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/ArchSpec.h" 21 #include "lldb/Core/Debugger.h" 22 #include "lldb/Core/FileSpecList.h" 23 #include "lldb/Core/Module.h" 24 #include "lldb/Core/ModuleSpec.h" 25 #include "lldb/Core/PluginManager.h" 26 #include "lldb/Core/RangeMap.h" 27 #include "lldb/Core/RegisterValue.h" 28 #include "lldb/Core/Section.h" 29 #include "lldb/Core/StreamFile.h" 30 #include "lldb/Host/Host.h" 31 #include "lldb/Symbol/DWARFCallFrameInfo.h" 32 #include "lldb/Symbol/ObjectFile.h" 33 #include "lldb/Target/DynamicLoader.h" 34 #include "lldb/Target/MemoryRegionInfo.h" 35 #include "lldb/Target/Platform.h" 36 #include "lldb/Target/Process.h" 37 #include "lldb/Target/SectionLoadList.h" 38 #include "lldb/Target/Target.h" 39 #include "lldb/Target/Thread.h" 40 #include "lldb/Target/ThreadList.h" 41 #include "lldb/Utility/DataBufferLLVM.h" 42 #include "lldb/Utility/FileSpec.h" 43 #include "lldb/Utility/Log.h" 44 #include "lldb/Utility/Status.h" 45 #include "lldb/Utility/StreamString.h" 46 #include "lldb/Utility/Timer.h" 47 #include "lldb/Utility/UUID.h" 48 49 #include "lldb/Utility/SafeMachO.h" 50 51 #include "llvm/Support/MemoryBuffer.h" 52 53 #include "ObjectFileMachO.h" 54 55 #if defined(__APPLE__) && \ 56 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 57 // GetLLDBSharedCacheUUID() needs to call dlsym() 58 #include <dlfcn.h> 59 #endif 60 61 #ifndef __APPLE__ 62 #include "Utility/UuidCompatibility.h" 63 #else 64 #include <uuid/uuid.h> 65 #endif 66 67 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 68 using namespace lldb; 69 using namespace lldb_private; 70 using namespace llvm::MachO; 71 72 // Some structure definitions needed for parsing the dyld shared cache files 73 // found on iOS devices. 74 75 struct lldb_copy_dyld_cache_header_v1 { 76 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 77 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 78 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 79 uint32_t imagesOffset; 80 uint32_t imagesCount; 81 uint64_t dyldBaseAddress; 82 uint64_t codeSignatureOffset; 83 uint64_t codeSignatureSize; 84 uint64_t slideInfoOffset; 85 uint64_t slideInfoSize; 86 uint64_t localSymbolsOffset; 87 uint64_t localSymbolsSize; 88 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 89 // and later 90 }; 91 92 struct lldb_copy_dyld_cache_mapping_info { 93 uint64_t address; 94 uint64_t size; 95 uint64_t fileOffset; 96 uint32_t maxProt; 97 uint32_t initProt; 98 }; 99 100 struct lldb_copy_dyld_cache_local_symbols_info { 101 uint32_t nlistOffset; 102 uint32_t nlistCount; 103 uint32_t stringsOffset; 104 uint32_t stringsSize; 105 uint32_t entriesOffset; 106 uint32_t entriesCount; 107 }; 108 struct lldb_copy_dyld_cache_local_symbols_entry { 109 uint32_t dylibOffset; 110 uint32_t nlistStartIndex; 111 uint32_t nlistCount; 112 }; 113 114 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 115 public: 116 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 117 const DataExtractor &data) 118 : RegisterContextDarwin_x86_64(thread, 0) { 119 SetRegisterDataFrom_LC_THREAD(data); 120 } 121 122 void InvalidateAllRegisters() override { 123 // Do nothing... registers are always valid... 124 } 125 126 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 127 lldb::offset_t offset = 0; 128 SetError(GPRRegSet, Read, -1); 129 SetError(FPURegSet, Read, -1); 130 SetError(EXCRegSet, Read, -1); 131 bool done = false; 132 133 while (!done) { 134 int flavor = data.GetU32(&offset); 135 if (flavor == 0) 136 done = true; 137 else { 138 uint32_t i; 139 uint32_t count = data.GetU32(&offset); 140 switch (flavor) { 141 case GPRRegSet: 142 for (i = 0; i < count; ++i) 143 (&gpr.rax)[i] = data.GetU64(&offset); 144 SetError(GPRRegSet, Read, 0); 145 done = true; 146 147 break; 148 case FPURegSet: 149 // TODO: fill in FPU regs.... 150 // SetError (FPURegSet, Read, -1); 151 done = true; 152 153 break; 154 case EXCRegSet: 155 exc.trapno = data.GetU32(&offset); 156 exc.err = data.GetU32(&offset); 157 exc.faultvaddr = data.GetU64(&offset); 158 SetError(EXCRegSet, Read, 0); 159 done = true; 160 break; 161 case 7: 162 case 8: 163 case 9: 164 // fancy flavors that encapsulate of the above 165 // flavors... 166 break; 167 168 default: 169 done = true; 170 break; 171 } 172 } 173 } 174 } 175 176 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 177 const char *alt_name, size_t reg_byte_size, 178 Stream &data) { 179 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 180 if (reg_info == NULL) 181 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 182 if (reg_info) { 183 lldb_private::RegisterValue reg_value; 184 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 185 if (reg_info->byte_size >= reg_byte_size) 186 data.Write(reg_value.GetBytes(), reg_byte_size); 187 else { 188 data.Write(reg_value.GetBytes(), reg_info->byte_size); 189 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 190 ++i) 191 data.PutChar(0); 192 } 193 return reg_byte_size; 194 } 195 } 196 // Just write zeros if all else fails 197 for (size_t i = 0; i < reg_byte_size; ++i) 198 data.PutChar(0); 199 return reg_byte_size; 200 } 201 202 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 203 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 204 if (reg_ctx_sp) { 205 RegisterContext *reg_ctx = reg_ctx_sp.get(); 206 207 data.PutHex32(GPRRegSet); // Flavor 208 data.PutHex32(GPRWordCount); 209 WriteRegister(reg_ctx, "rax", NULL, 8, data); 210 WriteRegister(reg_ctx, "rbx", NULL, 8, data); 211 WriteRegister(reg_ctx, "rcx", NULL, 8, data); 212 WriteRegister(reg_ctx, "rdx", NULL, 8, data); 213 WriteRegister(reg_ctx, "rdi", NULL, 8, data); 214 WriteRegister(reg_ctx, "rsi", NULL, 8, data); 215 WriteRegister(reg_ctx, "rbp", NULL, 8, data); 216 WriteRegister(reg_ctx, "rsp", NULL, 8, data); 217 WriteRegister(reg_ctx, "r8", NULL, 8, data); 218 WriteRegister(reg_ctx, "r9", NULL, 8, data); 219 WriteRegister(reg_ctx, "r10", NULL, 8, data); 220 WriteRegister(reg_ctx, "r11", NULL, 8, data); 221 WriteRegister(reg_ctx, "r12", NULL, 8, data); 222 WriteRegister(reg_ctx, "r13", NULL, 8, data); 223 WriteRegister(reg_ctx, "r14", NULL, 8, data); 224 WriteRegister(reg_ctx, "r15", NULL, 8, data); 225 WriteRegister(reg_ctx, "rip", NULL, 8, data); 226 WriteRegister(reg_ctx, "rflags", NULL, 8, data); 227 WriteRegister(reg_ctx, "cs", NULL, 8, data); 228 WriteRegister(reg_ctx, "fs", NULL, 8, data); 229 WriteRegister(reg_ctx, "gs", NULL, 8, data); 230 231 // // Write out the FPU registers 232 // const size_t fpu_byte_size = sizeof(FPU); 233 // size_t bytes_written = 0; 234 // data.PutHex32 (FPURegSet); 235 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 236 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 237 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 238 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 239 // data); // uint16_t fcw; // "fctrl" 240 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 241 // data); // uint16_t fsw; // "fstat" 242 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 243 // data); // uint8_t ftw; // "ftag" 244 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 245 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 246 // data); // uint16_t fop; // "fop" 247 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 248 // data); // uint32_t ip; // "fioff" 249 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 250 // data); // uint16_t cs; // "fiseg" 251 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 252 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 253 // data); // uint32_t dp; // "fooff" 254 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 255 // data); // uint16_t ds; // "foseg" 256 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 257 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 258 // data); // uint32_t mxcsr; 259 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 260 // 4, data);// uint32_t mxcsrmask; 261 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 262 // sizeof(MMSReg), data); 263 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 264 // sizeof(MMSReg), data); 265 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 266 // sizeof(MMSReg), data); 267 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 268 // sizeof(MMSReg), data); 269 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 270 // sizeof(MMSReg), data); 271 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 272 // sizeof(MMSReg), data); 273 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 274 // sizeof(MMSReg), data); 275 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 276 // sizeof(MMSReg), data); 277 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 278 // sizeof(XMMReg), data); 279 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 280 // sizeof(XMMReg), data); 281 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 282 // sizeof(XMMReg), data); 283 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 284 // sizeof(XMMReg), data); 285 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 286 // sizeof(XMMReg), data); 287 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 288 // sizeof(XMMReg), data); 289 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 290 // sizeof(XMMReg), data); 291 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 292 // sizeof(XMMReg), data); 293 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 294 // sizeof(XMMReg), data); 295 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 296 // sizeof(XMMReg), data); 297 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 298 // sizeof(XMMReg), data); 299 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 300 // sizeof(XMMReg), data); 301 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 302 // sizeof(XMMReg), data); 303 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 304 // sizeof(XMMReg), data); 305 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 306 // sizeof(XMMReg), data); 307 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 308 // sizeof(XMMReg), data); 309 // 310 // // Fill rest with zeros 311 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 312 // i) 313 // data.PutChar(0); 314 315 // Write out the EXC registers 316 data.PutHex32(EXCRegSet); 317 data.PutHex32(EXCWordCount); 318 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 319 WriteRegister(reg_ctx, "err", NULL, 4, data); 320 WriteRegister(reg_ctx, "faultvaddr", NULL, 8, data); 321 return true; 322 } 323 return false; 324 } 325 326 protected: 327 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 328 329 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 330 331 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 332 333 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 334 return 0; 335 } 336 337 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 338 return 0; 339 } 340 341 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 342 return 0; 343 } 344 }; 345 346 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 347 public: 348 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 349 const DataExtractor &data) 350 : RegisterContextDarwin_i386(thread, 0) { 351 SetRegisterDataFrom_LC_THREAD(data); 352 } 353 354 void InvalidateAllRegisters() override { 355 // Do nothing... registers are always valid... 356 } 357 358 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 359 lldb::offset_t offset = 0; 360 SetError(GPRRegSet, Read, -1); 361 SetError(FPURegSet, Read, -1); 362 SetError(EXCRegSet, Read, -1); 363 bool done = false; 364 365 while (!done) { 366 int flavor = data.GetU32(&offset); 367 if (flavor == 0) 368 done = true; 369 else { 370 uint32_t i; 371 uint32_t count = data.GetU32(&offset); 372 switch (flavor) { 373 case GPRRegSet: 374 for (i = 0; i < count; ++i) 375 (&gpr.eax)[i] = data.GetU32(&offset); 376 SetError(GPRRegSet, Read, 0); 377 done = true; 378 379 break; 380 case FPURegSet: 381 // TODO: fill in FPU regs.... 382 // SetError (FPURegSet, Read, -1); 383 done = true; 384 385 break; 386 case EXCRegSet: 387 exc.trapno = data.GetU32(&offset); 388 exc.err = data.GetU32(&offset); 389 exc.faultvaddr = data.GetU32(&offset); 390 SetError(EXCRegSet, Read, 0); 391 done = true; 392 break; 393 case 7: 394 case 8: 395 case 9: 396 // fancy flavors that encapsulate of the above 397 // flavors... 398 break; 399 400 default: 401 done = true; 402 break; 403 } 404 } 405 } 406 } 407 408 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 409 const char *alt_name, size_t reg_byte_size, 410 Stream &data) { 411 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 412 if (reg_info == NULL) 413 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 414 if (reg_info) { 415 lldb_private::RegisterValue reg_value; 416 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 417 if (reg_info->byte_size >= reg_byte_size) 418 data.Write(reg_value.GetBytes(), reg_byte_size); 419 else { 420 data.Write(reg_value.GetBytes(), reg_info->byte_size); 421 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 422 ++i) 423 data.PutChar(0); 424 } 425 return reg_byte_size; 426 } 427 } 428 // Just write zeros if all else fails 429 for (size_t i = 0; i < reg_byte_size; ++i) 430 data.PutChar(0); 431 return reg_byte_size; 432 } 433 434 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 435 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 436 if (reg_ctx_sp) { 437 RegisterContext *reg_ctx = reg_ctx_sp.get(); 438 439 data.PutHex32(GPRRegSet); // Flavor 440 data.PutHex32(GPRWordCount); 441 WriteRegister(reg_ctx, "eax", NULL, 4, data); 442 WriteRegister(reg_ctx, "ebx", NULL, 4, data); 443 WriteRegister(reg_ctx, "ecx", NULL, 4, data); 444 WriteRegister(reg_ctx, "edx", NULL, 4, data); 445 WriteRegister(reg_ctx, "edi", NULL, 4, data); 446 WriteRegister(reg_ctx, "esi", NULL, 4, data); 447 WriteRegister(reg_ctx, "ebp", NULL, 4, data); 448 WriteRegister(reg_ctx, "esp", NULL, 4, data); 449 WriteRegister(reg_ctx, "ss", NULL, 4, data); 450 WriteRegister(reg_ctx, "eflags", NULL, 4, data); 451 WriteRegister(reg_ctx, "eip", NULL, 4, data); 452 WriteRegister(reg_ctx, "cs", NULL, 4, data); 453 WriteRegister(reg_ctx, "ds", NULL, 4, data); 454 WriteRegister(reg_ctx, "es", NULL, 4, data); 455 WriteRegister(reg_ctx, "fs", NULL, 4, data); 456 WriteRegister(reg_ctx, "gs", NULL, 4, data); 457 458 // Write out the EXC registers 459 data.PutHex32(EXCRegSet); 460 data.PutHex32(EXCWordCount); 461 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 462 WriteRegister(reg_ctx, "err", NULL, 4, data); 463 WriteRegister(reg_ctx, "faultvaddr", NULL, 4, data); 464 return true; 465 } 466 return false; 467 } 468 469 protected: 470 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 471 472 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 473 474 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 475 476 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 477 return 0; 478 } 479 480 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 481 return 0; 482 } 483 484 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 485 return 0; 486 } 487 }; 488 489 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 490 public: 491 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 492 const DataExtractor &data) 493 : RegisterContextDarwin_arm(thread, 0) { 494 SetRegisterDataFrom_LC_THREAD(data); 495 } 496 497 void InvalidateAllRegisters() override { 498 // Do nothing... registers are always valid... 499 } 500 501 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 502 lldb::offset_t offset = 0; 503 SetError(GPRRegSet, Read, -1); 504 SetError(FPURegSet, Read, -1); 505 SetError(EXCRegSet, Read, -1); 506 bool done = false; 507 508 while (!done) { 509 int flavor = data.GetU32(&offset); 510 uint32_t count = data.GetU32(&offset); 511 lldb::offset_t next_thread_state = offset + (count * 4); 512 switch (flavor) { 513 case GPRAltRegSet: 514 case GPRRegSet: 515 for (uint32_t i = 0; i < count; ++i) { 516 gpr.r[i] = data.GetU32(&offset); 517 } 518 519 // Note that gpr.cpsr is also copied by the above loop; this loop 520 // technically extends 521 // one element past the end of the gpr.r[] array. 522 523 SetError(GPRRegSet, Read, 0); 524 offset = next_thread_state; 525 break; 526 527 case FPURegSet: { 528 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 529 const int fpu_reg_buf_size = sizeof(fpu.floats); 530 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 531 fpu_reg_buf) == fpu_reg_buf_size) { 532 offset += fpu_reg_buf_size; 533 fpu.fpscr = data.GetU32(&offset); 534 SetError(FPURegSet, Read, 0); 535 } else { 536 done = true; 537 } 538 } 539 offset = next_thread_state; 540 break; 541 542 case EXCRegSet: 543 if (count == 3) { 544 exc.exception = data.GetU32(&offset); 545 exc.fsr = data.GetU32(&offset); 546 exc.far = data.GetU32(&offset); 547 SetError(EXCRegSet, Read, 0); 548 } 549 done = true; 550 offset = next_thread_state; 551 break; 552 553 // Unknown register set flavor, stop trying to parse. 554 default: 555 done = true; 556 } 557 } 558 } 559 560 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 561 const char *alt_name, size_t reg_byte_size, 562 Stream &data) { 563 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 564 if (reg_info == NULL) 565 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 566 if (reg_info) { 567 lldb_private::RegisterValue reg_value; 568 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 569 if (reg_info->byte_size >= reg_byte_size) 570 data.Write(reg_value.GetBytes(), reg_byte_size); 571 else { 572 data.Write(reg_value.GetBytes(), reg_info->byte_size); 573 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 574 ++i) 575 data.PutChar(0); 576 } 577 return reg_byte_size; 578 } 579 } 580 // Just write zeros if all else fails 581 for (size_t i = 0; i < reg_byte_size; ++i) 582 data.PutChar(0); 583 return reg_byte_size; 584 } 585 586 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 587 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 588 if (reg_ctx_sp) { 589 RegisterContext *reg_ctx = reg_ctx_sp.get(); 590 591 data.PutHex32(GPRRegSet); // Flavor 592 data.PutHex32(GPRWordCount); 593 WriteRegister(reg_ctx, "r0", NULL, 4, data); 594 WriteRegister(reg_ctx, "r1", NULL, 4, data); 595 WriteRegister(reg_ctx, "r2", NULL, 4, data); 596 WriteRegister(reg_ctx, "r3", NULL, 4, data); 597 WriteRegister(reg_ctx, "r4", NULL, 4, data); 598 WriteRegister(reg_ctx, "r5", NULL, 4, data); 599 WriteRegister(reg_ctx, "r6", NULL, 4, data); 600 WriteRegister(reg_ctx, "r7", NULL, 4, data); 601 WriteRegister(reg_ctx, "r8", NULL, 4, data); 602 WriteRegister(reg_ctx, "r9", NULL, 4, data); 603 WriteRegister(reg_ctx, "r10", NULL, 4, data); 604 WriteRegister(reg_ctx, "r11", NULL, 4, data); 605 WriteRegister(reg_ctx, "r12", NULL, 4, data); 606 WriteRegister(reg_ctx, "sp", NULL, 4, data); 607 WriteRegister(reg_ctx, "lr", NULL, 4, data); 608 WriteRegister(reg_ctx, "pc", NULL, 4, data); 609 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 610 611 // Write out the EXC registers 612 // data.PutHex32 (EXCRegSet); 613 // data.PutHex32 (EXCWordCount); 614 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 615 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 616 // WriteRegister (reg_ctx, "far", NULL, 4, data); 617 return true; 618 } 619 return false; 620 } 621 622 protected: 623 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 624 625 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 626 627 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 628 629 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 630 631 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 632 return 0; 633 } 634 635 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 636 return 0; 637 } 638 639 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 640 return 0; 641 } 642 643 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 644 return -1; 645 } 646 }; 647 648 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 649 public: 650 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 651 const DataExtractor &data) 652 : RegisterContextDarwin_arm64(thread, 0) { 653 SetRegisterDataFrom_LC_THREAD(data); 654 } 655 656 void InvalidateAllRegisters() override { 657 // Do nothing... registers are always valid... 658 } 659 660 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 661 lldb::offset_t offset = 0; 662 SetError(GPRRegSet, Read, -1); 663 SetError(FPURegSet, Read, -1); 664 SetError(EXCRegSet, Read, -1); 665 bool done = false; 666 while (!done) { 667 int flavor = data.GetU32(&offset); 668 uint32_t count = data.GetU32(&offset); 669 lldb::offset_t next_thread_state = offset + (count * 4); 670 switch (flavor) { 671 case GPRRegSet: 672 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 673 // 32-bit register) 674 if (count >= (33 * 2) + 1) { 675 for (uint32_t i = 0; i < 29; ++i) 676 gpr.x[i] = data.GetU64(&offset); 677 gpr.fp = data.GetU64(&offset); 678 gpr.lr = data.GetU64(&offset); 679 gpr.sp = data.GetU64(&offset); 680 gpr.pc = data.GetU64(&offset); 681 gpr.cpsr = data.GetU32(&offset); 682 SetError(GPRRegSet, Read, 0); 683 } 684 offset = next_thread_state; 685 break; 686 case FPURegSet: { 687 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 688 const int fpu_reg_buf_size = sizeof(fpu); 689 if (fpu_reg_buf_size == count && 690 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 691 fpu_reg_buf) == fpu_reg_buf_size) { 692 SetError(FPURegSet, Read, 0); 693 } else { 694 done = true; 695 } 696 } 697 offset = next_thread_state; 698 break; 699 case EXCRegSet: 700 if (count == 4) { 701 exc.far = data.GetU64(&offset); 702 exc.esr = data.GetU32(&offset); 703 exc.exception = data.GetU32(&offset); 704 SetError(EXCRegSet, Read, 0); 705 } 706 offset = next_thread_state; 707 break; 708 default: 709 done = true; 710 break; 711 } 712 } 713 } 714 715 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 716 const char *alt_name, size_t reg_byte_size, 717 Stream &data) { 718 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 719 if (reg_info == NULL) 720 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 721 if (reg_info) { 722 lldb_private::RegisterValue reg_value; 723 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 724 if (reg_info->byte_size >= reg_byte_size) 725 data.Write(reg_value.GetBytes(), reg_byte_size); 726 else { 727 data.Write(reg_value.GetBytes(), reg_info->byte_size); 728 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 729 ++i) 730 data.PutChar(0); 731 } 732 return reg_byte_size; 733 } 734 } 735 // Just write zeros if all else fails 736 for (size_t i = 0; i < reg_byte_size; ++i) 737 data.PutChar(0); 738 return reg_byte_size; 739 } 740 741 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 742 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 743 if (reg_ctx_sp) { 744 RegisterContext *reg_ctx = reg_ctx_sp.get(); 745 746 data.PutHex32(GPRRegSet); // Flavor 747 data.PutHex32(GPRWordCount); 748 WriteRegister(reg_ctx, "x0", NULL, 8, data); 749 WriteRegister(reg_ctx, "x1", NULL, 8, data); 750 WriteRegister(reg_ctx, "x2", NULL, 8, data); 751 WriteRegister(reg_ctx, "x3", NULL, 8, data); 752 WriteRegister(reg_ctx, "x4", NULL, 8, data); 753 WriteRegister(reg_ctx, "x5", NULL, 8, data); 754 WriteRegister(reg_ctx, "x6", NULL, 8, data); 755 WriteRegister(reg_ctx, "x7", NULL, 8, data); 756 WriteRegister(reg_ctx, "x8", NULL, 8, data); 757 WriteRegister(reg_ctx, "x9", NULL, 8, data); 758 WriteRegister(reg_ctx, "x10", NULL, 8, data); 759 WriteRegister(reg_ctx, "x11", NULL, 8, data); 760 WriteRegister(reg_ctx, "x12", NULL, 8, data); 761 WriteRegister(reg_ctx, "x13", NULL, 8, data); 762 WriteRegister(reg_ctx, "x14", NULL, 8, data); 763 WriteRegister(reg_ctx, "x15", NULL, 8, data); 764 WriteRegister(reg_ctx, "x16", NULL, 8, data); 765 WriteRegister(reg_ctx, "x17", NULL, 8, data); 766 WriteRegister(reg_ctx, "x18", NULL, 8, data); 767 WriteRegister(reg_ctx, "x19", NULL, 8, data); 768 WriteRegister(reg_ctx, "x20", NULL, 8, data); 769 WriteRegister(reg_ctx, "x21", NULL, 8, data); 770 WriteRegister(reg_ctx, "x22", NULL, 8, data); 771 WriteRegister(reg_ctx, "x23", NULL, 8, data); 772 WriteRegister(reg_ctx, "x24", NULL, 8, data); 773 WriteRegister(reg_ctx, "x25", NULL, 8, data); 774 WriteRegister(reg_ctx, "x26", NULL, 8, data); 775 WriteRegister(reg_ctx, "x27", NULL, 8, data); 776 WriteRegister(reg_ctx, "x28", NULL, 8, data); 777 WriteRegister(reg_ctx, "fp", NULL, 8, data); 778 WriteRegister(reg_ctx, "lr", NULL, 8, data); 779 WriteRegister(reg_ctx, "sp", NULL, 8, data); 780 WriteRegister(reg_ctx, "pc", NULL, 8, data); 781 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 782 783 // Write out the EXC registers 784 // data.PutHex32 (EXCRegSet); 785 // data.PutHex32 (EXCWordCount); 786 // WriteRegister (reg_ctx, "far", NULL, 8, data); 787 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 788 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 789 return true; 790 } 791 return false; 792 } 793 794 protected: 795 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 796 797 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 798 799 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 800 801 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 802 803 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 804 return 0; 805 } 806 807 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 808 return 0; 809 } 810 811 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 812 return 0; 813 } 814 815 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 816 return -1; 817 } 818 }; 819 820 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 821 switch (magic) { 822 case MH_MAGIC: 823 case MH_CIGAM: 824 return sizeof(struct mach_header); 825 826 case MH_MAGIC_64: 827 case MH_CIGAM_64: 828 return sizeof(struct mach_header_64); 829 break; 830 831 default: 832 break; 833 } 834 return 0; 835 } 836 837 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 838 839 void ObjectFileMachO::Initialize() { 840 PluginManager::RegisterPlugin( 841 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 842 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 843 } 844 845 void ObjectFileMachO::Terminate() { 846 PluginManager::UnregisterPlugin(CreateInstance); 847 } 848 849 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 850 static ConstString g_name("mach-o"); 851 return g_name; 852 } 853 854 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 855 return "Mach-o object file reader (32 and 64 bit)"; 856 } 857 858 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 859 DataBufferSP &data_sp, 860 lldb::offset_t data_offset, 861 const FileSpec *file, 862 lldb::offset_t file_offset, 863 lldb::offset_t length) { 864 if (!data_sp) { 865 data_sp = 866 DataBufferLLVM::CreateSliceFromPath(file->GetPath(), length, file_offset); 867 if (!data_sp) 868 return nullptr; 869 data_offset = 0; 870 } 871 872 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 873 return nullptr; 874 875 // Update the data to contain the entire file if it doesn't already 876 if (data_sp->GetByteSize() < length) { 877 data_sp = 878 DataBufferLLVM::CreateSliceFromPath(file->GetPath(), length, file_offset); 879 if (!data_sp) 880 return nullptr; 881 data_offset = 0; 882 } 883 auto objfile_ap = llvm::make_unique<ObjectFileMachO>( 884 module_sp, data_sp, data_offset, file, file_offset, length); 885 if (!objfile_ap || !objfile_ap->ParseHeader()) 886 return nullptr; 887 888 return objfile_ap.release(); 889 } 890 891 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 892 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 893 const ProcessSP &process_sp, lldb::addr_t header_addr) { 894 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 895 std::unique_ptr<ObjectFile> objfile_ap( 896 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 897 if (objfile_ap.get() && objfile_ap->ParseHeader()) 898 return objfile_ap.release(); 899 } 900 return NULL; 901 } 902 903 size_t ObjectFileMachO::GetModuleSpecifications( 904 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 905 lldb::offset_t data_offset, lldb::offset_t file_offset, 906 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 907 const size_t initial_count = specs.GetSize(); 908 909 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 910 DataExtractor data; 911 data.SetData(data_sp); 912 llvm::MachO::mach_header header; 913 if (ParseHeader(data, &data_offset, header)) { 914 size_t header_and_load_cmds = 915 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 916 if (header_and_load_cmds >= data_sp->GetByteSize()) { 917 data_sp = DataBufferLLVM::CreateSliceFromPath( 918 file.GetPath(), header_and_load_cmds, file_offset); 919 data.SetData(data_sp); 920 data_offset = MachHeaderSizeFromMagic(header.magic); 921 } 922 if (data_sp) { 923 ModuleSpec spec; 924 spec.GetFileSpec() = file; 925 spec.SetObjectOffset(file_offset); 926 spec.SetObjectSize(length); 927 928 if (GetArchitecture(header, data, data_offset, 929 spec.GetArchitecture())) { 930 if (spec.GetArchitecture().IsValid()) { 931 GetUUID(header, data, data_offset, spec.GetUUID()); 932 specs.Append(spec); 933 } 934 } 935 } 936 } 937 } 938 return specs.GetSize() - initial_count; 939 } 940 941 const ConstString &ObjectFileMachO::GetSegmentNameTEXT() { 942 static ConstString g_segment_name_TEXT("__TEXT"); 943 return g_segment_name_TEXT; 944 } 945 946 const ConstString &ObjectFileMachO::GetSegmentNameDATA() { 947 static ConstString g_segment_name_DATA("__DATA"); 948 return g_segment_name_DATA; 949 } 950 951 const ConstString &ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 952 static ConstString g_segment_name("__DATA_DIRTY"); 953 return g_segment_name; 954 } 955 956 const ConstString &ObjectFileMachO::GetSegmentNameDATA_CONST() { 957 static ConstString g_segment_name("__DATA_CONST"); 958 return g_segment_name; 959 } 960 961 const ConstString &ObjectFileMachO::GetSegmentNameOBJC() { 962 static ConstString g_segment_name_OBJC("__OBJC"); 963 return g_segment_name_OBJC; 964 } 965 966 const ConstString &ObjectFileMachO::GetSegmentNameLINKEDIT() { 967 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 968 return g_section_name_LINKEDIT; 969 } 970 971 const ConstString &ObjectFileMachO::GetSectionNameEHFrame() { 972 static ConstString g_section_name_eh_frame("__eh_frame"); 973 return g_section_name_eh_frame; 974 } 975 976 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 977 lldb::addr_t data_offset, 978 lldb::addr_t data_length) { 979 DataExtractor data; 980 data.SetData(data_sp, data_offset, data_length); 981 lldb::offset_t offset = 0; 982 uint32_t magic = data.GetU32(&offset); 983 return MachHeaderSizeFromMagic(magic) != 0; 984 } 985 986 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 987 DataBufferSP &data_sp, 988 lldb::offset_t data_offset, 989 const FileSpec *file, 990 lldb::offset_t file_offset, 991 lldb::offset_t length) 992 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 993 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 994 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 995 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 996 ::memset(&m_header, 0, sizeof(m_header)); 997 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 998 } 999 1000 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 1001 lldb::DataBufferSP &header_data_sp, 1002 const lldb::ProcessSP &process_sp, 1003 lldb::addr_t header_addr) 1004 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 1005 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 1006 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 1007 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 1008 ::memset(&m_header, 0, sizeof(m_header)); 1009 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1010 } 1011 1012 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 1013 lldb::offset_t *data_offset_ptr, 1014 llvm::MachO::mach_header &header) { 1015 data.SetByteOrder(endian::InlHostByteOrder()); 1016 // Leave magic in the original byte order 1017 header.magic = data.GetU32(data_offset_ptr); 1018 bool can_parse = false; 1019 bool is_64_bit = false; 1020 switch (header.magic) { 1021 case MH_MAGIC: 1022 data.SetByteOrder(endian::InlHostByteOrder()); 1023 data.SetAddressByteSize(4); 1024 can_parse = true; 1025 break; 1026 1027 case MH_MAGIC_64: 1028 data.SetByteOrder(endian::InlHostByteOrder()); 1029 data.SetAddressByteSize(8); 1030 can_parse = true; 1031 is_64_bit = true; 1032 break; 1033 1034 case MH_CIGAM: 1035 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1036 ? eByteOrderLittle 1037 : eByteOrderBig); 1038 data.SetAddressByteSize(4); 1039 can_parse = true; 1040 break; 1041 1042 case MH_CIGAM_64: 1043 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1044 ? eByteOrderLittle 1045 : eByteOrderBig); 1046 data.SetAddressByteSize(8); 1047 is_64_bit = true; 1048 can_parse = true; 1049 break; 1050 1051 default: 1052 break; 1053 } 1054 1055 if (can_parse) { 1056 data.GetU32(data_offset_ptr, &header.cputype, 6); 1057 if (is_64_bit) 1058 *data_offset_ptr += 4; 1059 return true; 1060 } else { 1061 memset(&header, 0, sizeof(header)); 1062 } 1063 return false; 1064 } 1065 1066 bool ObjectFileMachO::ParseHeader() { 1067 ModuleSP module_sp(GetModule()); 1068 if (module_sp) { 1069 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1070 bool can_parse = false; 1071 lldb::offset_t offset = 0; 1072 m_data.SetByteOrder(endian::InlHostByteOrder()); 1073 // Leave magic in the original byte order 1074 m_header.magic = m_data.GetU32(&offset); 1075 switch (m_header.magic) { 1076 case MH_MAGIC: 1077 m_data.SetByteOrder(endian::InlHostByteOrder()); 1078 m_data.SetAddressByteSize(4); 1079 can_parse = true; 1080 break; 1081 1082 case MH_MAGIC_64: 1083 m_data.SetByteOrder(endian::InlHostByteOrder()); 1084 m_data.SetAddressByteSize(8); 1085 can_parse = true; 1086 break; 1087 1088 case MH_CIGAM: 1089 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1090 ? eByteOrderLittle 1091 : eByteOrderBig); 1092 m_data.SetAddressByteSize(4); 1093 can_parse = true; 1094 break; 1095 1096 case MH_CIGAM_64: 1097 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1098 ? eByteOrderLittle 1099 : eByteOrderBig); 1100 m_data.SetAddressByteSize(8); 1101 can_parse = true; 1102 break; 1103 1104 default: 1105 break; 1106 } 1107 1108 if (can_parse) { 1109 m_data.GetU32(&offset, &m_header.cputype, 6); 1110 1111 ArchSpec mach_arch; 1112 1113 if (GetArchitecture(mach_arch)) { 1114 // Check if the module has a required architecture 1115 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1116 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1117 return false; 1118 1119 if (SetModulesArchitecture(mach_arch)) { 1120 const size_t header_and_lc_size = 1121 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1122 if (m_data.GetByteSize() < header_and_lc_size) { 1123 DataBufferSP data_sp; 1124 ProcessSP process_sp(m_process_wp.lock()); 1125 if (process_sp) { 1126 data_sp = 1127 ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1128 } else { 1129 // Read in all only the load command data from the file on disk 1130 data_sp = DataBufferLLVM::CreateSliceFromPath( 1131 m_file.GetPath(), header_and_lc_size, m_file_offset); 1132 if (data_sp->GetByteSize() != header_and_lc_size) 1133 return false; 1134 } 1135 if (data_sp) 1136 m_data.SetData(data_sp); 1137 } 1138 } 1139 return true; 1140 } 1141 } else { 1142 memset(&m_header, 0, sizeof(struct mach_header)); 1143 } 1144 } 1145 return false; 1146 } 1147 1148 ByteOrder ObjectFileMachO::GetByteOrder() const { 1149 return m_data.GetByteOrder(); 1150 } 1151 1152 bool ObjectFileMachO::IsExecutable() const { 1153 return m_header.filetype == MH_EXECUTE; 1154 } 1155 1156 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1157 return m_data.GetAddressByteSize(); 1158 } 1159 1160 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1161 Symtab *symtab = GetSymtab(); 1162 if (symtab) { 1163 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1164 if (symbol) { 1165 if (symbol->ValueIsAddress()) { 1166 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1167 if (section_sp) { 1168 const lldb::SectionType section_type = section_sp->GetType(); 1169 switch (section_type) { 1170 case eSectionTypeInvalid: 1171 return eAddressClassUnknown; 1172 1173 case eSectionTypeCode: 1174 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1175 // For ARM we have a bit in the n_desc field of the symbol 1176 // that tells us ARM/Thumb which is bit 0x0008. 1177 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1178 return eAddressClassCodeAlternateISA; 1179 } 1180 return eAddressClassCode; 1181 1182 case eSectionTypeContainer: 1183 return eAddressClassUnknown; 1184 1185 case eSectionTypeData: 1186 case eSectionTypeDataCString: 1187 case eSectionTypeDataCStringPointers: 1188 case eSectionTypeDataSymbolAddress: 1189 case eSectionTypeData4: 1190 case eSectionTypeData8: 1191 case eSectionTypeData16: 1192 case eSectionTypeDataPointers: 1193 case eSectionTypeZeroFill: 1194 case eSectionTypeDataObjCMessageRefs: 1195 case eSectionTypeDataObjCCFStrings: 1196 case eSectionTypeGoSymtab: 1197 return eAddressClassData; 1198 1199 case eSectionTypeDebug: 1200 case eSectionTypeDWARFDebugAbbrev: 1201 case eSectionTypeDWARFDebugAddr: 1202 case eSectionTypeDWARFDebugAranges: 1203 case eSectionTypeDWARFDebugCuIndex: 1204 case eSectionTypeDWARFDebugFrame: 1205 case eSectionTypeDWARFDebugInfo: 1206 case eSectionTypeDWARFDebugLine: 1207 case eSectionTypeDWARFDebugLoc: 1208 case eSectionTypeDWARFDebugMacInfo: 1209 case eSectionTypeDWARFDebugMacro: 1210 case eSectionTypeDWARFDebugPubNames: 1211 case eSectionTypeDWARFDebugPubTypes: 1212 case eSectionTypeDWARFDebugRanges: 1213 case eSectionTypeDWARFDebugStr: 1214 case eSectionTypeDWARFDebugStrOffsets: 1215 case eSectionTypeDWARFAppleNames: 1216 case eSectionTypeDWARFAppleTypes: 1217 case eSectionTypeDWARFAppleNamespaces: 1218 case eSectionTypeDWARFAppleObjC: 1219 return eAddressClassDebug; 1220 1221 case eSectionTypeEHFrame: 1222 case eSectionTypeARMexidx: 1223 case eSectionTypeARMextab: 1224 case eSectionTypeCompactUnwind: 1225 return eAddressClassRuntime; 1226 1227 case eSectionTypeAbsoluteAddress: 1228 case eSectionTypeELFSymbolTable: 1229 case eSectionTypeELFDynamicSymbols: 1230 case eSectionTypeELFRelocationEntries: 1231 case eSectionTypeELFDynamicLinkInfo: 1232 case eSectionTypeOther: 1233 return eAddressClassUnknown; 1234 } 1235 } 1236 } 1237 1238 const SymbolType symbol_type = symbol->GetType(); 1239 switch (symbol_type) { 1240 case eSymbolTypeAny: 1241 return eAddressClassUnknown; 1242 case eSymbolTypeAbsolute: 1243 return eAddressClassUnknown; 1244 1245 case eSymbolTypeCode: 1246 case eSymbolTypeTrampoline: 1247 case eSymbolTypeResolver: 1248 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1249 // For ARM we have a bit in the n_desc field of the symbol 1250 // that tells us ARM/Thumb which is bit 0x0008. 1251 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1252 return eAddressClassCodeAlternateISA; 1253 } 1254 return eAddressClassCode; 1255 1256 case eSymbolTypeData: 1257 return eAddressClassData; 1258 case eSymbolTypeRuntime: 1259 return eAddressClassRuntime; 1260 case eSymbolTypeException: 1261 return eAddressClassRuntime; 1262 case eSymbolTypeSourceFile: 1263 return eAddressClassDebug; 1264 case eSymbolTypeHeaderFile: 1265 return eAddressClassDebug; 1266 case eSymbolTypeObjectFile: 1267 return eAddressClassDebug; 1268 case eSymbolTypeCommonBlock: 1269 return eAddressClassDebug; 1270 case eSymbolTypeBlock: 1271 return eAddressClassDebug; 1272 case eSymbolTypeLocal: 1273 return eAddressClassData; 1274 case eSymbolTypeParam: 1275 return eAddressClassData; 1276 case eSymbolTypeVariable: 1277 return eAddressClassData; 1278 case eSymbolTypeVariableType: 1279 return eAddressClassDebug; 1280 case eSymbolTypeLineEntry: 1281 return eAddressClassDebug; 1282 case eSymbolTypeLineHeader: 1283 return eAddressClassDebug; 1284 case eSymbolTypeScopeBegin: 1285 return eAddressClassDebug; 1286 case eSymbolTypeScopeEnd: 1287 return eAddressClassDebug; 1288 case eSymbolTypeAdditional: 1289 return eAddressClassUnknown; 1290 case eSymbolTypeCompiler: 1291 return eAddressClassDebug; 1292 case eSymbolTypeInstrumentation: 1293 return eAddressClassDebug; 1294 case eSymbolTypeUndefined: 1295 return eAddressClassUnknown; 1296 case eSymbolTypeObjCClass: 1297 return eAddressClassRuntime; 1298 case eSymbolTypeObjCMetaClass: 1299 return eAddressClassRuntime; 1300 case eSymbolTypeObjCIVar: 1301 return eAddressClassRuntime; 1302 case eSymbolTypeReExported: 1303 return eAddressClassRuntime; 1304 } 1305 } 1306 } 1307 return eAddressClassUnknown; 1308 } 1309 1310 Symtab *ObjectFileMachO::GetSymtab() { 1311 ModuleSP module_sp(GetModule()); 1312 if (module_sp) { 1313 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1314 if (m_symtab_ap.get() == NULL) { 1315 m_symtab_ap.reset(new Symtab(this)); 1316 std::lock_guard<std::recursive_mutex> symtab_guard( 1317 m_symtab_ap->GetMutex()); 1318 ParseSymtab(); 1319 m_symtab_ap->Finalize(); 1320 } 1321 } 1322 return m_symtab_ap.get(); 1323 } 1324 1325 bool ObjectFileMachO::IsStripped() { 1326 if (m_dysymtab.cmd == 0) { 1327 ModuleSP module_sp(GetModule()); 1328 if (module_sp) { 1329 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1330 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1331 const lldb::offset_t load_cmd_offset = offset; 1332 1333 load_command lc; 1334 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1335 break; 1336 if (lc.cmd == LC_DYSYMTAB) { 1337 m_dysymtab.cmd = lc.cmd; 1338 m_dysymtab.cmdsize = lc.cmdsize; 1339 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1340 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1341 NULL) { 1342 // Clear m_dysymtab if we were unable to read all items from the 1343 // load command 1344 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1345 } 1346 } 1347 offset = load_cmd_offset + lc.cmdsize; 1348 } 1349 } 1350 } 1351 if (m_dysymtab.cmd) 1352 return m_dysymtab.nlocalsym <= 1; 1353 return false; 1354 } 1355 1356 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1357 if (!m_sections_ap.get()) { 1358 m_sections_ap.reset(new SectionList()); 1359 1360 const bool is_dsym = (m_header.filetype == MH_DSYM); 1361 lldb::user_id_t segID = 0; 1362 lldb::user_id_t sectID = 0; 1363 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1364 uint32_t i; 1365 const bool is_core = GetType() == eTypeCoreFile; 1366 // bool dump_sections = false; 1367 ModuleSP module_sp(GetModule()); 1368 // First look up any LC_ENCRYPTION_INFO load commands 1369 typedef RangeArray<uint32_t, uint32_t, 8> EncryptedFileRanges; 1370 EncryptedFileRanges encrypted_file_ranges; 1371 encryption_info_command encryption_cmd; 1372 for (i = 0; i < m_header.ncmds; ++i) { 1373 const lldb::offset_t load_cmd_offset = offset; 1374 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1375 break; 1376 1377 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for 1378 // the 3 fields we care about, so treat them the same. 1379 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1380 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1381 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1382 if (encryption_cmd.cryptid != 0) { 1383 EncryptedFileRanges::Entry entry; 1384 entry.SetRangeBase(encryption_cmd.cryptoff); 1385 entry.SetByteSize(encryption_cmd.cryptsize); 1386 encrypted_file_ranges.Append(entry); 1387 } 1388 } 1389 } 1390 offset = load_cmd_offset + encryption_cmd.cmdsize; 1391 } 1392 1393 bool section_file_addresses_changed = false; 1394 1395 offset = MachHeaderSizeFromMagic(m_header.magic); 1396 1397 struct segment_command_64 load_cmd; 1398 for (i = 0; i < m_header.ncmds; ++i) { 1399 const lldb::offset_t load_cmd_offset = offset; 1400 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1401 break; 1402 1403 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) { 1404 if (m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) { 1405 bool add_section = true; 1406 bool add_to_unified = true; 1407 ConstString const_segname(load_cmd.segname, 1408 std::min<size_t>(strlen(load_cmd.segname), 1409 sizeof(load_cmd.segname))); 1410 1411 SectionSP unified_section_sp( 1412 unified_section_list.FindSectionByName(const_segname)); 1413 if (is_dsym && unified_section_sp) { 1414 if (const_segname == GetSegmentNameLINKEDIT()) { 1415 // We need to keep the __LINKEDIT segment private to this object 1416 // file only 1417 add_to_unified = false; 1418 } else { 1419 // This is the dSYM file and this section has already been created 1420 // by 1421 // the object file, no need to create it. 1422 add_section = false; 1423 } 1424 } 1425 load_cmd.vmaddr = m_data.GetAddress(&offset); 1426 load_cmd.vmsize = m_data.GetAddress(&offset); 1427 load_cmd.fileoff = m_data.GetAddress(&offset); 1428 load_cmd.filesize = m_data.GetAddress(&offset); 1429 if (m_length != 0 && load_cmd.filesize != 0) { 1430 if (load_cmd.fileoff > m_length) { 1431 // We have a load command that says it extends past the end of the 1432 // file. This is likely 1433 // a corrupt file. We don't have any way to return an error 1434 // condition here (this method 1435 // was likely invoked from something like 1436 // ObjectFile::GetSectionList()) -- all we can do 1437 // is null out the SectionList vector and if a process has been 1438 // set up, dump a message 1439 // to stdout. The most common case here is core file debugging 1440 // with a truncated file. 1441 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 1442 ? "LC_SEGMENT_64" 1443 : "LC_SEGMENT"; 1444 module_sp->ReportWarning( 1445 "load command %u %s has a fileoff (0x%" PRIx64 1446 ") that extends beyond the end of the file (0x%" PRIx64 1447 "), ignoring this section", 1448 i, lc_segment_name, load_cmd.fileoff, m_length); 1449 1450 load_cmd.fileoff = 0; 1451 load_cmd.filesize = 0; 1452 } 1453 1454 if (load_cmd.fileoff + load_cmd.filesize > m_length) { 1455 // We have a load command that says it extends past the end of the 1456 // file. This is likely 1457 // a corrupt file. We don't have any way to return an error 1458 // condition here (this method 1459 // was likely invoked from something like 1460 // ObjectFile::GetSectionList()) -- all we can do 1461 // is null out the SectionList vector and if a process has been 1462 // set up, dump a message 1463 // to stdout. The most common case here is core file debugging 1464 // with a truncated file. 1465 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 1466 ? "LC_SEGMENT_64" 1467 : "LC_SEGMENT"; 1468 GetModule()->ReportWarning( 1469 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1470 ") that extends beyond the end of the file (0x%" PRIx64 1471 "), the segment will be truncated to match", 1472 i, lc_segment_name, load_cmd.fileoff + load_cmd.filesize, 1473 m_length); 1474 1475 // Tuncase the length 1476 load_cmd.filesize = m_length - load_cmd.fileoff; 1477 } 1478 } 1479 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) { 1480 uint32_t segment_permissions = 0; 1481 if (load_cmd.initprot & VM_PROT_READ) 1482 segment_permissions |= ePermissionsReadable; 1483 if (load_cmd.initprot & VM_PROT_WRITE) 1484 segment_permissions |= ePermissionsWritable; 1485 if (load_cmd.initprot & VM_PROT_EXECUTE) 1486 segment_permissions |= ePermissionsExecutable; 1487 1488 const bool segment_is_encrypted = 1489 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1490 1491 // Keep a list of mach segments around in case we need to 1492 // get at data that isn't stored in the abstracted Sections. 1493 m_mach_segments.push_back(load_cmd); 1494 1495 // Use a segment ID of the segment index shifted left by 8 so they 1496 // never conflict with any of the sections. 1497 SectionSP segment_sp; 1498 if (add_section && (const_segname || is_core)) { 1499 segment_sp.reset(new Section( 1500 module_sp, // Module to which this section belongs 1501 this, // Object file to which this sections belongs 1502 ++segID << 8, // Section ID is the 1 based segment index 1503 // shifted right by 8 bits as not to collide 1504 // with any of the 256 section IDs that are 1505 // possible 1506 const_segname, // Name of this section 1507 eSectionTypeContainer, // This section is a container of other 1508 // sections. 1509 load_cmd.vmaddr, // File VM address == addresses as they are 1510 // found in the object file 1511 load_cmd.vmsize, // VM size in bytes of this section 1512 load_cmd.fileoff, // Offset to the data for this section in 1513 // the file 1514 load_cmd.filesize, // Size in bytes of this section as found 1515 // in the file 1516 0, // Segments have no alignment information 1517 load_cmd.flags)); // Flags for this section 1518 1519 segment_sp->SetIsEncrypted(segment_is_encrypted); 1520 m_sections_ap->AddSection(segment_sp); 1521 segment_sp->SetPermissions(segment_permissions); 1522 if (add_to_unified) 1523 unified_section_list.AddSection(segment_sp); 1524 } else if (unified_section_sp) { 1525 if (is_dsym && 1526 unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1527 // Check to see if the module was read from memory? 1528 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) { 1529 // We have a module that is in memory and needs to have its 1530 // file address adjusted. We need to do this because when we 1531 // load a file from memory, its addresses will be slid 1532 // already, 1533 // yet the addresses in the new symbol file will still be 1534 // unslid. 1535 // Since everything is stored as section offset, this 1536 // shouldn't 1537 // cause any problems. 1538 1539 // Make sure we've parsed the symbol table from the 1540 // ObjectFile before we go around changing its Sections. 1541 module_sp->GetObjectFile()->GetSymtab(); 1542 // eh_frame would present the same problems but we parse that 1543 // on 1544 // a per-function basis as-needed so it's more difficult to 1545 // remove its use of the Sections. Realistically, the 1546 // environments 1547 // where this code path will be taken will not have eh_frame 1548 // sections. 1549 1550 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1551 1552 // Notify the module that the section addresses have been 1553 // changed once 1554 // we're done so any file-address caches can be updated. 1555 section_file_addresses_changed = true; 1556 } 1557 } 1558 m_sections_ap->AddSection(unified_section_sp); 1559 } 1560 1561 struct section_64 sect64; 1562 ::memset(§64, 0, sizeof(sect64)); 1563 // Push a section into our mach sections for the section at 1564 // index zero (NO_SECT) if we don't have any mach sections yet... 1565 if (m_mach_sections.empty()) 1566 m_mach_sections.push_back(sect64); 1567 uint32_t segment_sect_idx; 1568 const lldb::user_id_t first_segment_sectID = sectID + 1; 1569 1570 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1571 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1572 ++segment_sect_idx) { 1573 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1574 sizeof(sect64.sectname)) == NULL) 1575 break; 1576 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1577 sizeof(sect64.segname)) == NULL) 1578 break; 1579 sect64.addr = m_data.GetAddress(&offset); 1580 sect64.size = m_data.GetAddress(&offset); 1581 1582 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1583 break; 1584 1585 // Keep a list of mach sections around in case we need to 1586 // get at data that isn't stored in the abstracted Sections. 1587 m_mach_sections.push_back(sect64); 1588 1589 if (add_section) { 1590 ConstString section_name( 1591 sect64.sectname, std::min<size_t>(strlen(sect64.sectname), 1592 sizeof(sect64.sectname))); 1593 if (!const_segname) { 1594 // We have a segment with no name so we need to conjure up 1595 // segments that correspond to the section's segname if there 1596 // isn't already such a section. If there is such a section, 1597 // we resize the section so that it spans all sections. 1598 // We also mark these sections as fake so address matches 1599 // don't 1600 // hit if they land in the gaps between the child sections. 1601 const_segname.SetTrimmedCStringWithLength( 1602 sect64.segname, sizeof(sect64.segname)); 1603 segment_sp = 1604 unified_section_list.FindSectionByName(const_segname); 1605 if (segment_sp.get()) { 1606 Section *segment = segment_sp.get(); 1607 // Grow the section size as needed. 1608 const lldb::addr_t sect64_min_addr = sect64.addr; 1609 const lldb::addr_t sect64_max_addr = 1610 sect64_min_addr + sect64.size; 1611 const lldb::addr_t curr_seg_byte_size = 1612 segment->GetByteSize(); 1613 const lldb::addr_t curr_seg_min_addr = 1614 segment->GetFileAddress(); 1615 const lldb::addr_t curr_seg_max_addr = 1616 curr_seg_min_addr + curr_seg_byte_size; 1617 if (sect64_min_addr >= curr_seg_min_addr) { 1618 const lldb::addr_t new_seg_byte_size = 1619 sect64_max_addr - curr_seg_min_addr; 1620 // Only grow the section size if needed 1621 if (new_seg_byte_size > curr_seg_byte_size) 1622 segment->SetByteSize(new_seg_byte_size); 1623 } else { 1624 // We need to change the base address of the segment and 1625 // adjust the child section offsets for all existing 1626 // children. 1627 const lldb::addr_t slide_amount = 1628 sect64_min_addr - curr_seg_min_addr; 1629 segment->Slide(slide_amount, false); 1630 segment->GetChildren().Slide(-slide_amount, false); 1631 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1632 } 1633 1634 // Grow the section size as needed. 1635 if (sect64.offset) { 1636 const lldb::addr_t segment_min_file_offset = 1637 segment->GetFileOffset(); 1638 const lldb::addr_t segment_max_file_offset = 1639 segment_min_file_offset + segment->GetFileSize(); 1640 1641 const lldb::addr_t section_min_file_offset = 1642 sect64.offset; 1643 const lldb::addr_t section_max_file_offset = 1644 section_min_file_offset + sect64.size; 1645 const lldb::addr_t new_file_offset = std::min( 1646 section_min_file_offset, segment_min_file_offset); 1647 const lldb::addr_t new_file_size = 1648 std::max(section_max_file_offset, 1649 segment_max_file_offset) - 1650 new_file_offset; 1651 segment->SetFileOffset(new_file_offset); 1652 segment->SetFileSize(new_file_size); 1653 } 1654 } else { 1655 // Create a fake section for the section's named segment 1656 segment_sp.reset(new Section( 1657 segment_sp, // Parent section 1658 module_sp, // Module to which this section belongs 1659 this, // Object file to which this section belongs 1660 ++segID << 8, // Section ID is the 1 based segment index 1661 // shifted right by 8 bits as not to 1662 // collide with any of the 256 section IDs 1663 // that are possible 1664 const_segname, // Name of this section 1665 eSectionTypeContainer, // This section is a container of 1666 // other sections. 1667 sect64.addr, // File VM address == addresses as they are 1668 // found in the object file 1669 sect64.size, // VM size in bytes of this section 1670 sect64.offset, // Offset to the data for this section in 1671 // the file 1672 sect64.offset ? sect64.size : 0, // Size in bytes of 1673 // this section as 1674 // found in the file 1675 sect64.align, 1676 load_cmd.flags)); // Flags for this section 1677 segment_sp->SetIsFake(true); 1678 segment_sp->SetPermissions(segment_permissions); 1679 m_sections_ap->AddSection(segment_sp); 1680 if (add_to_unified) 1681 unified_section_list.AddSection(segment_sp); 1682 segment_sp->SetIsEncrypted(segment_is_encrypted); 1683 } 1684 } 1685 assert(segment_sp.get()); 1686 1687 lldb::SectionType sect_type = eSectionTypeOther; 1688 1689 if (sect64.flags & 1690 (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1691 sect_type = eSectionTypeCode; 1692 else { 1693 uint32_t mach_sect_type = sect64.flags & SECTION_TYPE; 1694 static ConstString g_sect_name_objc_data("__objc_data"); 1695 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1696 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1697 static ConstString g_sect_name_objc_classrefs( 1698 "__objc_classrefs"); 1699 static ConstString g_sect_name_objc_superrefs( 1700 "__objc_superrefs"); 1701 static ConstString g_sect_name_objc_const("__objc_const"); 1702 static ConstString g_sect_name_objc_classlist( 1703 "__objc_classlist"); 1704 static ConstString g_sect_name_cfstring("__cfstring"); 1705 1706 static ConstString g_sect_name_dwarf_debug_abbrev( 1707 "__debug_abbrev"); 1708 static ConstString g_sect_name_dwarf_debug_aranges( 1709 "__debug_aranges"); 1710 static ConstString g_sect_name_dwarf_debug_frame( 1711 "__debug_frame"); 1712 static ConstString g_sect_name_dwarf_debug_info( 1713 "__debug_info"); 1714 static ConstString g_sect_name_dwarf_debug_line( 1715 "__debug_line"); 1716 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1717 static ConstString g_sect_name_dwarf_debug_macinfo( 1718 "__debug_macinfo"); 1719 static ConstString g_sect_name_dwarf_debug_pubnames( 1720 "__debug_pubnames"); 1721 static ConstString g_sect_name_dwarf_debug_pubtypes( 1722 "__debug_pubtypes"); 1723 static ConstString g_sect_name_dwarf_debug_ranges( 1724 "__debug_ranges"); 1725 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1726 static ConstString g_sect_name_dwarf_apple_names( 1727 "__apple_names"); 1728 static ConstString g_sect_name_dwarf_apple_types( 1729 "__apple_types"); 1730 static ConstString g_sect_name_dwarf_apple_namespaces( 1731 "__apple_namespac"); 1732 static ConstString g_sect_name_dwarf_apple_objc( 1733 "__apple_objc"); 1734 static ConstString g_sect_name_eh_frame("__eh_frame"); 1735 static ConstString g_sect_name_compact_unwind( 1736 "__unwind_info"); 1737 static ConstString g_sect_name_text("__text"); 1738 static ConstString g_sect_name_data("__data"); 1739 static ConstString g_sect_name_go_symtab("__gosymtab"); 1740 1741 if (section_name == g_sect_name_dwarf_debug_abbrev) 1742 sect_type = eSectionTypeDWARFDebugAbbrev; 1743 else if (section_name == g_sect_name_dwarf_debug_aranges) 1744 sect_type = eSectionTypeDWARFDebugAranges; 1745 else if (section_name == g_sect_name_dwarf_debug_frame) 1746 sect_type = eSectionTypeDWARFDebugFrame; 1747 else if (section_name == g_sect_name_dwarf_debug_info) 1748 sect_type = eSectionTypeDWARFDebugInfo; 1749 else if (section_name == g_sect_name_dwarf_debug_line) 1750 sect_type = eSectionTypeDWARFDebugLine; 1751 else if (section_name == g_sect_name_dwarf_debug_loc) 1752 sect_type = eSectionTypeDWARFDebugLoc; 1753 else if (section_name == g_sect_name_dwarf_debug_macinfo) 1754 sect_type = eSectionTypeDWARFDebugMacInfo; 1755 else if (section_name == g_sect_name_dwarf_debug_pubnames) 1756 sect_type = eSectionTypeDWARFDebugPubNames; 1757 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 1758 sect_type = eSectionTypeDWARFDebugPubTypes; 1759 else if (section_name == g_sect_name_dwarf_debug_ranges) 1760 sect_type = eSectionTypeDWARFDebugRanges; 1761 else if (section_name == g_sect_name_dwarf_debug_str) 1762 sect_type = eSectionTypeDWARFDebugStr; 1763 else if (section_name == g_sect_name_dwarf_apple_names) 1764 sect_type = eSectionTypeDWARFAppleNames; 1765 else if (section_name == g_sect_name_dwarf_apple_types) 1766 sect_type = eSectionTypeDWARFAppleTypes; 1767 else if (section_name == g_sect_name_dwarf_apple_namespaces) 1768 sect_type = eSectionTypeDWARFAppleNamespaces; 1769 else if (section_name == g_sect_name_dwarf_apple_objc) 1770 sect_type = eSectionTypeDWARFAppleObjC; 1771 else if (section_name == g_sect_name_objc_selrefs) 1772 sect_type = eSectionTypeDataCStringPointers; 1773 else if (section_name == g_sect_name_objc_msgrefs) 1774 sect_type = eSectionTypeDataObjCMessageRefs; 1775 else if (section_name == g_sect_name_eh_frame) 1776 sect_type = eSectionTypeEHFrame; 1777 else if (section_name == g_sect_name_compact_unwind) 1778 sect_type = eSectionTypeCompactUnwind; 1779 else if (section_name == g_sect_name_cfstring) 1780 sect_type = eSectionTypeDataObjCCFStrings; 1781 else if (section_name == g_sect_name_go_symtab) 1782 sect_type = eSectionTypeGoSymtab; 1783 else if (section_name == g_sect_name_objc_data || 1784 section_name == g_sect_name_objc_classrefs || 1785 section_name == g_sect_name_objc_superrefs || 1786 section_name == g_sect_name_objc_const || 1787 section_name == g_sect_name_objc_classlist) { 1788 sect_type = eSectionTypeDataPointers; 1789 } 1790 1791 if (sect_type == eSectionTypeOther) { 1792 switch (mach_sect_type) { 1793 // TODO: categorize sections by other flags for regular 1794 // sections 1795 case S_REGULAR: 1796 if (section_name == g_sect_name_text) 1797 sect_type = eSectionTypeCode; 1798 else if (section_name == g_sect_name_data) 1799 sect_type = eSectionTypeData; 1800 else 1801 sect_type = eSectionTypeOther; 1802 break; 1803 case S_ZEROFILL: 1804 sect_type = eSectionTypeZeroFill; 1805 break; 1806 case S_CSTRING_LITERALS: 1807 sect_type = eSectionTypeDataCString; 1808 break; // section with only literal C strings 1809 case S_4BYTE_LITERALS: 1810 sect_type = eSectionTypeData4; 1811 break; // section with only 4 byte literals 1812 case S_8BYTE_LITERALS: 1813 sect_type = eSectionTypeData8; 1814 break; // section with only 8 byte literals 1815 case S_LITERAL_POINTERS: 1816 sect_type = eSectionTypeDataPointers; 1817 break; // section with only pointers to literals 1818 case S_NON_LAZY_SYMBOL_POINTERS: 1819 sect_type = eSectionTypeDataPointers; 1820 break; // section with only non-lazy symbol pointers 1821 case S_LAZY_SYMBOL_POINTERS: 1822 sect_type = eSectionTypeDataPointers; 1823 break; // section with only lazy symbol pointers 1824 case S_SYMBOL_STUBS: 1825 sect_type = eSectionTypeCode; 1826 break; // section with only symbol stubs, byte size of 1827 // stub in the reserved2 field 1828 case S_MOD_INIT_FUNC_POINTERS: 1829 sect_type = eSectionTypeDataPointers; 1830 break; // section with only function pointers for 1831 // initialization 1832 case S_MOD_TERM_FUNC_POINTERS: 1833 sect_type = eSectionTypeDataPointers; 1834 break; // section with only function pointers for 1835 // termination 1836 case S_COALESCED: 1837 sect_type = eSectionTypeOther; 1838 break; 1839 case S_GB_ZEROFILL: 1840 sect_type = eSectionTypeZeroFill; 1841 break; 1842 case S_INTERPOSING: 1843 sect_type = eSectionTypeCode; 1844 break; // section with only pairs of function pointers for 1845 // interposing 1846 case S_16BYTE_LITERALS: 1847 sect_type = eSectionTypeData16; 1848 break; // section with only 16 byte literals 1849 case S_DTRACE_DOF: 1850 sect_type = eSectionTypeDebug; 1851 break; 1852 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1853 sect_type = eSectionTypeDataPointers; 1854 break; 1855 default: 1856 break; 1857 } 1858 } 1859 } 1860 1861 SectionSP section_sp(new Section( 1862 segment_sp, module_sp, this, ++sectID, section_name, 1863 sect_type, sect64.addr - segment_sp->GetFileAddress(), 1864 sect64.size, sect64.offset, 1865 sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1866 sect64.flags)); 1867 // Set the section to be encrypted to match the segment 1868 1869 bool section_is_encrypted = false; 1870 if (!segment_is_encrypted && load_cmd.filesize != 0) 1871 section_is_encrypted = 1872 encrypted_file_ranges.FindEntryThatContains( 1873 sect64.offset) != NULL; 1874 1875 section_sp->SetIsEncrypted(segment_is_encrypted || 1876 section_is_encrypted); 1877 section_sp->SetPermissions(segment_permissions); 1878 segment_sp->GetChildren().AddSection(section_sp); 1879 1880 if (segment_sp->IsFake()) { 1881 segment_sp.reset(); 1882 const_segname.Clear(); 1883 } 1884 } 1885 } 1886 if (segment_sp && is_dsym) { 1887 if (first_segment_sectID <= sectID) { 1888 lldb::user_id_t sect_uid; 1889 for (sect_uid = first_segment_sectID; sect_uid <= sectID; 1890 ++sect_uid) { 1891 SectionSP curr_section_sp( 1892 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1893 SectionSP next_section_sp; 1894 if (sect_uid + 1 <= sectID) 1895 next_section_sp = 1896 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1897 1898 if (curr_section_sp.get()) { 1899 if (curr_section_sp->GetByteSize() == 0) { 1900 if (next_section_sp.get() != NULL) 1901 curr_section_sp->SetByteSize( 1902 next_section_sp->GetFileAddress() - 1903 curr_section_sp->GetFileAddress()); 1904 else 1905 curr_section_sp->SetByteSize(load_cmd.vmsize); 1906 } 1907 } 1908 } 1909 } 1910 } 1911 } 1912 } 1913 } else if (load_cmd.cmd == LC_DYSYMTAB) { 1914 m_dysymtab.cmd = load_cmd.cmd; 1915 m_dysymtab.cmdsize = load_cmd.cmdsize; 1916 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1917 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1918 } 1919 1920 offset = load_cmd_offset + load_cmd.cmdsize; 1921 } 1922 1923 if (section_file_addresses_changed && module_sp.get()) { 1924 module_sp->SectionFileAddressesChanged(); 1925 } 1926 } 1927 } 1928 1929 class MachSymtabSectionInfo { 1930 public: 1931 MachSymtabSectionInfo(SectionList *section_list) 1932 : m_section_list(section_list), m_section_infos() { 1933 // Get the number of sections down to a depth of 1 to include 1934 // all segments and their sections, but no other sections that 1935 // may be added for debug map or 1936 m_section_infos.resize(section_list->GetNumSections(1)); 1937 } 1938 1939 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1940 if (n_sect == 0) 1941 return SectionSP(); 1942 if (n_sect < m_section_infos.size()) { 1943 if (!m_section_infos[n_sect].section_sp) { 1944 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1945 m_section_infos[n_sect].section_sp = section_sp; 1946 if (section_sp) { 1947 m_section_infos[n_sect].vm_range.SetBaseAddress( 1948 section_sp->GetFileAddress()); 1949 m_section_infos[n_sect].vm_range.SetByteSize( 1950 section_sp->GetByteSize()); 1951 } else { 1952 Host::SystemLog(Host::eSystemLogError, 1953 "error: unable to find section for section %u\n", 1954 n_sect); 1955 } 1956 } 1957 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1958 // Symbol is in section. 1959 return m_section_infos[n_sect].section_sp; 1960 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1961 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1962 file_addr) { 1963 // Symbol is in section with zero size, but has the same start 1964 // address as the section. This can happen with linker symbols 1965 // (symbols that start with the letter 'l' or 'L'. 1966 return m_section_infos[n_sect].section_sp; 1967 } 1968 } 1969 return m_section_list->FindSectionContainingFileAddress(file_addr); 1970 } 1971 1972 protected: 1973 struct SectionInfo { 1974 SectionInfo() : vm_range(), section_sp() {} 1975 1976 VMRange vm_range; 1977 SectionSP section_sp; 1978 }; 1979 SectionList *m_section_list; 1980 std::vector<SectionInfo> m_section_infos; 1981 }; 1982 1983 struct TrieEntry { 1984 TrieEntry() 1985 : name(), address(LLDB_INVALID_ADDRESS), flags(0), other(0), 1986 import_name() {} 1987 1988 void Clear() { 1989 name.Clear(); 1990 address = LLDB_INVALID_ADDRESS; 1991 flags = 0; 1992 other = 0; 1993 import_name.Clear(); 1994 } 1995 1996 void Dump() const { 1997 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1998 static_cast<unsigned long long>(address), 1999 static_cast<unsigned long long>(flags), 2000 static_cast<unsigned long long>(other), name.GetCString()); 2001 if (import_name) 2002 printf(" -> \"%s\"\n", import_name.GetCString()); 2003 else 2004 printf("\n"); 2005 } 2006 ConstString name; 2007 uint64_t address; 2008 uint64_t flags; 2009 uint64_t other; 2010 ConstString import_name; 2011 }; 2012 2013 struct TrieEntryWithOffset { 2014 lldb::offset_t nodeOffset; 2015 TrieEntry entry; 2016 2017 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 2018 2019 void Dump(uint32_t idx) const { 2020 printf("[%3u] 0x%16.16llx: ", idx, 2021 static_cast<unsigned long long>(nodeOffset)); 2022 entry.Dump(); 2023 } 2024 2025 bool operator<(const TrieEntryWithOffset &other) const { 2026 return (nodeOffset < other.nodeOffset); 2027 } 2028 }; 2029 2030 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2031 const bool is_arm, 2032 std::vector<llvm::StringRef> &nameSlices, 2033 std::set<lldb::addr_t> &resolver_addresses, 2034 std::vector<TrieEntryWithOffset> &output) { 2035 if (!data.ValidOffset(offset)) 2036 return true; 2037 2038 const uint64_t terminalSize = data.GetULEB128(&offset); 2039 lldb::offset_t children_offset = offset + terminalSize; 2040 if (terminalSize != 0) { 2041 TrieEntryWithOffset e(offset); 2042 e.entry.flags = data.GetULEB128(&offset); 2043 const char *import_name = NULL; 2044 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2045 e.entry.address = 0; 2046 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2047 import_name = data.GetCStr(&offset); 2048 } else { 2049 e.entry.address = data.GetULEB128(&offset); 2050 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2051 e.entry.other = data.GetULEB128(&offset); 2052 uint64_t resolver_addr = e.entry.other; 2053 if (is_arm) 2054 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2055 resolver_addresses.insert(resolver_addr); 2056 } else 2057 e.entry.other = 0; 2058 } 2059 // Only add symbols that are reexport symbols with a valid import name 2060 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && 2061 import_name[0]) { 2062 std::string name; 2063 if (!nameSlices.empty()) { 2064 for (auto name_slice : nameSlices) 2065 name.append(name_slice.data(), name_slice.size()); 2066 } 2067 if (name.size() > 1) { 2068 // Skip the leading '_' 2069 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2070 } 2071 if (import_name) { 2072 // Skip the leading '_' 2073 e.entry.import_name.SetCString(import_name + 1); 2074 } 2075 output.push_back(e); 2076 } 2077 } 2078 2079 const uint8_t childrenCount = data.GetU8(&children_offset); 2080 for (uint8_t i = 0; i < childrenCount; ++i) { 2081 const char *cstr = data.GetCStr(&children_offset); 2082 if (cstr) 2083 nameSlices.push_back(llvm::StringRef(cstr)); 2084 else 2085 return false; // Corrupt data 2086 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2087 if (childNodeOffset) { 2088 if (!ParseTrieEntries(data, childNodeOffset, is_arm, nameSlices, 2089 resolver_addresses, output)) { 2090 return false; 2091 } 2092 } 2093 nameSlices.pop_back(); 2094 } 2095 return true; 2096 } 2097 2098 // Read the UUID out of a dyld_shared_cache file on-disk. 2099 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2100 const ByteOrder byte_order, 2101 const uint32_t addr_byte_size) { 2102 UUID dsc_uuid; 2103 DataBufferSP DscData = DataBufferLLVM::CreateSliceFromPath( 2104 dyld_shared_cache.GetPath(), 2105 sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2106 if (!DscData) 2107 return dsc_uuid; 2108 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2109 2110 char version_str[7]; 2111 lldb::offset_t offset = 0; 2112 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2113 version_str[6] = '\0'; 2114 if (strcmp(version_str, "dyld_v") == 0) { 2115 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2116 uint8_t uuid_bytes[sizeof(uuid_t)]; 2117 memcpy(uuid_bytes, dsc_header_data.GetData(&offset, sizeof(uuid_t)), 2118 sizeof(uuid_t)); 2119 dsc_uuid.SetBytes(uuid_bytes); 2120 } 2121 return dsc_uuid; 2122 } 2123 2124 size_t ObjectFileMachO::ParseSymtab() { 2125 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2126 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2127 m_file.GetFilename().AsCString("")); 2128 ModuleSP module_sp(GetModule()); 2129 if (!module_sp) 2130 return 0; 2131 2132 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2133 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2134 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2135 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2136 FunctionStarts function_starts; 2137 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2138 uint32_t i; 2139 FileSpecList dylib_files; 2140 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2141 static const llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2142 static const llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2143 static const llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2144 2145 for (i = 0; i < m_header.ncmds; ++i) { 2146 const lldb::offset_t cmd_offset = offset; 2147 // Read in the load command and load command size 2148 struct load_command lc; 2149 if (m_data.GetU32(&offset, &lc, 2) == NULL) 2150 break; 2151 // Watch for the symbol table load command 2152 switch (lc.cmd) { 2153 case LC_SYMTAB: 2154 symtab_load_command.cmd = lc.cmd; 2155 symtab_load_command.cmdsize = lc.cmdsize; 2156 // Read in the rest of the symtab load command 2157 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2158 0) // fill in symoff, nsyms, stroff, strsize fields 2159 return 0; 2160 if (symtab_load_command.symoff == 0) { 2161 if (log) 2162 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2163 return 0; 2164 } 2165 2166 if (symtab_load_command.stroff == 0) { 2167 if (log) 2168 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2169 return 0; 2170 } 2171 2172 if (symtab_load_command.nsyms == 0) { 2173 if (log) 2174 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2175 return 0; 2176 } 2177 2178 if (symtab_load_command.strsize == 0) { 2179 if (log) 2180 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2181 return 0; 2182 } 2183 break; 2184 2185 case LC_DYLD_INFO: 2186 case LC_DYLD_INFO_ONLY: 2187 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2188 dyld_info.cmd = lc.cmd; 2189 dyld_info.cmdsize = lc.cmdsize; 2190 } else { 2191 memset(&dyld_info, 0, sizeof(dyld_info)); 2192 } 2193 break; 2194 2195 case LC_LOAD_DYLIB: 2196 case LC_LOAD_WEAK_DYLIB: 2197 case LC_REEXPORT_DYLIB: 2198 case LC_LOADFVMLIB: 2199 case LC_LOAD_UPWARD_DYLIB: { 2200 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2201 const char *path = m_data.PeekCStr(name_offset); 2202 if (path) { 2203 FileSpec file_spec(path, false); 2204 // Strip the path if there is @rpath, @executable, etc so we just use 2205 // the basename 2206 if (path[0] == '@') 2207 file_spec.GetDirectory().Clear(); 2208 2209 if (lc.cmd == LC_REEXPORT_DYLIB) { 2210 m_reexported_dylibs.AppendIfUnique(file_spec); 2211 } 2212 2213 dylib_files.Append(file_spec); 2214 } 2215 } break; 2216 2217 case LC_FUNCTION_STARTS: 2218 function_starts_load_command.cmd = lc.cmd; 2219 function_starts_load_command.cmdsize = lc.cmdsize; 2220 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2221 NULL) // fill in symoff, nsyms, stroff, strsize fields 2222 memset(&function_starts_load_command, 0, 2223 sizeof(function_starts_load_command)); 2224 break; 2225 2226 default: 2227 break; 2228 } 2229 offset = cmd_offset + lc.cmdsize; 2230 } 2231 2232 if (symtab_load_command.cmd) { 2233 Symtab *symtab = m_symtab_ap.get(); 2234 SectionList *section_list = GetSectionList(); 2235 if (section_list == NULL) 2236 return 0; 2237 2238 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2239 const ByteOrder byte_order = m_data.GetByteOrder(); 2240 bool bit_width_32 = addr_byte_size == 4; 2241 const size_t nlist_byte_size = 2242 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2243 2244 DataExtractor nlist_data(NULL, 0, byte_order, addr_byte_size); 2245 DataExtractor strtab_data(NULL, 0, byte_order, addr_byte_size); 2246 DataExtractor function_starts_data(NULL, 0, byte_order, addr_byte_size); 2247 DataExtractor indirect_symbol_index_data(NULL, 0, byte_order, 2248 addr_byte_size); 2249 DataExtractor dyld_trie_data(NULL, 0, byte_order, addr_byte_size); 2250 2251 const addr_t nlist_data_byte_size = 2252 symtab_load_command.nsyms * nlist_byte_size; 2253 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2254 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2255 2256 ProcessSP process_sp(m_process_wp.lock()); 2257 Process *process = process_sp.get(); 2258 2259 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2260 2261 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2262 Target &target = process->GetTarget(); 2263 2264 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2265 2266 SectionSP linkedit_section_sp( 2267 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2268 // Reading mach file from memory in a process or core file... 2269 2270 if (linkedit_section_sp) { 2271 addr_t linkedit_load_addr = 2272 linkedit_section_sp->GetLoadBaseAddress(&target); 2273 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2274 // We might be trying to access the symbol table before the 2275 // __LINKEDIT's load 2276 // address has been set in the target. We can't fail to read the 2277 // symbol table, 2278 // so calculate the right address manually 2279 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2280 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2281 } 2282 2283 const addr_t linkedit_file_offset = 2284 linkedit_section_sp->GetFileOffset(); 2285 const addr_t symoff_addr = linkedit_load_addr + 2286 symtab_load_command.symoff - 2287 linkedit_file_offset; 2288 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2289 linkedit_file_offset; 2290 2291 bool data_was_read = false; 2292 2293 #if defined(__APPLE__) && \ 2294 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2295 if (m_header.flags & 0x80000000u && 2296 process->GetAddressByteSize() == sizeof(void *)) { 2297 // This mach-o memory file is in the dyld shared cache. If this 2298 // program is not remote and this is iOS, then this process will 2299 // share the same shared cache as the process we are debugging and 2300 // we can read the entire __LINKEDIT from the address space in this 2301 // process. This is a needed optimization that is used for local iOS 2302 // debugging only since all shared libraries in the shared cache do 2303 // not have corresponding files that exist in the file system of the 2304 // device. They have been combined into a single file. This means we 2305 // always have to load these files from memory. All of the symbol and 2306 // string tables from all of the __LINKEDIT sections from the shared 2307 // libraries in the shared cache have been merged into a single large 2308 // symbol and string table. Reading all of this symbol and string 2309 // table 2310 // data across can slow down debug launch times, so we optimize this 2311 // by 2312 // reading the memory for the __LINKEDIT section from this process. 2313 2314 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 2315 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 2316 bool use_lldb_cache = true; 2317 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2318 lldb_shared_cache != process_shared_cache) { 2319 use_lldb_cache = false; 2320 ModuleSP module_sp(GetModule()); 2321 if (module_sp) 2322 module_sp->ReportWarning("shared cache in process does not match " 2323 "lldb's own shared cache, startup will " 2324 "be slow."); 2325 } 2326 2327 PlatformSP platform_sp(target.GetPlatform()); 2328 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2329 data_was_read = true; 2330 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2331 eByteOrderLittle); 2332 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2333 eByteOrderLittle); 2334 if (function_starts_load_command.cmd) { 2335 const addr_t func_start_addr = 2336 linkedit_load_addr + function_starts_load_command.dataoff - 2337 linkedit_file_offset; 2338 function_starts_data.SetData( 2339 (void *)func_start_addr, 2340 function_starts_load_command.datasize, eByteOrderLittle); 2341 } 2342 } 2343 } 2344 #endif 2345 2346 if (!data_was_read) { 2347 // Always load dyld - the dynamic linker - from memory if we didn't 2348 // find a binary anywhere else. 2349 // lldb will not register dylib/framework/bundle loads/unloads if we 2350 // don't have the dyld symbols, 2351 // we force dyld to load from memory despite the user's 2352 // target.memory-module-load-level setting. 2353 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2354 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2355 DataBufferSP nlist_data_sp( 2356 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2357 if (nlist_data_sp) 2358 nlist_data.SetData(nlist_data_sp, 0, 2359 nlist_data_sp->GetByteSize()); 2360 // Load strings individually from memory when loading from memory 2361 // since shared cache 2362 // string tables contain strings for all symbols from all shared 2363 // cached libraries 2364 // DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, 2365 // strtab_data_byte_size)); 2366 // if (strtab_data_sp) 2367 // strtab_data.SetData (strtab_data_sp, 0, 2368 // strtab_data_sp->GetByteSize()); 2369 if (m_dysymtab.nindirectsyms != 0) { 2370 const addr_t indirect_syms_addr = linkedit_load_addr + 2371 m_dysymtab.indirectsymoff - 2372 linkedit_file_offset; 2373 DataBufferSP indirect_syms_data_sp( 2374 ReadMemory(process_sp, indirect_syms_addr, 2375 m_dysymtab.nindirectsyms * 4)); 2376 if (indirect_syms_data_sp) 2377 indirect_symbol_index_data.SetData( 2378 indirect_syms_data_sp, 0, 2379 indirect_syms_data_sp->GetByteSize()); 2380 } 2381 } else if (memory_module_load_level >= 2382 eMemoryModuleLoadLevelPartial) { 2383 if (function_starts_load_command.cmd) { 2384 const addr_t func_start_addr = 2385 linkedit_load_addr + function_starts_load_command.dataoff - 2386 linkedit_file_offset; 2387 DataBufferSP func_start_data_sp( 2388 ReadMemory(process_sp, func_start_addr, 2389 function_starts_load_command.datasize)); 2390 if (func_start_data_sp) 2391 function_starts_data.SetData(func_start_data_sp, 0, 2392 func_start_data_sp->GetByteSize()); 2393 } 2394 } 2395 } 2396 } 2397 } else { 2398 nlist_data.SetData(m_data, symtab_load_command.symoff, 2399 nlist_data_byte_size); 2400 strtab_data.SetData(m_data, symtab_load_command.stroff, 2401 strtab_data_byte_size); 2402 2403 if (dyld_info.export_size > 0) { 2404 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2405 dyld_info.export_size); 2406 } 2407 2408 if (m_dysymtab.nindirectsyms != 0) { 2409 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2410 m_dysymtab.nindirectsyms * 4); 2411 } 2412 if (function_starts_load_command.cmd) { 2413 function_starts_data.SetData(m_data, 2414 function_starts_load_command.dataoff, 2415 function_starts_load_command.datasize); 2416 } 2417 } 2418 2419 if (nlist_data.GetByteSize() == 0 && 2420 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2421 if (log) 2422 module_sp->LogMessage(log, "failed to read nlist data"); 2423 return 0; 2424 } 2425 2426 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2427 if (!have_strtab_data) { 2428 if (process) { 2429 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2430 if (log) 2431 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2432 return 0; 2433 } 2434 } else { 2435 if (log) 2436 module_sp->LogMessage(log, "failed to read strtab data"); 2437 return 0; 2438 } 2439 } 2440 2441 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2442 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2443 const ConstString &g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2444 const ConstString &g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2445 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2446 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2447 SectionSP text_section_sp( 2448 section_list->FindSectionByName(g_segment_name_TEXT)); 2449 SectionSP data_section_sp( 2450 section_list->FindSectionByName(g_segment_name_DATA)); 2451 SectionSP data_dirty_section_sp( 2452 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2453 SectionSP data_const_section_sp( 2454 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2455 SectionSP objc_section_sp( 2456 section_list->FindSectionByName(g_segment_name_OBJC)); 2457 SectionSP eh_frame_section_sp; 2458 if (text_section_sp.get()) 2459 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2460 g_section_name_eh_frame); 2461 else 2462 eh_frame_section_sp = 2463 section_list->FindSectionByName(g_section_name_eh_frame); 2464 2465 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2466 2467 // lldb works best if it knows the start address of all functions in a 2468 // module. 2469 // Linker symbols or debug info are normally the best source of information 2470 // for start addr / size but 2471 // they may be stripped in a released binary. 2472 // Two additional sources of information exist in Mach-O binaries: 2473 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2474 // function's start address in the 2475 // binary, relative to the text section. 2476 // eh_frame - the eh_frame FDEs have the start addr & size of 2477 // each function 2478 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2479 // all modern binaries. 2480 // Binaries built to run on older releases may need to use eh_frame 2481 // information. 2482 2483 if (text_section_sp && function_starts_data.GetByteSize()) { 2484 FunctionStarts::Entry function_start_entry; 2485 function_start_entry.data = false; 2486 lldb::offset_t function_start_offset = 0; 2487 function_start_entry.addr = text_section_sp->GetFileAddress(); 2488 uint64_t delta; 2489 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2490 0) { 2491 // Now append the current entry 2492 function_start_entry.addr += delta; 2493 function_starts.Append(function_start_entry); 2494 } 2495 } else { 2496 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2497 // load command claiming an eh_frame 2498 // but it doesn't actually have the eh_frame content. And if we have a 2499 // dSYM, we don't need to do any 2500 // of this fill-in-the-missing-symbols works anyway - the debug info 2501 // should give us all the functions in 2502 // the module. 2503 if (text_section_sp.get() && eh_frame_section_sp.get() && 2504 m_type != eTypeDebugInfo) { 2505 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2506 DWARFCallFrameInfo::EH); 2507 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2508 eh_frame.GetFunctionAddressAndSizeVector(functions); 2509 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2510 size_t count = functions.GetSize(); 2511 for (size_t i = 0; i < count; ++i) { 2512 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2513 functions.GetEntryAtIndex(i); 2514 if (func) { 2515 FunctionStarts::Entry function_start_entry; 2516 function_start_entry.addr = func->base - text_base_addr; 2517 function_starts.Append(function_start_entry); 2518 } 2519 } 2520 } 2521 } 2522 2523 const size_t function_starts_count = function_starts.GetSize(); 2524 2525 // For user process binaries (executables, dylibs, frameworks, bundles), if 2526 // we don't have 2527 // LC_FUNCTION_STARTS/eh_frame section in this binary, we're going to assume 2528 // the binary 2529 // has been stripped. Don't allow assembly language instruction emulation 2530 // because we don't 2531 // know proper function start boundaries. 2532 // 2533 // For all other types of binaries (kernels, stand-alone bare board 2534 // binaries, kexts), they 2535 // may not have LC_FUNCTION_STARTS / eh_frame sections - we should not make 2536 // any assumptions 2537 // about them based on that. 2538 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2539 m_allow_assembly_emulation_unwind_plans = false; 2540 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2541 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2542 2543 if (unwind_or_symbol_log) 2544 module_sp->LogMessage( 2545 unwind_or_symbol_log, 2546 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2547 } 2548 2549 const user_id_t TEXT_eh_frame_sectID = 2550 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2551 : static_cast<user_id_t>(NO_SECT); 2552 2553 lldb::offset_t nlist_data_offset = 0; 2554 2555 uint32_t N_SO_index = UINT32_MAX; 2556 2557 MachSymtabSectionInfo section_info(section_list); 2558 std::vector<uint32_t> N_FUN_indexes; 2559 std::vector<uint32_t> N_NSYM_indexes; 2560 std::vector<uint32_t> N_INCL_indexes; 2561 std::vector<uint32_t> N_BRAC_indexes; 2562 std::vector<uint32_t> N_COMM_indexes; 2563 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2564 typedef std::map<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2565 typedef std::map<const char *, uint32_t> ConstNameToSymbolIndexMap; 2566 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2567 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2568 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2569 // Any symbols that get merged into another will get an entry 2570 // in this map so we know 2571 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2572 uint32_t nlist_idx = 0; 2573 Symbol *symbol_ptr = NULL; 2574 2575 uint32_t sym_idx = 0; 2576 Symbol *sym = NULL; 2577 size_t num_syms = 0; 2578 std::string memory_symbol_name; 2579 uint32_t unmapped_local_symbols_found = 0; 2580 2581 std::vector<TrieEntryWithOffset> trie_entries; 2582 std::set<lldb::addr_t> resolver_addresses; 2583 2584 if (dyld_trie_data.GetByteSize() > 0) { 2585 std::vector<llvm::StringRef> nameSlices; 2586 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, 2587 resolver_addresses, trie_entries); 2588 2589 ConstString text_segment_name("__TEXT"); 2590 SectionSP text_segment_sp = 2591 GetSectionList()->FindSectionByName(text_segment_name); 2592 if (text_segment_sp) { 2593 const lldb::addr_t text_segment_file_addr = 2594 text_segment_sp->GetFileAddress(); 2595 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2596 for (auto &e : trie_entries) 2597 e.entry.address += text_segment_file_addr; 2598 } 2599 } 2600 } 2601 2602 typedef std::set<ConstString> IndirectSymbols; 2603 IndirectSymbols indirect_symbol_names; 2604 2605 #if defined(__APPLE__) && \ 2606 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2607 2608 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2609 // optimized by moving LOCAL 2610 // symbols out of the memory mapped portion of the DSC. The symbol 2611 // information has all been retained, 2612 // but it isn't available in the normal nlist data. However, there *are* 2613 // duplicate entries of *some* 2614 // LOCAL symbols in the normal nlist data. To handle this situation 2615 // correctly, we must first attempt 2616 // to parse any DSC unmapped symbol information. If we find any, we set a 2617 // flag that tells the normal 2618 // nlist parser to ignore all LOCAL symbols. 2619 2620 if (m_header.flags & 0x80000000u) { 2621 // Before we can start mapping the DSC, we need to make certain the target 2622 // process is actually 2623 // using the cache we can find. 2624 2625 // Next we need to determine the correct path for the dyld shared cache. 2626 2627 ArchSpec header_arch; 2628 GetArchitecture(header_arch); 2629 char dsc_path[PATH_MAX]; 2630 char dsc_path_development[PATH_MAX]; 2631 2632 snprintf( 2633 dsc_path, sizeof(dsc_path), "%s%s%s", 2634 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2635 */ 2636 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2637 header_arch.GetArchitectureName()); 2638 2639 snprintf( 2640 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2641 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2642 */ 2643 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2644 header_arch.GetArchitectureName(), ".development"); 2645 2646 FileSpec dsc_nondevelopment_filespec(dsc_path, false); 2647 FileSpec dsc_development_filespec(dsc_path_development, false); 2648 FileSpec dsc_filespec; 2649 2650 UUID dsc_uuid; 2651 UUID process_shared_cache_uuid; 2652 2653 if (process) { 2654 process_shared_cache_uuid = GetProcessSharedCacheUUID(process); 2655 } 2656 2657 // First see if we can find an exact match for the inferior process shared 2658 // cache UUID in 2659 // the development or non-development shared caches on disk. 2660 if (process_shared_cache_uuid.IsValid()) { 2661 if (dsc_development_filespec.Exists()) { 2662 UUID dsc_development_uuid = GetSharedCacheUUID( 2663 dsc_development_filespec, byte_order, addr_byte_size); 2664 if (dsc_development_uuid.IsValid() && 2665 dsc_development_uuid == process_shared_cache_uuid) { 2666 dsc_filespec = dsc_development_filespec; 2667 dsc_uuid = dsc_development_uuid; 2668 } 2669 } 2670 if (!dsc_uuid.IsValid() && dsc_nondevelopment_filespec.Exists()) { 2671 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2672 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2673 if (dsc_nondevelopment_uuid.IsValid() && 2674 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2675 dsc_filespec = dsc_nondevelopment_filespec; 2676 dsc_uuid = dsc_nondevelopment_uuid; 2677 } 2678 } 2679 } 2680 2681 // Failing a UUID match, prefer the development dyld_shared cache if both 2682 // are present. 2683 if (!dsc_filespec.Exists()) { 2684 if (dsc_development_filespec.Exists()) { 2685 dsc_filespec = dsc_development_filespec; 2686 } else { 2687 dsc_filespec = dsc_nondevelopment_filespec; 2688 } 2689 } 2690 2691 /* The dyld_cache_header has a pointer to the 2692 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2693 The dyld_cache_local_symbols_info structure gives us three things: 2694 1. The start and count of the nlist records in the dyld_shared_cache 2695 file 2696 2. The start and size of the strings for these nlist records 2697 3. The start and count of dyld_cache_local_symbols_entry entries 2698 2699 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2700 dyld shared cache. 2701 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2702 the dyld shared cache. 2703 The dyld_cache_local_symbols_entry also lists the start of this 2704 dylib/framework's nlist records 2705 and the count of how many nlist records there are for this 2706 dylib/framework. 2707 */ 2708 2709 // Process the dyld shared cache header to find the unmapped symbols 2710 2711 DataBufferSP dsc_data_sp = DataBufferLLVM::CreateSliceFromPath( 2712 dsc_filespec.GetPath(), sizeof(struct lldb_copy_dyld_cache_header_v1), 2713 0); 2714 if (!dsc_uuid.IsValid()) { 2715 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2716 } 2717 if (dsc_data_sp) { 2718 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2719 2720 bool uuid_match = true; 2721 if (dsc_uuid.IsValid() && process) { 2722 if (process_shared_cache_uuid.IsValid() && 2723 dsc_uuid != process_shared_cache_uuid) { 2724 // The on-disk dyld_shared_cache file is not the same as the one in 2725 // this 2726 // process' memory, don't use it. 2727 uuid_match = false; 2728 ModuleSP module_sp(GetModule()); 2729 if (module_sp) 2730 module_sp->ReportWarning("process shared cache does not match " 2731 "on-disk dyld_shared_cache file, some " 2732 "symbol names will be missing."); 2733 } 2734 } 2735 2736 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2737 2738 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2739 2740 // If the mappingOffset points to a location inside the header, we've 2741 // opened an old dyld shared cache, and should not proceed further. 2742 if (uuid_match && 2743 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2744 2745 DataBufferSP dsc_mapping_info_data_sp = 2746 DataBufferLLVM::CreateSliceFromPath( 2747 dsc_filespec.GetPath(), 2748 sizeof(struct lldb_copy_dyld_cache_mapping_info), 2749 mappingOffset); 2750 2751 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2752 byte_order, addr_byte_size); 2753 offset = 0; 2754 2755 // The File addresses (from the in-memory Mach-O load commands) for 2756 // the shared libraries 2757 // in the shared library cache need to be adjusted by an offset to 2758 // match up with the 2759 // dylibOffset identifying field in the 2760 // dyld_cache_local_symbol_entry's. This offset is 2761 // recorded in mapping_offset_value. 2762 const uint64_t mapping_offset_value = 2763 dsc_mapping_info_data.GetU64(&offset); 2764 2765 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, 2766 localSymbolsOffset); 2767 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2768 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2769 2770 if (localSymbolsOffset && localSymbolsSize) { 2771 // Map the local symbols 2772 DataBufferSP dsc_local_symbols_data_sp = 2773 DataBufferLLVM::CreateSliceFromPath(dsc_filespec.GetPath(), 2774 localSymbolsSize, 2775 localSymbolsOffset); 2776 2777 if (dsc_local_symbols_data_sp) { 2778 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2779 byte_order, addr_byte_size); 2780 2781 offset = 0; 2782 2783 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2784 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2785 UndefinedNameToDescMap undefined_name_to_desc; 2786 SymbolIndexToName reexport_shlib_needs_fixup; 2787 2788 // Read the local_symbols_infos struct in one shot 2789 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2790 dsc_local_symbols_data.GetU32(&offset, 2791 &local_symbols_info.nlistOffset, 6); 2792 2793 SectionSP text_section_sp( 2794 section_list->FindSectionByName(GetSegmentNameTEXT())); 2795 2796 uint32_t header_file_offset = 2797 (text_section_sp->GetFileAddress() - mapping_offset_value); 2798 2799 offset = local_symbols_info.entriesOffset; 2800 for (uint32_t entry_index = 0; 2801 entry_index < local_symbols_info.entriesCount; 2802 entry_index++) { 2803 struct lldb_copy_dyld_cache_local_symbols_entry 2804 local_symbols_entry; 2805 local_symbols_entry.dylibOffset = 2806 dsc_local_symbols_data.GetU32(&offset); 2807 local_symbols_entry.nlistStartIndex = 2808 dsc_local_symbols_data.GetU32(&offset); 2809 local_symbols_entry.nlistCount = 2810 dsc_local_symbols_data.GetU32(&offset); 2811 2812 if (header_file_offset == local_symbols_entry.dylibOffset) { 2813 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2814 2815 // The normal nlist code cannot correctly size the Symbols 2816 // array, we need to allocate it here. 2817 sym = symtab->Resize( 2818 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2819 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2820 num_syms = symtab->GetNumSymbols(); 2821 2822 nlist_data_offset = 2823 local_symbols_info.nlistOffset + 2824 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2825 uint32_t string_table_offset = 2826 local_symbols_info.stringsOffset; 2827 2828 for (uint32_t nlist_index = 0; 2829 nlist_index < local_symbols_entry.nlistCount; 2830 nlist_index++) { 2831 ///////////////////////////// 2832 { 2833 struct nlist_64 nlist; 2834 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize( 2835 nlist_data_offset, nlist_byte_size)) 2836 break; 2837 2838 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked( 2839 &nlist_data_offset); 2840 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked( 2841 &nlist_data_offset); 2842 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked( 2843 &nlist_data_offset); 2844 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked( 2845 &nlist_data_offset); 2846 nlist.n_value = 2847 dsc_local_symbols_data.GetAddress_unchecked( 2848 &nlist_data_offset); 2849 2850 SymbolType type = eSymbolTypeInvalid; 2851 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2852 string_table_offset + nlist.n_strx); 2853 2854 if (symbol_name == NULL) { 2855 // No symbol should be NULL, even the symbols with no 2856 // string values should have an offset zero which points 2857 // to an empty C-string 2858 Host::SystemLog( 2859 Host::eSystemLogError, 2860 "error: DSC unmapped local symbol[%u] has invalid " 2861 "string table offset 0x%x in %s, ignoring symbol\n", 2862 entry_index, nlist.n_strx, 2863 module_sp->GetFileSpec().GetPath().c_str()); 2864 continue; 2865 } 2866 if (symbol_name[0] == '\0') 2867 symbol_name = NULL; 2868 2869 const char *symbol_name_non_abi_mangled = NULL; 2870 2871 SectionSP symbol_section; 2872 uint32_t symbol_byte_size = 0; 2873 bool add_nlist = true; 2874 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2875 bool demangled_is_synthesized = false; 2876 bool is_gsym = false; 2877 bool set_value = true; 2878 2879 assert(sym_idx < num_syms); 2880 2881 sym[sym_idx].SetDebug(is_debug); 2882 2883 if (is_debug) { 2884 switch (nlist.n_type) { 2885 case N_GSYM: 2886 // global symbol: name,,NO_SECT,type,0 2887 // Sometimes the N_GSYM value contains the address. 2888 2889 // FIXME: In the .o files, we have a GSYM and a debug 2890 // symbol for all the ObjC data. They 2891 // have the same address, but we want to ensure that 2892 // we always find only the real symbol, 2893 // 'cause we don't currently correctly attribute the 2894 // GSYM one to the ObjCClass/Ivar/MetaClass 2895 // symbol type. This is a temporary hack to make sure 2896 // the ObjectiveC symbols get treated 2897 // correctly. To do this right, we should coalesce 2898 // all the GSYM & global symbols that have the 2899 // same address. 2900 2901 is_gsym = true; 2902 sym[sym_idx].SetExternal(true); 2903 2904 if (symbol_name && symbol_name[0] == '_' && 2905 symbol_name[1] == 'O') { 2906 llvm::StringRef symbol_name_ref(symbol_name); 2907 if (symbol_name_ref.startswith( 2908 g_objc_v2_prefix_class)) { 2909 symbol_name_non_abi_mangled = symbol_name + 1; 2910 symbol_name = 2911 symbol_name + g_objc_v2_prefix_class.size(); 2912 type = eSymbolTypeObjCClass; 2913 demangled_is_synthesized = true; 2914 2915 } else if (symbol_name_ref.startswith( 2916 g_objc_v2_prefix_metaclass)) { 2917 symbol_name_non_abi_mangled = symbol_name + 1; 2918 symbol_name = symbol_name + 2919 g_objc_v2_prefix_metaclass.size(); 2920 type = eSymbolTypeObjCMetaClass; 2921 demangled_is_synthesized = true; 2922 } else if (symbol_name_ref.startswith( 2923 g_objc_v2_prefix_ivar)) { 2924 symbol_name_non_abi_mangled = symbol_name + 1; 2925 symbol_name = 2926 symbol_name + g_objc_v2_prefix_ivar.size(); 2927 type = eSymbolTypeObjCIVar; 2928 demangled_is_synthesized = true; 2929 } 2930 } else { 2931 if (nlist.n_value != 0) 2932 symbol_section = section_info.GetSection( 2933 nlist.n_sect, nlist.n_value); 2934 type = eSymbolTypeData; 2935 } 2936 break; 2937 2938 case N_FNAME: 2939 // procedure name (f77 kludge): name,,NO_SECT,0,0 2940 type = eSymbolTypeCompiler; 2941 break; 2942 2943 case N_FUN: 2944 // procedure: name,,n_sect,linenumber,address 2945 if (symbol_name) { 2946 type = eSymbolTypeCode; 2947 symbol_section = section_info.GetSection( 2948 nlist.n_sect, nlist.n_value); 2949 2950 N_FUN_addr_to_sym_idx.insert( 2951 std::make_pair(nlist.n_value, sym_idx)); 2952 // We use the current number of symbols in the 2953 // symbol table in lieu of 2954 // using nlist_idx in case we ever start trimming 2955 // entries out 2956 N_FUN_indexes.push_back(sym_idx); 2957 } else { 2958 type = eSymbolTypeCompiler; 2959 2960 if (!N_FUN_indexes.empty()) { 2961 // Copy the size of the function into the original 2962 // STAB entry so we don't have 2963 // to hunt for it later 2964 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2965 ->SetByteSize(nlist.n_value); 2966 N_FUN_indexes.pop_back(); 2967 // We don't really need the end function STAB as 2968 // it contains the size which 2969 // we already placed with the original symbol, so 2970 // don't add it if we want a 2971 // minimal symbol table 2972 add_nlist = false; 2973 } 2974 } 2975 break; 2976 2977 case N_STSYM: 2978 // static symbol: name,,n_sect,type,address 2979 N_STSYM_addr_to_sym_idx.insert( 2980 std::make_pair(nlist.n_value, sym_idx)); 2981 symbol_section = section_info.GetSection( 2982 nlist.n_sect, nlist.n_value); 2983 if (symbol_name && symbol_name[0]) { 2984 type = ObjectFile::GetSymbolTypeFromName( 2985 symbol_name + 1, eSymbolTypeData); 2986 } 2987 break; 2988 2989 case N_LCSYM: 2990 // .lcomm symbol: name,,n_sect,type,address 2991 symbol_section = section_info.GetSection( 2992 nlist.n_sect, nlist.n_value); 2993 type = eSymbolTypeCommonBlock; 2994 break; 2995 2996 case N_BNSYM: 2997 // We use the current number of symbols in the symbol 2998 // table in lieu of 2999 // using nlist_idx in case we ever start trimming 3000 // entries out 3001 // Skip these if we want minimal symbol tables 3002 add_nlist = false; 3003 break; 3004 3005 case N_ENSYM: 3006 // Set the size of the N_BNSYM to the terminating 3007 // index of this N_ENSYM 3008 // so that we can always skip the entire symbol if we 3009 // need to navigate 3010 // more quickly at the source level when parsing STABS 3011 // Skip these if we want minimal symbol tables 3012 add_nlist = false; 3013 break; 3014 3015 case N_OPT: 3016 // emitted with gcc2_compiled and in gcc source 3017 type = eSymbolTypeCompiler; 3018 break; 3019 3020 case N_RSYM: 3021 // register sym: name,,NO_SECT,type,register 3022 type = eSymbolTypeVariable; 3023 break; 3024 3025 case N_SLINE: 3026 // src line: 0,,n_sect,linenumber,address 3027 symbol_section = section_info.GetSection( 3028 nlist.n_sect, nlist.n_value); 3029 type = eSymbolTypeLineEntry; 3030 break; 3031 3032 case N_SSYM: 3033 // structure elt: name,,NO_SECT,type,struct_offset 3034 type = eSymbolTypeVariableType; 3035 break; 3036 3037 case N_SO: 3038 // source file name 3039 type = eSymbolTypeSourceFile; 3040 if (symbol_name == NULL) { 3041 add_nlist = false; 3042 if (N_SO_index != UINT32_MAX) { 3043 // Set the size of the N_SO to the terminating 3044 // index of this N_SO 3045 // so that we can always skip the entire N_SO if 3046 // we need to navigate 3047 // more quickly at the source level when parsing 3048 // STABS 3049 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3050 symbol_ptr->SetByteSize(sym_idx); 3051 symbol_ptr->SetSizeIsSibling(true); 3052 } 3053 N_NSYM_indexes.clear(); 3054 N_INCL_indexes.clear(); 3055 N_BRAC_indexes.clear(); 3056 N_COMM_indexes.clear(); 3057 N_FUN_indexes.clear(); 3058 N_SO_index = UINT32_MAX; 3059 } else { 3060 // We use the current number of symbols in the 3061 // symbol table in lieu of 3062 // using nlist_idx in case we ever start trimming 3063 // entries out 3064 const bool N_SO_has_full_path = 3065 symbol_name[0] == '/'; 3066 if (N_SO_has_full_path) { 3067 if ((N_SO_index == sym_idx - 1) && 3068 ((sym_idx - 1) < num_syms)) { 3069 // We have two consecutive N_SO entries where 3070 // the first contains a directory 3071 // and the second contains a full path. 3072 sym[sym_idx - 1].GetMangled().SetValue( 3073 ConstString(symbol_name), false); 3074 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3075 add_nlist = false; 3076 } else { 3077 // This is the first entry in a N_SO that 3078 // contains a directory or 3079 // a full path to the source file 3080 N_SO_index = sym_idx; 3081 } 3082 } else if ((N_SO_index == sym_idx - 1) && 3083 ((sym_idx - 1) < num_syms)) { 3084 // This is usually the second N_SO entry that 3085 // contains just the filename, 3086 // so here we combine it with the first one if we 3087 // are minimizing the symbol table 3088 const char *so_path = 3089 sym[sym_idx - 1] 3090 .GetMangled() 3091 .GetDemangledName( 3092 lldb::eLanguageTypeUnknown) 3093 .AsCString(); 3094 if (so_path && so_path[0]) { 3095 std::string full_so_path(so_path); 3096 const size_t double_slash_pos = 3097 full_so_path.find("//"); 3098 if (double_slash_pos != std::string::npos) { 3099 // The linker has been generating bad N_SO 3100 // entries with doubled up paths 3101 // in the format "%s%s" where the first string 3102 // in the DW_AT_comp_dir, 3103 // and the second is the directory for the 3104 // source file so you end up with 3105 // a path that looks like "/tmp/src//tmp/src/" 3106 FileSpec so_dir(so_path, false); 3107 if (!so_dir.Exists()) { 3108 so_dir.SetFile( 3109 &full_so_path[double_slash_pos + 1], 3110 false); 3111 if (so_dir.Exists()) { 3112 // Trim off the incorrect path 3113 full_so_path.erase(0, 3114 double_slash_pos + 1); 3115 } 3116 } 3117 } 3118 if (*full_so_path.rbegin() != '/') 3119 full_so_path += '/'; 3120 full_so_path += symbol_name; 3121 sym[sym_idx - 1].GetMangled().SetValue( 3122 ConstString(full_so_path.c_str()), false); 3123 add_nlist = false; 3124 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3125 } 3126 } else { 3127 // This could be a relative path to a N_SO 3128 N_SO_index = sym_idx; 3129 } 3130 } 3131 break; 3132 3133 case N_OSO: 3134 // object file name: name,,0,0,st_mtime 3135 type = eSymbolTypeObjectFile; 3136 break; 3137 3138 case N_LSYM: 3139 // local sym: name,,NO_SECT,type,offset 3140 type = eSymbolTypeLocal; 3141 break; 3142 3143 //---------------------------------------------------------------------- 3144 // INCL scopes 3145 //---------------------------------------------------------------------- 3146 case N_BINCL: 3147 // include file beginning: name,,NO_SECT,0,sum 3148 // We use the current number of symbols in the symbol 3149 // table in lieu of 3150 // using nlist_idx in case we ever start trimming 3151 // entries out 3152 N_INCL_indexes.push_back(sym_idx); 3153 type = eSymbolTypeScopeBegin; 3154 break; 3155 3156 case N_EINCL: 3157 // include file end: name,,NO_SECT,0,0 3158 // Set the size of the N_BINCL to the terminating 3159 // index of this N_EINCL 3160 // so that we can always skip the entire symbol if we 3161 // need to navigate 3162 // more quickly at the source level when parsing STABS 3163 if (!N_INCL_indexes.empty()) { 3164 symbol_ptr = 3165 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3166 symbol_ptr->SetByteSize(sym_idx + 1); 3167 symbol_ptr->SetSizeIsSibling(true); 3168 N_INCL_indexes.pop_back(); 3169 } 3170 type = eSymbolTypeScopeEnd; 3171 break; 3172 3173 case N_SOL: 3174 // #included file name: name,,n_sect,0,address 3175 type = eSymbolTypeHeaderFile; 3176 3177 // We currently don't use the header files on darwin 3178 add_nlist = false; 3179 break; 3180 3181 case N_PARAMS: 3182 // compiler parameters: name,,NO_SECT,0,0 3183 type = eSymbolTypeCompiler; 3184 break; 3185 3186 case N_VERSION: 3187 // compiler version: name,,NO_SECT,0,0 3188 type = eSymbolTypeCompiler; 3189 break; 3190 3191 case N_OLEVEL: 3192 // compiler -O level: name,,NO_SECT,0,0 3193 type = eSymbolTypeCompiler; 3194 break; 3195 3196 case N_PSYM: 3197 // parameter: name,,NO_SECT,type,offset 3198 type = eSymbolTypeVariable; 3199 break; 3200 3201 case N_ENTRY: 3202 // alternate entry: name,,n_sect,linenumber,address 3203 symbol_section = section_info.GetSection( 3204 nlist.n_sect, nlist.n_value); 3205 type = eSymbolTypeLineEntry; 3206 break; 3207 3208 //---------------------------------------------------------------------- 3209 // Left and Right Braces 3210 //---------------------------------------------------------------------- 3211 case N_LBRAC: 3212 // left bracket: 0,,NO_SECT,nesting level,address 3213 // We use the current number of symbols in the symbol 3214 // table in lieu of 3215 // using nlist_idx in case we ever start trimming 3216 // entries out 3217 symbol_section = section_info.GetSection( 3218 nlist.n_sect, nlist.n_value); 3219 N_BRAC_indexes.push_back(sym_idx); 3220 type = eSymbolTypeScopeBegin; 3221 break; 3222 3223 case N_RBRAC: 3224 // right bracket: 0,,NO_SECT,nesting level,address 3225 // Set the size of the N_LBRAC to the terminating 3226 // index of this N_RBRAC 3227 // so that we can always skip the entire symbol if we 3228 // need to navigate 3229 // more quickly at the source level when parsing STABS 3230 symbol_section = section_info.GetSection( 3231 nlist.n_sect, nlist.n_value); 3232 if (!N_BRAC_indexes.empty()) { 3233 symbol_ptr = 3234 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3235 symbol_ptr->SetByteSize(sym_idx + 1); 3236 symbol_ptr->SetSizeIsSibling(true); 3237 N_BRAC_indexes.pop_back(); 3238 } 3239 type = eSymbolTypeScopeEnd; 3240 break; 3241 3242 case N_EXCL: 3243 // deleted include file: name,,NO_SECT,0,sum 3244 type = eSymbolTypeHeaderFile; 3245 break; 3246 3247 //---------------------------------------------------------------------- 3248 // COMM scopes 3249 //---------------------------------------------------------------------- 3250 case N_BCOMM: 3251 // begin common: name,,NO_SECT,0,0 3252 // We use the current number of symbols in the symbol 3253 // table in lieu of 3254 // using nlist_idx in case we ever start trimming 3255 // entries out 3256 type = eSymbolTypeScopeBegin; 3257 N_COMM_indexes.push_back(sym_idx); 3258 break; 3259 3260 case N_ECOML: 3261 // end common (local name): 0,,n_sect,0,address 3262 symbol_section = section_info.GetSection( 3263 nlist.n_sect, nlist.n_value); 3264 // Fall through 3265 3266 case N_ECOMM: 3267 // end common: name,,n_sect,0,0 3268 // Set the size of the N_BCOMM to the terminating 3269 // index of this N_ECOMM/N_ECOML 3270 // so that we can always skip the entire symbol if we 3271 // need to navigate 3272 // more quickly at the source level when parsing STABS 3273 if (!N_COMM_indexes.empty()) { 3274 symbol_ptr = 3275 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3276 symbol_ptr->SetByteSize(sym_idx + 1); 3277 symbol_ptr->SetSizeIsSibling(true); 3278 N_COMM_indexes.pop_back(); 3279 } 3280 type = eSymbolTypeScopeEnd; 3281 break; 3282 3283 case N_LENG: 3284 // second stab entry with length information 3285 type = eSymbolTypeAdditional; 3286 break; 3287 3288 default: 3289 break; 3290 } 3291 } else { 3292 // uint8_t n_pext = N_PEXT & nlist.n_type; 3293 uint8_t n_type = N_TYPE & nlist.n_type; 3294 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3295 3296 switch (n_type) { 3297 case N_INDR: { 3298 const char *reexport_name_cstr = 3299 strtab_data.PeekCStr(nlist.n_value); 3300 if (reexport_name_cstr && reexport_name_cstr[0]) { 3301 type = eSymbolTypeReExported; 3302 ConstString reexport_name( 3303 reexport_name_cstr + 3304 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3305 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3306 set_value = false; 3307 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3308 indirect_symbol_names.insert( 3309 ConstString(symbol_name + 3310 ((symbol_name[0] == '_') ? 1 : 0))); 3311 } else 3312 type = eSymbolTypeUndefined; 3313 } break; 3314 3315 case N_UNDF: 3316 if (symbol_name && symbol_name[0]) { 3317 ConstString undefined_name( 3318 symbol_name + 3319 ((symbol_name[0] == '_') ? 1 : 0)); 3320 undefined_name_to_desc[undefined_name] = 3321 nlist.n_desc; 3322 } 3323 // Fall through 3324 case N_PBUD: 3325 type = eSymbolTypeUndefined; 3326 break; 3327 3328 case N_ABS: 3329 type = eSymbolTypeAbsolute; 3330 break; 3331 3332 case N_SECT: { 3333 symbol_section = section_info.GetSection( 3334 nlist.n_sect, nlist.n_value); 3335 3336 if (symbol_section == NULL) { 3337 // TODO: warn about this? 3338 add_nlist = false; 3339 break; 3340 } 3341 3342 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3343 type = eSymbolTypeException; 3344 } else { 3345 uint32_t section_type = 3346 symbol_section->Get() & SECTION_TYPE; 3347 3348 switch (section_type) { 3349 case S_CSTRING_LITERALS: 3350 type = eSymbolTypeData; 3351 break; // section with only literal C strings 3352 case S_4BYTE_LITERALS: 3353 type = eSymbolTypeData; 3354 break; // section with only 4 byte literals 3355 case S_8BYTE_LITERALS: 3356 type = eSymbolTypeData; 3357 break; // section with only 8 byte literals 3358 case S_LITERAL_POINTERS: 3359 type = eSymbolTypeTrampoline; 3360 break; // section with only pointers to literals 3361 case S_NON_LAZY_SYMBOL_POINTERS: 3362 type = eSymbolTypeTrampoline; 3363 break; // section with only non-lazy symbol 3364 // pointers 3365 case S_LAZY_SYMBOL_POINTERS: 3366 type = eSymbolTypeTrampoline; 3367 break; // section with only lazy symbol pointers 3368 case S_SYMBOL_STUBS: 3369 type = eSymbolTypeTrampoline; 3370 break; // section with only symbol stubs, byte 3371 // size of stub in the reserved2 field 3372 case S_MOD_INIT_FUNC_POINTERS: 3373 type = eSymbolTypeCode; 3374 break; // section with only function pointers for 3375 // initialization 3376 case S_MOD_TERM_FUNC_POINTERS: 3377 type = eSymbolTypeCode; 3378 break; // section with only function pointers for 3379 // termination 3380 case S_INTERPOSING: 3381 type = eSymbolTypeTrampoline; 3382 break; // section with only pairs of function 3383 // pointers for interposing 3384 case S_16BYTE_LITERALS: 3385 type = eSymbolTypeData; 3386 break; // section with only 16 byte literals 3387 case S_DTRACE_DOF: 3388 type = eSymbolTypeInstrumentation; 3389 break; 3390 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3391 type = eSymbolTypeTrampoline; 3392 break; 3393 default: 3394 switch (symbol_section->GetType()) { 3395 case lldb::eSectionTypeCode: 3396 type = eSymbolTypeCode; 3397 break; 3398 case eSectionTypeData: 3399 case eSectionTypeDataCString: // Inlined C string 3400 // data 3401 case eSectionTypeDataCStringPointers: // Pointers 3402 // to C 3403 // string 3404 // data 3405 case eSectionTypeDataSymbolAddress: // Address of 3406 // a symbol in 3407 // the symbol 3408 // table 3409 case eSectionTypeData4: 3410 case eSectionTypeData8: 3411 case eSectionTypeData16: 3412 type = eSymbolTypeData; 3413 break; 3414 default: 3415 break; 3416 } 3417 break; 3418 } 3419 3420 if (type == eSymbolTypeInvalid) { 3421 const char *symbol_sect_name = 3422 symbol_section->GetName().AsCString(); 3423 if (symbol_section->IsDescendant( 3424 text_section_sp.get())) { 3425 if (symbol_section->IsClear( 3426 S_ATTR_PURE_INSTRUCTIONS | 3427 S_ATTR_SELF_MODIFYING_CODE | 3428 S_ATTR_SOME_INSTRUCTIONS)) 3429 type = eSymbolTypeData; 3430 else 3431 type = eSymbolTypeCode; 3432 } else if (symbol_section->IsDescendant( 3433 data_section_sp.get()) || 3434 symbol_section->IsDescendant( 3435 data_dirty_section_sp.get()) || 3436 symbol_section->IsDescendant( 3437 data_const_section_sp.get())) { 3438 if (symbol_sect_name && 3439 ::strstr(symbol_sect_name, "__objc") == 3440 symbol_sect_name) { 3441 type = eSymbolTypeRuntime; 3442 3443 if (symbol_name) { 3444 llvm::StringRef symbol_name_ref( 3445 symbol_name); 3446 if (symbol_name_ref.startswith("_OBJC_")) { 3447 static const llvm::StringRef 3448 g_objc_v2_prefix_class( 3449 "_OBJC_CLASS_$_"); 3450 static const llvm::StringRef 3451 g_objc_v2_prefix_metaclass( 3452 "_OBJC_METACLASS_$_"); 3453 static const llvm::StringRef 3454 g_objc_v2_prefix_ivar( 3455 "_OBJC_IVAR_$_"); 3456 if (symbol_name_ref.startswith( 3457 g_objc_v2_prefix_class)) { 3458 symbol_name_non_abi_mangled = 3459 symbol_name + 1; 3460 symbol_name = 3461 symbol_name + 3462 g_objc_v2_prefix_class.size(); 3463 type = eSymbolTypeObjCClass; 3464 demangled_is_synthesized = true; 3465 } else if ( 3466 symbol_name_ref.startswith( 3467 g_objc_v2_prefix_metaclass)) { 3468 symbol_name_non_abi_mangled = 3469 symbol_name + 1; 3470 symbol_name = 3471 symbol_name + 3472 g_objc_v2_prefix_metaclass.size(); 3473 type = eSymbolTypeObjCMetaClass; 3474 demangled_is_synthesized = true; 3475 } else if (symbol_name_ref.startswith( 3476 g_objc_v2_prefix_ivar)) { 3477 symbol_name_non_abi_mangled = 3478 symbol_name + 1; 3479 symbol_name = 3480 symbol_name + 3481 g_objc_v2_prefix_ivar.size(); 3482 type = eSymbolTypeObjCIVar; 3483 demangled_is_synthesized = true; 3484 } 3485 } 3486 } 3487 } else if (symbol_sect_name && 3488 ::strstr(symbol_sect_name, 3489 "__gcc_except_tab") == 3490 symbol_sect_name) { 3491 type = eSymbolTypeException; 3492 } else { 3493 type = eSymbolTypeData; 3494 } 3495 } else if (symbol_sect_name && 3496 ::strstr(symbol_sect_name, 3497 "__IMPORT") == 3498 symbol_sect_name) { 3499 type = eSymbolTypeTrampoline; 3500 } else if (symbol_section->IsDescendant( 3501 objc_section_sp.get())) { 3502 type = eSymbolTypeRuntime; 3503 if (symbol_name && symbol_name[0] == '.') { 3504 llvm::StringRef symbol_name_ref(symbol_name); 3505 static const llvm::StringRef 3506 g_objc_v1_prefix_class( 3507 ".objc_class_name_"); 3508 if (symbol_name_ref.startswith( 3509 g_objc_v1_prefix_class)) { 3510 symbol_name_non_abi_mangled = symbol_name; 3511 symbol_name = symbol_name + 3512 g_objc_v1_prefix_class.size(); 3513 type = eSymbolTypeObjCClass; 3514 demangled_is_synthesized = true; 3515 } 3516 } 3517 } 3518 } 3519 } 3520 } break; 3521 } 3522 } 3523 3524 if (add_nlist) { 3525 uint64_t symbol_value = nlist.n_value; 3526 if (symbol_name_non_abi_mangled) { 3527 sym[sym_idx].GetMangled().SetMangledName( 3528 ConstString(symbol_name_non_abi_mangled)); 3529 sym[sym_idx].GetMangled().SetDemangledName( 3530 ConstString(symbol_name)); 3531 } else { 3532 bool symbol_name_is_mangled = false; 3533 3534 if (symbol_name && symbol_name[0] == '_') { 3535 symbol_name_is_mangled = symbol_name[1] == '_'; 3536 symbol_name++; // Skip the leading underscore 3537 } 3538 3539 if (symbol_name) { 3540 ConstString const_symbol_name(symbol_name); 3541 sym[sym_idx].GetMangled().SetValue( 3542 const_symbol_name, symbol_name_is_mangled); 3543 if (is_gsym && is_debug) { 3544 const char *gsym_name = 3545 sym[sym_idx] 3546 .GetMangled() 3547 .GetName(lldb::eLanguageTypeUnknown, 3548 Mangled::ePreferMangled) 3549 .GetCString(); 3550 if (gsym_name) 3551 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3552 } 3553 } 3554 } 3555 if (symbol_section) { 3556 const addr_t section_file_addr = 3557 symbol_section->GetFileAddress(); 3558 if (symbol_byte_size == 0 && 3559 function_starts_count > 0) { 3560 addr_t symbol_lookup_file_addr = nlist.n_value; 3561 // Do an exact address match for non-ARM addresses, 3562 // else get the closest since 3563 // the symbol might be a thumb symbol which has an 3564 // address with bit zero set 3565 FunctionStarts::Entry *func_start_entry = 3566 function_starts.FindEntry( 3567 symbol_lookup_file_addr, !is_arm); 3568 if (is_arm && func_start_entry) { 3569 // Verify that the function start address is the 3570 // symbol address (ARM) 3571 // or the symbol address + 1 (thumb) 3572 if (func_start_entry->addr != 3573 symbol_lookup_file_addr && 3574 func_start_entry->addr != 3575 (symbol_lookup_file_addr + 1)) { 3576 // Not the right entry, NULL it out... 3577 func_start_entry = NULL; 3578 } 3579 } 3580 if (func_start_entry) { 3581 func_start_entry->data = true; 3582 3583 addr_t symbol_file_addr = func_start_entry->addr; 3584 uint32_t symbol_flags = 0; 3585 if (is_arm) { 3586 if (symbol_file_addr & 1) 3587 symbol_flags = 3588 MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3589 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3590 } 3591 3592 const FunctionStarts::Entry 3593 *next_func_start_entry = 3594 function_starts.FindNextEntry( 3595 func_start_entry); 3596 const addr_t section_end_file_addr = 3597 section_file_addr + 3598 symbol_section->GetByteSize(); 3599 if (next_func_start_entry) { 3600 addr_t next_symbol_file_addr = 3601 next_func_start_entry->addr; 3602 // Be sure the clear the Thumb address bit when 3603 // we calculate the size 3604 // from the current and next address 3605 if (is_arm) 3606 next_symbol_file_addr &= 3607 THUMB_ADDRESS_BIT_MASK; 3608 symbol_byte_size = std::min<lldb::addr_t>( 3609 next_symbol_file_addr - symbol_file_addr, 3610 section_end_file_addr - symbol_file_addr); 3611 } else { 3612 symbol_byte_size = 3613 section_end_file_addr - symbol_file_addr; 3614 } 3615 } 3616 } 3617 symbol_value -= section_file_addr; 3618 } 3619 3620 if (is_debug == false) { 3621 if (type == eSymbolTypeCode) { 3622 // See if we can find a N_FUN entry for any code 3623 // symbols. 3624 // If we do find a match, and the name matches, then 3625 // we 3626 // can merge the two into just the function symbol 3627 // to avoid 3628 // duplicate entries in the symbol table 3629 std::pair<ValueToSymbolIndexMap::const_iterator, 3630 ValueToSymbolIndexMap::const_iterator> 3631 range; 3632 range = N_FUN_addr_to_sym_idx.equal_range( 3633 nlist.n_value); 3634 if (range.first != range.second) { 3635 bool found_it = false; 3636 for (ValueToSymbolIndexMap::const_iterator pos = 3637 range.first; 3638 pos != range.second; ++pos) { 3639 if (sym[sym_idx].GetMangled().GetName( 3640 lldb::eLanguageTypeUnknown, 3641 Mangled::ePreferMangled) == 3642 sym[pos->second].GetMangled().GetName( 3643 lldb::eLanguageTypeUnknown, 3644 Mangled::ePreferMangled)) { 3645 m_nlist_idx_to_sym_idx[nlist_idx] = 3646 pos->second; 3647 // We just need the flags from the linker 3648 // symbol, so put these flags 3649 // into the N_FUN flags to avoid duplicate 3650 // symbols in the symbol table 3651 sym[pos->second].SetExternal( 3652 sym[sym_idx].IsExternal()); 3653 sym[pos->second].SetFlags(nlist.n_type << 16 | 3654 nlist.n_desc); 3655 if (resolver_addresses.find(nlist.n_value) != 3656 resolver_addresses.end()) 3657 sym[pos->second].SetType( 3658 eSymbolTypeResolver); 3659 sym[sym_idx].Clear(); 3660 found_it = true; 3661 break; 3662 } 3663 } 3664 if (found_it) 3665 continue; 3666 } else { 3667 if (resolver_addresses.find(nlist.n_value) != 3668 resolver_addresses.end()) 3669 type = eSymbolTypeResolver; 3670 } 3671 } else if (type == eSymbolTypeData || 3672 type == eSymbolTypeObjCClass || 3673 type == eSymbolTypeObjCMetaClass || 3674 type == eSymbolTypeObjCIVar) { 3675 // See if we can find a N_STSYM entry for any data 3676 // symbols. 3677 // If we do find a match, and the name matches, then 3678 // we 3679 // can merge the two into just the Static symbol to 3680 // avoid 3681 // duplicate entries in the symbol table 3682 std::pair<ValueToSymbolIndexMap::const_iterator, 3683 ValueToSymbolIndexMap::const_iterator> 3684 range; 3685 range = N_STSYM_addr_to_sym_idx.equal_range( 3686 nlist.n_value); 3687 if (range.first != range.second) { 3688 bool found_it = false; 3689 for (ValueToSymbolIndexMap::const_iterator pos = 3690 range.first; 3691 pos != range.second; ++pos) { 3692 if (sym[sym_idx].GetMangled().GetName( 3693 lldb::eLanguageTypeUnknown, 3694 Mangled::ePreferMangled) == 3695 sym[pos->second].GetMangled().GetName( 3696 lldb::eLanguageTypeUnknown, 3697 Mangled::ePreferMangled)) { 3698 m_nlist_idx_to_sym_idx[nlist_idx] = 3699 pos->second; 3700 // We just need the flags from the linker 3701 // symbol, so put these flags 3702 // into the N_STSYM flags to avoid duplicate 3703 // symbols in the symbol table 3704 sym[pos->second].SetExternal( 3705 sym[sym_idx].IsExternal()); 3706 sym[pos->second].SetFlags(nlist.n_type << 16 | 3707 nlist.n_desc); 3708 sym[sym_idx].Clear(); 3709 found_it = true; 3710 break; 3711 } 3712 } 3713 if (found_it) 3714 continue; 3715 } else { 3716 const char *gsym_name = 3717 sym[sym_idx] 3718 .GetMangled() 3719 .GetName(lldb::eLanguageTypeUnknown, 3720 Mangled::ePreferMangled) 3721 .GetCString(); 3722 if (gsym_name) { 3723 // Combine N_GSYM stab entries with the non stab 3724 // symbol 3725 ConstNameToSymbolIndexMap::const_iterator pos = 3726 N_GSYM_name_to_sym_idx.find(gsym_name); 3727 if (pos != N_GSYM_name_to_sym_idx.end()) { 3728 const uint32_t GSYM_sym_idx = pos->second; 3729 m_nlist_idx_to_sym_idx[nlist_idx] = 3730 GSYM_sym_idx; 3731 // Copy the address, because often the N_GSYM 3732 // address has an invalid address of zero 3733 // when the global is a common symbol 3734 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3735 symbol_section); 3736 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3737 symbol_value); 3738 // We just need the flags from the linker 3739 // symbol, so put these flags 3740 // into the N_GSYM flags to avoid duplicate 3741 // symbols in the symbol table 3742 sym[GSYM_sym_idx].SetFlags( 3743 nlist.n_type << 16 | nlist.n_desc); 3744 sym[sym_idx].Clear(); 3745 continue; 3746 } 3747 } 3748 } 3749 } 3750 } 3751 3752 sym[sym_idx].SetID(nlist_idx); 3753 sym[sym_idx].SetType(type); 3754 if (set_value) { 3755 sym[sym_idx].GetAddressRef().SetSection( 3756 symbol_section); 3757 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3758 } 3759 sym[sym_idx].SetFlags(nlist.n_type << 16 | 3760 nlist.n_desc); 3761 3762 if (symbol_byte_size > 0) 3763 sym[sym_idx].SetByteSize(symbol_byte_size); 3764 3765 if (demangled_is_synthesized) 3766 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3767 ++sym_idx; 3768 } else { 3769 sym[sym_idx].Clear(); 3770 } 3771 } 3772 ///////////////////////////// 3773 } 3774 break; // No more entries to consider 3775 } 3776 } 3777 3778 for (const auto &pos : reexport_shlib_needs_fixup) { 3779 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3780 if (undef_pos != undefined_name_to_desc.end()) { 3781 const uint8_t dylib_ordinal = 3782 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3783 if (dylib_ordinal > 0 && 3784 dylib_ordinal < dylib_files.GetSize()) 3785 sym[pos.first].SetReExportedSymbolSharedLibrary( 3786 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3787 } 3788 } 3789 } 3790 } 3791 } 3792 } 3793 } 3794 3795 // Must reset this in case it was mutated above! 3796 nlist_data_offset = 0; 3797 #endif 3798 3799 if (nlist_data.GetByteSize() > 0) { 3800 3801 // If the sym array was not created while parsing the DSC unmapped 3802 // symbols, create it now. 3803 if (sym == NULL) { 3804 sym = symtab->Resize(symtab_load_command.nsyms + 3805 m_dysymtab.nindirectsyms); 3806 num_syms = symtab->GetNumSymbols(); 3807 } 3808 3809 if (unmapped_local_symbols_found) { 3810 assert(m_dysymtab.ilocalsym == 0); 3811 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3812 nlist_idx = m_dysymtab.nlocalsym; 3813 } else { 3814 nlist_idx = 0; 3815 } 3816 3817 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3818 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3819 UndefinedNameToDescMap undefined_name_to_desc; 3820 SymbolIndexToName reexport_shlib_needs_fixup; 3821 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 3822 struct nlist_64 nlist; 3823 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, 3824 nlist_byte_size)) 3825 break; 3826 3827 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3828 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 3829 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 3830 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 3831 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 3832 3833 SymbolType type = eSymbolTypeInvalid; 3834 const char *symbol_name = NULL; 3835 3836 if (have_strtab_data) { 3837 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3838 3839 if (symbol_name == NULL) { 3840 // No symbol should be NULL, even the symbols with no 3841 // string values should have an offset zero which points 3842 // to an empty C-string 3843 Host::SystemLog(Host::eSystemLogError, 3844 "error: symbol[%u] has invalid string table offset " 3845 "0x%x in %s, ignoring symbol\n", 3846 nlist_idx, nlist.n_strx, 3847 module_sp->GetFileSpec().GetPath().c_str()); 3848 continue; 3849 } 3850 if (symbol_name[0] == '\0') 3851 symbol_name = NULL; 3852 } else { 3853 const addr_t str_addr = strtab_addr + nlist.n_strx; 3854 Status str_error; 3855 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3856 str_error)) 3857 symbol_name = memory_symbol_name.c_str(); 3858 } 3859 const char *symbol_name_non_abi_mangled = NULL; 3860 3861 SectionSP symbol_section; 3862 lldb::addr_t symbol_byte_size = 0; 3863 bool add_nlist = true; 3864 bool is_gsym = false; 3865 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3866 bool demangled_is_synthesized = false; 3867 bool set_value = true; 3868 assert(sym_idx < num_syms); 3869 3870 sym[sym_idx].SetDebug(is_debug); 3871 3872 if (is_debug) { 3873 switch (nlist.n_type) { 3874 case N_GSYM: 3875 // global symbol: name,,NO_SECT,type,0 3876 // Sometimes the N_GSYM value contains the address. 3877 3878 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3879 // the ObjC data. They 3880 // have the same address, but we want to ensure that we always find 3881 // only the real symbol, 3882 // 'cause we don't currently correctly attribute the GSYM one to the 3883 // ObjCClass/Ivar/MetaClass 3884 // symbol type. This is a temporary hack to make sure the 3885 // ObjectiveC symbols get treated 3886 // correctly. To do this right, we should coalesce all the GSYM & 3887 // global symbols that have the 3888 // same address. 3889 is_gsym = true; 3890 sym[sym_idx].SetExternal(true); 3891 3892 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3893 llvm::StringRef symbol_name_ref(symbol_name); 3894 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3895 symbol_name_non_abi_mangled = symbol_name + 1; 3896 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3897 type = eSymbolTypeObjCClass; 3898 demangled_is_synthesized = true; 3899 3900 } else if (symbol_name_ref.startswith( 3901 g_objc_v2_prefix_metaclass)) { 3902 symbol_name_non_abi_mangled = symbol_name + 1; 3903 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3904 type = eSymbolTypeObjCMetaClass; 3905 demangled_is_synthesized = true; 3906 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3907 symbol_name_non_abi_mangled = symbol_name + 1; 3908 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3909 type = eSymbolTypeObjCIVar; 3910 demangled_is_synthesized = true; 3911 } 3912 } else { 3913 if (nlist.n_value != 0) 3914 symbol_section = 3915 section_info.GetSection(nlist.n_sect, nlist.n_value); 3916 type = eSymbolTypeData; 3917 } 3918 break; 3919 3920 case N_FNAME: 3921 // procedure name (f77 kludge): name,,NO_SECT,0,0 3922 type = eSymbolTypeCompiler; 3923 break; 3924 3925 case N_FUN: 3926 // procedure: name,,n_sect,linenumber,address 3927 if (symbol_name) { 3928 type = eSymbolTypeCode; 3929 symbol_section = 3930 section_info.GetSection(nlist.n_sect, nlist.n_value); 3931 3932 N_FUN_addr_to_sym_idx.insert( 3933 std::make_pair(nlist.n_value, sym_idx)); 3934 // We use the current number of symbols in the symbol table in 3935 // lieu of 3936 // using nlist_idx in case we ever start trimming entries out 3937 N_FUN_indexes.push_back(sym_idx); 3938 } else { 3939 type = eSymbolTypeCompiler; 3940 3941 if (!N_FUN_indexes.empty()) { 3942 // Copy the size of the function into the original STAB entry so 3943 // we don't have 3944 // to hunt for it later 3945 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3946 ->SetByteSize(nlist.n_value); 3947 N_FUN_indexes.pop_back(); 3948 // We don't really need the end function STAB as it contains the 3949 // size which 3950 // we already placed with the original symbol, so don't add it 3951 // if we want a 3952 // minimal symbol table 3953 add_nlist = false; 3954 } 3955 } 3956 break; 3957 3958 case N_STSYM: 3959 // static symbol: name,,n_sect,type,address 3960 N_STSYM_addr_to_sym_idx.insert( 3961 std::make_pair(nlist.n_value, sym_idx)); 3962 symbol_section = 3963 section_info.GetSection(nlist.n_sect, nlist.n_value); 3964 if (symbol_name && symbol_name[0]) { 3965 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3966 eSymbolTypeData); 3967 } 3968 break; 3969 3970 case N_LCSYM: 3971 // .lcomm symbol: name,,n_sect,type,address 3972 symbol_section = 3973 section_info.GetSection(nlist.n_sect, nlist.n_value); 3974 type = eSymbolTypeCommonBlock; 3975 break; 3976 3977 case N_BNSYM: 3978 // We use the current number of symbols in the symbol table in lieu 3979 // of 3980 // using nlist_idx in case we ever start trimming entries out 3981 // Skip these if we want minimal symbol tables 3982 add_nlist = false; 3983 break; 3984 3985 case N_ENSYM: 3986 // Set the size of the N_BNSYM to the terminating index of this 3987 // N_ENSYM 3988 // so that we can always skip the entire symbol if we need to 3989 // navigate 3990 // more quickly at the source level when parsing STABS 3991 // Skip these if we want minimal symbol tables 3992 add_nlist = false; 3993 break; 3994 3995 case N_OPT: 3996 // emitted with gcc2_compiled and in gcc source 3997 type = eSymbolTypeCompiler; 3998 break; 3999 4000 case N_RSYM: 4001 // register sym: name,,NO_SECT,type,register 4002 type = eSymbolTypeVariable; 4003 break; 4004 4005 case N_SLINE: 4006 // src line: 0,,n_sect,linenumber,address 4007 symbol_section = 4008 section_info.GetSection(nlist.n_sect, nlist.n_value); 4009 type = eSymbolTypeLineEntry; 4010 break; 4011 4012 case N_SSYM: 4013 // structure elt: name,,NO_SECT,type,struct_offset 4014 type = eSymbolTypeVariableType; 4015 break; 4016 4017 case N_SO: 4018 // source file name 4019 type = eSymbolTypeSourceFile; 4020 if (symbol_name == NULL) { 4021 add_nlist = false; 4022 if (N_SO_index != UINT32_MAX) { 4023 // Set the size of the N_SO to the terminating index of this 4024 // N_SO 4025 // so that we can always skip the entire N_SO if we need to 4026 // navigate 4027 // more quickly at the source level when parsing STABS 4028 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 4029 symbol_ptr->SetByteSize(sym_idx); 4030 symbol_ptr->SetSizeIsSibling(true); 4031 } 4032 N_NSYM_indexes.clear(); 4033 N_INCL_indexes.clear(); 4034 N_BRAC_indexes.clear(); 4035 N_COMM_indexes.clear(); 4036 N_FUN_indexes.clear(); 4037 N_SO_index = UINT32_MAX; 4038 } else { 4039 // We use the current number of symbols in the symbol table in 4040 // lieu of 4041 // using nlist_idx in case we ever start trimming entries out 4042 const bool N_SO_has_full_path = symbol_name[0] == '/'; 4043 if (N_SO_has_full_path) { 4044 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 4045 // We have two consecutive N_SO entries where the first 4046 // contains a directory 4047 // and the second contains a full path. 4048 sym[sym_idx - 1].GetMangled().SetValue( 4049 ConstString(symbol_name), false); 4050 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4051 add_nlist = false; 4052 } else { 4053 // This is the first entry in a N_SO that contains a directory 4054 // or 4055 // a full path to the source file 4056 N_SO_index = sym_idx; 4057 } 4058 } else if ((N_SO_index == sym_idx - 1) && 4059 ((sym_idx - 1) < num_syms)) { 4060 // This is usually the second N_SO entry that contains just the 4061 // filename, 4062 // so here we combine it with the first one if we are minimizing 4063 // the symbol table 4064 const char *so_path = 4065 sym[sym_idx - 1] 4066 .GetMangled() 4067 .GetDemangledName(lldb::eLanguageTypeUnknown) 4068 .AsCString(); 4069 if (so_path && so_path[0]) { 4070 std::string full_so_path(so_path); 4071 const size_t double_slash_pos = full_so_path.find("//"); 4072 if (double_slash_pos != std::string::npos) { 4073 // The linker has been generating bad N_SO entries with 4074 // doubled up paths 4075 // in the format "%s%s" where the first string in the 4076 // DW_AT_comp_dir, 4077 // and the second is the directory for the source file so 4078 // you end up with 4079 // a path that looks like "/tmp/src//tmp/src/" 4080 FileSpec so_dir(so_path, false); 4081 if (!so_dir.Exists()) { 4082 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 4083 false); 4084 if (so_dir.Exists()) { 4085 // Trim off the incorrect path 4086 full_so_path.erase(0, double_slash_pos + 1); 4087 } 4088 } 4089 } 4090 if (*full_so_path.rbegin() != '/') 4091 full_so_path += '/'; 4092 full_so_path += symbol_name; 4093 sym[sym_idx - 1].GetMangled().SetValue( 4094 ConstString(full_so_path.c_str()), false); 4095 add_nlist = false; 4096 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4097 } 4098 } else { 4099 // This could be a relative path to a N_SO 4100 N_SO_index = sym_idx; 4101 } 4102 } 4103 break; 4104 4105 case N_OSO: 4106 // object file name: name,,0,0,st_mtime 4107 type = eSymbolTypeObjectFile; 4108 break; 4109 4110 case N_LSYM: 4111 // local sym: name,,NO_SECT,type,offset 4112 type = eSymbolTypeLocal; 4113 break; 4114 4115 //---------------------------------------------------------------------- 4116 // INCL scopes 4117 //---------------------------------------------------------------------- 4118 case N_BINCL: 4119 // include file beginning: name,,NO_SECT,0,sum 4120 // We use the current number of symbols in the symbol table in lieu 4121 // of 4122 // using nlist_idx in case we ever start trimming entries out 4123 N_INCL_indexes.push_back(sym_idx); 4124 type = eSymbolTypeScopeBegin; 4125 break; 4126 4127 case N_EINCL: 4128 // include file end: name,,NO_SECT,0,0 4129 // Set the size of the N_BINCL to the terminating index of this 4130 // N_EINCL 4131 // so that we can always skip the entire symbol if we need to 4132 // navigate 4133 // more quickly at the source level when parsing STABS 4134 if (!N_INCL_indexes.empty()) { 4135 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4136 symbol_ptr->SetByteSize(sym_idx + 1); 4137 symbol_ptr->SetSizeIsSibling(true); 4138 N_INCL_indexes.pop_back(); 4139 } 4140 type = eSymbolTypeScopeEnd; 4141 break; 4142 4143 case N_SOL: 4144 // #included file name: name,,n_sect,0,address 4145 type = eSymbolTypeHeaderFile; 4146 4147 // We currently don't use the header files on darwin 4148 add_nlist = false; 4149 break; 4150 4151 case N_PARAMS: 4152 // compiler parameters: name,,NO_SECT,0,0 4153 type = eSymbolTypeCompiler; 4154 break; 4155 4156 case N_VERSION: 4157 // compiler version: name,,NO_SECT,0,0 4158 type = eSymbolTypeCompiler; 4159 break; 4160 4161 case N_OLEVEL: 4162 // compiler -O level: name,,NO_SECT,0,0 4163 type = eSymbolTypeCompiler; 4164 break; 4165 4166 case N_PSYM: 4167 // parameter: name,,NO_SECT,type,offset 4168 type = eSymbolTypeVariable; 4169 break; 4170 4171 case N_ENTRY: 4172 // alternate entry: name,,n_sect,linenumber,address 4173 symbol_section = 4174 section_info.GetSection(nlist.n_sect, nlist.n_value); 4175 type = eSymbolTypeLineEntry; 4176 break; 4177 4178 //---------------------------------------------------------------------- 4179 // Left and Right Braces 4180 //---------------------------------------------------------------------- 4181 case N_LBRAC: 4182 // left bracket: 0,,NO_SECT,nesting level,address 4183 // We use the current number of symbols in the symbol table in lieu 4184 // of 4185 // using nlist_idx in case we ever start trimming entries out 4186 symbol_section = 4187 section_info.GetSection(nlist.n_sect, nlist.n_value); 4188 N_BRAC_indexes.push_back(sym_idx); 4189 type = eSymbolTypeScopeBegin; 4190 break; 4191 4192 case N_RBRAC: 4193 // right bracket: 0,,NO_SECT,nesting level,address 4194 // Set the size of the N_LBRAC to the terminating index of this 4195 // N_RBRAC 4196 // so that we can always skip the entire symbol if we need to 4197 // navigate 4198 // more quickly at the source level when parsing STABS 4199 symbol_section = 4200 section_info.GetSection(nlist.n_sect, nlist.n_value); 4201 if (!N_BRAC_indexes.empty()) { 4202 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4203 symbol_ptr->SetByteSize(sym_idx + 1); 4204 symbol_ptr->SetSizeIsSibling(true); 4205 N_BRAC_indexes.pop_back(); 4206 } 4207 type = eSymbolTypeScopeEnd; 4208 break; 4209 4210 case N_EXCL: 4211 // deleted include file: name,,NO_SECT,0,sum 4212 type = eSymbolTypeHeaderFile; 4213 break; 4214 4215 //---------------------------------------------------------------------- 4216 // COMM scopes 4217 //---------------------------------------------------------------------- 4218 case N_BCOMM: 4219 // begin common: name,,NO_SECT,0,0 4220 // We use the current number of symbols in the symbol table in lieu 4221 // of 4222 // using nlist_idx in case we ever start trimming entries out 4223 type = eSymbolTypeScopeBegin; 4224 N_COMM_indexes.push_back(sym_idx); 4225 break; 4226 4227 case N_ECOML: 4228 // end common (local name): 0,,n_sect,0,address 4229 symbol_section = 4230 section_info.GetSection(nlist.n_sect, nlist.n_value); 4231 LLVM_FALLTHROUGH; 4232 4233 case N_ECOMM: 4234 // end common: name,,n_sect,0,0 4235 // Set the size of the N_BCOMM to the terminating index of this 4236 // N_ECOMM/N_ECOML 4237 // so that we can always skip the entire symbol if we need to 4238 // navigate 4239 // more quickly at the source level when parsing STABS 4240 if (!N_COMM_indexes.empty()) { 4241 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4242 symbol_ptr->SetByteSize(sym_idx + 1); 4243 symbol_ptr->SetSizeIsSibling(true); 4244 N_COMM_indexes.pop_back(); 4245 } 4246 type = eSymbolTypeScopeEnd; 4247 break; 4248 4249 case N_LENG: 4250 // second stab entry with length information 4251 type = eSymbolTypeAdditional; 4252 break; 4253 4254 default: 4255 break; 4256 } 4257 } else { 4258 // uint8_t n_pext = N_PEXT & nlist.n_type; 4259 uint8_t n_type = N_TYPE & nlist.n_type; 4260 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4261 4262 switch (n_type) { 4263 case N_INDR: { 4264 const char *reexport_name_cstr = 4265 strtab_data.PeekCStr(nlist.n_value); 4266 if (reexport_name_cstr && reexport_name_cstr[0]) { 4267 type = eSymbolTypeReExported; 4268 ConstString reexport_name( 4269 reexport_name_cstr + 4270 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4271 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4272 set_value = false; 4273 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4274 indirect_symbol_names.insert( 4275 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4276 } else 4277 type = eSymbolTypeUndefined; 4278 } break; 4279 4280 case N_UNDF: 4281 if (symbol_name && symbol_name[0]) { 4282 ConstString undefined_name(symbol_name + 4283 ((symbol_name[0] == '_') ? 1 : 0)); 4284 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4285 } 4286 LLVM_FALLTHROUGH; 4287 4288 case N_PBUD: 4289 type = eSymbolTypeUndefined; 4290 break; 4291 4292 case N_ABS: 4293 type = eSymbolTypeAbsolute; 4294 break; 4295 4296 case N_SECT: { 4297 symbol_section = 4298 section_info.GetSection(nlist.n_sect, nlist.n_value); 4299 4300 if (!symbol_section) { 4301 // TODO: warn about this? 4302 add_nlist = false; 4303 break; 4304 } 4305 4306 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4307 type = eSymbolTypeException; 4308 } else { 4309 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4310 4311 switch (section_type) { 4312 case S_CSTRING_LITERALS: 4313 type = eSymbolTypeData; 4314 break; // section with only literal C strings 4315 case S_4BYTE_LITERALS: 4316 type = eSymbolTypeData; 4317 break; // section with only 4 byte literals 4318 case S_8BYTE_LITERALS: 4319 type = eSymbolTypeData; 4320 break; // section with only 8 byte literals 4321 case S_LITERAL_POINTERS: 4322 type = eSymbolTypeTrampoline; 4323 break; // section with only pointers to literals 4324 case S_NON_LAZY_SYMBOL_POINTERS: 4325 type = eSymbolTypeTrampoline; 4326 break; // section with only non-lazy symbol pointers 4327 case S_LAZY_SYMBOL_POINTERS: 4328 type = eSymbolTypeTrampoline; 4329 break; // section with only lazy symbol pointers 4330 case S_SYMBOL_STUBS: 4331 type = eSymbolTypeTrampoline; 4332 break; // section with only symbol stubs, byte size of stub in 4333 // the reserved2 field 4334 case S_MOD_INIT_FUNC_POINTERS: 4335 type = eSymbolTypeCode; 4336 break; // section with only function pointers for initialization 4337 case S_MOD_TERM_FUNC_POINTERS: 4338 type = eSymbolTypeCode; 4339 break; // section with only function pointers for termination 4340 case S_INTERPOSING: 4341 type = eSymbolTypeTrampoline; 4342 break; // section with only pairs of function pointers for 4343 // interposing 4344 case S_16BYTE_LITERALS: 4345 type = eSymbolTypeData; 4346 break; // section with only 16 byte literals 4347 case S_DTRACE_DOF: 4348 type = eSymbolTypeInstrumentation; 4349 break; 4350 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4351 type = eSymbolTypeTrampoline; 4352 break; 4353 default: 4354 switch (symbol_section->GetType()) { 4355 case lldb::eSectionTypeCode: 4356 type = eSymbolTypeCode; 4357 break; 4358 case eSectionTypeData: 4359 case eSectionTypeDataCString: // Inlined C string data 4360 case eSectionTypeDataCStringPointers: // Pointers to C string 4361 // data 4362 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4363 // the symbol table 4364 case eSectionTypeData4: 4365 case eSectionTypeData8: 4366 case eSectionTypeData16: 4367 type = eSymbolTypeData; 4368 break; 4369 default: 4370 break; 4371 } 4372 break; 4373 } 4374 4375 if (type == eSymbolTypeInvalid) { 4376 const char *symbol_sect_name = 4377 symbol_section->GetName().AsCString(); 4378 if (symbol_section->IsDescendant(text_section_sp.get())) { 4379 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4380 S_ATTR_SELF_MODIFYING_CODE | 4381 S_ATTR_SOME_INSTRUCTIONS)) 4382 type = eSymbolTypeData; 4383 else 4384 type = eSymbolTypeCode; 4385 } else if (symbol_section->IsDescendant( 4386 data_section_sp.get()) || 4387 symbol_section->IsDescendant( 4388 data_dirty_section_sp.get()) || 4389 symbol_section->IsDescendant( 4390 data_const_section_sp.get())) { 4391 if (symbol_sect_name && 4392 ::strstr(symbol_sect_name, "__objc") == 4393 symbol_sect_name) { 4394 type = eSymbolTypeRuntime; 4395 4396 if (symbol_name) { 4397 llvm::StringRef symbol_name_ref(symbol_name); 4398 if (symbol_name_ref.startswith("_OBJC_")) { 4399 static const llvm::StringRef g_objc_v2_prefix_class( 4400 "_OBJC_CLASS_$_"); 4401 static const llvm::StringRef g_objc_v2_prefix_metaclass( 4402 "_OBJC_METACLASS_$_"); 4403 static const llvm::StringRef g_objc_v2_prefix_ivar( 4404 "_OBJC_IVAR_$_"); 4405 if (symbol_name_ref.startswith( 4406 g_objc_v2_prefix_class)) { 4407 symbol_name_non_abi_mangled = symbol_name + 1; 4408 symbol_name = 4409 symbol_name + g_objc_v2_prefix_class.size(); 4410 type = eSymbolTypeObjCClass; 4411 demangled_is_synthesized = true; 4412 } else if (symbol_name_ref.startswith( 4413 g_objc_v2_prefix_metaclass)) { 4414 symbol_name_non_abi_mangled = symbol_name + 1; 4415 symbol_name = 4416 symbol_name + g_objc_v2_prefix_metaclass.size(); 4417 type = eSymbolTypeObjCMetaClass; 4418 demangled_is_synthesized = true; 4419 } else if (symbol_name_ref.startswith( 4420 g_objc_v2_prefix_ivar)) { 4421 symbol_name_non_abi_mangled = symbol_name + 1; 4422 symbol_name = 4423 symbol_name + g_objc_v2_prefix_ivar.size(); 4424 type = eSymbolTypeObjCIVar; 4425 demangled_is_synthesized = true; 4426 } 4427 } 4428 } 4429 } else if (symbol_sect_name && 4430 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4431 symbol_sect_name) { 4432 type = eSymbolTypeException; 4433 } else { 4434 type = eSymbolTypeData; 4435 } 4436 } else if (symbol_sect_name && 4437 ::strstr(symbol_sect_name, "__IMPORT") == 4438 symbol_sect_name) { 4439 type = eSymbolTypeTrampoline; 4440 } else if (symbol_section->IsDescendant( 4441 objc_section_sp.get())) { 4442 type = eSymbolTypeRuntime; 4443 if (symbol_name && symbol_name[0] == '.') { 4444 llvm::StringRef symbol_name_ref(symbol_name); 4445 static const llvm::StringRef g_objc_v1_prefix_class( 4446 ".objc_class_name_"); 4447 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4448 symbol_name_non_abi_mangled = symbol_name; 4449 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4450 type = eSymbolTypeObjCClass; 4451 demangled_is_synthesized = true; 4452 } 4453 } 4454 } 4455 } 4456 } 4457 } break; 4458 } 4459 } 4460 4461 if (add_nlist) { 4462 uint64_t symbol_value = nlist.n_value; 4463 4464 if (symbol_name_non_abi_mangled) { 4465 sym[sym_idx].GetMangled().SetMangledName( 4466 ConstString(symbol_name_non_abi_mangled)); 4467 sym[sym_idx].GetMangled().SetDemangledName( 4468 ConstString(symbol_name)); 4469 } else { 4470 bool symbol_name_is_mangled = false; 4471 4472 if (symbol_name && symbol_name[0] == '_') { 4473 symbol_name_is_mangled = symbol_name[1] == '_'; 4474 symbol_name++; // Skip the leading underscore 4475 } 4476 4477 if (symbol_name) { 4478 ConstString const_symbol_name(symbol_name); 4479 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4480 symbol_name_is_mangled); 4481 } 4482 } 4483 4484 if (is_gsym) { 4485 const char *gsym_name = sym[sym_idx] 4486 .GetMangled() 4487 .GetName(lldb::eLanguageTypeUnknown, 4488 Mangled::ePreferMangled) 4489 .GetCString(); 4490 if (gsym_name) 4491 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4492 } 4493 4494 if (symbol_section) { 4495 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4496 if (symbol_byte_size == 0 && function_starts_count > 0) { 4497 addr_t symbol_lookup_file_addr = nlist.n_value; 4498 // Do an exact address match for non-ARM addresses, else get the 4499 // closest since 4500 // the symbol might be a thumb symbol which has an address with 4501 // bit zero set 4502 FunctionStarts::Entry *func_start_entry = 4503 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4504 if (is_arm && func_start_entry) { 4505 // Verify that the function start address is the symbol address 4506 // (ARM) 4507 // or the symbol address + 1 (thumb) 4508 if (func_start_entry->addr != symbol_lookup_file_addr && 4509 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4510 // Not the right entry, NULL it out... 4511 func_start_entry = NULL; 4512 } 4513 } 4514 if (func_start_entry) { 4515 func_start_entry->data = true; 4516 4517 addr_t symbol_file_addr = func_start_entry->addr; 4518 if (is_arm) 4519 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4520 4521 const FunctionStarts::Entry *next_func_start_entry = 4522 function_starts.FindNextEntry(func_start_entry); 4523 const addr_t section_end_file_addr = 4524 section_file_addr + symbol_section->GetByteSize(); 4525 if (next_func_start_entry) { 4526 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4527 // Be sure the clear the Thumb address bit when we calculate 4528 // the size 4529 // from the current and next address 4530 if (is_arm) 4531 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4532 symbol_byte_size = std::min<lldb::addr_t>( 4533 next_symbol_file_addr - symbol_file_addr, 4534 section_end_file_addr - symbol_file_addr); 4535 } else { 4536 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4537 } 4538 } 4539 } 4540 symbol_value -= section_file_addr; 4541 } 4542 4543 if (is_debug == false) { 4544 if (type == eSymbolTypeCode) { 4545 // See if we can find a N_FUN entry for any code symbols. 4546 // If we do find a match, and the name matches, then we 4547 // can merge the two into just the function symbol to avoid 4548 // duplicate entries in the symbol table 4549 std::pair<ValueToSymbolIndexMap::const_iterator, 4550 ValueToSymbolIndexMap::const_iterator> 4551 range; 4552 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4553 if (range.first != range.second) { 4554 bool found_it = false; 4555 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4556 pos != range.second; ++pos) { 4557 if (sym[sym_idx].GetMangled().GetName( 4558 lldb::eLanguageTypeUnknown, 4559 Mangled::ePreferMangled) == 4560 sym[pos->second].GetMangled().GetName( 4561 lldb::eLanguageTypeUnknown, 4562 Mangled::ePreferMangled)) { 4563 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4564 // We just need the flags from the linker symbol, so put 4565 // these flags 4566 // into the N_FUN flags to avoid duplicate symbols in the 4567 // symbol table 4568 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4569 sym[pos->second].SetFlags(nlist.n_type << 16 | 4570 nlist.n_desc); 4571 if (resolver_addresses.find(nlist.n_value) != 4572 resolver_addresses.end()) 4573 sym[pos->second].SetType(eSymbolTypeResolver); 4574 sym[sym_idx].Clear(); 4575 found_it = true; 4576 break; 4577 } 4578 } 4579 if (found_it) 4580 continue; 4581 } else { 4582 if (resolver_addresses.find(nlist.n_value) != 4583 resolver_addresses.end()) 4584 type = eSymbolTypeResolver; 4585 } 4586 } else if (type == eSymbolTypeData || 4587 type == eSymbolTypeObjCClass || 4588 type == eSymbolTypeObjCMetaClass || 4589 type == eSymbolTypeObjCIVar) { 4590 // See if we can find a N_STSYM entry for any data symbols. 4591 // If we do find a match, and the name matches, then we 4592 // can merge the two into just the Static symbol to avoid 4593 // duplicate entries in the symbol table 4594 std::pair<ValueToSymbolIndexMap::const_iterator, 4595 ValueToSymbolIndexMap::const_iterator> 4596 range; 4597 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4598 if (range.first != range.second) { 4599 bool found_it = false; 4600 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4601 pos != range.second; ++pos) { 4602 if (sym[sym_idx].GetMangled().GetName( 4603 lldb::eLanguageTypeUnknown, 4604 Mangled::ePreferMangled) == 4605 sym[pos->second].GetMangled().GetName( 4606 lldb::eLanguageTypeUnknown, 4607 Mangled::ePreferMangled)) { 4608 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4609 // We just need the flags from the linker symbol, so put 4610 // these flags 4611 // into the N_STSYM flags to avoid duplicate symbols in the 4612 // symbol table 4613 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4614 sym[pos->second].SetFlags(nlist.n_type << 16 | 4615 nlist.n_desc); 4616 sym[sym_idx].Clear(); 4617 found_it = true; 4618 break; 4619 } 4620 } 4621 if (found_it) 4622 continue; 4623 } else { 4624 // Combine N_GSYM stab entries with the non stab symbol 4625 const char *gsym_name = sym[sym_idx] 4626 .GetMangled() 4627 .GetName(lldb::eLanguageTypeUnknown, 4628 Mangled::ePreferMangled) 4629 .GetCString(); 4630 if (gsym_name) { 4631 ConstNameToSymbolIndexMap::const_iterator pos = 4632 N_GSYM_name_to_sym_idx.find(gsym_name); 4633 if (pos != N_GSYM_name_to_sym_idx.end()) { 4634 const uint32_t GSYM_sym_idx = pos->second; 4635 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4636 // Copy the address, because often the N_GSYM address has an 4637 // invalid address of zero 4638 // when the global is a common symbol 4639 sym[GSYM_sym_idx].GetAddressRef().SetSection( 4640 symbol_section); 4641 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4642 // We just need the flags from the linker symbol, so put 4643 // these flags 4644 // into the N_GSYM flags to avoid duplicate symbols in the 4645 // symbol table 4646 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 4647 nlist.n_desc); 4648 sym[sym_idx].Clear(); 4649 continue; 4650 } 4651 } 4652 } 4653 } 4654 } 4655 4656 sym[sym_idx].SetID(nlist_idx); 4657 sym[sym_idx].SetType(type); 4658 if (set_value) { 4659 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4660 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4661 } 4662 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4663 4664 if (symbol_byte_size > 0) 4665 sym[sym_idx].SetByteSize(symbol_byte_size); 4666 4667 if (demangled_is_synthesized) 4668 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4669 4670 ++sym_idx; 4671 } else { 4672 sym[sym_idx].Clear(); 4673 } 4674 } 4675 4676 for (const auto &pos : reexport_shlib_needs_fixup) { 4677 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4678 if (undef_pos != undefined_name_to_desc.end()) { 4679 const uint8_t dylib_ordinal = 4680 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4681 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4682 sym[pos.first].SetReExportedSymbolSharedLibrary( 4683 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4684 } 4685 } 4686 } 4687 4688 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4689 4690 if (function_starts_count > 0) { 4691 uint32_t num_synthetic_function_symbols = 0; 4692 for (i = 0; i < function_starts_count; ++i) { 4693 if (function_starts.GetEntryRef(i).data == false) 4694 ++num_synthetic_function_symbols; 4695 } 4696 4697 if (num_synthetic_function_symbols > 0) { 4698 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4699 num_syms = sym_idx + num_synthetic_function_symbols; 4700 sym = symtab->Resize(num_syms); 4701 } 4702 for (i = 0; i < function_starts_count; ++i) { 4703 const FunctionStarts::Entry *func_start_entry = 4704 function_starts.GetEntryAtIndex(i); 4705 if (func_start_entry->data == false) { 4706 addr_t symbol_file_addr = func_start_entry->addr; 4707 uint32_t symbol_flags = 0; 4708 if (is_arm) { 4709 if (symbol_file_addr & 1) 4710 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4711 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4712 } 4713 Address symbol_addr; 4714 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4715 SectionSP symbol_section(symbol_addr.GetSection()); 4716 uint32_t symbol_byte_size = 0; 4717 if (symbol_section) { 4718 const addr_t section_file_addr = 4719 symbol_section->GetFileAddress(); 4720 const FunctionStarts::Entry *next_func_start_entry = 4721 function_starts.FindNextEntry(func_start_entry); 4722 const addr_t section_end_file_addr = 4723 section_file_addr + symbol_section->GetByteSize(); 4724 if (next_func_start_entry) { 4725 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4726 if (is_arm) 4727 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4728 symbol_byte_size = std::min<lldb::addr_t>( 4729 next_symbol_file_addr - symbol_file_addr, 4730 section_end_file_addr - symbol_file_addr); 4731 } else { 4732 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4733 } 4734 sym[sym_idx].SetID(synthetic_sym_id++); 4735 sym[sym_idx].GetMangled().SetDemangledName( 4736 GetNextSyntheticSymbolName()); 4737 sym[sym_idx].SetType(eSymbolTypeCode); 4738 sym[sym_idx].SetIsSynthetic(true); 4739 sym[sym_idx].GetAddressRef() = symbol_addr; 4740 if (symbol_flags) 4741 sym[sym_idx].SetFlags(symbol_flags); 4742 if (symbol_byte_size) 4743 sym[sym_idx].SetByteSize(symbol_byte_size); 4744 ++sym_idx; 4745 } 4746 } 4747 } 4748 } 4749 } 4750 } 4751 4752 // Trim our symbols down to just what we ended up with after 4753 // removing any symbols. 4754 if (sym_idx < num_syms) { 4755 num_syms = sym_idx; 4756 sym = symtab->Resize(num_syms); 4757 } 4758 4759 // Now synthesize indirect symbols 4760 if (m_dysymtab.nindirectsyms != 0) { 4761 if (indirect_symbol_index_data.GetByteSize()) { 4762 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4763 m_nlist_idx_to_sym_idx.end(); 4764 4765 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4766 ++sect_idx) { 4767 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4768 S_SYMBOL_STUBS) { 4769 uint32_t symbol_stub_byte_size = 4770 m_mach_sections[sect_idx].reserved2; 4771 if (symbol_stub_byte_size == 0) 4772 continue; 4773 4774 const uint32_t num_symbol_stubs = 4775 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4776 4777 if (num_symbol_stubs == 0) 4778 continue; 4779 4780 const uint32_t symbol_stub_index_offset = 4781 m_mach_sections[sect_idx].reserved1; 4782 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; 4783 ++stub_idx) { 4784 const uint32_t symbol_stub_index = 4785 symbol_stub_index_offset + stub_idx; 4786 const lldb::addr_t symbol_stub_addr = 4787 m_mach_sections[sect_idx].addr + 4788 (stub_idx * symbol_stub_byte_size); 4789 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4790 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4791 symbol_stub_offset, 4)) { 4792 const uint32_t stub_sym_id = 4793 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4794 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4795 continue; 4796 4797 NListIndexToSymbolIndexMap::const_iterator index_pos = 4798 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4799 Symbol *stub_symbol = NULL; 4800 if (index_pos != end_index_pos) { 4801 // We have a remapping from the original nlist index to 4802 // a current symbol index, so just look this up by index 4803 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4804 } else { 4805 // We need to lookup a symbol using the original nlist 4806 // symbol index since this index is coming from the 4807 // S_SYMBOL_STUBS 4808 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4809 } 4810 4811 if (stub_symbol) { 4812 Address so_addr(symbol_stub_addr, section_list); 4813 4814 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4815 // Change the external symbol into a trampoline that makes 4816 // sense 4817 // These symbols were N_UNDF N_EXT, and are useless to us, 4818 // so we 4819 // can re-use them so we don't have to make up a synthetic 4820 // symbol 4821 // for no good reason. 4822 if (resolver_addresses.find(symbol_stub_addr) == 4823 resolver_addresses.end()) 4824 stub_symbol->SetType(eSymbolTypeTrampoline); 4825 else 4826 stub_symbol->SetType(eSymbolTypeResolver); 4827 stub_symbol->SetExternal(false); 4828 stub_symbol->GetAddressRef() = so_addr; 4829 stub_symbol->SetByteSize(symbol_stub_byte_size); 4830 } else { 4831 // Make a synthetic symbol to describe the trampoline stub 4832 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4833 if (sym_idx >= num_syms) { 4834 sym = symtab->Resize(++num_syms); 4835 stub_symbol = NULL; // this pointer no longer valid 4836 } 4837 sym[sym_idx].SetID(synthetic_sym_id++); 4838 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4839 if (resolver_addresses.find(symbol_stub_addr) == 4840 resolver_addresses.end()) 4841 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4842 else 4843 sym[sym_idx].SetType(eSymbolTypeResolver); 4844 sym[sym_idx].SetIsSynthetic(true); 4845 sym[sym_idx].GetAddressRef() = so_addr; 4846 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4847 ++sym_idx; 4848 } 4849 } else { 4850 if (log) 4851 log->Warning("symbol stub referencing symbol table symbol " 4852 "%u that isn't in our minimal symbol table, " 4853 "fix this!!!", 4854 stub_sym_id); 4855 } 4856 } 4857 } 4858 } 4859 } 4860 } 4861 } 4862 4863 if (!trie_entries.empty()) { 4864 for (const auto &e : trie_entries) { 4865 if (e.entry.import_name) { 4866 // Only add indirect symbols from the Trie entries if we 4867 // didn't have a N_INDR nlist entry for this already 4868 if (indirect_symbol_names.find(e.entry.name) == 4869 indirect_symbol_names.end()) { 4870 // Make a synthetic symbol to describe re-exported symbol. 4871 if (sym_idx >= num_syms) 4872 sym = symtab->Resize(++num_syms); 4873 sym[sym_idx].SetID(synthetic_sym_id++); 4874 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4875 sym[sym_idx].SetType(eSymbolTypeReExported); 4876 sym[sym_idx].SetIsSynthetic(true); 4877 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4878 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4879 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4880 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4881 } 4882 ++sym_idx; 4883 } 4884 } 4885 } 4886 } 4887 4888 // StreamFile s(stdout, false); 4889 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4890 // symtab->Dump(&s, NULL, eSortOrderNone); 4891 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4892 // sizes 4893 symtab->CalculateSymbolSizes(); 4894 4895 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4896 // symtab->Dump(&s, NULL, eSortOrderNone); 4897 4898 return symtab->GetNumSymbols(); 4899 } 4900 return 0; 4901 } 4902 4903 void ObjectFileMachO::Dump(Stream *s) { 4904 ModuleSP module_sp(GetModule()); 4905 if (module_sp) { 4906 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4907 s->Printf("%p: ", static_cast<void *>(this)); 4908 s->Indent(); 4909 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4910 s->PutCString("ObjectFileMachO64"); 4911 else 4912 s->PutCString("ObjectFileMachO32"); 4913 4914 ArchSpec header_arch; 4915 GetArchitecture(header_arch); 4916 4917 *s << ", file = '" << m_file 4918 << "', arch = " << header_arch.GetArchitectureName() << "\n"; 4919 4920 SectionList *sections = GetSectionList(); 4921 if (sections) 4922 sections->Dump(s, NULL, true, UINT32_MAX); 4923 4924 if (m_symtab_ap.get()) 4925 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4926 } 4927 } 4928 4929 bool ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4930 const lldb_private::DataExtractor &data, 4931 lldb::offset_t lc_offset, 4932 lldb_private::UUID &uuid) { 4933 uint32_t i; 4934 struct uuid_command load_cmd; 4935 4936 lldb::offset_t offset = lc_offset; 4937 for (i = 0; i < header.ncmds; ++i) { 4938 const lldb::offset_t cmd_offset = offset; 4939 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4940 break; 4941 4942 if (load_cmd.cmd == LC_UUID) { 4943 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4944 4945 if (uuid_bytes) { 4946 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4947 // We pretend these object files have no UUID to prevent crashing. 4948 4949 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4950 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4951 0xbb, 0x14, 0xf0, 0x0d}; 4952 4953 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4954 return false; 4955 4956 uuid.SetBytes(uuid_bytes); 4957 return true; 4958 } 4959 return false; 4960 } 4961 offset = cmd_offset + load_cmd.cmdsize; 4962 } 4963 return false; 4964 } 4965 4966 bool ObjectFileMachO::GetArchitecture(const llvm::MachO::mach_header &header, 4967 const lldb_private::DataExtractor &data, 4968 lldb::offset_t lc_offset, 4969 ArchSpec &arch) { 4970 arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4971 4972 if (arch.IsValid()) { 4973 llvm::Triple &triple = arch.GetTriple(); 4974 4975 // Set OS to an unspecified unknown or a "*" so it can match any OS 4976 triple.setOS(llvm::Triple::UnknownOS); 4977 triple.setOSName(llvm::StringRef()); 4978 4979 if (header.filetype == MH_PRELOAD) { 4980 if (header.cputype == CPU_TYPE_ARM) { 4981 // If this is a 32-bit arm binary, and it's a standalone binary, 4982 // force the Vendor to Apple so we don't accidentally pick up 4983 // the generic armv7 ABI at runtime. Apple's armv7 ABI always uses 4984 // r7 for the frame pointer register; most other armv7 ABIs use a 4985 // combination of r7 and r11. 4986 triple.setVendor(llvm::Triple::Apple); 4987 } else { 4988 // Set vendor to an unspecified unknown or a "*" so it can match any 4989 // vendor 4990 // This is required for correct behavior of EFI debugging on x86_64 4991 triple.setVendor(llvm::Triple::UnknownVendor); 4992 triple.setVendorName(llvm::StringRef()); 4993 } 4994 return true; 4995 } else { 4996 struct load_command load_cmd; 4997 4998 lldb::offset_t offset = lc_offset; 4999 for (uint32_t i = 0; i < header.ncmds; ++i) { 5000 const lldb::offset_t cmd_offset = offset; 5001 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 5002 break; 5003 5004 switch (load_cmd.cmd) { 5005 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 5006 triple.setOS(llvm::Triple::IOS); 5007 return true; 5008 5009 case llvm::MachO::LC_VERSION_MIN_MACOSX: 5010 triple.setOS(llvm::Triple::MacOSX); 5011 return true; 5012 5013 case llvm::MachO::LC_VERSION_MIN_TVOS: 5014 triple.setOS(llvm::Triple::TvOS); 5015 return true; 5016 5017 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 5018 triple.setOS(llvm::Triple::WatchOS); 5019 return true; 5020 5021 default: 5022 break; 5023 } 5024 5025 offset = cmd_offset + load_cmd.cmdsize; 5026 } 5027 5028 if (header.filetype != MH_KEXT_BUNDLE) { 5029 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 5030 // so lets not say our Vendor is Apple, leave it as an unspecified 5031 // unknown 5032 triple.setVendor(llvm::Triple::UnknownVendor); 5033 triple.setVendorName(llvm::StringRef()); 5034 } 5035 } 5036 } 5037 return arch.IsValid(); 5038 } 5039 5040 bool ObjectFileMachO::GetUUID(lldb_private::UUID *uuid) { 5041 ModuleSP module_sp(GetModule()); 5042 if (module_sp) { 5043 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5044 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5045 return GetUUID(m_header, m_data, offset, *uuid); 5046 } 5047 return false; 5048 } 5049 5050 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5051 uint32_t count = 0; 5052 ModuleSP module_sp(GetModule()); 5053 if (module_sp) { 5054 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5055 struct load_command load_cmd; 5056 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5057 std::vector<std::string> rpath_paths; 5058 std::vector<std::string> rpath_relative_paths; 5059 std::vector<std::string> at_exec_relative_paths; 5060 const bool resolve_path = false; // Don't resolve the dependent file paths 5061 // since they may not reside on this system 5062 uint32_t i; 5063 for (i = 0; i < m_header.ncmds; ++i) { 5064 const uint32_t cmd_offset = offset; 5065 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5066 break; 5067 5068 switch (load_cmd.cmd) { 5069 case LC_RPATH: 5070 case LC_LOAD_DYLIB: 5071 case LC_LOAD_WEAK_DYLIB: 5072 case LC_REEXPORT_DYLIB: 5073 case LC_LOAD_DYLINKER: 5074 case LC_LOADFVMLIB: 5075 case LC_LOAD_UPWARD_DYLIB: { 5076 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5077 const char *path = m_data.PeekCStr(name_offset); 5078 if (path) { 5079 if (load_cmd.cmd == LC_RPATH) 5080 rpath_paths.push_back(path); 5081 else { 5082 if (path[0] == '@') { 5083 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5084 rpath_relative_paths.push_back(path + strlen("@rpath")); 5085 else if (strncmp(path, "@executable_path", 5086 strlen("@executable_path")) == 0) 5087 at_exec_relative_paths.push_back(path 5088 + strlen("@executable_path")); 5089 } else { 5090 FileSpec file_spec(path, resolve_path); 5091 if (files.AppendIfUnique(file_spec)) 5092 count++; 5093 } 5094 } 5095 } 5096 } break; 5097 5098 default: 5099 break; 5100 } 5101 offset = cmd_offset + load_cmd.cmdsize; 5102 } 5103 5104 FileSpec this_file_spec(m_file); 5105 this_file_spec.ResolvePath(); 5106 5107 if (!rpath_paths.empty()) { 5108 // Fixup all LC_RPATH values to be absolute paths 5109 std::string loader_path("@loader_path"); 5110 std::string executable_path("@executable_path"); 5111 for (auto &rpath : rpath_paths) { 5112 if (rpath.find(loader_path) == 0) { 5113 rpath.erase(0, loader_path.size()); 5114 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5115 } else if (rpath.find(executable_path) == 0) { 5116 rpath.erase(0, executable_path.size()); 5117 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5118 } 5119 } 5120 5121 for (const auto &rpath_relative_path : rpath_relative_paths) { 5122 for (const auto &rpath : rpath_paths) { 5123 std::string path = rpath; 5124 path += rpath_relative_path; 5125 // It is OK to resolve this path because we must find a file on 5126 // disk for us to accept it anyway if it is rpath relative. 5127 FileSpec file_spec(path, true); 5128 // Remove any redundant parts of the path (like "../foo") since 5129 // LC_RPATH values often contain "..". 5130 file_spec = file_spec.GetNormalizedPath(); 5131 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) { 5132 count++; 5133 break; 5134 } 5135 } 5136 } 5137 } 5138 5139 // We may have @executable_paths but no RPATHS. Figure those out here. 5140 // Only do this if this object file is the executable. We have no way to 5141 // get back to the actual executable otherwise, so we won't get the right 5142 // path. 5143 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5144 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5145 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5146 FileSpec file_spec = 5147 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5148 file_spec = file_spec.GetNormalizedPath(); 5149 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) { 5150 count++; 5151 break; 5152 } 5153 } 5154 } 5155 } 5156 return count; 5157 } 5158 5159 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5160 // If the object file is not an executable it can't hold the entry point. 5161 // m_entry_point_address 5162 // is initialized to an invalid address, so we can just return that. 5163 // If m_entry_point_address is valid it means we've found it already, so 5164 // return the cached value. 5165 5166 if (!IsExecutable() || m_entry_point_address.IsValid()) 5167 return m_entry_point_address; 5168 5169 // Otherwise, look for the UnixThread or Thread command. The data for the 5170 // Thread command is given in 5171 // /usr/include/mach-o.h, but it is basically: 5172 // 5173 // uint32_t flavor - this is the flavor argument you would pass to 5174 // thread_get_state 5175 // uint32_t count - this is the count of longs in the thread state data 5176 // struct XXX_thread_state state - this is the structure from 5177 // <machine/thread_status.h> corresponding to the flavor. 5178 // <repeat this trio> 5179 // 5180 // So we just keep reading the various register flavors till we find the GPR 5181 // one, then read the PC out of there. 5182 // FIXME: We will need to have a "RegisterContext data provider" class at some 5183 // point that can get all the registers 5184 // out of data in this form & attach them to a given thread. That should 5185 // underlie the MacOS X User process plugin, 5186 // and we'll also need it for the MacOS X Core File process plugin. When we 5187 // have that we can also use it here. 5188 // 5189 // For now we hard-code the offsets and flavors we need: 5190 // 5191 // 5192 5193 ModuleSP module_sp(GetModule()); 5194 if (module_sp) { 5195 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5196 struct load_command load_cmd; 5197 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5198 uint32_t i; 5199 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5200 bool done = false; 5201 5202 for (i = 0; i < m_header.ncmds; ++i) { 5203 const lldb::offset_t cmd_offset = offset; 5204 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5205 break; 5206 5207 switch (load_cmd.cmd) { 5208 case LC_UNIXTHREAD: 5209 case LC_THREAD: { 5210 while (offset < cmd_offset + load_cmd.cmdsize) { 5211 uint32_t flavor = m_data.GetU32(&offset); 5212 uint32_t count = m_data.GetU32(&offset); 5213 if (count == 0) { 5214 // We've gotten off somehow, log and exit; 5215 return m_entry_point_address; 5216 } 5217 5218 switch (m_header.cputype) { 5219 case llvm::MachO::CPU_TYPE_ARM: 5220 if (flavor == 1 || 5221 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 from 5222 // mach/arm/thread_status.h 5223 { 5224 offset += 60; // This is the offset of pc in the GPR thread state 5225 // data structure. 5226 start_address = m_data.GetU32(&offset); 5227 done = true; 5228 } 5229 break; 5230 case llvm::MachO::CPU_TYPE_ARM64: 5231 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5232 { 5233 offset += 256; // This is the offset of pc in the GPR thread state 5234 // data structure. 5235 start_address = m_data.GetU64(&offset); 5236 done = true; 5237 } 5238 break; 5239 case llvm::MachO::CPU_TYPE_I386: 5240 if (flavor == 5241 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5242 { 5243 offset += 40; // This is the offset of eip in the GPR thread state 5244 // data structure. 5245 start_address = m_data.GetU32(&offset); 5246 done = true; 5247 } 5248 break; 5249 case llvm::MachO::CPU_TYPE_X86_64: 5250 if (flavor == 5251 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5252 { 5253 offset += 16 * 8; // This is the offset of rip in the GPR thread 5254 // state data structure. 5255 start_address = m_data.GetU64(&offset); 5256 done = true; 5257 } 5258 break; 5259 default: 5260 return m_entry_point_address; 5261 } 5262 // Haven't found the GPR flavor yet, skip over the data for this 5263 // flavor: 5264 if (done) 5265 break; 5266 offset += count * 4; 5267 } 5268 } break; 5269 case LC_MAIN: { 5270 ConstString text_segment_name("__TEXT"); 5271 uint64_t entryoffset = m_data.GetU64(&offset); 5272 SectionSP text_segment_sp = 5273 GetSectionList()->FindSectionByName(text_segment_name); 5274 if (text_segment_sp) { 5275 done = true; 5276 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5277 } 5278 } break; 5279 5280 default: 5281 break; 5282 } 5283 if (done) 5284 break; 5285 5286 // Go to the next load command: 5287 offset = cmd_offset + load_cmd.cmdsize; 5288 } 5289 5290 if (start_address != LLDB_INVALID_ADDRESS) { 5291 // We got the start address from the load commands, so now resolve that 5292 // address in the sections 5293 // of this ObjectFile: 5294 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5295 start_address, GetSectionList())) { 5296 m_entry_point_address.Clear(); 5297 } 5298 } else { 5299 // We couldn't read the UnixThread load command - maybe it wasn't there. 5300 // As a fallback look for the 5301 // "start" symbol in the main executable. 5302 5303 ModuleSP module_sp(GetModule()); 5304 5305 if (module_sp) { 5306 SymbolContextList contexts; 5307 SymbolContext context; 5308 if (module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5309 eSymbolTypeCode, contexts)) { 5310 if (contexts.GetContextAtIndex(0, context)) 5311 m_entry_point_address = context.symbol->GetAddress(); 5312 } 5313 } 5314 } 5315 } 5316 5317 return m_entry_point_address; 5318 } 5319 5320 lldb_private::Address ObjectFileMachO::GetHeaderAddress() { 5321 lldb_private::Address header_addr; 5322 SectionList *section_list = GetSectionList(); 5323 if (section_list) { 5324 SectionSP text_segment_sp( 5325 section_list->FindSectionByName(GetSegmentNameTEXT())); 5326 if (text_segment_sp) { 5327 header_addr.SetSection(text_segment_sp); 5328 header_addr.SetOffset(0); 5329 } 5330 } 5331 return header_addr; 5332 } 5333 5334 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5335 ModuleSP module_sp(GetModule()); 5336 if (module_sp) { 5337 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5338 if (!m_thread_context_offsets_valid) { 5339 m_thread_context_offsets_valid = true; 5340 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5341 FileRangeArray::Entry file_range; 5342 thread_command thread_cmd; 5343 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5344 const uint32_t cmd_offset = offset; 5345 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 5346 break; 5347 5348 if (thread_cmd.cmd == LC_THREAD) { 5349 file_range.SetRangeBase(offset); 5350 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5351 m_thread_context_offsets.Append(file_range); 5352 } 5353 offset = cmd_offset + thread_cmd.cmdsize; 5354 } 5355 } 5356 } 5357 return m_thread_context_offsets.GetSize(); 5358 } 5359 5360 std::string ObjectFileMachO::GetIdentifierString() { 5361 std::string result; 5362 ModuleSP module_sp(GetModule()); 5363 if (module_sp) { 5364 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5365 5366 // First, look over the load commands for an LC_NOTE load command 5367 // with data_owner string "kern ver str" & use that if found. 5368 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5369 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5370 const uint32_t cmd_offset = offset; 5371 load_command lc; 5372 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5373 break; 5374 if (lc.cmd == LC_NOTE) 5375 { 5376 char data_owner[17]; 5377 m_data.CopyData (offset, 16, data_owner); 5378 data_owner[16] = '\0'; 5379 offset += 16; 5380 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5381 uint64_t size = m_data.GetU64_unchecked (&offset); 5382 5383 // "kern ver str" has a uint32_t version and then a 5384 // nul terminated c-string. 5385 if (strcmp ("kern ver str", data_owner) == 0) 5386 { 5387 offset = fileoff; 5388 uint32_t version; 5389 if (m_data.GetU32 (&offset, &version, 1) != nullptr) 5390 { 5391 if (version == 1) 5392 { 5393 uint32_t strsize = size - sizeof (uint32_t); 5394 char *buf = (char*) malloc (strsize); 5395 if (buf) 5396 { 5397 m_data.CopyData (offset, strsize, buf); 5398 buf[strsize - 1] = '\0'; 5399 result = buf; 5400 if (buf) 5401 free (buf); 5402 return result; 5403 } 5404 } 5405 } 5406 } 5407 } 5408 offset = cmd_offset + lc.cmdsize; 5409 } 5410 5411 // Second, make a pass over the load commands looking for an 5412 // obsolete LC_IDENT load command. 5413 offset = MachHeaderSizeFromMagic(m_header.magic); 5414 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5415 const uint32_t cmd_offset = offset; 5416 struct ident_command ident_command; 5417 if (m_data.GetU32(&offset, &ident_command, 2) == NULL) 5418 break; 5419 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5420 char *buf = (char *) malloc (ident_command.cmdsize); 5421 if (buf != nullptr 5422 && m_data.CopyData (offset, ident_command.cmdsize, buf) == ident_command.cmdsize) { 5423 buf[ident_command.cmdsize - 1] = '\0'; 5424 result = buf; 5425 } 5426 if (buf) 5427 free (buf); 5428 } 5429 offset = cmd_offset + ident_command.cmdsize; 5430 } 5431 5432 } 5433 return result; 5434 } 5435 5436 bool ObjectFileMachO::GetCorefileMainBinaryInfo (addr_t &address, UUID &uuid) { 5437 address = LLDB_INVALID_ADDRESS; 5438 uuid.Clear(); 5439 ModuleSP module_sp(GetModule()); 5440 if (module_sp) { 5441 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5442 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5443 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5444 const uint32_t cmd_offset = offset; 5445 load_command lc; 5446 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5447 break; 5448 if (lc.cmd == LC_NOTE) 5449 { 5450 char data_owner[17]; 5451 memset (data_owner, 0, sizeof (data_owner)); 5452 m_data.CopyData (offset, 16, data_owner); 5453 offset += 16; 5454 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5455 uint64_t size = m_data.GetU64_unchecked (&offset); 5456 5457 // "main bin spec" (main binary specification) data payload is formatted: 5458 // uint32_t version [currently 1] 5459 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user process] 5460 // uint64_t address [ UINT64_MAX if address not specified ] 5461 // uuid_t uuid [ all zero's if uuid not specified ] 5462 // uint32_t log2_pagesize [ process page size in log base 2, e.g. 4k pages are 12. 0 for unspecified ] 5463 5464 if (strcmp ("main bin spec", data_owner) == 0 && size >= 32) 5465 { 5466 offset = fileoff; 5467 uint32_t version; 5468 if (m_data.GetU32 (&offset, &version, 1) != nullptr && version == 1) 5469 { 5470 uint32_t type = 0; 5471 uuid_t raw_uuid; 5472 memset (raw_uuid, 0, sizeof (uuid_t)); 5473 5474 if (m_data.GetU32 (&offset, &type, 1) 5475 && m_data.GetU64 (&offset, &address, 1) 5476 && m_data.CopyData (offset, sizeof (uuid_t), raw_uuid) != 0 5477 && uuid.SetBytes (raw_uuid, sizeof (uuid_t))) 5478 { 5479 return true; 5480 } 5481 } 5482 } 5483 } 5484 offset = cmd_offset + lc.cmdsize; 5485 } 5486 } 5487 return false; 5488 } 5489 5490 lldb::RegisterContextSP 5491 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5492 lldb_private::Thread &thread) { 5493 lldb::RegisterContextSP reg_ctx_sp; 5494 5495 ModuleSP module_sp(GetModule()); 5496 if (module_sp) { 5497 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5498 if (!m_thread_context_offsets_valid) 5499 GetNumThreadContexts(); 5500 5501 const FileRangeArray::Entry *thread_context_file_range = 5502 m_thread_context_offsets.GetEntryAtIndex(idx); 5503 if (thread_context_file_range) { 5504 5505 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5506 thread_context_file_range->GetByteSize()); 5507 5508 switch (m_header.cputype) { 5509 case llvm::MachO::CPU_TYPE_ARM64: 5510 reg_ctx_sp.reset(new RegisterContextDarwin_arm64_Mach(thread, data)); 5511 break; 5512 5513 case llvm::MachO::CPU_TYPE_ARM: 5514 reg_ctx_sp.reset(new RegisterContextDarwin_arm_Mach(thread, data)); 5515 break; 5516 5517 case llvm::MachO::CPU_TYPE_I386: 5518 reg_ctx_sp.reset(new RegisterContextDarwin_i386_Mach(thread, data)); 5519 break; 5520 5521 case llvm::MachO::CPU_TYPE_X86_64: 5522 reg_ctx_sp.reset(new RegisterContextDarwin_x86_64_Mach(thread, data)); 5523 break; 5524 } 5525 } 5526 } 5527 return reg_ctx_sp; 5528 } 5529 5530 ObjectFile::Type ObjectFileMachO::CalculateType() { 5531 switch (m_header.filetype) { 5532 case MH_OBJECT: // 0x1u 5533 if (GetAddressByteSize() == 4) { 5534 // 32 bit kexts are just object files, but they do have a valid 5535 // UUID load command. 5536 UUID uuid; 5537 if (GetUUID(&uuid)) { 5538 // this checking for the UUID load command is not enough 5539 // we could eventually look for the symbol named 5540 // "OSKextGetCurrentIdentifier" as this is required of kexts 5541 if (m_strata == eStrataInvalid) 5542 m_strata = eStrataKernel; 5543 return eTypeSharedLibrary; 5544 } 5545 } 5546 return eTypeObjectFile; 5547 5548 case MH_EXECUTE: 5549 return eTypeExecutable; // 0x2u 5550 case MH_FVMLIB: 5551 return eTypeSharedLibrary; // 0x3u 5552 case MH_CORE: 5553 return eTypeCoreFile; // 0x4u 5554 case MH_PRELOAD: 5555 return eTypeSharedLibrary; // 0x5u 5556 case MH_DYLIB: 5557 return eTypeSharedLibrary; // 0x6u 5558 case MH_DYLINKER: 5559 return eTypeDynamicLinker; // 0x7u 5560 case MH_BUNDLE: 5561 return eTypeSharedLibrary; // 0x8u 5562 case MH_DYLIB_STUB: 5563 return eTypeStubLibrary; // 0x9u 5564 case MH_DSYM: 5565 return eTypeDebugInfo; // 0xAu 5566 case MH_KEXT_BUNDLE: 5567 return eTypeSharedLibrary; // 0xBu 5568 default: 5569 break; 5570 } 5571 return eTypeUnknown; 5572 } 5573 5574 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5575 switch (m_header.filetype) { 5576 case MH_OBJECT: // 0x1u 5577 { 5578 // 32 bit kexts are just object files, but they do have a valid 5579 // UUID load command. 5580 UUID uuid; 5581 if (GetUUID(&uuid)) { 5582 // this checking for the UUID load command is not enough 5583 // we could eventually look for the symbol named 5584 // "OSKextGetCurrentIdentifier" as this is required of kexts 5585 if (m_type == eTypeInvalid) 5586 m_type = eTypeSharedLibrary; 5587 5588 return eStrataKernel; 5589 } 5590 } 5591 return eStrataUnknown; 5592 5593 case MH_EXECUTE: // 0x2u 5594 // Check for the MH_DYLDLINK bit in the flags 5595 if (m_header.flags & MH_DYLDLINK) { 5596 return eStrataUser; 5597 } else { 5598 SectionList *section_list = GetSectionList(); 5599 if (section_list) { 5600 static ConstString g_kld_section_name("__KLD"); 5601 if (section_list->FindSectionByName(g_kld_section_name)) 5602 return eStrataKernel; 5603 } 5604 } 5605 return eStrataRawImage; 5606 5607 case MH_FVMLIB: 5608 return eStrataUser; // 0x3u 5609 case MH_CORE: 5610 return eStrataUnknown; // 0x4u 5611 case MH_PRELOAD: 5612 return eStrataRawImage; // 0x5u 5613 case MH_DYLIB: 5614 return eStrataUser; // 0x6u 5615 case MH_DYLINKER: 5616 return eStrataUser; // 0x7u 5617 case MH_BUNDLE: 5618 return eStrataUser; // 0x8u 5619 case MH_DYLIB_STUB: 5620 return eStrataUser; // 0x9u 5621 case MH_DSYM: 5622 return eStrataUnknown; // 0xAu 5623 case MH_KEXT_BUNDLE: 5624 return eStrataKernel; // 0xBu 5625 default: 5626 break; 5627 } 5628 return eStrataUnknown; 5629 } 5630 5631 uint32_t ObjectFileMachO::GetVersion(uint32_t *versions, 5632 uint32_t num_versions) { 5633 ModuleSP module_sp(GetModule()); 5634 if (module_sp) { 5635 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5636 struct dylib_command load_cmd; 5637 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5638 uint32_t version_cmd = 0; 5639 uint64_t version = 0; 5640 uint32_t i; 5641 for (i = 0; i < m_header.ncmds; ++i) { 5642 const lldb::offset_t cmd_offset = offset; 5643 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5644 break; 5645 5646 if (load_cmd.cmd == LC_ID_DYLIB) { 5647 if (version_cmd == 0) { 5648 version_cmd = load_cmd.cmd; 5649 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5650 break; 5651 version = load_cmd.dylib.current_version; 5652 } 5653 break; // Break for now unless there is another more complete version 5654 // number load command in the future. 5655 } 5656 offset = cmd_offset + load_cmd.cmdsize; 5657 } 5658 5659 if (version_cmd == LC_ID_DYLIB) { 5660 if (versions != NULL && num_versions > 0) { 5661 if (num_versions > 0) 5662 versions[0] = (version & 0xFFFF0000ull) >> 16; 5663 if (num_versions > 1) 5664 versions[1] = (version & 0x0000FF00ull) >> 8; 5665 if (num_versions > 2) 5666 versions[2] = (version & 0x000000FFull); 5667 // Fill in an remaining version numbers with invalid values 5668 for (i = 3; i < num_versions; ++i) 5669 versions[i] = UINT32_MAX; 5670 } 5671 // The LC_ID_DYLIB load command has a version with 3 version numbers 5672 // in it, so always return 3 5673 return 3; 5674 } 5675 } 5676 return false; 5677 } 5678 5679 bool ObjectFileMachO::GetArchitecture(ArchSpec &arch) { 5680 ModuleSP module_sp(GetModule()); 5681 if (module_sp) { 5682 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5683 return GetArchitecture(m_header, m_data, 5684 MachHeaderSizeFromMagic(m_header.magic), arch); 5685 } 5686 return false; 5687 } 5688 5689 UUID ObjectFileMachO::GetProcessSharedCacheUUID(Process *process) { 5690 UUID uuid; 5691 if (process && process->GetDynamicLoader()) { 5692 DynamicLoader *dl = process->GetDynamicLoader(); 5693 addr_t load_address; 5694 LazyBool using_shared_cache; 5695 LazyBool private_shared_cache; 5696 dl->GetSharedCacheInformation(load_address, uuid, using_shared_cache, 5697 private_shared_cache); 5698 } 5699 return uuid; 5700 } 5701 5702 UUID ObjectFileMachO::GetLLDBSharedCacheUUID() { 5703 UUID uuid; 5704 #if defined(__APPLE__) && \ 5705 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5706 uint8_t *(*dyld_get_all_image_infos)(void); 5707 dyld_get_all_image_infos = 5708 (uint8_t * (*)())dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5709 if (dyld_get_all_image_infos) { 5710 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5711 if (dyld_all_image_infos_address) { 5712 uint32_t *version = (uint32_t *) 5713 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5714 if (*version >= 13) { 5715 uuid_t *sharedCacheUUID_address = 0; 5716 int wordsize = sizeof(uint8_t *); 5717 if (wordsize == 8) { 5718 sharedCacheUUID_address = 5719 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5720 160); // sharedCacheUUID <mach-o/dyld_images.h> 5721 } else { 5722 sharedCacheUUID_address = 5723 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5724 84); // sharedCacheUUID <mach-o/dyld_images.h> 5725 } 5726 uuid.SetBytes(sharedCacheUUID_address); 5727 } 5728 } 5729 } 5730 #endif 5731 return uuid; 5732 } 5733 5734 uint32_t ObjectFileMachO::GetMinimumOSVersion(uint32_t *versions, 5735 uint32_t num_versions) { 5736 if (m_min_os_versions.empty()) { 5737 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5738 bool success = false; 5739 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5740 const lldb::offset_t load_cmd_offset = offset; 5741 5742 version_min_command lc; 5743 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5744 break; 5745 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5746 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5747 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5748 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5749 if (m_data.GetU32(&offset, &lc.version, 5750 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5751 const uint32_t xxxx = lc.version >> 16; 5752 const uint32_t yy = (lc.version >> 8) & 0xffu; 5753 const uint32_t zz = lc.version & 0xffu; 5754 if (xxxx) { 5755 m_min_os_versions.push_back(xxxx); 5756 m_min_os_versions.push_back(yy); 5757 m_min_os_versions.push_back(zz); 5758 } 5759 success = true; 5760 } 5761 } 5762 offset = load_cmd_offset + lc.cmdsize; 5763 } 5764 5765 if (success == false) { 5766 // Push an invalid value so we don't keep trying to 5767 m_min_os_versions.push_back(UINT32_MAX); 5768 } 5769 } 5770 5771 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) { 5772 if (versions != NULL && num_versions > 0) { 5773 for (size_t i = 0; i < num_versions; ++i) { 5774 if (i < m_min_os_versions.size()) 5775 versions[i] = m_min_os_versions[i]; 5776 else 5777 versions[i] = 0; 5778 } 5779 } 5780 return m_min_os_versions.size(); 5781 } 5782 // Call the superclasses version that will empty out the data 5783 return ObjectFile::GetMinimumOSVersion(versions, num_versions); 5784 } 5785 5786 uint32_t ObjectFileMachO::GetSDKVersion(uint32_t *versions, 5787 uint32_t num_versions) { 5788 if (m_sdk_versions.empty()) { 5789 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5790 bool success = false; 5791 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5792 const lldb::offset_t load_cmd_offset = offset; 5793 5794 version_min_command lc; 5795 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5796 break; 5797 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5798 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5799 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5800 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5801 if (m_data.GetU32(&offset, &lc.version, 5802 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5803 const uint32_t xxxx = lc.sdk >> 16; 5804 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5805 const uint32_t zz = lc.sdk & 0xffu; 5806 if (xxxx) { 5807 m_sdk_versions.push_back(xxxx); 5808 m_sdk_versions.push_back(yy); 5809 m_sdk_versions.push_back(zz); 5810 } 5811 success = true; 5812 } 5813 } 5814 offset = load_cmd_offset + lc.cmdsize; 5815 } 5816 5817 if (success == false) { 5818 // Push an invalid value so we don't keep trying to 5819 m_sdk_versions.push_back(UINT32_MAX); 5820 } 5821 } 5822 5823 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) { 5824 if (versions != NULL && num_versions > 0) { 5825 for (size_t i = 0; i < num_versions; ++i) { 5826 if (i < m_sdk_versions.size()) 5827 versions[i] = m_sdk_versions[i]; 5828 else 5829 versions[i] = 0; 5830 } 5831 } 5832 return m_sdk_versions.size(); 5833 } 5834 // Call the superclasses version that will empty out the data 5835 return ObjectFile::GetSDKVersion(versions, num_versions); 5836 } 5837 5838 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5839 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5840 } 5841 5842 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5843 return m_allow_assembly_emulation_unwind_plans; 5844 } 5845 5846 //------------------------------------------------------------------ 5847 // PluginInterface protocol 5848 //------------------------------------------------------------------ 5849 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5850 return GetPluginNameStatic(); 5851 } 5852 5853 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5854 5855 Section *ObjectFileMachO::GetMachHeaderSection() { 5856 // Find the first address of the mach header which is the first non-zero 5857 // file sized section whose file offset is zero. This is the base file address 5858 // of the mach-o file which can be subtracted from the vmaddr of the other 5859 // segments found in memory and added to the load address 5860 ModuleSP module_sp = GetModule(); 5861 if (module_sp) { 5862 SectionList *section_list = GetSectionList(); 5863 if (section_list) { 5864 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5865 const size_t num_sections = section_list->GetSize(); 5866 5867 for (size_t sect_idx = 0; sect_idx < num_sections && 5868 mach_base_file_addr == LLDB_INVALID_ADDRESS; 5869 ++sect_idx) { 5870 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5871 if (section && section->GetFileSize() > 0 && 5872 section->GetFileOffset() == 0 && 5873 section->IsThreadSpecific() == false && 5874 module_sp.get() == section->GetModule().get()) { 5875 return section; 5876 } 5877 } 5878 } 5879 } 5880 return nullptr; 5881 } 5882 5883 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 5884 lldb::addr_t mach_header_load_address, const Section *mach_header_section, 5885 const Section *section) { 5886 ModuleSP module_sp = GetModule(); 5887 if (module_sp && mach_header_section && section && 5888 mach_header_load_address != LLDB_INVALID_ADDRESS) { 5889 lldb::addr_t mach_header_file_addr = mach_header_section->GetFileAddress(); 5890 if (mach_header_file_addr != LLDB_INVALID_ADDRESS) { 5891 if (section && section->GetFileSize() > 0 && 5892 section->IsThreadSpecific() == false && 5893 module_sp.get() == section->GetModule().get()) { 5894 // Ignore __LINKEDIT and __DWARF segments 5895 if (section->GetName() == GetSegmentNameLINKEDIT()) { 5896 // Only map __LINKEDIT if we have an in memory image and this isn't 5897 // a kernel binary like a kext or mach_kernel. 5898 const bool is_memory_image = (bool)m_process_wp.lock(); 5899 const Strata strata = GetStrata(); 5900 if (is_memory_image == false || strata == eStrataKernel) 5901 return LLDB_INVALID_ADDRESS; 5902 } 5903 return section->GetFileAddress() - mach_header_file_addr + 5904 mach_header_load_address; 5905 } 5906 } 5907 } 5908 return LLDB_INVALID_ADDRESS; 5909 } 5910 5911 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 5912 bool value_is_offset) { 5913 ModuleSP module_sp = GetModule(); 5914 if (module_sp) { 5915 size_t num_loaded_sections = 0; 5916 SectionList *section_list = GetSectionList(); 5917 if (section_list) { 5918 const size_t num_sections = section_list->GetSize(); 5919 5920 if (value_is_offset) { 5921 // "value" is an offset to apply to each top level segment 5922 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5923 // Iterate through the object file sections to find all 5924 // of the sections that size on disk (to avoid __PAGEZERO) 5925 // and load them 5926 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5927 if (section_sp && section_sp->GetFileSize() > 0 && 5928 section_sp->IsThreadSpecific() == false && 5929 module_sp.get() == section_sp->GetModule().get()) { 5930 // Ignore __LINKEDIT and __DWARF segments 5931 if (section_sp->GetName() == GetSegmentNameLINKEDIT()) { 5932 // Only map __LINKEDIT if we have an in memory image and this 5933 // isn't 5934 // a kernel binary like a kext or mach_kernel. 5935 const bool is_memory_image = (bool)m_process_wp.lock(); 5936 const Strata strata = GetStrata(); 5937 if (is_memory_image == false || strata == eStrataKernel) 5938 continue; 5939 } 5940 if (target.GetSectionLoadList().SetSectionLoadAddress( 5941 section_sp, section_sp->GetFileAddress() + value)) 5942 ++num_loaded_sections; 5943 } 5944 } 5945 } else { 5946 // "value" is the new base address of the mach_header, adjust each 5947 // section accordingly 5948 5949 Section *mach_header_section = GetMachHeaderSection(); 5950 if (mach_header_section) { 5951 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5952 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5953 5954 lldb::addr_t section_load_addr = 5955 CalculateSectionLoadAddressForMemoryImage( 5956 value, mach_header_section, section_sp.get()); 5957 if (section_load_addr != LLDB_INVALID_ADDRESS) { 5958 if (target.GetSectionLoadList().SetSectionLoadAddress( 5959 section_sp, section_load_addr)) 5960 ++num_loaded_sections; 5961 } 5962 } 5963 } 5964 } 5965 } 5966 return num_loaded_sections > 0; 5967 } 5968 return false; 5969 } 5970 5971 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 5972 const FileSpec &outfile, Status &error) { 5973 if (process_sp) { 5974 Target &target = process_sp->GetTarget(); 5975 const ArchSpec target_arch = target.GetArchitecture(); 5976 const llvm::Triple &target_triple = target_arch.GetTriple(); 5977 if (target_triple.getVendor() == llvm::Triple::Apple && 5978 (target_triple.getOS() == llvm::Triple::MacOSX || 5979 target_triple.getOS() == llvm::Triple::IOS || 5980 target_triple.getOS() == llvm::Triple::WatchOS || 5981 target_triple.getOS() == llvm::Triple::TvOS)) { 5982 bool make_core = false; 5983 switch (target_arch.GetMachine()) { 5984 case llvm::Triple::aarch64: 5985 case llvm::Triple::arm: 5986 case llvm::Triple::thumb: 5987 case llvm::Triple::x86: 5988 case llvm::Triple::x86_64: 5989 make_core = true; 5990 break; 5991 default: 5992 error.SetErrorStringWithFormat("unsupported core architecture: %s", 5993 target_triple.str().c_str()); 5994 break; 5995 } 5996 5997 if (make_core) { 5998 std::vector<segment_command_64> segment_load_commands; 5999 // uint32_t range_info_idx = 0; 6000 MemoryRegionInfo range_info; 6001 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6002 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6003 const ByteOrder byte_order = target_arch.GetByteOrder(); 6004 if (range_error.Success()) { 6005 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6006 const addr_t addr = range_info.GetRange().GetRangeBase(); 6007 const addr_t size = range_info.GetRange().GetByteSize(); 6008 6009 if (size == 0) 6010 break; 6011 6012 // Calculate correct protections 6013 uint32_t prot = 0; 6014 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6015 prot |= VM_PROT_READ; 6016 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6017 prot |= VM_PROT_WRITE; 6018 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6019 prot |= VM_PROT_EXECUTE; 6020 6021 // printf ("[%3u] [0x%16.16" PRIx64 " - 6022 // 0x%16.16" PRIx64 ") %c%c%c\n", 6023 // range_info_idx, 6024 // addr, 6025 // size, 6026 // (prot & VM_PROT_READ ) ? 'r' : 6027 // '-', 6028 // (prot & VM_PROT_WRITE ) ? 'w' : 6029 // '-', 6030 // (prot & VM_PROT_EXECUTE) ? 'x' : 6031 // '-'); 6032 6033 if (prot != 0) { 6034 uint32_t cmd_type = LC_SEGMENT_64; 6035 uint32_t segment_size = sizeof(segment_command_64); 6036 if (addr_byte_size == 4) { 6037 cmd_type = LC_SEGMENT; 6038 segment_size = sizeof(segment_command); 6039 } 6040 segment_command_64 segment = { 6041 cmd_type, // uint32_t cmd; 6042 segment_size, // uint32_t cmdsize; 6043 {0}, // char segname[16]; 6044 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6045 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6046 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6047 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6048 prot, // uint32_t maxprot; 6049 prot, // uint32_t initprot; 6050 0, // uint32_t nsects; 6051 0}; // uint32_t flags; 6052 segment_load_commands.push_back(segment); 6053 } else { 6054 // No protections and a size of 1 used to be returned from old 6055 // debugservers when we asked about a region that was past the 6056 // last memory region and it indicates the end... 6057 if (size == 1) 6058 break; 6059 } 6060 6061 range_error = process_sp->GetMemoryRegionInfo( 6062 range_info.GetRange().GetRangeEnd(), range_info); 6063 if (range_error.Fail()) 6064 break; 6065 } 6066 6067 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6068 6069 mach_header_64 mach_header; 6070 if (addr_byte_size == 8) { 6071 mach_header.magic = MH_MAGIC_64; 6072 } else { 6073 mach_header.magic = MH_MAGIC; 6074 } 6075 mach_header.cputype = target_arch.GetMachOCPUType(); 6076 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6077 mach_header.filetype = MH_CORE; 6078 mach_header.ncmds = segment_load_commands.size(); 6079 mach_header.flags = 0; 6080 mach_header.reserved = 0; 6081 ThreadList &thread_list = process_sp->GetThreadList(); 6082 const uint32_t num_threads = thread_list.GetSize(); 6083 6084 // Make an array of LC_THREAD data items. Each one contains 6085 // the contents of the LC_THREAD load command. The data doesn't 6086 // contain the load command + load command size, we will 6087 // add the load command and load command size as we emit the data. 6088 std::vector<StreamString> LC_THREAD_datas(num_threads); 6089 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6090 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6091 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6092 LC_THREAD_data.SetByteOrder(byte_order); 6093 } 6094 for (uint32_t thread_idx = 0; thread_idx < num_threads; 6095 ++thread_idx) { 6096 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6097 if (thread_sp) { 6098 switch (mach_header.cputype) { 6099 case llvm::MachO::CPU_TYPE_ARM64: 6100 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6101 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6102 break; 6103 6104 case llvm::MachO::CPU_TYPE_ARM: 6105 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6106 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6107 break; 6108 6109 case llvm::MachO::CPU_TYPE_I386: 6110 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6111 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6112 break; 6113 6114 case llvm::MachO::CPU_TYPE_X86_64: 6115 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6116 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6117 break; 6118 } 6119 } 6120 } 6121 6122 // The size of the load command is the size of the segments... 6123 if (addr_byte_size == 8) { 6124 mach_header.sizeofcmds = segment_load_commands.size() * 6125 sizeof(struct segment_command_64); 6126 } else { 6127 mach_header.sizeofcmds = 6128 segment_load_commands.size() * sizeof(struct segment_command); 6129 } 6130 6131 // and the size of all LC_THREAD load command 6132 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6133 ++mach_header.ncmds; 6134 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6135 } 6136 6137 printf("mach_header: 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x " 6138 "0x%8.8x 0x%8.8x\n", 6139 mach_header.magic, mach_header.cputype, mach_header.cpusubtype, 6140 mach_header.filetype, mach_header.ncmds, 6141 mach_header.sizeofcmds, mach_header.flags, 6142 mach_header.reserved); 6143 6144 // Write the mach header 6145 buffer.PutHex32(mach_header.magic); 6146 buffer.PutHex32(mach_header.cputype); 6147 buffer.PutHex32(mach_header.cpusubtype); 6148 buffer.PutHex32(mach_header.filetype); 6149 buffer.PutHex32(mach_header.ncmds); 6150 buffer.PutHex32(mach_header.sizeofcmds); 6151 buffer.PutHex32(mach_header.flags); 6152 if (addr_byte_size == 8) { 6153 buffer.PutHex32(mach_header.reserved); 6154 } 6155 6156 // Skip the mach header and all load commands and align to the next 6157 // 0x1000 byte boundary 6158 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6159 if (file_offset & 0x00000fff) { 6160 file_offset += 0x00001000ull; 6161 file_offset &= (~0x00001000ull + 1); 6162 } 6163 6164 for (auto &segment : segment_load_commands) { 6165 segment.fileoff = file_offset; 6166 file_offset += segment.filesize; 6167 } 6168 6169 // Write out all of the LC_THREAD load commands 6170 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6171 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6172 buffer.PutHex32(LC_THREAD); 6173 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6174 buffer.Write(LC_THREAD_data.GetString().data(), 6175 LC_THREAD_data_size); 6176 } 6177 6178 // Write out all of the segment load commands 6179 for (const auto &segment : segment_load_commands) { 6180 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6181 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6182 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6183 segment.cmd, segment.cmdsize, segment.vmaddr, 6184 segment.vmaddr + segment.vmsize, segment.fileoff, 6185 segment.filesize, segment.maxprot, segment.initprot, 6186 segment.nsects, segment.flags); 6187 6188 buffer.PutHex32(segment.cmd); 6189 buffer.PutHex32(segment.cmdsize); 6190 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6191 if (addr_byte_size == 8) { 6192 buffer.PutHex64(segment.vmaddr); 6193 buffer.PutHex64(segment.vmsize); 6194 buffer.PutHex64(segment.fileoff); 6195 buffer.PutHex64(segment.filesize); 6196 } else { 6197 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6198 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6199 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6200 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6201 } 6202 buffer.PutHex32(segment.maxprot); 6203 buffer.PutHex32(segment.initprot); 6204 buffer.PutHex32(segment.nsects); 6205 buffer.PutHex32(segment.flags); 6206 } 6207 6208 File core_file; 6209 std::string core_file_path(outfile.GetPath()); 6210 error = core_file.Open(core_file_path.c_str(), 6211 File::eOpenOptionWrite | 6212 File::eOpenOptionTruncate | 6213 File::eOpenOptionCanCreate); 6214 if (error.Success()) { 6215 // Read 1 page at a time 6216 uint8_t bytes[0x1000]; 6217 // Write the mach header and load commands out to the core file 6218 size_t bytes_written = buffer.GetString().size(); 6219 error = core_file.Write(buffer.GetString().data(), bytes_written); 6220 if (error.Success()) { 6221 // Now write the file data for all memory segments in the process 6222 for (const auto &segment : segment_load_commands) { 6223 if (core_file.SeekFromStart(segment.fileoff) == -1) { 6224 error.SetErrorStringWithFormat( 6225 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6226 segment.fileoff, core_file_path.c_str()); 6227 break; 6228 } 6229 6230 printf("Saving %" PRId64 6231 " bytes of data for memory region at 0x%" PRIx64 "\n", 6232 segment.vmsize, segment.vmaddr); 6233 addr_t bytes_left = segment.vmsize; 6234 addr_t addr = segment.vmaddr; 6235 Status memory_read_error; 6236 while (bytes_left > 0 && error.Success()) { 6237 const size_t bytes_to_read = 6238 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6239 const size_t bytes_read = process_sp->ReadMemory( 6240 addr, bytes, bytes_to_read, memory_read_error); 6241 if (bytes_read == bytes_to_read) { 6242 size_t bytes_written = bytes_read; 6243 error = core_file.Write(bytes, bytes_written); 6244 bytes_left -= bytes_read; 6245 addr += bytes_read; 6246 } else { 6247 // Some pages within regions are not readable, those 6248 // should be zero filled 6249 memset(bytes, 0, bytes_to_read); 6250 size_t bytes_written = bytes_to_read; 6251 error = core_file.Write(bytes, bytes_written); 6252 bytes_left -= bytes_to_read; 6253 addr += bytes_to_read; 6254 } 6255 } 6256 } 6257 } 6258 } 6259 } else { 6260 error.SetErrorString( 6261 "process doesn't support getting memory region info"); 6262 } 6263 } 6264 return true; // This is the right plug to handle saving core files for 6265 // this process 6266 } 6267 } 6268 return false; 6269 } 6270