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/Section.h" 27 #include "lldb/Core/StreamFile.h" 28 #include "lldb/Host/Host.h" 29 #include "lldb/Symbol/DWARFCallFrameInfo.h" 30 #include "lldb/Symbol/ObjectFile.h" 31 #include "lldb/Target/DynamicLoader.h" 32 #include "lldb/Target/MemoryRegionInfo.h" 33 #include "lldb/Target/Platform.h" 34 #include "lldb/Target/Process.h" 35 #include "lldb/Target/SectionLoadList.h" 36 #include "lldb/Target/Target.h" 37 #include "lldb/Target/Thread.h" 38 #include "lldb/Target/ThreadList.h" 39 #include "lldb/Utility/ArchSpec.h" 40 #include "lldb/Utility/DataBuffer.h" 41 #include "lldb/Utility/FileSpec.h" 42 #include "lldb/Utility/Log.h" 43 #include "lldb/Utility/RegisterValue.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 AddressClass::eUnknown; 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 AddressClass::eCodeAlternateISA; 1172 } 1173 return AddressClass::eCode; 1174 1175 case eSectionTypeContainer: 1176 return AddressClass::eUnknown; 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 AddressClass::eData; 1191 1192 case eSectionTypeDebug: 1193 case eSectionTypeDWARFDebugAbbrev: 1194 case eSectionTypeDWARFDebugAddr: 1195 case eSectionTypeDWARFDebugAranges: 1196 case eSectionTypeDWARFDebugCuIndex: 1197 case eSectionTypeDWARFDebugFrame: 1198 case eSectionTypeDWARFDebugInfo: 1199 case eSectionTypeDWARFDebugLine: 1200 case eSectionTypeDWARFDebugLoc: 1201 case eSectionTypeDWARFDebugMacInfo: 1202 case eSectionTypeDWARFDebugMacro: 1203 case eSectionTypeDWARFDebugNames: 1204 case eSectionTypeDWARFDebugPubNames: 1205 case eSectionTypeDWARFDebugPubTypes: 1206 case eSectionTypeDWARFDebugRanges: 1207 case eSectionTypeDWARFDebugStr: 1208 case eSectionTypeDWARFDebugStrOffsets: 1209 case eSectionTypeDWARFDebugTypes: 1210 case eSectionTypeDWARFAppleNames: 1211 case eSectionTypeDWARFAppleTypes: 1212 case eSectionTypeDWARFAppleNamespaces: 1213 case eSectionTypeDWARFAppleObjC: 1214 case eSectionTypeDWARFGNUDebugAltLink: 1215 return AddressClass::eDebug; 1216 1217 case eSectionTypeEHFrame: 1218 case eSectionTypeARMexidx: 1219 case eSectionTypeARMextab: 1220 case eSectionTypeCompactUnwind: 1221 return AddressClass::eRuntime; 1222 1223 case eSectionTypeAbsoluteAddress: 1224 case eSectionTypeELFSymbolTable: 1225 case eSectionTypeELFDynamicSymbols: 1226 case eSectionTypeELFRelocationEntries: 1227 case eSectionTypeELFDynamicLinkInfo: 1228 case eSectionTypeOther: 1229 return AddressClass::eUnknown; 1230 } 1231 } 1232 } 1233 1234 const SymbolType symbol_type = symbol->GetType(); 1235 switch (symbol_type) { 1236 case eSymbolTypeAny: 1237 return AddressClass::eUnknown; 1238 case eSymbolTypeAbsolute: 1239 return AddressClass::eUnknown; 1240 1241 case eSymbolTypeCode: 1242 case eSymbolTypeTrampoline: 1243 case eSymbolTypeResolver: 1244 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1245 // For ARM we have a bit in the n_desc field of the symbol that tells 1246 // us ARM/Thumb which is bit 0x0008. 1247 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1248 return AddressClass::eCodeAlternateISA; 1249 } 1250 return AddressClass::eCode; 1251 1252 case eSymbolTypeData: 1253 return AddressClass::eData; 1254 case eSymbolTypeRuntime: 1255 return AddressClass::eRuntime; 1256 case eSymbolTypeException: 1257 return AddressClass::eRuntime; 1258 case eSymbolTypeSourceFile: 1259 return AddressClass::eDebug; 1260 case eSymbolTypeHeaderFile: 1261 return AddressClass::eDebug; 1262 case eSymbolTypeObjectFile: 1263 return AddressClass::eDebug; 1264 case eSymbolTypeCommonBlock: 1265 return AddressClass::eDebug; 1266 case eSymbolTypeBlock: 1267 return AddressClass::eDebug; 1268 case eSymbolTypeLocal: 1269 return AddressClass::eData; 1270 case eSymbolTypeParam: 1271 return AddressClass::eData; 1272 case eSymbolTypeVariable: 1273 return AddressClass::eData; 1274 case eSymbolTypeVariableType: 1275 return AddressClass::eDebug; 1276 case eSymbolTypeLineEntry: 1277 return AddressClass::eDebug; 1278 case eSymbolTypeLineHeader: 1279 return AddressClass::eDebug; 1280 case eSymbolTypeScopeBegin: 1281 return AddressClass::eDebug; 1282 case eSymbolTypeScopeEnd: 1283 return AddressClass::eDebug; 1284 case eSymbolTypeAdditional: 1285 return AddressClass::eUnknown; 1286 case eSymbolTypeCompiler: 1287 return AddressClass::eDebug; 1288 case eSymbolTypeInstrumentation: 1289 return AddressClass::eDebug; 1290 case eSymbolTypeUndefined: 1291 return AddressClass::eUnknown; 1292 case eSymbolTypeObjCClass: 1293 return AddressClass::eRuntime; 1294 case eSymbolTypeObjCMetaClass: 1295 return AddressClass::eRuntime; 1296 case eSymbolTypeObjCIVar: 1297 return AddressClass::eRuntime; 1298 case eSymbolTypeReExported: 1299 return AddressClass::eRuntime; 1300 } 1301 } 1302 } 1303 return AddressClass::eUnknown; 1304 } 1305 1306 Symtab *ObjectFileMachO::GetSymtab() { 1307 ModuleSP module_sp(GetModule()); 1308 if (module_sp) { 1309 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1310 if (m_symtab_ap.get() == NULL) { 1311 m_symtab_ap.reset(new Symtab(this)); 1312 std::lock_guard<std::recursive_mutex> symtab_guard( 1313 m_symtab_ap->GetMutex()); 1314 ParseSymtab(); 1315 m_symtab_ap->Finalize(); 1316 } 1317 } 1318 return m_symtab_ap.get(); 1319 } 1320 1321 bool ObjectFileMachO::IsStripped() { 1322 if (m_dysymtab.cmd == 0) { 1323 ModuleSP module_sp(GetModule()); 1324 if (module_sp) { 1325 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1326 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1327 const lldb::offset_t load_cmd_offset = offset; 1328 1329 load_command lc; 1330 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1331 break; 1332 if (lc.cmd == LC_DYSYMTAB) { 1333 m_dysymtab.cmd = lc.cmd; 1334 m_dysymtab.cmdsize = lc.cmdsize; 1335 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1336 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1337 NULL) { 1338 // Clear m_dysymtab if we were unable to read all items from the 1339 // load command 1340 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1341 } 1342 } 1343 offset = load_cmd_offset + lc.cmdsize; 1344 } 1345 } 1346 } 1347 if (m_dysymtab.cmd) 1348 return m_dysymtab.nlocalsym <= 1; 1349 return false; 1350 } 1351 1352 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1353 EncryptedFileRanges result; 1354 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1355 1356 encryption_info_command encryption_cmd; 1357 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1358 const lldb::offset_t load_cmd_offset = offset; 1359 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1360 break; 1361 1362 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the 1363 // 3 fields we care about, so treat them the same. 1364 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1365 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1366 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1367 if (encryption_cmd.cryptid != 0) { 1368 EncryptedFileRanges::Entry entry; 1369 entry.SetRangeBase(encryption_cmd.cryptoff); 1370 entry.SetByteSize(encryption_cmd.cryptsize); 1371 result.Append(entry); 1372 } 1373 } 1374 } 1375 offset = load_cmd_offset + encryption_cmd.cmdsize; 1376 } 1377 1378 return result; 1379 } 1380 1381 void ObjectFileMachO::SanitizeSegmentCommand(segment_command_64 &seg_cmd, 1382 uint32_t cmd_idx) { 1383 if (m_length == 0 || seg_cmd.filesize == 0) 1384 return; 1385 1386 if (seg_cmd.fileoff > m_length) { 1387 // We have a load command that says it extends past the end of the file. 1388 // This is likely a corrupt file. We don't have any way to return an error 1389 // condition here (this method was likely invoked from something like 1390 // ObjectFile::GetSectionList()), so we just null out the section contents, 1391 // and dump a message to stdout. The most common case here is core file 1392 // debugging with a truncated file. 1393 const char *lc_segment_name = 1394 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1395 GetModule()->ReportWarning( 1396 "load command %u %s has a fileoff (0x%" PRIx64 1397 ") that extends beyond the end of the file (0x%" PRIx64 1398 "), ignoring this section", 1399 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1400 1401 seg_cmd.fileoff = 0; 1402 seg_cmd.filesize = 0; 1403 } 1404 1405 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1406 // We have a load command that says it extends past the end of the file. 1407 // This is likely a corrupt file. We don't have any way to return an error 1408 // condition here (this method was likely invoked from something like 1409 // ObjectFile::GetSectionList()), so we just null out the section contents, 1410 // and dump a message to stdout. The most common case here is core file 1411 // debugging with a truncated file. 1412 const char *lc_segment_name = 1413 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1414 GetModule()->ReportWarning( 1415 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1416 ") that extends beyond the end of the file (0x%" PRIx64 1417 "), the segment will be truncated to match", 1418 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1419 1420 // Truncate the length 1421 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1422 } 1423 } 1424 1425 static uint32_t GetSegmentPermissions(const segment_command_64 &seg_cmd) { 1426 uint32_t result = 0; 1427 if (seg_cmd.initprot & VM_PROT_READ) 1428 result |= ePermissionsReadable; 1429 if (seg_cmd.initprot & VM_PROT_WRITE) 1430 result |= ePermissionsWritable; 1431 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1432 result |= ePermissionsExecutable; 1433 return result; 1434 } 1435 1436 static lldb::SectionType GetSectionType(uint32_t flags, 1437 ConstString section_name) { 1438 1439 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1440 return eSectionTypeCode; 1441 1442 uint32_t mach_sect_type = flags & SECTION_TYPE; 1443 static ConstString g_sect_name_objc_data("__objc_data"); 1444 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1445 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1446 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1447 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1448 static ConstString g_sect_name_objc_const("__objc_const"); 1449 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1450 static ConstString g_sect_name_cfstring("__cfstring"); 1451 1452 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1453 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1454 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1455 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1456 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1457 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1458 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1459 static ConstString g_sect_name_dwarf_debug_names("__debug_names"); 1460 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1461 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1462 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1463 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1464 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1465 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1466 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1467 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1468 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1469 static ConstString g_sect_name_eh_frame("__eh_frame"); 1470 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1471 static ConstString g_sect_name_text("__text"); 1472 static ConstString g_sect_name_data("__data"); 1473 static ConstString g_sect_name_go_symtab("__gosymtab"); 1474 1475 if (section_name == g_sect_name_dwarf_debug_abbrev) 1476 return eSectionTypeDWARFDebugAbbrev; 1477 if (section_name == g_sect_name_dwarf_debug_aranges) 1478 return eSectionTypeDWARFDebugAranges; 1479 if (section_name == g_sect_name_dwarf_debug_frame) 1480 return eSectionTypeDWARFDebugFrame; 1481 if (section_name == g_sect_name_dwarf_debug_info) 1482 return eSectionTypeDWARFDebugInfo; 1483 if (section_name == g_sect_name_dwarf_debug_line) 1484 return eSectionTypeDWARFDebugLine; 1485 if (section_name == g_sect_name_dwarf_debug_loc) 1486 return eSectionTypeDWARFDebugLoc; 1487 if (section_name == g_sect_name_dwarf_debug_macinfo) 1488 return eSectionTypeDWARFDebugMacInfo; 1489 if (section_name == g_sect_name_dwarf_debug_names) 1490 return eSectionTypeDWARFDebugNames; 1491 if (section_name == g_sect_name_dwarf_debug_pubnames) 1492 return eSectionTypeDWARFDebugPubNames; 1493 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1494 return eSectionTypeDWARFDebugPubTypes; 1495 if (section_name == g_sect_name_dwarf_debug_ranges) 1496 return eSectionTypeDWARFDebugRanges; 1497 if (section_name == g_sect_name_dwarf_debug_str) 1498 return eSectionTypeDWARFDebugStr; 1499 if (section_name == g_sect_name_dwarf_debug_types) 1500 return eSectionTypeDWARFDebugTypes; 1501 if (section_name == g_sect_name_dwarf_apple_names) 1502 return eSectionTypeDWARFAppleNames; 1503 if (section_name == g_sect_name_dwarf_apple_types) 1504 return eSectionTypeDWARFAppleTypes; 1505 if (section_name == g_sect_name_dwarf_apple_namespaces) 1506 return eSectionTypeDWARFAppleNamespaces; 1507 if (section_name == g_sect_name_dwarf_apple_objc) 1508 return eSectionTypeDWARFAppleObjC; 1509 if (section_name == g_sect_name_objc_selrefs) 1510 return eSectionTypeDataCStringPointers; 1511 if (section_name == g_sect_name_objc_msgrefs) 1512 return eSectionTypeDataObjCMessageRefs; 1513 if (section_name == g_sect_name_eh_frame) 1514 return eSectionTypeEHFrame; 1515 if (section_name == g_sect_name_compact_unwind) 1516 return eSectionTypeCompactUnwind; 1517 if (section_name == g_sect_name_cfstring) 1518 return eSectionTypeDataObjCCFStrings; 1519 if (section_name == g_sect_name_go_symtab) 1520 return eSectionTypeGoSymtab; 1521 if (section_name == g_sect_name_objc_data || 1522 section_name == g_sect_name_objc_classrefs || 1523 section_name == g_sect_name_objc_superrefs || 1524 section_name == g_sect_name_objc_const || 1525 section_name == g_sect_name_objc_classlist) { 1526 return eSectionTypeDataPointers; 1527 } 1528 1529 switch (mach_sect_type) { 1530 // TODO: categorize sections by other flags for regular sections 1531 case S_REGULAR: 1532 if (section_name == g_sect_name_text) 1533 return eSectionTypeCode; 1534 if (section_name == g_sect_name_data) 1535 return eSectionTypeData; 1536 return eSectionTypeOther; 1537 case S_ZEROFILL: 1538 return eSectionTypeZeroFill; 1539 case S_CSTRING_LITERALS: // section with only literal C strings 1540 return eSectionTypeDataCString; 1541 case S_4BYTE_LITERALS: // section with only 4 byte literals 1542 return eSectionTypeData4; 1543 case S_8BYTE_LITERALS: // section with only 8 byte literals 1544 return eSectionTypeData8; 1545 case S_LITERAL_POINTERS: // section with only pointers to literals 1546 return eSectionTypeDataPointers; 1547 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1548 return eSectionTypeDataPointers; 1549 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1550 return eSectionTypeDataPointers; 1551 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1552 // the reserved2 field 1553 return eSectionTypeCode; 1554 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1555 // initialization 1556 return eSectionTypeDataPointers; 1557 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1558 // termination 1559 return eSectionTypeDataPointers; 1560 case S_COALESCED: 1561 return eSectionTypeOther; 1562 case S_GB_ZEROFILL: 1563 return eSectionTypeZeroFill; 1564 case S_INTERPOSING: // section with only pairs of function pointers for 1565 // interposing 1566 return eSectionTypeCode; 1567 case S_16BYTE_LITERALS: // section with only 16 byte literals 1568 return eSectionTypeData16; 1569 case S_DTRACE_DOF: 1570 return eSectionTypeDebug; 1571 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1572 return eSectionTypeDataPointers; 1573 default: 1574 return eSectionTypeOther; 1575 } 1576 } 1577 1578 struct ObjectFileMachO::SegmentParsingContext { 1579 const EncryptedFileRanges EncryptedRanges; 1580 lldb_private::SectionList &UnifiedList; 1581 uint32_t NextSegmentIdx = 0; 1582 uint32_t NextSectionIdx = 0; 1583 bool FileAddressesChanged = false; 1584 1585 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1586 lldb_private::SectionList &UnifiedList) 1587 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1588 }; 1589 1590 void ObjectFileMachO::ProcessSegmentCommand(const load_command &load_cmd_, 1591 lldb::offset_t offset, 1592 uint32_t cmd_idx, 1593 SegmentParsingContext &context) { 1594 segment_command_64 load_cmd; 1595 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1596 1597 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1598 return; 1599 1600 ModuleSP module_sp = GetModule(); 1601 const bool is_core = GetType() == eTypeCoreFile; 1602 const bool is_dsym = (m_header.filetype == MH_DSYM); 1603 bool add_section = true; 1604 bool add_to_unified = true; 1605 ConstString const_segname( 1606 load_cmd.segname, 1607 std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 1608 1609 SectionSP unified_section_sp( 1610 context.UnifiedList.FindSectionByName(const_segname)); 1611 if (is_dsym && unified_section_sp) { 1612 if (const_segname == GetSegmentNameLINKEDIT()) { 1613 // We need to keep the __LINKEDIT segment private to this object file 1614 // only 1615 add_to_unified = false; 1616 } else { 1617 // This is the dSYM file and this section has already been created by the 1618 // object file, no need to create it. 1619 add_section = false; 1620 } 1621 } 1622 load_cmd.vmaddr = m_data.GetAddress(&offset); 1623 load_cmd.vmsize = m_data.GetAddress(&offset); 1624 load_cmd.fileoff = m_data.GetAddress(&offset); 1625 load_cmd.filesize = m_data.GetAddress(&offset); 1626 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1627 return; 1628 1629 SanitizeSegmentCommand(load_cmd, cmd_idx); 1630 1631 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1632 const bool segment_is_encrypted = 1633 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1634 1635 // Keep a list of mach segments around in case we need to get at data that 1636 // isn't stored in the abstracted Sections. 1637 m_mach_segments.push_back(load_cmd); 1638 1639 // Use a segment ID of the segment index shifted left by 8 so they never 1640 // conflict with any of the sections. 1641 SectionSP segment_sp; 1642 if (add_section && (const_segname || is_core)) { 1643 segment_sp.reset(new Section( 1644 module_sp, // Module to which this section belongs 1645 this, // Object file to which this sections belongs 1646 ++context.NextSegmentIdx 1647 << 8, // Section ID is the 1 based segment index 1648 // shifted right by 8 bits as not to collide with any of the 256 1649 // section IDs that are possible 1650 const_segname, // Name of this section 1651 eSectionTypeContainer, // This section is a container of other 1652 // sections. 1653 load_cmd.vmaddr, // File VM address == addresses as they are 1654 // found in the object file 1655 load_cmd.vmsize, // VM size in bytes of this section 1656 load_cmd.fileoff, // Offset to the data for this section in 1657 // the file 1658 load_cmd.filesize, // Size in bytes of this section as found 1659 // in the file 1660 0, // Segments have no alignment information 1661 load_cmd.flags)); // Flags for this section 1662 1663 segment_sp->SetIsEncrypted(segment_is_encrypted); 1664 m_sections_ap->AddSection(segment_sp); 1665 segment_sp->SetPermissions(segment_permissions); 1666 if (add_to_unified) 1667 context.UnifiedList.AddSection(segment_sp); 1668 } else if (unified_section_sp) { 1669 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1670 // Check to see if the module was read from memory? 1671 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) { 1672 // We have a module that is in memory and needs to have its file 1673 // address adjusted. We need to do this because when we load a file 1674 // from memory, its addresses will be slid already, yet the addresses 1675 // in the new symbol file will still be unslid. Since everything is 1676 // stored as section offset, this shouldn't cause any problems. 1677 1678 // Make sure we've parsed the symbol table from the ObjectFile before 1679 // we go around changing its Sections. 1680 module_sp->GetObjectFile()->GetSymtab(); 1681 // eh_frame would present the same problems but we parse that on a per- 1682 // function basis as-needed so it's more difficult to remove its use of 1683 // the Sections. Realistically, the environments where this code path 1684 // will be taken will not have eh_frame sections. 1685 1686 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1687 1688 // Notify the module that the section addresses have been changed once 1689 // we're done so any file-address caches can be updated. 1690 context.FileAddressesChanged = true; 1691 } 1692 } 1693 m_sections_ap->AddSection(unified_section_sp); 1694 } 1695 1696 struct section_64 sect64; 1697 ::memset(§64, 0, sizeof(sect64)); 1698 // Push a section into our mach sections for the section at index zero 1699 // (NO_SECT) if we don't have any mach sections yet... 1700 if (m_mach_sections.empty()) 1701 m_mach_sections.push_back(sect64); 1702 uint32_t segment_sect_idx; 1703 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1704 1705 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1706 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1707 ++segment_sect_idx) { 1708 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1709 sizeof(sect64.sectname)) == NULL) 1710 break; 1711 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1712 sizeof(sect64.segname)) == NULL) 1713 break; 1714 sect64.addr = m_data.GetAddress(&offset); 1715 sect64.size = m_data.GetAddress(&offset); 1716 1717 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1718 break; 1719 1720 // Keep a list of mach sections around in case we need to get at data that 1721 // isn't stored in the abstracted Sections. 1722 m_mach_sections.push_back(sect64); 1723 1724 if (add_section) { 1725 ConstString section_name( 1726 sect64.sectname, 1727 std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1728 if (!const_segname) { 1729 // We have a segment with no name so we need to conjure up segments 1730 // that correspond to the section's segname if there isn't already such 1731 // a section. If there is such a section, we resize the section so that 1732 // it spans all sections. We also mark these sections as fake so 1733 // address matches don't hit if they land in the gaps between the child 1734 // sections. 1735 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1736 sizeof(sect64.segname)); 1737 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1738 if (segment_sp.get()) { 1739 Section *segment = segment_sp.get(); 1740 // Grow the section size as needed. 1741 const lldb::addr_t sect64_min_addr = sect64.addr; 1742 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1743 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1744 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1745 const lldb::addr_t curr_seg_max_addr = 1746 curr_seg_min_addr + curr_seg_byte_size; 1747 if (sect64_min_addr >= curr_seg_min_addr) { 1748 const lldb::addr_t new_seg_byte_size = 1749 sect64_max_addr - curr_seg_min_addr; 1750 // Only grow the section size if needed 1751 if (new_seg_byte_size > curr_seg_byte_size) 1752 segment->SetByteSize(new_seg_byte_size); 1753 } else { 1754 // We need to change the base address of the segment and adjust the 1755 // child section offsets for all existing children. 1756 const lldb::addr_t slide_amount = 1757 sect64_min_addr - curr_seg_min_addr; 1758 segment->Slide(slide_amount, false); 1759 segment->GetChildren().Slide(-slide_amount, false); 1760 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1761 } 1762 1763 // Grow the section size as needed. 1764 if (sect64.offset) { 1765 const lldb::addr_t segment_min_file_offset = 1766 segment->GetFileOffset(); 1767 const lldb::addr_t segment_max_file_offset = 1768 segment_min_file_offset + segment->GetFileSize(); 1769 1770 const lldb::addr_t section_min_file_offset = sect64.offset; 1771 const lldb::addr_t section_max_file_offset = 1772 section_min_file_offset + sect64.size; 1773 const lldb::addr_t new_file_offset = 1774 std::min(section_min_file_offset, segment_min_file_offset); 1775 const lldb::addr_t new_file_size = 1776 std::max(section_max_file_offset, segment_max_file_offset) - 1777 new_file_offset; 1778 segment->SetFileOffset(new_file_offset); 1779 segment->SetFileSize(new_file_size); 1780 } 1781 } else { 1782 // Create a fake section for the section's named segment 1783 segment_sp.reset(new Section( 1784 segment_sp, // Parent section 1785 module_sp, // Module to which this section belongs 1786 this, // Object file to which this section belongs 1787 ++context.NextSegmentIdx 1788 << 8, // Section ID is the 1 based segment index 1789 // shifted right by 8 bits as not to 1790 // collide with any of the 256 section IDs 1791 // that are possible 1792 const_segname, // Name of this section 1793 eSectionTypeContainer, // This section is a container of 1794 // other sections. 1795 sect64.addr, // File VM address == addresses as they are 1796 // found in the object file 1797 sect64.size, // VM size in bytes of this section 1798 sect64.offset, // Offset to the data for this section in 1799 // the file 1800 sect64.offset ? sect64.size : 0, // Size in bytes of 1801 // this section as 1802 // found in the file 1803 sect64.align, 1804 load_cmd.flags)); // Flags for this section 1805 segment_sp->SetIsFake(true); 1806 segment_sp->SetPermissions(segment_permissions); 1807 m_sections_ap->AddSection(segment_sp); 1808 if (add_to_unified) 1809 context.UnifiedList.AddSection(segment_sp); 1810 segment_sp->SetIsEncrypted(segment_is_encrypted); 1811 } 1812 } 1813 assert(segment_sp.get()); 1814 1815 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1816 1817 SectionSP section_sp(new Section( 1818 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1819 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1820 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1821 sect64.flags)); 1822 // Set the section to be encrypted to match the segment 1823 1824 bool section_is_encrypted = false; 1825 if (!segment_is_encrypted && load_cmd.filesize != 0) 1826 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1827 sect64.offset) != NULL; 1828 1829 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1830 section_sp->SetPermissions(segment_permissions); 1831 segment_sp->GetChildren().AddSection(section_sp); 1832 1833 if (segment_sp->IsFake()) { 1834 segment_sp.reset(); 1835 const_segname.Clear(); 1836 } 1837 } 1838 } 1839 if (segment_sp && is_dsym) { 1840 if (first_segment_sectID <= context.NextSectionIdx) { 1841 lldb::user_id_t sect_uid; 1842 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1843 ++sect_uid) { 1844 SectionSP curr_section_sp( 1845 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1846 SectionSP next_section_sp; 1847 if (sect_uid + 1 <= context.NextSectionIdx) 1848 next_section_sp = 1849 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1850 1851 if (curr_section_sp.get()) { 1852 if (curr_section_sp->GetByteSize() == 0) { 1853 if (next_section_sp.get() != NULL) 1854 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1855 curr_section_sp->GetFileAddress()); 1856 else 1857 curr_section_sp->SetByteSize(load_cmd.vmsize); 1858 } 1859 } 1860 } 1861 } 1862 } 1863 } 1864 1865 void ObjectFileMachO::ProcessDysymtabCommand(const load_command &load_cmd, 1866 lldb::offset_t offset) { 1867 m_dysymtab.cmd = load_cmd.cmd; 1868 m_dysymtab.cmdsize = load_cmd.cmdsize; 1869 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1870 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1871 } 1872 1873 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1874 if (m_sections_ap) 1875 return; 1876 1877 m_sections_ap.reset(new SectionList()); 1878 1879 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1880 // bool dump_sections = false; 1881 ModuleSP module_sp(GetModule()); 1882 1883 offset = MachHeaderSizeFromMagic(m_header.magic); 1884 1885 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1886 struct load_command load_cmd; 1887 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1888 const lldb::offset_t load_cmd_offset = offset; 1889 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1890 break; 1891 1892 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1893 ProcessSegmentCommand(load_cmd, offset, i, context); 1894 else if (load_cmd.cmd == LC_DYSYMTAB) 1895 ProcessDysymtabCommand(load_cmd, offset); 1896 1897 offset = load_cmd_offset + load_cmd.cmdsize; 1898 } 1899 1900 if (context.FileAddressesChanged && module_sp) 1901 module_sp->SectionFileAddressesChanged(); 1902 } 1903 1904 class MachSymtabSectionInfo { 1905 public: 1906 MachSymtabSectionInfo(SectionList *section_list) 1907 : m_section_list(section_list), m_section_infos() { 1908 // Get the number of sections down to a depth of 1 to include all segments 1909 // and their sections, but no other sections that may be added for debug 1910 // map or 1911 m_section_infos.resize(section_list->GetNumSections(1)); 1912 } 1913 1914 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1915 if (n_sect == 0) 1916 return SectionSP(); 1917 if (n_sect < m_section_infos.size()) { 1918 if (!m_section_infos[n_sect].section_sp) { 1919 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1920 m_section_infos[n_sect].section_sp = section_sp; 1921 if (section_sp) { 1922 m_section_infos[n_sect].vm_range.SetBaseAddress( 1923 section_sp->GetFileAddress()); 1924 m_section_infos[n_sect].vm_range.SetByteSize( 1925 section_sp->GetByteSize()); 1926 } else { 1927 Host::SystemLog(Host::eSystemLogError, 1928 "error: unable to find section for section %u\n", 1929 n_sect); 1930 } 1931 } 1932 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1933 // Symbol is in section. 1934 return m_section_infos[n_sect].section_sp; 1935 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1936 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1937 file_addr) { 1938 // Symbol is in section with zero size, but has the same start address 1939 // as the section. This can happen with linker symbols (symbols that 1940 // start with the letter 'l' or 'L'. 1941 return m_section_infos[n_sect].section_sp; 1942 } 1943 } 1944 return m_section_list->FindSectionContainingFileAddress(file_addr); 1945 } 1946 1947 protected: 1948 struct SectionInfo { 1949 SectionInfo() : vm_range(), section_sp() {} 1950 1951 VMRange vm_range; 1952 SectionSP section_sp; 1953 }; 1954 SectionList *m_section_list; 1955 std::vector<SectionInfo> m_section_infos; 1956 }; 1957 1958 struct TrieEntry { 1959 TrieEntry() 1960 : name(), address(LLDB_INVALID_ADDRESS), flags(0), other(0), 1961 import_name() {} 1962 1963 void Clear() { 1964 name.Clear(); 1965 address = LLDB_INVALID_ADDRESS; 1966 flags = 0; 1967 other = 0; 1968 import_name.Clear(); 1969 } 1970 1971 void Dump() const { 1972 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1973 static_cast<unsigned long long>(address), 1974 static_cast<unsigned long long>(flags), 1975 static_cast<unsigned long long>(other), name.GetCString()); 1976 if (import_name) 1977 printf(" -> \"%s\"\n", import_name.GetCString()); 1978 else 1979 printf("\n"); 1980 } 1981 ConstString name; 1982 uint64_t address; 1983 uint64_t flags; 1984 uint64_t other; 1985 ConstString import_name; 1986 }; 1987 1988 struct TrieEntryWithOffset { 1989 lldb::offset_t nodeOffset; 1990 TrieEntry entry; 1991 1992 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1993 1994 void Dump(uint32_t idx) const { 1995 printf("[%3u] 0x%16.16llx: ", idx, 1996 static_cast<unsigned long long>(nodeOffset)); 1997 entry.Dump(); 1998 } 1999 2000 bool operator<(const TrieEntryWithOffset &other) const { 2001 return (nodeOffset < other.nodeOffset); 2002 } 2003 }; 2004 2005 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2006 const bool is_arm, 2007 std::vector<llvm::StringRef> &nameSlices, 2008 std::set<lldb::addr_t> &resolver_addresses, 2009 std::vector<TrieEntryWithOffset> &output) { 2010 if (!data.ValidOffset(offset)) 2011 return true; 2012 2013 const uint64_t terminalSize = data.GetULEB128(&offset); 2014 lldb::offset_t children_offset = offset + terminalSize; 2015 if (terminalSize != 0) { 2016 TrieEntryWithOffset e(offset); 2017 e.entry.flags = data.GetULEB128(&offset); 2018 const char *import_name = NULL; 2019 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2020 e.entry.address = 0; 2021 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2022 import_name = data.GetCStr(&offset); 2023 } else { 2024 e.entry.address = data.GetULEB128(&offset); 2025 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2026 e.entry.other = data.GetULEB128(&offset); 2027 uint64_t resolver_addr = e.entry.other; 2028 if (is_arm) 2029 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2030 resolver_addresses.insert(resolver_addr); 2031 } else 2032 e.entry.other = 0; 2033 } 2034 // Only add symbols that are reexport symbols with a valid import name 2035 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && 2036 import_name[0]) { 2037 std::string name; 2038 if (!nameSlices.empty()) { 2039 for (auto name_slice : nameSlices) 2040 name.append(name_slice.data(), name_slice.size()); 2041 } 2042 if (name.size() > 1) { 2043 // Skip the leading '_' 2044 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2045 } 2046 if (import_name) { 2047 // Skip the leading '_' 2048 e.entry.import_name.SetCString(import_name + 1); 2049 } 2050 output.push_back(e); 2051 } 2052 } 2053 2054 const uint8_t childrenCount = data.GetU8(&children_offset); 2055 for (uint8_t i = 0; i < childrenCount; ++i) { 2056 const char *cstr = data.GetCStr(&children_offset); 2057 if (cstr) 2058 nameSlices.push_back(llvm::StringRef(cstr)); 2059 else 2060 return false; // Corrupt data 2061 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2062 if (childNodeOffset) { 2063 if (!ParseTrieEntries(data, childNodeOffset, is_arm, nameSlices, 2064 resolver_addresses, output)) { 2065 return false; 2066 } 2067 } 2068 nameSlices.pop_back(); 2069 } 2070 return true; 2071 } 2072 2073 // Read the UUID out of a dyld_shared_cache file on-disk. 2074 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2075 const ByteOrder byte_order, 2076 const uint32_t addr_byte_size) { 2077 UUID dsc_uuid; 2078 DataBufferSP DscData = MapFileData( 2079 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2080 if (!DscData) 2081 return dsc_uuid; 2082 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2083 2084 char version_str[7]; 2085 lldb::offset_t offset = 0; 2086 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2087 version_str[6] = '\0'; 2088 if (strcmp(version_str, "dyld_v") == 0) { 2089 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2090 dsc_uuid = UUID::fromOptionalData( 2091 dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t)); 2092 } 2093 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2094 if (log && dsc_uuid.IsValid()) { 2095 log->Printf("Shared cache %s has UUID %s", dyld_shared_cache.GetPath().c_str(), 2096 dsc_uuid.GetAsString().c_str()); 2097 } 2098 return dsc_uuid; 2099 } 2100 2101 size_t ObjectFileMachO::ParseSymtab() { 2102 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2103 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2104 m_file.GetFilename().AsCString("")); 2105 ModuleSP module_sp(GetModule()); 2106 if (!module_sp) 2107 return 0; 2108 2109 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2110 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2111 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2112 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2113 FunctionStarts function_starts; 2114 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2115 uint32_t i; 2116 FileSpecList dylib_files; 2117 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2118 static const llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2119 static const llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2120 static const llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2121 2122 for (i = 0; i < m_header.ncmds; ++i) { 2123 const lldb::offset_t cmd_offset = offset; 2124 // Read in the load command and load command size 2125 struct load_command lc; 2126 if (m_data.GetU32(&offset, &lc, 2) == NULL) 2127 break; 2128 // Watch for the symbol table load command 2129 switch (lc.cmd) { 2130 case LC_SYMTAB: 2131 symtab_load_command.cmd = lc.cmd; 2132 symtab_load_command.cmdsize = lc.cmdsize; 2133 // Read in the rest of the symtab load command 2134 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2135 0) // fill in symoff, nsyms, stroff, strsize fields 2136 return 0; 2137 if (symtab_load_command.symoff == 0) { 2138 if (log) 2139 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2140 return 0; 2141 } 2142 2143 if (symtab_load_command.stroff == 0) { 2144 if (log) 2145 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2146 return 0; 2147 } 2148 2149 if (symtab_load_command.nsyms == 0) { 2150 if (log) 2151 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2152 return 0; 2153 } 2154 2155 if (symtab_load_command.strsize == 0) { 2156 if (log) 2157 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2158 return 0; 2159 } 2160 break; 2161 2162 case LC_DYLD_INFO: 2163 case LC_DYLD_INFO_ONLY: 2164 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2165 dyld_info.cmd = lc.cmd; 2166 dyld_info.cmdsize = lc.cmdsize; 2167 } else { 2168 memset(&dyld_info, 0, sizeof(dyld_info)); 2169 } 2170 break; 2171 2172 case LC_LOAD_DYLIB: 2173 case LC_LOAD_WEAK_DYLIB: 2174 case LC_REEXPORT_DYLIB: 2175 case LC_LOADFVMLIB: 2176 case LC_LOAD_UPWARD_DYLIB: { 2177 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2178 const char *path = m_data.PeekCStr(name_offset); 2179 if (path) { 2180 FileSpec file_spec(path, false); 2181 // Strip the path if there is @rpath, @executable, etc so we just use 2182 // the basename 2183 if (path[0] == '@') 2184 file_spec.GetDirectory().Clear(); 2185 2186 if (lc.cmd == LC_REEXPORT_DYLIB) { 2187 m_reexported_dylibs.AppendIfUnique(file_spec); 2188 } 2189 2190 dylib_files.Append(file_spec); 2191 } 2192 } break; 2193 2194 case LC_FUNCTION_STARTS: 2195 function_starts_load_command.cmd = lc.cmd; 2196 function_starts_load_command.cmdsize = lc.cmdsize; 2197 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2198 NULL) // fill in symoff, nsyms, stroff, strsize fields 2199 memset(&function_starts_load_command, 0, 2200 sizeof(function_starts_load_command)); 2201 break; 2202 2203 default: 2204 break; 2205 } 2206 offset = cmd_offset + lc.cmdsize; 2207 } 2208 2209 if (symtab_load_command.cmd) { 2210 Symtab *symtab = m_symtab_ap.get(); 2211 SectionList *section_list = GetSectionList(); 2212 if (section_list == NULL) 2213 return 0; 2214 2215 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2216 const ByteOrder byte_order = m_data.GetByteOrder(); 2217 bool bit_width_32 = addr_byte_size == 4; 2218 const size_t nlist_byte_size = 2219 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2220 2221 DataExtractor nlist_data(NULL, 0, byte_order, addr_byte_size); 2222 DataExtractor strtab_data(NULL, 0, byte_order, addr_byte_size); 2223 DataExtractor function_starts_data(NULL, 0, byte_order, addr_byte_size); 2224 DataExtractor indirect_symbol_index_data(NULL, 0, byte_order, 2225 addr_byte_size); 2226 DataExtractor dyld_trie_data(NULL, 0, byte_order, addr_byte_size); 2227 2228 const addr_t nlist_data_byte_size = 2229 symtab_load_command.nsyms * nlist_byte_size; 2230 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2231 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2232 2233 ProcessSP process_sp(m_process_wp.lock()); 2234 Process *process = process_sp.get(); 2235 2236 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2237 2238 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2239 Target &target = process->GetTarget(); 2240 2241 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2242 2243 SectionSP linkedit_section_sp( 2244 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2245 // Reading mach file from memory in a process or core file... 2246 2247 if (linkedit_section_sp) { 2248 addr_t linkedit_load_addr = 2249 linkedit_section_sp->GetLoadBaseAddress(&target); 2250 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2251 // We might be trying to access the symbol table before the 2252 // __LINKEDIT's load address has been set in the target. We can't 2253 // fail to read the symbol table, so calculate the right address 2254 // manually 2255 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2256 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2257 } 2258 2259 const addr_t linkedit_file_offset = 2260 linkedit_section_sp->GetFileOffset(); 2261 const addr_t symoff_addr = linkedit_load_addr + 2262 symtab_load_command.symoff - 2263 linkedit_file_offset; 2264 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2265 linkedit_file_offset; 2266 2267 bool data_was_read = false; 2268 2269 #if defined(__APPLE__) && \ 2270 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2271 if (m_header.flags & 0x80000000u && 2272 process->GetAddressByteSize() == sizeof(void *)) { 2273 // This mach-o memory file is in the dyld shared cache. If this 2274 // program is not remote and this is iOS, then this process will 2275 // share the same shared cache as the process we are debugging and we 2276 // can read the entire __LINKEDIT from the address space in this 2277 // process. This is a needed optimization that is used for local iOS 2278 // debugging only since all shared libraries in the shared cache do 2279 // not have corresponding files that exist in the file system of the 2280 // device. They have been combined into a single file. This means we 2281 // always have to load these files from memory. All of the symbol and 2282 // string tables from all of the __LINKEDIT sections from the shared 2283 // libraries in the shared cache have been merged into a single large 2284 // symbol and string table. Reading all of this symbol and string 2285 // table data across can slow down debug launch times, so we optimize 2286 // this by reading the memory for the __LINKEDIT section from this 2287 // process. 2288 2289 UUID lldb_shared_cache; 2290 addr_t lldb_shared_cache_addr; 2291 GetLLDBSharedCacheUUID (lldb_shared_cache_addr, lldb_shared_cache); 2292 UUID process_shared_cache; 2293 addr_t process_shared_cache_addr; 2294 GetProcessSharedCacheUUID(process, process_shared_cache_addr, process_shared_cache); 2295 bool use_lldb_cache = true; 2296 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2297 (lldb_shared_cache != process_shared_cache 2298 || process_shared_cache_addr != lldb_shared_cache_addr)) { 2299 use_lldb_cache = false; 2300 } 2301 2302 PlatformSP platform_sp(target.GetPlatform()); 2303 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2304 data_was_read = true; 2305 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2306 eByteOrderLittle); 2307 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2308 eByteOrderLittle); 2309 if (function_starts_load_command.cmd) { 2310 const addr_t func_start_addr = 2311 linkedit_load_addr + function_starts_load_command.dataoff - 2312 linkedit_file_offset; 2313 function_starts_data.SetData( 2314 (void *)func_start_addr, 2315 function_starts_load_command.datasize, eByteOrderLittle); 2316 } 2317 } 2318 } 2319 #endif 2320 2321 if (!data_was_read) { 2322 // Always load dyld - the dynamic linker - from memory if we didn't 2323 // find a binary anywhere else. lldb will not register 2324 // dylib/framework/bundle loads/unloads if we don't have the dyld 2325 // symbols, we force dyld to load from memory despite the user's 2326 // target.memory-module-load-level setting. 2327 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2328 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2329 DataBufferSP nlist_data_sp( 2330 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2331 if (nlist_data_sp) 2332 nlist_data.SetData(nlist_data_sp, 0, 2333 nlist_data_sp->GetByteSize()); 2334 // Load strings individually from memory when loading from memory 2335 // since shared cache string tables contain strings for all symbols 2336 // from all shared cached libraries DataBufferSP strtab_data_sp 2337 // (ReadMemory (process_sp, strtab_addr, 2338 // strtab_data_byte_size)); 2339 // if (strtab_data_sp) 2340 // strtab_data.SetData (strtab_data_sp, 0, 2341 // strtab_data_sp->GetByteSize()); 2342 if (m_dysymtab.nindirectsyms != 0) { 2343 const addr_t indirect_syms_addr = linkedit_load_addr + 2344 m_dysymtab.indirectsymoff - 2345 linkedit_file_offset; 2346 DataBufferSP indirect_syms_data_sp( 2347 ReadMemory(process_sp, indirect_syms_addr, 2348 m_dysymtab.nindirectsyms * 4)); 2349 if (indirect_syms_data_sp) 2350 indirect_symbol_index_data.SetData( 2351 indirect_syms_data_sp, 0, 2352 indirect_syms_data_sp->GetByteSize()); 2353 } 2354 } else if (memory_module_load_level >= 2355 eMemoryModuleLoadLevelPartial) { 2356 if (function_starts_load_command.cmd) { 2357 const addr_t func_start_addr = 2358 linkedit_load_addr + function_starts_load_command.dataoff - 2359 linkedit_file_offset; 2360 DataBufferSP func_start_data_sp( 2361 ReadMemory(process_sp, func_start_addr, 2362 function_starts_load_command.datasize)); 2363 if (func_start_data_sp) 2364 function_starts_data.SetData(func_start_data_sp, 0, 2365 func_start_data_sp->GetByteSize()); 2366 } 2367 } 2368 } 2369 } 2370 } else { 2371 nlist_data.SetData(m_data, symtab_load_command.symoff, 2372 nlist_data_byte_size); 2373 strtab_data.SetData(m_data, symtab_load_command.stroff, 2374 strtab_data_byte_size); 2375 2376 if (dyld_info.export_size > 0) { 2377 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2378 dyld_info.export_size); 2379 } 2380 2381 if (m_dysymtab.nindirectsyms != 0) { 2382 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2383 m_dysymtab.nindirectsyms * 4); 2384 } 2385 if (function_starts_load_command.cmd) { 2386 function_starts_data.SetData(m_data, 2387 function_starts_load_command.dataoff, 2388 function_starts_load_command.datasize); 2389 } 2390 } 2391 2392 if (nlist_data.GetByteSize() == 0 && 2393 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2394 if (log) 2395 module_sp->LogMessage(log, "failed to read nlist data"); 2396 return 0; 2397 } 2398 2399 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2400 if (!have_strtab_data) { 2401 if (process) { 2402 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2403 if (log) 2404 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2405 return 0; 2406 } 2407 } else { 2408 if (log) 2409 module_sp->LogMessage(log, "failed to read strtab data"); 2410 return 0; 2411 } 2412 } 2413 2414 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2415 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2416 const ConstString &g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2417 const ConstString &g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2418 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2419 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2420 SectionSP text_section_sp( 2421 section_list->FindSectionByName(g_segment_name_TEXT)); 2422 SectionSP data_section_sp( 2423 section_list->FindSectionByName(g_segment_name_DATA)); 2424 SectionSP data_dirty_section_sp( 2425 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2426 SectionSP data_const_section_sp( 2427 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2428 SectionSP objc_section_sp( 2429 section_list->FindSectionByName(g_segment_name_OBJC)); 2430 SectionSP eh_frame_section_sp; 2431 if (text_section_sp.get()) 2432 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2433 g_section_name_eh_frame); 2434 else 2435 eh_frame_section_sp = 2436 section_list->FindSectionByName(g_section_name_eh_frame); 2437 2438 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2439 2440 // lldb works best if it knows the start address of all functions in a 2441 // module. Linker symbols or debug info are normally the best source of 2442 // information for start addr / size but they may be stripped in a released 2443 // binary. Two additional sources of information exist in Mach-O binaries: 2444 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2445 // function's start address in the 2446 // binary, relative to the text section. 2447 // eh_frame - the eh_frame FDEs have the start addr & size of 2448 // each function 2449 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2450 // all modern binaries. 2451 // Binaries built to run on older releases may need to use eh_frame 2452 // information. 2453 2454 if (text_section_sp && function_starts_data.GetByteSize()) { 2455 FunctionStarts::Entry function_start_entry; 2456 function_start_entry.data = false; 2457 lldb::offset_t function_start_offset = 0; 2458 function_start_entry.addr = text_section_sp->GetFileAddress(); 2459 uint64_t delta; 2460 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2461 0) { 2462 // Now append the current entry 2463 function_start_entry.addr += delta; 2464 function_starts.Append(function_start_entry); 2465 } 2466 } else { 2467 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2468 // load command claiming an eh_frame but it doesn't actually have the 2469 // eh_frame content. And if we have a dSYM, we don't need to do any of 2470 // this fill-in-the-missing-symbols works anyway - the debug info should 2471 // give us all the functions in the module. 2472 if (text_section_sp.get() && eh_frame_section_sp.get() && 2473 m_type != eTypeDebugInfo) { 2474 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2475 DWARFCallFrameInfo::EH); 2476 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2477 eh_frame.GetFunctionAddressAndSizeVector(functions); 2478 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2479 size_t count = functions.GetSize(); 2480 for (size_t i = 0; i < count; ++i) { 2481 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2482 functions.GetEntryAtIndex(i); 2483 if (func) { 2484 FunctionStarts::Entry function_start_entry; 2485 function_start_entry.addr = func->base - text_base_addr; 2486 function_starts.Append(function_start_entry); 2487 } 2488 } 2489 } 2490 } 2491 2492 const size_t function_starts_count = function_starts.GetSize(); 2493 2494 // For user process binaries (executables, dylibs, frameworks, bundles), if 2495 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2496 // going to assume the binary has been stripped. Don't allow assembly 2497 // language instruction emulation because we don't know proper function 2498 // start boundaries. 2499 // 2500 // For all other types of binaries (kernels, stand-alone bare board 2501 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2502 // sections - we should not make any assumptions about them based on that. 2503 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2504 m_allow_assembly_emulation_unwind_plans = false; 2505 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2506 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2507 2508 if (unwind_or_symbol_log) 2509 module_sp->LogMessage( 2510 unwind_or_symbol_log, 2511 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2512 } 2513 2514 const user_id_t TEXT_eh_frame_sectID = 2515 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2516 : static_cast<user_id_t>(NO_SECT); 2517 2518 lldb::offset_t nlist_data_offset = 0; 2519 2520 uint32_t N_SO_index = UINT32_MAX; 2521 2522 MachSymtabSectionInfo section_info(section_list); 2523 std::vector<uint32_t> N_FUN_indexes; 2524 std::vector<uint32_t> N_NSYM_indexes; 2525 std::vector<uint32_t> N_INCL_indexes; 2526 std::vector<uint32_t> N_BRAC_indexes; 2527 std::vector<uint32_t> N_COMM_indexes; 2528 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2529 typedef std::map<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2530 typedef std::map<const char *, uint32_t> ConstNameToSymbolIndexMap; 2531 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2532 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2533 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2534 // Any symbols that get merged into another will get an entry in this map 2535 // so we know 2536 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2537 uint32_t nlist_idx = 0; 2538 Symbol *symbol_ptr = NULL; 2539 2540 uint32_t sym_idx = 0; 2541 Symbol *sym = NULL; 2542 size_t num_syms = 0; 2543 std::string memory_symbol_name; 2544 uint32_t unmapped_local_symbols_found = 0; 2545 2546 std::vector<TrieEntryWithOffset> trie_entries; 2547 std::set<lldb::addr_t> resolver_addresses; 2548 2549 if (dyld_trie_data.GetByteSize() > 0) { 2550 std::vector<llvm::StringRef> nameSlices; 2551 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, 2552 resolver_addresses, trie_entries); 2553 2554 ConstString text_segment_name("__TEXT"); 2555 SectionSP text_segment_sp = 2556 GetSectionList()->FindSectionByName(text_segment_name); 2557 if (text_segment_sp) { 2558 const lldb::addr_t text_segment_file_addr = 2559 text_segment_sp->GetFileAddress(); 2560 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2561 for (auto &e : trie_entries) 2562 e.entry.address += text_segment_file_addr; 2563 } 2564 } 2565 } 2566 2567 typedef std::set<ConstString> IndirectSymbols; 2568 IndirectSymbols indirect_symbol_names; 2569 2570 #if defined(__APPLE__) && \ 2571 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2572 2573 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2574 // optimized by moving LOCAL symbols out of the memory mapped portion of 2575 // the DSC. The symbol information has all been retained, but it isn't 2576 // available in the normal nlist data. However, there *are* duplicate 2577 // entries of *some* 2578 // LOCAL symbols in the normal nlist data. To handle this situation 2579 // correctly, we must first attempt 2580 // to parse any DSC unmapped symbol information. If we find any, we set a 2581 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2582 2583 if (m_header.flags & 0x80000000u) { 2584 // Before we can start mapping the DSC, we need to make certain the 2585 // target process is actually using the cache we can find. 2586 2587 // Next we need to determine the correct path for the dyld shared cache. 2588 2589 ArchSpec header_arch; 2590 GetArchitecture(header_arch); 2591 char dsc_path[PATH_MAX]; 2592 char dsc_path_development[PATH_MAX]; 2593 2594 snprintf( 2595 dsc_path, sizeof(dsc_path), "%s%s%s", 2596 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2597 */ 2598 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2599 header_arch.GetArchitectureName()); 2600 2601 snprintf( 2602 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2603 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2604 */ 2605 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2606 header_arch.GetArchitectureName(), ".development"); 2607 2608 FileSpec dsc_nondevelopment_filespec(dsc_path, false); 2609 FileSpec dsc_development_filespec(dsc_path_development, false); 2610 FileSpec dsc_filespec; 2611 2612 UUID dsc_uuid; 2613 UUID process_shared_cache_uuid; 2614 addr_t process_shared_cache_base_addr; 2615 2616 if (process) { 2617 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, process_shared_cache_uuid); 2618 } 2619 2620 // First see if we can find an exact match for the inferior process 2621 // shared cache UUID in the development or non-development shared caches 2622 // on disk. 2623 if (process_shared_cache_uuid.IsValid()) { 2624 if (dsc_development_filespec.Exists()) { 2625 UUID dsc_development_uuid = GetSharedCacheUUID( 2626 dsc_development_filespec, byte_order, addr_byte_size); 2627 if (dsc_development_uuid.IsValid() && 2628 dsc_development_uuid == process_shared_cache_uuid) { 2629 dsc_filespec = dsc_development_filespec; 2630 dsc_uuid = dsc_development_uuid; 2631 } 2632 } 2633 if (!dsc_uuid.IsValid() && dsc_nondevelopment_filespec.Exists()) { 2634 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2635 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2636 if (dsc_nondevelopment_uuid.IsValid() && 2637 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2638 dsc_filespec = dsc_nondevelopment_filespec; 2639 dsc_uuid = dsc_nondevelopment_uuid; 2640 } 2641 } 2642 } 2643 2644 // Failing a UUID match, prefer the development dyld_shared cache if both 2645 // are present. 2646 if (!dsc_filespec.Exists()) { 2647 if (dsc_development_filespec.Exists()) { 2648 dsc_filespec = dsc_development_filespec; 2649 } else { 2650 dsc_filespec = dsc_nondevelopment_filespec; 2651 } 2652 } 2653 2654 /* The dyld_cache_header has a pointer to the 2655 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2656 The dyld_cache_local_symbols_info structure gives us three things: 2657 1. The start and count of the nlist records in the dyld_shared_cache 2658 file 2659 2. The start and size of the strings for these nlist records 2660 3. The start and count of dyld_cache_local_symbols_entry entries 2661 2662 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2663 dyld shared cache. 2664 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2665 the dyld shared cache. 2666 The dyld_cache_local_symbols_entry also lists the start of this 2667 dylib/framework's nlist records 2668 and the count of how many nlist records there are for this 2669 dylib/framework. 2670 */ 2671 2672 // Process the dyld shared cache header to find the unmapped symbols 2673 2674 DataBufferSP dsc_data_sp = MapFileData( 2675 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2676 if (!dsc_uuid.IsValid()) { 2677 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2678 } 2679 if (dsc_data_sp) { 2680 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2681 2682 bool uuid_match = true; 2683 if (dsc_uuid.IsValid() && process) { 2684 if (process_shared_cache_uuid.IsValid() && 2685 dsc_uuid != process_shared_cache_uuid) { 2686 // The on-disk dyld_shared_cache file is not the same as the one in 2687 // this process' memory, don't use it. 2688 uuid_match = false; 2689 ModuleSP module_sp(GetModule()); 2690 if (module_sp) 2691 module_sp->ReportWarning("process shared cache does not match " 2692 "on-disk dyld_shared_cache file, some " 2693 "symbol names will be missing."); 2694 } 2695 } 2696 2697 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2698 2699 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2700 2701 // If the mappingOffset points to a location inside the header, we've 2702 // opened an old dyld shared cache, and should not proceed further. 2703 if (uuid_match && 2704 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2705 2706 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2707 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2708 mappingOffset); 2709 2710 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2711 byte_order, addr_byte_size); 2712 offset = 0; 2713 2714 // The File addresses (from the in-memory Mach-O load commands) for 2715 // the shared libraries in the shared library cache need to be 2716 // adjusted by an offset to match up with the dylibOffset identifying 2717 // field in the dyld_cache_local_symbol_entry's. This offset is 2718 // recorded in mapping_offset_value. 2719 const uint64_t mapping_offset_value = 2720 dsc_mapping_info_data.GetU64(&offset); 2721 2722 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, 2723 localSymbolsOffset); 2724 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2725 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2726 2727 if (localSymbolsOffset && localSymbolsSize) { 2728 // Map the local symbols 2729 DataBufferSP dsc_local_symbols_data_sp = 2730 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2731 2732 if (dsc_local_symbols_data_sp) { 2733 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2734 byte_order, addr_byte_size); 2735 2736 offset = 0; 2737 2738 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2739 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2740 UndefinedNameToDescMap undefined_name_to_desc; 2741 SymbolIndexToName reexport_shlib_needs_fixup; 2742 2743 // Read the local_symbols_infos struct in one shot 2744 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2745 dsc_local_symbols_data.GetU32(&offset, 2746 &local_symbols_info.nlistOffset, 6); 2747 2748 SectionSP text_section_sp( 2749 section_list->FindSectionByName(GetSegmentNameTEXT())); 2750 2751 uint32_t header_file_offset = 2752 (text_section_sp->GetFileAddress() - mapping_offset_value); 2753 2754 offset = local_symbols_info.entriesOffset; 2755 for (uint32_t entry_index = 0; 2756 entry_index < local_symbols_info.entriesCount; 2757 entry_index++) { 2758 struct lldb_copy_dyld_cache_local_symbols_entry 2759 local_symbols_entry; 2760 local_symbols_entry.dylibOffset = 2761 dsc_local_symbols_data.GetU32(&offset); 2762 local_symbols_entry.nlistStartIndex = 2763 dsc_local_symbols_data.GetU32(&offset); 2764 local_symbols_entry.nlistCount = 2765 dsc_local_symbols_data.GetU32(&offset); 2766 2767 if (header_file_offset == local_symbols_entry.dylibOffset) { 2768 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2769 2770 // The normal nlist code cannot correctly size the Symbols 2771 // array, we need to allocate it here. 2772 sym = symtab->Resize( 2773 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2774 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2775 num_syms = symtab->GetNumSymbols(); 2776 2777 nlist_data_offset = 2778 local_symbols_info.nlistOffset + 2779 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2780 uint32_t string_table_offset = 2781 local_symbols_info.stringsOffset; 2782 2783 for (uint32_t nlist_index = 0; 2784 nlist_index < local_symbols_entry.nlistCount; 2785 nlist_index++) { 2786 ///////////////////////////// 2787 { 2788 struct nlist_64 nlist; 2789 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize( 2790 nlist_data_offset, nlist_byte_size)) 2791 break; 2792 2793 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked( 2794 &nlist_data_offset); 2795 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked( 2796 &nlist_data_offset); 2797 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked( 2798 &nlist_data_offset); 2799 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked( 2800 &nlist_data_offset); 2801 nlist.n_value = 2802 dsc_local_symbols_data.GetAddress_unchecked( 2803 &nlist_data_offset); 2804 2805 SymbolType type = eSymbolTypeInvalid; 2806 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2807 string_table_offset + nlist.n_strx); 2808 2809 if (symbol_name == NULL) { 2810 // No symbol should be NULL, even the symbols with no 2811 // string values should have an offset zero which 2812 // points to an empty C-string 2813 Host::SystemLog( 2814 Host::eSystemLogError, 2815 "error: DSC unmapped local symbol[%u] has invalid " 2816 "string table offset 0x%x in %s, ignoring symbol\n", 2817 entry_index, nlist.n_strx, 2818 module_sp->GetFileSpec().GetPath().c_str()); 2819 continue; 2820 } 2821 if (symbol_name[0] == '\0') 2822 symbol_name = NULL; 2823 2824 const char *symbol_name_non_abi_mangled = NULL; 2825 2826 SectionSP symbol_section; 2827 uint32_t symbol_byte_size = 0; 2828 bool add_nlist = true; 2829 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2830 bool demangled_is_synthesized = false; 2831 bool is_gsym = false; 2832 bool set_value = true; 2833 2834 assert(sym_idx < num_syms); 2835 2836 sym[sym_idx].SetDebug(is_debug); 2837 2838 if (is_debug) { 2839 switch (nlist.n_type) { 2840 case N_GSYM: 2841 // global symbol: name,,NO_SECT,type,0 2842 // Sometimes the N_GSYM value contains the address. 2843 2844 // FIXME: In the .o files, we have a GSYM and a debug 2845 // symbol for all the ObjC data. They 2846 // have the same address, but we want to ensure that 2847 // we always find only the real symbol, 'cause we 2848 // don't currently correctly attribute the 2849 // GSYM one to the ObjCClass/Ivar/MetaClass 2850 // symbol type. This is a temporary hack to make 2851 // sure the ObjectiveC symbols get treated correctly. 2852 // To do this right, we should coalesce all the GSYM 2853 // & global symbols that have the same address. 2854 2855 is_gsym = true; 2856 sym[sym_idx].SetExternal(true); 2857 2858 if (symbol_name && symbol_name[0] == '_' && 2859 symbol_name[1] == 'O') { 2860 llvm::StringRef symbol_name_ref(symbol_name); 2861 if (symbol_name_ref.startswith( 2862 g_objc_v2_prefix_class)) { 2863 symbol_name_non_abi_mangled = symbol_name + 1; 2864 symbol_name = 2865 symbol_name + g_objc_v2_prefix_class.size(); 2866 type = eSymbolTypeObjCClass; 2867 demangled_is_synthesized = true; 2868 2869 } else if (symbol_name_ref.startswith( 2870 g_objc_v2_prefix_metaclass)) { 2871 symbol_name_non_abi_mangled = symbol_name + 1; 2872 symbol_name = symbol_name + 2873 g_objc_v2_prefix_metaclass.size(); 2874 type = eSymbolTypeObjCMetaClass; 2875 demangled_is_synthesized = true; 2876 } else if (symbol_name_ref.startswith( 2877 g_objc_v2_prefix_ivar)) { 2878 symbol_name_non_abi_mangled = symbol_name + 1; 2879 symbol_name = 2880 symbol_name + g_objc_v2_prefix_ivar.size(); 2881 type = eSymbolTypeObjCIVar; 2882 demangled_is_synthesized = true; 2883 } 2884 } else { 2885 if (nlist.n_value != 0) 2886 symbol_section = section_info.GetSection( 2887 nlist.n_sect, nlist.n_value); 2888 type = eSymbolTypeData; 2889 } 2890 break; 2891 2892 case N_FNAME: 2893 // procedure name (f77 kludge): name,,NO_SECT,0,0 2894 type = eSymbolTypeCompiler; 2895 break; 2896 2897 case N_FUN: 2898 // procedure: name,,n_sect,linenumber,address 2899 if (symbol_name) { 2900 type = eSymbolTypeCode; 2901 symbol_section = section_info.GetSection( 2902 nlist.n_sect, nlist.n_value); 2903 2904 N_FUN_addr_to_sym_idx.insert( 2905 std::make_pair(nlist.n_value, sym_idx)); 2906 // We use the current number of symbols in the 2907 // symbol table in lieu of using nlist_idx in case 2908 // we ever start trimming entries out 2909 N_FUN_indexes.push_back(sym_idx); 2910 } else { 2911 type = eSymbolTypeCompiler; 2912 2913 if (!N_FUN_indexes.empty()) { 2914 // Copy the size of the function into the 2915 // original 2916 // STAB entry so we don't have 2917 // to hunt for it later 2918 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2919 ->SetByteSize(nlist.n_value); 2920 N_FUN_indexes.pop_back(); 2921 // We don't really need the end function STAB as 2922 // it contains the size which we already placed 2923 // with the original symbol, so don't add it if 2924 // we want a minimal symbol table 2925 add_nlist = false; 2926 } 2927 } 2928 break; 2929 2930 case N_STSYM: 2931 // static symbol: name,,n_sect,type,address 2932 N_STSYM_addr_to_sym_idx.insert( 2933 std::make_pair(nlist.n_value, sym_idx)); 2934 symbol_section = section_info.GetSection( 2935 nlist.n_sect, nlist.n_value); 2936 if (symbol_name && symbol_name[0]) { 2937 type = ObjectFile::GetSymbolTypeFromName( 2938 symbol_name + 1, eSymbolTypeData); 2939 } 2940 break; 2941 2942 case N_LCSYM: 2943 // .lcomm symbol: name,,n_sect,type,address 2944 symbol_section = section_info.GetSection( 2945 nlist.n_sect, nlist.n_value); 2946 type = eSymbolTypeCommonBlock; 2947 break; 2948 2949 case N_BNSYM: 2950 // We use the current number of symbols in the symbol 2951 // table in lieu of using nlist_idx in case we ever 2952 // start trimming entries out Skip these if we want 2953 // minimal symbol tables 2954 add_nlist = false; 2955 break; 2956 2957 case N_ENSYM: 2958 // Set the size of the N_BNSYM to the terminating 2959 // index of this N_ENSYM so that we can always skip 2960 // the entire symbol if we need to navigate more 2961 // quickly at the source level when parsing STABS 2962 // Skip these if we want minimal symbol tables 2963 add_nlist = false; 2964 break; 2965 2966 case N_OPT: 2967 // emitted with gcc2_compiled and in gcc source 2968 type = eSymbolTypeCompiler; 2969 break; 2970 2971 case N_RSYM: 2972 // register sym: name,,NO_SECT,type,register 2973 type = eSymbolTypeVariable; 2974 break; 2975 2976 case N_SLINE: 2977 // src line: 0,,n_sect,linenumber,address 2978 symbol_section = section_info.GetSection( 2979 nlist.n_sect, nlist.n_value); 2980 type = eSymbolTypeLineEntry; 2981 break; 2982 2983 case N_SSYM: 2984 // structure elt: name,,NO_SECT,type,struct_offset 2985 type = eSymbolTypeVariableType; 2986 break; 2987 2988 case N_SO: 2989 // source file name 2990 type = eSymbolTypeSourceFile; 2991 if (symbol_name == NULL) { 2992 add_nlist = false; 2993 if (N_SO_index != UINT32_MAX) { 2994 // Set the size of the N_SO to the terminating 2995 // index of this N_SO so that we can always skip 2996 // the entire N_SO if we need to navigate more 2997 // quickly at the source level when parsing STABS 2998 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2999 symbol_ptr->SetByteSize(sym_idx); 3000 symbol_ptr->SetSizeIsSibling(true); 3001 } 3002 N_NSYM_indexes.clear(); 3003 N_INCL_indexes.clear(); 3004 N_BRAC_indexes.clear(); 3005 N_COMM_indexes.clear(); 3006 N_FUN_indexes.clear(); 3007 N_SO_index = UINT32_MAX; 3008 } else { 3009 // We use the current number of symbols in the 3010 // symbol table in lieu of using nlist_idx in case 3011 // we ever start trimming entries out 3012 const bool N_SO_has_full_path = 3013 symbol_name[0] == '/'; 3014 if (N_SO_has_full_path) { 3015 if ((N_SO_index == sym_idx - 1) && 3016 ((sym_idx - 1) < num_syms)) { 3017 // We have two consecutive N_SO entries where 3018 // the first contains a directory and the 3019 // second contains a full path. 3020 sym[sym_idx - 1].GetMangled().SetValue( 3021 ConstString(symbol_name), false); 3022 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3023 add_nlist = false; 3024 } else { 3025 // This is the first entry in a N_SO that 3026 // contains a directory or 3027 // a full path to the source file 3028 N_SO_index = sym_idx; 3029 } 3030 } else if ((N_SO_index == sym_idx - 1) && 3031 ((sym_idx - 1) < num_syms)) { 3032 // This is usually the second N_SO entry that 3033 // contains just the filename, so here we combine 3034 // it with the first one if we are minimizing the 3035 // symbol table 3036 const char *so_path = 3037 sym[sym_idx - 1] 3038 .GetMangled() 3039 .GetDemangledName( 3040 lldb::eLanguageTypeUnknown) 3041 .AsCString(); 3042 if (so_path && so_path[0]) { 3043 std::string full_so_path(so_path); 3044 const size_t double_slash_pos = 3045 full_so_path.find("//"); 3046 if (double_slash_pos != std::string::npos) { 3047 // The linker has been generating bad N_SO 3048 // entries with doubled up paths 3049 // in the format "%s%s" where the first 3050 // string in the DW_AT_comp_dir, and the 3051 // second is the directory for the source 3052 // file so you end up with a path that looks 3053 // like "/tmp/src//tmp/src/" 3054 FileSpec so_dir(so_path, false); 3055 if (!so_dir.Exists()) { 3056 so_dir.SetFile( 3057 &full_so_path[double_slash_pos + 1], 3058 false); 3059 if (so_dir.Exists()) { 3060 // Trim off the incorrect path 3061 full_so_path.erase(0, 3062 double_slash_pos + 1); 3063 } 3064 } 3065 } 3066 if (*full_so_path.rbegin() != '/') 3067 full_so_path += '/'; 3068 full_so_path += symbol_name; 3069 sym[sym_idx - 1].GetMangled().SetValue( 3070 ConstString(full_so_path.c_str()), false); 3071 add_nlist = false; 3072 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3073 } 3074 } else { 3075 // This could be a relative path to a N_SO 3076 N_SO_index = sym_idx; 3077 } 3078 } 3079 break; 3080 3081 case N_OSO: 3082 // object file name: name,,0,0,st_mtime 3083 type = eSymbolTypeObjectFile; 3084 break; 3085 3086 case N_LSYM: 3087 // local sym: name,,NO_SECT,type,offset 3088 type = eSymbolTypeLocal; 3089 break; 3090 3091 //---------------------------------------------------------------------- 3092 // INCL scopes 3093 //---------------------------------------------------------------------- 3094 case N_BINCL: 3095 // include file beginning: name,,NO_SECT,0,sum We use 3096 // the current number of symbols in the symbol table 3097 // in lieu of using nlist_idx in case we ever start 3098 // trimming entries out 3099 N_INCL_indexes.push_back(sym_idx); 3100 type = eSymbolTypeScopeBegin; 3101 break; 3102 3103 case N_EINCL: 3104 // include file end: name,,NO_SECT,0,0 3105 // Set the size of the N_BINCL to the terminating 3106 // index of this N_EINCL so that we can always skip 3107 // the entire symbol if we need to navigate more 3108 // quickly at the source level when parsing STABS 3109 if (!N_INCL_indexes.empty()) { 3110 symbol_ptr = 3111 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3112 symbol_ptr->SetByteSize(sym_idx + 1); 3113 symbol_ptr->SetSizeIsSibling(true); 3114 N_INCL_indexes.pop_back(); 3115 } 3116 type = eSymbolTypeScopeEnd; 3117 break; 3118 3119 case N_SOL: 3120 // #included file name: name,,n_sect,0,address 3121 type = eSymbolTypeHeaderFile; 3122 3123 // We currently don't use the header files on darwin 3124 add_nlist = false; 3125 break; 3126 3127 case N_PARAMS: 3128 // compiler parameters: name,,NO_SECT,0,0 3129 type = eSymbolTypeCompiler; 3130 break; 3131 3132 case N_VERSION: 3133 // compiler version: name,,NO_SECT,0,0 3134 type = eSymbolTypeCompiler; 3135 break; 3136 3137 case N_OLEVEL: 3138 // compiler -O level: name,,NO_SECT,0,0 3139 type = eSymbolTypeCompiler; 3140 break; 3141 3142 case N_PSYM: 3143 // parameter: name,,NO_SECT,type,offset 3144 type = eSymbolTypeVariable; 3145 break; 3146 3147 case N_ENTRY: 3148 // alternate entry: name,,n_sect,linenumber,address 3149 symbol_section = section_info.GetSection( 3150 nlist.n_sect, nlist.n_value); 3151 type = eSymbolTypeLineEntry; 3152 break; 3153 3154 //---------------------------------------------------------------------- 3155 // Left and Right Braces 3156 //---------------------------------------------------------------------- 3157 case N_LBRAC: 3158 // left bracket: 0,,NO_SECT,nesting level,address We 3159 // use the current number of symbols in the symbol 3160 // table in lieu of using nlist_idx in case we ever 3161 // start trimming entries out 3162 symbol_section = section_info.GetSection( 3163 nlist.n_sect, nlist.n_value); 3164 N_BRAC_indexes.push_back(sym_idx); 3165 type = eSymbolTypeScopeBegin; 3166 break; 3167 3168 case N_RBRAC: 3169 // right bracket: 0,,NO_SECT,nesting level,address 3170 // Set the size of the N_LBRAC to the terminating 3171 // index of this N_RBRAC so that we can always skip 3172 // the entire symbol if we need to navigate more 3173 // quickly at the source level when parsing STABS 3174 symbol_section = section_info.GetSection( 3175 nlist.n_sect, nlist.n_value); 3176 if (!N_BRAC_indexes.empty()) { 3177 symbol_ptr = 3178 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3179 symbol_ptr->SetByteSize(sym_idx + 1); 3180 symbol_ptr->SetSizeIsSibling(true); 3181 N_BRAC_indexes.pop_back(); 3182 } 3183 type = eSymbolTypeScopeEnd; 3184 break; 3185 3186 case N_EXCL: 3187 // deleted include file: name,,NO_SECT,0,sum 3188 type = eSymbolTypeHeaderFile; 3189 break; 3190 3191 //---------------------------------------------------------------------- 3192 // COMM scopes 3193 //---------------------------------------------------------------------- 3194 case N_BCOMM: 3195 // begin common: name,,NO_SECT,0,0 3196 // We use the current number of symbols in the symbol 3197 // table in lieu of using nlist_idx in case we ever 3198 // start trimming entries out 3199 type = eSymbolTypeScopeBegin; 3200 N_COMM_indexes.push_back(sym_idx); 3201 break; 3202 3203 case N_ECOML: 3204 // end common (local name): 0,,n_sect,0,address 3205 symbol_section = section_info.GetSection( 3206 nlist.n_sect, nlist.n_value); 3207 // Fall through 3208 3209 case N_ECOMM: 3210 // end common: name,,n_sect,0,0 3211 // Set the size of the N_BCOMM to the terminating 3212 // index of this N_ECOMM/N_ECOML so that we can 3213 // always skip the entire symbol if we need to 3214 // navigate more quickly at the source level when 3215 // parsing STABS 3216 if (!N_COMM_indexes.empty()) { 3217 symbol_ptr = 3218 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3219 symbol_ptr->SetByteSize(sym_idx + 1); 3220 symbol_ptr->SetSizeIsSibling(true); 3221 N_COMM_indexes.pop_back(); 3222 } 3223 type = eSymbolTypeScopeEnd; 3224 break; 3225 3226 case N_LENG: 3227 // second stab entry with length information 3228 type = eSymbolTypeAdditional; 3229 break; 3230 3231 default: 3232 break; 3233 } 3234 } else { 3235 // uint8_t n_pext = N_PEXT & nlist.n_type; 3236 uint8_t n_type = N_TYPE & nlist.n_type; 3237 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3238 3239 switch (n_type) { 3240 case N_INDR: { 3241 const char *reexport_name_cstr = 3242 strtab_data.PeekCStr(nlist.n_value); 3243 if (reexport_name_cstr && reexport_name_cstr[0]) { 3244 type = eSymbolTypeReExported; 3245 ConstString reexport_name( 3246 reexport_name_cstr + 3247 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3248 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3249 set_value = false; 3250 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3251 indirect_symbol_names.insert( 3252 ConstString(symbol_name + 3253 ((symbol_name[0] == '_') ? 1 : 0))); 3254 } else 3255 type = eSymbolTypeUndefined; 3256 } break; 3257 3258 case N_UNDF: 3259 if (symbol_name && symbol_name[0]) { 3260 ConstString undefined_name( 3261 symbol_name + 3262 ((symbol_name[0] == '_') ? 1 : 0)); 3263 undefined_name_to_desc[undefined_name] = 3264 nlist.n_desc; 3265 } 3266 // Fall through 3267 case N_PBUD: 3268 type = eSymbolTypeUndefined; 3269 break; 3270 3271 case N_ABS: 3272 type = eSymbolTypeAbsolute; 3273 break; 3274 3275 case N_SECT: { 3276 symbol_section = section_info.GetSection( 3277 nlist.n_sect, nlist.n_value); 3278 3279 if (symbol_section == NULL) { 3280 // TODO: warn about this? 3281 add_nlist = false; 3282 break; 3283 } 3284 3285 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3286 type = eSymbolTypeException; 3287 } else { 3288 uint32_t section_type = 3289 symbol_section->Get() & SECTION_TYPE; 3290 3291 switch (section_type) { 3292 case S_CSTRING_LITERALS: 3293 type = eSymbolTypeData; 3294 break; // section with only literal C strings 3295 case S_4BYTE_LITERALS: 3296 type = eSymbolTypeData; 3297 break; // section with only 4 byte literals 3298 case S_8BYTE_LITERALS: 3299 type = eSymbolTypeData; 3300 break; // section with only 8 byte literals 3301 case S_LITERAL_POINTERS: 3302 type = eSymbolTypeTrampoline; 3303 break; // section with only pointers to literals 3304 case S_NON_LAZY_SYMBOL_POINTERS: 3305 type = eSymbolTypeTrampoline; 3306 break; // section with only non-lazy symbol 3307 // pointers 3308 case S_LAZY_SYMBOL_POINTERS: 3309 type = eSymbolTypeTrampoline; 3310 break; // section with only lazy symbol pointers 3311 case S_SYMBOL_STUBS: 3312 type = eSymbolTypeTrampoline; 3313 break; // section with only symbol stubs, byte 3314 // size of stub in the reserved2 field 3315 case S_MOD_INIT_FUNC_POINTERS: 3316 type = eSymbolTypeCode; 3317 break; // section with only function pointers for 3318 // initialization 3319 case S_MOD_TERM_FUNC_POINTERS: 3320 type = eSymbolTypeCode; 3321 break; // section with only function pointers for 3322 // termination 3323 case S_INTERPOSING: 3324 type = eSymbolTypeTrampoline; 3325 break; // section with only pairs of function 3326 // pointers for interposing 3327 case S_16BYTE_LITERALS: 3328 type = eSymbolTypeData; 3329 break; // section with only 16 byte literals 3330 case S_DTRACE_DOF: 3331 type = eSymbolTypeInstrumentation; 3332 break; 3333 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3334 type = eSymbolTypeTrampoline; 3335 break; 3336 default: 3337 switch (symbol_section->GetType()) { 3338 case lldb::eSectionTypeCode: 3339 type = eSymbolTypeCode; 3340 break; 3341 case eSectionTypeData: 3342 case eSectionTypeDataCString: // Inlined C string 3343 // data 3344 case eSectionTypeDataCStringPointers: // Pointers 3345 // to C 3346 // string 3347 // data 3348 case eSectionTypeDataSymbolAddress: // Address of 3349 // a symbol in 3350 // the symbol 3351 // table 3352 case eSectionTypeData4: 3353 case eSectionTypeData8: 3354 case eSectionTypeData16: 3355 type = eSymbolTypeData; 3356 break; 3357 default: 3358 break; 3359 } 3360 break; 3361 } 3362 3363 if (type == eSymbolTypeInvalid) { 3364 const char *symbol_sect_name = 3365 symbol_section->GetName().AsCString(); 3366 if (symbol_section->IsDescendant( 3367 text_section_sp.get())) { 3368 if (symbol_section->IsClear( 3369 S_ATTR_PURE_INSTRUCTIONS | 3370 S_ATTR_SELF_MODIFYING_CODE | 3371 S_ATTR_SOME_INSTRUCTIONS)) 3372 type = eSymbolTypeData; 3373 else 3374 type = eSymbolTypeCode; 3375 } else if (symbol_section->IsDescendant( 3376 data_section_sp.get()) || 3377 symbol_section->IsDescendant( 3378 data_dirty_section_sp.get()) || 3379 symbol_section->IsDescendant( 3380 data_const_section_sp.get())) { 3381 if (symbol_sect_name && 3382 ::strstr(symbol_sect_name, "__objc") == 3383 symbol_sect_name) { 3384 type = eSymbolTypeRuntime; 3385 3386 if (symbol_name) { 3387 llvm::StringRef symbol_name_ref( 3388 symbol_name); 3389 if (symbol_name_ref.startswith("_OBJC_")) { 3390 static const llvm::StringRef 3391 g_objc_v2_prefix_class( 3392 "_OBJC_CLASS_$_"); 3393 static const llvm::StringRef 3394 g_objc_v2_prefix_metaclass( 3395 "_OBJC_METACLASS_$_"); 3396 static const llvm::StringRef 3397 g_objc_v2_prefix_ivar( 3398 "_OBJC_IVAR_$_"); 3399 if (symbol_name_ref.startswith( 3400 g_objc_v2_prefix_class)) { 3401 symbol_name_non_abi_mangled = 3402 symbol_name + 1; 3403 symbol_name = 3404 symbol_name + 3405 g_objc_v2_prefix_class.size(); 3406 type = eSymbolTypeObjCClass; 3407 demangled_is_synthesized = true; 3408 } else if ( 3409 symbol_name_ref.startswith( 3410 g_objc_v2_prefix_metaclass)) { 3411 symbol_name_non_abi_mangled = 3412 symbol_name + 1; 3413 symbol_name = 3414 symbol_name + 3415 g_objc_v2_prefix_metaclass.size(); 3416 type = eSymbolTypeObjCMetaClass; 3417 demangled_is_synthesized = true; 3418 } else if (symbol_name_ref.startswith( 3419 g_objc_v2_prefix_ivar)) { 3420 symbol_name_non_abi_mangled = 3421 symbol_name + 1; 3422 symbol_name = 3423 symbol_name + 3424 g_objc_v2_prefix_ivar.size(); 3425 type = eSymbolTypeObjCIVar; 3426 demangled_is_synthesized = true; 3427 } 3428 } 3429 } 3430 } else if (symbol_sect_name && 3431 ::strstr(symbol_sect_name, 3432 "__gcc_except_tab") == 3433 symbol_sect_name) { 3434 type = eSymbolTypeException; 3435 } else { 3436 type = eSymbolTypeData; 3437 } 3438 } else if (symbol_sect_name && 3439 ::strstr(symbol_sect_name, 3440 "__IMPORT") == 3441 symbol_sect_name) { 3442 type = eSymbolTypeTrampoline; 3443 } else if (symbol_section->IsDescendant( 3444 objc_section_sp.get())) { 3445 type = eSymbolTypeRuntime; 3446 if (symbol_name && symbol_name[0] == '.') { 3447 llvm::StringRef symbol_name_ref(symbol_name); 3448 static const llvm::StringRef 3449 g_objc_v1_prefix_class( 3450 ".objc_class_name_"); 3451 if (symbol_name_ref.startswith( 3452 g_objc_v1_prefix_class)) { 3453 symbol_name_non_abi_mangled = symbol_name; 3454 symbol_name = symbol_name + 3455 g_objc_v1_prefix_class.size(); 3456 type = eSymbolTypeObjCClass; 3457 demangled_is_synthesized = true; 3458 } 3459 } 3460 } 3461 } 3462 } 3463 } break; 3464 } 3465 } 3466 3467 if (add_nlist) { 3468 uint64_t symbol_value = nlist.n_value; 3469 if (symbol_name_non_abi_mangled) { 3470 sym[sym_idx].GetMangled().SetMangledName( 3471 ConstString(symbol_name_non_abi_mangled)); 3472 sym[sym_idx].GetMangled().SetDemangledName( 3473 ConstString(symbol_name)); 3474 } else { 3475 bool symbol_name_is_mangled = false; 3476 3477 if (symbol_name && symbol_name[0] == '_') { 3478 symbol_name_is_mangled = symbol_name[1] == '_'; 3479 symbol_name++; // Skip the leading underscore 3480 } 3481 3482 if (symbol_name) { 3483 ConstString const_symbol_name(symbol_name); 3484 sym[sym_idx].GetMangled().SetValue( 3485 const_symbol_name, symbol_name_is_mangled); 3486 if (is_gsym && is_debug) { 3487 const char *gsym_name = 3488 sym[sym_idx] 3489 .GetMangled() 3490 .GetName(lldb::eLanguageTypeUnknown, 3491 Mangled::ePreferMangled) 3492 .GetCString(); 3493 if (gsym_name) 3494 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3495 } 3496 } 3497 } 3498 if (symbol_section) { 3499 const addr_t section_file_addr = 3500 symbol_section->GetFileAddress(); 3501 if (symbol_byte_size == 0 && 3502 function_starts_count > 0) { 3503 addr_t symbol_lookup_file_addr = nlist.n_value; 3504 // Do an exact address match for non-ARM addresses, 3505 // else get the closest since the symbol might be a 3506 // thumb symbol which has an address with bit zero 3507 // set 3508 FunctionStarts::Entry *func_start_entry = 3509 function_starts.FindEntry( 3510 symbol_lookup_file_addr, !is_arm); 3511 if (is_arm && func_start_entry) { 3512 // Verify that the function start address is the 3513 // symbol address (ARM) or the symbol address + 1 3514 // (thumb) 3515 if (func_start_entry->addr != 3516 symbol_lookup_file_addr && 3517 func_start_entry->addr != 3518 (symbol_lookup_file_addr + 1)) { 3519 // Not the right entry, NULL it out... 3520 func_start_entry = NULL; 3521 } 3522 } 3523 if (func_start_entry) { 3524 func_start_entry->data = true; 3525 3526 addr_t symbol_file_addr = func_start_entry->addr; 3527 uint32_t symbol_flags = 0; 3528 if (is_arm) { 3529 if (symbol_file_addr & 1) 3530 symbol_flags = 3531 MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3532 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3533 } 3534 3535 const FunctionStarts::Entry 3536 *next_func_start_entry = 3537 function_starts.FindNextEntry( 3538 func_start_entry); 3539 const addr_t section_end_file_addr = 3540 section_file_addr + 3541 symbol_section->GetByteSize(); 3542 if (next_func_start_entry) { 3543 addr_t next_symbol_file_addr = 3544 next_func_start_entry->addr; 3545 // Be sure the clear the Thumb address bit when 3546 // we calculate the size from the current and 3547 // next address 3548 if (is_arm) 3549 next_symbol_file_addr &= 3550 THUMB_ADDRESS_BIT_MASK; 3551 symbol_byte_size = std::min<lldb::addr_t>( 3552 next_symbol_file_addr - symbol_file_addr, 3553 section_end_file_addr - symbol_file_addr); 3554 } else { 3555 symbol_byte_size = 3556 section_end_file_addr - symbol_file_addr; 3557 } 3558 } 3559 } 3560 symbol_value -= section_file_addr; 3561 } 3562 3563 if (is_debug == false) { 3564 if (type == eSymbolTypeCode) { 3565 // See if we can find a N_FUN entry for any code 3566 // symbols. If we do find a match, and the name 3567 // matches, then we can merge the two into just the 3568 // function symbol to avoid duplicate entries in 3569 // the symbol table 3570 std::pair<ValueToSymbolIndexMap::const_iterator, 3571 ValueToSymbolIndexMap::const_iterator> 3572 range; 3573 range = N_FUN_addr_to_sym_idx.equal_range( 3574 nlist.n_value); 3575 if (range.first != range.second) { 3576 bool found_it = false; 3577 for (ValueToSymbolIndexMap::const_iterator pos = 3578 range.first; 3579 pos != range.second; ++pos) { 3580 if (sym[sym_idx].GetMangled().GetName( 3581 lldb::eLanguageTypeUnknown, 3582 Mangled::ePreferMangled) == 3583 sym[pos->second].GetMangled().GetName( 3584 lldb::eLanguageTypeUnknown, 3585 Mangled::ePreferMangled)) { 3586 m_nlist_idx_to_sym_idx[nlist_idx] = 3587 pos->second; 3588 // We just need the flags from the linker 3589 // symbol, so put these flags 3590 // into the N_FUN flags to avoid duplicate 3591 // symbols in the symbol table 3592 sym[pos->second].SetExternal( 3593 sym[sym_idx].IsExternal()); 3594 sym[pos->second].SetFlags(nlist.n_type << 16 | 3595 nlist.n_desc); 3596 if (resolver_addresses.find(nlist.n_value) != 3597 resolver_addresses.end()) 3598 sym[pos->second].SetType( 3599 eSymbolTypeResolver); 3600 sym[sym_idx].Clear(); 3601 found_it = true; 3602 break; 3603 } 3604 } 3605 if (found_it) 3606 continue; 3607 } else { 3608 if (resolver_addresses.find(nlist.n_value) != 3609 resolver_addresses.end()) 3610 type = eSymbolTypeResolver; 3611 } 3612 } else if (type == eSymbolTypeData || 3613 type == eSymbolTypeObjCClass || 3614 type == eSymbolTypeObjCMetaClass || 3615 type == eSymbolTypeObjCIVar) { 3616 // See if we can find a N_STSYM entry for any data 3617 // symbols. If we do find a match, and the name 3618 // matches, then we can merge the two into just the 3619 // Static symbol to avoid duplicate entries in the 3620 // symbol table 3621 std::pair<ValueToSymbolIndexMap::const_iterator, 3622 ValueToSymbolIndexMap::const_iterator> 3623 range; 3624 range = N_STSYM_addr_to_sym_idx.equal_range( 3625 nlist.n_value); 3626 if (range.first != range.second) { 3627 bool found_it = false; 3628 for (ValueToSymbolIndexMap::const_iterator pos = 3629 range.first; 3630 pos != range.second; ++pos) { 3631 if (sym[sym_idx].GetMangled().GetName( 3632 lldb::eLanguageTypeUnknown, 3633 Mangled::ePreferMangled) == 3634 sym[pos->second].GetMangled().GetName( 3635 lldb::eLanguageTypeUnknown, 3636 Mangled::ePreferMangled)) { 3637 m_nlist_idx_to_sym_idx[nlist_idx] = 3638 pos->second; 3639 // We just need the flags from the linker 3640 // symbol, so put these flags 3641 // into the N_STSYM flags to avoid duplicate 3642 // symbols in the symbol table 3643 sym[pos->second].SetExternal( 3644 sym[sym_idx].IsExternal()); 3645 sym[pos->second].SetFlags(nlist.n_type << 16 | 3646 nlist.n_desc); 3647 sym[sym_idx].Clear(); 3648 found_it = true; 3649 break; 3650 } 3651 } 3652 if (found_it) 3653 continue; 3654 } else { 3655 const char *gsym_name = 3656 sym[sym_idx] 3657 .GetMangled() 3658 .GetName(lldb::eLanguageTypeUnknown, 3659 Mangled::ePreferMangled) 3660 .GetCString(); 3661 if (gsym_name) { 3662 // Combine N_GSYM stab entries with the non 3663 // stab symbol 3664 ConstNameToSymbolIndexMap::const_iterator pos = 3665 N_GSYM_name_to_sym_idx.find(gsym_name); 3666 if (pos != N_GSYM_name_to_sym_idx.end()) { 3667 const uint32_t GSYM_sym_idx = pos->second; 3668 m_nlist_idx_to_sym_idx[nlist_idx] = 3669 GSYM_sym_idx; 3670 // Copy the address, because often the N_GSYM 3671 // address has an invalid address of zero 3672 // when the global is a common symbol 3673 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3674 symbol_section); 3675 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3676 symbol_value); 3677 // We just need the flags from the linker 3678 // symbol, so put these flags 3679 // into the N_GSYM flags to avoid duplicate 3680 // symbols in the symbol table 3681 sym[GSYM_sym_idx].SetFlags( 3682 nlist.n_type << 16 | nlist.n_desc); 3683 sym[sym_idx].Clear(); 3684 continue; 3685 } 3686 } 3687 } 3688 } 3689 } 3690 3691 sym[sym_idx].SetID(nlist_idx); 3692 sym[sym_idx].SetType(type); 3693 if (set_value) { 3694 sym[sym_idx].GetAddressRef().SetSection( 3695 symbol_section); 3696 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3697 } 3698 sym[sym_idx].SetFlags(nlist.n_type << 16 | 3699 nlist.n_desc); 3700 3701 if (symbol_byte_size > 0) 3702 sym[sym_idx].SetByteSize(symbol_byte_size); 3703 3704 if (demangled_is_synthesized) 3705 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3706 ++sym_idx; 3707 } else { 3708 sym[sym_idx].Clear(); 3709 } 3710 } 3711 ///////////////////////////// 3712 } 3713 break; // No more entries to consider 3714 } 3715 } 3716 3717 for (const auto &pos : reexport_shlib_needs_fixup) { 3718 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3719 if (undef_pos != undefined_name_to_desc.end()) { 3720 const uint8_t dylib_ordinal = 3721 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3722 if (dylib_ordinal > 0 && 3723 dylib_ordinal < dylib_files.GetSize()) 3724 sym[pos.first].SetReExportedSymbolSharedLibrary( 3725 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3726 } 3727 } 3728 } 3729 } 3730 } 3731 } 3732 } 3733 3734 // Must reset this in case it was mutated above! 3735 nlist_data_offset = 0; 3736 #endif 3737 3738 if (nlist_data.GetByteSize() > 0) { 3739 3740 // If the sym array was not created while parsing the DSC unmapped 3741 // symbols, create it now. 3742 if (sym == NULL) { 3743 sym = symtab->Resize(symtab_load_command.nsyms + 3744 m_dysymtab.nindirectsyms); 3745 num_syms = symtab->GetNumSymbols(); 3746 } 3747 3748 if (unmapped_local_symbols_found) { 3749 assert(m_dysymtab.ilocalsym == 0); 3750 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3751 nlist_idx = m_dysymtab.nlocalsym; 3752 } else { 3753 nlist_idx = 0; 3754 } 3755 3756 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3757 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3758 UndefinedNameToDescMap undefined_name_to_desc; 3759 SymbolIndexToName reexport_shlib_needs_fixup; 3760 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 3761 struct nlist_64 nlist; 3762 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, 3763 nlist_byte_size)) 3764 break; 3765 3766 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3767 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 3768 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 3769 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 3770 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 3771 3772 SymbolType type = eSymbolTypeInvalid; 3773 const char *symbol_name = NULL; 3774 3775 if (have_strtab_data) { 3776 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3777 3778 if (symbol_name == NULL) { 3779 // No symbol should be NULL, even the symbols with no string values 3780 // should have an offset zero which points to an empty C-string 3781 Host::SystemLog(Host::eSystemLogError, 3782 "error: symbol[%u] has invalid string table offset " 3783 "0x%x in %s, ignoring symbol\n", 3784 nlist_idx, nlist.n_strx, 3785 module_sp->GetFileSpec().GetPath().c_str()); 3786 continue; 3787 } 3788 if (symbol_name[0] == '\0') 3789 symbol_name = NULL; 3790 } else { 3791 const addr_t str_addr = strtab_addr + nlist.n_strx; 3792 Status str_error; 3793 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3794 str_error)) 3795 symbol_name = memory_symbol_name.c_str(); 3796 } 3797 const char *symbol_name_non_abi_mangled = NULL; 3798 3799 SectionSP symbol_section; 3800 lldb::addr_t symbol_byte_size = 0; 3801 bool add_nlist = true; 3802 bool is_gsym = false; 3803 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3804 bool demangled_is_synthesized = false; 3805 bool set_value = true; 3806 assert(sym_idx < num_syms); 3807 3808 sym[sym_idx].SetDebug(is_debug); 3809 3810 if (is_debug) { 3811 switch (nlist.n_type) { 3812 case N_GSYM: 3813 // global symbol: name,,NO_SECT,type,0 3814 // Sometimes the N_GSYM value contains the address. 3815 3816 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3817 // the ObjC data. They 3818 // have the same address, but we want to ensure that we always find 3819 // only the real symbol, 'cause we don't currently correctly 3820 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3821 // type. This is a temporary hack to make sure the ObjectiveC 3822 // symbols get treated correctly. To do this right, we should 3823 // coalesce all the GSYM & global symbols that have the same 3824 // address. 3825 is_gsym = true; 3826 sym[sym_idx].SetExternal(true); 3827 3828 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3829 llvm::StringRef symbol_name_ref(symbol_name); 3830 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3831 symbol_name_non_abi_mangled = symbol_name + 1; 3832 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3833 type = eSymbolTypeObjCClass; 3834 demangled_is_synthesized = true; 3835 3836 } else if (symbol_name_ref.startswith( 3837 g_objc_v2_prefix_metaclass)) { 3838 symbol_name_non_abi_mangled = symbol_name + 1; 3839 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3840 type = eSymbolTypeObjCMetaClass; 3841 demangled_is_synthesized = true; 3842 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3843 symbol_name_non_abi_mangled = symbol_name + 1; 3844 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3845 type = eSymbolTypeObjCIVar; 3846 demangled_is_synthesized = true; 3847 } 3848 } else { 3849 if (nlist.n_value != 0) 3850 symbol_section = 3851 section_info.GetSection(nlist.n_sect, nlist.n_value); 3852 type = eSymbolTypeData; 3853 } 3854 break; 3855 3856 case N_FNAME: 3857 // procedure name (f77 kludge): name,,NO_SECT,0,0 3858 type = eSymbolTypeCompiler; 3859 break; 3860 3861 case N_FUN: 3862 // procedure: name,,n_sect,linenumber,address 3863 if (symbol_name) { 3864 type = eSymbolTypeCode; 3865 symbol_section = 3866 section_info.GetSection(nlist.n_sect, nlist.n_value); 3867 3868 N_FUN_addr_to_sym_idx.insert( 3869 std::make_pair(nlist.n_value, sym_idx)); 3870 // We use the current number of symbols in the symbol table in 3871 // lieu of using nlist_idx in case we ever start trimming entries 3872 // out 3873 N_FUN_indexes.push_back(sym_idx); 3874 } else { 3875 type = eSymbolTypeCompiler; 3876 3877 if (!N_FUN_indexes.empty()) { 3878 // Copy the size of the function into the original STAB entry 3879 // so we don't have to hunt for it later 3880 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3881 ->SetByteSize(nlist.n_value); 3882 N_FUN_indexes.pop_back(); 3883 // We don't really need the end function STAB as it contains 3884 // the size which we already placed with the original symbol, 3885 // so don't add it if we want a minimal symbol table 3886 add_nlist = false; 3887 } 3888 } 3889 break; 3890 3891 case N_STSYM: 3892 // static symbol: name,,n_sect,type,address 3893 N_STSYM_addr_to_sym_idx.insert( 3894 std::make_pair(nlist.n_value, sym_idx)); 3895 symbol_section = 3896 section_info.GetSection(nlist.n_sect, nlist.n_value); 3897 if (symbol_name && symbol_name[0]) { 3898 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3899 eSymbolTypeData); 3900 } 3901 break; 3902 3903 case N_LCSYM: 3904 // .lcomm symbol: name,,n_sect,type,address 3905 symbol_section = 3906 section_info.GetSection(nlist.n_sect, nlist.n_value); 3907 type = eSymbolTypeCommonBlock; 3908 break; 3909 3910 case N_BNSYM: 3911 // We use the current number of symbols in the symbol table in lieu 3912 // of using nlist_idx in case we ever start trimming entries out 3913 // Skip these if we want minimal symbol tables 3914 add_nlist = false; 3915 break; 3916 3917 case N_ENSYM: 3918 // Set the size of the N_BNSYM to the terminating index of this 3919 // N_ENSYM so that we can always skip the entire symbol if we need 3920 // to navigate more quickly at the source level when parsing STABS 3921 // Skip these if we want minimal symbol tables 3922 add_nlist = false; 3923 break; 3924 3925 case N_OPT: 3926 // emitted with gcc2_compiled and in gcc source 3927 type = eSymbolTypeCompiler; 3928 break; 3929 3930 case N_RSYM: 3931 // register sym: name,,NO_SECT,type,register 3932 type = eSymbolTypeVariable; 3933 break; 3934 3935 case N_SLINE: 3936 // src line: 0,,n_sect,linenumber,address 3937 symbol_section = 3938 section_info.GetSection(nlist.n_sect, nlist.n_value); 3939 type = eSymbolTypeLineEntry; 3940 break; 3941 3942 case N_SSYM: 3943 // structure elt: name,,NO_SECT,type,struct_offset 3944 type = eSymbolTypeVariableType; 3945 break; 3946 3947 case N_SO: 3948 // source file name 3949 type = eSymbolTypeSourceFile; 3950 if (symbol_name == NULL) { 3951 add_nlist = false; 3952 if (N_SO_index != UINT32_MAX) { 3953 // Set the size of the N_SO to the terminating index of this 3954 // N_SO so that we can always skip the entire N_SO if we need 3955 // to navigate more quickly at the source level when parsing 3956 // STABS 3957 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3958 symbol_ptr->SetByteSize(sym_idx); 3959 symbol_ptr->SetSizeIsSibling(true); 3960 } 3961 N_NSYM_indexes.clear(); 3962 N_INCL_indexes.clear(); 3963 N_BRAC_indexes.clear(); 3964 N_COMM_indexes.clear(); 3965 N_FUN_indexes.clear(); 3966 N_SO_index = UINT32_MAX; 3967 } else { 3968 // We use the current number of symbols in the symbol table in 3969 // lieu of using nlist_idx in case we ever start trimming entries 3970 // out 3971 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3972 if (N_SO_has_full_path) { 3973 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3974 // We have two consecutive N_SO entries where the first 3975 // contains a directory and the second contains a full path. 3976 sym[sym_idx - 1].GetMangled().SetValue( 3977 ConstString(symbol_name), false); 3978 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3979 add_nlist = false; 3980 } else { 3981 // This is the first entry in a N_SO that contains a 3982 // directory or a full path to the source file 3983 N_SO_index = sym_idx; 3984 } 3985 } else if ((N_SO_index == sym_idx - 1) && 3986 ((sym_idx - 1) < num_syms)) { 3987 // This is usually the second N_SO entry that contains just the 3988 // filename, so here we combine it with the first one if we are 3989 // minimizing the symbol table 3990 const char *so_path = 3991 sym[sym_idx - 1] 3992 .GetMangled() 3993 .GetDemangledName(lldb::eLanguageTypeUnknown) 3994 .AsCString(); 3995 if (so_path && so_path[0]) { 3996 std::string full_so_path(so_path); 3997 const size_t double_slash_pos = full_so_path.find("//"); 3998 if (double_slash_pos != std::string::npos) { 3999 // The linker has been generating bad N_SO entries with 4000 // doubled up paths in the format "%s%s" where the first 4001 // string in the DW_AT_comp_dir, and the second is the 4002 // directory for the source file so you end up with a path 4003 // that looks like "/tmp/src//tmp/src/" 4004 FileSpec so_dir(so_path, false); 4005 if (!so_dir.Exists()) { 4006 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false, 4007 FileSpec::Style::native); 4008 if (so_dir.Exists()) { 4009 // Trim off the incorrect path 4010 full_so_path.erase(0, double_slash_pos + 1); 4011 } 4012 } 4013 } 4014 if (*full_so_path.rbegin() != '/') 4015 full_so_path += '/'; 4016 full_so_path += symbol_name; 4017 sym[sym_idx - 1].GetMangled().SetValue( 4018 ConstString(full_so_path.c_str()), false); 4019 add_nlist = false; 4020 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4021 } 4022 } else { 4023 // This could be a relative path to a N_SO 4024 N_SO_index = sym_idx; 4025 } 4026 } 4027 break; 4028 4029 case N_OSO: 4030 // object file name: name,,0,0,st_mtime 4031 type = eSymbolTypeObjectFile; 4032 break; 4033 4034 case N_LSYM: 4035 // local sym: name,,NO_SECT,type,offset 4036 type = eSymbolTypeLocal; 4037 break; 4038 4039 //---------------------------------------------------------------------- 4040 // INCL scopes 4041 //---------------------------------------------------------------------- 4042 case N_BINCL: 4043 // include file beginning: name,,NO_SECT,0,sum We use the current 4044 // number of symbols in the symbol table in lieu of using nlist_idx 4045 // in case we ever start trimming entries out 4046 N_INCL_indexes.push_back(sym_idx); 4047 type = eSymbolTypeScopeBegin; 4048 break; 4049 4050 case N_EINCL: 4051 // include file end: name,,NO_SECT,0,0 4052 // Set the size of the N_BINCL to the terminating index of this 4053 // N_EINCL so that we can always skip the entire symbol if we need 4054 // to navigate more quickly at the source level when parsing STABS 4055 if (!N_INCL_indexes.empty()) { 4056 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4057 symbol_ptr->SetByteSize(sym_idx + 1); 4058 symbol_ptr->SetSizeIsSibling(true); 4059 N_INCL_indexes.pop_back(); 4060 } 4061 type = eSymbolTypeScopeEnd; 4062 break; 4063 4064 case N_SOL: 4065 // #included file name: name,,n_sect,0,address 4066 type = eSymbolTypeHeaderFile; 4067 4068 // We currently don't use the header files on darwin 4069 add_nlist = false; 4070 break; 4071 4072 case N_PARAMS: 4073 // compiler parameters: name,,NO_SECT,0,0 4074 type = eSymbolTypeCompiler; 4075 break; 4076 4077 case N_VERSION: 4078 // compiler version: name,,NO_SECT,0,0 4079 type = eSymbolTypeCompiler; 4080 break; 4081 4082 case N_OLEVEL: 4083 // compiler -O level: name,,NO_SECT,0,0 4084 type = eSymbolTypeCompiler; 4085 break; 4086 4087 case N_PSYM: 4088 // parameter: name,,NO_SECT,type,offset 4089 type = eSymbolTypeVariable; 4090 break; 4091 4092 case N_ENTRY: 4093 // alternate entry: name,,n_sect,linenumber,address 4094 symbol_section = 4095 section_info.GetSection(nlist.n_sect, nlist.n_value); 4096 type = eSymbolTypeLineEntry; 4097 break; 4098 4099 //---------------------------------------------------------------------- 4100 // Left and Right Braces 4101 //---------------------------------------------------------------------- 4102 case N_LBRAC: 4103 // left bracket: 0,,NO_SECT,nesting level,address We use the 4104 // current number of symbols in the symbol table in lieu of using 4105 // nlist_idx in case we ever start trimming entries out 4106 symbol_section = 4107 section_info.GetSection(nlist.n_sect, nlist.n_value); 4108 N_BRAC_indexes.push_back(sym_idx); 4109 type = eSymbolTypeScopeBegin; 4110 break; 4111 4112 case N_RBRAC: 4113 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4114 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4115 // can always skip the entire symbol if we need to navigate more 4116 // quickly at the source level when parsing STABS 4117 symbol_section = 4118 section_info.GetSection(nlist.n_sect, nlist.n_value); 4119 if (!N_BRAC_indexes.empty()) { 4120 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4121 symbol_ptr->SetByteSize(sym_idx + 1); 4122 symbol_ptr->SetSizeIsSibling(true); 4123 N_BRAC_indexes.pop_back(); 4124 } 4125 type = eSymbolTypeScopeEnd; 4126 break; 4127 4128 case N_EXCL: 4129 // deleted include file: name,,NO_SECT,0,sum 4130 type = eSymbolTypeHeaderFile; 4131 break; 4132 4133 //---------------------------------------------------------------------- 4134 // COMM scopes 4135 //---------------------------------------------------------------------- 4136 case N_BCOMM: 4137 // begin common: name,,NO_SECT,0,0 4138 // We use the current number of symbols in the symbol table in lieu 4139 // of using nlist_idx in case we ever start trimming entries out 4140 type = eSymbolTypeScopeBegin; 4141 N_COMM_indexes.push_back(sym_idx); 4142 break; 4143 4144 case N_ECOML: 4145 // end common (local name): 0,,n_sect,0,address 4146 symbol_section = 4147 section_info.GetSection(nlist.n_sect, nlist.n_value); 4148 LLVM_FALLTHROUGH; 4149 4150 case N_ECOMM: 4151 // end common: name,,n_sect,0,0 4152 // Set the size of the N_BCOMM to the terminating index of this 4153 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4154 // we need to navigate more quickly at the source level when 4155 // parsing STABS 4156 if (!N_COMM_indexes.empty()) { 4157 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4158 symbol_ptr->SetByteSize(sym_idx + 1); 4159 symbol_ptr->SetSizeIsSibling(true); 4160 N_COMM_indexes.pop_back(); 4161 } 4162 type = eSymbolTypeScopeEnd; 4163 break; 4164 4165 case N_LENG: 4166 // second stab entry with length information 4167 type = eSymbolTypeAdditional; 4168 break; 4169 4170 default: 4171 break; 4172 } 4173 } else { 4174 // uint8_t n_pext = N_PEXT & nlist.n_type; 4175 uint8_t n_type = N_TYPE & nlist.n_type; 4176 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4177 4178 switch (n_type) { 4179 case N_INDR: { 4180 const char *reexport_name_cstr = 4181 strtab_data.PeekCStr(nlist.n_value); 4182 if (reexport_name_cstr && reexport_name_cstr[0]) { 4183 type = eSymbolTypeReExported; 4184 ConstString reexport_name( 4185 reexport_name_cstr + 4186 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4187 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4188 set_value = false; 4189 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4190 indirect_symbol_names.insert( 4191 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4192 } else 4193 type = eSymbolTypeUndefined; 4194 } break; 4195 4196 case N_UNDF: 4197 if (symbol_name && symbol_name[0]) { 4198 ConstString undefined_name(symbol_name + 4199 ((symbol_name[0] == '_') ? 1 : 0)); 4200 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4201 } 4202 LLVM_FALLTHROUGH; 4203 4204 case N_PBUD: 4205 type = eSymbolTypeUndefined; 4206 break; 4207 4208 case N_ABS: 4209 type = eSymbolTypeAbsolute; 4210 break; 4211 4212 case N_SECT: { 4213 symbol_section = 4214 section_info.GetSection(nlist.n_sect, nlist.n_value); 4215 4216 if (!symbol_section) { 4217 // TODO: warn about this? 4218 add_nlist = false; 4219 break; 4220 } 4221 4222 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4223 type = eSymbolTypeException; 4224 } else { 4225 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4226 4227 switch (section_type) { 4228 case S_CSTRING_LITERALS: 4229 type = eSymbolTypeData; 4230 break; // section with only literal C strings 4231 case S_4BYTE_LITERALS: 4232 type = eSymbolTypeData; 4233 break; // section with only 4 byte literals 4234 case S_8BYTE_LITERALS: 4235 type = eSymbolTypeData; 4236 break; // section with only 8 byte literals 4237 case S_LITERAL_POINTERS: 4238 type = eSymbolTypeTrampoline; 4239 break; // section with only pointers to literals 4240 case S_NON_LAZY_SYMBOL_POINTERS: 4241 type = eSymbolTypeTrampoline; 4242 break; // section with only non-lazy symbol pointers 4243 case S_LAZY_SYMBOL_POINTERS: 4244 type = eSymbolTypeTrampoline; 4245 break; // section with only lazy symbol pointers 4246 case S_SYMBOL_STUBS: 4247 type = eSymbolTypeTrampoline; 4248 break; // section with only symbol stubs, byte size of stub in 4249 // the reserved2 field 4250 case S_MOD_INIT_FUNC_POINTERS: 4251 type = eSymbolTypeCode; 4252 break; // section with only function pointers for initialization 4253 case S_MOD_TERM_FUNC_POINTERS: 4254 type = eSymbolTypeCode; 4255 break; // section with only function pointers for termination 4256 case S_INTERPOSING: 4257 type = eSymbolTypeTrampoline; 4258 break; // section with only pairs of function pointers for 4259 // interposing 4260 case S_16BYTE_LITERALS: 4261 type = eSymbolTypeData; 4262 break; // section with only 16 byte literals 4263 case S_DTRACE_DOF: 4264 type = eSymbolTypeInstrumentation; 4265 break; 4266 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4267 type = eSymbolTypeTrampoline; 4268 break; 4269 default: 4270 switch (symbol_section->GetType()) { 4271 case lldb::eSectionTypeCode: 4272 type = eSymbolTypeCode; 4273 break; 4274 case eSectionTypeData: 4275 case eSectionTypeDataCString: // Inlined C string data 4276 case eSectionTypeDataCStringPointers: // Pointers to C string 4277 // data 4278 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4279 // the symbol table 4280 case eSectionTypeData4: 4281 case eSectionTypeData8: 4282 case eSectionTypeData16: 4283 type = eSymbolTypeData; 4284 break; 4285 default: 4286 break; 4287 } 4288 break; 4289 } 4290 4291 if (type == eSymbolTypeInvalid) { 4292 const char *symbol_sect_name = 4293 symbol_section->GetName().AsCString(); 4294 if (symbol_section->IsDescendant(text_section_sp.get())) { 4295 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4296 S_ATTR_SELF_MODIFYING_CODE | 4297 S_ATTR_SOME_INSTRUCTIONS)) 4298 type = eSymbolTypeData; 4299 else 4300 type = eSymbolTypeCode; 4301 } else if (symbol_section->IsDescendant( 4302 data_section_sp.get()) || 4303 symbol_section->IsDescendant( 4304 data_dirty_section_sp.get()) || 4305 symbol_section->IsDescendant( 4306 data_const_section_sp.get())) { 4307 if (symbol_sect_name && 4308 ::strstr(symbol_sect_name, "__objc") == 4309 symbol_sect_name) { 4310 type = eSymbolTypeRuntime; 4311 4312 if (symbol_name) { 4313 llvm::StringRef symbol_name_ref(symbol_name); 4314 if (symbol_name_ref.startswith("_OBJC_")) { 4315 static const llvm::StringRef g_objc_v2_prefix_class( 4316 "_OBJC_CLASS_$_"); 4317 static const llvm::StringRef g_objc_v2_prefix_metaclass( 4318 "_OBJC_METACLASS_$_"); 4319 static const llvm::StringRef g_objc_v2_prefix_ivar( 4320 "_OBJC_IVAR_$_"); 4321 if (symbol_name_ref.startswith( 4322 g_objc_v2_prefix_class)) { 4323 symbol_name_non_abi_mangled = symbol_name + 1; 4324 symbol_name = 4325 symbol_name + g_objc_v2_prefix_class.size(); 4326 type = eSymbolTypeObjCClass; 4327 demangled_is_synthesized = true; 4328 } else if (symbol_name_ref.startswith( 4329 g_objc_v2_prefix_metaclass)) { 4330 symbol_name_non_abi_mangled = symbol_name + 1; 4331 symbol_name = 4332 symbol_name + g_objc_v2_prefix_metaclass.size(); 4333 type = eSymbolTypeObjCMetaClass; 4334 demangled_is_synthesized = true; 4335 } else if (symbol_name_ref.startswith( 4336 g_objc_v2_prefix_ivar)) { 4337 symbol_name_non_abi_mangled = symbol_name + 1; 4338 symbol_name = 4339 symbol_name + g_objc_v2_prefix_ivar.size(); 4340 type = eSymbolTypeObjCIVar; 4341 demangled_is_synthesized = true; 4342 } 4343 } 4344 } 4345 } else if (symbol_sect_name && 4346 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4347 symbol_sect_name) { 4348 type = eSymbolTypeException; 4349 } else { 4350 type = eSymbolTypeData; 4351 } 4352 } else if (symbol_sect_name && 4353 ::strstr(symbol_sect_name, "__IMPORT") == 4354 symbol_sect_name) { 4355 type = eSymbolTypeTrampoline; 4356 } else if (symbol_section->IsDescendant( 4357 objc_section_sp.get())) { 4358 type = eSymbolTypeRuntime; 4359 if (symbol_name && symbol_name[0] == '.') { 4360 llvm::StringRef symbol_name_ref(symbol_name); 4361 static const llvm::StringRef g_objc_v1_prefix_class( 4362 ".objc_class_name_"); 4363 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4364 symbol_name_non_abi_mangled = symbol_name; 4365 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4366 type = eSymbolTypeObjCClass; 4367 demangled_is_synthesized = true; 4368 } 4369 } 4370 } 4371 } 4372 } 4373 } break; 4374 } 4375 } 4376 4377 if (add_nlist) { 4378 uint64_t symbol_value = nlist.n_value; 4379 4380 if (symbol_name_non_abi_mangled) { 4381 sym[sym_idx].GetMangled().SetMangledName( 4382 ConstString(symbol_name_non_abi_mangled)); 4383 sym[sym_idx].GetMangled().SetDemangledName( 4384 ConstString(symbol_name)); 4385 } else { 4386 bool symbol_name_is_mangled = false; 4387 4388 if (symbol_name && symbol_name[0] == '_') { 4389 symbol_name_is_mangled = symbol_name[1] == '_'; 4390 symbol_name++; // Skip the leading underscore 4391 } 4392 4393 if (symbol_name) { 4394 ConstString const_symbol_name(symbol_name); 4395 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4396 symbol_name_is_mangled); 4397 } 4398 } 4399 4400 if (is_gsym) { 4401 const char *gsym_name = sym[sym_idx] 4402 .GetMangled() 4403 .GetName(lldb::eLanguageTypeUnknown, 4404 Mangled::ePreferMangled) 4405 .GetCString(); 4406 if (gsym_name) 4407 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4408 } 4409 4410 if (symbol_section) { 4411 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4412 if (symbol_byte_size == 0 && function_starts_count > 0) { 4413 addr_t symbol_lookup_file_addr = nlist.n_value; 4414 // Do an exact address match for non-ARM addresses, else get the 4415 // closest since the symbol might be a thumb symbol which has an 4416 // address with bit zero set 4417 FunctionStarts::Entry *func_start_entry = 4418 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4419 if (is_arm && func_start_entry) { 4420 // Verify that the function start address is the symbol address 4421 // (ARM) or the symbol address + 1 (thumb) 4422 if (func_start_entry->addr != symbol_lookup_file_addr && 4423 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4424 // Not the right entry, NULL it out... 4425 func_start_entry = NULL; 4426 } 4427 } 4428 if (func_start_entry) { 4429 func_start_entry->data = true; 4430 4431 addr_t symbol_file_addr = func_start_entry->addr; 4432 if (is_arm) 4433 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4434 4435 const FunctionStarts::Entry *next_func_start_entry = 4436 function_starts.FindNextEntry(func_start_entry); 4437 const addr_t section_end_file_addr = 4438 section_file_addr + symbol_section->GetByteSize(); 4439 if (next_func_start_entry) { 4440 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4441 // Be sure the clear the Thumb address bit when we calculate 4442 // the size from the current and next address 4443 if (is_arm) 4444 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4445 symbol_byte_size = std::min<lldb::addr_t>( 4446 next_symbol_file_addr - symbol_file_addr, 4447 section_end_file_addr - symbol_file_addr); 4448 } else { 4449 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4450 } 4451 } 4452 } 4453 symbol_value -= section_file_addr; 4454 } 4455 4456 if (is_debug == false) { 4457 if (type == eSymbolTypeCode) { 4458 // See if we can find a N_FUN entry for any code symbols. If we 4459 // do find a match, and the name matches, then we can merge the 4460 // two into just the function symbol to avoid duplicate entries 4461 // in the symbol table 4462 std::pair<ValueToSymbolIndexMap::const_iterator, 4463 ValueToSymbolIndexMap::const_iterator> 4464 range; 4465 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4466 if (range.first != range.second) { 4467 bool found_it = false; 4468 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4469 pos != range.second; ++pos) { 4470 if (sym[sym_idx].GetMangled().GetName( 4471 lldb::eLanguageTypeUnknown, 4472 Mangled::ePreferMangled) == 4473 sym[pos->second].GetMangled().GetName( 4474 lldb::eLanguageTypeUnknown, 4475 Mangled::ePreferMangled)) { 4476 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4477 // We just need the flags from the linker symbol, so put 4478 // these flags into the N_FUN flags to avoid duplicate 4479 // symbols in the symbol table 4480 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4481 sym[pos->second].SetFlags(nlist.n_type << 16 | 4482 nlist.n_desc); 4483 if (resolver_addresses.find(nlist.n_value) != 4484 resolver_addresses.end()) 4485 sym[pos->second].SetType(eSymbolTypeResolver); 4486 sym[sym_idx].Clear(); 4487 found_it = true; 4488 break; 4489 } 4490 } 4491 if (found_it) 4492 continue; 4493 } else { 4494 if (resolver_addresses.find(nlist.n_value) != 4495 resolver_addresses.end()) 4496 type = eSymbolTypeResolver; 4497 } 4498 } else if (type == eSymbolTypeData || 4499 type == eSymbolTypeObjCClass || 4500 type == eSymbolTypeObjCMetaClass || 4501 type == eSymbolTypeObjCIVar) { 4502 // See if we can find a N_STSYM entry for any data symbols. If we 4503 // do find a match, and the name matches, then we can merge the 4504 // two into just the Static symbol to avoid duplicate entries in 4505 // the symbol table 4506 std::pair<ValueToSymbolIndexMap::const_iterator, 4507 ValueToSymbolIndexMap::const_iterator> 4508 range; 4509 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4510 if (range.first != range.second) { 4511 bool found_it = false; 4512 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4513 pos != range.second; ++pos) { 4514 if (sym[sym_idx].GetMangled().GetName( 4515 lldb::eLanguageTypeUnknown, 4516 Mangled::ePreferMangled) == 4517 sym[pos->second].GetMangled().GetName( 4518 lldb::eLanguageTypeUnknown, 4519 Mangled::ePreferMangled)) { 4520 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4521 // We just need the flags from the linker symbol, so put 4522 // these flags into the N_STSYM flags to avoid duplicate 4523 // symbols in the symbol table 4524 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4525 sym[pos->second].SetFlags(nlist.n_type << 16 | 4526 nlist.n_desc); 4527 sym[sym_idx].Clear(); 4528 found_it = true; 4529 break; 4530 } 4531 } 4532 if (found_it) 4533 continue; 4534 } else { 4535 // Combine N_GSYM stab entries with the non stab symbol 4536 const char *gsym_name = sym[sym_idx] 4537 .GetMangled() 4538 .GetName(lldb::eLanguageTypeUnknown, 4539 Mangled::ePreferMangled) 4540 .GetCString(); 4541 if (gsym_name) { 4542 ConstNameToSymbolIndexMap::const_iterator pos = 4543 N_GSYM_name_to_sym_idx.find(gsym_name); 4544 if (pos != N_GSYM_name_to_sym_idx.end()) { 4545 const uint32_t GSYM_sym_idx = pos->second; 4546 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4547 // Copy the address, because often the N_GSYM address has 4548 // an invalid address of zero when the global is a common 4549 // symbol 4550 sym[GSYM_sym_idx].GetAddressRef().SetSection( 4551 symbol_section); 4552 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4553 // We just need the flags from the linker symbol, so put 4554 // these flags into the N_GSYM flags to avoid duplicate 4555 // symbols in the symbol table 4556 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 4557 nlist.n_desc); 4558 sym[sym_idx].Clear(); 4559 continue; 4560 } 4561 } 4562 } 4563 } 4564 } 4565 4566 sym[sym_idx].SetID(nlist_idx); 4567 sym[sym_idx].SetType(type); 4568 if (set_value) { 4569 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4570 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4571 } 4572 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4573 4574 if (symbol_byte_size > 0) 4575 sym[sym_idx].SetByteSize(symbol_byte_size); 4576 4577 if (demangled_is_synthesized) 4578 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4579 4580 ++sym_idx; 4581 } else { 4582 sym[sym_idx].Clear(); 4583 } 4584 } 4585 4586 for (const auto &pos : reexport_shlib_needs_fixup) { 4587 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4588 if (undef_pos != undefined_name_to_desc.end()) { 4589 const uint8_t dylib_ordinal = 4590 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4591 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4592 sym[pos.first].SetReExportedSymbolSharedLibrary( 4593 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4594 } 4595 } 4596 } 4597 4598 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4599 4600 if (function_starts_count > 0) { 4601 uint32_t num_synthetic_function_symbols = 0; 4602 for (i = 0; i < function_starts_count; ++i) { 4603 if (function_starts.GetEntryRef(i).data == false) 4604 ++num_synthetic_function_symbols; 4605 } 4606 4607 if (num_synthetic_function_symbols > 0) { 4608 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4609 num_syms = sym_idx + num_synthetic_function_symbols; 4610 sym = symtab->Resize(num_syms); 4611 } 4612 for (i = 0; i < function_starts_count; ++i) { 4613 const FunctionStarts::Entry *func_start_entry = 4614 function_starts.GetEntryAtIndex(i); 4615 if (func_start_entry->data == false) { 4616 addr_t symbol_file_addr = func_start_entry->addr; 4617 uint32_t symbol_flags = 0; 4618 if (is_arm) { 4619 if (symbol_file_addr & 1) 4620 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4621 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4622 } 4623 Address symbol_addr; 4624 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4625 SectionSP symbol_section(symbol_addr.GetSection()); 4626 uint32_t symbol_byte_size = 0; 4627 if (symbol_section) { 4628 const addr_t section_file_addr = 4629 symbol_section->GetFileAddress(); 4630 const FunctionStarts::Entry *next_func_start_entry = 4631 function_starts.FindNextEntry(func_start_entry); 4632 const addr_t section_end_file_addr = 4633 section_file_addr + symbol_section->GetByteSize(); 4634 if (next_func_start_entry) { 4635 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4636 if (is_arm) 4637 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4638 symbol_byte_size = std::min<lldb::addr_t>( 4639 next_symbol_file_addr - symbol_file_addr, 4640 section_end_file_addr - symbol_file_addr); 4641 } else { 4642 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4643 } 4644 sym[sym_idx].SetID(synthetic_sym_id++); 4645 sym[sym_idx].GetMangled().SetDemangledName( 4646 GetNextSyntheticSymbolName()); 4647 sym[sym_idx].SetType(eSymbolTypeCode); 4648 sym[sym_idx].SetIsSynthetic(true); 4649 sym[sym_idx].GetAddressRef() = symbol_addr; 4650 if (symbol_flags) 4651 sym[sym_idx].SetFlags(symbol_flags); 4652 if (symbol_byte_size) 4653 sym[sym_idx].SetByteSize(symbol_byte_size); 4654 ++sym_idx; 4655 } 4656 } 4657 } 4658 } 4659 } 4660 } 4661 4662 // Trim our symbols down to just what we ended up with after removing any 4663 // symbols. 4664 if (sym_idx < num_syms) { 4665 num_syms = sym_idx; 4666 sym = symtab->Resize(num_syms); 4667 } 4668 4669 // Now synthesize indirect symbols 4670 if (m_dysymtab.nindirectsyms != 0) { 4671 if (indirect_symbol_index_data.GetByteSize()) { 4672 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4673 m_nlist_idx_to_sym_idx.end(); 4674 4675 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4676 ++sect_idx) { 4677 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4678 S_SYMBOL_STUBS) { 4679 uint32_t symbol_stub_byte_size = 4680 m_mach_sections[sect_idx].reserved2; 4681 if (symbol_stub_byte_size == 0) 4682 continue; 4683 4684 const uint32_t num_symbol_stubs = 4685 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4686 4687 if (num_symbol_stubs == 0) 4688 continue; 4689 4690 const uint32_t symbol_stub_index_offset = 4691 m_mach_sections[sect_idx].reserved1; 4692 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; 4693 ++stub_idx) { 4694 const uint32_t symbol_stub_index = 4695 symbol_stub_index_offset + stub_idx; 4696 const lldb::addr_t symbol_stub_addr = 4697 m_mach_sections[sect_idx].addr + 4698 (stub_idx * symbol_stub_byte_size); 4699 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4700 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4701 symbol_stub_offset, 4)) { 4702 const uint32_t stub_sym_id = 4703 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4704 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4705 continue; 4706 4707 NListIndexToSymbolIndexMap::const_iterator index_pos = 4708 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4709 Symbol *stub_symbol = NULL; 4710 if (index_pos != end_index_pos) { 4711 // We have a remapping from the original nlist index to a 4712 // current symbol index, so just look this up by index 4713 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4714 } else { 4715 // We need to lookup a symbol using the original nlist symbol 4716 // index since this index is coming from the S_SYMBOL_STUBS 4717 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4718 } 4719 4720 if (stub_symbol) { 4721 Address so_addr(symbol_stub_addr, section_list); 4722 4723 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4724 // Change the external symbol into a trampoline that makes 4725 // sense These symbols were N_UNDF N_EXT, and are useless 4726 // to us, so we can re-use them so we don't have to make up 4727 // a synthetic symbol for no good reason. 4728 if (resolver_addresses.find(symbol_stub_addr) == 4729 resolver_addresses.end()) 4730 stub_symbol->SetType(eSymbolTypeTrampoline); 4731 else 4732 stub_symbol->SetType(eSymbolTypeResolver); 4733 stub_symbol->SetExternal(false); 4734 stub_symbol->GetAddressRef() = so_addr; 4735 stub_symbol->SetByteSize(symbol_stub_byte_size); 4736 } else { 4737 // Make a synthetic symbol to describe the trampoline stub 4738 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4739 if (sym_idx >= num_syms) { 4740 sym = symtab->Resize(++num_syms); 4741 stub_symbol = NULL; // this pointer no longer valid 4742 } 4743 sym[sym_idx].SetID(synthetic_sym_id++); 4744 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4745 if (resolver_addresses.find(symbol_stub_addr) == 4746 resolver_addresses.end()) 4747 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4748 else 4749 sym[sym_idx].SetType(eSymbolTypeResolver); 4750 sym[sym_idx].SetIsSynthetic(true); 4751 sym[sym_idx].GetAddressRef() = so_addr; 4752 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4753 ++sym_idx; 4754 } 4755 } else { 4756 if (log) 4757 log->Warning("symbol stub referencing symbol table symbol " 4758 "%u that isn't in our minimal symbol table, " 4759 "fix this!!!", 4760 stub_sym_id); 4761 } 4762 } 4763 } 4764 } 4765 } 4766 } 4767 } 4768 4769 if (!trie_entries.empty()) { 4770 for (const auto &e : trie_entries) { 4771 if (e.entry.import_name) { 4772 // Only add indirect symbols from the Trie entries if we didn't have 4773 // a N_INDR nlist entry for this already 4774 if (indirect_symbol_names.find(e.entry.name) == 4775 indirect_symbol_names.end()) { 4776 // Make a synthetic symbol to describe re-exported symbol. 4777 if (sym_idx >= num_syms) 4778 sym = symtab->Resize(++num_syms); 4779 sym[sym_idx].SetID(synthetic_sym_id++); 4780 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4781 sym[sym_idx].SetType(eSymbolTypeReExported); 4782 sym[sym_idx].SetIsSynthetic(true); 4783 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4784 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4785 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4786 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4787 } 4788 ++sym_idx; 4789 } 4790 } 4791 } 4792 } 4793 4794 // StreamFile s(stdout, false); 4795 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4796 // symtab->Dump(&s, NULL, eSortOrderNone); 4797 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4798 // sizes 4799 symtab->CalculateSymbolSizes(); 4800 4801 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4802 // symtab->Dump(&s, NULL, eSortOrderNone); 4803 4804 return symtab->GetNumSymbols(); 4805 } 4806 return 0; 4807 } 4808 4809 void ObjectFileMachO::Dump(Stream *s) { 4810 ModuleSP module_sp(GetModule()); 4811 if (module_sp) { 4812 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4813 s->Printf("%p: ", static_cast<void *>(this)); 4814 s->Indent(); 4815 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4816 s->PutCString("ObjectFileMachO64"); 4817 else 4818 s->PutCString("ObjectFileMachO32"); 4819 4820 ArchSpec header_arch; 4821 GetArchitecture(header_arch); 4822 4823 *s << ", file = '" << m_file 4824 << "', triple = " << header_arch.GetTriple().getTriple() << "\n"; 4825 4826 SectionList *sections = GetSectionList(); 4827 if (sections) 4828 sections->Dump(s, NULL, true, UINT32_MAX); 4829 4830 if (m_symtab_ap.get()) 4831 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4832 } 4833 } 4834 4835 bool ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4836 const lldb_private::DataExtractor &data, 4837 lldb::offset_t lc_offset, 4838 lldb_private::UUID &uuid) { 4839 uint32_t i; 4840 struct uuid_command load_cmd; 4841 4842 lldb::offset_t offset = lc_offset; 4843 for (i = 0; i < header.ncmds; ++i) { 4844 const lldb::offset_t cmd_offset = offset; 4845 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4846 break; 4847 4848 if (load_cmd.cmd == LC_UUID) { 4849 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4850 4851 if (uuid_bytes) { 4852 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4853 // We pretend these object files have no UUID to prevent crashing. 4854 4855 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4856 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4857 0xbb, 0x14, 0xf0, 0x0d}; 4858 4859 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4860 return false; 4861 4862 uuid = UUID::fromOptionalData(uuid_bytes, 16); 4863 return true; 4864 } 4865 return false; 4866 } 4867 offset = cmd_offset + load_cmd.cmdsize; 4868 } 4869 return false; 4870 } 4871 4872 static const char *GetOSName(uint32_t cmd) { 4873 switch (cmd) { 4874 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4875 return "ios"; 4876 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4877 return "macosx"; 4878 case llvm::MachO::LC_VERSION_MIN_TVOS: 4879 return "tvos"; 4880 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4881 return "watchos"; 4882 default: 4883 llvm_unreachable("unexpected LC_VERSION load command"); 4884 } 4885 } 4886 4887 bool ObjectFileMachO::GetArchitecture(const llvm::MachO::mach_header &header, 4888 const lldb_private::DataExtractor &data, 4889 lldb::offset_t lc_offset, 4890 ArchSpec &arch) { 4891 arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4892 4893 if (arch.IsValid()) { 4894 llvm::Triple &triple = arch.GetTriple(); 4895 4896 // Set OS to an unspecified unknown or a "*" so it can match any OS 4897 triple.setOS(llvm::Triple::UnknownOS); 4898 triple.setOSName(llvm::StringRef()); 4899 4900 if (header.filetype == MH_PRELOAD) { 4901 if (header.cputype == CPU_TYPE_ARM) { 4902 // If this is a 32-bit arm binary, and it's a standalone binary, force 4903 // the Vendor to Apple so we don't accidentally pick up the generic 4904 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4905 // frame pointer register; most other armv7 ABIs use a combination of 4906 // r7 and r11. 4907 triple.setVendor(llvm::Triple::Apple); 4908 } else { 4909 // Set vendor to an unspecified unknown or a "*" so it can match any 4910 // vendor This is required for correct behavior of EFI debugging on 4911 // x86_64 4912 triple.setVendor(llvm::Triple::UnknownVendor); 4913 triple.setVendorName(llvm::StringRef()); 4914 } 4915 return true; 4916 } else { 4917 struct load_command load_cmd; 4918 4919 lldb::offset_t offset = lc_offset; 4920 for (uint32_t i = 0; i < header.ncmds; ++i) { 4921 const lldb::offset_t cmd_offset = offset; 4922 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4923 break; 4924 4925 uint32_t major, minor, patch; 4926 struct version_min_command version_min; 4927 4928 llvm::SmallString<16> os_name; 4929 llvm::raw_svector_ostream os(os_name); 4930 4931 switch (load_cmd.cmd) { 4932 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4933 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4934 case llvm::MachO::LC_VERSION_MIN_TVOS: 4935 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4936 if (load_cmd.cmdsize != sizeof(version_min)) 4937 break; 4938 data.ExtractBytes(cmd_offset, 4939 sizeof(version_min), data.GetByteOrder(), 4940 &version_min); 4941 major = version_min.version >> 16; 4942 minor = (version_min.version >> 8) & 0xffu; 4943 patch = version_min.version & 0xffu; 4944 os << GetOSName(load_cmd.cmd) << major << '.' << minor << '.' 4945 << patch; 4946 triple.setOSName(os.str()); 4947 return true; 4948 default: 4949 break; 4950 } 4951 4952 offset = cmd_offset + load_cmd.cmdsize; 4953 } 4954 4955 if (header.filetype != MH_KEXT_BUNDLE) { 4956 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 4957 // so lets not say our Vendor is Apple, leave it as an unspecified 4958 // unknown 4959 triple.setVendor(llvm::Triple::UnknownVendor); 4960 triple.setVendorName(llvm::StringRef()); 4961 } 4962 } 4963 } 4964 return arch.IsValid(); 4965 } 4966 4967 bool ObjectFileMachO::GetUUID(lldb_private::UUID *uuid) { 4968 ModuleSP module_sp(GetModule()); 4969 if (module_sp) { 4970 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4971 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4972 return GetUUID(m_header, m_data, offset, *uuid); 4973 } 4974 return false; 4975 } 4976 4977 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 4978 uint32_t count = 0; 4979 ModuleSP module_sp(GetModule()); 4980 if (module_sp) { 4981 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4982 struct load_command load_cmd; 4983 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4984 std::vector<std::string> rpath_paths; 4985 std::vector<std::string> rpath_relative_paths; 4986 std::vector<std::string> at_exec_relative_paths; 4987 const bool resolve_path = false; // Don't resolve the dependent file paths 4988 // since they may not reside on this 4989 // system 4990 uint32_t i; 4991 for (i = 0; i < m_header.ncmds; ++i) { 4992 const uint32_t cmd_offset = offset; 4993 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4994 break; 4995 4996 switch (load_cmd.cmd) { 4997 case LC_RPATH: 4998 case LC_LOAD_DYLIB: 4999 case LC_LOAD_WEAK_DYLIB: 5000 case LC_REEXPORT_DYLIB: 5001 case LC_LOAD_DYLINKER: 5002 case LC_LOADFVMLIB: 5003 case LC_LOAD_UPWARD_DYLIB: { 5004 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5005 const char *path = m_data.PeekCStr(name_offset); 5006 if (path) { 5007 if (load_cmd.cmd == LC_RPATH) 5008 rpath_paths.push_back(path); 5009 else { 5010 if (path[0] == '@') { 5011 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5012 rpath_relative_paths.push_back(path + strlen("@rpath")); 5013 else if (strncmp(path, "@executable_path", 5014 strlen("@executable_path")) == 0) 5015 at_exec_relative_paths.push_back(path 5016 + strlen("@executable_path")); 5017 } else { 5018 FileSpec file_spec(path, resolve_path); 5019 if (files.AppendIfUnique(file_spec)) 5020 count++; 5021 } 5022 } 5023 } 5024 } break; 5025 5026 default: 5027 break; 5028 } 5029 offset = cmd_offset + load_cmd.cmdsize; 5030 } 5031 5032 FileSpec this_file_spec(m_file); 5033 this_file_spec.ResolvePath(); 5034 5035 if (!rpath_paths.empty()) { 5036 // Fixup all LC_RPATH values to be absolute paths 5037 std::string loader_path("@loader_path"); 5038 std::string executable_path("@executable_path"); 5039 for (auto &rpath : rpath_paths) { 5040 if (rpath.find(loader_path) == 0) { 5041 rpath.erase(0, loader_path.size()); 5042 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5043 } else if (rpath.find(executable_path) == 0) { 5044 rpath.erase(0, executable_path.size()); 5045 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5046 } 5047 } 5048 5049 for (const auto &rpath_relative_path : rpath_relative_paths) { 5050 for (const auto &rpath : rpath_paths) { 5051 std::string path = rpath; 5052 path += rpath_relative_path; 5053 // It is OK to resolve this path because we must find a file on disk 5054 // for us to accept it anyway if it is rpath relative. 5055 FileSpec file_spec(path, true); 5056 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) { 5057 count++; 5058 break; 5059 } 5060 } 5061 } 5062 } 5063 5064 // We may have @executable_paths but no RPATHS. Figure those out here. 5065 // Only do this if this object file is the executable. We have no way to 5066 // get back to the actual executable otherwise, so we won't get the right 5067 // path. 5068 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5069 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5070 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5071 FileSpec file_spec = 5072 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5073 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) 5074 count++; 5075 } 5076 } 5077 } 5078 return count; 5079 } 5080 5081 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5082 // If the object file is not an executable it can't hold the entry point. 5083 // m_entry_point_address is initialized to an invalid address, so we can just 5084 // return that. If m_entry_point_address is valid it means we've found it 5085 // already, so return the cached value. 5086 5087 if (!IsExecutable() || m_entry_point_address.IsValid()) 5088 return m_entry_point_address; 5089 5090 // Otherwise, look for the UnixThread or Thread command. The data for the 5091 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5092 // 5093 // uint32_t flavor - this is the flavor argument you would pass to 5094 // thread_get_state 5095 // uint32_t count - this is the count of longs in the thread state data 5096 // struct XXX_thread_state state - this is the structure from 5097 // <machine/thread_status.h> corresponding to the flavor. 5098 // <repeat this trio> 5099 // 5100 // So we just keep reading the various register flavors till we find the GPR 5101 // one, then read the PC out of there. 5102 // FIXME: We will need to have a "RegisterContext data provider" class at some 5103 // point that can get all the registers 5104 // out of data in this form & attach them to a given thread. That should 5105 // underlie the MacOS X User process plugin, and we'll also need it for the 5106 // MacOS X Core File process plugin. When we have that we can also use it 5107 // here. 5108 // 5109 // For now we hard-code the offsets and flavors we need: 5110 // 5111 // 5112 5113 ModuleSP module_sp(GetModule()); 5114 if (module_sp) { 5115 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5116 struct load_command load_cmd; 5117 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5118 uint32_t i; 5119 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5120 bool done = false; 5121 5122 for (i = 0; i < m_header.ncmds; ++i) { 5123 const lldb::offset_t cmd_offset = offset; 5124 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5125 break; 5126 5127 switch (load_cmd.cmd) { 5128 case LC_UNIXTHREAD: 5129 case LC_THREAD: { 5130 while (offset < cmd_offset + load_cmd.cmdsize) { 5131 uint32_t flavor = m_data.GetU32(&offset); 5132 uint32_t count = m_data.GetU32(&offset); 5133 if (count == 0) { 5134 // We've gotten off somehow, log and exit; 5135 return m_entry_point_address; 5136 } 5137 5138 switch (m_header.cputype) { 5139 case llvm::MachO::CPU_TYPE_ARM: 5140 if (flavor == 1 || 5141 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 from 5142 // mach/arm/thread_status.h 5143 { 5144 offset += 60; // This is the offset of pc in the GPR thread state 5145 // data structure. 5146 start_address = m_data.GetU32(&offset); 5147 done = true; 5148 } 5149 break; 5150 case llvm::MachO::CPU_TYPE_ARM64: 5151 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5152 { 5153 offset += 256; // This is the offset of pc in the GPR thread state 5154 // data structure. 5155 start_address = m_data.GetU64(&offset); 5156 done = true; 5157 } 5158 break; 5159 case llvm::MachO::CPU_TYPE_I386: 5160 if (flavor == 5161 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5162 { 5163 offset += 40; // This is the offset of eip in the GPR thread state 5164 // data structure. 5165 start_address = m_data.GetU32(&offset); 5166 done = true; 5167 } 5168 break; 5169 case llvm::MachO::CPU_TYPE_X86_64: 5170 if (flavor == 5171 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5172 { 5173 offset += 16 * 8; // This is the offset of rip in the GPR thread 5174 // state data structure. 5175 start_address = m_data.GetU64(&offset); 5176 done = true; 5177 } 5178 break; 5179 default: 5180 return m_entry_point_address; 5181 } 5182 // Haven't found the GPR flavor yet, skip over the data for this 5183 // flavor: 5184 if (done) 5185 break; 5186 offset += count * 4; 5187 } 5188 } break; 5189 case LC_MAIN: { 5190 ConstString text_segment_name("__TEXT"); 5191 uint64_t entryoffset = m_data.GetU64(&offset); 5192 SectionSP text_segment_sp = 5193 GetSectionList()->FindSectionByName(text_segment_name); 5194 if (text_segment_sp) { 5195 done = true; 5196 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5197 } 5198 } break; 5199 5200 default: 5201 break; 5202 } 5203 if (done) 5204 break; 5205 5206 // Go to the next load command: 5207 offset = cmd_offset + load_cmd.cmdsize; 5208 } 5209 5210 if (start_address != LLDB_INVALID_ADDRESS) { 5211 // We got the start address from the load commands, so now resolve that 5212 // address in the sections of this ObjectFile: 5213 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5214 start_address, GetSectionList())) { 5215 m_entry_point_address.Clear(); 5216 } 5217 } else { 5218 // We couldn't read the UnixThread load command - maybe it wasn't there. 5219 // As a fallback look for the "start" symbol in the main executable. 5220 5221 ModuleSP module_sp(GetModule()); 5222 5223 if (module_sp) { 5224 SymbolContextList contexts; 5225 SymbolContext context; 5226 if (module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5227 eSymbolTypeCode, contexts)) { 5228 if (contexts.GetContextAtIndex(0, context)) 5229 m_entry_point_address = context.symbol->GetAddress(); 5230 } 5231 } 5232 } 5233 } 5234 5235 return m_entry_point_address; 5236 } 5237 5238 lldb_private::Address ObjectFileMachO::GetHeaderAddress() { 5239 lldb_private::Address header_addr; 5240 SectionList *section_list = GetSectionList(); 5241 if (section_list) { 5242 SectionSP text_segment_sp( 5243 section_list->FindSectionByName(GetSegmentNameTEXT())); 5244 if (text_segment_sp) { 5245 header_addr.SetSection(text_segment_sp); 5246 header_addr.SetOffset(0); 5247 } 5248 } 5249 return header_addr; 5250 } 5251 5252 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5253 ModuleSP module_sp(GetModule()); 5254 if (module_sp) { 5255 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5256 if (!m_thread_context_offsets_valid) { 5257 m_thread_context_offsets_valid = true; 5258 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5259 FileRangeArray::Entry file_range; 5260 thread_command thread_cmd; 5261 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5262 const uint32_t cmd_offset = offset; 5263 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 5264 break; 5265 5266 if (thread_cmd.cmd == LC_THREAD) { 5267 file_range.SetRangeBase(offset); 5268 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5269 m_thread_context_offsets.Append(file_range); 5270 } 5271 offset = cmd_offset + thread_cmd.cmdsize; 5272 } 5273 } 5274 } 5275 return m_thread_context_offsets.GetSize(); 5276 } 5277 5278 std::string ObjectFileMachO::GetIdentifierString() { 5279 std::string result; 5280 ModuleSP module_sp(GetModule()); 5281 if (module_sp) { 5282 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5283 5284 // First, look over the load commands for an LC_NOTE load command with 5285 // data_owner string "kern ver str" & use that if found. 5286 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5287 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5288 const uint32_t cmd_offset = offset; 5289 load_command lc; 5290 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5291 break; 5292 if (lc.cmd == LC_NOTE) 5293 { 5294 char data_owner[17]; 5295 m_data.CopyData (offset, 16, data_owner); 5296 data_owner[16] = '\0'; 5297 offset += 16; 5298 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5299 uint64_t size = m_data.GetU64_unchecked (&offset); 5300 5301 // "kern ver str" has a uint32_t version and then a nul terminated 5302 // c-string. 5303 if (strcmp ("kern ver str", data_owner) == 0) 5304 { 5305 offset = fileoff; 5306 uint32_t version; 5307 if (m_data.GetU32 (&offset, &version, 1) != nullptr) 5308 { 5309 if (version == 1) 5310 { 5311 uint32_t strsize = size - sizeof (uint32_t); 5312 char *buf = (char*) malloc (strsize); 5313 if (buf) 5314 { 5315 m_data.CopyData (offset, strsize, buf); 5316 buf[strsize - 1] = '\0'; 5317 result = buf; 5318 if (buf) 5319 free (buf); 5320 return result; 5321 } 5322 } 5323 } 5324 } 5325 } 5326 offset = cmd_offset + lc.cmdsize; 5327 } 5328 5329 // Second, make a pass over the load commands looking for an obsolete 5330 // LC_IDENT load command. 5331 offset = MachHeaderSizeFromMagic(m_header.magic); 5332 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5333 const uint32_t cmd_offset = offset; 5334 struct ident_command ident_command; 5335 if (m_data.GetU32(&offset, &ident_command, 2) == NULL) 5336 break; 5337 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5338 char *buf = (char *) malloc (ident_command.cmdsize); 5339 if (buf != nullptr 5340 && m_data.CopyData (offset, ident_command.cmdsize, buf) == ident_command.cmdsize) { 5341 buf[ident_command.cmdsize - 1] = '\0'; 5342 result = buf; 5343 } 5344 if (buf) 5345 free (buf); 5346 } 5347 offset = cmd_offset + ident_command.cmdsize; 5348 } 5349 5350 } 5351 return result; 5352 } 5353 5354 bool ObjectFileMachO::GetCorefileMainBinaryInfo (addr_t &address, UUID &uuid) { 5355 address = LLDB_INVALID_ADDRESS; 5356 uuid.Clear(); 5357 ModuleSP module_sp(GetModule()); 5358 if (module_sp) { 5359 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5360 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5361 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5362 const uint32_t cmd_offset = offset; 5363 load_command lc; 5364 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5365 break; 5366 if (lc.cmd == LC_NOTE) 5367 { 5368 char data_owner[17]; 5369 memset (data_owner, 0, sizeof (data_owner)); 5370 m_data.CopyData (offset, 16, data_owner); 5371 offset += 16; 5372 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5373 uint64_t size = m_data.GetU64_unchecked (&offset); 5374 5375 // "main bin spec" (main binary specification) data payload is 5376 // formatted: 5377 // uint32_t version [currently 1] 5378 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user process] 5379 // uint64_t address [ UINT64_MAX if address not specified ] 5380 // uuid_t uuid [ all zero's if uuid not specified ] 5381 // uint32_t log2_pagesize [ process page size in log base 2, e.g. 4k pages are 12. 0 for unspecified ] 5382 5383 if (strcmp ("main bin spec", data_owner) == 0 && size >= 32) 5384 { 5385 offset = fileoff; 5386 uint32_t version; 5387 if (m_data.GetU32 (&offset, &version, 1) != nullptr && version == 1) 5388 { 5389 uint32_t type = 0; 5390 uuid_t raw_uuid; 5391 memset (raw_uuid, 0, sizeof (uuid_t)); 5392 5393 if (m_data.GetU32(&offset, &type, 1) && 5394 m_data.GetU64(&offset, &address, 1) && 5395 m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5396 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5397 return true; 5398 } 5399 } 5400 } 5401 } 5402 offset = cmd_offset + lc.cmdsize; 5403 } 5404 } 5405 return false; 5406 } 5407 5408 lldb::RegisterContextSP 5409 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5410 lldb_private::Thread &thread) { 5411 lldb::RegisterContextSP reg_ctx_sp; 5412 5413 ModuleSP module_sp(GetModule()); 5414 if (module_sp) { 5415 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5416 if (!m_thread_context_offsets_valid) 5417 GetNumThreadContexts(); 5418 5419 const FileRangeArray::Entry *thread_context_file_range = 5420 m_thread_context_offsets.GetEntryAtIndex(idx); 5421 if (thread_context_file_range) { 5422 5423 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5424 thread_context_file_range->GetByteSize()); 5425 5426 switch (m_header.cputype) { 5427 case llvm::MachO::CPU_TYPE_ARM64: 5428 reg_ctx_sp.reset(new RegisterContextDarwin_arm64_Mach(thread, data)); 5429 break; 5430 5431 case llvm::MachO::CPU_TYPE_ARM: 5432 reg_ctx_sp.reset(new RegisterContextDarwin_arm_Mach(thread, data)); 5433 break; 5434 5435 case llvm::MachO::CPU_TYPE_I386: 5436 reg_ctx_sp.reset(new RegisterContextDarwin_i386_Mach(thread, data)); 5437 break; 5438 5439 case llvm::MachO::CPU_TYPE_X86_64: 5440 reg_ctx_sp.reset(new RegisterContextDarwin_x86_64_Mach(thread, data)); 5441 break; 5442 } 5443 } 5444 } 5445 return reg_ctx_sp; 5446 } 5447 5448 ObjectFile::Type ObjectFileMachO::CalculateType() { 5449 switch (m_header.filetype) { 5450 case MH_OBJECT: // 0x1u 5451 if (GetAddressByteSize() == 4) { 5452 // 32 bit kexts are just object files, but they do have a valid 5453 // UUID load command. 5454 UUID uuid; 5455 if (GetUUID(&uuid)) { 5456 // this checking for the UUID load command is not enough we could 5457 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5458 // this is required of kexts 5459 if (m_strata == eStrataInvalid) 5460 m_strata = eStrataKernel; 5461 return eTypeSharedLibrary; 5462 } 5463 } 5464 return eTypeObjectFile; 5465 5466 case MH_EXECUTE: 5467 return eTypeExecutable; // 0x2u 5468 case MH_FVMLIB: 5469 return eTypeSharedLibrary; // 0x3u 5470 case MH_CORE: 5471 return eTypeCoreFile; // 0x4u 5472 case MH_PRELOAD: 5473 return eTypeSharedLibrary; // 0x5u 5474 case MH_DYLIB: 5475 return eTypeSharedLibrary; // 0x6u 5476 case MH_DYLINKER: 5477 return eTypeDynamicLinker; // 0x7u 5478 case MH_BUNDLE: 5479 return eTypeSharedLibrary; // 0x8u 5480 case MH_DYLIB_STUB: 5481 return eTypeStubLibrary; // 0x9u 5482 case MH_DSYM: 5483 return eTypeDebugInfo; // 0xAu 5484 case MH_KEXT_BUNDLE: 5485 return eTypeSharedLibrary; // 0xBu 5486 default: 5487 break; 5488 } 5489 return eTypeUnknown; 5490 } 5491 5492 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5493 switch (m_header.filetype) { 5494 case MH_OBJECT: // 0x1u 5495 { 5496 // 32 bit kexts are just object files, but they do have a valid 5497 // UUID load command. 5498 UUID uuid; 5499 if (GetUUID(&uuid)) { 5500 // this checking for the UUID load command is not enough we could 5501 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5502 // this is required of kexts 5503 if (m_type == eTypeInvalid) 5504 m_type = eTypeSharedLibrary; 5505 5506 return eStrataKernel; 5507 } 5508 } 5509 return eStrataUnknown; 5510 5511 case MH_EXECUTE: // 0x2u 5512 // Check for the MH_DYLDLINK bit in the flags 5513 if (m_header.flags & MH_DYLDLINK) { 5514 return eStrataUser; 5515 } else { 5516 SectionList *section_list = GetSectionList(); 5517 if (section_list) { 5518 static ConstString g_kld_section_name("__KLD"); 5519 if (section_list->FindSectionByName(g_kld_section_name)) 5520 return eStrataKernel; 5521 } 5522 } 5523 return eStrataRawImage; 5524 5525 case MH_FVMLIB: 5526 return eStrataUser; // 0x3u 5527 case MH_CORE: 5528 return eStrataUnknown; // 0x4u 5529 case MH_PRELOAD: 5530 return eStrataRawImage; // 0x5u 5531 case MH_DYLIB: 5532 return eStrataUser; // 0x6u 5533 case MH_DYLINKER: 5534 return eStrataUser; // 0x7u 5535 case MH_BUNDLE: 5536 return eStrataUser; // 0x8u 5537 case MH_DYLIB_STUB: 5538 return eStrataUser; // 0x9u 5539 case MH_DSYM: 5540 return eStrataUnknown; // 0xAu 5541 case MH_KEXT_BUNDLE: 5542 return eStrataKernel; // 0xBu 5543 default: 5544 break; 5545 } 5546 return eStrataUnknown; 5547 } 5548 5549 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5550 ModuleSP module_sp(GetModule()); 5551 if (module_sp) { 5552 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5553 struct dylib_command load_cmd; 5554 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5555 uint32_t version_cmd = 0; 5556 uint64_t version = 0; 5557 uint32_t i; 5558 for (i = 0; i < m_header.ncmds; ++i) { 5559 const lldb::offset_t cmd_offset = offset; 5560 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5561 break; 5562 5563 if (load_cmd.cmd == LC_ID_DYLIB) { 5564 if (version_cmd == 0) { 5565 version_cmd = load_cmd.cmd; 5566 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5567 break; 5568 version = load_cmd.dylib.current_version; 5569 } 5570 break; // Break for now unless there is another more complete version 5571 // number load command in the future. 5572 } 5573 offset = cmd_offset + load_cmd.cmdsize; 5574 } 5575 5576 if (version_cmd == LC_ID_DYLIB) { 5577 unsigned major = (version & 0xFFFF0000ull) >> 16; 5578 unsigned minor = (version & 0x0000FF00ull) >> 8; 5579 unsigned subminor = (version & 0x000000FFull); 5580 return llvm::VersionTuple(major, minor, subminor); 5581 } 5582 } 5583 return llvm::VersionTuple(); 5584 } 5585 5586 bool ObjectFileMachO::GetArchitecture(ArchSpec &arch) { 5587 ModuleSP module_sp(GetModule()); 5588 if (module_sp) { 5589 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5590 return GetArchitecture(m_header, m_data, 5591 MachHeaderSizeFromMagic(m_header.magic), arch); 5592 } 5593 return false; 5594 } 5595 5596 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, addr_t &base_addr, UUID &uuid) { 5597 uuid.Clear(); 5598 base_addr = LLDB_INVALID_ADDRESS; 5599 if (process && process->GetDynamicLoader()) { 5600 DynamicLoader *dl = process->GetDynamicLoader(); 5601 LazyBool using_shared_cache; 5602 LazyBool private_shared_cache; 5603 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5604 private_shared_cache); 5605 } 5606 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5607 if (log) 5608 log->Printf("inferior process shared cache has a UUID of %s, base address 0x%" PRIx64 , uuid.GetAsString().c_str(), base_addr); 5609 } 5610 5611 // From dyld SPI header dyld_process_info.h 5612 typedef void *dyld_process_info; 5613 struct lldb_copy__dyld_process_cache_info { 5614 uuid_t cacheUUID; // UUID of cache used by process 5615 uint64_t cacheBaseAddress; // load address of dyld shared cache 5616 bool noCache; // process is running without a dyld cache 5617 bool privateCache; // process is using a private copy of its dyld cache 5618 }; 5619 5620 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with llvm 5621 // enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile errors. 5622 // So we need to use the actual underlying types of task_t and kern_return_t 5623 // below. 5624 extern "C" unsigned int /*task_t*/ mach_task_self(); 5625 5626 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5627 uuid.Clear(); 5628 base_addr = LLDB_INVALID_ADDRESS; 5629 5630 #if defined(__APPLE__) && \ 5631 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5632 uint8_t *(*dyld_get_all_image_infos)(void); 5633 dyld_get_all_image_infos = 5634 (uint8_t * (*)())dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5635 if (dyld_get_all_image_infos) { 5636 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5637 if (dyld_all_image_infos_address) { 5638 uint32_t *version = (uint32_t *) 5639 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5640 if (*version >= 13) { 5641 uuid_t *sharedCacheUUID_address = 0; 5642 int wordsize = sizeof(uint8_t *); 5643 if (wordsize == 8) { 5644 sharedCacheUUID_address = 5645 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5646 160); // sharedCacheUUID <mach-o/dyld_images.h> 5647 if (*version >= 15) 5648 base_addr = *(uint64_t *) ((uint8_t *) dyld_all_image_infos_address 5649 + 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5650 } else { 5651 sharedCacheUUID_address = 5652 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5653 84); // sharedCacheUUID <mach-o/dyld_images.h> 5654 if (*version >= 15) { 5655 base_addr = 0; 5656 base_addr = *(uint32_t *) ((uint8_t *) dyld_all_image_infos_address 5657 + 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5658 } 5659 } 5660 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5661 } 5662 } 5663 } else { 5664 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5665 dyld_process_info (*dyld_process_info_create)(unsigned int /* task_t */ task, uint64_t timestamp, unsigned int /*kern_return_t*/ *kernelError); 5666 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5667 void (*dyld_process_info_release)(dyld_process_info info); 5668 5669 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, unsigned int /*kern_return_t*/ *)) 5670 dlsym (RTLD_DEFAULT, "_dyld_process_info_create"); 5671 dyld_process_info_get_cache = (void (*)(void *, void *)) 5672 dlsym (RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5673 dyld_process_info_release = (void (*)(void *)) 5674 dlsym (RTLD_DEFAULT, "_dyld_process_info_release"); 5675 5676 if (dyld_process_info_create && dyld_process_info_get_cache) { 5677 unsigned int /*kern_return_t */ kern_ret; 5678 dyld_process_info process_info = dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5679 if (process_info) { 5680 struct lldb_copy__dyld_process_cache_info sc_info; 5681 memset (&sc_info, 0, sizeof (struct lldb_copy__dyld_process_cache_info)); 5682 dyld_process_info_get_cache (process_info, &sc_info); 5683 if (sc_info.cacheBaseAddress != 0) { 5684 base_addr = sc_info.cacheBaseAddress; 5685 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 5686 } 5687 dyld_process_info_release (process_info); 5688 } 5689 } 5690 } 5691 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5692 if (log && uuid.IsValid()) 5693 log->Printf("lldb's in-memory shared cache has a UUID of %s base address of 0x%" PRIx64, uuid.GetAsString().c_str(), base_addr); 5694 #endif 5695 } 5696 5697 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5698 if (!m_min_os_version) { 5699 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5700 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5701 const lldb::offset_t load_cmd_offset = offset; 5702 5703 version_min_command lc; 5704 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5705 break; 5706 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5707 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5708 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5709 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5710 if (m_data.GetU32(&offset, &lc.version, 5711 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5712 const uint32_t xxxx = lc.version >> 16; 5713 const uint32_t yy = (lc.version >> 8) & 0xffu; 5714 const uint32_t zz = lc.version & 0xffu; 5715 if (xxxx) { 5716 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5717 break; 5718 } 5719 } 5720 } 5721 offset = load_cmd_offset + lc.cmdsize; 5722 } 5723 5724 if (!m_min_os_version) { 5725 // Set version to an empty value so we don't keep trying to 5726 m_min_os_version = llvm::VersionTuple(); 5727 } 5728 } 5729 5730 return *m_min_os_version; 5731 } 5732 5733 uint32_t ObjectFileMachO::GetSDKVersion(uint32_t *versions, 5734 uint32_t num_versions) { 5735 if (m_sdk_versions.empty()) { 5736 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5737 bool success = false; 5738 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5739 const lldb::offset_t load_cmd_offset = offset; 5740 5741 version_min_command lc; 5742 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5743 break; 5744 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5745 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5746 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5747 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5748 if (m_data.GetU32(&offset, &lc.version, 5749 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5750 const uint32_t xxxx = lc.sdk >> 16; 5751 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5752 const uint32_t zz = lc.sdk & 0xffu; 5753 if (xxxx) { 5754 m_sdk_versions.push_back(xxxx); 5755 m_sdk_versions.push_back(yy); 5756 m_sdk_versions.push_back(zz); 5757 success = true; 5758 } else { 5759 GetModule()->ReportWarning( 5760 "minimum OS version load command with invalid (0) version found."); 5761 } 5762 } 5763 } 5764 offset = load_cmd_offset + lc.cmdsize; 5765 } 5766 5767 if (success == false) { 5768 // Push an invalid value so we don't try to find 5769 // the version # again on the next call to this 5770 // method. 5771 m_sdk_versions.push_back(UINT32_MAX); 5772 } 5773 } 5774 5775 // Legitimate version numbers will have 3 entries pushed 5776 // on to m_sdk_versions. If we only have one value, it's 5777 // the sentinel value indicating that this object file 5778 // does not have a valid minimum os version #. 5779 if (m_sdk_versions.size() > 1) { 5780 if (versions != NULL && num_versions > 0) { 5781 for (size_t i = 0; i < num_versions; ++i) { 5782 if (i < m_sdk_versions.size()) 5783 versions[i] = m_sdk_versions[i]; 5784 else 5785 versions[i] = 0; 5786 } 5787 } 5788 return m_sdk_versions.size(); 5789 } 5790 // Call the superclasses version that will empty out the data 5791 return ObjectFile::GetSDKVersion(versions, num_versions); 5792 } 5793 5794 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5795 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5796 } 5797 5798 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5799 return m_allow_assembly_emulation_unwind_plans; 5800 } 5801 5802 //------------------------------------------------------------------ 5803 // PluginInterface protocol 5804 //------------------------------------------------------------------ 5805 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5806 return GetPluginNameStatic(); 5807 } 5808 5809 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5810 5811 Section *ObjectFileMachO::GetMachHeaderSection() { 5812 // Find the first address of the mach header which is the first non-zero file 5813 // sized section whose file offset is zero. This is the base file address of 5814 // the mach-o file which can be subtracted from the vmaddr of the other 5815 // segments found in memory and added to the load address 5816 ModuleSP module_sp = GetModule(); 5817 if (module_sp) { 5818 SectionList *section_list = GetSectionList(); 5819 if (section_list) { 5820 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5821 const size_t num_sections = section_list->GetSize(); 5822 5823 for (size_t sect_idx = 0; sect_idx < num_sections && 5824 mach_base_file_addr == LLDB_INVALID_ADDRESS; 5825 ++sect_idx) { 5826 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5827 if (section && section->GetFileSize() > 0 && 5828 section->GetFileOffset() == 0 && 5829 section->IsThreadSpecific() == false && 5830 module_sp.get() == section->GetModule().get()) { 5831 return section; 5832 } 5833 } 5834 } 5835 } 5836 return nullptr; 5837 } 5838 5839 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 5840 lldb::addr_t mach_header_load_address, const Section *mach_header_section, 5841 const Section *section) { 5842 ModuleSP module_sp = GetModule(); 5843 if (module_sp && mach_header_section && section && 5844 mach_header_load_address != LLDB_INVALID_ADDRESS) { 5845 lldb::addr_t mach_header_file_addr = mach_header_section->GetFileAddress(); 5846 if (mach_header_file_addr != LLDB_INVALID_ADDRESS) { 5847 if (section && section->GetFileSize() > 0 && 5848 section->IsThreadSpecific() == false && 5849 module_sp.get() == section->GetModule().get()) { 5850 // Ignore __LINKEDIT and __DWARF segments 5851 if (section->GetName() == GetSegmentNameLINKEDIT()) { 5852 // Only map __LINKEDIT if we have an in memory image and this isn't a 5853 // kernel binary like a kext or mach_kernel. 5854 const bool is_memory_image = (bool)m_process_wp.lock(); 5855 const Strata strata = GetStrata(); 5856 if (is_memory_image == false || strata == eStrataKernel) 5857 return LLDB_INVALID_ADDRESS; 5858 } 5859 return section->GetFileAddress() - mach_header_file_addr + 5860 mach_header_load_address; 5861 } 5862 } 5863 } 5864 return LLDB_INVALID_ADDRESS; 5865 } 5866 5867 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 5868 bool value_is_offset) { 5869 ModuleSP module_sp = GetModule(); 5870 if (module_sp) { 5871 size_t num_loaded_sections = 0; 5872 SectionList *section_list = GetSectionList(); 5873 if (section_list) { 5874 const size_t num_sections = section_list->GetSize(); 5875 5876 if (value_is_offset) { 5877 // "value" is an offset to apply to each top level segment 5878 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5879 // Iterate through the object file sections to find all of the 5880 // sections that size on disk (to avoid __PAGEZERO) and load them 5881 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5882 if (section_sp && section_sp->GetFileSize() > 0 && 5883 section_sp->IsThreadSpecific() == false && 5884 module_sp.get() == section_sp->GetModule().get()) { 5885 // Ignore __LINKEDIT and __DWARF segments 5886 if (section_sp->GetName() == GetSegmentNameLINKEDIT()) { 5887 // Only map __LINKEDIT if we have an in memory image and this 5888 // isn't a kernel binary like a kext or mach_kernel. 5889 const bool is_memory_image = (bool)m_process_wp.lock(); 5890 const Strata strata = GetStrata(); 5891 if (is_memory_image == false || strata == eStrataKernel) 5892 continue; 5893 } 5894 if (target.GetSectionLoadList().SetSectionLoadAddress( 5895 section_sp, section_sp->GetFileAddress() + value)) 5896 ++num_loaded_sections; 5897 } 5898 } 5899 } else { 5900 // "value" is the new base address of the mach_header, adjust each 5901 // section accordingly 5902 5903 Section *mach_header_section = GetMachHeaderSection(); 5904 if (mach_header_section) { 5905 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5906 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5907 5908 lldb::addr_t section_load_addr = 5909 CalculateSectionLoadAddressForMemoryImage( 5910 value, mach_header_section, section_sp.get()); 5911 if (section_load_addr != LLDB_INVALID_ADDRESS) { 5912 if (target.GetSectionLoadList().SetSectionLoadAddress( 5913 section_sp, section_load_addr)) 5914 ++num_loaded_sections; 5915 } 5916 } 5917 } 5918 } 5919 } 5920 return num_loaded_sections > 0; 5921 } 5922 return false; 5923 } 5924 5925 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 5926 const FileSpec &outfile, Status &error) { 5927 if (process_sp) { 5928 Target &target = process_sp->GetTarget(); 5929 const ArchSpec target_arch = target.GetArchitecture(); 5930 const llvm::Triple &target_triple = target_arch.GetTriple(); 5931 if (target_triple.getVendor() == llvm::Triple::Apple && 5932 (target_triple.getOS() == llvm::Triple::MacOSX || 5933 target_triple.getOS() == llvm::Triple::IOS || 5934 target_triple.getOS() == llvm::Triple::WatchOS || 5935 target_triple.getOS() == llvm::Triple::TvOS)) { 5936 bool make_core = false; 5937 switch (target_arch.GetMachine()) { 5938 case llvm::Triple::aarch64: 5939 case llvm::Triple::arm: 5940 case llvm::Triple::thumb: 5941 case llvm::Triple::x86: 5942 case llvm::Triple::x86_64: 5943 make_core = true; 5944 break; 5945 default: 5946 error.SetErrorStringWithFormat("unsupported core architecture: %s", 5947 target_triple.str().c_str()); 5948 break; 5949 } 5950 5951 if (make_core) { 5952 std::vector<segment_command_64> segment_load_commands; 5953 // uint32_t range_info_idx = 0; 5954 MemoryRegionInfo range_info; 5955 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 5956 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 5957 const ByteOrder byte_order = target_arch.GetByteOrder(); 5958 if (range_error.Success()) { 5959 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 5960 const addr_t addr = range_info.GetRange().GetRangeBase(); 5961 const addr_t size = range_info.GetRange().GetByteSize(); 5962 5963 if (size == 0) 5964 break; 5965 5966 // Calculate correct protections 5967 uint32_t prot = 0; 5968 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 5969 prot |= VM_PROT_READ; 5970 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 5971 prot |= VM_PROT_WRITE; 5972 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 5973 prot |= VM_PROT_EXECUTE; 5974 5975 // printf ("[%3u] [0x%16.16" PRIx64 " - 5976 // 0x%16.16" PRIx64 ") %c%c%c\n", 5977 // range_info_idx, 5978 // addr, 5979 // size, 5980 // (prot & VM_PROT_READ ) ? 'r' : 5981 // '-', 5982 // (prot & VM_PROT_WRITE ) ? 'w' : 5983 // '-', 5984 // (prot & VM_PROT_EXECUTE) ? 'x' : 5985 // '-'); 5986 5987 if (prot != 0) { 5988 uint32_t cmd_type = LC_SEGMENT_64; 5989 uint32_t segment_size = sizeof(segment_command_64); 5990 if (addr_byte_size == 4) { 5991 cmd_type = LC_SEGMENT; 5992 segment_size = sizeof(segment_command); 5993 } 5994 segment_command_64 segment = { 5995 cmd_type, // uint32_t cmd; 5996 segment_size, // uint32_t cmdsize; 5997 {0}, // char segname[16]; 5998 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 5999 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6000 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6001 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6002 prot, // uint32_t maxprot; 6003 prot, // uint32_t initprot; 6004 0, // uint32_t nsects; 6005 0}; // uint32_t flags; 6006 segment_load_commands.push_back(segment); 6007 } else { 6008 // No protections and a size of 1 used to be returned from old 6009 // debugservers when we asked about a region that was past the 6010 // last memory region and it indicates the end... 6011 if (size == 1) 6012 break; 6013 } 6014 6015 range_error = process_sp->GetMemoryRegionInfo( 6016 range_info.GetRange().GetRangeEnd(), range_info); 6017 if (range_error.Fail()) 6018 break; 6019 } 6020 6021 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6022 6023 mach_header_64 mach_header; 6024 if (addr_byte_size == 8) { 6025 mach_header.magic = MH_MAGIC_64; 6026 } else { 6027 mach_header.magic = MH_MAGIC; 6028 } 6029 mach_header.cputype = target_arch.GetMachOCPUType(); 6030 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6031 mach_header.filetype = MH_CORE; 6032 mach_header.ncmds = segment_load_commands.size(); 6033 mach_header.flags = 0; 6034 mach_header.reserved = 0; 6035 ThreadList &thread_list = process_sp->GetThreadList(); 6036 const uint32_t num_threads = thread_list.GetSize(); 6037 6038 // Make an array of LC_THREAD data items. Each one contains the 6039 // contents of the LC_THREAD load command. The data doesn't contain 6040 // the load command + load command size, we will add the load command 6041 // and load command size as we emit the data. 6042 std::vector<StreamString> LC_THREAD_datas(num_threads); 6043 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6044 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6045 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6046 LC_THREAD_data.SetByteOrder(byte_order); 6047 } 6048 for (uint32_t thread_idx = 0; thread_idx < num_threads; 6049 ++thread_idx) { 6050 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6051 if (thread_sp) { 6052 switch (mach_header.cputype) { 6053 case llvm::MachO::CPU_TYPE_ARM64: 6054 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6055 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6056 break; 6057 6058 case llvm::MachO::CPU_TYPE_ARM: 6059 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6060 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6061 break; 6062 6063 case llvm::MachO::CPU_TYPE_I386: 6064 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6065 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6066 break; 6067 6068 case llvm::MachO::CPU_TYPE_X86_64: 6069 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6070 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6071 break; 6072 } 6073 } 6074 } 6075 6076 // The size of the load command is the size of the segments... 6077 if (addr_byte_size == 8) { 6078 mach_header.sizeofcmds = segment_load_commands.size() * 6079 sizeof(struct segment_command_64); 6080 } else { 6081 mach_header.sizeofcmds = 6082 segment_load_commands.size() * sizeof(struct segment_command); 6083 } 6084 6085 // and the size of all LC_THREAD load command 6086 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6087 ++mach_header.ncmds; 6088 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6089 } 6090 6091 printf("mach_header: 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x " 6092 "0x%8.8x 0x%8.8x\n", 6093 mach_header.magic, mach_header.cputype, mach_header.cpusubtype, 6094 mach_header.filetype, mach_header.ncmds, 6095 mach_header.sizeofcmds, mach_header.flags, 6096 mach_header.reserved); 6097 6098 // Write the mach header 6099 buffer.PutHex32(mach_header.magic); 6100 buffer.PutHex32(mach_header.cputype); 6101 buffer.PutHex32(mach_header.cpusubtype); 6102 buffer.PutHex32(mach_header.filetype); 6103 buffer.PutHex32(mach_header.ncmds); 6104 buffer.PutHex32(mach_header.sizeofcmds); 6105 buffer.PutHex32(mach_header.flags); 6106 if (addr_byte_size == 8) { 6107 buffer.PutHex32(mach_header.reserved); 6108 } 6109 6110 // Skip the mach header and all load commands and align to the next 6111 // 0x1000 byte boundary 6112 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6113 if (file_offset & 0x00000fff) { 6114 file_offset += 0x00001000ull; 6115 file_offset &= (~0x00001000ull + 1); 6116 } 6117 6118 for (auto &segment : segment_load_commands) { 6119 segment.fileoff = file_offset; 6120 file_offset += segment.filesize; 6121 } 6122 6123 // Write out all of the LC_THREAD load commands 6124 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6125 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6126 buffer.PutHex32(LC_THREAD); 6127 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6128 buffer.Write(LC_THREAD_data.GetString().data(), 6129 LC_THREAD_data_size); 6130 } 6131 6132 // Write out all of the segment load commands 6133 for (const auto &segment : segment_load_commands) { 6134 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6135 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6136 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6137 segment.cmd, segment.cmdsize, segment.vmaddr, 6138 segment.vmaddr + segment.vmsize, segment.fileoff, 6139 segment.filesize, segment.maxprot, segment.initprot, 6140 segment.nsects, segment.flags); 6141 6142 buffer.PutHex32(segment.cmd); 6143 buffer.PutHex32(segment.cmdsize); 6144 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6145 if (addr_byte_size == 8) { 6146 buffer.PutHex64(segment.vmaddr); 6147 buffer.PutHex64(segment.vmsize); 6148 buffer.PutHex64(segment.fileoff); 6149 buffer.PutHex64(segment.filesize); 6150 } else { 6151 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6152 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6153 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6154 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6155 } 6156 buffer.PutHex32(segment.maxprot); 6157 buffer.PutHex32(segment.initprot); 6158 buffer.PutHex32(segment.nsects); 6159 buffer.PutHex32(segment.flags); 6160 } 6161 6162 File core_file; 6163 std::string core_file_path(outfile.GetPath()); 6164 error = core_file.Open(core_file_path.c_str(), 6165 File::eOpenOptionWrite | 6166 File::eOpenOptionTruncate | 6167 File::eOpenOptionCanCreate); 6168 if (error.Success()) { 6169 // Read 1 page at a time 6170 uint8_t bytes[0x1000]; 6171 // Write the mach header and load commands out to the core file 6172 size_t bytes_written = buffer.GetString().size(); 6173 error = core_file.Write(buffer.GetString().data(), bytes_written); 6174 if (error.Success()) { 6175 // Now write the file data for all memory segments in the process 6176 for (const auto &segment : segment_load_commands) { 6177 if (core_file.SeekFromStart(segment.fileoff) == -1) { 6178 error.SetErrorStringWithFormat( 6179 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6180 segment.fileoff, core_file_path.c_str()); 6181 break; 6182 } 6183 6184 printf("Saving %" PRId64 6185 " bytes of data for memory region at 0x%" PRIx64 "\n", 6186 segment.vmsize, segment.vmaddr); 6187 addr_t bytes_left = segment.vmsize; 6188 addr_t addr = segment.vmaddr; 6189 Status memory_read_error; 6190 while (bytes_left > 0 && error.Success()) { 6191 const size_t bytes_to_read = 6192 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6193 const size_t bytes_read = process_sp->ReadMemory( 6194 addr, bytes, bytes_to_read, memory_read_error); 6195 if (bytes_read == bytes_to_read) { 6196 size_t bytes_written = bytes_read; 6197 error = core_file.Write(bytes, bytes_written); 6198 bytes_left -= bytes_read; 6199 addr += bytes_read; 6200 } else { 6201 // Some pages within regions are not readable, those should 6202 // be zero filled 6203 memset(bytes, 0, bytes_to_read); 6204 size_t bytes_written = bytes_to_read; 6205 error = core_file.Write(bytes, bytes_written); 6206 bytes_left -= bytes_to_read; 6207 addr += bytes_to_read; 6208 } 6209 } 6210 } 6211 } 6212 } 6213 } else { 6214 error.SetErrorString( 6215 "process doesn't support getting memory region info"); 6216 } 6217 } 6218 return true; // This is the right plug to handle saving core files for 6219 // this process 6220 } 6221 } 6222 return false; 6223 } 6224