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