSymbolTypeInstrumentation; 4258 break; 4259 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4260 type = eSymbolTypeTrampoline; 4261 break; 4262 default: 4263 switch (symbol_section->GetType()) { 4264 case lldb::eSectionTypeCode: 4265 type = eSymbolTypeCode; 4266 break; 4267 case eSectionTypeData: 4268 case eSectionTypeDataCString: // Inlined C string data 4269 case eSectionTypeDataCStringPointers: // Pointers to C string 4270 // data 4271 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4272 // the symbol table 4273 case eSectionTypeData4: 4274 case eSectionTypeData8: 4275 case eSectionTypeData16: 4276 type = eSymbolTypeData; 4277 break; 4278 default: 4279 break; 4280 } 4281 break; 4282 } 4283 4284 if (type == eSymbolTypeInvalid) { 4285 const char *symbol_sect_name = 4286 symbol_section->GetName().AsCString(); 4287 if (symbol_section->IsDescendant(text_section_sp.get())) { 4288 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4289 S_ATTR_SELF_MODIFYING_CODE | 4290 S_ATTR_SOME_INSTRUCTIONS)) 4291 type = eSymbolTypeData; 4292 else 4293 type = eSymbolTypeCode; 4294 } else if (symbol_section->IsDescendant( 4295 data_section_sp.get()) || 4296 symbol_section->IsDescendant( 4297 data_dirty_section_sp.get()) || 4298 symbol_section->IsDescendant( 4299 data_const_section_sp.get())) { 4300 if (symbol_sect_name && 4301 ::strstr(symbol_sect_name, "__objc") == 4302 symbol_sect_name) { 4303 type = eSymbolTypeRuntime; 4304 4305 if (symbol_name) { 4306 llvm::StringRef symbol_name_ref(symbol_name); 4307 if (symbol_name_ref.startswith("_OBJC_")) { 4308 static const llvm::StringRef g_objc_v2_prefix_class( 4309 "_OBJC_CLASS_$_"); 4310 static const llvm::StringRef g_objc_v2_prefix_metaclass( 4311 "_OBJC_METACLASS_$_"); 4312 static const llvm::StringRef g_objc_v2_prefix_ivar( 4313 "_OBJC_IVAR_$_"); 4314 if (symbol_name_ref.startswith( 4315 g_objc_v2_prefix_class)) { 4316 symbol_name_non_abi_mangled = symbol_name + 1; 4317 symbol_name = 4318 symbol_name + g_objc_v2_prefix_class.size(); 4319 type = eSymbolTypeObjCClass; 4320 demangled_is_synthesized = true; 4321 } else if (symbol_name_ref.startswith( 4322 g_objc_v2_prefix_metaclass)) { 4323 symbol_name_non_abi_mangled = symbol_name + 1; 4324 symbol_name = 4325 symbol_name + g_objc_v2_prefix_metaclass.size(); 4326 type = eSymbolTypeObjCMetaClass; 4327 demangled_is_synthesized = true; 4328 } else if (symbol_name_ref.startswith( 4329 g_objc_v2_prefix_ivar)) { 4330 symbol_name_non_abi_mangled = symbol_name + 1; 4331 symbol_name = 4332 symbol_name + g_objc_v2_prefix_ivar.size(); 4333 type = eSymbolTypeObjCIVar; 4334 demangled_is_synthesized = true; 4335 } 4336 } 4337 } 4338 } else if (symbol_sect_name && 4339 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4340 symbol_sect_name) { 4341 type = eSymbolTypeException; 4342 } else { 4343 type = eSymbolTypeData; 4344 } 4345 } else if (symbol_sect_name && 4346 ::strstr(symbol_sect_name, "__IMPORT") == 4347 symbol_sect_name) { 4348 type = eSymbolTypeTrampoline; 4349 } else if (symbol_section->IsDescendant( 4350 objc_section_sp.get())) { 4351 type = eSymbolTypeRuntime; 4352 if (symbol_name && symbol_name[0] == '.') { 4353 llvm::StringRef symbol_name_ref(symbol_name); 4354 static const llvm::StringRef g_objc_v1_prefix_class( 4355 ".objc_class_name_"); 4356 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4357 symbol_name_non_abi_mangled = symbol_name; 4358 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4359 type = eSymbolTypeObjCClass; 4360 demangled_is_synthesized = true; 4361 } 4362 } 4363 } 4364 } 4365 } 4366 } break; 4367 } 4368 } 4369 4370 if (add_nlist) { 4371 uint64_t symbol_value = nlist.n_value; 4372 4373 if (symbol_name_non_abi_mangled) { 4374 sym[sym_idx].GetMangled().SetMangledName( 4375 ConstString(symbol_name_non_abi_mangled)); 4376 sym[sym_idx].GetMangled().SetDemangledName( 4377 ConstString(symbol_name)); 4378 } else { 4379 bool symbol_name_is_mangled = false; 4380 4381 if (symbol_name && symbol_name[0] == '_') { 4382 symbol_name_is_mangled = symbol_name[1] == '_'; 4383 symbol_name++; // Skip the leading underscore 4384 } 4385 4386 if (symbol_name) { 4387 ConstString const_symbol_name(symbol_name); 4388 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4389 symbol_name_is_mangled); 4390 } 4391 } 4392 4393 if (is_gsym) { 4394 const char *gsym_name = sym[sym_idx] 4395 .GetMangled() 4396 .GetName(lldb::eLanguageTypeUnknown, 4397 Mangled::ePreferMangled) 4398 .GetCString(); 4399 if (gsym_name) 4400 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4401 } 4402 4403 if (symbol_section) { 4404 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4405 if (symbol_byte_size == 0 && function_starts_count > 0) { 4406 addr_t symbol_lookup_file_addr = nlist.n_value; 4407 // Do an exact address match for non-ARM addresses, else get the 4408 // closest since the symbol might be a thumb symbol which has an 4409 // address with bit zero set 4410 FunctionStarts::Entry *func_start_entry = 4411 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4412 if (is_arm && func_start_entry) { 4413 // Verify that the function start address is the symbol address 4414 // (ARM) or the symbol address + 1 (thumb) 4415 if (func_start_entry->addr != symbol_lookup_file_addr && 4416 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4417 // Not the right entry, NULL it out... 4418 func_start_entry = NULL; 4419 } 4420 } 4421 if (func_start_entry) { 4422 func_start_entry->data = true; 4423 4424 addr_t symbol_file_addr = func_start_entry->addr; 4425 if (is_arm) 4426 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4427 4428 const FunctionStarts::Entry *next_func_start_entry = 4429 function_starts.FindNextEntry(func_start_entry); 4430 const addr_t section_end_file_addr = 4431 section_file_addr + symbol_section->GetByteSize(); 4432 if (next_func_start_entry) { 4433 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4434 // Be sure the clear the Thumb address bit when we calculate 4435 // the size from the current and next address 4436 if (is_arm) 4437 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4438 symbol_byte_size = std::min<lldb::addr_t>( 4439 next_symbol_file_addr - symbol_file_addr, 4440 section_end_file_addr - symbol_file_addr); 4441 } else { 4442 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4443 } 4444 } 4445 } 4446 symbol_value -= section_file_addr; 4447 } 4448 4449 if (is_debug == false) { 4450 if (type == eSymbolTypeCode) { 4451 // See if we can find a N_FUN entry for any code symbols. If we 4452 // do find a match, and the name matches, then we can merge the 4453 // two into just the function symbol to avoid duplicate entries 4454 // in the symbol table 4455 std::pair<ValueToSymbolIndexMap::const_iterator, 4456 ValueToSymbolIndexMap::const_iterator> 4457 range; 4458 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4459 if (range.first != range.second) { 4460 bool found_it = false; 4461 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4462 pos != range.second; ++pos) { 4463 if (sym[sym_idx].GetMangled().GetName( 4464 lldb::eLanguageTypeUnknown, 4465 Mangled::ePreferMangled) == 4466 sym[pos->second].GetMangled().GetName( 4467 lldb::eLanguageTypeUnknown, 4468 Mangled::ePreferMangled)) { 4469 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4470 // We just need the flags from the linker symbol, so put 4471 // these flags into the N_FUN flags to avoid duplicate 4472 // symbols in the symbol table 4473 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4474 sym[pos->second].SetFlags(nlist.n_type << 16 | 4475 nlist.n_desc); 4476 if (resolver_addresses.find(nlist.n_value) != 4477 resolver_addresses.end()) 4478 sym[pos->second].SetType(eSymbolTypeResolver); 4479 sym[sym_idx].Clear(); 4480 found_it = true; 4481 break; 4482 } 4483 } 4484 if (found_it) 4485 continue; 4486 } else { 4487 if (resolver_addresses.find(nlist.n_value) != 4488 resolver_addresses.end()) 4489 type = eSymbolTypeResolver; 4490 } 4491 } else if (type == eSymbolTypeData || 4492 type == eSymbolTypeObjCClass || 4493 type == eSymbolTypeObjCMetaClass || 4494 type == eSymbolTypeObjCIVar) { 4495 // See if we can find a N_STSYM entry for any data symbols. If we 4496 // do find a match, and the name matches, then we can merge the 4497 // two into just the Static symbol to avoid duplicate entries in 4498 // the symbol table 4499 std::pair<ValueToSymbolIndexMap::const_iterator, 4500 ValueToSymbolIndexMap::const_iterator> 4501 range; 4502 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4503 if (range.first != range.second) { 4504 bool found_it = false; 4505 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4506 pos != range.second; ++pos) { 4507 if (sym[sym_idx].GetMangled().GetName( 4508 lldb::eLanguageTypeUnknown, 4509 Mangled::ePreferMangled) == 4510 sym[pos->second].GetMangled().GetName( 4511 lldb::eLanguageTypeUnknown, 4512 Mangled::ePreferMangled)) { 4513 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4514 // We just need the flags from the linker symbol, so put 4515 // these flags into the N_STSYM flags to avoid duplicate 4516 // symbols in the symbol table 4517 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4518 sym[pos->second].SetFlags(nlist.n_type << 16 | 4519 nlist.n_desc); 4520 sym[sym_idx].Clear(); 4521 found_it = true; 4522 break; 4523 } 4524 } 4525 if (found_it) 4526 continue; 4527 } else { 4528 // Combine N_GSYM stab entries with the non stab symbol 4529 const char *gsym_name = sym[sym_idx] 4530 .GetMangled() 4531 .GetName(lldb::eLanguageTypeUnknown, 4532 Mangled::ePreferMangled) 4533 .GetCString(); 4534 if (gsym_name) { 4535 ConstNameToSymbolIndexMap::const_iterator pos = 4536 N_GSYM_name_to_sym_idx.find(gsym_name); 4537 if (pos != N_GSYM_name_to_sym_idx.end()) { 4538 const uint32_t GSYM_sym_idx = pos->second; 4539 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4540 // Copy the address, because often the N_GSYM address has 4541 // an invalid address of zero when the global is a common 4542 // symbol 4543 sym[GSYM_sym_idx].GetAddressRef().SetSection( 4544 symbol_section); 4545 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4546 // We just need the flags from the linker symbol, so put 4547 // these flags into the N_GSYM flags to avoid duplicate 4548 // symbols in the symbol table 4549 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 4550 nlist.n_desc); 4551 sym[sym_idx].Clear(); 4552 continue; 4553 } 4554 } 4555 } 4556 } 4557 } 4558 4559 sym[sym_idx].SetID(nlist_idx); 4560 sym[sym_idx].SetType(type); 4561 if (set_value) { 4562 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4563 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4564 } 4565 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4566 4567 if (symbol_byte_size > 0) 4568 sym[sym_idx].SetByteSize(symbol_byte_size); 4569 4570 if (demangled_is_synthesized) 4571 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4572 4573 ++sym_idx; 4574 } else { 4575 sym[sym_idx].Clear(); 4576 } 4577 } 4578 4579 for (const auto &pos : reexport_shlib_needs_fixup) { 4580 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4581 if (undef_pos != undefined_name_to_desc.end()) { 4582 const uint8_t dylib_ordinal = 4583 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4584 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4585 sym[pos.first].SetReExportedSymbolSharedLibrary( 4586 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4587 } 4588 } 4589 } 4590 4591 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4592 4593 if (function_starts_count > 0) { 4594 uint32_t num_synthetic_function_symbols = 0; 4595 for (i = 0; i < function_starts_count; ++i) { 4596 if (function_starts.GetEntryRef(i).data == false) 4597 ++num_synthetic_function_symbols; 4598 } 4599 4600 if (num_synthetic_function_symbols > 0) { 4601 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4602 num_syms = sym_idx + num_synthetic_function_symbols; 4603 sym = symtab->Resize(num_syms); 4604 } 4605 for (i = 0; i < function_starts_count; ++i) { 4606 const FunctionStarts::Entry *func_start_entry = 4607 function_starts.GetEntryAtIndex(i); 4608 if (func_start_entry->data == false) { 4609 addr_t symbol_file_addr = func_start_entry->addr; 4610 uint32_t symbol_flags = 0; 4611 if (is_arm) { 4612 if (symbol_file_addr & 1) 4613 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4614 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4615 } 4616 Address symbol_addr; 4617 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4618 SectionSP symbol_section(symbol_addr.GetSection()); 4619 uint32_t symbol_byte_size = 0; 4620 if (symbol_section) { 4621 const addr_t section_file_addr = 4622 symbol_section->GetFileAddress(); 4623 const FunctionStarts::Entry *next_func_start_entry = 4624 function_starts.FindNextEntry(func_start_entry); 4625 const addr_t section_end_file_addr = 4626 section_file_addr + symbol_section->GetByteSize(); 4627 if (next_func_start_entry) { 4628 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4629 if (is_arm) 4630 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4631 symbol_byte_size = std::min<lldb::addr_t>( 4632 next_symbol_file_addr - symbol_file_addr, 4633 section_end_file_addr - symbol_file_addr); 4634 } else { 4635 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4636 } 4637 sym[sym_idx].SetID(synthetic_sym_id++); 4638 sym[sym_idx].GetMangled().SetDemangledName( 4639 GetNextSyntheticSymbolName()); 4640 sym[sym_idx].SetType(eSymbolTypeCode); 4641 sym[sym_idx].SetIsSynthetic(true); 4642 sym[sym_idx].GetAddressRef() = symbol_addr; 4643 if (symbol_flags) 4644 sym[sym_idx].SetFlags(symbol_flags); 4645 if (symbol_byte_size) 4646 sym[sym_idx].SetByteSize(symbol_byte_size); 4647 ++sym_idx; 4648 } 4649 } 4650 } 4651 } 4652 } 4653 } 4654 4655 // Trim our symbols down to just what we ended up with after removing any 4656 // symbols. 4657 if (sym_idx < num_syms) { 4658 num_syms = sym_idx; 4659 sym = symtab->Resize(num_syms); 4660 } 4661 4662 // Now synthesize indirect symbols 4663 if (m_dysymtab.nindirectsyms != 0) { 4664 if (indirect_symbol_index_data.GetByteSize()) { 4665 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4666 m_nlist_idx_to_sym_idx.end(); 4667 4668 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4669 ++sect_idx) { 4670 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4671 S_SYMBOL_STUBS) { 4672 uint32_t symbol_stub_byte_size = 4673 m_mach_sections[sect_idx].reserved2; 4674 if (symbol_stub_byte_size == 0) 4675 continue; 4676 4677 const uint32_t num_symbol_stubs = 4678 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4679 4680 if (num_symbol_stubs == 0) 4681 continue; 4682 4683 const uint32_t symbol_stub_index_offset = 4684 m_mach_sections[sect_idx].reserved1; 4685 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; 4686 ++stub_idx) { 4687 const uint32_t symbol_stub_index = 4688 symbol_stub_index_offset + stub_idx; 4689 const lldb::addr_t symbol_stub_addr = 4690 m_mach_sections[sect_idx].addr + 4691 (stub_idx * symbol_stub_byte_size); 4692 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4693 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4694 symbol_stub_offset, 4)) { 4695 const uint32_t stub_sym_id = 4696 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4697 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4698 continue; 4699 4700 NListIndexToSymbolIndexMap::const_iterator index_pos = 4701 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4702 Symbol *stub_symbol = NULL; 4703 if (index_pos != end_index_pos) { 4704 // We have a remapping from the original nlist index to a 4705 // current symbol index, so just look this up by index 4706 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4707 } else { 4708 // We need to lookup a symbol using the original nlist symbol 4709 // index since this index is coming from the S_SYMBOL_STUBS 4710 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4711 } 4712 4713 if (stub_symbol) { 4714 Address so_addr(symbol_stub_addr, section_list); 4715 4716 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4717 // Change the external symbol into a trampoline that makes 4718 // sense These symbols were N_UNDF N_EXT, and are useless 4719 // to us, so we can re-use them so we don't have to make up 4720 // a synthetic symbol for no good reason. 4721 if (resolver_addresses.find(symbol_stub_addr) == 4722 resolver_addresses.end()) 4723 stub_symbol->SetType(eSymbolTypeTrampoline); 4724 else 4725 stub_symbol->SetType(eSymbolTypeResolver); 4726 stub_symbol->SetExternal(false); 4727 stub_symbol->GetAddressRef() = so_addr; 4728 stub_symbol->SetByteSize(symbol_stub_byte_size); 4729 } else { 4730 // Make a synthetic symbol to describe the trampoline stub 4731 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4732 if (sym_idx >= num_syms) { 4733 sym = symtab->Resize(++num_syms); 4734 stub_symbol = NULL; // this pointer no longer valid 4735 } 4736 sym[sym_idx].SetID(synthetic_sym_id++); 4737 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4738 if (resolver_addresses.find(symbol_stub_addr) == 4739 resolver_addresses.end()) 4740 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4741 else 4742 sym[sym_idx].SetType(eSymbolTypeResolver); 4743 sym[sym_idx].SetIsSynthetic(true); 4744 sym[sym_idx].GetAddressRef() = so_addr; 4745 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4746 ++sym_idx; 4747 } 4748 } else { 4749 if (log) 4750 log->Warning("symbol stub referencing symbol table symbol " 4751 "%u that isn't in our minimal symbol table, " 4752 "fix this!!!", 4753 stub_sym_id); 4754 } 4755 } 4756 } 4757 } 4758 } 4759 } 4760 } 4761 4762 if (!trie_entries.empty()) { 4763 for (const auto &e : trie_entries) { 4764 if (e.entry.import_name) { 4765 // Only add indirect symbols from the Trie entries if we didn't have 4766 // a N_INDR nlist entry for this already 4767 if (indirect_symbol_names.find(e.entry.name) == 4768 indirect_symbol_names.end()) { 4769 // Make a synthetic symbol to describe re-exported symbol. 4770 if (sym_idx >= num_syms) 4771 sym = symtab->Resize(++num_syms); 4772 sym[sym_idx].SetID(synthetic_sym_id++); 4773 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4774 sym[sym_idx].SetType(eSymbolTypeReExported); 4775 sym[sym_idx].SetIsSynthetic(true); 4776 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4777 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4778 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4779 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4780 } 4781 ++sym_idx; 4782 } 4783 } 4784 } 4785 } 4786 4787 // StreamFile s(stdout, false); 4788 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4789 // symtab->Dump(&s, NULL, eSortOrderNone); 4790 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4791 // sizes 4792 symtab->CalculateSymbolSizes(); 4793 4794 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4795 // symtab->Dump(&s, NULL, eSortOrderNone); 4796 4797 return symtab->GetNumSymbols(); 4798 } 4799 return 0; 4800 } 4801 4802 void ObjectFileMachO::Dump(Stream *s) { 4803 ModuleSP module_sp(GetModule()); 4804 if (module_sp) { 4805 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4806 s->Printf("%p: ", static_cast<void *>(this)); 4807 s->Indent(); 4808 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4809 s->PutCString("ObjectFileMachO64"); 4810 else 4811 s->PutCString("ObjectFileMachO32"); 4812 4813 ArchSpec header_arch; 4814 GetArchitecture(header_arch); 4815 4816 *s << ", file = '" << m_file 4817 << "', arch = " << header_arch.GetArchitectureName() << "\n"; 4818 4819 SectionList *sections = GetSectionList(); 4820 if (sections) 4821 sections->Dump(s, NULL, true, UINT32_MAX); 4822 4823 if (m_symtab_ap.get()) 4824 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4825 } 4826 } 4827 4828 bool ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4829 const lldb_private::DataExtractor &data, 4830 lldb::offset_t lc_offset, 4831 lldb_private::UUID &uuid) { 4832 uint32_t i; 4833 struct uuid_command load_cmd; 4834 4835 lldb::offset_t offset = lc_offset; 4836 for (i = 0; i < header.ncmds; ++i) { 4837 const lldb::offset_t cmd_offset = offset; 4838 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4839 break; 4840 4841 if (load_cmd.cmd == LC_UUID) { 4842 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4843 4844 if (uuid_bytes) { 4845 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4846 // We pretend these object files have no UUID to prevent crashing. 4847 4848 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4849 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4850 0xbb, 0x14, 0xf0, 0x0d}; 4851 4852 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4853 return false; 4854 4855 uuid.SetBytes(uuid_bytes); 4856 return true; 4857 } 4858 return false; 4859 } 4860 offset = cmd_offset + load_cmd.cmdsize; 4861 } 4862 return false; 4863 } 4864 4865 bool ObjectFileMachO::GetArchitecture(const llvm::MachO::mach_header &header, 4866 const lldb_private::DataExtractor &data, 4867 lldb::offset_t lc_offset, 4868 ArchSpec &arch) { 4869 arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4870 4871 if (arch.IsValid()) { 4872 llvm::Triple &triple = arch.GetTriple(); 4873 4874 // Set OS to an unspecified unknown or a "*" so it can match any OS 4875 triple.setOS(llvm::Triple::UnknownOS); 4876 triple.setOSName(llvm::StringRef()); 4877 4878 if (header.filetype == MH_PRELOAD) { 4879 if (header.cputype == CPU_TYPE_ARM) { 4880 // If this is a 32-bit arm binary, and it's a standalone binary, force 4881 // the Vendor to Apple so we don't accidentally pick up the generic 4882 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4883 // frame pointer register; most other armv7 ABIs use a combination of 4884 // r7 and r11. 4885 triple.setVendor(llvm::Triple::Apple); 4886 } else { 4887 // Set vendor to an unspecified unknown or a "*" so it can match any 4888 // vendor This is required for correct behavior of EFI debugging on 4889 // x86_64 4890 triple.setVendor(llvm::Triple::UnknownVendor); 4891 triple.setVendorName(llvm::StringRef()); 4892 } 4893 return true; 4894 } else { 4895 struct load_command load_cmd; 4896 4897 lldb::offset_t offset = lc_offset; 4898 for (uint32_t i = 0; i < header.ncmds; ++i) { 4899 const lldb::offset_t cmd_offset = offset; 4900 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4901 break; 4902 4903 switch (load_cmd.cmd) { 4904 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4905 triple.setOS(llvm::Triple::IOS); 4906 return true; 4907 4908 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4909 triple.setOS(llvm::Triple::MacOSX); 4910 return true; 4911 4912 case llvm::MachO::LC_VERSION_MIN_TVOS: 4913 triple.setOS(llvm::Triple::TvOS); 4914 return true; 4915 4916 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4917 triple.setOS(llvm::Triple::WatchOS); 4918 return true; 4919 4920 default: 4921 break; 4922 } 4923 4924 offset = cmd_offset + load_cmd.cmdsize; 4925 } 4926 4927 if (header.filetype != MH_KEXT_BUNDLE) { 4928 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 4929 // so lets not say our Vendor is Apple, leave it as an unspecified 4930 // unknown 4931 triple.setVendor(llvm::Triple::UnknownVendor); 4932 triple.setVendorName(llvm::StringRef()); 4933 } 4934 } 4935 } 4936 return arch.IsValid(); 4937 } 4938 4939 bool ObjectFileMachO::GetUUID(lldb_private::UUID *uuid) { 4940 ModuleSP module_sp(GetModule()); 4941 if (module_sp) { 4942 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4943 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4944 return GetUUID(m_header, m_data, offset, *uuid); 4945 } 4946 return false; 4947 } 4948 4949 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 4950 uint32_t count = 0; 4951 ModuleSP module_sp(GetModule()); 4952 if (module_sp) { 4953 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4954 struct load_command load_cmd; 4955 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4956 std::vector<std::string> rpath_paths; 4957 std::vector<std::string> rpath_relative_paths; 4958 std::vector<std::string> at_exec_relative_paths; 4959 const bool resolve_path = false; // Don't resolve the dependent file paths 4960 // since they may not reside on this 4961 // system 4962 uint32_t i; 4963 for (i = 0; i < m_header.ncmds; ++i) { 4964 const uint32_t cmd_offset = offset; 4965 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4966 break; 4967 4968 switch (load_cmd.cmd) { 4969 case LC_RPATH: 4970 case LC_LOAD_DYLIB: 4971 case LC_LOAD_WEAK_DYLIB: 4972 case LC_REEXPORT_DYLIB: 4973 case LC_LOAD_DYLINKER: 4974 case LC_LOADFVMLIB: 4975 case LC_LOAD_UPWARD_DYLIB: { 4976 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 4977 const char *path = m_data.PeekCStr(name_offset); 4978 if (path) { 4979 if (load_cmd.cmd == LC_RPATH) 4980 rpath_paths.push_back(path); 4981 else { 4982 if (path[0] == '@') { 4983 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 4984 rpath_relative_paths.push_back(path + strlen("@rpath")); 4985 else if (strncmp(path, "@executable_path", 4986 strlen("@executable_path")) == 0) 4987 at_exec_relative_paths.push_back(path 4988 + strlen("@executable_path")); 4989 } else { 4990 FileSpec file_spec(path, resolve_path); 4991 if (files.AppendIfUnique(file_spec)) 4992 count++; 4993 } 4994 } 4995 } 4996 } break; 4997 4998 default: 4999 break; 5000 } 5001 offset = cmd_offset + load_cmd.cmdsize; 5002 } 5003 5004 FileSpec this_file_spec(m_file); 5005 this_file_spec.ResolvePath(); 5006 5007 if (!rpath_paths.empty()) { 5008 // Fixup all LC_RPATH values to be absolute paths 5009 std::string loader_path("@loader_path"); 5010 std::string executable_path("@executable_path"); 5011 for (auto &rpath : rpath_paths) { 5012 if (rpath.find(loader_path) == 0) { 5013 rpath.erase(0, loader_path.size()); 5014 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5015 } else if (rpath.find(executable_path) == 0) { 5016 rpath.erase(0, executable_path.size()); 5017 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5018 } 5019 } 5020 5021 for (const auto &rpath_relative_path : rpath_relative_paths) { 5022 for (const auto &rpath : rpath_paths) { 5023 std::string path = rpath; 5024 path += rpath_relative_path; 5025 // It is OK to resolve this path because we must find a file on disk 5026 // for us to accept it anyway if it is rpath relative. 5027 FileSpec file_spec(path, true); 5028 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) { 5029 count++; 5030 break; 5031 } 5032 } 5033 } 5034 } 5035 5036 // We may have @executable_paths but no RPATHS. Figure those out here. 5037 // Only do this if this object file is the executable. We have no way to 5038 // get back to the actual executable otherwise, so we won't get the right 5039 // path. 5040 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5041 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5042 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5043 FileSpec file_spec = 5044 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5045 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) 5046 count++; 5047 } 5048 } 5049 } 5050 return count; 5051 } 5052 5053 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5054 // If the object file is not an executable it can't hold the entry point. 5055 // m_entry_point_address is initialized to an invalid address, so we can just 5056 // return that. If m_entry_point_address is valid it means we've found it 5057 // already, so return the cached value. 5058 5059 if (!IsExecutable() || m_entry_point_address.IsValid()) 5060 return m_entry_point_address; 5061 5062 // Otherwise, look for the UnixThread or Thread command. The data for the 5063 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5064 // 5065 // uint32_t flavor - this is the flavor argument you would pass to 5066 // thread_get_state 5067 // uint32_t count - this is the count of longs in the thread state data 5068 // struct XXX_thread_state state - this is the structure from 5069 // <machine/thread_status.h> corresponding to the flavor. 5070 // <repeat this trio> 5071 // 5072 // So we just keep reading the various register flavors till we find the GPR 5073 // one, then read the PC out of there. 5074 // FIXME: We will need to have a "RegisterContext data provider" class at some 5075 // point that can get all the registers 5076 // out of data in this form & attach them to a given thread. That should 5077 // underlie the MacOS X User process plugin, and we'll also need it for the 5078 // MacOS X Core File process plugin. When we have that we can also use it 5079 // here. 5080 // 5081 // For now we hard-code the offsets and flavors we need: 5082 // 5083 // 5084 5085 ModuleSP module_sp(GetModule()); 5086 if (module_sp) { 5087 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5088 struct load_command load_cmd; 5089 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5090 uint32_t i; 5091 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5092 bool done = false; 5093 5094 for (i = 0; i < m_header.ncmds; ++i) { 5095 const lldb::offset_t cmd_offset = offset; 5096 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5097 break; 5098 5099 switch (load_cmd.cmd) { 5100 case LC_UNIXTHREAD: 5101 case LC_THREAD: { 5102 while (offset < cmd_offset + load_cmd.cmdsize) { 5103 uint32_t flavor = m_data.GetU32(&offset); 5104 uint32_t count = m_data.GetU32(&offset); 5105 if (count == 0) { 5106 // We've gotten off somehow, log and exit; 5107 return m_entry_point_address; 5108 } 5109 5110 switch (m_header.cputype) { 5111 case llvm::MachO::CPU_TYPE_ARM: 5112 if (flavor == 1 || 5113 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 from 5114 // mach/arm/thread_status.h 5115 { 5116 offset += 60; // This is the offset of pc in the GPR thread state 5117 // data structure. 5118 start_address = m_data.GetU32(&offset); 5119 done = true; 5120 } 5121 break; 5122 case llvm::MachO::CPU_TYPE_ARM64: 5123 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5124 { 5125 offset += 256; // This is the offset of pc in the GPR thread state 5126 // data structure. 5127 start_address = m_data.GetU64(&offset); 5128 done = true; 5129 } 5130 break; 5131 case llvm::MachO::CPU_TYPE_I386: 5132 if (flavor == 5133 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5134 { 5135 offset += 40; // This is the offset of eip in the GPR thread state 5136 // data structure. 5137 start_address = m_data.GetU32(&offset); 5138 done = true; 5139 } 5140 break; 5141 case llvm::MachO::CPU_TYPE_X86_64: 5142 if (flavor == 5143 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5144 { 5145 offset += 16 * 8; // This is the offset of rip in the GPR thread 5146 // state data structure. 5147 start_address = m_data.GetU64(&offset); 5148 done = true; 5149 } 5150 break; 5151 default: 5152 return m_entry_point_address; 5153 } 5154 // Haven't found the GPR flavor yet, skip over the data for this 5155 // flavor: 5156 if (done) 5157 break; 5158 offset += count * 4; 5159 } 5160 } break; 5161 case LC_MAIN: { 5162 ConstString text_segment_name("__TEXT"); 5163 uint64_t entryoffset = m_data.GetU64(&offset); 5164 SectionSP text_segment_sp = 5165 GetSectionList()->FindSectionByName(text_segment_name); 5166 if (text_segment_sp) { 5167 done = true; 5168 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5169 } 5170 } break; 5171 5172 default: 5173 break; 5174 } 5175 if (done) 5176 break; 5177 5178 // Go to the next load command: 5179 offset = cmd_offset + load_cmd.cmdsize; 5180 } 5181 5182 if (start_address != LLDB_INVALID_ADDRESS) { 5183 // We got the start address from the load commands, so now resolve that 5184 // address in the sections of this ObjectFile: 5185 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5186 start_address, GetSectionList())) { 5187 m_entry_point_address.Clear(); 5188 } 5189 } else { 5190 // We couldn't read the UnixThread load command - maybe it wasn't there. 5191 // As a fallback look for the "start" symbol in the main executable. 5192 5193 ModuleSP module_sp(GetModule()); 5194 5195 if (module_sp) { 5196 SymbolContextList contexts; 5197 SymbolContext context; 5198 if (module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5199 eSymbolTypeCode, contexts)) { 5200 if (contexts.GetContextAtIndex(0, context)) 5201 m_entry_point_address = context.symbol->GetAddress(); 5202 } 5203 } 5204 } 5205 } 5206 5207 return m_entry_point_address; 5208 } 5209 5210 lldb_private::Address ObjectFileMachO::GetHeaderAddress() { 5211 lldb_private::Address header_addr; 5212 SectionList *section_list = GetSectionList(); 5213 if (section_list) { 5214 SectionSP text_segment_sp( 5215 section_list->FindSectionByName(GetSegmentNameTEXT())); 5216 if (text_segment_sp) { 5217 header_addr.SetSection(text_segment_sp); 5218 header_addr.SetOffset(0); 5219 } 5220 } 5221 return header_addr; 5222 } 5223 5224 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5225 ModuleSP module_sp(GetModule()); 5226 if (module_sp) { 5227 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5228 if (!m_thread_context_offsets_valid) { 5229 m_thread_context_offsets_valid = true; 5230 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5231 FileRangeArray::Entry file_range; 5232 thread_command thread_cmd; 5233 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5234 const uint32_t cmd_offset = offset; 5235 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 5236 break; 5237 5238 if (thread_cmd.cmd == LC_THREAD) { 5239 file_range.SetRangeBase(offset); 5240 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5241 m_thread_context_offsets.Append(file_range); 5242 } 5243 offset = cmd_offset + thread_cmd.cmdsize; 5244 } 5245 } 5246 } 5247 return m_thread_context_offsets.GetSize(); 5248 } 5249 5250 std::string ObjectFileMachO::GetIdentifierString() { 5251 std::string result; 5252 ModuleSP module_sp(GetModule()); 5253 if (module_sp) { 5254 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5255 5256 // First, look over the load commands for an LC_NOTE load command with 5257 // data_owner string "kern ver str" & use that if found. 5258 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5259 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5260 const uint32_t cmd_offset = offset; 5261 load_command lc; 5262 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5263 break; 5264 if (lc.cmd == LC_NOTE) 5265 { 5266 char data_owner[17]; 5267 m_data.CopyData (offset, 16, data_owner); 5268 data_owner[16] = '\0'; 5269 offset += 16; 5270 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5271 uint64_t size = m_data.GetU64_unchecked (&offset); 5272 5273 // "kern ver str" has a uint32_t version and then a nul terminated 5274 // c-string. 5275 if (strcmp ("kern ver str", data_owner) == 0) 5276 { 5277 offset = fileoff; 5278 uint32_t version; 5279 if (m_data.GetU32 (&offset, &version, 1) != nullptr) 5280 { 5281 if (version == 1) 5282 { 5283 uint32_t strsize = size - sizeof (uint32_t); 5284 char *buf = (char*) malloc (strsize); 5285 if (buf) 5286 { 5287 m_data.CopyData (offset, strsize, buf); 5288 buf[strsize - 1] = '\0'; 5289 result = buf; 5290 if (buf) 5291 free (buf); 5292 return result; 5293 } 5294 } 5295 } 5296 } 5297 } 5298 offset = cmd_offset + lc.cmdsize; 5299 } 5300 5301 // Second, make a pass over the load commands looking for an obsolete 5302 // LC_IDENT load command. 5303 offset = MachHeaderSizeFromMagic(m_header.magic); 5304 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5305 const uint32_t cmd_offset = offset; 5306 struct ident_command ident_command; 5307 if (m_data.GetU32(&offset, &ident_command, 2) == NULL) 5308 break; 5309 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5310 char *buf = (char *) malloc (ident_command.cmdsize); 5311 if (buf != nullptr 5312 && m_data.CopyData (offset, ident_command.cmdsize, buf) == ident_command.cmdsize) { 5313 buf[ident_command.cmdsize - 1] = '\0'; 5314 result = buf; 5315 } 5316 if (buf) 5317 free (buf); 5318 } 5319 offset = cmd_offset + ident_command.cmdsize; 5320 } 5321 5322 } 5323 return result; 5324 } 5325 5326 bool ObjectFileMachO::GetCorefileMainBinaryInfo (addr_t &address, UUID &uuid) { 5327 address = LLDB_INVALID_ADDRESS; 5328 uuid.Clear(); 5329 ModuleSP module_sp(GetModule()); 5330 if (module_sp) { 5331 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5332 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5333 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5334 const uint32_t cmd_offset = offset; 5335 load_command lc; 5336 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5337 break; 5338 if (lc.cmd == LC_NOTE) 5339 { 5340 char data_owner[17]; 5341 memset (data_owner, 0, sizeof (data_owner)); 5342 m_data.CopyData (offset, 16, data_owner); 5343 offset += 16; 5344 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5345 uint64_t size = m_data.GetU64_unchecked (&offset); 5346 5347 // "main bin spec" (main binary specification) data payload is 5348 // formatted: 5349 // uint32_t version [currently 1] 5350 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user process] 5351 // uint64_t address [ UINT64_MAX if address not specified ] 5352 // uuid_t uuid [ all zero's if uuid not specified ] 5353 // uint32_t log2_pagesize [ process page size in log base 2, e.g. 4k pages are 12. 0 for unspecified ] 5354 5355 if (strcmp ("main bin spec", data_owner) == 0 && size >= 32) 5356 { 5357 offset = fileoff; 5358 uint32_t version; 5359 if (m_data.GetU32 (&offset, &version, 1) != nullptr && version == 1) 5360 { 5361 uint32_t type = 0; 5362 uuid_t raw_uuid; 5363 memset (raw_uuid, 0, sizeof (uuid_t)); 5364 5365 if (m_data.GetU32 (&offset, &type, 1) 5366 && m_data.GetU64 (&offset, &address, 1) 5367 && m_data.CopyData (offset, sizeof (uuid_t), raw_uuid) != 0 5368 && uuid.SetBytes (raw_uuid, sizeof (uuid_t))) 5369 { 5370 return true; 5371 } 5372 } 5373 } 5374 } 5375 offset = cmd_offset + lc.cmdsize; 5376 } 5377 } 5378 return false; 5379 } 5380 5381 lldb::RegisterContextSP 5382 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5383 lldb_private::Thread &thread) { 5384 lldb::RegisterContextSP reg_ctx_sp; 5385 5386 ModuleSP module_sp(GetModule()); 5387 if (module_sp) { 5388 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5389 if (!m_thread_context_offsets_valid) 5390 GetNumThreadContexts(); 5391 5392 const FileRangeArray::Entry *thread_context_file_range = 5393 m_thread_context_offsets.GetEntryAtIndex(idx); 5394 if (thread_context_file_range) { 5395 5396 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5397 thread_context_file_range->GetByteSize()); 5398 5399 switch (m_header.cputype) { 5400 case llvm::MachO::CPU_TYPE_ARM64: 5401 reg_ctx_sp.reset(new RegisterContextDarwin_arm64_Mach(thread, data)); 5402 break; 5403 5404 case llvm::MachO::CPU_TYPE_ARM: 5405 reg_ctx_sp.reset(new RegisterContextDarwin_arm_Mach(thread, data)); 5406 break; 5407 5408 case llvm::MachO::CPU_TYPE_I386: 5409 reg_ctx_sp.reset(new RegisterContextDarwin_i386_Mach(thread, data)); 5410 break; 5411 5412 case llvm::MachO::CPU_TYPE_X86_64: 5413 reg_ctx_sp.reset(new RegisterContextDarwin_x86_64_Mach(thread, data)); 5414 break; 5415 } 5416 } 5417 } 5418 return reg_ctx_sp; 5419 } 5420 5421 ObjectFile::Type ObjectFileMachO::CalculateType() { 5422 switch (m_header.filetype) { 5423 case MH_OBJECT: // 0x1u 5424 if (GetAddressByteSize() == 4) { 5425 // 32 bit kexts are just object files, but they do have a valid 5426 // UUID load command. 5427 UUID uuid; 5428 if (GetUUID(&uuid)) { 5429 // this checking for the UUID load command is not enough we could 5430 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5431 // this is required of kexts 5432 if (m_strata == eStrataInvalid) 5433 m_strata = eStrataKernel; 5434 return eTypeSharedLibrary; 5435 } 5436 } 5437 return eTypeObjectFile; 5438 5439 case MH_EXECUTE: 5440 return eTypeExecutable; // 0x2u 5441 case MH_FVMLIB: 5442 return eTypeSharedLibrary; // 0x3u 5443 case MH_CORE: 5444 return eTypeCoreFile; // 0x4u 5445 case MH_PRELOAD: 5446 return eTypeSharedLibrary; // 0x5u 5447 case MH_DYLIB: 5448 return eTypeSharedLibrary; // 0x6u 5449 case MH_DYLINKER: 5450 return eTypeDynamicLinker; // 0x7u 5451 case MH_BUNDLE: 5452 return eTypeSharedLibrary; // 0x8u 5453 case MH_DYLIB_STUB: 5454 return eTypeStubLibrary; // 0x9u 5455 case MH_DSYM: 5456 return eTypeDebugInfo; // 0xAu 5457 case MH_KEXT_BUNDLE: 5458 return eTypeSharedLibrary; // 0xBu 5459 default: 5460 break; 5461 } 5462 return eTypeUnknown; 5463 } 5464 5465 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5466 switch (m_header.filetype) { 5467 case MH_OBJECT: // 0x1u 5468 { 5469 // 32 bit kexts are just object files, but they do have a valid 5470 // UUID load command. 5471 UUID uuid; 5472 if (GetUUID(&uuid)) { 5473 // this checking for the UUID load command is not enough we could 5474 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5475 // this is required of kexts 5476 if (m_type == eTypeInvalid) 5477 m_type = eTypeSharedLibrary; 5478 5479 return eStrataKernel; 5480 } 5481 } 5482 return eStrataUnknown; 5483 5484 case MH_EXECUTE: // 0x2u 5485 // Check for the MH_DYLDLINK bit in the flags 5486 if (m_header.flags & MH_DYLDLINK) { 5487 return eStrataUser; 5488 } else { 5489 SectionList *section_list = GetSectionList(); 5490 if (section_list) { 5491 static ConstString g_kld_section_name("__KLD"); 5492 if (section_list->FindSectionByName(g_kld_section_name)) 5493 return eStrataKernel; 5494 } 5495 } 5496 return eStrataRawImage; 5497 5498 case MH_FVMLIB: 5499 return eStrataUser; // 0x3u 5500 case MH_CORE: 5501 return eStrataUnknown; // 0x4u 5502 case MH_PRELOAD: 5503 return eStrataRawImage; // 0x5u 5504 case MH_DYLIB: 5505 return eStrataUser; // 0x6u 5506 case MH_DYLINKER: 5507 return eStrataUser; // 0x7u 5508 case MH_BUNDLE: 5509 return eStrataUser; // 0x8u 5510 case MH_DYLIB_STUB: 5511 return eStrataUser; // 0x9u 5512 case MH_DSYM: 5513 return eStrataUnknown; // 0xAu 5514 case MH_KEXT_BUNDLE: 5515 return eStrataKernel; // 0xBu 5516 default: 5517 break; 5518 } 5519 return eStrataUnknown; 5520 } 5521 5522 uint32_t ObjectFileMachO::GetVersion(uint32_t *versions, 5523 uint32_t num_versions) { 5524 ModuleSP module_sp(GetModule()); 5525 if (module_sp) { 5526 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5527 struct dylib_command load_cmd; 5528 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5529 uint32_t version_cmd = 0; 5530 uint64_t version = 0; 5531 uint32_t i; 5532 for (i = 0; i < m_header.ncmds; ++i) { 5533 const lldb::offset_t cmd_offset = offset; 5534 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5535 break; 5536 5537 if (load_cmd.cmd == LC_ID_DYLIB) { 5538 if (version_cmd == 0) { 5539 version_cmd = load_cmd.cmd; 5540 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5541 break; 5542 version = load_cmd.dylib.current_version; 5543 } 5544 break; // Break for now unless there is another more complete version 5545 // number load command in the future. 5546 } 5547 offset = cmd_offset + load_cmd.cmdsize; 5548 } 5549 5550 if (version_cmd == LC_ID_DYLIB) { 5551 if (versions != NULL && num_versions > 0) { 5552 if (num_versions > 0) 5553 versions[0] = (version & 0xFFFF0000ull) >> 16; 5554 if (num_versions > 1) 5555 versions[1] = (version & 0x0000FF00ull) >> 8; 5556 if (num_versions > 2) 5557 versions[2] = (version & 0x000000FFull); 5558 // Fill in an remaining version numbers with invalid values 5559 for (i = 3; i < num_versions; ++i) 5560 versions[i] = UINT32_MAX; 5561 } 5562 // The LC_ID_DYLIB load command has a version with 3 version numbers in 5563 // it, so always return 3 5564 return 3; 5565 } 5566 } 5567 return false; 5568 } 5569 5570 bool ObjectFileMachO::GetArchitecture(ArchSpec &arch) { 5571 ModuleSP module_sp(GetModule()); 5572 if (module_sp) { 5573 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5574 return GetArchitecture(m_header, m_data, 5575 MachHeaderSizeFromMagic(m_header.magic), arch); 5576 } 5577 return false; 5578 } 5579 5580 UUID ObjectFileMachO::GetProcessSharedCacheUUID(Process *process) { 5581 UUID uuid; 5582 if (process && process->GetDynamicLoader()) { 5583 DynamicLoader *dl = process->GetDynamicLoader(); 5584 addr_t load_address; 5585 LazyBool using_shared_cache; 5586 LazyBool private_shared_cache; 5587 dl->GetSharedCacheInformation(load_address, uuid, using_shared_cache, 5588 private_shared_cache); 5589 } 5590 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5591 if (log) 5592 log->Printf("inferior process shared cache has a UUID of %s", uuid.GetAsString().c_str()); 5593 return uuid; 5594 } 5595 5596 UUID ObjectFileMachO::GetLLDBSharedCacheUUID() { 5597 UUID uuid; 5598 #if defined(__APPLE__) && \ 5599 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5600 uint8_t *(*dyld_get_all_image_infos)(void); 5601 dyld_get_all_image_infos = 5602 (uint8_t * (*)())dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5603 if (dyld_get_all_image_infos) { 5604 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5605 if (dyld_all_image_infos_address) { 5606 uint32_t *version = (uint32_t *) 5607 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5608 if (*version >= 13) { 5609 uuid_t *sharedCacheUUID_address = 0; 5610 int wordsize = sizeof(uint8_t *); 5611 if (wordsize == 8) { 5612 sharedCacheUUID_address = 5613 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5614 160); // sharedCacheUUID <mach-o/dyld_images.h> 5615 } else { 5616 sharedCacheUUID_address = 5617 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5618 84); // sharedCacheUUID <mach-o/dyld_images.h> 5619 } 5620 uuid.SetBytes(sharedCacheUUID_address); 5621 } 5622 } 5623 } else { 5624 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5625 bool *(*dyld_get_shared_cache_uuid)(uuid_t); 5626 dyld_get_shared_cache_uuid = (bool *(*)(uuid_t)) 5627 dlsym(RTLD_DEFAULT, "_dyld_get_shared_cache_uuid"); 5628 if (dyld_get_shared_cache_uuid) { 5629 uuid_t tmp_uuid; 5630 if (dyld_get_shared_cache_uuid(tmp_uuid)) 5631 uuid.SetBytes(tmp_uuid); 5632 } 5633 } 5634 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5635 if (log && uuid.IsValid()) 5636 log->Printf("lldb's in-memory shared cache has a UUID of %s", uuid.GetAsString().c_str()); 5637 #endif 5638 return uuid; 5639 } 5640 5641 uint32_t ObjectFileMachO::GetMinimumOSVersion(uint32_t *versions, 5642 uint32_t num_versions) { 5643 if (m_min_os_versions.empty()) { 5644 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5645 bool success = false; 5646 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5647 const lldb::offset_t load_cmd_offset = offset; 5648 5649 version_min_command lc; 5650 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5651 break; 5652 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5653 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5654 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5655 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5656 if (m_data.GetU32(&offset, &lc.version, 5657 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5658 const uint32_t xxxx = lc.version >> 16; 5659 const uint32_t yy = (lc.version >> 8) & 0xffu; 5660 const uint32_t zz = lc.version & 0xffu; 5661 if (xxxx) { 5662 m_min_os_versions.push_back(xxxx); 5663 m_min_os_versions.push_back(yy); 5664 m_min_os_versions.push_back(zz); 5665 } 5666 success = true; 5667 } 5668 } 5669 offset = load_cmd_offset + lc.cmdsize; 5670 } 5671 5672 if (success == false) { 5673 // Push an invalid value so we don't keep trying to 5674 m_min_os_versions.push_back(UINT32_MAX); 5675 } 5676 } 5677 5678 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) { 5679 if (versions != NULL && num_versions > 0) { 5680 for (size_t i = 0; i < num_versions; ++i) { 5681 if (i < m_min_os_versions.size()) 5682 versions[i] = m_min_os_versions[i]; 5683 else 5684 versions[i] = 0; 5685 } 5686 } 5687 return m_min_os_versions.size(); 5688 } 5689 // Call the superclasses version that will empty out the data 5690 return ObjectFile::GetMinimumOSVersion(versions, num_versions); 5691 } 5692 5693 uint32_t ObjectFileMachO::GetSDKVersion(uint32_t *versions, 5694 uint32_t num_versions) { 5695 if (m_sdk_versions.empty()) { 5696 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5697 bool success = false; 5698 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5699 const lldb::offset_t load_cmd_offset = offset; 5700 5701 version_min_command lc; 5702 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5703 break; 5704 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5705 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5706 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5707 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5708 if (m_data.GetU32(&offset, &lc.version, 5709 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5710 const uint32_t xxxx = lc.sdk >> 16; 5711 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5712 const uint32_t zz = lc.sdk & 0xffu; 5713 if (xxxx) { 5714 m_sdk_versions.push_back(xxxx); 5715 m_sdk_versions.push_back(yy); 5716 m_sdk_versions.push_back(zz); 5717 } 5718 success = true; 5719 } 5720 } 5721 offset = load_cmd_offset + lc.cmdsize; 5722 } 5723 5724 if (success == false) { 5725 // Push an invalid value so we don't keep trying to 5726 m_sdk_versions.push_back(UINT32_MAX); 5727 } 5728 } 5729 5730 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) { 5731 if (versions != NULL && num_versions > 0) { 5732 for (size_t i = 0; i < num_versions; ++i) { 5733 if (i < m_sdk_versions.size()) 5734 versions[i] = m_sdk_versions[i]; 5735 else 5736 versions[i] = 0; 5737 } 5738 } 5739 return m_sdk_versions.size(); 5740 } 5741 // Call the superclasses version that will empty out the data 5742 return ObjectFile::GetSDKVersion(versions, num_versions); 5743 } 5744 5745 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5746 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5747 } 5748 5749 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5750 return m_allow_assembly_emulation_unwind_plans; 5751 } 5752 5753 //------------------------------------------------------------------ 5754 // PluginInterface protocol 5755 //------------------------------------------------------------------ 5756 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5757 return GetPluginNameStatic(); 5758 } 5759 5760 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5761 5762 Section *ObjectFileMachO::GetMachHeaderSection() { 5763 // Find the first address of the mach header which is the first non-zero file 5764 // sized section whose file offset is zero. This is the base file address of 5765 // the mach-o file which can be subtracted from the vmaddr of the other 5766 // segments found in memory and added to the load address 5767 ModuleSP module_sp = GetModule(); 5768 if (module_sp) { 5769 SectionList *section_list = GetSectionList(); 5770 if (section_list) { 5771 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5772 const size_t num_sections = section_list->GetSize(); 5773 5774 for (size_t sect_idx = 0; sect_idx < num_sections && 5775 mach_base_file_addr == LLDB_INVALID_ADDRESS; 5776 ++sect_idx) { 5777 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5778 if (section && section->GetFileSize() > 0 && 5779 section->GetFileOffset() == 0 && 5780 section->IsThreadSpecific() == false && 5781 module_sp.get() == section->GetModule().get()) { 5782 return section; 5783 } 5784 } 5785 } 5786 } 5787 return nullptr; 5788 } 5789 5790 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 5791 lldb::addr_t mach_header_load_address, const Section *mach_header_section, 5792 const Section *section) { 5793 ModuleSP module_sp = GetModule(); 5794 if (module_sp && mach_header_section && section && 5795 mach_header_load_address != LLDB_INVALID_ADDRESS) { 5796 lldb::addr_t mach_header_file_addr = mach_header_section->GetFileAddress(); 5797 if (mach_header_file_addr != LLDB_INVALID_ADDRESS) { 5798 if (section && section->GetFileSize() > 0 && 5799 section->IsThreadSpecific() == false && 5800 module_sp.get() == section->GetModule().get()) { 5801 // Ignore __LINKEDIT and __DWARF segments 5802 if (section->GetName() == GetSegmentNameLINKEDIT()) { 5803 // Only map __LINKEDIT if we have an in memory image and this isn't a 5804 // kernel binary like a kext or mach_kernel. 5805 const bool is_memory_image = (bool)m_process_wp.lock(); 5806 const Strata strata = GetStrata(); 5807 if (is_memory_image == false || strata == eStrataKernel) 5808 return LLDB_INVALID_ADDRESS; 5809 } 5810 return section->GetFileAddress() - mach_header_file_addr + 5811 mach_header_load_address; 5812 } 5813 } 5814 } 5815 return LLDB_INVALID_ADDRESS; 5816 } 5817 5818 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 5819 bool value_is_offset) { 5820 ModuleSP module_sp = GetModule(); 5821 if (module_sp) { 5822 size_t num_loaded_sections = 0; 5823 SectionList *section_list = GetSectionList(); 5824 if (section_list) { 5825 const size_t num_sections = section_list->GetSize(); 5826 5827 if (value_is_offset) { 5828 // "value" is an offset to apply to each top level segment 5829 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5830 // Iterate through the object file sections to find all of the 5831 // sections that size on disk (to avoid __PAGEZERO) and load them 5832 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5833 if (section_sp && section_sp->GetFileSize() > 0 && 5834 section_sp->IsThreadSpecific() == false && 5835 module_sp.get() == section_sp->GetModule().get()) { 5836 // Ignore __LINKEDIT and __DWARF segments 5837 if (section_sp->GetName() == GetSegmentNameLINKEDIT()) { 5838 // Only map __LINKEDIT if we have an in memory image and this 5839 // isn't a kernel binary like a kext or mach_kernel. 5840 const bool is_memory_image = (bool)m_process_wp.lock(); 5841 const Strata strata = GetStrata(); 5842 if (is_memory_image == false || strata == eStrataKernel) 5843 continue; 5844 } 5845 if (target.GetSectionLoadList().SetSectionLoadAddress( 5846 section_sp, section_sp->GetFileAddress() + value)) 5847 ++num_loaded_sections; 5848 } 5849 } 5850 } else { 5851 // "value" is the new base address of the mach_header, adjust each 5852 // section accordingly 5853 5854 Section *mach_header_section = GetMachHeaderSection(); 5855 if (mach_header_section) { 5856 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5857 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5858 5859 lldb::addr_t section_load_addr = 5860 CalculateSectionLoadAddressForMemoryImage( 5861 value, mach_header_section, section_sp.get()); 5862 if (section_load_addr != LLDB_INVALID_ADDRESS) { 5863 if (target.GetSectionLoadList().SetSectionLoadAddress( 5864 section_sp, section_load_addr)) 5865 ++num_loaded_sections; 5866 } 5867 } 5868 } 5869 } 5870 } 5871 return num_loaded_sections > 0; 5872 } 5873 return false; 5874 } 5875 5876 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 5877 const FileSpec &outfile, Status &error) { 5878 if (process_sp) { 5879 Target &target = process_sp->GetTarget(); 5880 const ArchSpec target_arch = target.GetArchitecture(); 5881 const llvm::Triple &target_triple = target_arch.GetTriple(); 5882 if (target_triple.getVendor() == llvm::Triple::Apple && 5883 (target_triple.getOS() == llvm::Triple::MacOSX || 5884 target_triple.getOS() == llvm::Triple::IOS || 5885 target_triple.getOS() == llvm::Triple::WatchOS || 5886 target_triple.getOS() == llvm::Triple::TvOS)) { 5887 bool make_core = false; 5888 switch (target_arch.GetMachine()) { 5889 case llvm::Triple::aarch64: 5890 case llvm::Triple::arm: 5891 case llvm::Triple::thumb: 5892 case llvm::Triple::x86: 5893 case llvm::Triple::x86_64: 5894 make_core = true; 5895 break; 5896 default: 5897 error.SetErrorStringWithFormat("unsupported core architecture: %s", 5898 target_triple.str().c_str()); 5899 break; 5900 } 5901 5902 if (make_core) { 5903 std::vector<segment_command_64> segment_load_commands; 5904 // uint32_t range_info_idx = 0; 5905 MemoryRegionInfo range_info; 5906 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 5907 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 5908 const ByteOrder byte_order = target_arch.GetByteOrder(); 5909 if (range_error.Success()) { 5910 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 5911 const addr_t addr = range_info.GetRange().GetRangeBase(); 5912 const addr_t size = range_info.GetRange().GetByteSize(); 5913 5914 if (size == 0) 5915 break; 5916 5917 // Calculate correct protections 5918 uint32_t prot = 0; 5919 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 5920 prot |= VM_PROT_READ; 5921 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 5922 prot |= VM_PROT_WRITE; 5923 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 5924 prot |= VM_PROT_EXECUTE; 5925 5926 // printf ("[%3u] [0x%16.16" PRIx64 " - 5927 // 0x%16.16" PRIx64 ") %c%c%c\n", 5928 // range_info_idx, 5929 // addr, 5930 // size, 5931 // (prot & VM_PROT_READ ) ? 'r' : 5932 // '-', 5933 // (prot & VM_PROT_WRITE ) ? 'w' : 5934 // '-', 5935 // (prot & VM_PROT_EXECUTE) ? 'x' : 5936 // '-'); 5937 5938 if (prot != 0) { 5939 uint32_t cmd_type = LC_SEGMENT_64; 5940 uint32_t segment_size = sizeof(segment_command_64); 5941 if (addr_byte_size == 4) { 5942 cmd_type = LC_SEGMENT; 5943 segment_size = sizeof(segment_command); 5944 } 5945 segment_command_64 segment = { 5946 cmd_type, // uint32_t cmd; 5947 segment_size, // uint32_t cmdsize; 5948 {0}, // char segname[16]; 5949 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 5950 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 5951 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 5952 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 5953 prot, // uint32_t maxprot; 5954 prot, // uint32_t initprot; 5955 0, // uint32_t nsects; 5956 0}; // uint32_t flags; 5957 segment_load_commands.push_back(segment); 5958 } else { 5959 // No protections and a size of 1 used to be returned from old 5960 // debugservers when we asked about a region that was past the 5961 // last memory region and it indicates the end... 5962 if (size == 1) 5963 break; 5964 } 5965 5966 range_error = process_sp->GetMemoryRegionInfo( 5967 range_info.GetRange().GetRangeEnd(), range_info); 5968 if (range_error.Fail()) 5969 break; 5970 } 5971 5972 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 5973 5974 mach_header_64 mach_header; 5975 if (addr_byte_size == 8) { 5976 mach_header.magic = MH_MAGIC_64; 5977 } else { 5978 mach_header.magic = MH_MAGIC; 5979 } 5980 mach_header.cputype = target_arch.GetMachOCPUType(); 5981 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 5982 mach_header.filetype = MH_CORE; 5983 mach_header.ncmds = segment_load_commands.size(); 5984 mach_header.flags = 0; 5985 mach_header.reserved = 0; 5986 ThreadList &thread_list = process_sp->GetThreadList(); 5987 const uint32_t num_threads = thread_list.GetSize(); 5988 5989 // Make an array of LC_THREAD data items. Each one contains the 5990 // contents of the LC_THREAD load command. The data doesn't contain 5991 // the load command + load command size, we will add the load command 5992 // and load command size as we emit the data. 5993 std::vector<StreamString> LC_THREAD_datas(num_threads); 5994 for (auto &LC_THREAD_data : LC_THREAD_datas) { 5995 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 5996 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 5997 LC_THREAD_data.SetByteOrder(byte_order); 5998 } 5999 for (uint32_t thread_idx = 0; thread_idx < num_threads; 6000 ++thread_idx) { 6001 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6002 if (thread_sp) { 6003 switch (mach_header.cputype) { 6004 case llvm::MachO::CPU_TYPE_ARM64: 6005 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6006 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6007 break; 6008 6009 case llvm::MachO::CPU_TYPE_ARM: 6010 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6011 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6012 break; 6013 6014 case llvm::MachO::CPU_TYPE_I386: 6015 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6016 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6017 break; 6018 6019 case llvm::MachO::CPU_TYPE_X86_64: 6020 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6021 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6022 break; 6023 } 6024 } 6025 } 6026 6027 // The size of the load command is the size of the segments... 6028 if (addr_byte_size == 8) { 6029 mach_header.sizeofcmds = segment_load_commands.size() * 6030 sizeof(struct segment_command_64); 6031 } else { 6032 mach_header.sizeofcmds = 6033 segment_load_commands.size() * sizeof(struct segment_command); 6034 } 6035 6036 // and the size of all LC_THREAD load command 6037 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6038 ++mach_header.ncmds; 6039 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6040 } 6041 6042 printf("mach_header: 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x " 6043 "0x%8.8x 0x%8.8x\n", 6044 mach_header.magic, mach_header.cputype, mach_header.cpusubtype, 6045 mach_header.filetype, mach_header.ncmds, 6046 mach_header.sizeofcmds, mach_header.flags, 6047 mach_header.reserved); 6048 6049 // Write the mach header 6050 buffer.PutHex32(mach_header.magic); 6051 buffer.PutHex32(mach_header.cputype); 6052 buffer.PutHex32(mach_header.cpusubtype); 6053 buffer.PutHex32(mach_header.filetype); 6054 buffer.PutHex32(mach_header.ncmds); 6055 buffer.PutHex32(mach_header.sizeofcmds); 6056 buffer.PutHex32(mach_header.flags); 6057 if (addr_byte_size == 8) { 6058 buffer.PutHex32(mach_header.reserved); 6059 } 6060 6061 // Skip the mach header and all load commands and align to the next 6062 // 0x1000 byte boundary 6063 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6064 if (file_offset & 0x00000fff) { 6065 file_offset += 0x00001000ull; 6066 file_offset &= (~0x00001000ull + 1); 6067 } 6068 6069 for (auto &segment : segment_load_commands) { 6070 segment.fileoff = file_offset; 6071 file_offset += segment.filesize; 6072 } 6073 6074 // Write out all of the LC_THREAD load commands 6075 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6076 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6077 buffer.PutHex32(LC_THREAD); 6078 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6079 buffer.Write(LC_THREAD_data.GetString().data(), 6080 LC_THREAD_data_size); 6081 } 6082 6083 // Write out all of the segment load commands 6084 for (const auto &segment : segment_load_commands) { 6085 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6086 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6087 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6088 segment.cmd, segment.cmdsize, segment.vmaddr, 6089 segment.vmaddr + segment.vmsize, segment.fileoff, 6090 segment.filesize, segment.maxprot, segment.initprot, 6091 segment.nsects, segment.flags); 6092 6093 buffer.PutHex32(segment.cmd); 6094 buffer.PutHex32(segment.cmdsize); 6095 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6096 if (addr_byte_size == 8) { 6097 buffer.PutHex64(segment.vmaddr); 6098 buffer.PutHex64(segment.vmsize); 6099 buffer.PutHex64(segment.fileoff); 6100 buffer.PutHex64(segment.filesize); 6101 } else { 6102 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6103 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6104 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6105 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6106 } 6107 buffer.PutHex32(segment.maxprot); 6108 buffer.PutHex32(segment.initprot); 6109 buffer.PutHex32(segment.nsects); 6110 buffer.PutHex32(segment.flags); 6111 } 6112 6113 File core_file; 6114 std::string core_file_path(outfile.GetPath()); 6115 error = core_file.Open(core_file_path.c_str(), 6116 File::eOpenOptionWrite | 6117 File::eOpenOptionTruncate | 6118 File::eOpenOptionCanCreate); 6119 if (error.Success()) { 6120 // Read 1 page at a time 6121 uint8_t bytes[0x1000]; 6122 // Write the mach header and load commands out to the core file 6123 size_t bytes_written = buffer.GetString().size(); 6124 error = core_file.Write(buffer.GetString().data(), bytes_written); 6125 if (error.Success()) { 6126 // Now write the file data for all memory segments in the process 6127 for (const auto &segment : segment_load_commands) { 6128 if (core_file.SeekFromStart(segment.fileoff) == -1) { 6129 error.SetErrorStringWithFormat( 6130 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6131 segment.fileoff, core_file_path.c_str()); 6132 break; 6133 } 6134 6135 printf("Saving %" PRId64 6136 " bytes of data for memory region at 0x%" PRIx64 "\n", 6137 segment.vmsize, segment.vmaddr); 6138 addr_t bytes_left = segment.vmsize; 6139 addr_t addr = segment.vmaddr; 6140 Status memory_read_error; 6141 while (bytes_left > 0 && error.Success()) { 6142 const size_t bytes_to_read = 6143 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6144 const size_t bytes_read = process_sp->ReadMemory( 6145 addr, bytes, bytes_to_read, memory_read_error); 6146 if (bytes_read == bytes_to_read) { 6147 size_t bytes_written = bytes_read; 6148 error = core_file.Write(bytes, bytes_written); 6149 bytes_left -= bytes_read; 6150 addr += bytes_read; 6151 } else { 6152 // Some pages within regions are not readable, those should 6153 // be zero filled 6154 memset(bytes, 0, bytes_to_read); 6155 size_t bytes_written = bytes_to_read; 6156 error = core_file.Write(bytes, bytes_written); 6157 bytes_left -= bytes_to_read; 6158 addr += bytes_to_read; 6159 } 6160 } 6161 } 6162 } 6163 } 6164 } else { 6165 error.SetErrorString( 6166 "process doesn't support getting memory region info"); 6167 } 6168 } 6169 return true; // This is the right plug to handle saving core files for 6170 // this process 6171 } 6172 } 6173 return false; 6174 } 6175