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 #include "llvm/ADT/StringRef.h" 11 12 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 13 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 14 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 15 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 16 #include "lldb/Core/Debugger.h" 17 #include "lldb/Core/FileSpecList.h" 18 #include "lldb/Core/Module.h" 19 #include "lldb/Core/ModuleSpec.h" 20 #include "lldb/Core/PluginManager.h" 21 #include "lldb/Core/RangeMap.h" 22 #include "lldb/Core/Section.h" 23 #include "lldb/Core/StreamFile.h" 24 #include "lldb/Host/Host.h" 25 #include "lldb/Symbol/DWARFCallFrameInfo.h" 26 #include "lldb/Symbol/ObjectFile.h" 27 #include "lldb/Target/DynamicLoader.h" 28 #include "lldb/Target/MemoryRegionInfo.h" 29 #include "lldb/Target/Platform.h" 30 #include "lldb/Target/Process.h" 31 #include "lldb/Target/SectionLoadList.h" 32 #include "lldb/Target/Target.h" 33 #include "lldb/Target/Thread.h" 34 #include "lldb/Target/ThreadList.h" 35 #include "lldb/Utility/ArchSpec.h" 36 #include "lldb/Utility/DataBuffer.h" 37 #include "lldb/Utility/FileSpec.h" 38 #include "lldb/Utility/Log.h" 39 #include "lldb/Utility/RegisterValue.h" 40 #include "lldb/Utility/Status.h" 41 #include "lldb/Utility/StreamString.h" 42 #include "lldb/Utility/Timer.h" 43 #include "lldb/Utility/UUID.h" 44 45 #include "lldb/Host/SafeMachO.h" 46 47 #include "llvm/Support/MemoryBuffer.h" 48 49 #include "ObjectFileMachO.h" 50 51 #if defined(__APPLE__) && \ 52 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 53 // GetLLDBSharedCacheUUID() needs to call dlsym() 54 #include <dlfcn.h> 55 #endif 56 57 #ifndef __APPLE__ 58 #include "Utility/UuidCompatibility.h" 59 #else 60 #include <uuid/uuid.h> 61 #endif 62 63 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 64 using namespace lldb; 65 using namespace lldb_private; 66 using namespace llvm::MachO; 67 68 // Some structure definitions needed for parsing the dyld shared cache files 69 // found on iOS devices. 70 71 struct lldb_copy_dyld_cache_header_v1 { 72 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 73 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 74 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 75 uint32_t imagesOffset; 76 uint32_t imagesCount; 77 uint64_t dyldBaseAddress; 78 uint64_t codeSignatureOffset; 79 uint64_t codeSignatureSize; 80 uint64_t slideInfoOffset; 81 uint64_t slideInfoSize; 82 uint64_t localSymbolsOffset; 83 uint64_t localSymbolsSize; 84 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 85 // and later 86 }; 87 88 struct lldb_copy_dyld_cache_mapping_info { 89 uint64_t address; 90 uint64_t size; 91 uint64_t fileOffset; 92 uint32_t maxProt; 93 uint32_t initProt; 94 }; 95 96 struct lldb_copy_dyld_cache_local_symbols_info { 97 uint32_t nlistOffset; 98 uint32_t nlistCount; 99 uint32_t stringsOffset; 100 uint32_t stringsSize; 101 uint32_t entriesOffset; 102 uint32_t entriesCount; 103 }; 104 struct lldb_copy_dyld_cache_local_symbols_entry { 105 uint32_t dylibOffset; 106 uint32_t nlistStartIndex; 107 uint32_t nlistCount; 108 }; 109 110 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 111 public: 112 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 113 const DataExtractor &data) 114 : RegisterContextDarwin_x86_64(thread, 0) { 115 SetRegisterDataFrom_LC_THREAD(data); 116 } 117 118 void InvalidateAllRegisters() override { 119 // Do nothing... registers are always valid... 120 } 121 122 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 123 lldb::offset_t offset = 0; 124 SetError(GPRRegSet, Read, -1); 125 SetError(FPURegSet, Read, -1); 126 SetError(EXCRegSet, Read, -1); 127 bool done = false; 128 129 while (!done) { 130 int flavor = data.GetU32(&offset); 131 if (flavor == 0) 132 done = true; 133 else { 134 uint32_t i; 135 uint32_t count = data.GetU32(&offset); 136 switch (flavor) { 137 case GPRRegSet: 138 for (i = 0; i < count; ++i) 139 (&gpr.rax)[i] = data.GetU64(&offset); 140 SetError(GPRRegSet, Read, 0); 141 done = true; 142 143 break; 144 case FPURegSet: 145 // TODO: fill in FPU regs.... 146 // SetError (FPURegSet, Read, -1); 147 done = true; 148 149 break; 150 case EXCRegSet: 151 exc.trapno = data.GetU32(&offset); 152 exc.err = data.GetU32(&offset); 153 exc.faultvaddr = data.GetU64(&offset); 154 SetError(EXCRegSet, Read, 0); 155 done = true; 156 break; 157 case 7: 158 case 8: 159 case 9: 160 // fancy flavors that encapsulate of the above flavors... 161 break; 162 163 default: 164 done = true; 165 break; 166 } 167 } 168 } 169 } 170 171 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 172 const char *alt_name, size_t reg_byte_size, 173 Stream &data) { 174 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 175 if (reg_info == NULL) 176 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 177 if (reg_info) { 178 lldb_private::RegisterValue reg_value; 179 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 180 if (reg_info->byte_size >= reg_byte_size) 181 data.Write(reg_value.GetBytes(), reg_byte_size); 182 else { 183 data.Write(reg_value.GetBytes(), reg_info->byte_size); 184 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 185 ++i) 186 data.PutChar(0); 187 } 188 return reg_byte_size; 189 } 190 } 191 // Just write zeros if all else fails 192 for (size_t i = 0; i < reg_byte_size; ++i) 193 data.PutChar(0); 194 return reg_byte_size; 195 } 196 197 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 198 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 199 if (reg_ctx_sp) { 200 RegisterContext *reg_ctx = reg_ctx_sp.get(); 201 202 data.PutHex32(GPRRegSet); // Flavor 203 data.PutHex32(GPRWordCount); 204 WriteRegister(reg_ctx, "rax", NULL, 8, data); 205 WriteRegister(reg_ctx, "rbx", NULL, 8, data); 206 WriteRegister(reg_ctx, "rcx", NULL, 8, data); 207 WriteRegister(reg_ctx, "rdx", NULL, 8, data); 208 WriteRegister(reg_ctx, "rdi", NULL, 8, data); 209 WriteRegister(reg_ctx, "rsi", NULL, 8, data); 210 WriteRegister(reg_ctx, "rbp", NULL, 8, data); 211 WriteRegister(reg_ctx, "rsp", NULL, 8, data); 212 WriteRegister(reg_ctx, "r8", NULL, 8, data); 213 WriteRegister(reg_ctx, "r9", NULL, 8, data); 214 WriteRegister(reg_ctx, "r10", NULL, 8, data); 215 WriteRegister(reg_ctx, "r11", NULL, 8, data); 216 WriteRegister(reg_ctx, "r12", NULL, 8, data); 217 WriteRegister(reg_ctx, "r13", NULL, 8, data); 218 WriteRegister(reg_ctx, "r14", NULL, 8, data); 219 WriteRegister(reg_ctx, "r15", NULL, 8, data); 220 WriteRegister(reg_ctx, "rip", NULL, 8, data); 221 WriteRegister(reg_ctx, "rflags", NULL, 8, data); 222 WriteRegister(reg_ctx, "cs", NULL, 8, data); 223 WriteRegister(reg_ctx, "fs", NULL, 8, data); 224 WriteRegister(reg_ctx, "gs", NULL, 8, data); 225 226 // // Write out the FPU registers 227 // const size_t fpu_byte_size = sizeof(FPU); 228 // size_t bytes_written = 0; 229 // data.PutHex32 (FPURegSet); 230 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 231 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 232 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 233 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 234 // data); // uint16_t fcw; // "fctrl" 235 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 236 // data); // uint16_t fsw; // "fstat" 237 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 238 // data); // uint8_t ftw; // "ftag" 239 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 240 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 241 // data); // uint16_t fop; // "fop" 242 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 243 // data); // uint32_t ip; // "fioff" 244 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 245 // data); // uint16_t cs; // "fiseg" 246 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 247 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 248 // data); // uint32_t dp; // "fooff" 249 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 250 // data); // uint16_t ds; // "foseg" 251 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 252 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 253 // data); // uint32_t mxcsr; 254 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 255 // 4, data);// uint32_t mxcsrmask; 256 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 257 // sizeof(MMSReg), data); 258 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 259 // sizeof(MMSReg), data); 260 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 261 // sizeof(MMSReg), data); 262 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 263 // sizeof(MMSReg), data); 264 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 265 // sizeof(MMSReg), data); 266 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 267 // sizeof(MMSReg), data); 268 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 269 // sizeof(MMSReg), data); 270 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 271 // sizeof(MMSReg), data); 272 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 273 // sizeof(XMMReg), data); 274 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 275 // sizeof(XMMReg), data); 276 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 277 // sizeof(XMMReg), data); 278 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 279 // sizeof(XMMReg), data); 280 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 281 // sizeof(XMMReg), data); 282 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 283 // sizeof(XMMReg), data); 284 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 285 // sizeof(XMMReg), data); 286 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 287 // sizeof(XMMReg), data); 288 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 289 // sizeof(XMMReg), data); 290 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 291 // sizeof(XMMReg), data); 292 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 293 // sizeof(XMMReg), data); 294 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 295 // sizeof(XMMReg), data); 296 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 297 // sizeof(XMMReg), data); 298 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 299 // sizeof(XMMReg), data); 300 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 301 // sizeof(XMMReg), data); 302 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 303 // sizeof(XMMReg), data); 304 // 305 // // Fill rest with zeros 306 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 307 // i) 308 // data.PutChar(0); 309 310 // Write out the EXC registers 311 data.PutHex32(EXCRegSet); 312 data.PutHex32(EXCWordCount); 313 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 314 WriteRegister(reg_ctx, "err", NULL, 4, data); 315 WriteRegister(reg_ctx, "faultvaddr", NULL, 8, data); 316 return true; 317 } 318 return false; 319 } 320 321 protected: 322 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 323 324 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 325 326 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 327 328 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 329 return 0; 330 } 331 332 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 333 return 0; 334 } 335 336 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 337 return 0; 338 } 339 }; 340 341 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 342 public: 343 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 344 const DataExtractor &data) 345 : RegisterContextDarwin_i386(thread, 0) { 346 SetRegisterDataFrom_LC_THREAD(data); 347 } 348 349 void InvalidateAllRegisters() override { 350 // Do nothing... registers are always valid... 351 } 352 353 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 354 lldb::offset_t offset = 0; 355 SetError(GPRRegSet, Read, -1); 356 SetError(FPURegSet, Read, -1); 357 SetError(EXCRegSet, Read, -1); 358 bool done = false; 359 360 while (!done) { 361 int flavor = data.GetU32(&offset); 362 if (flavor == 0) 363 done = true; 364 else { 365 uint32_t i; 366 uint32_t count = data.GetU32(&offset); 367 switch (flavor) { 368 case GPRRegSet: 369 for (i = 0; i < count; ++i) 370 (&gpr.eax)[i] = data.GetU32(&offset); 371 SetError(GPRRegSet, Read, 0); 372 done = true; 373 374 break; 375 case FPURegSet: 376 // TODO: fill in FPU regs.... 377 // SetError (FPURegSet, Read, -1); 378 done = true; 379 380 break; 381 case EXCRegSet: 382 exc.trapno = data.GetU32(&offset); 383 exc.err = data.GetU32(&offset); 384 exc.faultvaddr = data.GetU32(&offset); 385 SetError(EXCRegSet, Read, 0); 386 done = true; 387 break; 388 case 7: 389 case 8: 390 case 9: 391 // fancy flavors that encapsulate of the above flavors... 392 break; 393 394 default: 395 done = true; 396 break; 397 } 398 } 399 } 400 } 401 402 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 403 const char *alt_name, size_t reg_byte_size, 404 Stream &data) { 405 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 406 if (reg_info == NULL) 407 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 408 if (reg_info) { 409 lldb_private::RegisterValue reg_value; 410 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 411 if (reg_info->byte_size >= reg_byte_size) 412 data.Write(reg_value.GetBytes(), reg_byte_size); 413 else { 414 data.Write(reg_value.GetBytes(), reg_info->byte_size); 415 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 416 ++i) 417 data.PutChar(0); 418 } 419 return reg_byte_size; 420 } 421 } 422 // Just write zeros if all else fails 423 for (size_t i = 0; i < reg_byte_size; ++i) 424 data.PutChar(0); 425 return reg_byte_size; 426 } 427 428 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 429 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 430 if (reg_ctx_sp) { 431 RegisterContext *reg_ctx = reg_ctx_sp.get(); 432 433 data.PutHex32(GPRRegSet); // Flavor 434 data.PutHex32(GPRWordCount); 435 WriteRegister(reg_ctx, "eax", NULL, 4, data); 436 WriteRegister(reg_ctx, "ebx", NULL, 4, data); 437 WriteRegister(reg_ctx, "ecx", NULL, 4, data); 438 WriteRegister(reg_ctx, "edx", NULL, 4, data); 439 WriteRegister(reg_ctx, "edi", NULL, 4, data); 440 WriteRegister(reg_ctx, "esi", NULL, 4, data); 441 WriteRegister(reg_ctx, "ebp", NULL, 4, data); 442 WriteRegister(reg_ctx, "esp", NULL, 4, data); 443 WriteRegister(reg_ctx, "ss", NULL, 4, data); 444 WriteRegister(reg_ctx, "eflags", NULL, 4, data); 445 WriteRegister(reg_ctx, "eip", NULL, 4, data); 446 WriteRegister(reg_ctx, "cs", NULL, 4, data); 447 WriteRegister(reg_ctx, "ds", NULL, 4, data); 448 WriteRegister(reg_ctx, "es", NULL, 4, data); 449 WriteRegister(reg_ctx, "fs", NULL, 4, data); 450 WriteRegister(reg_ctx, "gs", NULL, 4, data); 451 452 // Write out the EXC registers 453 data.PutHex32(EXCRegSet); 454 data.PutHex32(EXCWordCount); 455 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 456 WriteRegister(reg_ctx, "err", NULL, 4, data); 457 WriteRegister(reg_ctx, "faultvaddr", NULL, 4, data); 458 return true; 459 } 460 return false; 461 } 462 463 protected: 464 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 465 466 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 467 468 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 469 470 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 471 return 0; 472 } 473 474 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 475 return 0; 476 } 477 478 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 479 return 0; 480 } 481 }; 482 483 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 484 public: 485 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 486 const DataExtractor &data) 487 : RegisterContextDarwin_arm(thread, 0) { 488 SetRegisterDataFrom_LC_THREAD(data); 489 } 490 491 void InvalidateAllRegisters() override { 492 // Do nothing... registers are always valid... 493 } 494 495 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 496 lldb::offset_t offset = 0; 497 SetError(GPRRegSet, Read, -1); 498 SetError(FPURegSet, Read, -1); 499 SetError(EXCRegSet, Read, -1); 500 bool done = false; 501 502 while (!done) { 503 int flavor = data.GetU32(&offset); 504 uint32_t count = data.GetU32(&offset); 505 lldb::offset_t next_thread_state = offset + (count * 4); 506 switch (flavor) { 507 case GPRAltRegSet: 508 case GPRRegSet: 509 for (uint32_t i = 0; i < count; ++i) { 510 gpr.r[i] = data.GetU32(&offset); 511 } 512 513 // Note that gpr.cpsr is also copied by the above loop; this loop 514 // technically extends one element past the end of the gpr.r[] array. 515 516 SetError(GPRRegSet, Read, 0); 517 offset = next_thread_state; 518 break; 519 520 case FPURegSet: { 521 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 522 const int fpu_reg_buf_size = sizeof(fpu.floats); 523 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 524 fpu_reg_buf) == fpu_reg_buf_size) { 525 offset += fpu_reg_buf_size; 526 fpu.fpscr = data.GetU32(&offset); 527 SetError(FPURegSet, Read, 0); 528 } else { 529 done = true; 530 } 531 } 532 offset = next_thread_state; 533 break; 534 535 case EXCRegSet: 536 if (count == 3) { 537 exc.exception = data.GetU32(&offset); 538 exc.fsr = data.GetU32(&offset); 539 exc.far = data.GetU32(&offset); 540 SetError(EXCRegSet, Read, 0); 541 } 542 done = true; 543 offset = next_thread_state; 544 break; 545 546 // Unknown register set flavor, stop trying to parse. 547 default: 548 done = true; 549 } 550 } 551 } 552 553 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 554 const char *alt_name, size_t reg_byte_size, 555 Stream &data) { 556 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 557 if (reg_info == NULL) 558 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 559 if (reg_info) { 560 lldb_private::RegisterValue reg_value; 561 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 562 if (reg_info->byte_size >= reg_byte_size) 563 data.Write(reg_value.GetBytes(), reg_byte_size); 564 else { 565 data.Write(reg_value.GetBytes(), reg_info->byte_size); 566 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 567 ++i) 568 data.PutChar(0); 569 } 570 return reg_byte_size; 571 } 572 } 573 // Just write zeros if all else fails 574 for (size_t i = 0; i < reg_byte_size; ++i) 575 data.PutChar(0); 576 return reg_byte_size; 577 } 578 579 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 580 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 581 if (reg_ctx_sp) { 582 RegisterContext *reg_ctx = reg_ctx_sp.get(); 583 584 data.PutHex32(GPRRegSet); // Flavor 585 data.PutHex32(GPRWordCount); 586 WriteRegister(reg_ctx, "r0", NULL, 4, data); 587 WriteRegister(reg_ctx, "r1", NULL, 4, data); 588 WriteRegister(reg_ctx, "r2", NULL, 4, data); 589 WriteRegister(reg_ctx, "r3", NULL, 4, data); 590 WriteRegister(reg_ctx, "r4", NULL, 4, data); 591 WriteRegister(reg_ctx, "r5", NULL, 4, data); 592 WriteRegister(reg_ctx, "r6", NULL, 4, data); 593 WriteRegister(reg_ctx, "r7", NULL, 4, data); 594 WriteRegister(reg_ctx, "r8", NULL, 4, data); 595 WriteRegister(reg_ctx, "r9", NULL, 4, data); 596 WriteRegister(reg_ctx, "r10", NULL, 4, data); 597 WriteRegister(reg_ctx, "r11", NULL, 4, data); 598 WriteRegister(reg_ctx, "r12", NULL, 4, data); 599 WriteRegister(reg_ctx, "sp", NULL, 4, data); 600 WriteRegister(reg_ctx, "lr", NULL, 4, data); 601 WriteRegister(reg_ctx, "pc", NULL, 4, data); 602 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 603 604 // Write out the EXC registers 605 // data.PutHex32 (EXCRegSet); 606 // data.PutHex32 (EXCWordCount); 607 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 608 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 609 // WriteRegister (reg_ctx, "far", NULL, 4, data); 610 return true; 611 } 612 return false; 613 } 614 615 protected: 616 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 617 618 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 619 620 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 621 622 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 623 624 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 625 return 0; 626 } 627 628 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 629 return 0; 630 } 631 632 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 633 return 0; 634 } 635 636 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 637 return -1; 638 } 639 }; 640 641 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 642 public: 643 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 644 const DataExtractor &data) 645 : RegisterContextDarwin_arm64(thread, 0) { 646 SetRegisterDataFrom_LC_THREAD(data); 647 } 648 649 void InvalidateAllRegisters() override { 650 // Do nothing... registers are always valid... 651 } 652 653 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 654 lldb::offset_t offset = 0; 655 SetError(GPRRegSet, Read, -1); 656 SetError(FPURegSet, Read, -1); 657 SetError(EXCRegSet, Read, -1); 658 bool done = false; 659 while (!done) { 660 int flavor = data.GetU32(&offset); 661 uint32_t count = data.GetU32(&offset); 662 lldb::offset_t next_thread_state = offset + (count * 4); 663 switch (flavor) { 664 case GPRRegSet: 665 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 666 // 32-bit register) 667 if (count >= (33 * 2) + 1) { 668 for (uint32_t i = 0; i < 29; ++i) 669 gpr.x[i] = data.GetU64(&offset); 670 gpr.fp = data.GetU64(&offset); 671 gpr.lr = data.GetU64(&offset); 672 gpr.sp = data.GetU64(&offset); 673 gpr.pc = data.GetU64(&offset); 674 gpr.cpsr = data.GetU32(&offset); 675 SetError(GPRRegSet, Read, 0); 676 } 677 offset = next_thread_state; 678 break; 679 case FPURegSet: { 680 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 681 const int fpu_reg_buf_size = sizeof(fpu); 682 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 683 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 684 fpu_reg_buf) == fpu_reg_buf_size) { 685 SetError(FPURegSet, Read, 0); 686 } else { 687 done = true; 688 } 689 } 690 offset = next_thread_state; 691 break; 692 case EXCRegSet: 693 if (count == 4) { 694 exc.far = data.GetU64(&offset); 695 exc.esr = data.GetU32(&offset); 696 exc.exception = data.GetU32(&offset); 697 SetError(EXCRegSet, Read, 0); 698 } 699 offset = next_thread_state; 700 break; 701 default: 702 done = true; 703 break; 704 } 705 } 706 } 707 708 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 709 const char *alt_name, size_t reg_byte_size, 710 Stream &data) { 711 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 712 if (reg_info == NULL) 713 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 714 if (reg_info) { 715 lldb_private::RegisterValue reg_value; 716 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 717 if (reg_info->byte_size >= reg_byte_size) 718 data.Write(reg_value.GetBytes(), reg_byte_size); 719 else { 720 data.Write(reg_value.GetBytes(), reg_info->byte_size); 721 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 722 ++i) 723 data.PutChar(0); 724 } 725 return reg_byte_size; 726 } 727 } 728 // Just write zeros if all else fails 729 for (size_t i = 0; i < reg_byte_size; ++i) 730 data.PutChar(0); 731 return reg_byte_size; 732 } 733 734 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 735 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 736 if (reg_ctx_sp) { 737 RegisterContext *reg_ctx = reg_ctx_sp.get(); 738 739 data.PutHex32(GPRRegSet); // Flavor 740 data.PutHex32(GPRWordCount); 741 WriteRegister(reg_ctx, "x0", NULL, 8, data); 742 WriteRegister(reg_ctx, "x1", NULL, 8, data); 743 WriteRegister(reg_ctx, "x2", NULL, 8, data); 744 WriteRegister(reg_ctx, "x3", NULL, 8, data); 745 WriteRegister(reg_ctx, "x4", NULL, 8, data); 746 WriteRegister(reg_ctx, "x5", NULL, 8, data); 747 WriteRegister(reg_ctx, "x6", NULL, 8, data); 748 WriteRegister(reg_ctx, "x7", NULL, 8, data); 749 WriteRegister(reg_ctx, "x8", NULL, 8, data); 750 WriteRegister(reg_ctx, "x9", NULL, 8, data); 751 WriteRegister(reg_ctx, "x10", NULL, 8, data); 752 WriteRegister(reg_ctx, "x11", NULL, 8, data); 753 WriteRegister(reg_ctx, "x12", NULL, 8, data); 754 WriteRegister(reg_ctx, "x13", NULL, 8, data); 755 WriteRegister(reg_ctx, "x14", NULL, 8, data); 756 WriteRegister(reg_ctx, "x15", NULL, 8, data); 757 WriteRegister(reg_ctx, "x16", NULL, 8, data); 758 WriteRegister(reg_ctx, "x17", NULL, 8, data); 759 WriteRegister(reg_ctx, "x18", NULL, 8, data); 760 WriteRegister(reg_ctx, "x19", NULL, 8, data); 761 WriteRegister(reg_ctx, "x20", NULL, 8, data); 762 WriteRegister(reg_ctx, "x21", NULL, 8, data); 763 WriteRegister(reg_ctx, "x22", NULL, 8, data); 764 WriteRegister(reg_ctx, "x23", NULL, 8, data); 765 WriteRegister(reg_ctx, "x24", NULL, 8, data); 766 WriteRegister(reg_ctx, "x25", NULL, 8, data); 767 WriteRegister(reg_ctx, "x26", NULL, 8, data); 768 WriteRegister(reg_ctx, "x27", NULL, 8, data); 769 WriteRegister(reg_ctx, "x28", NULL, 8, data); 770 WriteRegister(reg_ctx, "fp", NULL, 8, data); 771 WriteRegister(reg_ctx, "lr", NULL, 8, data); 772 WriteRegister(reg_ctx, "sp", NULL, 8, data); 773 WriteRegister(reg_ctx, "pc", NULL, 8, data); 774 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 775 776 // Write out the EXC registers 777 // data.PutHex32 (EXCRegSet); 778 // data.PutHex32 (EXCWordCount); 779 // WriteRegister (reg_ctx, "far", NULL, 8, data); 780 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 781 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 782 return true; 783 } 784 return false; 785 } 786 787 protected: 788 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 789 790 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 791 792 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 793 794 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 795 796 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 797 return 0; 798 } 799 800 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 801 return 0; 802 } 803 804 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 805 return 0; 806 } 807 808 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 809 return -1; 810 } 811 }; 812 813 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 814 switch (magic) { 815 case MH_MAGIC: 816 case MH_CIGAM: 817 return sizeof(struct mach_header); 818 819 case MH_MAGIC_64: 820 case MH_CIGAM_64: 821 return sizeof(struct mach_header_64); 822 break; 823 824 default: 825 break; 826 } 827 return 0; 828 } 829 830 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 831 832 void ObjectFileMachO::Initialize() { 833 PluginManager::RegisterPlugin( 834 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 835 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 836 } 837 838 void ObjectFileMachO::Terminate() { 839 PluginManager::UnregisterPlugin(CreateInstance); 840 } 841 842 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 843 static ConstString g_name("mach-o"); 844 return g_name; 845 } 846 847 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 848 return "Mach-o object file reader (32 and 64 bit)"; 849 } 850 851 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 852 DataBufferSP &data_sp, 853 lldb::offset_t data_offset, 854 const FileSpec *file, 855 lldb::offset_t file_offset, 856 lldb::offset_t length) { 857 if (!data_sp) { 858 data_sp = MapFileData(*file, length, file_offset); 859 if (!data_sp) 860 return nullptr; 861 data_offset = 0; 862 } 863 864 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 865 return nullptr; 866 867 // Update the data to contain the entire file if it doesn't already 868 if (data_sp->GetByteSize() < length) { 869 data_sp = MapFileData(*file, length, file_offset); 870 if (!data_sp) 871 return nullptr; 872 data_offset = 0; 873 } 874 auto objfile_ap = llvm::make_unique<ObjectFileMachO>( 875 module_sp, data_sp, data_offset, file, file_offset, length); 876 if (!objfile_ap || !objfile_ap->ParseHeader()) 877 return nullptr; 878 879 return objfile_ap.release(); 880 } 881 882 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 883 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 884 const ProcessSP &process_sp, lldb::addr_t header_addr) { 885 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 886 std::unique_ptr<ObjectFile> objfile_ap( 887 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 888 if (objfile_ap.get() && objfile_ap->ParseHeader()) 889 return objfile_ap.release(); 890 } 891 return NULL; 892 } 893 894 size_t ObjectFileMachO::GetModuleSpecifications( 895 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 896 lldb::offset_t data_offset, lldb::offset_t file_offset, 897 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 898 const size_t initial_count = specs.GetSize(); 899 900 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 901 DataExtractor data; 902 data.SetData(data_sp); 903 llvm::MachO::mach_header header; 904 if (ParseHeader(data, &data_offset, header)) { 905 size_t header_and_load_cmds = 906 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 907 if (header_and_load_cmds >= data_sp->GetByteSize()) { 908 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 909 data.SetData(data_sp); 910 data_offset = MachHeaderSizeFromMagic(header.magic); 911 } 912 if (data_sp) { 913 ModuleSpec spec; 914 spec.GetFileSpec() = file; 915 spec.SetObjectOffset(file_offset); 916 spec.SetObjectSize(length); 917 918 if (GetArchitecture(header, data, data_offset, 919 spec.GetArchitecture())) { 920 if (spec.GetArchitecture().IsValid()) { 921 GetUUID(header, data, data_offset, spec.GetUUID()); 922 specs.Append(spec); 923 } 924 } 925 } 926 } 927 } 928 return specs.GetSize() - initial_count; 929 } 930 931 const ConstString &ObjectFileMachO::GetSegmentNameTEXT() { 932 static ConstString g_segment_name_TEXT("__TEXT"); 933 return g_segment_name_TEXT; 934 } 935 936 const ConstString &ObjectFileMachO::GetSegmentNameDATA() { 937 static ConstString g_segment_name_DATA("__DATA"); 938 return g_segment_name_DATA; 939 } 940 941 const ConstString &ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 942 static ConstString g_segment_name("__DATA_DIRTY"); 943 return g_segment_name; 944 } 945 946 const ConstString &ObjectFileMachO::GetSegmentNameDATA_CONST() { 947 static ConstString g_segment_name("__DATA_CONST"); 948 return g_segment_name; 949 } 950 951 const ConstString &ObjectFileMachO::GetSegmentNameOBJC() { 952 static ConstString g_segment_name_OBJC("__OBJC"); 953 return g_segment_name_OBJC; 954 } 955 956 const ConstString &ObjectFileMachO::GetSegmentNameLINKEDIT() { 957 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 958 return g_section_name_LINKEDIT; 959 } 960 961 const ConstString &ObjectFileMachO::GetSectionNameEHFrame() { 962 static ConstString g_section_name_eh_frame("__eh_frame"); 963 return g_section_name_eh_frame; 964 } 965 966 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 967 lldb::addr_t data_offset, 968 lldb::addr_t data_length) { 969 DataExtractor data; 970 data.SetData(data_sp, data_offset, data_length); 971 lldb::offset_t offset = 0; 972 uint32_t magic = data.GetU32(&offset); 973 return MachHeaderSizeFromMagic(magic) != 0; 974 } 975 976 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 977 DataBufferSP &data_sp, 978 lldb::offset_t data_offset, 979 const FileSpec *file, 980 lldb::offset_t file_offset, 981 lldb::offset_t length) 982 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 983 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 984 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 985 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 986 ::memset(&m_header, 0, sizeof(m_header)); 987 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 988 } 989 990 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 991 lldb::DataBufferSP &header_data_sp, 992 const lldb::ProcessSP &process_sp, 993 lldb::addr_t header_addr) 994 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 995 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 996 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 997 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 998 ::memset(&m_header, 0, sizeof(m_header)); 999 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1000 } 1001 1002 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 1003 lldb::offset_t *data_offset_ptr, 1004 llvm::MachO::mach_header &header) { 1005 data.SetByteOrder(endian::InlHostByteOrder()); 1006 // Leave magic in the original byte order 1007 header.magic = data.GetU32(data_offset_ptr); 1008 bool can_parse = false; 1009 bool is_64_bit = false; 1010 switch (header.magic) { 1011 case MH_MAGIC: 1012 data.SetByteOrder(endian::InlHostByteOrder()); 1013 data.SetAddressByteSize(4); 1014 can_parse = true; 1015 break; 1016 1017 case MH_MAGIC_64: 1018 data.SetByteOrder(endian::InlHostByteOrder()); 1019 data.SetAddressByteSize(8); 1020 can_parse = true; 1021 is_64_bit = true; 1022 break; 1023 1024 case MH_CIGAM: 1025 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1026 ? eByteOrderLittle 1027 : eByteOrderBig); 1028 data.SetAddressByteSize(4); 1029 can_parse = true; 1030 break; 1031 1032 case MH_CIGAM_64: 1033 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1034 ? eByteOrderLittle 1035 : eByteOrderBig); 1036 data.SetAddressByteSize(8); 1037 is_64_bit = true; 1038 can_parse = true; 1039 break; 1040 1041 default: 1042 break; 1043 } 1044 1045 if (can_parse) { 1046 data.GetU32(data_offset_ptr, &header.cputype, 6); 1047 if (is_64_bit) 1048 *data_offset_ptr += 4; 1049 return true; 1050 } else { 1051 memset(&header, 0, sizeof(header)); 1052 } 1053 return false; 1054 } 1055 1056 bool ObjectFileMachO::ParseHeader() { 1057 ModuleSP module_sp(GetModule()); 1058 if (module_sp) { 1059 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1060 bool can_parse = false; 1061 lldb::offset_t offset = 0; 1062 m_data.SetByteOrder(endian::InlHostByteOrder()); 1063 // Leave magic in the original byte order 1064 m_header.magic = m_data.GetU32(&offset); 1065 switch (m_header.magic) { 1066 case MH_MAGIC: 1067 m_data.SetByteOrder(endian::InlHostByteOrder()); 1068 m_data.SetAddressByteSize(4); 1069 can_parse = true; 1070 break; 1071 1072 case MH_MAGIC_64: 1073 m_data.SetByteOrder(endian::InlHostByteOrder()); 1074 m_data.SetAddressByteSize(8); 1075 can_parse = true; 1076 break; 1077 1078 case MH_CIGAM: 1079 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1080 ? eByteOrderLittle 1081 : eByteOrderBig); 1082 m_data.SetAddressByteSize(4); 1083 can_parse = true; 1084 break; 1085 1086 case MH_CIGAM_64: 1087 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1088 ? eByteOrderLittle 1089 : eByteOrderBig); 1090 m_data.SetAddressByteSize(8); 1091 can_parse = true; 1092 break; 1093 1094 default: 1095 break; 1096 } 1097 1098 if (can_parse) { 1099 m_data.GetU32(&offset, &m_header.cputype, 6); 1100 1101 ArchSpec mach_arch; 1102 1103 if (GetArchitecture(mach_arch)) { 1104 // Check if the module has a required architecture 1105 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1106 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1107 return false; 1108 1109 if (SetModulesArchitecture(mach_arch)) { 1110 const size_t header_and_lc_size = 1111 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1112 if (m_data.GetByteSize() < header_and_lc_size) { 1113 DataBufferSP data_sp; 1114 ProcessSP process_sp(m_process_wp.lock()); 1115 if (process_sp) { 1116 data_sp = 1117 ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1118 } else { 1119 // Read in all only the load command data from the file on disk 1120 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset); 1121 if (data_sp->GetByteSize() != header_and_lc_size) 1122 return false; 1123 } 1124 if (data_sp) 1125 m_data.SetData(data_sp); 1126 } 1127 } 1128 return true; 1129 } 1130 } else { 1131 memset(&m_header, 0, sizeof(struct mach_header)); 1132 } 1133 } 1134 return false; 1135 } 1136 1137 ByteOrder ObjectFileMachO::GetByteOrder() const { 1138 return m_data.GetByteOrder(); 1139 } 1140 1141 bool ObjectFileMachO::IsExecutable() const { 1142 return m_header.filetype == MH_EXECUTE; 1143 } 1144 1145 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1146 return m_data.GetAddressByteSize(); 1147 } 1148 1149 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1150 Symtab *symtab = GetSymtab(); 1151 if (symtab) { 1152 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1153 if (symbol) { 1154 if (symbol->ValueIsAddress()) { 1155 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1156 if (section_sp) { 1157 const lldb::SectionType section_type = section_sp->GetType(); 1158 switch (section_type) { 1159 case eSectionTypeInvalid: 1160 return AddressClass::eUnknown; 1161 1162 case eSectionTypeCode: 1163 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1164 // For ARM we have a bit in the n_desc field of the symbol that 1165 // tells us ARM/Thumb which is bit 0x0008. 1166 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1167 return AddressClass::eCodeAlternateISA; 1168 } 1169 return AddressClass::eCode; 1170 1171 case eSectionTypeContainer: 1172 return AddressClass::eUnknown; 1173 1174 case eSectionTypeData: 1175 case eSectionTypeDataCString: 1176 case eSectionTypeDataCStringPointers: 1177 case eSectionTypeDataSymbolAddress: 1178 case eSectionTypeData4: 1179 case eSectionTypeData8: 1180 case eSectionTypeData16: 1181 case eSectionTypeDataPointers: 1182 case eSectionTypeZeroFill: 1183 case eSectionTypeDataObjCMessageRefs: 1184 case eSectionTypeDataObjCCFStrings: 1185 case eSectionTypeGoSymtab: 1186 return AddressClass::eData; 1187 1188 case eSectionTypeDebug: 1189 case eSectionTypeDWARFDebugAbbrev: 1190 case eSectionTypeDWARFDebugAddr: 1191 case eSectionTypeDWARFDebugAranges: 1192 case eSectionTypeDWARFDebugCuIndex: 1193 case eSectionTypeDWARFDebugFrame: 1194 case eSectionTypeDWARFDebugInfo: 1195 case eSectionTypeDWARFDebugLine: 1196 case eSectionTypeDWARFDebugLineStr: 1197 case eSectionTypeDWARFDebugLoc: 1198 case eSectionTypeDWARFDebugLocLists: 1199 case eSectionTypeDWARFDebugMacInfo: 1200 case eSectionTypeDWARFDebugMacro: 1201 case eSectionTypeDWARFDebugNames: 1202 case eSectionTypeDWARFDebugPubNames: 1203 case eSectionTypeDWARFDebugPubTypes: 1204 case eSectionTypeDWARFDebugRanges: 1205 case eSectionTypeDWARFDebugRngLists: 1206 case eSectionTypeDWARFDebugStr: 1207 case eSectionTypeDWARFDebugStrOffsets: 1208 case eSectionTypeDWARFDebugTypes: 1209 case eSectionTypeDWARFAppleNames: 1210 case eSectionTypeDWARFAppleTypes: 1211 case eSectionTypeDWARFAppleNamespaces: 1212 case eSectionTypeDWARFAppleObjC: 1213 case eSectionTypeDWARFGNUDebugAltLink: 1214 return AddressClass::eDebug; 1215 1216 case eSectionTypeEHFrame: 1217 case eSectionTypeARMexidx: 1218 case eSectionTypeARMextab: 1219 case eSectionTypeCompactUnwind: 1220 return AddressClass::eRuntime; 1221 1222 case eSectionTypeAbsoluteAddress: 1223 case eSectionTypeELFSymbolTable: 1224 case eSectionTypeELFDynamicSymbols: 1225 case eSectionTypeELFRelocationEntries: 1226 case eSectionTypeELFDynamicLinkInfo: 1227 case eSectionTypeOther: 1228 return AddressClass::eUnknown; 1229 } 1230 } 1231 } 1232 1233 const SymbolType symbol_type = symbol->GetType(); 1234 switch (symbol_type) { 1235 case eSymbolTypeAny: 1236 return AddressClass::eUnknown; 1237 case eSymbolTypeAbsolute: 1238 return AddressClass::eUnknown; 1239 1240 case eSymbolTypeCode: 1241 case eSymbolTypeTrampoline: 1242 case eSymbolTypeResolver: 1243 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1244 // For ARM we have a bit in the n_desc field of the symbol that tells 1245 // us ARM/Thumb which is bit 0x0008. 1246 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1247 return AddressClass::eCodeAlternateISA; 1248 } 1249 return AddressClass::eCode; 1250 1251 case eSymbolTypeData: 1252 return AddressClass::eData; 1253 case eSymbolTypeRuntime: 1254 return AddressClass::eRuntime; 1255 case eSymbolTypeException: 1256 return AddressClass::eRuntime; 1257 case eSymbolTypeSourceFile: 1258 return AddressClass::eDebug; 1259 case eSymbolTypeHeaderFile: 1260 return AddressClass::eDebug; 1261 case eSymbolTypeObjectFile: 1262 return AddressClass::eDebug; 1263 case eSymbolTypeCommonBlock: 1264 return AddressClass::eDebug; 1265 case eSymbolTypeBlock: 1266 return AddressClass::eDebug; 1267 case eSymbolTypeLocal: 1268 return AddressClass::eData; 1269 case eSymbolTypeParam: 1270 return AddressClass::eData; 1271 case eSymbolTypeVariable: 1272 return AddressClass::eData; 1273 case eSymbolTypeVariableType: 1274 return AddressClass::eDebug; 1275 case eSymbolTypeLineEntry: 1276 return AddressClass::eDebug; 1277 case eSymbolTypeLineHeader: 1278 return AddressClass::eDebug; 1279 case eSymbolTypeScopeBegin: 1280 return AddressClass::eDebug; 1281 case eSymbolTypeScopeEnd: 1282 return AddressClass::eDebug; 1283 case eSymbolTypeAdditional: 1284 return AddressClass::eUnknown; 1285 case eSymbolTypeCompiler: 1286 return AddressClass::eDebug; 1287 case eSymbolTypeInstrumentation: 1288 return AddressClass::eDebug; 1289 case eSymbolTypeUndefined: 1290 return AddressClass::eUnknown; 1291 case eSymbolTypeObjCClass: 1292 return AddressClass::eRuntime; 1293 case eSymbolTypeObjCMetaClass: 1294 return AddressClass::eRuntime; 1295 case eSymbolTypeObjCIVar: 1296 return AddressClass::eRuntime; 1297 case eSymbolTypeReExported: 1298 return AddressClass::eRuntime; 1299 } 1300 } 1301 } 1302 return AddressClass::eUnknown; 1303 } 1304 1305 Symtab *ObjectFileMachO::GetSymtab() { 1306 ModuleSP module_sp(GetModule()); 1307 if (module_sp) { 1308 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1309 if (m_symtab_ap.get() == NULL) { 1310 m_symtab_ap.reset(new Symtab(this)); 1311 std::lock_guard<std::recursive_mutex> symtab_guard( 1312 m_symtab_ap->GetMutex()); 1313 ParseSymtab(); 1314 m_symtab_ap->Finalize(); 1315 } 1316 } 1317 return m_symtab_ap.get(); 1318 } 1319 1320 bool ObjectFileMachO::IsStripped() { 1321 if (m_dysymtab.cmd == 0) { 1322 ModuleSP module_sp(GetModule()); 1323 if (module_sp) { 1324 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1325 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1326 const lldb::offset_t load_cmd_offset = offset; 1327 1328 load_command lc; 1329 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1330 break; 1331 if (lc.cmd == LC_DYSYMTAB) { 1332 m_dysymtab.cmd = lc.cmd; 1333 m_dysymtab.cmdsize = lc.cmdsize; 1334 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1335 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1336 NULL) { 1337 // Clear m_dysymtab if we were unable to read all items from the 1338 // load command 1339 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1340 } 1341 } 1342 offset = load_cmd_offset + lc.cmdsize; 1343 } 1344 } 1345 } 1346 if (m_dysymtab.cmd) 1347 return m_dysymtab.nlocalsym <= 1; 1348 return false; 1349 } 1350 1351 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1352 EncryptedFileRanges result; 1353 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1354 1355 encryption_info_command encryption_cmd; 1356 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1357 const lldb::offset_t load_cmd_offset = offset; 1358 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1359 break; 1360 1361 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the 1362 // 3 fields we care about, so treat them the same. 1363 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1364 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1365 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1366 if (encryption_cmd.cryptid != 0) { 1367 EncryptedFileRanges::Entry entry; 1368 entry.SetRangeBase(encryption_cmd.cryptoff); 1369 entry.SetByteSize(encryption_cmd.cryptsize); 1370 result.Append(entry); 1371 } 1372 } 1373 } 1374 offset = load_cmd_offset + encryption_cmd.cmdsize; 1375 } 1376 1377 return result; 1378 } 1379 1380 void ObjectFileMachO::SanitizeSegmentCommand(segment_command_64 &seg_cmd, 1381 uint32_t cmd_idx) { 1382 if (m_length == 0 || seg_cmd.filesize == 0) 1383 return; 1384 1385 if (seg_cmd.fileoff > m_length) { 1386 // We have a load command that says it extends past the end of the file. 1387 // This is likely a corrupt file. We don't have any way to return an error 1388 // condition here (this method was likely invoked from something like 1389 // ObjectFile::GetSectionList()), so we just null out the section contents, 1390 // and dump a message to stdout. The most common case here is core file 1391 // debugging with a truncated file. 1392 const char *lc_segment_name = 1393 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1394 GetModule()->ReportWarning( 1395 "load command %u %s has a fileoff (0x%" PRIx64 1396 ") that extends beyond the end of the file (0x%" PRIx64 1397 "), ignoring this section", 1398 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1399 1400 seg_cmd.fileoff = 0; 1401 seg_cmd.filesize = 0; 1402 } 1403 1404 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1405 // We have a load command that says it extends past the end of the file. 1406 // This is likely a corrupt file. We don't have any way to return an error 1407 // condition here (this method was likely invoked from something like 1408 // ObjectFile::GetSectionList()), so we just null out the section contents, 1409 // and dump a message to stdout. The most common case here is core file 1410 // debugging with a truncated file. 1411 const char *lc_segment_name = 1412 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1413 GetModule()->ReportWarning( 1414 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1415 ") that extends beyond the end of the file (0x%" PRIx64 1416 "), the segment will be truncated to match", 1417 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1418 1419 // Truncate the length 1420 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1421 } 1422 } 1423 1424 static uint32_t GetSegmentPermissions(const segment_command_64 &seg_cmd) { 1425 uint32_t result = 0; 1426 if (seg_cmd.initprot & VM_PROT_READ) 1427 result |= ePermissionsReadable; 1428 if (seg_cmd.initprot & VM_PROT_WRITE) 1429 result |= ePermissionsWritable; 1430 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1431 result |= ePermissionsExecutable; 1432 return result; 1433 } 1434 1435 static lldb::SectionType GetSectionType(uint32_t flags, 1436 ConstString section_name) { 1437 1438 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1439 return eSectionTypeCode; 1440 1441 uint32_t mach_sect_type = flags & SECTION_TYPE; 1442 static ConstString g_sect_name_objc_data("__objc_data"); 1443 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1444 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1445 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1446 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1447 static ConstString g_sect_name_objc_const("__objc_const"); 1448 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1449 static ConstString g_sect_name_cfstring("__cfstring"); 1450 1451 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1452 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1453 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1454 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1455 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1456 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1457 static ConstString g_sect_name_dwarf_debug_loclists("__debug_loclists"); 1458 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1459 static ConstString g_sect_name_dwarf_debug_names("__debug_names"); 1460 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1461 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1462 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1463 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1464 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1465 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1466 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1467 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1468 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1469 static ConstString g_sect_name_eh_frame("__eh_frame"); 1470 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1471 static ConstString g_sect_name_text("__text"); 1472 static ConstString g_sect_name_data("__data"); 1473 static ConstString g_sect_name_go_symtab("__gosymtab"); 1474 1475 if (section_name == g_sect_name_dwarf_debug_abbrev) 1476 return eSectionTypeDWARFDebugAbbrev; 1477 if (section_name == g_sect_name_dwarf_debug_aranges) 1478 return eSectionTypeDWARFDebugAranges; 1479 if (section_name == g_sect_name_dwarf_debug_frame) 1480 return eSectionTypeDWARFDebugFrame; 1481 if (section_name == g_sect_name_dwarf_debug_info) 1482 return eSectionTypeDWARFDebugInfo; 1483 if (section_name == g_sect_name_dwarf_debug_line) 1484 return eSectionTypeDWARFDebugLine; 1485 if (section_name == g_sect_name_dwarf_debug_loc) 1486 return eSectionTypeDWARFDebugLoc; 1487 if (section_name == g_sect_name_dwarf_debug_loclists) 1488 return eSectionTypeDWARFDebugLocLists; 1489 if (section_name == g_sect_name_dwarf_debug_macinfo) 1490 return eSectionTypeDWARFDebugMacInfo; 1491 if (section_name == g_sect_name_dwarf_debug_names) 1492 return eSectionTypeDWARFDebugNames; 1493 if (section_name == g_sect_name_dwarf_debug_pubnames) 1494 return eSectionTypeDWARFDebugPubNames; 1495 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1496 return eSectionTypeDWARFDebugPubTypes; 1497 if (section_name == g_sect_name_dwarf_debug_ranges) 1498 return eSectionTypeDWARFDebugRanges; 1499 if (section_name == g_sect_name_dwarf_debug_str) 1500 return eSectionTypeDWARFDebugStr; 1501 if (section_name == g_sect_name_dwarf_debug_types) 1502 return eSectionTypeDWARFDebugTypes; 1503 if (section_name == g_sect_name_dwarf_apple_names) 1504 return eSectionTypeDWARFAppleNames; 1505 if (section_name == g_sect_name_dwarf_apple_types) 1506 return eSectionTypeDWARFAppleTypes; 1507 if (section_name == g_sect_name_dwarf_apple_namespaces) 1508 return eSectionTypeDWARFAppleNamespaces; 1509 if (section_name == g_sect_name_dwarf_apple_objc) 1510 return eSectionTypeDWARFAppleObjC; 1511 if (section_name == g_sect_name_objc_selrefs) 1512 return eSectionTypeDataCStringPointers; 1513 if (section_name == g_sect_name_objc_msgrefs) 1514 return eSectionTypeDataObjCMessageRefs; 1515 if (section_name == g_sect_name_eh_frame) 1516 return eSectionTypeEHFrame; 1517 if (section_name == g_sect_name_compact_unwind) 1518 return eSectionTypeCompactUnwind; 1519 if (section_name == g_sect_name_cfstring) 1520 return eSectionTypeDataObjCCFStrings; 1521 if (section_name == g_sect_name_go_symtab) 1522 return eSectionTypeGoSymtab; 1523 if (section_name == g_sect_name_objc_data || 1524 section_name == g_sect_name_objc_classrefs || 1525 section_name == g_sect_name_objc_superrefs || 1526 section_name == g_sect_name_objc_const || 1527 section_name == g_sect_name_objc_classlist) { 1528 return eSectionTypeDataPointers; 1529 } 1530 1531 switch (mach_sect_type) { 1532 // TODO: categorize sections by other flags for regular sections 1533 case S_REGULAR: 1534 if (section_name == g_sect_name_text) 1535 return eSectionTypeCode; 1536 if (section_name == g_sect_name_data) 1537 return eSectionTypeData; 1538 return eSectionTypeOther; 1539 case S_ZEROFILL: 1540 return eSectionTypeZeroFill; 1541 case S_CSTRING_LITERALS: // section with only literal C strings 1542 return eSectionTypeDataCString; 1543 case S_4BYTE_LITERALS: // section with only 4 byte literals 1544 return eSectionTypeData4; 1545 case S_8BYTE_LITERALS: // section with only 8 byte literals 1546 return eSectionTypeData8; 1547 case S_LITERAL_POINTERS: // section with only pointers to literals 1548 return eSectionTypeDataPointers; 1549 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1550 return eSectionTypeDataPointers; 1551 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1552 return eSectionTypeDataPointers; 1553 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1554 // the reserved2 field 1555 return eSectionTypeCode; 1556 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1557 // initialization 1558 return eSectionTypeDataPointers; 1559 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1560 // termination 1561 return eSectionTypeDataPointers; 1562 case S_COALESCED: 1563 return eSectionTypeOther; 1564 case S_GB_ZEROFILL: 1565 return eSectionTypeZeroFill; 1566 case S_INTERPOSING: // section with only pairs of function pointers for 1567 // interposing 1568 return eSectionTypeCode; 1569 case S_16BYTE_LITERALS: // section with only 16 byte literals 1570 return eSectionTypeData16; 1571 case S_DTRACE_DOF: 1572 return eSectionTypeDebug; 1573 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1574 return eSectionTypeDataPointers; 1575 default: 1576 return eSectionTypeOther; 1577 } 1578 } 1579 1580 struct ObjectFileMachO::SegmentParsingContext { 1581 const EncryptedFileRanges EncryptedRanges; 1582 lldb_private::SectionList &UnifiedList; 1583 uint32_t NextSegmentIdx = 0; 1584 uint32_t NextSectionIdx = 0; 1585 bool FileAddressesChanged = false; 1586 1587 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1588 lldb_private::SectionList &UnifiedList) 1589 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1590 }; 1591 1592 void ObjectFileMachO::ProcessSegmentCommand(const load_command &load_cmd_, 1593 lldb::offset_t offset, 1594 uint32_t cmd_idx, 1595 SegmentParsingContext &context) { 1596 segment_command_64 load_cmd; 1597 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1598 1599 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1600 return; 1601 1602 ModuleSP module_sp = GetModule(); 1603 const bool is_core = GetType() == eTypeCoreFile; 1604 const bool is_dsym = (m_header.filetype == MH_DSYM); 1605 bool add_section = true; 1606 bool add_to_unified = true; 1607 ConstString const_segname( 1608 load_cmd.segname, 1609 std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 1610 1611 SectionSP unified_section_sp( 1612 context.UnifiedList.FindSectionByName(const_segname)); 1613 if (is_dsym && unified_section_sp) { 1614 if (const_segname == GetSegmentNameLINKEDIT()) { 1615 // We need to keep the __LINKEDIT segment private to this object file 1616 // only 1617 add_to_unified = false; 1618 } else { 1619 // This is the dSYM file and this section has already been created by the 1620 // object file, no need to create it. 1621 add_section = false; 1622 } 1623 } 1624 load_cmd.vmaddr = m_data.GetAddress(&offset); 1625 load_cmd.vmsize = m_data.GetAddress(&offset); 1626 load_cmd.fileoff = m_data.GetAddress(&offset); 1627 load_cmd.filesize = m_data.GetAddress(&offset); 1628 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1629 return; 1630 1631 SanitizeSegmentCommand(load_cmd, cmd_idx); 1632 1633 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1634 const bool segment_is_encrypted = 1635 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1636 1637 // Keep a list of mach segments around in case we need to get at data that 1638 // isn't stored in the abstracted Sections. 1639 m_mach_segments.push_back(load_cmd); 1640 1641 // Use a segment ID of the segment index shifted left by 8 so they never 1642 // conflict with any of the sections. 1643 SectionSP segment_sp; 1644 if (add_section && (const_segname || is_core)) { 1645 segment_sp.reset(new Section( 1646 module_sp, // Module to which this section belongs 1647 this, // Object file to which this sections belongs 1648 ++context.NextSegmentIdx 1649 << 8, // Section ID is the 1 based segment index 1650 // shifted right by 8 bits as not to collide with any of the 256 1651 // section IDs that are possible 1652 const_segname, // Name of this section 1653 eSectionTypeContainer, // This section is a container of other 1654 // sections. 1655 load_cmd.vmaddr, // File VM address == addresses as they are 1656 // found in the object file 1657 load_cmd.vmsize, // VM size in bytes of this section 1658 load_cmd.fileoff, // Offset to the data for this section in 1659 // the file 1660 load_cmd.filesize, // Size in bytes of this section as found 1661 // in the file 1662 0, // Segments have no alignment information 1663 load_cmd.flags)); // Flags for this section 1664 1665 segment_sp->SetIsEncrypted(segment_is_encrypted); 1666 m_sections_ap->AddSection(segment_sp); 1667 segment_sp->SetPermissions(segment_permissions); 1668 if (add_to_unified) 1669 context.UnifiedList.AddSection(segment_sp); 1670 } else if (unified_section_sp) { 1671 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1672 // Check to see if the module was read from memory? 1673 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) { 1674 // We have a module that is in memory and needs to have its file 1675 // address adjusted. We need to do this because when we load a file 1676 // from memory, its addresses will be slid already, yet the addresses 1677 // in the new symbol file will still be unslid. Since everything is 1678 // stored as section offset, this shouldn't cause any problems. 1679 1680 // Make sure we've parsed the symbol table from the ObjectFile before 1681 // we go around changing its Sections. 1682 module_sp->GetObjectFile()->GetSymtab(); 1683 // eh_frame would present the same problems but we parse that on a per- 1684 // function basis as-needed so it's more difficult to remove its use of 1685 // the Sections. Realistically, the environments where this code path 1686 // will be taken will not have eh_frame sections. 1687 1688 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1689 1690 // Notify the module that the section addresses have been changed once 1691 // we're done so any file-address caches can be updated. 1692 context.FileAddressesChanged = true; 1693 } 1694 } 1695 m_sections_ap->AddSection(unified_section_sp); 1696 } 1697 1698 struct section_64 sect64; 1699 ::memset(§64, 0, sizeof(sect64)); 1700 // Push a section into our mach sections for the section at index zero 1701 // (NO_SECT) if we don't have any mach sections yet... 1702 if (m_mach_sections.empty()) 1703 m_mach_sections.push_back(sect64); 1704 uint32_t segment_sect_idx; 1705 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1706 1707 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1708 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1709 ++segment_sect_idx) { 1710 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1711 sizeof(sect64.sectname)) == NULL) 1712 break; 1713 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1714 sizeof(sect64.segname)) == NULL) 1715 break; 1716 sect64.addr = m_data.GetAddress(&offset); 1717 sect64.size = m_data.GetAddress(&offset); 1718 1719 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1720 break; 1721 1722 // Keep a list of mach sections around in case we need to get at data that 1723 // isn't stored in the abstracted Sections. 1724 m_mach_sections.push_back(sect64); 1725 1726 if (add_section) { 1727 ConstString section_name( 1728 sect64.sectname, 1729 std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1730 if (!const_segname) { 1731 // We have a segment with no name so we need to conjure up segments 1732 // that correspond to the section's segname if there isn't already such 1733 // a section. If there is such a section, we resize the section so that 1734 // it spans all sections. We also mark these sections as fake so 1735 // address matches don't hit if they land in the gaps between the child 1736 // sections. 1737 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1738 sizeof(sect64.segname)); 1739 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1740 if (segment_sp.get()) { 1741 Section *segment = segment_sp.get(); 1742 // Grow the section size as needed. 1743 const lldb::addr_t sect64_min_addr = sect64.addr; 1744 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1745 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1746 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1747 const lldb::addr_t curr_seg_max_addr = 1748 curr_seg_min_addr + curr_seg_byte_size; 1749 if (sect64_min_addr >= curr_seg_min_addr) { 1750 const lldb::addr_t new_seg_byte_size = 1751 sect64_max_addr - curr_seg_min_addr; 1752 // Only grow the section size if needed 1753 if (new_seg_byte_size > curr_seg_byte_size) 1754 segment->SetByteSize(new_seg_byte_size); 1755 } else { 1756 // We need to change the base address of the segment and adjust the 1757 // child section offsets for all existing children. 1758 const lldb::addr_t slide_amount = 1759 sect64_min_addr - curr_seg_min_addr; 1760 segment->Slide(slide_amount, false); 1761 segment->GetChildren().Slide(-slide_amount, false); 1762 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1763 } 1764 1765 // Grow the section size as needed. 1766 if (sect64.offset) { 1767 const lldb::addr_t segment_min_file_offset = 1768 segment->GetFileOffset(); 1769 const lldb::addr_t segment_max_file_offset = 1770 segment_min_file_offset + segment->GetFileSize(); 1771 1772 const lldb::addr_t section_min_file_offset = sect64.offset; 1773 const lldb::addr_t section_max_file_offset = 1774 section_min_file_offset + sect64.size; 1775 const lldb::addr_t new_file_offset = 1776 std::min(section_min_file_offset, segment_min_file_offset); 1777 const lldb::addr_t new_file_size = 1778 std::max(section_max_file_offset, segment_max_file_offset) - 1779 new_file_offset; 1780 segment->SetFileOffset(new_file_offset); 1781 segment->SetFileSize(new_file_size); 1782 } 1783 } else { 1784 // Create a fake section for the section's named segment 1785 segment_sp.reset(new Section( 1786 segment_sp, // Parent section 1787 module_sp, // Module to which this section belongs 1788 this, // Object file to which this section belongs 1789 ++context.NextSegmentIdx 1790 << 8, // Section ID is the 1 based segment index 1791 // shifted right by 8 bits as not to 1792 // collide with any of the 256 section IDs 1793 // that are possible 1794 const_segname, // Name of this section 1795 eSectionTypeContainer, // This section is a container of 1796 // other sections. 1797 sect64.addr, // File VM address == addresses as they are 1798 // found in the object file 1799 sect64.size, // VM size in bytes of this section 1800 sect64.offset, // Offset to the data for this section in 1801 // the file 1802 sect64.offset ? sect64.size : 0, // Size in bytes of 1803 // this section as 1804 // found in the file 1805 sect64.align, 1806 load_cmd.flags)); // Flags for this section 1807 segment_sp->SetIsFake(true); 1808 segment_sp->SetPermissions(segment_permissions); 1809 m_sections_ap->AddSection(segment_sp); 1810 if (add_to_unified) 1811 context.UnifiedList.AddSection(segment_sp); 1812 segment_sp->SetIsEncrypted(segment_is_encrypted); 1813 } 1814 } 1815 assert(segment_sp.get()); 1816 1817 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1818 1819 SectionSP section_sp(new Section( 1820 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1821 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1822 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1823 sect64.flags)); 1824 // Set the section to be encrypted to match the segment 1825 1826 bool section_is_encrypted = false; 1827 if (!segment_is_encrypted && load_cmd.filesize != 0) 1828 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1829 sect64.offset) != NULL; 1830 1831 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1832 section_sp->SetPermissions(segment_permissions); 1833 segment_sp->GetChildren().AddSection(section_sp); 1834 1835 if (segment_sp->IsFake()) { 1836 segment_sp.reset(); 1837 const_segname.Clear(); 1838 } 1839 } 1840 } 1841 if (segment_sp && is_dsym) { 1842 if (first_segment_sectID <= context.NextSectionIdx) { 1843 lldb::user_id_t sect_uid; 1844 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1845 ++sect_uid) { 1846 SectionSP curr_section_sp( 1847 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1848 SectionSP next_section_sp; 1849 if (sect_uid + 1 <= context.NextSectionIdx) 1850 next_section_sp = 1851 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1852 1853 if (curr_section_sp.get()) { 1854 if (curr_section_sp->GetByteSize() == 0) { 1855 if (next_section_sp.get() != NULL) 1856 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1857 curr_section_sp->GetFileAddress()); 1858 else 1859 curr_section_sp->SetByteSize(load_cmd.vmsize); 1860 } 1861 } 1862 } 1863 } 1864 } 1865 } 1866 1867 void ObjectFileMachO::ProcessDysymtabCommand(const load_command &load_cmd, 1868 lldb::offset_t offset) { 1869 m_dysymtab.cmd = load_cmd.cmd; 1870 m_dysymtab.cmdsize = load_cmd.cmdsize; 1871 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1872 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1873 } 1874 1875 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1876 if (m_sections_ap) 1877 return; 1878 1879 m_sections_ap.reset(new SectionList()); 1880 1881 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1882 // bool dump_sections = false; 1883 ModuleSP module_sp(GetModule()); 1884 1885 offset = MachHeaderSizeFromMagic(m_header.magic); 1886 1887 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1888 struct load_command load_cmd; 1889 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1890 const lldb::offset_t load_cmd_offset = offset; 1891 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1892 break; 1893 1894 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1895 ProcessSegmentCommand(load_cmd, offset, i, context); 1896 else if (load_cmd.cmd == LC_DYSYMTAB) 1897 ProcessDysymtabCommand(load_cmd, offset); 1898 1899 offset = load_cmd_offset + load_cmd.cmdsize; 1900 } 1901 1902 if (context.FileAddressesChanged && module_sp) 1903 module_sp->SectionFileAddressesChanged(); 1904 } 1905 1906 class MachSymtabSectionInfo { 1907 public: 1908 MachSymtabSectionInfo(SectionList *section_list) 1909 : m_section_list(section_list), m_section_infos() { 1910 // Get the number of sections down to a depth of 1 to include all segments 1911 // and their sections, but no other sections that may be added for debug 1912 // map or 1913 m_section_infos.resize(section_list->GetNumSections(1)); 1914 } 1915 1916 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1917 if (n_sect == 0) 1918 return SectionSP(); 1919 if (n_sect < m_section_infos.size()) { 1920 if (!m_section_infos[n_sect].section_sp) { 1921 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1922 m_section_infos[n_sect].section_sp = section_sp; 1923 if (section_sp) { 1924 m_section_infos[n_sect].vm_range.SetBaseAddress( 1925 section_sp->GetFileAddress()); 1926 m_section_infos[n_sect].vm_range.SetByteSize( 1927 section_sp->GetByteSize()); 1928 } else { 1929 Host::SystemLog(Host::eSystemLogError, 1930 "error: unable to find section for section %u\n", 1931 n_sect); 1932 } 1933 } 1934 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1935 // Symbol is in section. 1936 return m_section_infos[n_sect].section_sp; 1937 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1938 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1939 file_addr) { 1940 // Symbol is in section with zero size, but has the same start address 1941 // as the section. This can happen with linker symbols (symbols that 1942 // start with the letter 'l' or 'L'. 1943 return m_section_infos[n_sect].section_sp; 1944 } 1945 } 1946 return m_section_list->FindSectionContainingFileAddress(file_addr); 1947 } 1948 1949 protected: 1950 struct SectionInfo { 1951 SectionInfo() : vm_range(), section_sp() {} 1952 1953 VMRange vm_range; 1954 SectionSP section_sp; 1955 }; 1956 SectionList *m_section_list; 1957 std::vector<SectionInfo> m_section_infos; 1958 }; 1959 1960 struct TrieEntry { 1961 TrieEntry() 1962 : name(), address(LLDB_INVALID_ADDRESS), flags(0), other(0), 1963 import_name() {} 1964 1965 void Clear() { 1966 name.Clear(); 1967 address = LLDB_INVALID_ADDRESS; 1968 flags = 0; 1969 other = 0; 1970 import_name.Clear(); 1971 } 1972 1973 void Dump() const { 1974 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1975 static_cast<unsigned long long>(address), 1976 static_cast<unsigned long long>(flags), 1977 static_cast<unsigned long long>(other), name.GetCString()); 1978 if (import_name) 1979 printf(" -> \"%s\"\n", import_name.GetCString()); 1980 else 1981 printf("\n"); 1982 } 1983 ConstString name; 1984 uint64_t address; 1985 uint64_t flags; 1986 uint64_t other; 1987 ConstString import_name; 1988 }; 1989 1990 struct TrieEntryWithOffset { 1991 lldb::offset_t nodeOffset; 1992 TrieEntry entry; 1993 1994 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1995 1996 void Dump(uint32_t idx) const { 1997 printf("[%3u] 0x%16.16llx: ", idx, 1998 static_cast<unsigned long long>(nodeOffset)); 1999 entry.Dump(); 2000 } 2001 2002 bool operator<(const TrieEntryWithOffset &other) const { 2003 return (nodeOffset < other.nodeOffset); 2004 } 2005 }; 2006 2007 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2008 const bool is_arm, 2009 std::vector<llvm::StringRef> &nameSlices, 2010 std::set<lldb::addr_t> &resolver_addresses, 2011 std::vector<TrieEntryWithOffset> &output) { 2012 if (!data.ValidOffset(offset)) 2013 return true; 2014 2015 const uint64_t terminalSize = data.GetULEB128(&offset); 2016 lldb::offset_t children_offset = offset + terminalSize; 2017 if (terminalSize != 0) { 2018 TrieEntryWithOffset e(offset); 2019 e.entry.flags = data.GetULEB128(&offset); 2020 const char *import_name = NULL; 2021 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2022 e.entry.address = 0; 2023 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2024 import_name = data.GetCStr(&offset); 2025 } else { 2026 e.entry.address = data.GetULEB128(&offset); 2027 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2028 e.entry.other = data.GetULEB128(&offset); 2029 uint64_t resolver_addr = e.entry.other; 2030 if (is_arm) 2031 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2032 resolver_addresses.insert(resolver_addr); 2033 } else 2034 e.entry.other = 0; 2035 } 2036 // Only add symbols that are reexport symbols with a valid import name 2037 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && 2038 import_name[0]) { 2039 std::string name; 2040 if (!nameSlices.empty()) { 2041 for (auto name_slice : nameSlices) 2042 name.append(name_slice.data(), name_slice.size()); 2043 } 2044 if (name.size() > 1) { 2045 // Skip the leading '_' 2046 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2047 } 2048 if (import_name) { 2049 // Skip the leading '_' 2050 e.entry.import_name.SetCString(import_name + 1); 2051 } 2052 output.push_back(e); 2053 } 2054 } 2055 2056 const uint8_t childrenCount = data.GetU8(&children_offset); 2057 for (uint8_t i = 0; i < childrenCount; ++i) { 2058 const char *cstr = data.GetCStr(&children_offset); 2059 if (cstr) 2060 nameSlices.push_back(llvm::StringRef(cstr)); 2061 else 2062 return false; // Corrupt data 2063 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2064 if (childNodeOffset) { 2065 if (!ParseTrieEntries(data, childNodeOffset, is_arm, nameSlices, 2066 resolver_addresses, output)) { 2067 return false; 2068 } 2069 } 2070 nameSlices.pop_back(); 2071 } 2072 return true; 2073 } 2074 2075 // Read the UUID out of a dyld_shared_cache file on-disk. 2076 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2077 const ByteOrder byte_order, 2078 const uint32_t addr_byte_size) { 2079 UUID dsc_uuid; 2080 DataBufferSP DscData = MapFileData( 2081 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2082 if (!DscData) 2083 return dsc_uuid; 2084 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2085 2086 char version_str[7]; 2087 lldb::offset_t offset = 0; 2088 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2089 version_str[6] = '\0'; 2090 if (strcmp(version_str, "dyld_v") == 0) { 2091 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2092 dsc_uuid = UUID::fromOptionalData( 2093 dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t)); 2094 } 2095 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2096 if (log && dsc_uuid.IsValid()) { 2097 log->Printf("Shared cache %s has UUID %s", dyld_shared_cache.GetPath().c_str(), 2098 dsc_uuid.GetAsString().c_str()); 2099 } 2100 return dsc_uuid; 2101 } 2102 2103 size_t ObjectFileMachO::ParseSymtab() { 2104 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2105 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2106 m_file.GetFilename().AsCString("")); 2107 ModuleSP module_sp(GetModule()); 2108 if (!module_sp) 2109 return 0; 2110 2111 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2112 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2113 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2114 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2115 FunctionStarts function_starts; 2116 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2117 uint32_t i; 2118 FileSpecList dylib_files; 2119 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2120 static const llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2121 static const llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2122 static const llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2123 2124 for (i = 0; i < m_header.ncmds; ++i) { 2125 const lldb::offset_t cmd_offset = offset; 2126 // Read in the load command and load command size 2127 struct load_command lc; 2128 if (m_data.GetU32(&offset, &lc, 2) == NULL) 2129 break; 2130 // Watch for the symbol table load command 2131 switch (lc.cmd) { 2132 case LC_SYMTAB: 2133 symtab_load_command.cmd = lc.cmd; 2134 symtab_load_command.cmdsize = lc.cmdsize; 2135 // Read in the rest of the symtab load command 2136 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2137 0) // fill in symoff, nsyms, stroff, strsize fields 2138 return 0; 2139 if (symtab_load_command.symoff == 0) { 2140 if (log) 2141 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2142 return 0; 2143 } 2144 2145 if (symtab_load_command.stroff == 0) { 2146 if (log) 2147 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2148 return 0; 2149 } 2150 2151 if (symtab_load_command.nsyms == 0) { 2152 if (log) 2153 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2154 return 0; 2155 } 2156 2157 if (symtab_load_command.strsize == 0) { 2158 if (log) 2159 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2160 return 0; 2161 } 2162 break; 2163 2164 case LC_DYLD_INFO: 2165 case LC_DYLD_INFO_ONLY: 2166 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2167 dyld_info.cmd = lc.cmd; 2168 dyld_info.cmdsize = lc.cmdsize; 2169 } else { 2170 memset(&dyld_info, 0, sizeof(dyld_info)); 2171 } 2172 break; 2173 2174 case LC_LOAD_DYLIB: 2175 case LC_LOAD_WEAK_DYLIB: 2176 case LC_REEXPORT_DYLIB: 2177 case LC_LOADFVMLIB: 2178 case LC_LOAD_UPWARD_DYLIB: { 2179 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2180 const char *path = m_data.PeekCStr(name_offset); 2181 if (path) { 2182 FileSpec file_spec(path); 2183 // Strip the path if there is @rpath, @executable, etc so we just use 2184 // the basename 2185 if (path[0] == '@') 2186 file_spec.GetDirectory().Clear(); 2187 2188 if (lc.cmd == LC_REEXPORT_DYLIB) { 2189 m_reexported_dylibs.AppendIfUnique(file_spec); 2190 } 2191 2192 dylib_files.Append(file_spec); 2193 } 2194 } break; 2195 2196 case LC_FUNCTION_STARTS: 2197 function_starts_load_command.cmd = lc.cmd; 2198 function_starts_load_command.cmdsize = lc.cmdsize; 2199 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2200 NULL) // fill in symoff, nsyms, stroff, strsize fields 2201 memset(&function_starts_load_command, 0, 2202 sizeof(function_starts_load_command)); 2203 break; 2204 2205 default: 2206 break; 2207 } 2208 offset = cmd_offset + lc.cmdsize; 2209 } 2210 2211 if (symtab_load_command.cmd) { 2212 Symtab *symtab = m_symtab_ap.get(); 2213 SectionList *section_list = GetSectionList(); 2214 if (section_list == NULL) 2215 return 0; 2216 2217 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2218 const ByteOrder byte_order = m_data.GetByteOrder(); 2219 bool bit_width_32 = addr_byte_size == 4; 2220 const size_t nlist_byte_size = 2221 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2222 2223 DataExtractor nlist_data(NULL, 0, byte_order, addr_byte_size); 2224 DataExtractor strtab_data(NULL, 0, byte_order, addr_byte_size); 2225 DataExtractor function_starts_data(NULL, 0, byte_order, addr_byte_size); 2226 DataExtractor indirect_symbol_index_data(NULL, 0, byte_order, 2227 addr_byte_size); 2228 DataExtractor dyld_trie_data(NULL, 0, byte_order, addr_byte_size); 2229 2230 const addr_t nlist_data_byte_size = 2231 symtab_load_command.nsyms * nlist_byte_size; 2232 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2233 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2234 2235 ProcessSP process_sp(m_process_wp.lock()); 2236 Process *process = process_sp.get(); 2237 2238 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2239 2240 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2241 Target &target = process->GetTarget(); 2242 2243 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2244 2245 SectionSP linkedit_section_sp( 2246 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2247 // Reading mach file from memory in a process or core file... 2248 2249 if (linkedit_section_sp) { 2250 addr_t linkedit_load_addr = 2251 linkedit_section_sp->GetLoadBaseAddress(&target); 2252 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2253 // We might be trying to access the symbol table before the 2254 // __LINKEDIT's load address has been set in the target. We can't 2255 // fail to read the symbol table, so calculate the right address 2256 // manually 2257 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2258 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2259 } 2260 2261 const addr_t linkedit_file_offset = 2262 linkedit_section_sp->GetFileOffset(); 2263 const addr_t symoff_addr = linkedit_load_addr + 2264 symtab_load_command.symoff - 2265 linkedit_file_offset; 2266 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2267 linkedit_file_offset; 2268 2269 bool data_was_read = false; 2270 2271 #if defined(__APPLE__) && \ 2272 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2273 if (m_header.flags & 0x80000000u && 2274 process->GetAddressByteSize() == sizeof(void *)) { 2275 // This mach-o memory file is in the dyld shared cache. If this 2276 // program is not remote and this is iOS, then this process will 2277 // share the same shared cache as the process we are debugging and we 2278 // can read the entire __LINKEDIT from the address space in this 2279 // process. This is a needed optimization that is used for local iOS 2280 // debugging only since all shared libraries in the shared cache do 2281 // not have corresponding files that exist in the file system of the 2282 // device. They have been combined into a single file. This means we 2283 // always have to load these files from memory. All of the symbol and 2284 // string tables from all of the __LINKEDIT sections from the shared 2285 // libraries in the shared cache have been merged into a single large 2286 // symbol and string table. Reading all of this symbol and string 2287 // table data across can slow down debug launch times, so we optimize 2288 // this by reading the memory for the __LINKEDIT section from this 2289 // process. 2290 2291 UUID lldb_shared_cache; 2292 addr_t lldb_shared_cache_addr; 2293 GetLLDBSharedCacheUUID (lldb_shared_cache_addr, lldb_shared_cache); 2294 UUID process_shared_cache; 2295 addr_t process_shared_cache_addr; 2296 GetProcessSharedCacheUUID(process, process_shared_cache_addr, process_shared_cache); 2297 bool use_lldb_cache = true; 2298 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2299 (lldb_shared_cache != process_shared_cache 2300 || process_shared_cache_addr != lldb_shared_cache_addr)) { 2301 use_lldb_cache = false; 2302 } 2303 2304 PlatformSP platform_sp(target.GetPlatform()); 2305 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2306 data_was_read = true; 2307 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2308 eByteOrderLittle); 2309 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2310 eByteOrderLittle); 2311 if (function_starts_load_command.cmd) { 2312 const addr_t func_start_addr = 2313 linkedit_load_addr + function_starts_load_command.dataoff - 2314 linkedit_file_offset; 2315 function_starts_data.SetData( 2316 (void *)func_start_addr, 2317 function_starts_load_command.datasize, eByteOrderLittle); 2318 } 2319 } 2320 } 2321 #endif 2322 2323 if (!data_was_read) { 2324 // Always load dyld - the dynamic linker - from memory if we didn't 2325 // find a binary anywhere else. lldb will not register 2326 // dylib/framework/bundle loads/unloads if we don't have the dyld 2327 // symbols, we force dyld to load from memory despite the user's 2328 // target.memory-module-load-level setting. 2329 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2330 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2331 DataBufferSP nlist_data_sp( 2332 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2333 if (nlist_data_sp) 2334 nlist_data.SetData(nlist_data_sp, 0, 2335 nlist_data_sp->GetByteSize()); 2336 if (m_dysymtab.nindirectsyms != 0) { 2337 const addr_t indirect_syms_addr = linkedit_load_addr + 2338 m_dysymtab.indirectsymoff - 2339 linkedit_file_offset; 2340 DataBufferSP indirect_syms_data_sp( 2341 ReadMemory(process_sp, indirect_syms_addr, 2342 m_dysymtab.nindirectsyms * 4)); 2343 if (indirect_syms_data_sp) 2344 indirect_symbol_index_data.SetData( 2345 indirect_syms_data_sp, 0, 2346 indirect_syms_data_sp->GetByteSize()); 2347 // If this binary is outside the shared cache, 2348 // cache the string table. 2349 // Binaries in the shared cache all share a giant string table, and 2350 // we can't share the string tables across multiple ObjectFileMachO's, 2351 // so we'd end up re-reading this mega-strtab for every binary 2352 // in the shared cache - it would be a big perf problem. 2353 // For binaries outside the shared cache, it's faster to read the 2354 // entire strtab at once instead of piece-by-piece as we process 2355 // the nlist records. 2356 if ((m_header.flags & 0x80000000u) == 0) { 2357 DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, 2358 strtab_data_byte_size)); 2359 if (strtab_data_sp) { 2360 strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 2361 } 2362 } 2363 } 2364 } 2365 if (memory_module_load_level >= 2366 eMemoryModuleLoadLevelPartial) { 2367 if (function_starts_load_command.cmd) { 2368 const addr_t func_start_addr = 2369 linkedit_load_addr + function_starts_load_command.dataoff - 2370 linkedit_file_offset; 2371 DataBufferSP func_start_data_sp( 2372 ReadMemory(process_sp, func_start_addr, 2373 function_starts_load_command.datasize)); 2374 if (func_start_data_sp) 2375 function_starts_data.SetData(func_start_data_sp, 0, 2376 func_start_data_sp->GetByteSize()); 2377 } 2378 } 2379 } 2380 } 2381 } else { 2382 nlist_data.SetData(m_data, symtab_load_command.symoff, 2383 nlist_data_byte_size); 2384 strtab_data.SetData(m_data, symtab_load_command.stroff, 2385 strtab_data_byte_size); 2386 2387 if (dyld_info.export_size > 0) { 2388 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2389 dyld_info.export_size); 2390 } 2391 2392 if (m_dysymtab.nindirectsyms != 0) { 2393 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2394 m_dysymtab.nindirectsyms * 4); 2395 } 2396 if (function_starts_load_command.cmd) { 2397 function_starts_data.SetData(m_data, 2398 function_starts_load_command.dataoff, 2399 function_starts_load_command.datasize); 2400 } 2401 } 2402 2403 if (nlist_data.GetByteSize() == 0 && 2404 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2405 if (log) 2406 module_sp->LogMessage(log, "failed to read nlist data"); 2407 return 0; 2408 } 2409 2410 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2411 if (!have_strtab_data) { 2412 if (process) { 2413 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2414 if (log) 2415 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2416 return 0; 2417 } 2418 } else { 2419 if (log) 2420 module_sp->LogMessage(log, "failed to read strtab data"); 2421 return 0; 2422 } 2423 } 2424 2425 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2426 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2427 const ConstString &g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2428 const ConstString &g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2429 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2430 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2431 SectionSP text_section_sp( 2432 section_list->FindSectionByName(g_segment_name_TEXT)); 2433 SectionSP data_section_sp( 2434 section_list->FindSectionByName(g_segment_name_DATA)); 2435 SectionSP data_dirty_section_sp( 2436 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2437 SectionSP data_const_section_sp( 2438 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2439 SectionSP objc_section_sp( 2440 section_list->FindSectionByName(g_segment_name_OBJC)); 2441 SectionSP eh_frame_section_sp; 2442 if (text_section_sp.get()) 2443 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2444 g_section_name_eh_frame); 2445 else 2446 eh_frame_section_sp = 2447 section_list->FindSectionByName(g_section_name_eh_frame); 2448 2449 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2450 2451 // lldb works best if it knows the start address of all functions in a 2452 // module. Linker symbols or debug info are normally the best source of 2453 // information for start addr / size but they may be stripped in a released 2454 // binary. Two additional sources of information exist in Mach-O binaries: 2455 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2456 // function's start address in the 2457 // binary, relative to the text section. 2458 // eh_frame - the eh_frame FDEs have the start addr & size of 2459 // each function 2460 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2461 // all modern binaries. 2462 // Binaries built to run on older releases may need to use eh_frame 2463 // information. 2464 2465 if (text_section_sp && function_starts_data.GetByteSize()) { 2466 FunctionStarts::Entry function_start_entry; 2467 function_start_entry.data = false; 2468 lldb::offset_t function_start_offset = 0; 2469 function_start_entry.addr = text_section_sp->GetFileAddress(); 2470 uint64_t delta; 2471 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2472 0) { 2473 // Now append the current entry 2474 function_start_entry.addr += delta; 2475 function_starts.Append(function_start_entry); 2476 } 2477 } else { 2478 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2479 // load command claiming an eh_frame but it doesn't actually have the 2480 // eh_frame content. And if we have a dSYM, we don't need to do any of 2481 // this fill-in-the-missing-symbols works anyway - the debug info should 2482 // give us all the functions in the module. 2483 if (text_section_sp.get() && eh_frame_section_sp.get() && 2484 m_type != eTypeDebugInfo) { 2485 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2486 DWARFCallFrameInfo::EH); 2487 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2488 eh_frame.GetFunctionAddressAndSizeVector(functions); 2489 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2490 size_t count = functions.GetSize(); 2491 for (size_t i = 0; i < count; ++i) { 2492 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2493 functions.GetEntryAtIndex(i); 2494 if (func) { 2495 FunctionStarts::Entry function_start_entry; 2496 function_start_entry.addr = func->base - text_base_addr; 2497 function_starts.Append(function_start_entry); 2498 } 2499 } 2500 } 2501 } 2502 2503 const size_t function_starts_count = function_starts.GetSize(); 2504 2505 // For user process binaries (executables, dylibs, frameworks, bundles), if 2506 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2507 // going to assume the binary has been stripped. Don't allow assembly 2508 // language instruction emulation because we don't know proper function 2509 // start boundaries. 2510 // 2511 // For all other types of binaries (kernels, stand-alone bare board 2512 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2513 // sections - we should not make any assumptions about them based on that. 2514 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2515 m_allow_assembly_emulation_unwind_plans = false; 2516 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2517 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2518 2519 if (unwind_or_symbol_log) 2520 module_sp->LogMessage( 2521 unwind_or_symbol_log, 2522 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2523 } 2524 2525 const user_id_t TEXT_eh_frame_sectID = 2526 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2527 : static_cast<user_id_t>(NO_SECT); 2528 2529 lldb::offset_t nlist_data_offset = 0; 2530 2531 uint32_t N_SO_index = UINT32_MAX; 2532 2533 MachSymtabSectionInfo section_info(section_list); 2534 std::vector<uint32_t> N_FUN_indexes; 2535 std::vector<uint32_t> N_NSYM_indexes; 2536 std::vector<uint32_t> N_INCL_indexes; 2537 std::vector<uint32_t> N_BRAC_indexes; 2538 std::vector<uint32_t> N_COMM_indexes; 2539 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2540 typedef std::map<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2541 typedef std::map<const char *, uint32_t> ConstNameToSymbolIndexMap; 2542 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2543 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2544 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2545 // Any symbols that get merged into another will get an entry in this map 2546 // so we know 2547 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2548 uint32_t nlist_idx = 0; 2549 Symbol *symbol_ptr = NULL; 2550 2551 uint32_t sym_idx = 0; 2552 Symbol *sym = NULL; 2553 size_t num_syms = 0; 2554 std::string memory_symbol_name; 2555 uint32_t unmapped_local_symbols_found = 0; 2556 2557 std::vector<TrieEntryWithOffset> trie_entries; 2558 std::set<lldb::addr_t> resolver_addresses; 2559 2560 if (dyld_trie_data.GetByteSize() > 0) { 2561 std::vector<llvm::StringRef> nameSlices; 2562 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, 2563 resolver_addresses, trie_entries); 2564 2565 ConstString text_segment_name("__TEXT"); 2566 SectionSP text_segment_sp = 2567 GetSectionList()->FindSectionByName(text_segment_name); 2568 if (text_segment_sp) { 2569 const lldb::addr_t text_segment_file_addr = 2570 text_segment_sp->GetFileAddress(); 2571 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2572 for (auto &e : trie_entries) 2573 e.entry.address += text_segment_file_addr; 2574 } 2575 } 2576 } 2577 2578 typedef std::set<ConstString> IndirectSymbols; 2579 IndirectSymbols indirect_symbol_names; 2580 2581 #if defined(__APPLE__) && \ 2582 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2583 2584 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2585 // optimized by moving LOCAL symbols out of the memory mapped portion of 2586 // the DSC. The symbol information has all been retained, but it isn't 2587 // available in the normal nlist data. However, there *are* duplicate 2588 // entries of *some* 2589 // LOCAL symbols in the normal nlist data. To handle this situation 2590 // correctly, we must first attempt 2591 // to parse any DSC unmapped symbol information. If we find any, we set a 2592 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2593 2594 if (m_header.flags & 0x80000000u) { 2595 // Before we can start mapping the DSC, we need to make certain the 2596 // target process is actually using the cache we can find. 2597 2598 // Next we need to determine the correct path for the dyld shared cache. 2599 2600 ArchSpec header_arch; 2601 GetArchitecture(header_arch); 2602 char dsc_path[PATH_MAX]; 2603 char dsc_path_development[PATH_MAX]; 2604 2605 snprintf( 2606 dsc_path, sizeof(dsc_path), "%s%s%s", 2607 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2608 */ 2609 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2610 header_arch.GetArchitectureName()); 2611 2612 snprintf( 2613 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2614 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2615 */ 2616 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2617 header_arch.GetArchitectureName(), ".development"); 2618 2619 FileSpec dsc_nondevelopment_filespec(dsc_path, false); 2620 FileSpec dsc_development_filespec(dsc_path_development, false); 2621 FileSpec dsc_filespec; 2622 2623 UUID dsc_uuid; 2624 UUID process_shared_cache_uuid; 2625 addr_t process_shared_cache_base_addr; 2626 2627 if (process) { 2628 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, process_shared_cache_uuid); 2629 } 2630 2631 // First see if we can find an exact match for the inferior process 2632 // shared cache UUID in the development or non-development shared caches 2633 // on disk. 2634 if (process_shared_cache_uuid.IsValid()) { 2635 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2636 UUID dsc_development_uuid = GetSharedCacheUUID( 2637 dsc_development_filespec, byte_order, addr_byte_size); 2638 if (dsc_development_uuid.IsValid() && 2639 dsc_development_uuid == process_shared_cache_uuid) { 2640 dsc_filespec = dsc_development_filespec; 2641 dsc_uuid = dsc_development_uuid; 2642 } 2643 } 2644 if (!dsc_uuid.IsValid() && 2645 FileSystem::Instance().Exists(dsc_nondevelopment_filespec)) { 2646 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2647 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2648 if (dsc_nondevelopment_uuid.IsValid() && 2649 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2650 dsc_filespec = dsc_nondevelopment_filespec; 2651 dsc_uuid = dsc_nondevelopment_uuid; 2652 } 2653 } 2654 } 2655 2656 // Failing a UUID match, prefer the development dyld_shared cache if both 2657 // are present. 2658 if (!FileSystem::Instance().Exists(dsc_filespec)) { 2659 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2660 dsc_filespec = dsc_development_filespec; 2661 } else { 2662 dsc_filespec = dsc_nondevelopment_filespec; 2663 } 2664 } 2665 2666 /* The dyld_cache_header has a pointer to the 2667 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2668 The dyld_cache_local_symbols_info structure gives us three things: 2669 1. The start and count of the nlist records in the dyld_shared_cache 2670 file 2671 2. The start and size of the strings for these nlist records 2672 3. The start and count of dyld_cache_local_symbols_entry entries 2673 2674 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2675 dyld shared cache. 2676 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2677 the dyld shared cache. 2678 The dyld_cache_local_symbols_entry also lists the start of this 2679 dylib/framework's nlist records 2680 and the count of how many nlist records there are for this 2681 dylib/framework. 2682 */ 2683 2684 // Process the dyld shared cache header to find the unmapped symbols 2685 2686 DataBufferSP dsc_data_sp = MapFileData( 2687 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2688 if (!dsc_uuid.IsValid()) { 2689 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2690 } 2691 if (dsc_data_sp) { 2692 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2693 2694 bool uuid_match = true; 2695 if (dsc_uuid.IsValid() && process) { 2696 if (process_shared_cache_uuid.IsValid() && 2697 dsc_uuid != process_shared_cache_uuid) { 2698 // The on-disk dyld_shared_cache file is not the same as the one in 2699 // this process' memory, don't use it. 2700 uuid_match = false; 2701 ModuleSP module_sp(GetModule()); 2702 if (module_sp) 2703 module_sp->ReportWarning("process shared cache does not match " 2704 "on-disk dyld_shared_cache file, some " 2705 "symbol names will be missing."); 2706 } 2707 } 2708 2709 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2710 2711 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2712 2713 // If the mappingOffset points to a location inside the header, we've 2714 // opened an old dyld shared cache, and should not proceed further. 2715 if (uuid_match && 2716 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2717 2718 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2719 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2720 mappingOffset); 2721 2722 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2723 byte_order, addr_byte_size); 2724 offset = 0; 2725 2726 // The File addresses (from the in-memory Mach-O load commands) for 2727 // the shared libraries in the shared library cache need to be 2728 // adjusted by an offset to match up with the dylibOffset identifying 2729 // field in the dyld_cache_local_symbol_entry's. This offset is 2730 // recorded in mapping_offset_value. 2731 const uint64_t mapping_offset_value = 2732 dsc_mapping_info_data.GetU64(&offset); 2733 2734 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, 2735 localSymbolsOffset); 2736 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2737 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2738 2739 if (localSymbolsOffset && localSymbolsSize) { 2740 // Map the local symbols 2741 DataBufferSP dsc_local_symbols_data_sp = 2742 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2743 2744 if (dsc_local_symbols_data_sp) { 2745 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2746 byte_order, addr_byte_size); 2747 2748 offset = 0; 2749 2750 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2751 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2752 UndefinedNameToDescMap undefined_name_to_desc; 2753 SymbolIndexToName reexport_shlib_needs_fixup; 2754 2755 // Read the local_symbols_infos struct in one shot 2756 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2757 dsc_local_symbols_data.GetU32(&offset, 2758 &local_symbols_info.nlistOffset, 6); 2759 2760 SectionSP text_section_sp( 2761 section_list->FindSectionByName(GetSegmentNameTEXT())); 2762 2763 uint32_t header_file_offset = 2764 (text_section_sp->GetFileAddress() - mapping_offset_value); 2765 2766 offset = local_symbols_info.entriesOffset; 2767 for (uint32_t entry_index = 0; 2768 entry_index < local_symbols_info.entriesCount; 2769 entry_index++) { 2770 struct lldb_copy_dyld_cache_local_symbols_entry 2771 local_symbols_entry; 2772 local_symbols_entry.dylibOffset = 2773 dsc_local_symbols_data.GetU32(&offset); 2774 local_symbols_entry.nlistStartIndex = 2775 dsc_local_symbols_data.GetU32(&offset); 2776 local_symbols_entry.nlistCount = 2777 dsc_local_symbols_data.GetU32(&offset); 2778 2779 if (header_file_offset == local_symbols_entry.dylibOffset) { 2780 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2781 2782 // The normal nlist code cannot correctly size the Symbols 2783 // array, we need to allocate it here. 2784 sym = symtab->Resize( 2785 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2786 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2787 num_syms = symtab->GetNumSymbols(); 2788 2789 nlist_data_offset = 2790 local_symbols_info.nlistOffset + 2791 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2792 uint32_t string_table_offset = 2793 local_symbols_info.stringsOffset; 2794 2795 for (uint32_t nlist_index = 0; 2796 nlist_index < local_symbols_entry.nlistCount; 2797 nlist_index++) { 2798 ///////////////////////////// 2799 { 2800 struct nlist_64 nlist; 2801 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize( 2802 nlist_data_offset, nlist_byte_size)) 2803 break; 2804 2805 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked( 2806 &nlist_data_offset); 2807 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked( 2808 &nlist_data_offset); 2809 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked( 2810 &nlist_data_offset); 2811 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked( 2812 &nlist_data_offset); 2813 nlist.n_value = 2814 dsc_local_symbols_data.GetAddress_unchecked( 2815 &nlist_data_offset); 2816 2817 SymbolType type = eSymbolTypeInvalid; 2818 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2819 string_table_offset + nlist.n_strx); 2820 2821 if (symbol_name == NULL) { 2822 // No symbol should be NULL, even the symbols with no 2823 // string values should have an offset zero which 2824 // points to an empty C-string 2825 Host::SystemLog( 2826 Host::eSystemLogError, 2827 "error: DSC unmapped local symbol[%u] has invalid " 2828 "string table offset 0x%x in %s, ignoring symbol\n", 2829 entry_index, nlist.n_strx, 2830 module_sp->GetFileSpec().GetPath().c_str()); 2831 continue; 2832 } 2833 if (symbol_name[0] == '\0') 2834 symbol_name = NULL; 2835 2836 const char *symbol_name_non_abi_mangled = NULL; 2837 2838 SectionSP symbol_section; 2839 uint32_t symbol_byte_size = 0; 2840 bool add_nlist = true; 2841 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2842 bool demangled_is_synthesized = false; 2843 bool is_gsym = false; 2844 bool set_value = true; 2845 2846 assert(sym_idx < num_syms); 2847 2848 sym[sym_idx].SetDebug(is_debug); 2849 2850 if (is_debug) { 2851 switch (nlist.n_type) { 2852 case N_GSYM: 2853 // global symbol: name,,NO_SECT,type,0 2854 // Sometimes the N_GSYM value contains the address. 2855 2856 // FIXME: In the .o files, we have a GSYM and a debug 2857 // symbol for all the ObjC data. They 2858 // have the same address, but we want to ensure that 2859 // we always find only the real symbol, 'cause we 2860 // don't currently correctly attribute the 2861 // GSYM one to the ObjCClass/Ivar/MetaClass 2862 // symbol type. This is a temporary hack to make 2863 // sure the ObjectiveC symbols get treated correctly. 2864 // To do this right, we should coalesce all the GSYM 2865 // & global symbols that have the same address. 2866 2867 is_gsym = true; 2868 sym[sym_idx].SetExternal(true); 2869 2870 if (symbol_name && symbol_name[0] == '_' && 2871 symbol_name[1] == 'O') { 2872 llvm::StringRef symbol_name_ref(symbol_name); 2873 if (symbol_name_ref.startswith( 2874 g_objc_v2_prefix_class)) { 2875 symbol_name_non_abi_mangled = symbol_name + 1; 2876 symbol_name = 2877 symbol_name + g_objc_v2_prefix_class.size(); 2878 type = eSymbolTypeObjCClass; 2879 demangled_is_synthesized = true; 2880 2881 } else if (symbol_name_ref.startswith( 2882 g_objc_v2_prefix_metaclass)) { 2883 symbol_name_non_abi_mangled = symbol_name + 1; 2884 symbol_name = symbol_name + 2885 g_objc_v2_prefix_metaclass.size(); 2886 type = eSymbolTypeObjCMetaClass; 2887 demangled_is_synthesized = true; 2888 } else if (symbol_name_ref.startswith( 2889 g_objc_v2_prefix_ivar)) { 2890 symbol_name_non_abi_mangled = symbol_name + 1; 2891 symbol_name = 2892 symbol_name + g_objc_v2_prefix_ivar.size(); 2893 type = eSymbolTypeObjCIVar; 2894 demangled_is_synthesized = true; 2895 } 2896 } else { 2897 if (nlist.n_value != 0) 2898 symbol_section = section_info.GetSection( 2899 nlist.n_sect, nlist.n_value); 2900 type = eSymbolTypeData; 2901 } 2902 break; 2903 2904 case N_FNAME: 2905 // procedure name (f77 kludge): name,,NO_SECT,0,0 2906 type = eSymbolTypeCompiler; 2907 break; 2908 2909 case N_FUN: 2910 // procedure: name,,n_sect,linenumber,address 2911 if (symbol_name) { 2912 type = eSymbolTypeCode; 2913 symbol_section = section_info.GetSection( 2914 nlist.n_sect, nlist.n_value); 2915 2916 N_FUN_addr_to_sym_idx.insert( 2917 std::make_pair(nlist.n_value, sym_idx)); 2918 // We use the current number of symbols in the 2919 // symbol table in lieu of using nlist_idx in case 2920 // we ever start trimming entries out 2921 N_FUN_indexes.push_back(sym_idx); 2922 } else { 2923 type = eSymbolTypeCompiler; 2924 2925 if (!N_FUN_indexes.empty()) { 2926 // Copy the size of the function into the 2927 // original 2928 // STAB entry so we don't have 2929 // to hunt for it later 2930 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2931 ->SetByteSize(nlist.n_value); 2932 N_FUN_indexes.pop_back(); 2933 // We don't really need the end function STAB as 2934 // it contains the size which we already placed 2935 // with the original symbol, so don't add it if 2936 // we want a minimal symbol table 2937 add_nlist = false; 2938 } 2939 } 2940 break; 2941 2942 case N_STSYM: 2943 // static symbol: name,,n_sect,type,address 2944 N_STSYM_addr_to_sym_idx.insert( 2945 std::make_pair(nlist.n_value, sym_idx)); 2946 symbol_section = section_info.GetSection( 2947 nlist.n_sect, nlist.n_value); 2948 if (symbol_name && symbol_name[0]) { 2949 type = ObjectFile::GetSymbolTypeFromName( 2950 symbol_name + 1, eSymbolTypeData); 2951 } 2952 break; 2953 2954 case N_LCSYM: 2955 // .lcomm symbol: name,,n_sect,type,address 2956 symbol_section = section_info.GetSection( 2957 nlist.n_sect, nlist.n_value); 2958 type = eSymbolTypeCommonBlock; 2959 break; 2960 2961 case N_BNSYM: 2962 // We use the current number of symbols in the symbol 2963 // table in lieu of using nlist_idx in case we ever 2964 // start trimming entries out Skip these if we want 2965 // minimal symbol tables 2966 add_nlist = false; 2967 break; 2968 2969 case N_ENSYM: 2970 // Set the size of the N_BNSYM to the terminating 2971 // index of this N_ENSYM so that we can always skip 2972 // the entire symbol if we need to navigate more 2973 // quickly at the source level when parsing STABS 2974 // Skip these if we want minimal symbol tables 2975 add_nlist = false; 2976 break; 2977 2978 case N_OPT: 2979 // emitted with gcc2_compiled and in gcc source 2980 type = eSymbolTypeCompiler; 2981 break; 2982 2983 case N_RSYM: 2984 // register sym: name,,NO_SECT,type,register 2985 type = eSymbolTypeVariable; 2986 break; 2987 2988 case N_SLINE: 2989 // src line: 0,,n_sect,linenumber,address 2990 symbol_section = section_info.GetSection( 2991 nlist.n_sect, nlist.n_value); 2992 type = eSymbolTypeLineEntry; 2993 break; 2994 2995 case N_SSYM: 2996 // structure elt: name,,NO_SECT,type,struct_offset 2997 type = eSymbolTypeVariableType; 2998 break; 2999 3000 case N_SO: 3001 // source file name 3002 type = eSymbolTypeSourceFile; 3003 if (symbol_name == NULL) { 3004 add_nlist = false; 3005 if (N_SO_index != UINT32_MAX) { 3006 // Set the size of the N_SO to the terminating 3007 // index of this N_SO so that we can always skip 3008 // the entire N_SO if we need to navigate more 3009 // quickly at the source level when parsing STABS 3010 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3011 symbol_ptr->SetByteSize(sym_idx); 3012 symbol_ptr->SetSizeIsSibling(true); 3013 } 3014 N_NSYM_indexes.clear(); 3015 N_INCL_indexes.clear(); 3016 N_BRAC_indexes.clear(); 3017 N_COMM_indexes.clear(); 3018 N_FUN_indexes.clear(); 3019 N_SO_index = UINT32_MAX; 3020 } else { 3021 // We use the current number of symbols in the 3022 // symbol table in lieu of using nlist_idx in case 3023 // we ever start trimming entries out 3024 const bool N_SO_has_full_path = 3025 symbol_name[0] == '/'; 3026 if (N_SO_has_full_path) { 3027 if ((N_SO_index == sym_idx - 1) && 3028 ((sym_idx - 1) < num_syms)) { 3029 // We have two consecutive N_SO entries where 3030 // the first contains a directory and the 3031 // second contains a full path. 3032 sym[sym_idx - 1].GetMangled().SetValue( 3033 ConstString(symbol_name), false); 3034 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3035 add_nlist = false; 3036 } else { 3037 // This is the first entry in a N_SO that 3038 // contains a directory or 3039 // a full path to the source file 3040 N_SO_index = sym_idx; 3041 } 3042 } else if ((N_SO_index == sym_idx - 1) && 3043 ((sym_idx - 1) < num_syms)) { 3044 // This is usually the second N_SO entry that 3045 // contains just the filename, so here we combine 3046 // it with the first one if we are minimizing the 3047 // symbol table 3048 const char *so_path = 3049 sym[sym_idx - 1] 3050 .GetMangled() 3051 .GetDemangledName( 3052 lldb::eLanguageTypeUnknown) 3053 .AsCString(); 3054 if (so_path && so_path[0]) { 3055 std::string full_so_path(so_path); 3056 const size_t double_slash_pos = 3057 full_so_path.find("//"); 3058 if (double_slash_pos != std::string::npos) { 3059 // The linker has been generating bad N_SO 3060 // entries with doubled up paths 3061 // in the format "%s%s" where the first 3062 // string in the DW_AT_comp_dir, and the 3063 // second is the directory for the source 3064 // file so you end up with a path that looks 3065 // like "/tmp/src//tmp/src/" 3066 FileSpec so_dir(so_path, false); 3067 if (!FileSystem::Instance().Exists(so_dir)) { 3068 so_dir.SetFile( 3069 &full_so_path[double_slash_pos + 1], 3070 false); 3071 if (FileSystem::Instance().Exists(so_dir)) { 3072 // Trim off the incorrect path 3073 full_so_path.erase(0, 3074 double_slash_pos + 1); 3075 } 3076 } 3077 } 3078 if (*full_so_path.rbegin() != '/') 3079 full_so_path += '/'; 3080 full_so_path += symbol_name; 3081 sym[sym_idx - 1].GetMangled().SetValue( 3082 ConstString(full_so_path.c_str()), false); 3083 add_nlist = false; 3084 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3085 } 3086 } else { 3087 // This could be a relative path to a N_SO 3088 N_SO_index = sym_idx; 3089 } 3090 } 3091 break; 3092 3093 case N_OSO: 3094 // object file name: name,,0,0,st_mtime 3095 type = eSymbolTypeObjectFile; 3096 break; 3097 3098 case N_LSYM: 3099 // local sym: name,,NO_SECT,type,offset 3100 type = eSymbolTypeLocal; 3101 break; 3102 3103 //---------------------------------------------------------------------- 3104 // INCL scopes 3105 //---------------------------------------------------------------------- 3106 case N_BINCL: 3107 // include file beginning: name,,NO_SECT,0,sum We use 3108 // the current number of symbols in the symbol table 3109 // in lieu of using nlist_idx in case we ever start 3110 // trimming entries out 3111 N_INCL_indexes.push_back(sym_idx); 3112 type = eSymbolTypeScopeBegin; 3113 break; 3114 3115 case N_EINCL: 3116 // include file end: name,,NO_SECT,0,0 3117 // Set the size of the N_BINCL to the terminating 3118 // index of this N_EINCL so that we can always skip 3119 // the entire symbol if we need to navigate more 3120 // quickly at the source level when parsing STABS 3121 if (!N_INCL_indexes.empty()) { 3122 symbol_ptr = 3123 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3124 symbol_ptr->SetByteSize(sym_idx + 1); 3125 symbol_ptr->SetSizeIsSibling(true); 3126 N_INCL_indexes.pop_back(); 3127 } 3128 type = eSymbolTypeScopeEnd; 3129 break; 3130 3131 case N_SOL: 3132 // #included file name: name,,n_sect,0,address 3133 type = eSymbolTypeHeaderFile; 3134 3135 // We currently don't use the header files on darwin 3136 add_nlist = false; 3137 break; 3138 3139 case N_PARAMS: 3140 // compiler parameters: name,,NO_SECT,0,0 3141 type = eSymbolTypeCompiler; 3142 break; 3143 3144 case N_VERSION: 3145 // compiler version: name,,NO_SECT,0,0 3146 type = eSymbolTypeCompiler; 3147 break; 3148 3149 case N_OLEVEL: 3150 // compiler -O level: name,,NO_SECT,0,0 3151 type = eSymbolTypeCompiler; 3152 break; 3153 3154 case N_PSYM: 3155 // parameter: name,,NO_SECT,type,offset 3156 type = eSymbolTypeVariable; 3157 break; 3158 3159 case N_ENTRY: 3160 // alternate entry: name,,n_sect,linenumber,address 3161 symbol_section = section_info.GetSection( 3162 nlist.n_sect, nlist.n_value); 3163 type = eSymbolTypeLineEntry; 3164 break; 3165 3166 //---------------------------------------------------------------------- 3167 // Left and Right Braces 3168 //---------------------------------------------------------------------- 3169 case N_LBRAC: 3170 // left bracket: 0,,NO_SECT,nesting level,address We 3171 // use the current number of symbols in the symbol 3172 // table in lieu of using nlist_idx in case we ever 3173 // start trimming entries out 3174 symbol_section = section_info.GetSection( 3175 nlist.n_sect, nlist.n_value); 3176 N_BRAC_indexes.push_back(sym_idx); 3177 type = eSymbolTypeScopeBegin; 3178 break; 3179 3180 case N_RBRAC: 3181 // right bracket: 0,,NO_SECT,nesting level,address 3182 // Set the size of the N_LBRAC to the terminating 3183 // index of this N_RBRAC so that we can always skip 3184 // the entire symbol if we need to navigate more 3185 // quickly at the source level when parsing STABS 3186 symbol_section = section_info.GetSection( 3187 nlist.n_sect, nlist.n_value); 3188 if (!N_BRAC_indexes.empty()) { 3189 symbol_ptr = 3190 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3191 symbol_ptr->SetByteSize(sym_idx + 1); 3192 symbol_ptr->SetSizeIsSibling(true); 3193 N_BRAC_indexes.pop_back(); 3194 } 3195 type = eSymbolTypeScopeEnd; 3196 break; 3197 3198 case N_EXCL: 3199 // deleted include file: name,,NO_SECT,0,sum 3200 type = eSymbolTypeHeaderFile; 3201 break; 3202 3203 //---------------------------------------------------------------------- 3204 // COMM scopes 3205 //---------------------------------------------------------------------- 3206 case N_BCOMM: 3207 // begin common: name,,NO_SECT,0,0 3208 // We use the current number of symbols in the symbol 3209 // table in lieu of using nlist_idx in case we ever 3210 // start trimming entries out 3211 type = eSymbolTypeScopeBegin; 3212 N_COMM_indexes.push_back(sym_idx); 3213 break; 3214 3215 case N_ECOML: 3216 // end common (local name): 0,,n_sect,0,address 3217 symbol_section = section_info.GetSection( 3218 nlist.n_sect, nlist.n_value); 3219 // Fall through 3220 3221 case N_ECOMM: 3222 // end common: name,,n_sect,0,0 3223 // Set the size of the N_BCOMM to the terminating 3224 // index of this N_ECOMM/N_ECOML so that we can 3225 // always skip the entire symbol if we need to 3226 // navigate more quickly at the source level when 3227 // parsing STABS 3228 if (!N_COMM_indexes.empty()) { 3229 symbol_ptr = 3230 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3231 symbol_ptr->SetByteSize(sym_idx + 1); 3232 symbol_ptr->SetSizeIsSibling(true); 3233 N_COMM_indexes.pop_back(); 3234 } 3235 type = eSymbolTypeScopeEnd; 3236 break; 3237 3238 case N_LENG: 3239 // second stab entry with length information 3240 type = eSymbolTypeAdditional; 3241 break; 3242 3243 default: 3244 break; 3245 } 3246 } else { 3247 // uint8_t n_pext = N_PEXT & nlist.n_type; 3248 uint8_t n_type = N_TYPE & nlist.n_type; 3249 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3250 3251 switch (n_type) { 3252 case N_INDR: { 3253 const char *reexport_name_cstr = 3254 strtab_data.PeekCStr(nlist.n_value); 3255 if (reexport_name_cstr && reexport_name_cstr[0]) { 3256 type = eSymbolTypeReExported; 3257 ConstString reexport_name( 3258 reexport_name_cstr + 3259 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3260 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3261 set_value = false; 3262 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3263 indirect_symbol_names.insert( 3264 ConstString(symbol_name + 3265 ((symbol_name[0] == '_') ? 1 : 0))); 3266 } else 3267 type = eSymbolTypeUndefined; 3268 } break; 3269 3270 case N_UNDF: 3271 if (symbol_name && symbol_name[0]) { 3272 ConstString undefined_name( 3273 symbol_name + 3274 ((symbol_name[0] == '_') ? 1 : 0)); 3275 undefined_name_to_desc[undefined_name] = 3276 nlist.n_desc; 3277 } 3278 // Fall through 3279 case N_PBUD: 3280 type = eSymbolTypeUndefined; 3281 break; 3282 3283 case N_ABS: 3284 type = eSymbolTypeAbsolute; 3285 break; 3286 3287 case N_SECT: { 3288 symbol_section = section_info.GetSection( 3289 nlist.n_sect, nlist.n_value); 3290 3291 if (symbol_section == NULL) { 3292 // TODO: warn about this? 3293 add_nlist = false; 3294 break; 3295 } 3296 3297 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3298 type = eSymbolTypeException; 3299 } else { 3300 uint32_t section_type = 3301 symbol_section->Get() & SECTION_TYPE; 3302 3303 switch (section_type) { 3304 case S_CSTRING_LITERALS: 3305 type = eSymbolTypeData; 3306 break; // section with only literal C strings 3307 case S_4BYTE_LITERALS: 3308 type = eSymbolTypeData; 3309 break; // section with only 4 byte literals 3310 case S_8BYTE_LITERALS: 3311 type = eSymbolTypeData; 3312 break; // section with only 8 byte literals 3313 case S_LITERAL_POINTERS: 3314 type = eSymbolTypeTrampoline; 3315 break; // section with only pointers to literals 3316 case S_NON_LAZY_SYMBOL_POINTERS: 3317 type = eSymbolTypeTrampoline; 3318 break; // section with only non-lazy symbol 3319 // pointers 3320 case S_LAZY_SYMBOL_POINTERS: 3321 type = eSymbolTypeTrampoline; 3322 break; // section with only lazy symbol pointers 3323 case S_SYMBOL_STUBS: 3324 type = eSymbolTypeTrampoline; 3325 break; // section with only symbol stubs, byte 3326 // size of stub in the reserved2 field 3327 case S_MOD_INIT_FUNC_POINTERS: 3328 type = eSymbolTypeCode; 3329 break; // section with only function pointers for 3330 // initialization 3331 case S_MOD_TERM_FUNC_POINTERS: 3332 type = eSymbolTypeCode; 3333 break; // section with only function pointers for 3334 // termination 3335 case S_INTERPOSING: 3336 type = eSymbolTypeTrampoline; 3337 break; // section with only pairs of function 3338 // pointers for interposing 3339 case S_16BYTE_LITERALS: 3340 type = eSymbolTypeData; 3341 break; // section with only 16 byte literals 3342 case S_DTRACE_DOF: 3343 type = eSymbolTypeInstrumentation; 3344 break; 3345 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3346 type = eSymbolTypeTrampoline; 3347 break; 3348 default: 3349 switch (symbol_section->GetType()) { 3350 case lldb::eSectionTypeCode: 3351 type = eSymbolTypeCode; 3352 break; 3353 case eSectionTypeData: 3354 case eSectionTypeDataCString: // Inlined C string 3355 // data 3356 case eSectionTypeDataCStringPointers: // Pointers 3357 // to C 3358 // string 3359 // data 3360 case eSectionTypeDataSymbolAddress: // Address of 3361 // a symbol in 3362 // the symbol 3363 // table 3364 case eSectionTypeData4: 3365 case eSectionTypeData8: 3366 case eSectionTypeData16: 3367 type = eSymbolTypeData; 3368 break; 3369 default: 3370 break; 3371 } 3372 break; 3373 } 3374 3375 if (type == eSymbolTypeInvalid) { 3376 const char *symbol_sect_name = 3377 symbol_section->GetName().AsCString(); 3378 if (symbol_section->IsDescendant( 3379 text_section_sp.get())) { 3380 if (symbol_section->IsClear( 3381 S_ATTR_PURE_INSTRUCTIONS | 3382 S_ATTR_SELF_MODIFYING_CODE | 3383 S_ATTR_SOME_INSTRUCTIONS)) 3384 type = eSymbolTypeData; 3385 else 3386 type = eSymbolTypeCode; 3387 } else if (symbol_section->IsDescendant( 3388 data_section_sp.get()) || 3389 symbol_section->IsDescendant( 3390 data_dirty_section_sp.get()) || 3391 symbol_section->IsDescendant( 3392 data_const_section_sp.get())) { 3393 if (symbol_sect_name && 3394 ::strstr(symbol_sect_name, "__objc") == 3395 symbol_sect_name) { 3396 type = eSymbolTypeRuntime; 3397 3398 if (symbol_name) { 3399 llvm::StringRef symbol_name_ref( 3400 symbol_name); 3401 if (symbol_name_ref.startswith("_OBJC_")) { 3402 static const llvm::StringRef 3403 g_objc_v2_prefix_class( 3404 "_OBJC_CLASS_$_"); 3405 static const llvm::StringRef 3406 g_objc_v2_prefix_metaclass( 3407 "_OBJC_METACLASS_$_"); 3408 static const llvm::StringRef 3409 g_objc_v2_prefix_ivar( 3410 "_OBJC_IVAR_$_"); 3411 if (symbol_name_ref.startswith( 3412 g_objc_v2_prefix_class)) { 3413 symbol_name_non_abi_mangled = 3414 symbol_name + 1; 3415 symbol_name = 3416 symbol_name + 3417 g_objc_v2_prefix_class.size(); 3418 type = eSymbolTypeObjCClass; 3419 demangled_is_synthesized = true; 3420 } else if ( 3421 symbol_name_ref.startswith( 3422 g_objc_v2_prefix_metaclass)) { 3423 symbol_name_non_abi_mangled = 3424 symbol_name + 1; 3425 symbol_name = 3426 symbol_name + 3427 g_objc_v2_prefix_metaclass.size(); 3428 type = eSymbolTypeObjCMetaClass; 3429 demangled_is_synthesized = true; 3430 } else if (symbol_name_ref.startswith( 3431 g_objc_v2_prefix_ivar)) { 3432 symbol_name_non_abi_mangled = 3433 symbol_name + 1; 3434 symbol_name = 3435 symbol_name + 3436 g_objc_v2_prefix_ivar.size(); 3437 type = eSymbolTypeObjCIVar; 3438 demangled_is_synthesized = true; 3439 } 3440 } 3441 } 3442 } else if (symbol_sect_name && 3443 ::strstr(symbol_sect_name, 3444 "__gcc_except_tab") == 3445 symbol_sect_name) { 3446 type = eSymbolTypeException; 3447 } else { 3448 type = eSymbolTypeData; 3449 } 3450 } else if (symbol_sect_name && 3451 ::strstr(symbol_sect_name, 3452 "__IMPORT") == 3453 symbol_sect_name) { 3454 type = eSymbolTypeTrampoline; 3455 } else if (symbol_section->IsDescendant( 3456 objc_section_sp.get())) { 3457 type = eSymbolTypeRuntime; 3458 if (symbol_name && symbol_name[0] == '.') { 3459 llvm::StringRef symbol_name_ref(symbol_name); 3460 static const llvm::StringRef 3461 g_objc_v1_prefix_class( 3462 ".objc_class_name_"); 3463 if (symbol_name_ref.startswith( 3464 g_objc_v1_prefix_class)) { 3465 symbol_name_non_abi_mangled = symbol_name; 3466 symbol_name = symbol_name + 3467 g_objc_v1_prefix_class.size(); 3468 type = eSymbolTypeObjCClass; 3469 demangled_is_synthesized = true; 3470 } 3471 } 3472 } 3473 } 3474 } 3475 } break; 3476 } 3477 } 3478 3479 if (add_nlist) { 3480 uint64_t symbol_value = nlist.n_value; 3481 if (symbol_name_non_abi_mangled) { 3482 sym[sym_idx].GetMangled().SetMangledName( 3483 ConstString(symbol_name_non_abi_mangled)); 3484 sym[sym_idx].GetMangled().SetDemangledName( 3485 ConstString(symbol_name)); 3486 } else { 3487 bool symbol_name_is_mangled = false; 3488 3489 if (symbol_name && symbol_name[0] == '_') { 3490 symbol_name_is_mangled = symbol_name[1] == '_'; 3491 symbol_name++; // Skip the leading underscore 3492 } 3493 3494 if (symbol_name) { 3495 ConstString const_symbol_name(symbol_name); 3496 sym[sym_idx].GetMangled().SetValue( 3497 const_symbol_name, symbol_name_is_mangled); 3498 if (is_gsym && is_debug) { 3499 const char *gsym_name = 3500 sym[sym_idx] 3501 .GetMangled() 3502 .GetName(lldb::eLanguageTypeUnknown, 3503 Mangled::ePreferMangled) 3504 .GetCString(); 3505 if (gsym_name) 3506 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3507 } 3508 } 3509 } 3510 if (symbol_section) { 3511 const addr_t section_file_addr = 3512 symbol_section->GetFileAddress(); 3513 if (symbol_byte_size == 0 && 3514 function_starts_count > 0) { 3515 addr_t symbol_lookup_file_addr = nlist.n_value; 3516 // Do an exact address match for non-ARM addresses, 3517 // else get the closest since the symbol might be a 3518 // thumb symbol which has an address with bit zero 3519 // set 3520 FunctionStarts::Entry *func_start_entry = 3521 function_starts.FindEntry( 3522 symbol_lookup_file_addr, !is_arm); 3523 if (is_arm && func_start_entry) { 3524 // Verify that the function start address is the 3525 // symbol address (ARM) or the symbol address + 1 3526 // (thumb) 3527 if (func_start_entry->addr != 3528 symbol_lookup_file_addr && 3529 func_start_entry->addr != 3530 (symbol_lookup_file_addr + 1)) { 3531 // Not the right entry, NULL it out... 3532 func_start_entry = NULL; 3533 } 3534 } 3535 if (func_start_entry) { 3536 func_start_entry->data = true; 3537 3538 addr_t symbol_file_addr = func_start_entry->addr; 3539 uint32_t symbol_flags = 0; 3540 if (is_arm) { 3541 if (symbol_file_addr & 1) 3542 symbol_flags = 3543 MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3544 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3545 } 3546 3547 const FunctionStarts::Entry 3548 *next_func_start_entry = 3549 function_starts.FindNextEntry( 3550 func_start_entry); 3551 const addr_t section_end_file_addr = 3552 section_file_addr + 3553 symbol_section->GetByteSize(); 3554 if (next_func_start_entry) { 3555 addr_t next_symbol_file_addr = 3556 next_func_start_entry->addr; 3557 // Be sure the clear the Thumb address bit when 3558 // we calculate the size from the current and 3559 // next address 3560 if (is_arm) 3561 next_symbol_file_addr &= 3562 THUMB_ADDRESS_BIT_MASK; 3563 symbol_byte_size = std::min<lldb::addr_t>( 3564 next_symbol_file_addr - symbol_file_addr, 3565 section_end_file_addr - symbol_file_addr); 3566 } else { 3567 symbol_byte_size = 3568 section_end_file_addr - symbol_file_addr; 3569 } 3570 } 3571 } 3572 symbol_value -= section_file_addr; 3573 } 3574 3575 if (is_debug == false) { 3576 if (type == eSymbolTypeCode) { 3577 // See if we can find a N_FUN entry for any code 3578 // symbols. If we do find a match, and the name 3579 // matches, then we can merge the two into just the 3580 // function symbol to avoid duplicate entries in 3581 // the symbol table 3582 std::pair<ValueToSymbolIndexMap::const_iterator, 3583 ValueToSymbolIndexMap::const_iterator> 3584 range; 3585 range = N_FUN_addr_to_sym_idx.equal_range( 3586 nlist.n_value); 3587 if (range.first != range.second) { 3588 bool found_it = false; 3589 for (ValueToSymbolIndexMap::const_iterator pos = 3590 range.first; 3591 pos != range.second; ++pos) { 3592 if (sym[sym_idx].GetMangled().GetName( 3593 lldb::eLanguageTypeUnknown, 3594 Mangled::ePreferMangled) == 3595 sym[pos->second].GetMangled().GetName( 3596 lldb::eLanguageTypeUnknown, 3597 Mangled::ePreferMangled)) { 3598 m_nlist_idx_to_sym_idx[nlist_idx] = 3599 pos->second; 3600 // We just need the flags from the linker 3601 // symbol, so put these flags 3602 // into the N_FUN flags to avoid duplicate 3603 // symbols in the symbol table 3604 sym[pos->second].SetExternal( 3605 sym[sym_idx].IsExternal()); 3606 sym[pos->second].SetFlags(nlist.n_type << 16 | 3607 nlist.n_desc); 3608 if (resolver_addresses.find(nlist.n_value) != 3609 resolver_addresses.end()) 3610 sym[pos->second].SetType( 3611 eSymbolTypeResolver); 3612 sym[sym_idx].Clear(); 3613 found_it = true; 3614 break; 3615 } 3616 } 3617 if (found_it) 3618 continue; 3619 } else { 3620 if (resolver_addresses.find(nlist.n_value) != 3621 resolver_addresses.end()) 3622 type = eSymbolTypeResolver; 3623 } 3624 } else if (type == eSymbolTypeData || 3625 type == eSymbolTypeObjCClass || 3626 type == eSymbolTypeObjCMetaClass || 3627 type == eSymbolTypeObjCIVar) { 3628 // See if we can find a N_STSYM entry for any data 3629 // symbols. If we do find a match, and the name 3630 // matches, then we can merge the two into just the 3631 // Static symbol to avoid duplicate entries in the 3632 // symbol table 3633 std::pair<ValueToSymbolIndexMap::const_iterator, 3634 ValueToSymbolIndexMap::const_iterator> 3635 range; 3636 range = N_STSYM_addr_to_sym_idx.equal_range( 3637 nlist.n_value); 3638 if (range.first != range.second) { 3639 bool found_it = false; 3640 for (ValueToSymbolIndexMap::const_iterator pos = 3641 range.first; 3642 pos != range.second; ++pos) { 3643 if (sym[sym_idx].GetMangled().GetName( 3644 lldb::eLanguageTypeUnknown, 3645 Mangled::ePreferMangled) == 3646 sym[pos->second].GetMangled().GetName( 3647 lldb::eLanguageTypeUnknown, 3648 Mangled::ePreferMangled)) { 3649 m_nlist_idx_to_sym_idx[nlist_idx] = 3650 pos->second; 3651 // We just need the flags from the linker 3652 // symbol, so put these flags 3653 // into the N_STSYM flags to avoid duplicate 3654 // symbols in the symbol table 3655 sym[pos->second].SetExternal( 3656 sym[sym_idx].IsExternal()); 3657 sym[pos->second].SetFlags(nlist.n_type << 16 | 3658 nlist.n_desc); 3659 sym[sym_idx].Clear(); 3660 found_it = true; 3661 break; 3662 } 3663 } 3664 if (found_it) 3665 continue; 3666 } else { 3667 const char *gsym_name = 3668 sym[sym_idx] 3669 .GetMangled() 3670 .GetName(lldb::eLanguageTypeUnknown, 3671 Mangled::ePreferMangled) 3672 .GetCString(); 3673 if (gsym_name) { 3674 // Combine N_GSYM stab entries with the non 3675 // stab symbol 3676 ConstNameToSymbolIndexMap::const_iterator pos = 3677 N_GSYM_name_to_sym_idx.find(gsym_name); 3678 if (pos != N_GSYM_name_to_sym_idx.end()) { 3679 const uint32_t GSYM_sym_idx = pos->second; 3680 m_nlist_idx_to_sym_idx[nlist_idx] = 3681 GSYM_sym_idx; 3682 // Copy the address, because often the N_GSYM 3683 // address has an invalid address of zero 3684 // when the global is a common symbol 3685 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3686 symbol_section); 3687 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3688 symbol_value); 3689 // We just need the flags from the linker 3690 // symbol, so put these flags 3691 // into the N_GSYM flags to avoid duplicate 3692 // symbols in the symbol table 3693 sym[GSYM_sym_idx].SetFlags( 3694 nlist.n_type << 16 | nlist.n_desc); 3695 sym[sym_idx].Clear(); 3696 continue; 3697 } 3698 } 3699 } 3700 } 3701 } 3702 3703 sym[sym_idx].SetID(nlist_idx); 3704 sym[sym_idx].SetType(type); 3705 if (set_value) { 3706 sym[sym_idx].GetAddressRef().SetSection( 3707 symbol_section); 3708 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3709 } 3710 sym[sym_idx].SetFlags(nlist.n_type << 16 | 3711 nlist.n_desc); 3712 3713 if (symbol_byte_size > 0) 3714 sym[sym_idx].SetByteSize(symbol_byte_size); 3715 3716 if (demangled_is_synthesized) 3717 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3718 ++sym_idx; 3719 } else { 3720 sym[sym_idx].Clear(); 3721 } 3722 } 3723 ///////////////////////////// 3724 } 3725 break; // No more entries to consider 3726 } 3727 } 3728 3729 for (const auto &pos : reexport_shlib_needs_fixup) { 3730 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3731 if (undef_pos != undefined_name_to_desc.end()) { 3732 const uint8_t dylib_ordinal = 3733 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3734 if (dylib_ordinal > 0 && 3735 dylib_ordinal < dylib_files.GetSize()) 3736 sym[pos.first].SetReExportedSymbolSharedLibrary( 3737 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3738 } 3739 } 3740 } 3741 } 3742 } 3743 } 3744 } 3745 3746 // Must reset this in case it was mutated above! 3747 nlist_data_offset = 0; 3748 #endif 3749 3750 if (nlist_data.GetByteSize() > 0) { 3751 3752 // If the sym array was not created while parsing the DSC unmapped 3753 // symbols, create it now. 3754 if (sym == NULL) { 3755 sym = symtab->Resize(symtab_load_command.nsyms + 3756 m_dysymtab.nindirectsyms); 3757 num_syms = symtab->GetNumSymbols(); 3758 } 3759 3760 if (unmapped_local_symbols_found) { 3761 assert(m_dysymtab.ilocalsym == 0); 3762 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3763 nlist_idx = m_dysymtab.nlocalsym; 3764 } else { 3765 nlist_idx = 0; 3766 } 3767 3768 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3769 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3770 UndefinedNameToDescMap undefined_name_to_desc; 3771 SymbolIndexToName reexport_shlib_needs_fixup; 3772 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 3773 struct nlist_64 nlist; 3774 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, 3775 nlist_byte_size)) 3776 break; 3777 3778 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3779 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 3780 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 3781 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 3782 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 3783 3784 SymbolType type = eSymbolTypeInvalid; 3785 const char *symbol_name = NULL; 3786 3787 if (have_strtab_data) { 3788 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3789 3790 if (symbol_name == NULL) { 3791 // No symbol should be NULL, even the symbols with no string values 3792 // should have an offset zero which points to an empty C-string 3793 Host::SystemLog(Host::eSystemLogError, 3794 "error: symbol[%u] has invalid string table offset " 3795 "0x%x in %s, ignoring symbol\n", 3796 nlist_idx, nlist.n_strx, 3797 module_sp->GetFileSpec().GetPath().c_str()); 3798 continue; 3799 } 3800 if (symbol_name[0] == '\0') 3801 symbol_name = NULL; 3802 } else { 3803 const addr_t str_addr = strtab_addr + nlist.n_strx; 3804 Status str_error; 3805 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3806 str_error)) 3807 symbol_name = memory_symbol_name.c_str(); 3808 } 3809 const char *symbol_name_non_abi_mangled = NULL; 3810 3811 SectionSP symbol_section; 3812 lldb::addr_t symbol_byte_size = 0; 3813 bool add_nlist = true; 3814 bool is_gsym = false; 3815 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3816 bool demangled_is_synthesized = false; 3817 bool set_value = true; 3818 assert(sym_idx < num_syms); 3819 3820 sym[sym_idx].SetDebug(is_debug); 3821 3822 if (is_debug) { 3823 switch (nlist.n_type) { 3824 case N_GSYM: 3825 // global symbol: name,,NO_SECT,type,0 3826 // Sometimes the N_GSYM value contains the address. 3827 3828 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3829 // the ObjC data. They 3830 // have the same address, but we want to ensure that we always find 3831 // only the real symbol, 'cause we don't currently correctly 3832 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3833 // type. This is a temporary hack to make sure the ObjectiveC 3834 // symbols get treated correctly. To do this right, we should 3835 // coalesce all the GSYM & global symbols that have the same 3836 // address. 3837 is_gsym = true; 3838 sym[sym_idx].SetExternal(true); 3839 3840 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3841 llvm::StringRef symbol_name_ref(symbol_name); 3842 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3843 symbol_name_non_abi_mangled = symbol_name + 1; 3844 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3845 type = eSymbolTypeObjCClass; 3846 demangled_is_synthesized = true; 3847 3848 } else if (symbol_name_ref.startswith( 3849 g_objc_v2_prefix_metaclass)) { 3850 symbol_name_non_abi_mangled = symbol_name + 1; 3851 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3852 type = eSymbolTypeObjCMetaClass; 3853 demangled_is_synthesized = true; 3854 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3855 symbol_name_non_abi_mangled = symbol_name + 1; 3856 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3857 type = eSymbolTypeObjCIVar; 3858 demangled_is_synthesized = true; 3859 } 3860 } else { 3861 if (nlist.n_value != 0) 3862 symbol_section = 3863 section_info.GetSection(nlist.n_sect, nlist.n_value); 3864 type = eSymbolTypeData; 3865 } 3866 break; 3867 3868 case N_FNAME: 3869 // procedure name (f77 kludge): name,,NO_SECT,0,0 3870 type = eSymbolTypeCompiler; 3871 break; 3872 3873 case N_FUN: 3874 // procedure: name,,n_sect,linenumber,address 3875 if (symbol_name) { 3876 type = eSymbolTypeCode; 3877 symbol_section = 3878 section_info.GetSection(nlist.n_sect, nlist.n_value); 3879 3880 N_FUN_addr_to_sym_idx.insert( 3881 std::make_pair(nlist.n_value, sym_idx)); 3882 // We use the current number of symbols in the symbol table in 3883 // lieu of using nlist_idx in case we ever start trimming entries 3884 // out 3885 N_FUN_indexes.push_back(sym_idx); 3886 } else { 3887 type = eSymbolTypeCompiler; 3888 3889 if (!N_FUN_indexes.empty()) { 3890 // Copy the size of the function into the original STAB entry 3891 // so we don't have to hunt for it later 3892 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3893 ->SetByteSize(nlist.n_value); 3894 N_FUN_indexes.pop_back(); 3895 // We don't really need the end function STAB as it contains 3896 // the size which we already placed with the original symbol, 3897 // so don't add it if we want a minimal symbol table 3898 add_nlist = false; 3899 } 3900 } 3901 break; 3902 3903 case N_STSYM: 3904 // static symbol: name,,n_sect,type,address 3905 N_STSYM_addr_to_sym_idx.insert( 3906 std::make_pair(nlist.n_value, sym_idx)); 3907 symbol_section = 3908 section_info.GetSection(nlist.n_sect, nlist.n_value); 3909 if (symbol_name && symbol_name[0]) { 3910 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3911 eSymbolTypeData); 3912 } 3913 break; 3914 3915 case N_LCSYM: 3916 // .lcomm symbol: name,,n_sect,type,address 3917 symbol_section = 3918 section_info.GetSection(nlist.n_sect, nlist.n_value); 3919 type = eSymbolTypeCommonBlock; 3920 break; 3921 3922 case N_BNSYM: 3923 // We use the current number of symbols in the symbol table in lieu 3924 // of using nlist_idx in case we ever start trimming entries out 3925 // Skip these if we want minimal symbol tables 3926 add_nlist = false; 3927 break; 3928 3929 case N_ENSYM: 3930 // Set the size of the N_BNSYM to the terminating index of this 3931 // N_ENSYM so that we can always skip the entire symbol if we need 3932 // to navigate more quickly at the source level when parsing STABS 3933 // Skip these if we want minimal symbol tables 3934 add_nlist = false; 3935 break; 3936 3937 case N_OPT: 3938 // emitted with gcc2_compiled and in gcc source 3939 type = eSymbolTypeCompiler; 3940 break; 3941 3942 case N_RSYM: 3943 // register sym: name,,NO_SECT,type,register 3944 type = eSymbolTypeVariable; 3945 break; 3946 3947 case N_SLINE: 3948 // src line: 0,,n_sect,linenumber,address 3949 symbol_section = 3950 section_info.GetSection(nlist.n_sect, nlist.n_value); 3951 type = eSymbolTypeLineEntry; 3952 break; 3953 3954 case N_SSYM: 3955 // structure elt: name,,NO_SECT,type,struct_offset 3956 type = eSymbolTypeVariableType; 3957 break; 3958 3959 case N_SO: 3960 // source file name 3961 type = eSymbolTypeSourceFile; 3962 if (symbol_name == NULL) { 3963 add_nlist = false; 3964 if (N_SO_index != UINT32_MAX) { 3965 // Set the size of the N_SO to the terminating index of this 3966 // N_SO so that we can always skip the entire N_SO if we need 3967 // to navigate more quickly at the source level when parsing 3968 // STABS 3969 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3970 symbol_ptr->SetByteSize(sym_idx); 3971 symbol_ptr->SetSizeIsSibling(true); 3972 } 3973 N_NSYM_indexes.clear(); 3974 N_INCL_indexes.clear(); 3975 N_BRAC_indexes.clear(); 3976 N_COMM_indexes.clear(); 3977 N_FUN_indexes.clear(); 3978 N_SO_index = UINT32_MAX; 3979 } else { 3980 // We use the current number of symbols in the symbol table in 3981 // lieu of using nlist_idx in case we ever start trimming entries 3982 // out 3983 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3984 if (N_SO_has_full_path) { 3985 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3986 // We have two consecutive N_SO entries where the first 3987 // contains a directory and the second contains a full path. 3988 sym[sym_idx - 1].GetMangled().SetValue( 3989 ConstString(symbol_name), false); 3990 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3991 add_nlist = false; 3992 } else { 3993 // This is the first entry in a N_SO that contains a 3994 // directory or a full path to the source file 3995 N_SO_index = sym_idx; 3996 } 3997 } else if ((N_SO_index == sym_idx - 1) && 3998 ((sym_idx - 1) < num_syms)) { 3999 // This is usually the second N_SO entry that contains just the 4000 // filename, so here we combine it with the first one if we are 4001 // minimizing the symbol table 4002 const char *so_path = 4003 sym[sym_idx - 1] 4004 .GetMangled() 4005 .GetDemangledName(lldb::eLanguageTypeUnknown) 4006 .AsCString(); 4007 if (so_path && so_path[0]) { 4008 std::string full_so_path(so_path); 4009 const size_t double_slash_pos = full_so_path.find("//"); 4010 if (double_slash_pos != std::string::npos) { 4011 // The linker has been generating bad N_SO entries with 4012 // doubled up paths in the format "%s%s" where the first 4013 // string in the DW_AT_comp_dir, and the second is the 4014 // directory for the source file so you end up with a path 4015 // that looks like "/tmp/src//tmp/src/" 4016 FileSpec so_dir(so_path); 4017 if (!FileSystem::Instance().Exists(so_dir)) { 4018 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 4019 FileSpec::Style::native); 4020 if (FileSystem::Instance().Exists(so_dir)) { 4021 // Trim off the incorrect path 4022 full_so_path.erase(0, double_slash_pos + 1); 4023 } 4024 } 4025 } 4026 if (*full_so_path.rbegin() != '/') 4027 full_so_path += '/'; 4028 full_so_path += symbol_name; 4029 sym[sym_idx - 1].GetMangled().SetValue( 4030 ConstString(full_so_path.c_str()), false); 4031 add_nlist = false; 4032 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4033 } 4034 } else { 4035 // This could be a relative path to a N_SO 4036 N_SO_index = sym_idx; 4037 } 4038 } 4039 break; 4040 4041 case N_OSO: 4042 // object file name: name,,0,0,st_mtime 4043 type = eSymbolTypeObjectFile; 4044 break; 4045 4046 case N_LSYM: 4047 // local sym: name,,NO_SECT,type,offset 4048 type = eSymbolTypeLocal; 4049 break; 4050 4051 //---------------------------------------------------------------------- 4052 // INCL scopes 4053 //---------------------------------------------------------------------- 4054 case N_BINCL: 4055 // include file beginning: name,,NO_SECT,0,sum We use the current 4056 // number of symbols in the symbol table in lieu of using nlist_idx 4057 // in case we ever start trimming entries out 4058 N_INCL_indexes.push_back(sym_idx); 4059 type = eSymbolTypeScopeBegin; 4060 break; 4061 4062 case N_EINCL: 4063 // include file end: name,,NO_SECT,0,0 4064 // Set the size of the N_BINCL to the terminating index of this 4065 // N_EINCL so that we can always skip the entire symbol if we need 4066 // to navigate more quickly at the source level when parsing STABS 4067 if (!N_INCL_indexes.empty()) { 4068 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4069 symbol_ptr->SetByteSize(sym_idx + 1); 4070 symbol_ptr->SetSizeIsSibling(true); 4071 N_INCL_indexes.pop_back(); 4072 } 4073 type = eSymbolTypeScopeEnd; 4074 break; 4075 4076 case N_SOL: 4077 // #included file name: name,,n_sect,0,address 4078 type = eSymbolTypeHeaderFile; 4079 4080 // We currently don't use the header files on darwin 4081 add_nlist = false; 4082 break; 4083 4084 case N_PARAMS: 4085 // compiler parameters: name,,NO_SECT,0,0 4086 type = eSymbolTypeCompiler; 4087 break; 4088 4089 case N_VERSION: 4090 // compiler version: name,,NO_SECT,0,0 4091 type = eSymbolTypeCompiler; 4092 break; 4093 4094 case N_OLEVEL: 4095 // compiler -O level: name,,NO_SECT,0,0 4096 type = eSymbolTypeCompiler; 4097 break; 4098 4099 case N_PSYM: 4100 // parameter: name,,NO_SECT,type,offset 4101 type = eSymbolTypeVariable; 4102 break; 4103 4104 case N_ENTRY: 4105 // alternate entry: name,,n_sect,linenumber,address 4106 symbol_section = 4107 section_info.GetSection(nlist.n_sect, nlist.n_value); 4108 type = eSymbolTypeLineEntry; 4109 break; 4110 4111 //---------------------------------------------------------------------- 4112 // Left and Right Braces 4113 //---------------------------------------------------------------------- 4114 case N_LBRAC: 4115 // left bracket: 0,,NO_SECT,nesting level,address We use the 4116 // current number of symbols in the symbol table in lieu of using 4117 // nlist_idx in case we ever start trimming entries out 4118 symbol_section = 4119 section_info.GetSection(nlist.n_sect, nlist.n_value); 4120 N_BRAC_indexes.push_back(sym_idx); 4121 type = eSymbolTypeScopeBegin; 4122 break; 4123 4124 case N_RBRAC: 4125 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4126 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4127 // can always skip the entire symbol if we need to navigate more 4128 // quickly at the source level when parsing STABS 4129 symbol_section = 4130 section_info.GetSection(nlist.n_sect, nlist.n_value); 4131 if (!N_BRAC_indexes.empty()) { 4132 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4133 symbol_ptr->SetByteSize(sym_idx + 1); 4134 symbol_ptr->SetSizeIsSibling(true); 4135 N_BRAC_indexes.pop_back(); 4136 } 4137 type = eSymbolTypeScopeEnd; 4138 break; 4139 4140 case N_EXCL: 4141 // deleted include file: name,,NO_SECT,0,sum 4142 type = eSymbolTypeHeaderFile; 4143 break; 4144 4145 //---------------------------------------------------------------------- 4146 // COMM scopes 4147 //---------------------------------------------------------------------- 4148 case N_BCOMM: 4149 // begin common: name,,NO_SECT,0,0 4150 // We use the current number of symbols in the symbol table in lieu 4151 // of using nlist_idx in case we ever start trimming entries out 4152 type = eSymbolTypeScopeBegin; 4153 N_COMM_indexes.push_back(sym_idx); 4154 break; 4155 4156 case N_ECOML: 4157 // end common (local name): 0,,n_sect,0,address 4158 symbol_section = 4159 section_info.GetSection(nlist.n_sect, nlist.n_value); 4160 LLVM_FALLTHROUGH; 4161 4162 case N_ECOMM: 4163 // end common: name,,n_sect,0,0 4164 // Set the size of the N_BCOMM to the terminating index of this 4165 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4166 // we need to navigate more quickly at the source level when 4167 // parsing STABS 4168 if (!N_COMM_indexes.empty()) { 4169 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4170 symbol_ptr->SetByteSize(sym_idx + 1); 4171 symbol_ptr->SetSizeIsSibling(true); 4172 N_COMM_indexes.pop_back(); 4173 } 4174 type = eSymbolTypeScopeEnd; 4175 break; 4176 4177 case N_LENG: 4178 // second stab entry with length information 4179 type = eSymbolTypeAdditional; 4180 break; 4181 4182 default: 4183 break; 4184 } 4185 } else { 4186 // uint8_t n_pext = N_PEXT & nlist.n_type; 4187 uint8_t n_type = N_TYPE & nlist.n_type; 4188 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4189 4190 switch (n_type) { 4191 case N_INDR: { 4192 const char *reexport_name_cstr = 4193 strtab_data.PeekCStr(nlist.n_value); 4194 if (reexport_name_cstr && reexport_name_cstr[0]) { 4195 type = eSymbolTypeReExported; 4196 ConstString reexport_name( 4197 reexport_name_cstr + 4198 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4199 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4200 set_value = false; 4201 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4202 indirect_symbol_names.insert( 4203 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4204 } else 4205 type = eSymbolTypeUndefined; 4206 } break; 4207 4208 case N_UNDF: 4209 if (symbol_name && symbol_name[0]) { 4210 ConstString undefined_name(symbol_name + 4211 ((symbol_name[0] == '_') ? 1 : 0)); 4212 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4213 } 4214 LLVM_FALLTHROUGH; 4215 4216 case N_PBUD: 4217 type = eSymbolTypeUndefined; 4218 break; 4219 4220 case N_ABS: 4221 type = eSymbolTypeAbsolute; 4222 break; 4223 4224 case N_SECT: { 4225 symbol_section = 4226 section_info.GetSection(nlist.n_sect, nlist.n_value); 4227 4228 if (!symbol_section) { 4229 // TODO: warn about this? 4230 add_nlist = false; 4231 break; 4232 } 4233 4234 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4235 type = eSymbolTypeException; 4236 } else { 4237 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4238 4239 switch (section_type) { 4240 case S_CSTRING_LITERALS: 4241 type = eSymbolTypeData; 4242 break; // section with only literal C strings 4243 case S_4BYTE_LITERALS: 4244 type = eSymbolTypeData; 4245 break; // section with only 4 byte literals 4246 case S_8BYTE_LITERALS: 4247 type = eSymbolTypeData; 4248 break; // section with only 8 byte literals 4249 case S_LITERAL_POINTERS: 4250 type = eSymbolTypeTrampoline; 4251 break; // section with only pointers to literals 4252 case S_NON_LAZY_SYMBOL_POINTERS: 4253 type = eSymbolTypeTrampoline; 4254 break; // section with only non-lazy symbol pointers 4255 case S_LAZY_SYMBOL_POINTERS: 4256 type = eSymbolTypeTrampoline; 4257 break; // section with only lazy symbol pointers 4258 case S_SYMBOL_STUBS: 4259 type = eSymbolTypeTrampoline; 4260 break; // section with only symbol stubs, byte size of stub in 4261 // the reserved2 field 4262 case S_MOD_INIT_FUNC_POINTERS: 4263 type = eSymbolTypeCode; 4264 break; // section with only function pointers for initialization 4265 case S_MOD_TERM_FUNC_POINTERS: 4266 type = eSymbolTypeCode; 4267 break; // section with only function pointers for termination 4268 case S_INTERPOSING: 4269 type = eSymbolTypeTrampoline; 4270 break; // section with only pairs of function pointers for 4271 // interposing 4272 case S_16BYTE_LITERALS: 4273 type = eSymbolTypeData; 4274 break; // section with only 16 byte literals 4275 case S_DTRACE_DOF: 4276 type = eSymbolTypeInstrumentation; 4277 break; 4278 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4279 type = eSymbolTypeTrampoline; 4280 break; 4281 default: 4282 switch (symbol_section->GetType()) { 4283 case lldb::eSectionTypeCode: 4284 type = eSymbolTypeCode; 4285 break; 4286 case eSectionTypeData: 4287 case eSectionTypeDataCString: // Inlined C string data 4288 case eSectionTypeDataCStringPointers: // Pointers to C string 4289 // data 4290 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4291 // the symbol table 4292 case eSectionTypeData4: 4293 case eSectionTypeData8: 4294 case eSectionTypeData16: 4295 type = eSymbolTypeData; 4296 break; 4297 default: 4298 break; 4299 } 4300 break; 4301 } 4302 4303 if (type == eSymbolTypeInvalid) { 4304 const char *symbol_sect_name = 4305 symbol_section->GetName().AsCString(); 4306 if (symbol_section->IsDescendant(text_section_sp.get())) { 4307 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4308 S_ATTR_SELF_MODIFYING_CODE | 4309 S_ATTR_SOME_INSTRUCTIONS)) 4310 type = eSymbolTypeData; 4311 else 4312 type = eSymbolTypeCode; 4313 } else if (symbol_section->IsDescendant( 4314 data_section_sp.get()) || 4315 symbol_section->IsDescendant( 4316 data_dirty_section_sp.get()) || 4317 symbol_section->IsDescendant( 4318 data_const_section_sp.get())) { 4319 if (symbol_sect_name && 4320 ::strstr(symbol_sect_name, "__objc") == 4321 symbol_sect_name) { 4322 type = eSymbolTypeRuntime; 4323 4324 if (symbol_name) { 4325 llvm::StringRef symbol_name_ref(symbol_name); 4326 if (symbol_name_ref.startswith("_OBJC_")) { 4327 static const llvm::StringRef g_objc_v2_prefix_class( 4328 "_OBJC_CLASS_$_"); 4329 static const llvm::StringRef g_objc_v2_prefix_metaclass( 4330 "_OBJC_METACLASS_$_"); 4331 static const llvm::StringRef g_objc_v2_prefix_ivar( 4332 "_OBJC_IVAR_$_"); 4333 if (symbol_name_ref.startswith( 4334 g_objc_v2_prefix_class)) { 4335 symbol_name_non_abi_mangled = symbol_name + 1; 4336 symbol_name = 4337 symbol_name + g_objc_v2_prefix_class.size(); 4338 type = eSymbolTypeObjCClass; 4339 demangled_is_synthesized = true; 4340 } else if (symbol_name_ref.startswith( 4341 g_objc_v2_prefix_metaclass)) { 4342 symbol_name_non_abi_mangled = symbol_name + 1; 4343 symbol_name = 4344 symbol_name + g_objc_v2_prefix_metaclass.size(); 4345 type = eSymbolTypeObjCMetaClass; 4346 demangled_is_synthesized = true; 4347 } else if (symbol_name_ref.startswith( 4348 g_objc_v2_prefix_ivar)) { 4349 symbol_name_non_abi_mangled = symbol_name + 1; 4350 symbol_name = 4351 symbol_name + g_objc_v2_prefix_ivar.size(); 4352 type = eSymbolTypeObjCIVar; 4353 demangled_is_synthesized = true; 4354 } 4355 } 4356 } 4357 } else if (symbol_sect_name && 4358 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4359 symbol_sect_name) { 4360 type = eSymbolTypeException; 4361 } else { 4362 type = eSymbolTypeData; 4363 } 4364 } else if (symbol_sect_name && 4365 ::strstr(symbol_sect_name, "__IMPORT") == 4366 symbol_sect_name) { 4367 type = eSymbolTypeTrampoline; 4368 } else if (symbol_section->IsDescendant( 4369 objc_section_sp.get())) { 4370 type = eSymbolTypeRuntime; 4371 if (symbol_name && symbol_name[0] == '.') { 4372 llvm::StringRef symbol_name_ref(symbol_name); 4373 static const llvm::StringRef g_objc_v1_prefix_class( 4374 ".objc_class_name_"); 4375 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4376 symbol_name_non_abi_mangled = symbol_name; 4377 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4378 type = eSymbolTypeObjCClass; 4379 demangled_is_synthesized = true; 4380 } 4381 } 4382 } 4383 } 4384 } 4385 } break; 4386 } 4387 } 4388 4389 if (add_nlist) { 4390 uint64_t symbol_value = nlist.n_value; 4391 4392 if (symbol_name_non_abi_mangled) { 4393 sym[sym_idx].GetMangled().SetMangledName( 4394 ConstString(symbol_name_non_abi_mangled)); 4395 sym[sym_idx].GetMangled().SetDemangledName( 4396 ConstString(symbol_name)); 4397 } else { 4398 bool symbol_name_is_mangled = false; 4399 4400 if (symbol_name && symbol_name[0] == '_') { 4401 symbol_name_is_mangled = symbol_name[1] == '_'; 4402 symbol_name++; // Skip the leading underscore 4403 } 4404 4405 if (symbol_name) { 4406 ConstString const_symbol_name(symbol_name); 4407 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4408 symbol_name_is_mangled); 4409 } 4410 } 4411 4412 if (is_gsym) { 4413 const char *gsym_name = sym[sym_idx] 4414 .GetMangled() 4415 .GetName(lldb::eLanguageTypeUnknown, 4416 Mangled::ePreferMangled) 4417 .GetCString(); 4418 if (gsym_name) 4419 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4420 } 4421 4422 if (symbol_section) { 4423 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4424 if (symbol_byte_size == 0 && function_starts_count > 0) { 4425 addr_t symbol_lookup_file_addr = nlist.n_value; 4426 // Do an exact address match for non-ARM addresses, else get the 4427 // closest since the symbol might be a thumb symbol which has an 4428 // address with bit zero set 4429 FunctionStarts::Entry *func_start_entry = 4430 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4431 if (is_arm && func_start_entry) { 4432 // Verify that the function start address is the symbol address 4433 // (ARM) or the symbol address + 1 (thumb) 4434 if (func_start_entry->addr != symbol_lookup_file_addr && 4435 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4436 // Not the right entry, NULL it out... 4437 func_start_entry = NULL; 4438 } 4439 } 4440 if (func_start_entry) { 4441 func_start_entry->data = true; 4442 4443 addr_t symbol_file_addr = func_start_entry->addr; 4444 if (is_arm) 4445 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4446 4447 const FunctionStarts::Entry *next_func_start_entry = 4448 function_starts.FindNextEntry(func_start_entry); 4449 const addr_t section_end_file_addr = 4450 section_file_addr + symbol_section->GetByteSize(); 4451 if (next_func_start_entry) { 4452 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4453 // Be sure the clear the Thumb address bit when we calculate 4454 // the size from the current and next address 4455 if (is_arm) 4456 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4457 symbol_byte_size = std::min<lldb::addr_t>( 4458 next_symbol_file_addr - symbol_file_addr, 4459 section_end_file_addr - symbol_file_addr); 4460 } else { 4461 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4462 } 4463 } 4464 } 4465 symbol_value -= section_file_addr; 4466 } 4467 4468 if (is_debug == false) { 4469 if (type == eSymbolTypeCode) { 4470 // See if we can find a N_FUN entry for any code symbols. If we 4471 // do find a match, and the name matches, then we can merge the 4472 // two into just the function symbol to avoid duplicate entries 4473 // in the symbol table 4474 std::pair<ValueToSymbolIndexMap::const_iterator, 4475 ValueToSymbolIndexMap::const_iterator> 4476 range; 4477 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4478 if (range.first != range.second) { 4479 bool found_it = false; 4480 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4481 pos != range.second; ++pos) { 4482 if (sym[sym_idx].GetMangled().GetName( 4483 lldb::eLanguageTypeUnknown, 4484 Mangled::ePreferMangled) == 4485 sym[pos->second].GetMangled().GetName( 4486 lldb::eLanguageTypeUnknown, 4487 Mangled::ePreferMangled)) { 4488 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4489 // We just need the flags from the linker symbol, so put 4490 // these flags into the N_FUN flags to avoid duplicate 4491 // symbols in the symbol table 4492 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4493 sym[pos->second].SetFlags(nlist.n_type << 16 | 4494 nlist.n_desc); 4495 if (resolver_addresses.find(nlist.n_value) != 4496 resolver_addresses.end()) 4497 sym[pos->second].SetType(eSymbolTypeResolver); 4498 sym[sym_idx].Clear(); 4499 found_it = true; 4500 break; 4501 } 4502 } 4503 if (found_it) 4504 continue; 4505 } else { 4506 if (resolver_addresses.find(nlist.n_value) != 4507 resolver_addresses.end()) 4508 type = eSymbolTypeResolver; 4509 } 4510 } else if (type == eSymbolTypeData || 4511 type == eSymbolTypeObjCClass || 4512 type == eSymbolTypeObjCMetaClass || 4513 type == eSymbolTypeObjCIVar) { 4514 // See if we can find a N_STSYM entry for any data symbols. If we 4515 // do find a match, and the name matches, then we can merge the 4516 // two into just the Static symbol to avoid duplicate entries in 4517 // the symbol table 4518 std::pair<ValueToSymbolIndexMap::const_iterator, 4519 ValueToSymbolIndexMap::const_iterator> 4520 range; 4521 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4522 if (range.first != range.second) { 4523 bool found_it = false; 4524 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4525 pos != range.second; ++pos) { 4526 if (sym[sym_idx].GetMangled().GetName( 4527 lldb::eLanguageTypeUnknown, 4528 Mangled::ePreferMangled) == 4529 sym[pos->second].GetMangled().GetName( 4530 lldb::eLanguageTypeUnknown, 4531 Mangled::ePreferMangled)) { 4532 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4533 // We just need the flags from the linker symbol, so put 4534 // these flags into the N_STSYM flags to avoid duplicate 4535 // symbols in the symbol table 4536 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4537 sym[pos->second].SetFlags(nlist.n_type << 16 | 4538 nlist.n_desc); 4539 sym[sym_idx].Clear(); 4540 found_it = true; 4541 break; 4542 } 4543 } 4544 if (found_it) 4545 continue; 4546 } else { 4547 // Combine N_GSYM stab entries with the non stab symbol 4548 const char *gsym_name = sym[sym_idx] 4549 .GetMangled() 4550 .GetName(lldb::eLanguageTypeUnknown, 4551 Mangled::ePreferMangled) 4552 .GetCString(); 4553 if (gsym_name) { 4554 ConstNameToSymbolIndexMap::const_iterator pos = 4555 N_GSYM_name_to_sym_idx.find(gsym_name); 4556 if (pos != N_GSYM_name_to_sym_idx.end()) { 4557 const uint32_t GSYM_sym_idx = pos->second; 4558 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4559 // Copy the address, because often the N_GSYM address has 4560 // an invalid address of zero when the global is a common 4561 // symbol 4562 sym[GSYM_sym_idx].GetAddressRef().SetSection( 4563 symbol_section); 4564 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4565 // We just need the flags from the linker symbol, so put 4566 // these flags into the N_GSYM flags to avoid duplicate 4567 // symbols in the symbol table 4568 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 4569 nlist.n_desc); 4570 sym[sym_idx].Clear(); 4571 continue; 4572 } 4573 } 4574 } 4575 } 4576 } 4577 4578 sym[sym_idx].SetID(nlist_idx); 4579 sym[sym_idx].SetType(type); 4580 if (set_value) { 4581 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4582 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4583 } 4584 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4585 4586 if (symbol_byte_size > 0) 4587 sym[sym_idx].SetByteSize(symbol_byte_size); 4588 4589 if (demangled_is_synthesized) 4590 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4591 4592 ++sym_idx; 4593 } else { 4594 sym[sym_idx].Clear(); 4595 } 4596 } 4597 4598 for (const auto &pos : reexport_shlib_needs_fixup) { 4599 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4600 if (undef_pos != undefined_name_to_desc.end()) { 4601 const uint8_t dylib_ordinal = 4602 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4603 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4604 sym[pos.first].SetReExportedSymbolSharedLibrary( 4605 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4606 } 4607 } 4608 } 4609 4610 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4611 4612 if (function_starts_count > 0) { 4613 uint32_t num_synthetic_function_symbols = 0; 4614 for (i = 0; i < function_starts_count; ++i) { 4615 if (function_starts.GetEntryRef(i).data == false) 4616 ++num_synthetic_function_symbols; 4617 } 4618 4619 if (num_synthetic_function_symbols > 0) { 4620 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4621 num_syms = sym_idx + num_synthetic_function_symbols; 4622 sym = symtab->Resize(num_syms); 4623 } 4624 for (i = 0; i < function_starts_count; ++i) { 4625 const FunctionStarts::Entry *func_start_entry = 4626 function_starts.GetEntryAtIndex(i); 4627 if (func_start_entry->data == false) { 4628 addr_t symbol_file_addr = func_start_entry->addr; 4629 uint32_t symbol_flags = 0; 4630 if (is_arm) { 4631 if (symbol_file_addr & 1) 4632 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4633 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4634 } 4635 Address symbol_addr; 4636 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4637 SectionSP symbol_section(symbol_addr.GetSection()); 4638 uint32_t symbol_byte_size = 0; 4639 if (symbol_section) { 4640 const addr_t section_file_addr = 4641 symbol_section->GetFileAddress(); 4642 const FunctionStarts::Entry *next_func_start_entry = 4643 function_starts.FindNextEntry(func_start_entry); 4644 const addr_t section_end_file_addr = 4645 section_file_addr + symbol_section->GetByteSize(); 4646 if (next_func_start_entry) { 4647 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4648 if (is_arm) 4649 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4650 symbol_byte_size = std::min<lldb::addr_t>( 4651 next_symbol_file_addr - symbol_file_addr, 4652 section_end_file_addr - symbol_file_addr); 4653 } else { 4654 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4655 } 4656 sym[sym_idx].SetID(synthetic_sym_id++); 4657 sym[sym_idx].GetMangled().SetDemangledName( 4658 GetNextSyntheticSymbolName()); 4659 sym[sym_idx].SetType(eSymbolTypeCode); 4660 sym[sym_idx].SetIsSynthetic(true); 4661 sym[sym_idx].GetAddressRef() = symbol_addr; 4662 if (symbol_flags) 4663 sym[sym_idx].SetFlags(symbol_flags); 4664 if (symbol_byte_size) 4665 sym[sym_idx].SetByteSize(symbol_byte_size); 4666 ++sym_idx; 4667 } 4668 } 4669 } 4670 } 4671 } 4672 } 4673 4674 // Trim our symbols down to just what we ended up with after removing any 4675 // symbols. 4676 if (sym_idx < num_syms) { 4677 num_syms = sym_idx; 4678 sym = symtab->Resize(num_syms); 4679 } 4680 4681 // Now synthesize indirect symbols 4682 if (m_dysymtab.nindirectsyms != 0) { 4683 if (indirect_symbol_index_data.GetByteSize()) { 4684 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4685 m_nlist_idx_to_sym_idx.end(); 4686 4687 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4688 ++sect_idx) { 4689 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4690 S_SYMBOL_STUBS) { 4691 uint32_t symbol_stub_byte_size = 4692 m_mach_sections[sect_idx].reserved2; 4693 if (symbol_stub_byte_size == 0) 4694 continue; 4695 4696 const uint32_t num_symbol_stubs = 4697 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4698 4699 if (num_symbol_stubs == 0) 4700 continue; 4701 4702 const uint32_t symbol_stub_index_offset = 4703 m_mach_sections[sect_idx].reserved1; 4704 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; 4705 ++stub_idx) { 4706 const uint32_t symbol_stub_index = 4707 symbol_stub_index_offset + stub_idx; 4708 const lldb::addr_t symbol_stub_addr = 4709 m_mach_sections[sect_idx].addr + 4710 (stub_idx * symbol_stub_byte_size); 4711 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4712 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4713 symbol_stub_offset, 4)) { 4714 const uint32_t stub_sym_id = 4715 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4716 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4717 continue; 4718 4719 NListIndexToSymbolIndexMap::const_iterator index_pos = 4720 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4721 Symbol *stub_symbol = NULL; 4722 if (index_pos != end_index_pos) { 4723 // We have a remapping from the original nlist index to a 4724 // current symbol index, so just look this up by index 4725 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4726 } else { 4727 // We need to lookup a symbol using the original nlist symbol 4728 // index since this index is coming from the S_SYMBOL_STUBS 4729 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4730 } 4731 4732 if (stub_symbol) { 4733 Address so_addr(symbol_stub_addr, section_list); 4734 4735 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4736 // Change the external symbol into a trampoline that makes 4737 // sense These symbols were N_UNDF N_EXT, and are useless 4738 // to us, so we can re-use them so we don't have to make up 4739 // a synthetic symbol for no good reason. 4740 if (resolver_addresses.find(symbol_stub_addr) == 4741 resolver_addresses.end()) 4742 stub_symbol->SetType(eSymbolTypeTrampoline); 4743 else 4744 stub_symbol->SetType(eSymbolTypeResolver); 4745 stub_symbol->SetExternal(false); 4746 stub_symbol->GetAddressRef() = so_addr; 4747 stub_symbol->SetByteSize(symbol_stub_byte_size); 4748 } else { 4749 // Make a synthetic symbol to describe the trampoline stub 4750 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4751 if (sym_idx >= num_syms) { 4752 sym = symtab->Resize(++num_syms); 4753 stub_symbol = NULL; // this pointer no longer valid 4754 } 4755 sym[sym_idx].SetID(synthetic_sym_id++); 4756 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4757 if (resolver_addresses.find(symbol_stub_addr) == 4758 resolver_addresses.end()) 4759 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4760 else 4761 sym[sym_idx].SetType(eSymbolTypeResolver); 4762 sym[sym_idx].SetIsSynthetic(true); 4763 sym[sym_idx].GetAddressRef() = so_addr; 4764 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4765 ++sym_idx; 4766 } 4767 } else { 4768 if (log) 4769 log->Warning("symbol stub referencing symbol table symbol " 4770 "%u that isn't in our minimal symbol table, " 4771 "fix this!!!", 4772 stub_sym_id); 4773 } 4774 } 4775 } 4776 } 4777 } 4778 } 4779 } 4780 4781 if (!trie_entries.empty()) { 4782 for (const auto &e : trie_entries) { 4783 if (e.entry.import_name) { 4784 // Only add indirect symbols from the Trie entries if we didn't have 4785 // a N_INDR nlist entry for this already 4786 if (indirect_symbol_names.find(e.entry.name) == 4787 indirect_symbol_names.end()) { 4788 // Make a synthetic symbol to describe re-exported symbol. 4789 if (sym_idx >= num_syms) 4790 sym = symtab->Resize(++num_syms); 4791 sym[sym_idx].SetID(synthetic_sym_id++); 4792 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4793 sym[sym_idx].SetType(eSymbolTypeReExported); 4794 sym[sym_idx].SetIsSynthetic(true); 4795 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4796 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4797 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4798 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4799 } 4800 ++sym_idx; 4801 } 4802 } 4803 } 4804 } 4805 4806 // StreamFile s(stdout, false); 4807 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4808 // symtab->Dump(&s, NULL, eSortOrderNone); 4809 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4810 // sizes 4811 symtab->CalculateSymbolSizes(); 4812 4813 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4814 // symtab->Dump(&s, NULL, eSortOrderNone); 4815 4816 return symtab->GetNumSymbols(); 4817 } 4818 return 0; 4819 } 4820 4821 void ObjectFileMachO::Dump(Stream *s) { 4822 ModuleSP module_sp(GetModule()); 4823 if (module_sp) { 4824 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4825 s->Printf("%p: ", static_cast<void *>(this)); 4826 s->Indent(); 4827 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4828 s->PutCString("ObjectFileMachO64"); 4829 else 4830 s->PutCString("ObjectFileMachO32"); 4831 4832 ArchSpec header_arch; 4833 GetArchitecture(header_arch); 4834 4835 *s << ", file = '" << m_file 4836 << "', triple = " << header_arch.GetTriple().getTriple() << "\n"; 4837 4838 SectionList *sections = GetSectionList(); 4839 if (sections) 4840 sections->Dump(s, NULL, true, UINT32_MAX); 4841 4842 if (m_symtab_ap.get()) 4843 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4844 } 4845 } 4846 4847 bool ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4848 const lldb_private::DataExtractor &data, 4849 lldb::offset_t lc_offset, 4850 lldb_private::UUID &uuid) { 4851 uint32_t i; 4852 struct uuid_command load_cmd; 4853 4854 lldb::offset_t offset = lc_offset; 4855 for (i = 0; i < header.ncmds; ++i) { 4856 const lldb::offset_t cmd_offset = offset; 4857 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4858 break; 4859 4860 if (load_cmd.cmd == LC_UUID) { 4861 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4862 4863 if (uuid_bytes) { 4864 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4865 // We pretend these object files have no UUID to prevent crashing. 4866 4867 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4868 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4869 0xbb, 0x14, 0xf0, 0x0d}; 4870 4871 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4872 return false; 4873 4874 uuid = UUID::fromOptionalData(uuid_bytes, 16); 4875 return true; 4876 } 4877 return false; 4878 } 4879 offset = cmd_offset + load_cmd.cmdsize; 4880 } 4881 return false; 4882 } 4883 4884 static llvm::StringRef GetOSName(uint32_t cmd) { 4885 switch (cmd) { 4886 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4887 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4888 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4889 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4890 case llvm::MachO::LC_VERSION_MIN_TVOS: 4891 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4892 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4893 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4894 default: 4895 llvm_unreachable("unexpected LC_VERSION load command"); 4896 } 4897 } 4898 4899 namespace { 4900 struct OSEnv { 4901 llvm::StringRef os_type; 4902 llvm::StringRef environment; 4903 OSEnv(uint32_t cmd) { 4904 switch (cmd) { 4905 case PLATFORM_MACOS: 4906 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4907 return; 4908 case PLATFORM_IOS: 4909 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4910 return; 4911 case PLATFORM_TVOS: 4912 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4913 return; 4914 case PLATFORM_WATCHOS: 4915 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4916 return; 4917 // NEED_BRIDGEOS_TRIPLE case PLATFORM_BRIDGEOS: 4918 // NEED_BRIDGEOS_TRIPLE os_type = llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4919 // NEED_BRIDGEOS_TRIPLE return; 4920 #if defined (PLATFORM_IOSSIMULATOR) && defined (PLATFORM_TVOSSIMULATOR) && defined (PLATFORM_WATCHOSSIMULATOR) 4921 case PLATFORM_IOSSIMULATOR: 4922 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4923 environment = 4924 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4925 return; 4926 case PLATFORM_TVOSSIMULATOR: 4927 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4928 environment = 4929 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4930 return; 4931 case PLATFORM_WATCHOSSIMULATOR: 4932 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4933 environment = 4934 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4935 return; 4936 #endif 4937 default: { 4938 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 4939 LIBLLDB_LOG_PROCESS)); 4940 if (log) 4941 log->Printf("unsupported platform in LC_BUILD_VERSION"); 4942 } 4943 } 4944 } 4945 }; 4946 4947 struct MinOS { 4948 uint32_t major_version, minor_version, patch_version; 4949 MinOS(uint32_t version) 4950 : major_version(version >> 16), 4951 minor_version((version >> 8) & 0xffu), 4952 patch_version(version & 0xffu) {} 4953 }; 4954 } // namespace 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, force 4972 // the Vendor to Apple so we don't accidentally pick up the generic 4973 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4974 // frame pointer register; most other armv7 ABIs use a combination of 4975 // 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 This is required for correct behavior of EFI debugging on 4980 // 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 llvm::SmallString<16> os_name; 4988 llvm::raw_svector_ostream os(os_name); 4989 4990 // See if there is an LC_VERSION_MIN_* load command that can give 4991 // us the OS type. 4992 lldb::offset_t offset = lc_offset; 4993 for (uint32_t i = 0; i < header.ncmds; ++i) { 4994 const lldb::offset_t cmd_offset = offset; 4995 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4996 break; 4997 4998 struct version_min_command version_min; 4999 switch (load_cmd.cmd) { 5000 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 5001 case llvm::MachO::LC_VERSION_MIN_MACOSX: 5002 case llvm::MachO::LC_VERSION_MIN_TVOS: 5003 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 5004 if (load_cmd.cmdsize != sizeof(version_min)) 5005 break; 5006 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 5007 data.GetByteOrder(), &version_min) == 0) 5008 break; 5009 MinOS min_os(version_min.version); 5010 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 5011 << min_os.minor_version << '.' << min_os.patch_version; 5012 triple.setOSName(os.str()); 5013 return true; 5014 } 5015 default: 5016 break; 5017 } 5018 5019 offset = cmd_offset + load_cmd.cmdsize; 5020 } 5021 5022 // See if there is an LC_BUILD_VERSION load command that can give 5023 // us the OS type. 5024 5025 offset = lc_offset; 5026 for (uint32_t i = 0; i < header.ncmds; ++i) { 5027 const lldb::offset_t cmd_offset = offset; 5028 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 5029 break; 5030 5031 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 5032 struct build_version_command build_version; 5033 if (load_cmd.cmdsize != sizeof(build_version)) 5034 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 5035 data.GetByteOrder(), &build_version) == 0) 5036 continue; 5037 MinOS min_os(build_version.minos); 5038 OSEnv os_env(build_version.platform); 5039 if (os_env.os_type.empty()) 5040 continue; 5041 os << os_env.os_type << min_os.major_version << '.' 5042 << min_os.minor_version << '.' << min_os.patch_version; 5043 triple.setOSName(os.str()); 5044 if (!os_env.environment.empty()) 5045 triple.setEnvironmentName(os_env.environment); 5046 return true; 5047 } 5048 offset = cmd_offset + load_cmd.cmdsize; 5049 } 5050 5051 if (header.filetype != MH_KEXT_BUNDLE) { 5052 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 5053 // so lets not say our Vendor is Apple, leave it as an unspecified 5054 // unknown 5055 triple.setVendor(llvm::Triple::UnknownVendor); 5056 triple.setVendorName(llvm::StringRef()); 5057 } 5058 } 5059 } 5060 return arch.IsValid(); 5061 } 5062 5063 bool ObjectFileMachO::GetUUID(lldb_private::UUID *uuid) { 5064 ModuleSP module_sp(GetModule()); 5065 if (module_sp) { 5066 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5067 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5068 return GetUUID(m_header, m_data, offset, *uuid); 5069 } 5070 return false; 5071 } 5072 5073 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5074 uint32_t count = 0; 5075 ModuleSP module_sp(GetModule()); 5076 if (module_sp) { 5077 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5078 struct load_command load_cmd; 5079 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5080 std::vector<std::string> rpath_paths; 5081 std::vector<std::string> rpath_relative_paths; 5082 std::vector<std::string> at_exec_relative_paths; 5083 uint32_t i; 5084 for (i = 0; i < m_header.ncmds; ++i) { 5085 const uint32_t cmd_offset = offset; 5086 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5087 break; 5088 5089 switch (load_cmd.cmd) { 5090 case LC_RPATH: 5091 case LC_LOAD_DYLIB: 5092 case LC_LOAD_WEAK_DYLIB: 5093 case LC_REEXPORT_DYLIB: 5094 case LC_LOAD_DYLINKER: 5095 case LC_LOADFVMLIB: 5096 case LC_LOAD_UPWARD_DYLIB: { 5097 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5098 const char *path = m_data.PeekCStr(name_offset); 5099 if (path) { 5100 if (load_cmd.cmd == LC_RPATH) 5101 rpath_paths.push_back(path); 5102 else { 5103 if (path[0] == '@') { 5104 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5105 rpath_relative_paths.push_back(path + strlen("@rpath")); 5106 else if (strncmp(path, "@executable_path", 5107 strlen("@executable_path")) == 0) 5108 at_exec_relative_paths.push_back(path 5109 + strlen("@executable_path")); 5110 } else { 5111 FileSpec file_spec(path); 5112 if (files.AppendIfUnique(file_spec)) 5113 count++; 5114 } 5115 } 5116 } 5117 } break; 5118 5119 default: 5120 break; 5121 } 5122 offset = cmd_offset + load_cmd.cmdsize; 5123 } 5124 5125 FileSpec this_file_spec(m_file); 5126 FileSystem::Instance().Resolve(this_file_spec); 5127 5128 if (!rpath_paths.empty()) { 5129 // Fixup all LC_RPATH values to be absolute paths 5130 std::string loader_path("@loader_path"); 5131 std::string executable_path("@executable_path"); 5132 for (auto &rpath : rpath_paths) { 5133 if (rpath.find(loader_path) == 0) { 5134 rpath.erase(0, loader_path.size()); 5135 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5136 } else if (rpath.find(executable_path) == 0) { 5137 rpath.erase(0, executable_path.size()); 5138 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5139 } 5140 } 5141 5142 for (const auto &rpath_relative_path : rpath_relative_paths) { 5143 for (const auto &rpath : rpath_paths) { 5144 std::string path = rpath; 5145 path += rpath_relative_path; 5146 // It is OK to resolve this path because we must find a file on disk 5147 // for us to accept it anyway if it is rpath relative. 5148 FileSpec file_spec(path); 5149 FileSystem::Instance().Resolve(file_spec); 5150 if (FileSystem::Instance().Exists(file_spec) && 5151 files.AppendIfUnique(file_spec)) { 5152 count++; 5153 break; 5154 } 5155 } 5156 } 5157 } 5158 5159 // We may have @executable_paths but no RPATHS. Figure those out here. 5160 // Only do this if this object file is the executable. We have no way to 5161 // get back to the actual executable otherwise, so we won't get the right 5162 // path. 5163 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5164 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5165 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5166 FileSpec file_spec = 5167 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5168 if (FileSystem::Instance().Exists(file_spec) && 5169 files.AppendIfUnique(file_spec)) 5170 count++; 5171 } 5172 } 5173 } 5174 return count; 5175 } 5176 5177 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5178 // If the object file is not an executable it can't hold the entry point. 5179 // m_entry_point_address is initialized to an invalid address, so we can just 5180 // return that. If m_entry_point_address is valid it means we've found it 5181 // already, so return the cached value. 5182 5183 if (!IsExecutable() || m_entry_point_address.IsValid()) 5184 return m_entry_point_address; 5185 5186 // Otherwise, look for the UnixThread or Thread command. The data for the 5187 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5188 // 5189 // uint32_t flavor - this is the flavor argument you would pass to 5190 // thread_get_state 5191 // uint32_t count - this is the count of longs in the thread state data 5192 // struct XXX_thread_state state - this is the structure from 5193 // <machine/thread_status.h> corresponding to the flavor. 5194 // <repeat this trio> 5195 // 5196 // So we just keep reading the various register flavors till we find the GPR 5197 // one, then read the PC out of there. 5198 // FIXME: We will need to have a "RegisterContext data provider" class at some 5199 // point that can get all the registers 5200 // out of data in this form & attach them to a given thread. That should 5201 // underlie the MacOS X User process plugin, and we'll also need it for the 5202 // MacOS X Core File process plugin. When we have that we can also use it 5203 // here. 5204 // 5205 // For now we hard-code the offsets and flavors we need: 5206 // 5207 // 5208 5209 ModuleSP module_sp(GetModule()); 5210 if (module_sp) { 5211 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5212 struct load_command load_cmd; 5213 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5214 uint32_t i; 5215 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5216 bool done = false; 5217 5218 for (i = 0; i < m_header.ncmds; ++i) { 5219 const lldb::offset_t cmd_offset = offset; 5220 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5221 break; 5222 5223 switch (load_cmd.cmd) { 5224 case LC_UNIXTHREAD: 5225 case LC_THREAD: { 5226 while (offset < cmd_offset + load_cmd.cmdsize) { 5227 uint32_t flavor = m_data.GetU32(&offset); 5228 uint32_t count = m_data.GetU32(&offset); 5229 if (count == 0) { 5230 // We've gotten off somehow, log and exit; 5231 return m_entry_point_address; 5232 } 5233 5234 switch (m_header.cputype) { 5235 case llvm::MachO::CPU_TYPE_ARM: 5236 if (flavor == 1 || 5237 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 from 5238 // mach/arm/thread_status.h 5239 { 5240 offset += 60; // This is the offset of pc in the GPR thread state 5241 // data structure. 5242 start_address = m_data.GetU32(&offset); 5243 done = true; 5244 } 5245 break; 5246 case llvm::MachO::CPU_TYPE_ARM64: 5247 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5248 { 5249 offset += 256; // This is the offset of pc in the GPR thread state 5250 // data structure. 5251 start_address = m_data.GetU64(&offset); 5252 done = true; 5253 } 5254 break; 5255 case llvm::MachO::CPU_TYPE_I386: 5256 if (flavor == 5257 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5258 { 5259 offset += 40; // This is the offset of eip in the GPR thread state 5260 // data structure. 5261 start_address = m_data.GetU32(&offset); 5262 done = true; 5263 } 5264 break; 5265 case llvm::MachO::CPU_TYPE_X86_64: 5266 if (flavor == 5267 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5268 { 5269 offset += 16 * 8; // This is the offset of rip in the GPR thread 5270 // state data structure. 5271 start_address = m_data.GetU64(&offset); 5272 done = true; 5273 } 5274 break; 5275 default: 5276 return m_entry_point_address; 5277 } 5278 // Haven't found the GPR flavor yet, skip over the data for this 5279 // flavor: 5280 if (done) 5281 break; 5282 offset += count * 4; 5283 } 5284 } break; 5285 case LC_MAIN: { 5286 ConstString text_segment_name("__TEXT"); 5287 uint64_t entryoffset = m_data.GetU64(&offset); 5288 SectionSP text_segment_sp = 5289 GetSectionList()->FindSectionByName(text_segment_name); 5290 if (text_segment_sp) { 5291 done = true; 5292 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5293 } 5294 } break; 5295 5296 default: 5297 break; 5298 } 5299 if (done) 5300 break; 5301 5302 // Go to the next load command: 5303 offset = cmd_offset + load_cmd.cmdsize; 5304 } 5305 5306 if (start_address != LLDB_INVALID_ADDRESS) { 5307 // We got the start address from the load commands, so now resolve that 5308 // address in the sections of this ObjectFile: 5309 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5310 start_address, GetSectionList())) { 5311 m_entry_point_address.Clear(); 5312 } 5313 } else { 5314 // We couldn't read the UnixThread load command - maybe it wasn't there. 5315 // As a fallback look for the "start" symbol in the main executable. 5316 5317 ModuleSP module_sp(GetModule()); 5318 5319 if (module_sp) { 5320 SymbolContextList contexts; 5321 SymbolContext context; 5322 if (module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5323 eSymbolTypeCode, contexts)) { 5324 if (contexts.GetContextAtIndex(0, context)) 5325 m_entry_point_address = context.symbol->GetAddress(); 5326 } 5327 } 5328 } 5329 } 5330 5331 return m_entry_point_address; 5332 } 5333 5334 lldb_private::Address ObjectFileMachO::GetHeaderAddress() { 5335 lldb_private::Address header_addr; 5336 SectionList *section_list = GetSectionList(); 5337 if (section_list) { 5338 SectionSP text_segment_sp( 5339 section_list->FindSectionByName(GetSegmentNameTEXT())); 5340 if (text_segment_sp) { 5341 header_addr.SetSection(text_segment_sp); 5342 header_addr.SetOffset(0); 5343 } 5344 } 5345 return header_addr; 5346 } 5347 5348 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5349 ModuleSP module_sp(GetModule()); 5350 if (module_sp) { 5351 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5352 if (!m_thread_context_offsets_valid) { 5353 m_thread_context_offsets_valid = true; 5354 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5355 FileRangeArray::Entry file_range; 5356 thread_command thread_cmd; 5357 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5358 const uint32_t cmd_offset = offset; 5359 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 5360 break; 5361 5362 if (thread_cmd.cmd == LC_THREAD) { 5363 file_range.SetRangeBase(offset); 5364 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5365 m_thread_context_offsets.Append(file_range); 5366 } 5367 offset = cmd_offset + thread_cmd.cmdsize; 5368 } 5369 } 5370 } 5371 return m_thread_context_offsets.GetSize(); 5372 } 5373 5374 std::string ObjectFileMachO::GetIdentifierString() { 5375 std::string result; 5376 ModuleSP module_sp(GetModule()); 5377 if (module_sp) { 5378 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5379 5380 // First, look over the load commands for an LC_NOTE load command with 5381 // data_owner string "kern ver str" & use that if found. 5382 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5383 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5384 const uint32_t cmd_offset = offset; 5385 load_command lc; 5386 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5387 break; 5388 if (lc.cmd == LC_NOTE) 5389 { 5390 char data_owner[17]; 5391 m_data.CopyData (offset, 16, data_owner); 5392 data_owner[16] = '\0'; 5393 offset += 16; 5394 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5395 uint64_t size = m_data.GetU64_unchecked (&offset); 5396 5397 // "kern ver str" has a uint32_t version and then a nul terminated 5398 // c-string. 5399 if (strcmp ("kern ver str", data_owner) == 0) 5400 { 5401 offset = fileoff; 5402 uint32_t version; 5403 if (m_data.GetU32 (&offset, &version, 1) != nullptr) 5404 { 5405 if (version == 1) 5406 { 5407 uint32_t strsize = size - sizeof (uint32_t); 5408 char *buf = (char*) malloc (strsize); 5409 if (buf) 5410 { 5411 m_data.CopyData (offset, strsize, buf); 5412 buf[strsize - 1] = '\0'; 5413 result = buf; 5414 if (buf) 5415 free (buf); 5416 return result; 5417 } 5418 } 5419 } 5420 } 5421 } 5422 offset = cmd_offset + lc.cmdsize; 5423 } 5424 5425 // Second, make a pass over the load commands looking for an obsolete 5426 // LC_IDENT load command. 5427 offset = MachHeaderSizeFromMagic(m_header.magic); 5428 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5429 const uint32_t cmd_offset = offset; 5430 struct ident_command ident_command; 5431 if (m_data.GetU32(&offset, &ident_command, 2) == NULL) 5432 break; 5433 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5434 char *buf = (char *) malloc (ident_command.cmdsize); 5435 if (buf != nullptr 5436 && m_data.CopyData (offset, ident_command.cmdsize, buf) == ident_command.cmdsize) { 5437 buf[ident_command.cmdsize - 1] = '\0'; 5438 result = buf; 5439 } 5440 if (buf) 5441 free (buf); 5442 } 5443 offset = cmd_offset + ident_command.cmdsize; 5444 } 5445 5446 } 5447 return result; 5448 } 5449 5450 bool ObjectFileMachO::GetCorefileMainBinaryInfo (addr_t &address, UUID &uuid) { 5451 address = LLDB_INVALID_ADDRESS; 5452 uuid.Clear(); 5453 ModuleSP module_sp(GetModule()); 5454 if (module_sp) { 5455 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5456 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5457 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5458 const uint32_t cmd_offset = offset; 5459 load_command lc; 5460 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5461 break; 5462 if (lc.cmd == LC_NOTE) 5463 { 5464 char data_owner[17]; 5465 memset (data_owner, 0, sizeof (data_owner)); 5466 m_data.CopyData (offset, 16, data_owner); 5467 offset += 16; 5468 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5469 uint64_t size = m_data.GetU64_unchecked (&offset); 5470 5471 // "main bin spec" (main binary specification) data payload is 5472 // formatted: 5473 // uint32_t version [currently 1] 5474 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user process] 5475 // uint64_t address [ UINT64_MAX if address not specified ] 5476 // uuid_t uuid [ all zero's if uuid not specified ] 5477 // uint32_t log2_pagesize [ process page size in log base 2, e.g. 4k pages are 12. 0 for unspecified ] 5478 5479 if (strcmp ("main bin spec", data_owner) == 0 && size >= 32) 5480 { 5481 offset = fileoff; 5482 uint32_t version; 5483 if (m_data.GetU32 (&offset, &version, 1) != nullptr && version == 1) 5484 { 5485 uint32_t type = 0; 5486 uuid_t raw_uuid; 5487 memset (raw_uuid, 0, sizeof (uuid_t)); 5488 5489 if (m_data.GetU32(&offset, &type, 1) && 5490 m_data.GetU64(&offset, &address, 1) && 5491 m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5492 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5493 return true; 5494 } 5495 } 5496 } 5497 } 5498 offset = cmd_offset + lc.cmdsize; 5499 } 5500 } 5501 return false; 5502 } 5503 5504 lldb::RegisterContextSP 5505 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5506 lldb_private::Thread &thread) { 5507 lldb::RegisterContextSP reg_ctx_sp; 5508 5509 ModuleSP module_sp(GetModule()); 5510 if (module_sp) { 5511 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5512 if (!m_thread_context_offsets_valid) 5513 GetNumThreadContexts(); 5514 5515 const FileRangeArray::Entry *thread_context_file_range = 5516 m_thread_context_offsets.GetEntryAtIndex(idx); 5517 if (thread_context_file_range) { 5518 5519 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5520 thread_context_file_range->GetByteSize()); 5521 5522 switch (m_header.cputype) { 5523 case llvm::MachO::CPU_TYPE_ARM64: 5524 reg_ctx_sp.reset(new RegisterContextDarwin_arm64_Mach(thread, data)); 5525 break; 5526 5527 case llvm::MachO::CPU_TYPE_ARM: 5528 reg_ctx_sp.reset(new RegisterContextDarwin_arm_Mach(thread, data)); 5529 break; 5530 5531 case llvm::MachO::CPU_TYPE_I386: 5532 reg_ctx_sp.reset(new RegisterContextDarwin_i386_Mach(thread, data)); 5533 break; 5534 5535 case llvm::MachO::CPU_TYPE_X86_64: 5536 reg_ctx_sp.reset(new RegisterContextDarwin_x86_64_Mach(thread, data)); 5537 break; 5538 } 5539 } 5540 } 5541 return reg_ctx_sp; 5542 } 5543 5544 ObjectFile::Type ObjectFileMachO::CalculateType() { 5545 switch (m_header.filetype) { 5546 case MH_OBJECT: // 0x1u 5547 if (GetAddressByteSize() == 4) { 5548 // 32 bit kexts are just object files, but they do have a valid 5549 // UUID load command. 5550 UUID uuid; 5551 if (GetUUID(&uuid)) { 5552 // this checking for the UUID load command is not enough we could 5553 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5554 // this is required of kexts 5555 if (m_strata == eStrataInvalid) 5556 m_strata = eStrataKernel; 5557 return eTypeSharedLibrary; 5558 } 5559 } 5560 return eTypeObjectFile; 5561 5562 case MH_EXECUTE: 5563 return eTypeExecutable; // 0x2u 5564 case MH_FVMLIB: 5565 return eTypeSharedLibrary; // 0x3u 5566 case MH_CORE: 5567 return eTypeCoreFile; // 0x4u 5568 case MH_PRELOAD: 5569 return eTypeSharedLibrary; // 0x5u 5570 case MH_DYLIB: 5571 return eTypeSharedLibrary; // 0x6u 5572 case MH_DYLINKER: 5573 return eTypeDynamicLinker; // 0x7u 5574 case MH_BUNDLE: 5575 return eTypeSharedLibrary; // 0x8u 5576 case MH_DYLIB_STUB: 5577 return eTypeStubLibrary; // 0x9u 5578 case MH_DSYM: 5579 return eTypeDebugInfo; // 0xAu 5580 case MH_KEXT_BUNDLE: 5581 return eTypeSharedLibrary; // 0xBu 5582 default: 5583 break; 5584 } 5585 return eTypeUnknown; 5586 } 5587 5588 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5589 switch (m_header.filetype) { 5590 case MH_OBJECT: // 0x1u 5591 { 5592 // 32 bit kexts are just object files, but they do have a valid 5593 // UUID load command. 5594 UUID uuid; 5595 if (GetUUID(&uuid)) { 5596 // this checking for the UUID load command is not enough we could 5597 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5598 // this is required of kexts 5599 if (m_type == eTypeInvalid) 5600 m_type = eTypeSharedLibrary; 5601 5602 return eStrataKernel; 5603 } 5604 } 5605 return eStrataUnknown; 5606 5607 case MH_EXECUTE: // 0x2u 5608 // Check for the MH_DYLDLINK bit in the flags 5609 if (m_header.flags & MH_DYLDLINK) { 5610 return eStrataUser; 5611 } else { 5612 SectionList *section_list = GetSectionList(); 5613 if (section_list) { 5614 static ConstString g_kld_section_name("__KLD"); 5615 if (section_list->FindSectionByName(g_kld_section_name)) 5616 return eStrataKernel; 5617 } 5618 } 5619 return eStrataRawImage; 5620 5621 case MH_FVMLIB: 5622 return eStrataUser; // 0x3u 5623 case MH_CORE: 5624 return eStrataUnknown; // 0x4u 5625 case MH_PRELOAD: 5626 return eStrataRawImage; // 0x5u 5627 case MH_DYLIB: 5628 return eStrataUser; // 0x6u 5629 case MH_DYLINKER: 5630 return eStrataUser; // 0x7u 5631 case MH_BUNDLE: 5632 return eStrataUser; // 0x8u 5633 case MH_DYLIB_STUB: 5634 return eStrataUser; // 0x9u 5635 case MH_DSYM: 5636 return eStrataUnknown; // 0xAu 5637 case MH_KEXT_BUNDLE: 5638 return eStrataKernel; // 0xBu 5639 default: 5640 break; 5641 } 5642 return eStrataUnknown; 5643 } 5644 5645 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5646 ModuleSP module_sp(GetModule()); 5647 if (module_sp) { 5648 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5649 struct dylib_command load_cmd; 5650 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5651 uint32_t version_cmd = 0; 5652 uint64_t version = 0; 5653 uint32_t i; 5654 for (i = 0; i < m_header.ncmds; ++i) { 5655 const lldb::offset_t cmd_offset = offset; 5656 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5657 break; 5658 5659 if (load_cmd.cmd == LC_ID_DYLIB) { 5660 if (version_cmd == 0) { 5661 version_cmd = load_cmd.cmd; 5662 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5663 break; 5664 version = load_cmd.dylib.current_version; 5665 } 5666 break; // Break for now unless there is another more complete version 5667 // number load command in the future. 5668 } 5669 offset = cmd_offset + load_cmd.cmdsize; 5670 } 5671 5672 if (version_cmd == LC_ID_DYLIB) { 5673 unsigned major = (version & 0xFFFF0000ull) >> 16; 5674 unsigned minor = (version & 0x0000FF00ull) >> 8; 5675 unsigned subminor = (version & 0x000000FFull); 5676 return llvm::VersionTuple(major, minor, subminor); 5677 } 5678 } 5679 return llvm::VersionTuple(); 5680 } 5681 5682 bool ObjectFileMachO::GetArchitecture(ArchSpec &arch) { 5683 ModuleSP module_sp(GetModule()); 5684 if (module_sp) { 5685 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5686 return GetArchitecture(m_header, m_data, 5687 MachHeaderSizeFromMagic(m_header.magic), arch); 5688 } 5689 return false; 5690 } 5691 5692 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, addr_t &base_addr, UUID &uuid) { 5693 uuid.Clear(); 5694 base_addr = LLDB_INVALID_ADDRESS; 5695 if (process && process->GetDynamicLoader()) { 5696 DynamicLoader *dl = process->GetDynamicLoader(); 5697 LazyBool using_shared_cache; 5698 LazyBool private_shared_cache; 5699 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5700 private_shared_cache); 5701 } 5702 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5703 if (log) 5704 log->Printf("inferior process shared cache has a UUID of %s, base address 0x%" PRIx64 , uuid.GetAsString().c_str(), base_addr); 5705 } 5706 5707 // From dyld SPI header dyld_process_info.h 5708 typedef void *dyld_process_info; 5709 struct lldb_copy__dyld_process_cache_info { 5710 uuid_t cacheUUID; // UUID of cache used by process 5711 uint64_t cacheBaseAddress; // load address of dyld shared cache 5712 bool noCache; // process is running without a dyld cache 5713 bool privateCache; // process is using a private copy of its dyld cache 5714 }; 5715 5716 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with llvm 5717 // enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile errors. 5718 // So we need to use the actual underlying types of task_t and kern_return_t 5719 // below. 5720 extern "C" unsigned int /*task_t*/ mach_task_self(); 5721 5722 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5723 uuid.Clear(); 5724 base_addr = LLDB_INVALID_ADDRESS; 5725 5726 #if defined(__APPLE__) && \ 5727 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5728 uint8_t *(*dyld_get_all_image_infos)(void); 5729 dyld_get_all_image_infos = 5730 (uint8_t * (*)())dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5731 if (dyld_get_all_image_infos) { 5732 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5733 if (dyld_all_image_infos_address) { 5734 uint32_t *version = (uint32_t *) 5735 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5736 if (*version >= 13) { 5737 uuid_t *sharedCacheUUID_address = 0; 5738 int wordsize = sizeof(uint8_t *); 5739 if (wordsize == 8) { 5740 sharedCacheUUID_address = 5741 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5742 160); // sharedCacheUUID <mach-o/dyld_images.h> 5743 if (*version >= 15) 5744 base_addr = *(uint64_t *) ((uint8_t *) dyld_all_image_infos_address 5745 + 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5746 } else { 5747 sharedCacheUUID_address = 5748 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5749 84); // sharedCacheUUID <mach-o/dyld_images.h> 5750 if (*version >= 15) { 5751 base_addr = 0; 5752 base_addr = *(uint32_t *) ((uint8_t *) dyld_all_image_infos_address 5753 + 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5754 } 5755 } 5756 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5757 } 5758 } 5759 } else { 5760 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5761 dyld_process_info (*dyld_process_info_create)(unsigned int /* task_t */ task, uint64_t timestamp, unsigned int /*kern_return_t*/ *kernelError); 5762 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5763 void (*dyld_process_info_release)(dyld_process_info info); 5764 5765 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, unsigned int /*kern_return_t*/ *)) 5766 dlsym (RTLD_DEFAULT, "_dyld_process_info_create"); 5767 dyld_process_info_get_cache = (void (*)(void *, void *)) 5768 dlsym (RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5769 dyld_process_info_release = (void (*)(void *)) 5770 dlsym (RTLD_DEFAULT, "_dyld_process_info_release"); 5771 5772 if (dyld_process_info_create && dyld_process_info_get_cache) { 5773 unsigned int /*kern_return_t */ kern_ret; 5774 dyld_process_info process_info = dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5775 if (process_info) { 5776 struct lldb_copy__dyld_process_cache_info sc_info; 5777 memset (&sc_info, 0, sizeof (struct lldb_copy__dyld_process_cache_info)); 5778 dyld_process_info_get_cache (process_info, &sc_info); 5779 if (sc_info.cacheBaseAddress != 0) { 5780 base_addr = sc_info.cacheBaseAddress; 5781 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 5782 } 5783 dyld_process_info_release (process_info); 5784 } 5785 } 5786 } 5787 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5788 if (log && uuid.IsValid()) 5789 log->Printf("lldb's in-memory shared cache has a UUID of %s base address of 0x%" PRIx64, uuid.GetAsString().c_str(), base_addr); 5790 #endif 5791 } 5792 5793 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5794 if (!m_min_os_version) { 5795 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5796 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5797 const lldb::offset_t load_cmd_offset = offset; 5798 5799 version_min_command lc; 5800 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5801 break; 5802 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5803 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5804 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5805 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5806 if (m_data.GetU32(&offset, &lc.version, 5807 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5808 const uint32_t xxxx = lc.version >> 16; 5809 const uint32_t yy = (lc.version >> 8) & 0xffu; 5810 const uint32_t zz = lc.version & 0xffu; 5811 if (xxxx) { 5812 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5813 break; 5814 } 5815 } 5816 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5817 // struct build_version_command { 5818 // uint32_t cmd; /* LC_BUILD_VERSION */ 5819 // uint32_t cmdsize; /* sizeof(struct build_version_command) plus */ 5820 // /* ntools * sizeof(struct build_tool_version) */ 5821 // uint32_t platform; /* platform */ 5822 // uint32_t minos; /* X.Y.Z is encoded in nibbles xxxx.yy.zz */ 5823 // uint32_t sdk; /* X.Y.Z is encoded in nibbles xxxx.yy.zz */ 5824 // uint32_t ntools; /* number of tool entries following this */ 5825 // }; 5826 5827 offset += 4; // skip platform 5828 uint32_t minos = m_data.GetU32(&offset); 5829 5830 const uint32_t xxxx = minos >> 16; 5831 const uint32_t yy = (minos >> 8) & 0xffu; 5832 const uint32_t zz = minos & 0xffu; 5833 if (xxxx) { 5834 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5835 break; 5836 } 5837 } 5838 5839 offset = load_cmd_offset + lc.cmdsize; 5840 } 5841 5842 if (!m_min_os_version) { 5843 // Set version to an empty value so we don't keep trying to 5844 m_min_os_version = llvm::VersionTuple(); 5845 } 5846 } 5847 5848 return *m_min_os_version; 5849 } 5850 5851 uint32_t ObjectFileMachO::GetSDKVersion(uint32_t *versions, 5852 uint32_t num_versions) { 5853 if (m_sdk_versions.empty()) { 5854 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5855 bool success = false; 5856 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5857 const lldb::offset_t load_cmd_offset = offset; 5858 5859 version_min_command lc; 5860 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5861 break; 5862 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5863 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5864 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5865 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5866 if (m_data.GetU32(&offset, &lc.version, 5867 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5868 const uint32_t xxxx = lc.sdk >> 16; 5869 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5870 const uint32_t zz = lc.sdk & 0xffu; 5871 if (xxxx) { 5872 m_sdk_versions.push_back(xxxx); 5873 m_sdk_versions.push_back(yy); 5874 m_sdk_versions.push_back(zz); 5875 success = true; 5876 } else { 5877 GetModule()->ReportWarning( 5878 "minimum OS version load command with invalid (0) version found."); 5879 } 5880 } 5881 } 5882 offset = load_cmd_offset + lc.cmdsize; 5883 } 5884 5885 if (success == false) 5886 { 5887 offset = MachHeaderSizeFromMagic(m_header.magic); 5888 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) 5889 { 5890 const lldb::offset_t load_cmd_offset = offset; 5891 5892 version_min_command lc; 5893 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5894 break; 5895 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) 5896 { 5897 // struct build_version_command { 5898 // uint32_t cmd; /* LC_BUILD_VERSION */ 5899 // uint32_t cmdsize; /* sizeof(struct build_version_command) plus */ 5900 // /* ntools * sizeof(struct build_tool_version) */ 5901 // uint32_t platform; /* platform */ 5902 // uint32_t minos; /* X.Y.Z is encoded in nibbles xxxx.yy.zz */ 5903 // uint32_t sdk; /* X.Y.Z is encoded in nibbles xxxx.yy.zz */ 5904 // uint32_t ntools; /* number of tool entries following this */ 5905 // }; 5906 5907 offset += 4; // skip platform 5908 uint32_t minos = m_data.GetU32(&offset); 5909 5910 const uint32_t xxxx = minos >> 16; 5911 const uint32_t yy = (minos >> 8) & 0xffu; 5912 const uint32_t zz = minos & 0xffu; 5913 if (xxxx) 5914 { 5915 m_sdk_versions.push_back (xxxx); 5916 m_sdk_versions.push_back (yy); 5917 m_sdk_versions.push_back (zz); 5918 success = true; 5919 } 5920 } 5921 offset = load_cmd_offset + lc.cmdsize; 5922 } 5923 } 5924 5925 if (success == false) { 5926 // Push an invalid value so we don't try to find 5927 // the version # again on the next call to this 5928 // method. 5929 m_sdk_versions.push_back(UINT32_MAX); 5930 } 5931 } 5932 5933 // Legitimate version numbers will have 3 entries pushed 5934 // on to m_sdk_versions. If we only have one value, it's 5935 // the sentinel value indicating that this object file 5936 // does not have a valid minimum os version #. 5937 if (m_sdk_versions.size() > 1) { 5938 if (versions != NULL && num_versions > 0) { 5939 for (size_t i = 0; i < num_versions; ++i) { 5940 if (i < m_sdk_versions.size()) 5941 versions[i] = m_sdk_versions[i]; 5942 else 5943 versions[i] = 0; 5944 } 5945 } 5946 return m_sdk_versions.size(); 5947 } 5948 // Call the superclasses version that will empty out the data 5949 return ObjectFile::GetSDKVersion(versions, num_versions); 5950 } 5951 5952 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5953 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5954 } 5955 5956 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5957 return m_allow_assembly_emulation_unwind_plans; 5958 } 5959 5960 //------------------------------------------------------------------ 5961 // PluginInterface protocol 5962 //------------------------------------------------------------------ 5963 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5964 return GetPluginNameStatic(); 5965 } 5966 5967 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5968 5969 Section *ObjectFileMachO::GetMachHeaderSection() { 5970 // Find the first address of the mach header which is the first non-zero file 5971 // sized section whose file offset is zero. This is the base file address of 5972 // the mach-o file which can be subtracted from the vmaddr of the other 5973 // segments found in memory and added to the load address 5974 ModuleSP module_sp = GetModule(); 5975 if (module_sp) { 5976 SectionList *section_list = GetSectionList(); 5977 if (section_list) { 5978 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5979 const size_t num_sections = section_list->GetSize(); 5980 5981 for (size_t sect_idx = 0; sect_idx < num_sections && 5982 mach_base_file_addr == LLDB_INVALID_ADDRESS; 5983 ++sect_idx) { 5984 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5985 if (section && section->GetFileSize() > 0 && 5986 section->GetFileOffset() == 0 && 5987 section->IsThreadSpecific() == false && 5988 module_sp.get() == section->GetModule().get()) { 5989 return section; 5990 } 5991 } 5992 } 5993 } 5994 return nullptr; 5995 } 5996 5997 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 5998 lldb::addr_t mach_header_load_address, const Section *mach_header_section, 5999 const Section *section) { 6000 ModuleSP module_sp = GetModule(); 6001 if (module_sp && mach_header_section && section && 6002 mach_header_load_address != LLDB_INVALID_ADDRESS) { 6003 lldb::addr_t mach_header_file_addr = mach_header_section->GetFileAddress(); 6004 if (mach_header_file_addr != LLDB_INVALID_ADDRESS) { 6005 if (section && section->GetFileSize() > 0 && 6006 section->IsThreadSpecific() == false && 6007 module_sp.get() == section->GetModule().get()) { 6008 // Ignore __LINKEDIT and __DWARF segments 6009 if (section->GetName() == GetSegmentNameLINKEDIT()) { 6010 // Only map __LINKEDIT if we have an in memory image and this isn't a 6011 // kernel binary like a kext or mach_kernel. 6012 const bool is_memory_image = (bool)m_process_wp.lock(); 6013 const Strata strata = GetStrata(); 6014 if (is_memory_image == false || strata == eStrataKernel) 6015 return LLDB_INVALID_ADDRESS; 6016 } 6017 return section->GetFileAddress() - mach_header_file_addr + 6018 mach_header_load_address; 6019 } 6020 } 6021 } 6022 return LLDB_INVALID_ADDRESS; 6023 } 6024 6025 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 6026 bool value_is_offset) { 6027 ModuleSP module_sp = GetModule(); 6028 if (module_sp) { 6029 size_t num_loaded_sections = 0; 6030 SectionList *section_list = GetSectionList(); 6031 if (section_list) { 6032 const size_t num_sections = section_list->GetSize(); 6033 6034 if (value_is_offset) { 6035 // "value" is an offset to apply to each top level segment 6036 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6037 // Iterate through the object file sections to find all of the 6038 // sections that size on disk (to avoid __PAGEZERO) and load them 6039 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6040 if (section_sp && section_sp->GetFileSize() > 0 && 6041 section_sp->IsThreadSpecific() == false && 6042 module_sp.get() == section_sp->GetModule().get()) { 6043 // Ignore __LINKEDIT and __DWARF segments 6044 if (section_sp->GetName() == GetSegmentNameLINKEDIT()) { 6045 // Only map __LINKEDIT if we have an in memory image and this 6046 // isn't a kernel binary like a kext or mach_kernel. 6047 const bool is_memory_image = (bool)m_process_wp.lock(); 6048 const Strata strata = GetStrata(); 6049 if (is_memory_image == false || strata == eStrataKernel) 6050 continue; 6051 } 6052 if (target.GetSectionLoadList().SetSectionLoadAddress( 6053 section_sp, section_sp->GetFileAddress() + value)) 6054 ++num_loaded_sections; 6055 } 6056 } 6057 } else { 6058 // "value" is the new base address of the mach_header, adjust each 6059 // section accordingly 6060 6061 Section *mach_header_section = GetMachHeaderSection(); 6062 if (mach_header_section) { 6063 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6064 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6065 6066 lldb::addr_t section_load_addr = 6067 CalculateSectionLoadAddressForMemoryImage( 6068 value, mach_header_section, section_sp.get()); 6069 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6070 if (target.GetSectionLoadList().SetSectionLoadAddress( 6071 section_sp, section_load_addr)) 6072 ++num_loaded_sections; 6073 } 6074 } 6075 } 6076 } 6077 } 6078 return num_loaded_sections > 0; 6079 } 6080 return false; 6081 } 6082 6083 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6084 const FileSpec &outfile, Status &error) { 6085 if (process_sp) { 6086 Target &target = process_sp->GetTarget(); 6087 const ArchSpec target_arch = target.GetArchitecture(); 6088 const llvm::Triple &target_triple = target_arch.GetTriple(); 6089 if (target_triple.getVendor() == llvm::Triple::Apple && 6090 (target_triple.getOS() == llvm::Triple::MacOSX || 6091 target_triple.getOS() == llvm::Triple::IOS || 6092 target_triple.getOS() == llvm::Triple::WatchOS || 6093 target_triple.getOS() == llvm::Triple::TvOS)) { 6094 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) { 6095 bool make_core = false; 6096 switch (target_arch.GetMachine()) { 6097 case llvm::Triple::aarch64: 6098 case llvm::Triple::arm: 6099 case llvm::Triple::thumb: 6100 case llvm::Triple::x86: 6101 case llvm::Triple::x86_64: 6102 make_core = true; 6103 break; 6104 default: 6105 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6106 target_triple.str().c_str()); 6107 break; 6108 } 6109 6110 if (make_core) { 6111 std::vector<segment_command_64> segment_load_commands; 6112 // uint32_t range_info_idx = 0; 6113 MemoryRegionInfo range_info; 6114 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6115 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6116 const ByteOrder byte_order = target_arch.GetByteOrder(); 6117 if (range_error.Success()) { 6118 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6119 const addr_t addr = range_info.GetRange().GetRangeBase(); 6120 const addr_t size = range_info.GetRange().GetByteSize(); 6121 6122 if (size == 0) 6123 break; 6124 6125 // Calculate correct protections 6126 uint32_t prot = 0; 6127 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6128 prot |= VM_PROT_READ; 6129 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6130 prot |= VM_PROT_WRITE; 6131 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6132 prot |= VM_PROT_EXECUTE; 6133 6134 // printf ("[%3u] [0x%16.16" PRIx64 " - 6135 // 0x%16.16" PRIx64 ") %c%c%c\n", 6136 // range_info_idx, 6137 // addr, 6138 // size, 6139 // (prot & VM_PROT_READ ) ? 'r' : 6140 // '-', 6141 // (prot & VM_PROT_WRITE ) ? 'w' : 6142 // '-', 6143 // (prot & VM_PROT_EXECUTE) ? 'x' : 6144 // '-'); 6145 6146 if (prot != 0) { 6147 uint32_t cmd_type = LC_SEGMENT_64; 6148 uint32_t segment_size = sizeof(segment_command_64); 6149 if (addr_byte_size == 4) { 6150 cmd_type = LC_SEGMENT; 6151 segment_size = sizeof(segment_command); 6152 } 6153 segment_command_64 segment = { 6154 cmd_type, // uint32_t cmd; 6155 segment_size, // uint32_t cmdsize; 6156 {0}, // char segname[16]; 6157 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6158 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6159 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6160 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6161 prot, // uint32_t maxprot; 6162 prot, // uint32_t initprot; 6163 0, // uint32_t nsects; 6164 0}; // uint32_t flags; 6165 segment_load_commands.push_back(segment); 6166 } else { 6167 // No protections and a size of 1 used to be returned from old 6168 // debugservers when we asked about a region that was past the 6169 // last memory region and it indicates the end... 6170 if (size == 1) 6171 break; 6172 } 6173 6174 range_error = process_sp->GetMemoryRegionInfo( 6175 range_info.GetRange().GetRangeEnd(), range_info); 6176 if (range_error.Fail()) 6177 break; 6178 } 6179 6180 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6181 6182 mach_header_64 mach_header; 6183 if (addr_byte_size == 8) { 6184 mach_header.magic = MH_MAGIC_64; 6185 } else { 6186 mach_header.magic = MH_MAGIC; 6187 } 6188 mach_header.cputype = target_arch.GetMachOCPUType(); 6189 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6190 mach_header.filetype = MH_CORE; 6191 mach_header.ncmds = segment_load_commands.size(); 6192 mach_header.flags = 0; 6193 mach_header.reserved = 0; 6194 ThreadList &thread_list = process_sp->GetThreadList(); 6195 const uint32_t num_threads = thread_list.GetSize(); 6196 6197 // Make an array of LC_THREAD data items. Each one contains the 6198 // contents of the LC_THREAD load command. The data doesn't contain 6199 // the load command + load command size, we will add the load command 6200 // and load command size as we emit the data. 6201 std::vector<StreamString> LC_THREAD_datas(num_threads); 6202 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6203 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6204 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6205 LC_THREAD_data.SetByteOrder(byte_order); 6206 } 6207 for (uint32_t thread_idx = 0; thread_idx < num_threads; 6208 ++thread_idx) { 6209 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6210 if (thread_sp) { 6211 switch (mach_header.cputype) { 6212 case llvm::MachO::CPU_TYPE_ARM64: 6213 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6214 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6215 break; 6216 6217 case llvm::MachO::CPU_TYPE_ARM: 6218 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6219 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6220 break; 6221 6222 case llvm::MachO::CPU_TYPE_I386: 6223 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6224 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6225 break; 6226 6227 case llvm::MachO::CPU_TYPE_X86_64: 6228 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6229 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6230 break; 6231 } 6232 } 6233 } 6234 6235 // The size of the load command is the size of the segments... 6236 if (addr_byte_size == 8) { 6237 mach_header.sizeofcmds = segment_load_commands.size() * 6238 sizeof(struct segment_command_64); 6239 } else { 6240 mach_header.sizeofcmds = 6241 segment_load_commands.size() * sizeof(struct segment_command); 6242 } 6243 6244 // and the size of all LC_THREAD load command 6245 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6246 ++mach_header.ncmds; 6247 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6248 } 6249 6250 printf("mach_header: 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x " 6251 "0x%8.8x 0x%8.8x\n", 6252 mach_header.magic, mach_header.cputype, mach_header.cpusubtype, 6253 mach_header.filetype, mach_header.ncmds, 6254 mach_header.sizeofcmds, mach_header.flags, 6255 mach_header.reserved); 6256 6257 // Write the mach header 6258 buffer.PutHex32(mach_header.magic); 6259 buffer.PutHex32(mach_header.cputype); 6260 buffer.PutHex32(mach_header.cpusubtype); 6261 buffer.PutHex32(mach_header.filetype); 6262 buffer.PutHex32(mach_header.ncmds); 6263 buffer.PutHex32(mach_header.sizeofcmds); 6264 buffer.PutHex32(mach_header.flags); 6265 if (addr_byte_size == 8) { 6266 buffer.PutHex32(mach_header.reserved); 6267 } 6268 6269 // Skip the mach header and all load commands and align to the next 6270 // 0x1000 byte boundary 6271 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6272 if (file_offset & 0x00000fff) { 6273 file_offset += 0x00001000ull; 6274 file_offset &= (~0x00001000ull + 1); 6275 } 6276 6277 for (auto &segment : segment_load_commands) { 6278 segment.fileoff = file_offset; 6279 file_offset += segment.filesize; 6280 } 6281 6282 // Write out all of the LC_THREAD load commands 6283 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6284 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6285 buffer.PutHex32(LC_THREAD); 6286 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6287 buffer.Write(LC_THREAD_data.GetString().data(), 6288 LC_THREAD_data_size); 6289 } 6290 6291 // Write out all of the segment load commands 6292 for (const auto &segment : segment_load_commands) { 6293 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6294 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6295 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6296 segment.cmd, segment.cmdsize, segment.vmaddr, 6297 segment.vmaddr + segment.vmsize, segment.fileoff, 6298 segment.filesize, segment.maxprot, segment.initprot, 6299 segment.nsects, segment.flags); 6300 6301 buffer.PutHex32(segment.cmd); 6302 buffer.PutHex32(segment.cmdsize); 6303 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6304 if (addr_byte_size == 8) { 6305 buffer.PutHex64(segment.vmaddr); 6306 buffer.PutHex64(segment.vmsize); 6307 buffer.PutHex64(segment.fileoff); 6308 buffer.PutHex64(segment.filesize); 6309 } else { 6310 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6311 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6312 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6313 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6314 } 6315 buffer.PutHex32(segment.maxprot); 6316 buffer.PutHex32(segment.initprot); 6317 buffer.PutHex32(segment.nsects); 6318 buffer.PutHex32(segment.flags); 6319 } 6320 6321 File core_file; 6322 std::string core_file_path(outfile.GetPath()); 6323 error = FileSystem::Instance().Open(core_file, outfile, 6324 File::eOpenOptionWrite | 6325 File::eOpenOptionTruncate | 6326 File::eOpenOptionCanCreate); 6327 if (error.Success()) { 6328 // Read 1 page at a time 6329 uint8_t bytes[0x1000]; 6330 // Write the mach header and load commands out to the core file 6331 size_t bytes_written = buffer.GetString().size(); 6332 error = core_file.Write(buffer.GetString().data(), bytes_written); 6333 if (error.Success()) { 6334 // Now write the file data for all memory segments in the process 6335 for (const auto &segment : segment_load_commands) { 6336 if (core_file.SeekFromStart(segment.fileoff) == -1) { 6337 error.SetErrorStringWithFormat( 6338 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6339 segment.fileoff, core_file_path.c_str()); 6340 break; 6341 } 6342 6343 printf("Saving %" PRId64 6344 " bytes of data for memory region at 0x%" PRIx64 "\n", 6345 segment.vmsize, segment.vmaddr); 6346 addr_t bytes_left = segment.vmsize; 6347 addr_t addr = segment.vmaddr; 6348 Status memory_read_error; 6349 while (bytes_left > 0 && error.Success()) { 6350 const size_t bytes_to_read = 6351 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6352 const size_t bytes_read = process_sp->ReadMemory( 6353 addr, bytes, bytes_to_read, memory_read_error); 6354 if (bytes_read == bytes_to_read) { 6355 size_t bytes_written = bytes_read; 6356 error = core_file.Write(bytes, bytes_written); 6357 bytes_left -= bytes_read; 6358 addr += bytes_read; 6359 } else { 6360 // Some pages within regions are not readable, those should 6361 // be zero filled 6362 memset(bytes, 0, bytes_to_read); 6363 size_t bytes_written = bytes_to_read; 6364 error = core_file.Write(bytes, bytes_written); 6365 bytes_left -= bytes_to_read; 6366 addr += bytes_to_read; 6367 } 6368 } 6369 } 6370 } 6371 } 6372 } else { 6373 error.SetErrorString( 6374 "process doesn't support getting memory region info"); 6375 } 6376 } 6377 return true; // This is the right plug to handle saving core files for 6378 // this process 6379 } 6380 } 6381 return false; 6382 } 6383