1 //===-- ObjectFileMachO.cpp -----------------------------------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 9 #include "llvm/ADT/StringRef.h" 10 11 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 12 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 13 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 14 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 15 #include "lldb/Core/Debugger.h" 16 #include "lldb/Core/FileSpecList.h" 17 #include "lldb/Core/Module.h" 18 #include "lldb/Core/ModuleSpec.h" 19 #include "lldb/Core/PluginManager.h" 20 #include "lldb/Core/Progress.h" 21 #include "lldb/Core/Section.h" 22 #include "lldb/Core/StreamFile.h" 23 #include "lldb/Host/Host.h" 24 #include "lldb/Host/SafeMachO.h" 25 #include "lldb/Symbol/DWARFCallFrameInfo.h" 26 #include "lldb/Symbol/LocateSymbolFile.h" 27 #include "lldb/Symbol/ObjectFile.h" 28 #include "lldb/Target/DynamicLoader.h" 29 #include "lldb/Target/MemoryRegionInfo.h" 30 #include "lldb/Target/Platform.h" 31 #include "lldb/Target/Process.h" 32 #include "lldb/Target/SectionLoadList.h" 33 #include "lldb/Target/Target.h" 34 #include "lldb/Target/Thread.h" 35 #include "lldb/Target/ThreadList.h" 36 #include "lldb/Utility/ArchSpec.h" 37 #include "lldb/Utility/DataBuffer.h" 38 #include "lldb/Utility/FileSpec.h" 39 #include "lldb/Utility/Log.h" 40 #include "lldb/Utility/RangeMap.h" 41 #include "lldb/Utility/RegisterValue.h" 42 #include "lldb/Utility/Status.h" 43 #include "lldb/Utility/StreamString.h" 44 #include "lldb/Utility/Timer.h" 45 #include "lldb/Utility/UUID.h" 46 47 #include "llvm/ADT/DenseSet.h" 48 #include "llvm/Support/FormatVariadic.h" 49 #include "llvm/Support/MemoryBuffer.h" 50 51 #include "ObjectFileMachO.h" 52 53 #if defined(__APPLE__) 54 #include <TargetConditionals.h> 55 // GetLLDBSharedCacheUUID() needs to call dlsym() 56 #include <dlfcn.h> 57 #endif 58 59 #ifndef __APPLE__ 60 #include "Utility/UuidCompatibility.h" 61 #else 62 #include <uuid/uuid.h> 63 #endif 64 65 #include <bitset> 66 #include <memory> 67 68 #if LLVM_SUPPORT_XCODE_SIGNPOSTS 69 // Unfortunately the signpost header pulls in the system MachO header, too. 70 #undef CPU_TYPE_ARM 71 #undef CPU_TYPE_ARM64 72 #undef CPU_TYPE_ARM64_32 73 #undef CPU_TYPE_I386 74 #undef CPU_TYPE_X86_64 75 #undef MH_BINDATLOAD 76 #undef MH_BUNDLE 77 #undef MH_CIGAM 78 #undef MH_CIGAM_64 79 #undef MH_CORE 80 #undef MH_DSYM 81 #undef MH_DYLDLINK 82 #undef MH_DYLIB 83 #undef MH_DYLIB_STUB 84 #undef MH_DYLINKER 85 #undef MH_DYLINKER 86 #undef MH_EXECUTE 87 #undef MH_FVMLIB 88 #undef MH_INCRLINK 89 #undef MH_KEXT_BUNDLE 90 #undef MH_MAGIC 91 #undef MH_MAGIC_64 92 #undef MH_NOUNDEFS 93 #undef MH_OBJECT 94 #undef MH_OBJECT 95 #undef MH_PRELOAD 96 97 #undef LC_BUILD_VERSION 98 #undef LC_VERSION_MIN_MACOSX 99 #undef LC_VERSION_MIN_IPHONEOS 100 #undef LC_VERSION_MIN_TVOS 101 #undef LC_VERSION_MIN_WATCHOS 102 103 #undef PLATFORM_MACOS 104 #undef PLATFORM_MACCATALYST 105 #undef PLATFORM_IOS 106 #undef PLATFORM_IOSSIMULATOR 107 #undef PLATFORM_TVOS 108 #undef PLATFORM_TVOSSIMULATOR 109 #undef PLATFORM_WATCHOS 110 #undef PLATFORM_WATCHOSSIMULATOR 111 #endif 112 113 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 114 using namespace lldb; 115 using namespace lldb_private; 116 using namespace llvm::MachO; 117 118 LLDB_PLUGIN_DEFINE(ObjectFileMachO) 119 120 // Some structure definitions needed for parsing the dyld shared cache files 121 // found on iOS devices. 122 123 struct lldb_copy_dyld_cache_header_v1 { 124 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 125 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 126 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 127 uint32_t imagesOffset; 128 uint32_t imagesCount; 129 uint64_t dyldBaseAddress; 130 uint64_t codeSignatureOffset; 131 uint64_t codeSignatureSize; 132 uint64_t slideInfoOffset; 133 uint64_t slideInfoSize; 134 uint64_t localSymbolsOffset; 135 uint64_t localSymbolsSize; 136 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 137 // and later 138 }; 139 140 struct lldb_copy_dyld_cache_mapping_info { 141 uint64_t address; 142 uint64_t size; 143 uint64_t fileOffset; 144 uint32_t maxProt; 145 uint32_t initProt; 146 }; 147 148 struct lldb_copy_dyld_cache_local_symbols_info { 149 uint32_t nlistOffset; 150 uint32_t nlistCount; 151 uint32_t stringsOffset; 152 uint32_t stringsSize; 153 uint32_t entriesOffset; 154 uint32_t entriesCount; 155 }; 156 struct lldb_copy_dyld_cache_local_symbols_entry { 157 uint32_t dylibOffset; 158 uint32_t nlistStartIndex; 159 uint32_t nlistCount; 160 }; 161 162 static void PrintRegisterValue(RegisterContext *reg_ctx, const char *name, 163 const char *alt_name, size_t reg_byte_size, 164 Stream &data) { 165 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 166 if (reg_info == nullptr) 167 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 168 if (reg_info) { 169 lldb_private::RegisterValue reg_value; 170 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 171 if (reg_info->byte_size >= reg_byte_size) 172 data.Write(reg_value.GetBytes(), reg_byte_size); 173 else { 174 data.Write(reg_value.GetBytes(), reg_info->byte_size); 175 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; ++i) 176 data.PutChar(0); 177 } 178 return; 179 } 180 } 181 // Just write zeros if all else fails 182 for (size_t i = 0; i < reg_byte_size; ++i) 183 data.PutChar(0); 184 } 185 186 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 187 public: 188 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 189 const DataExtractor &data) 190 : RegisterContextDarwin_x86_64(thread, 0) { 191 SetRegisterDataFrom_LC_THREAD(data); 192 } 193 194 void InvalidateAllRegisters() override { 195 // Do nothing... registers are always valid... 196 } 197 198 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 199 lldb::offset_t offset = 0; 200 SetError(GPRRegSet, Read, -1); 201 SetError(FPURegSet, Read, -1); 202 SetError(EXCRegSet, Read, -1); 203 bool done = false; 204 205 while (!done) { 206 int flavor = data.GetU32(&offset); 207 if (flavor == 0) 208 done = true; 209 else { 210 uint32_t i; 211 uint32_t count = data.GetU32(&offset); 212 switch (flavor) { 213 case GPRRegSet: 214 for (i = 0; i < count; ++i) 215 (&gpr.rax)[i] = data.GetU64(&offset); 216 SetError(GPRRegSet, Read, 0); 217 done = true; 218 219 break; 220 case FPURegSet: 221 // TODO: fill in FPU regs.... 222 // SetError (FPURegSet, Read, -1); 223 done = true; 224 225 break; 226 case EXCRegSet: 227 exc.trapno = data.GetU32(&offset); 228 exc.err = data.GetU32(&offset); 229 exc.faultvaddr = data.GetU64(&offset); 230 SetError(EXCRegSet, Read, 0); 231 done = true; 232 break; 233 case 7: 234 case 8: 235 case 9: 236 // fancy flavors that encapsulate of the above flavors... 237 break; 238 239 default: 240 done = true; 241 break; 242 } 243 } 244 } 245 } 246 247 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 248 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 249 if (reg_ctx_sp) { 250 RegisterContext *reg_ctx = reg_ctx_sp.get(); 251 252 data.PutHex32(GPRRegSet); // Flavor 253 data.PutHex32(GPRWordCount); 254 PrintRegisterValue(reg_ctx, "rax", nullptr, 8, data); 255 PrintRegisterValue(reg_ctx, "rbx", nullptr, 8, data); 256 PrintRegisterValue(reg_ctx, "rcx", nullptr, 8, data); 257 PrintRegisterValue(reg_ctx, "rdx", nullptr, 8, data); 258 PrintRegisterValue(reg_ctx, "rdi", nullptr, 8, data); 259 PrintRegisterValue(reg_ctx, "rsi", nullptr, 8, data); 260 PrintRegisterValue(reg_ctx, "rbp", nullptr, 8, data); 261 PrintRegisterValue(reg_ctx, "rsp", nullptr, 8, data); 262 PrintRegisterValue(reg_ctx, "r8", nullptr, 8, data); 263 PrintRegisterValue(reg_ctx, "r9", nullptr, 8, data); 264 PrintRegisterValue(reg_ctx, "r10", nullptr, 8, data); 265 PrintRegisterValue(reg_ctx, "r11", nullptr, 8, data); 266 PrintRegisterValue(reg_ctx, "r12", nullptr, 8, data); 267 PrintRegisterValue(reg_ctx, "r13", nullptr, 8, data); 268 PrintRegisterValue(reg_ctx, "r14", nullptr, 8, data); 269 PrintRegisterValue(reg_ctx, "r15", nullptr, 8, data); 270 PrintRegisterValue(reg_ctx, "rip", nullptr, 8, data); 271 PrintRegisterValue(reg_ctx, "rflags", nullptr, 8, data); 272 PrintRegisterValue(reg_ctx, "cs", nullptr, 8, data); 273 PrintRegisterValue(reg_ctx, "fs", nullptr, 8, data); 274 PrintRegisterValue(reg_ctx, "gs", nullptr, 8, data); 275 276 // // Write out the FPU registers 277 // const size_t fpu_byte_size = sizeof(FPU); 278 // size_t bytes_written = 0; 279 // data.PutHex32 (FPURegSet); 280 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 281 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 282 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 283 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 284 // data); // uint16_t fcw; // "fctrl" 285 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 286 // data); // uint16_t fsw; // "fstat" 287 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 288 // data); // uint8_t ftw; // "ftag" 289 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 290 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 291 // data); // uint16_t fop; // "fop" 292 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 293 // data); // uint32_t ip; // "fioff" 294 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 295 // data); // uint16_t cs; // "fiseg" 296 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 297 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 298 // data); // uint32_t dp; // "fooff" 299 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 300 // data); // uint16_t ds; // "foseg" 301 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 302 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 303 // data); // uint32_t mxcsr; 304 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 305 // 4, data);// uint32_t mxcsrmask; 306 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 307 // sizeof(MMSReg), data); 308 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 309 // sizeof(MMSReg), data); 310 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 311 // sizeof(MMSReg), data); 312 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 313 // sizeof(MMSReg), data); 314 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 315 // sizeof(MMSReg), data); 316 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 317 // sizeof(MMSReg), data); 318 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 319 // sizeof(MMSReg), data); 320 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 321 // sizeof(MMSReg), data); 322 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 323 // sizeof(XMMReg), data); 324 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 325 // sizeof(XMMReg), data); 326 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 327 // sizeof(XMMReg), data); 328 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 329 // sizeof(XMMReg), data); 330 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 331 // sizeof(XMMReg), data); 332 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 333 // sizeof(XMMReg), data); 334 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 335 // sizeof(XMMReg), data); 336 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 337 // sizeof(XMMReg), data); 338 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 339 // sizeof(XMMReg), data); 340 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 341 // sizeof(XMMReg), data); 342 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 343 // sizeof(XMMReg), data); 344 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 345 // sizeof(XMMReg), data); 346 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 347 // sizeof(XMMReg), data); 348 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 349 // sizeof(XMMReg), data); 350 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 351 // sizeof(XMMReg), data); 352 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 353 // sizeof(XMMReg), data); 354 // 355 // // Fill rest with zeros 356 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 357 // i) 358 // data.PutChar(0); 359 360 // Write out the EXC registers 361 data.PutHex32(EXCRegSet); 362 data.PutHex32(EXCWordCount); 363 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 364 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 365 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 8, data); 366 return true; 367 } 368 return false; 369 } 370 371 protected: 372 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 373 374 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 375 376 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 377 378 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 379 return 0; 380 } 381 382 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 383 return 0; 384 } 385 386 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 387 return 0; 388 } 389 }; 390 391 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 392 public: 393 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 394 const DataExtractor &data) 395 : RegisterContextDarwin_i386(thread, 0) { 396 SetRegisterDataFrom_LC_THREAD(data); 397 } 398 399 void InvalidateAllRegisters() override { 400 // Do nothing... registers are always valid... 401 } 402 403 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 404 lldb::offset_t offset = 0; 405 SetError(GPRRegSet, Read, -1); 406 SetError(FPURegSet, Read, -1); 407 SetError(EXCRegSet, Read, -1); 408 bool done = false; 409 410 while (!done) { 411 int flavor = data.GetU32(&offset); 412 if (flavor == 0) 413 done = true; 414 else { 415 uint32_t i; 416 uint32_t count = data.GetU32(&offset); 417 switch (flavor) { 418 case GPRRegSet: 419 for (i = 0; i < count; ++i) 420 (&gpr.eax)[i] = data.GetU32(&offset); 421 SetError(GPRRegSet, Read, 0); 422 done = true; 423 424 break; 425 case FPURegSet: 426 // TODO: fill in FPU regs.... 427 // SetError (FPURegSet, Read, -1); 428 done = true; 429 430 break; 431 case EXCRegSet: 432 exc.trapno = data.GetU32(&offset); 433 exc.err = data.GetU32(&offset); 434 exc.faultvaddr = data.GetU32(&offset); 435 SetError(EXCRegSet, Read, 0); 436 done = true; 437 break; 438 case 7: 439 case 8: 440 case 9: 441 // fancy flavors that encapsulate of the above flavors... 442 break; 443 444 default: 445 done = true; 446 break; 447 } 448 } 449 } 450 } 451 452 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 453 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 454 if (reg_ctx_sp) { 455 RegisterContext *reg_ctx = reg_ctx_sp.get(); 456 457 data.PutHex32(GPRRegSet); // Flavor 458 data.PutHex32(GPRWordCount); 459 PrintRegisterValue(reg_ctx, "eax", nullptr, 4, data); 460 PrintRegisterValue(reg_ctx, "ebx", nullptr, 4, data); 461 PrintRegisterValue(reg_ctx, "ecx", nullptr, 4, data); 462 PrintRegisterValue(reg_ctx, "edx", nullptr, 4, data); 463 PrintRegisterValue(reg_ctx, "edi", nullptr, 4, data); 464 PrintRegisterValue(reg_ctx, "esi", nullptr, 4, data); 465 PrintRegisterValue(reg_ctx, "ebp", nullptr, 4, data); 466 PrintRegisterValue(reg_ctx, "esp", nullptr, 4, data); 467 PrintRegisterValue(reg_ctx, "ss", nullptr, 4, data); 468 PrintRegisterValue(reg_ctx, "eflags", nullptr, 4, data); 469 PrintRegisterValue(reg_ctx, "eip", nullptr, 4, data); 470 PrintRegisterValue(reg_ctx, "cs", nullptr, 4, data); 471 PrintRegisterValue(reg_ctx, "ds", nullptr, 4, data); 472 PrintRegisterValue(reg_ctx, "es", nullptr, 4, data); 473 PrintRegisterValue(reg_ctx, "fs", nullptr, 4, data); 474 PrintRegisterValue(reg_ctx, "gs", nullptr, 4, data); 475 476 // Write out the EXC registers 477 data.PutHex32(EXCRegSet); 478 data.PutHex32(EXCWordCount); 479 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 480 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 481 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 4, data); 482 return true; 483 } 484 return false; 485 } 486 487 protected: 488 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 489 490 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 491 492 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 493 494 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 495 return 0; 496 } 497 498 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 499 return 0; 500 } 501 502 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 503 return 0; 504 } 505 }; 506 507 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 508 public: 509 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 510 const DataExtractor &data) 511 : RegisterContextDarwin_arm(thread, 0) { 512 SetRegisterDataFrom_LC_THREAD(data); 513 } 514 515 void InvalidateAllRegisters() override { 516 // Do nothing... registers are always valid... 517 } 518 519 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 520 lldb::offset_t offset = 0; 521 SetError(GPRRegSet, Read, -1); 522 SetError(FPURegSet, Read, -1); 523 SetError(EXCRegSet, Read, -1); 524 bool done = false; 525 526 while (!done) { 527 int flavor = data.GetU32(&offset); 528 uint32_t count = data.GetU32(&offset); 529 lldb::offset_t next_thread_state = offset + (count * 4); 530 switch (flavor) { 531 case GPRAltRegSet: 532 case GPRRegSet: 533 // On ARM, the CPSR register is also included in the count but it is 534 // not included in gpr.r so loop until (count-1). 535 for (uint32_t i = 0; i < (count - 1); ++i) { 536 gpr.r[i] = data.GetU32(&offset); 537 } 538 // Save cpsr explicitly. 539 gpr.cpsr = data.GetU32(&offset); 540 541 SetError(GPRRegSet, Read, 0); 542 offset = next_thread_state; 543 break; 544 545 case FPURegSet: { 546 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 547 const int fpu_reg_buf_size = sizeof(fpu.floats); 548 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 549 fpu_reg_buf) == fpu_reg_buf_size) { 550 offset += fpu_reg_buf_size; 551 fpu.fpscr = data.GetU32(&offset); 552 SetError(FPURegSet, Read, 0); 553 } else { 554 done = true; 555 } 556 } 557 offset = next_thread_state; 558 break; 559 560 case EXCRegSet: 561 if (count == 3) { 562 exc.exception = data.GetU32(&offset); 563 exc.fsr = data.GetU32(&offset); 564 exc.far = data.GetU32(&offset); 565 SetError(EXCRegSet, Read, 0); 566 } 567 done = true; 568 offset = next_thread_state; 569 break; 570 571 // Unknown register set flavor, stop trying to parse. 572 default: 573 done = true; 574 } 575 } 576 } 577 578 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 579 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 580 if (reg_ctx_sp) { 581 RegisterContext *reg_ctx = reg_ctx_sp.get(); 582 583 data.PutHex32(GPRRegSet); // Flavor 584 data.PutHex32(GPRWordCount); 585 PrintRegisterValue(reg_ctx, "r0", nullptr, 4, data); 586 PrintRegisterValue(reg_ctx, "r1", nullptr, 4, data); 587 PrintRegisterValue(reg_ctx, "r2", nullptr, 4, data); 588 PrintRegisterValue(reg_ctx, "r3", nullptr, 4, data); 589 PrintRegisterValue(reg_ctx, "r4", nullptr, 4, data); 590 PrintRegisterValue(reg_ctx, "r5", nullptr, 4, data); 591 PrintRegisterValue(reg_ctx, "r6", nullptr, 4, data); 592 PrintRegisterValue(reg_ctx, "r7", nullptr, 4, data); 593 PrintRegisterValue(reg_ctx, "r8", nullptr, 4, data); 594 PrintRegisterValue(reg_ctx, "r9", nullptr, 4, data); 595 PrintRegisterValue(reg_ctx, "r10", nullptr, 4, data); 596 PrintRegisterValue(reg_ctx, "r11", nullptr, 4, data); 597 PrintRegisterValue(reg_ctx, "r12", nullptr, 4, data); 598 PrintRegisterValue(reg_ctx, "sp", nullptr, 4, data); 599 PrintRegisterValue(reg_ctx, "lr", nullptr, 4, data); 600 PrintRegisterValue(reg_ctx, "pc", nullptr, 4, data); 601 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 602 603 // Write out the EXC registers 604 // data.PutHex32 (EXCRegSet); 605 // data.PutHex32 (EXCWordCount); 606 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 607 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 608 // WriteRegister (reg_ctx, "far", NULL, 4, data); 609 return true; 610 } 611 return false; 612 } 613 614 protected: 615 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 616 617 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 618 619 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 620 621 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 622 623 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 624 return 0; 625 } 626 627 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 628 return 0; 629 } 630 631 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 632 return 0; 633 } 634 635 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 636 return -1; 637 } 638 }; 639 640 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 641 public: 642 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 643 const DataExtractor &data) 644 : RegisterContextDarwin_arm64(thread, 0) { 645 SetRegisterDataFrom_LC_THREAD(data); 646 } 647 648 void InvalidateAllRegisters() override { 649 // Do nothing... registers are always valid... 650 } 651 652 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 653 lldb::offset_t offset = 0; 654 SetError(GPRRegSet, Read, -1); 655 SetError(FPURegSet, Read, -1); 656 SetError(EXCRegSet, Read, -1); 657 bool done = false; 658 while (!done) { 659 int flavor = data.GetU32(&offset); 660 uint32_t count = data.GetU32(&offset); 661 lldb::offset_t next_thread_state = offset + (count * 4); 662 switch (flavor) { 663 case GPRRegSet: 664 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 665 // 32-bit register) 666 if (count >= (33 * 2) + 1) { 667 for (uint32_t i = 0; i < 29; ++i) 668 gpr.x[i] = data.GetU64(&offset); 669 gpr.fp = data.GetU64(&offset); 670 gpr.lr = data.GetU64(&offset); 671 gpr.sp = data.GetU64(&offset); 672 gpr.pc = data.GetU64(&offset); 673 gpr.cpsr = data.GetU32(&offset); 674 SetError(GPRRegSet, Read, 0); 675 } 676 offset = next_thread_state; 677 break; 678 case FPURegSet: { 679 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 680 const int fpu_reg_buf_size = sizeof(fpu); 681 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 682 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 683 fpu_reg_buf) == fpu_reg_buf_size) { 684 SetError(FPURegSet, Read, 0); 685 } else { 686 done = true; 687 } 688 } 689 offset = next_thread_state; 690 break; 691 case EXCRegSet: 692 if (count == 4) { 693 exc.far = data.GetU64(&offset); 694 exc.esr = data.GetU32(&offset); 695 exc.exception = data.GetU32(&offset); 696 SetError(EXCRegSet, Read, 0); 697 } 698 offset = next_thread_state; 699 break; 700 default: 701 done = true; 702 break; 703 } 704 } 705 } 706 707 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 708 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 709 if (reg_ctx_sp) { 710 RegisterContext *reg_ctx = reg_ctx_sp.get(); 711 712 data.PutHex32(GPRRegSet); // Flavor 713 data.PutHex32(GPRWordCount); 714 PrintRegisterValue(reg_ctx, "x0", nullptr, 8, data); 715 PrintRegisterValue(reg_ctx, "x1", nullptr, 8, data); 716 PrintRegisterValue(reg_ctx, "x2", nullptr, 8, data); 717 PrintRegisterValue(reg_ctx, "x3", nullptr, 8, data); 718 PrintRegisterValue(reg_ctx, "x4", nullptr, 8, data); 719 PrintRegisterValue(reg_ctx, "x5", nullptr, 8, data); 720 PrintRegisterValue(reg_ctx, "x6", nullptr, 8, data); 721 PrintRegisterValue(reg_ctx, "x7", nullptr, 8, data); 722 PrintRegisterValue(reg_ctx, "x8", nullptr, 8, data); 723 PrintRegisterValue(reg_ctx, "x9", nullptr, 8, data); 724 PrintRegisterValue(reg_ctx, "x10", nullptr, 8, data); 725 PrintRegisterValue(reg_ctx, "x11", nullptr, 8, data); 726 PrintRegisterValue(reg_ctx, "x12", nullptr, 8, data); 727 PrintRegisterValue(reg_ctx, "x13", nullptr, 8, data); 728 PrintRegisterValue(reg_ctx, "x14", nullptr, 8, data); 729 PrintRegisterValue(reg_ctx, "x15", nullptr, 8, data); 730 PrintRegisterValue(reg_ctx, "x16", nullptr, 8, data); 731 PrintRegisterValue(reg_ctx, "x17", nullptr, 8, data); 732 PrintRegisterValue(reg_ctx, "x18", nullptr, 8, data); 733 PrintRegisterValue(reg_ctx, "x19", nullptr, 8, data); 734 PrintRegisterValue(reg_ctx, "x20", nullptr, 8, data); 735 PrintRegisterValue(reg_ctx, "x21", nullptr, 8, data); 736 PrintRegisterValue(reg_ctx, "x22", nullptr, 8, data); 737 PrintRegisterValue(reg_ctx, "x23", nullptr, 8, data); 738 PrintRegisterValue(reg_ctx, "x24", nullptr, 8, data); 739 PrintRegisterValue(reg_ctx, "x25", nullptr, 8, data); 740 PrintRegisterValue(reg_ctx, "x26", nullptr, 8, data); 741 PrintRegisterValue(reg_ctx, "x27", nullptr, 8, data); 742 PrintRegisterValue(reg_ctx, "x28", nullptr, 8, data); 743 PrintRegisterValue(reg_ctx, "fp", nullptr, 8, data); 744 PrintRegisterValue(reg_ctx, "lr", nullptr, 8, data); 745 PrintRegisterValue(reg_ctx, "sp", nullptr, 8, data); 746 PrintRegisterValue(reg_ctx, "pc", nullptr, 8, data); 747 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 748 749 // Write out the EXC registers 750 // data.PutHex32 (EXCRegSet); 751 // data.PutHex32 (EXCWordCount); 752 // WriteRegister (reg_ctx, "far", NULL, 8, data); 753 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 754 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 755 return true; 756 } 757 return false; 758 } 759 760 protected: 761 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 762 763 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 764 765 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 766 767 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 768 769 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 770 return 0; 771 } 772 773 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 774 return 0; 775 } 776 777 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 778 return 0; 779 } 780 781 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 782 return -1; 783 } 784 }; 785 786 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 787 switch (magic) { 788 case MH_MAGIC: 789 case MH_CIGAM: 790 return sizeof(struct llvm::MachO::mach_header); 791 792 case MH_MAGIC_64: 793 case MH_CIGAM_64: 794 return sizeof(struct llvm::MachO::mach_header_64); 795 break; 796 797 default: 798 break; 799 } 800 return 0; 801 } 802 803 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 804 805 char ObjectFileMachO::ID; 806 807 void ObjectFileMachO::Initialize() { 808 PluginManager::RegisterPlugin( 809 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 810 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 811 } 812 813 void ObjectFileMachO::Terminate() { 814 PluginManager::UnregisterPlugin(CreateInstance); 815 } 816 817 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 818 static ConstString g_name("mach-o"); 819 return g_name; 820 } 821 822 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 823 return "Mach-o object file reader (32 and 64 bit)"; 824 } 825 826 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 827 DataBufferSP &data_sp, 828 lldb::offset_t data_offset, 829 const FileSpec *file, 830 lldb::offset_t file_offset, 831 lldb::offset_t length) { 832 if (!data_sp) { 833 data_sp = MapFileData(*file, length, file_offset); 834 if (!data_sp) 835 return nullptr; 836 data_offset = 0; 837 } 838 839 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 840 return nullptr; 841 842 // Update the data to contain the entire file if it doesn't already 843 if (data_sp->GetByteSize() < length) { 844 data_sp = MapFileData(*file, length, file_offset); 845 if (!data_sp) 846 return nullptr; 847 data_offset = 0; 848 } 849 auto objfile_up = std::make_unique<ObjectFileMachO>( 850 module_sp, data_sp, data_offset, file, file_offset, length); 851 if (!objfile_up || !objfile_up->ParseHeader()) 852 return nullptr; 853 854 return objfile_up.release(); 855 } 856 857 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 858 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 859 const ProcessSP &process_sp, lldb::addr_t header_addr) { 860 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 861 std::unique_ptr<ObjectFile> objfile_up( 862 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 863 if (objfile_up.get() && objfile_up->ParseHeader()) 864 return objfile_up.release(); 865 } 866 return nullptr; 867 } 868 869 size_t ObjectFileMachO::GetModuleSpecifications( 870 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 871 lldb::offset_t data_offset, lldb::offset_t file_offset, 872 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 873 const size_t initial_count = specs.GetSize(); 874 875 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 876 DataExtractor data; 877 data.SetData(data_sp); 878 llvm::MachO::mach_header header; 879 if (ParseHeader(data, &data_offset, header)) { 880 size_t header_and_load_cmds = 881 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 882 if (header_and_load_cmds >= data_sp->GetByteSize()) { 883 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 884 data.SetData(data_sp); 885 data_offset = MachHeaderSizeFromMagic(header.magic); 886 } 887 if (data_sp) { 888 ModuleSpec base_spec; 889 base_spec.GetFileSpec() = file; 890 base_spec.SetObjectOffset(file_offset); 891 base_spec.SetObjectSize(length); 892 GetAllArchSpecs(header, data, data_offset, base_spec, specs); 893 } 894 } 895 } 896 return specs.GetSize() - initial_count; 897 } 898 899 ConstString ObjectFileMachO::GetSegmentNameTEXT() { 900 static ConstString g_segment_name_TEXT("__TEXT"); 901 return g_segment_name_TEXT; 902 } 903 904 ConstString ObjectFileMachO::GetSegmentNameDATA() { 905 static ConstString g_segment_name_DATA("__DATA"); 906 return g_segment_name_DATA; 907 } 908 909 ConstString ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 910 static ConstString g_segment_name("__DATA_DIRTY"); 911 return g_segment_name; 912 } 913 914 ConstString ObjectFileMachO::GetSegmentNameDATA_CONST() { 915 static ConstString g_segment_name("__DATA_CONST"); 916 return g_segment_name; 917 } 918 919 ConstString ObjectFileMachO::GetSegmentNameOBJC() { 920 static ConstString g_segment_name_OBJC("__OBJC"); 921 return g_segment_name_OBJC; 922 } 923 924 ConstString ObjectFileMachO::GetSegmentNameLINKEDIT() { 925 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 926 return g_section_name_LINKEDIT; 927 } 928 929 ConstString ObjectFileMachO::GetSegmentNameDWARF() { 930 static ConstString g_section_name("__DWARF"); 931 return g_section_name; 932 } 933 934 ConstString ObjectFileMachO::GetSectionNameEHFrame() { 935 static ConstString g_section_name_eh_frame("__eh_frame"); 936 return g_section_name_eh_frame; 937 } 938 939 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 940 lldb::addr_t data_offset, 941 lldb::addr_t data_length) { 942 DataExtractor data; 943 data.SetData(data_sp, data_offset, data_length); 944 lldb::offset_t offset = 0; 945 uint32_t magic = data.GetU32(&offset); 946 return MachHeaderSizeFromMagic(magic) != 0; 947 } 948 949 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 950 DataBufferSP &data_sp, 951 lldb::offset_t data_offset, 952 const FileSpec *file, 953 lldb::offset_t file_offset, 954 lldb::offset_t length) 955 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 956 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 957 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 958 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 959 ::memset(&m_header, 0, sizeof(m_header)); 960 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 961 } 962 963 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 964 lldb::DataBufferSP &header_data_sp, 965 const lldb::ProcessSP &process_sp, 966 lldb::addr_t header_addr) 967 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 968 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 969 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 970 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 971 ::memset(&m_header, 0, sizeof(m_header)); 972 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 973 } 974 975 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 976 lldb::offset_t *data_offset_ptr, 977 llvm::MachO::mach_header &header) { 978 data.SetByteOrder(endian::InlHostByteOrder()); 979 // Leave magic in the original byte order 980 header.magic = data.GetU32(data_offset_ptr); 981 bool can_parse = false; 982 bool is_64_bit = false; 983 switch (header.magic) { 984 case MH_MAGIC: 985 data.SetByteOrder(endian::InlHostByteOrder()); 986 data.SetAddressByteSize(4); 987 can_parse = true; 988 break; 989 990 case MH_MAGIC_64: 991 data.SetByteOrder(endian::InlHostByteOrder()); 992 data.SetAddressByteSize(8); 993 can_parse = true; 994 is_64_bit = true; 995 break; 996 997 case MH_CIGAM: 998 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 999 ? eByteOrderLittle 1000 : eByteOrderBig); 1001 data.SetAddressByteSize(4); 1002 can_parse = true; 1003 break; 1004 1005 case MH_CIGAM_64: 1006 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1007 ? eByteOrderLittle 1008 : eByteOrderBig); 1009 data.SetAddressByteSize(8); 1010 is_64_bit = true; 1011 can_parse = true; 1012 break; 1013 1014 default: 1015 break; 1016 } 1017 1018 if (can_parse) { 1019 data.GetU32(data_offset_ptr, &header.cputype, 6); 1020 if (is_64_bit) 1021 *data_offset_ptr += 4; 1022 return true; 1023 } else { 1024 memset(&header, 0, sizeof(header)); 1025 } 1026 return false; 1027 } 1028 1029 bool ObjectFileMachO::ParseHeader() { 1030 ModuleSP module_sp(GetModule()); 1031 if (!module_sp) 1032 return false; 1033 1034 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1035 bool can_parse = false; 1036 lldb::offset_t offset = 0; 1037 m_data.SetByteOrder(endian::InlHostByteOrder()); 1038 // Leave magic in the original byte order 1039 m_header.magic = m_data.GetU32(&offset); 1040 switch (m_header.magic) { 1041 case MH_MAGIC: 1042 m_data.SetByteOrder(endian::InlHostByteOrder()); 1043 m_data.SetAddressByteSize(4); 1044 can_parse = true; 1045 break; 1046 1047 case MH_MAGIC_64: 1048 m_data.SetByteOrder(endian::InlHostByteOrder()); 1049 m_data.SetAddressByteSize(8); 1050 can_parse = true; 1051 break; 1052 1053 case MH_CIGAM: 1054 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1055 ? eByteOrderLittle 1056 : eByteOrderBig); 1057 m_data.SetAddressByteSize(4); 1058 can_parse = true; 1059 break; 1060 1061 case MH_CIGAM_64: 1062 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1063 ? eByteOrderLittle 1064 : eByteOrderBig); 1065 m_data.SetAddressByteSize(8); 1066 can_parse = true; 1067 break; 1068 1069 default: 1070 break; 1071 } 1072 1073 if (can_parse) { 1074 m_data.GetU32(&offset, &m_header.cputype, 6); 1075 1076 ModuleSpecList all_specs; 1077 ModuleSpec base_spec; 1078 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 1079 base_spec, all_specs); 1080 1081 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 1082 ArchSpec mach_arch = 1083 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 1084 1085 // Check if the module has a required architecture 1086 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1087 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1088 continue; 1089 1090 if (SetModulesArchitecture(mach_arch)) { 1091 const size_t header_and_lc_size = 1092 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1093 if (m_data.GetByteSize() < header_and_lc_size) { 1094 DataBufferSP data_sp; 1095 ProcessSP process_sp(m_process_wp.lock()); 1096 if (process_sp) { 1097 data_sp = ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1098 } else { 1099 // Read in all only the load command data from the file on disk 1100 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset); 1101 if (data_sp->GetByteSize() != header_and_lc_size) 1102 continue; 1103 } 1104 if (data_sp) 1105 m_data.SetData(data_sp); 1106 } 1107 } 1108 return true; 1109 } 1110 // None found. 1111 return false; 1112 } else { 1113 memset(&m_header, 0, sizeof(struct llvm::MachO::mach_header)); 1114 } 1115 return false; 1116 } 1117 1118 ByteOrder ObjectFileMachO::GetByteOrder() const { 1119 return m_data.GetByteOrder(); 1120 } 1121 1122 bool ObjectFileMachO::IsExecutable() const { 1123 return m_header.filetype == MH_EXECUTE; 1124 } 1125 1126 bool ObjectFileMachO::IsDynamicLoader() const { 1127 return m_header.filetype == MH_DYLINKER; 1128 } 1129 1130 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1131 return m_data.GetAddressByteSize(); 1132 } 1133 1134 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1135 Symtab *symtab = GetSymtab(); 1136 if (!symtab) 1137 return AddressClass::eUnknown; 1138 1139 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1140 if (symbol) { 1141 if (symbol->ValueIsAddress()) { 1142 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1143 if (section_sp) { 1144 const lldb::SectionType section_type = section_sp->GetType(); 1145 switch (section_type) { 1146 case eSectionTypeInvalid: 1147 return AddressClass::eUnknown; 1148 1149 case eSectionTypeCode: 1150 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1151 // For ARM we have a bit in the n_desc field of the symbol that 1152 // tells us ARM/Thumb which is bit 0x0008. 1153 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1154 return AddressClass::eCodeAlternateISA; 1155 } 1156 return AddressClass::eCode; 1157 1158 case eSectionTypeContainer: 1159 return AddressClass::eUnknown; 1160 1161 case eSectionTypeData: 1162 case eSectionTypeDataCString: 1163 case eSectionTypeDataCStringPointers: 1164 case eSectionTypeDataSymbolAddress: 1165 case eSectionTypeData4: 1166 case eSectionTypeData8: 1167 case eSectionTypeData16: 1168 case eSectionTypeDataPointers: 1169 case eSectionTypeZeroFill: 1170 case eSectionTypeDataObjCMessageRefs: 1171 case eSectionTypeDataObjCCFStrings: 1172 case eSectionTypeGoSymtab: 1173 return AddressClass::eData; 1174 1175 case eSectionTypeDebug: 1176 case eSectionTypeDWARFDebugAbbrev: 1177 case eSectionTypeDWARFDebugAbbrevDwo: 1178 case eSectionTypeDWARFDebugAddr: 1179 case eSectionTypeDWARFDebugAranges: 1180 case eSectionTypeDWARFDebugCuIndex: 1181 case eSectionTypeDWARFDebugFrame: 1182 case eSectionTypeDWARFDebugInfo: 1183 case eSectionTypeDWARFDebugInfoDwo: 1184 case eSectionTypeDWARFDebugLine: 1185 case eSectionTypeDWARFDebugLineStr: 1186 case eSectionTypeDWARFDebugLoc: 1187 case eSectionTypeDWARFDebugLocDwo: 1188 case eSectionTypeDWARFDebugLocLists: 1189 case eSectionTypeDWARFDebugLocListsDwo: 1190 case eSectionTypeDWARFDebugMacInfo: 1191 case eSectionTypeDWARFDebugMacro: 1192 case eSectionTypeDWARFDebugNames: 1193 case eSectionTypeDWARFDebugPubNames: 1194 case eSectionTypeDWARFDebugPubTypes: 1195 case eSectionTypeDWARFDebugRanges: 1196 case eSectionTypeDWARFDebugRngLists: 1197 case eSectionTypeDWARFDebugRngListsDwo: 1198 case eSectionTypeDWARFDebugStr: 1199 case eSectionTypeDWARFDebugStrDwo: 1200 case eSectionTypeDWARFDebugStrOffsets: 1201 case eSectionTypeDWARFDebugStrOffsetsDwo: 1202 case eSectionTypeDWARFDebugTuIndex: 1203 case eSectionTypeDWARFDebugTypes: 1204 case eSectionTypeDWARFDebugTypesDwo: 1205 case eSectionTypeDWARFAppleNames: 1206 case eSectionTypeDWARFAppleTypes: 1207 case eSectionTypeDWARFAppleNamespaces: 1208 case eSectionTypeDWARFAppleObjC: 1209 case eSectionTypeDWARFGNUDebugAltLink: 1210 return AddressClass::eDebug; 1211 1212 case eSectionTypeEHFrame: 1213 case eSectionTypeARMexidx: 1214 case eSectionTypeARMextab: 1215 case eSectionTypeCompactUnwind: 1216 return AddressClass::eRuntime; 1217 1218 case eSectionTypeAbsoluteAddress: 1219 case eSectionTypeELFSymbolTable: 1220 case eSectionTypeELFDynamicSymbols: 1221 case eSectionTypeELFRelocationEntries: 1222 case eSectionTypeELFDynamicLinkInfo: 1223 case eSectionTypeOther: 1224 return AddressClass::eUnknown; 1225 } 1226 } 1227 } 1228 1229 const SymbolType symbol_type = symbol->GetType(); 1230 switch (symbol_type) { 1231 case eSymbolTypeAny: 1232 return AddressClass::eUnknown; 1233 case eSymbolTypeAbsolute: 1234 return AddressClass::eUnknown; 1235 1236 case eSymbolTypeCode: 1237 case eSymbolTypeTrampoline: 1238 case eSymbolTypeResolver: 1239 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1240 // For ARM we have a bit in the n_desc field of the symbol that tells 1241 // us ARM/Thumb which is bit 0x0008. 1242 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1243 return AddressClass::eCodeAlternateISA; 1244 } 1245 return AddressClass::eCode; 1246 1247 case eSymbolTypeData: 1248 return AddressClass::eData; 1249 case eSymbolTypeRuntime: 1250 return AddressClass::eRuntime; 1251 case eSymbolTypeException: 1252 return AddressClass::eRuntime; 1253 case eSymbolTypeSourceFile: 1254 return AddressClass::eDebug; 1255 case eSymbolTypeHeaderFile: 1256 return AddressClass::eDebug; 1257 case eSymbolTypeObjectFile: 1258 return AddressClass::eDebug; 1259 case eSymbolTypeCommonBlock: 1260 return AddressClass::eDebug; 1261 case eSymbolTypeBlock: 1262 return AddressClass::eDebug; 1263 case eSymbolTypeLocal: 1264 return AddressClass::eData; 1265 case eSymbolTypeParam: 1266 return AddressClass::eData; 1267 case eSymbolTypeVariable: 1268 return AddressClass::eData; 1269 case eSymbolTypeVariableType: 1270 return AddressClass::eDebug; 1271 case eSymbolTypeLineEntry: 1272 return AddressClass::eDebug; 1273 case eSymbolTypeLineHeader: 1274 return AddressClass::eDebug; 1275 case eSymbolTypeScopeBegin: 1276 return AddressClass::eDebug; 1277 case eSymbolTypeScopeEnd: 1278 return AddressClass::eDebug; 1279 case eSymbolTypeAdditional: 1280 return AddressClass::eUnknown; 1281 case eSymbolTypeCompiler: 1282 return AddressClass::eDebug; 1283 case eSymbolTypeInstrumentation: 1284 return AddressClass::eDebug; 1285 case eSymbolTypeUndefined: 1286 return AddressClass::eUnknown; 1287 case eSymbolTypeObjCClass: 1288 return AddressClass::eRuntime; 1289 case eSymbolTypeObjCMetaClass: 1290 return AddressClass::eRuntime; 1291 case eSymbolTypeObjCIVar: 1292 return AddressClass::eRuntime; 1293 case eSymbolTypeReExported: 1294 return AddressClass::eRuntime; 1295 } 1296 } 1297 return AddressClass::eUnknown; 1298 } 1299 1300 Symtab *ObjectFileMachO::GetSymtab() { 1301 ModuleSP module_sp(GetModule()); 1302 if (module_sp) { 1303 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1304 if (m_symtab_up == nullptr) { 1305 m_symtab_up = std::make_unique<Symtab>(this); 1306 std::lock_guard<std::recursive_mutex> symtab_guard( 1307 m_symtab_up->GetMutex()); 1308 ParseSymtab(); 1309 m_symtab_up->Finalize(); 1310 } 1311 } 1312 return m_symtab_up.get(); 1313 } 1314 1315 bool ObjectFileMachO::IsStripped() { 1316 if (m_dysymtab.cmd == 0) { 1317 ModuleSP module_sp(GetModule()); 1318 if (module_sp) { 1319 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1320 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1321 const lldb::offset_t load_cmd_offset = offset; 1322 1323 llvm::MachO::load_command lc; 1324 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 1325 break; 1326 if (lc.cmd == LC_DYSYMTAB) { 1327 m_dysymtab.cmd = lc.cmd; 1328 m_dysymtab.cmdsize = lc.cmdsize; 1329 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1330 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1331 nullptr) { 1332 // Clear m_dysymtab if we were unable to read all items from the 1333 // load command 1334 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1335 } 1336 } 1337 offset = load_cmd_offset + lc.cmdsize; 1338 } 1339 } 1340 } 1341 if (m_dysymtab.cmd) 1342 return m_dysymtab.nlocalsym <= 1; 1343 return false; 1344 } 1345 1346 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1347 EncryptedFileRanges result; 1348 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1349 1350 llvm::MachO::encryption_info_command encryption_cmd; 1351 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1352 const lldb::offset_t load_cmd_offset = offset; 1353 if (m_data.GetU32(&offset, &encryption_cmd, 2) == nullptr) 1354 break; 1355 1356 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the 1357 // 3 fields we care about, so treat them the same. 1358 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1359 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1360 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1361 if (encryption_cmd.cryptid != 0) { 1362 EncryptedFileRanges::Entry entry; 1363 entry.SetRangeBase(encryption_cmd.cryptoff); 1364 entry.SetByteSize(encryption_cmd.cryptsize); 1365 result.Append(entry); 1366 } 1367 } 1368 } 1369 offset = load_cmd_offset + encryption_cmd.cmdsize; 1370 } 1371 1372 return result; 1373 } 1374 1375 void ObjectFileMachO::SanitizeSegmentCommand( 1376 llvm::MachO::segment_command_64 &seg_cmd, uint32_t cmd_idx) { 1377 if (m_length == 0 || seg_cmd.filesize == 0) 1378 return; 1379 1380 if ((m_header.flags & MH_DYLIB_IN_CACHE) && !IsInMemory()) { 1381 // In shared cache images, the load commands are relative to the 1382 // shared cache file, and not the specific image we are 1383 // examining. Let's fix this up so that it looks like a normal 1384 // image. 1385 if (strncmp(seg_cmd.segname, "__TEXT", sizeof(seg_cmd.segname)) == 0) 1386 m_text_address = seg_cmd.vmaddr; 1387 if (strncmp(seg_cmd.segname, "__LINKEDIT", sizeof(seg_cmd.segname)) == 0) 1388 m_linkedit_original_offset = seg_cmd.fileoff; 1389 1390 seg_cmd.fileoff = seg_cmd.vmaddr - m_text_address; 1391 } 1392 1393 if (seg_cmd.fileoff > m_length) { 1394 // We have a load command that says it extends past the end of the file. 1395 // This is likely a corrupt file. We don't have any way to return an error 1396 // condition here (this method was likely invoked from something like 1397 // ObjectFile::GetSectionList()), so we just null out the section contents, 1398 // and dump a message to stdout. The most common case here is core file 1399 // debugging with a truncated file. 1400 const char *lc_segment_name = 1401 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1402 GetModule()->ReportWarning( 1403 "load command %u %s has a fileoff (0x%" PRIx64 1404 ") that extends beyond the end of the file (0x%" PRIx64 1405 "), ignoring this section", 1406 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1407 1408 seg_cmd.fileoff = 0; 1409 seg_cmd.filesize = 0; 1410 } 1411 1412 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1413 // We have a load command that says it extends past the end of the file. 1414 // This is likely a corrupt file. We don't have any way to return an error 1415 // condition here (this method was likely invoked from something like 1416 // ObjectFile::GetSectionList()), so we just null out the section contents, 1417 // and dump a message to stdout. The most common case here is core file 1418 // debugging with a truncated file. 1419 const char *lc_segment_name = 1420 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1421 GetModule()->ReportWarning( 1422 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1423 ") that extends beyond the end of the file (0x%" PRIx64 1424 "), the segment will be truncated to match", 1425 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1426 1427 // Truncate the length 1428 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1429 } 1430 } 1431 1432 static uint32_t 1433 GetSegmentPermissions(const llvm::MachO::segment_command_64 &seg_cmd) { 1434 uint32_t result = 0; 1435 if (seg_cmd.initprot & VM_PROT_READ) 1436 result |= ePermissionsReadable; 1437 if (seg_cmd.initprot & VM_PROT_WRITE) 1438 result |= ePermissionsWritable; 1439 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1440 result |= ePermissionsExecutable; 1441 return result; 1442 } 1443 1444 static lldb::SectionType GetSectionType(uint32_t flags, 1445 ConstString section_name) { 1446 1447 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1448 return eSectionTypeCode; 1449 1450 uint32_t mach_sect_type = flags & SECTION_TYPE; 1451 static ConstString g_sect_name_objc_data("__objc_data"); 1452 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1453 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1454 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1455 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1456 static ConstString g_sect_name_objc_const("__objc_const"); 1457 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1458 static ConstString g_sect_name_cfstring("__cfstring"); 1459 1460 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1461 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1462 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1463 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1464 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1465 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1466 static ConstString g_sect_name_dwarf_debug_loclists("__debug_loclists"); 1467 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1468 static ConstString g_sect_name_dwarf_debug_names("__debug_names"); 1469 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1470 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1471 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1472 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1473 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1474 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1475 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1476 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1477 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1478 static ConstString g_sect_name_eh_frame("__eh_frame"); 1479 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1480 static ConstString g_sect_name_text("__text"); 1481 static ConstString g_sect_name_data("__data"); 1482 static ConstString g_sect_name_go_symtab("__gosymtab"); 1483 1484 if (section_name == g_sect_name_dwarf_debug_abbrev) 1485 return eSectionTypeDWARFDebugAbbrev; 1486 if (section_name == g_sect_name_dwarf_debug_aranges) 1487 return eSectionTypeDWARFDebugAranges; 1488 if (section_name == g_sect_name_dwarf_debug_frame) 1489 return eSectionTypeDWARFDebugFrame; 1490 if (section_name == g_sect_name_dwarf_debug_info) 1491 return eSectionTypeDWARFDebugInfo; 1492 if (section_name == g_sect_name_dwarf_debug_line) 1493 return eSectionTypeDWARFDebugLine; 1494 if (section_name == g_sect_name_dwarf_debug_loc) 1495 return eSectionTypeDWARFDebugLoc; 1496 if (section_name == g_sect_name_dwarf_debug_loclists) 1497 return eSectionTypeDWARFDebugLocLists; 1498 if (section_name == g_sect_name_dwarf_debug_macinfo) 1499 return eSectionTypeDWARFDebugMacInfo; 1500 if (section_name == g_sect_name_dwarf_debug_names) 1501 return eSectionTypeDWARFDebugNames; 1502 if (section_name == g_sect_name_dwarf_debug_pubnames) 1503 return eSectionTypeDWARFDebugPubNames; 1504 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1505 return eSectionTypeDWARFDebugPubTypes; 1506 if (section_name == g_sect_name_dwarf_debug_ranges) 1507 return eSectionTypeDWARFDebugRanges; 1508 if (section_name == g_sect_name_dwarf_debug_str) 1509 return eSectionTypeDWARFDebugStr; 1510 if (section_name == g_sect_name_dwarf_debug_types) 1511 return eSectionTypeDWARFDebugTypes; 1512 if (section_name == g_sect_name_dwarf_apple_names) 1513 return eSectionTypeDWARFAppleNames; 1514 if (section_name == g_sect_name_dwarf_apple_types) 1515 return eSectionTypeDWARFAppleTypes; 1516 if (section_name == g_sect_name_dwarf_apple_namespaces) 1517 return eSectionTypeDWARFAppleNamespaces; 1518 if (section_name == g_sect_name_dwarf_apple_objc) 1519 return eSectionTypeDWARFAppleObjC; 1520 if (section_name == g_sect_name_objc_selrefs) 1521 return eSectionTypeDataCStringPointers; 1522 if (section_name == g_sect_name_objc_msgrefs) 1523 return eSectionTypeDataObjCMessageRefs; 1524 if (section_name == g_sect_name_eh_frame) 1525 return eSectionTypeEHFrame; 1526 if (section_name == g_sect_name_compact_unwind) 1527 return eSectionTypeCompactUnwind; 1528 if (section_name == g_sect_name_cfstring) 1529 return eSectionTypeDataObjCCFStrings; 1530 if (section_name == g_sect_name_go_symtab) 1531 return eSectionTypeGoSymtab; 1532 if (section_name == g_sect_name_objc_data || 1533 section_name == g_sect_name_objc_classrefs || 1534 section_name == g_sect_name_objc_superrefs || 1535 section_name == g_sect_name_objc_const || 1536 section_name == g_sect_name_objc_classlist) { 1537 return eSectionTypeDataPointers; 1538 } 1539 1540 switch (mach_sect_type) { 1541 // TODO: categorize sections by other flags for regular sections 1542 case S_REGULAR: 1543 if (section_name == g_sect_name_text) 1544 return eSectionTypeCode; 1545 if (section_name == g_sect_name_data) 1546 return eSectionTypeData; 1547 return eSectionTypeOther; 1548 case S_ZEROFILL: 1549 return eSectionTypeZeroFill; 1550 case S_CSTRING_LITERALS: // section with only literal C strings 1551 return eSectionTypeDataCString; 1552 case S_4BYTE_LITERALS: // section with only 4 byte literals 1553 return eSectionTypeData4; 1554 case S_8BYTE_LITERALS: // section with only 8 byte literals 1555 return eSectionTypeData8; 1556 case S_LITERAL_POINTERS: // section with only pointers to literals 1557 return eSectionTypeDataPointers; 1558 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1559 return eSectionTypeDataPointers; 1560 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1561 return eSectionTypeDataPointers; 1562 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1563 // the reserved2 field 1564 return eSectionTypeCode; 1565 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1566 // initialization 1567 return eSectionTypeDataPointers; 1568 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1569 // termination 1570 return eSectionTypeDataPointers; 1571 case S_COALESCED: 1572 return eSectionTypeOther; 1573 case S_GB_ZEROFILL: 1574 return eSectionTypeZeroFill; 1575 case S_INTERPOSING: // section with only pairs of function pointers for 1576 // interposing 1577 return eSectionTypeCode; 1578 case S_16BYTE_LITERALS: // section with only 16 byte literals 1579 return eSectionTypeData16; 1580 case S_DTRACE_DOF: 1581 return eSectionTypeDebug; 1582 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1583 return eSectionTypeDataPointers; 1584 default: 1585 return eSectionTypeOther; 1586 } 1587 } 1588 1589 struct ObjectFileMachO::SegmentParsingContext { 1590 const EncryptedFileRanges EncryptedRanges; 1591 lldb_private::SectionList &UnifiedList; 1592 uint32_t NextSegmentIdx = 0; 1593 uint32_t NextSectionIdx = 0; 1594 bool FileAddressesChanged = false; 1595 1596 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1597 lldb_private::SectionList &UnifiedList) 1598 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1599 }; 1600 1601 void ObjectFileMachO::ProcessSegmentCommand( 1602 const llvm::MachO::load_command &load_cmd_, lldb::offset_t offset, 1603 uint32_t cmd_idx, SegmentParsingContext &context) { 1604 llvm::MachO::segment_command_64 load_cmd; 1605 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1606 1607 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1608 return; 1609 1610 ModuleSP module_sp = GetModule(); 1611 const bool is_core = GetType() == eTypeCoreFile; 1612 const bool is_dsym = (m_header.filetype == MH_DSYM); 1613 bool add_section = true; 1614 bool add_to_unified = true; 1615 ConstString const_segname( 1616 load_cmd.segname, strnlen(load_cmd.segname, sizeof(load_cmd.segname))); 1617 1618 SectionSP unified_section_sp( 1619 context.UnifiedList.FindSectionByName(const_segname)); 1620 if (is_dsym && unified_section_sp) { 1621 if (const_segname == GetSegmentNameLINKEDIT()) { 1622 // We need to keep the __LINKEDIT segment private to this object file 1623 // only 1624 add_to_unified = false; 1625 } else { 1626 // This is the dSYM file and this section has already been created by the 1627 // object file, no need to create it. 1628 add_section = false; 1629 } 1630 } 1631 load_cmd.vmaddr = m_data.GetAddress(&offset); 1632 load_cmd.vmsize = m_data.GetAddress(&offset); 1633 load_cmd.fileoff = m_data.GetAddress(&offset); 1634 load_cmd.filesize = m_data.GetAddress(&offset); 1635 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1636 return; 1637 1638 SanitizeSegmentCommand(load_cmd, cmd_idx); 1639 1640 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1641 const bool segment_is_encrypted = 1642 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1643 1644 // Keep a list of mach segments around in case we need to get at data that 1645 // isn't stored in the abstracted Sections. 1646 m_mach_segments.push_back(load_cmd); 1647 1648 // Use a segment ID of the segment index shifted left by 8 so they never 1649 // conflict with any of the sections. 1650 SectionSP segment_sp; 1651 if (add_section && (const_segname || is_core)) { 1652 segment_sp = std::make_shared<Section>( 1653 module_sp, // Module to which this section belongs 1654 this, // Object file to which this sections belongs 1655 ++context.NextSegmentIdx 1656 << 8, // Section ID is the 1 based segment index 1657 // shifted right by 8 bits as not to collide with any of the 256 1658 // section IDs that are possible 1659 const_segname, // Name of this section 1660 eSectionTypeContainer, // This section is a container of other 1661 // sections. 1662 load_cmd.vmaddr, // File VM address == addresses as they are 1663 // found in the object file 1664 load_cmd.vmsize, // VM size in bytes of this section 1665 load_cmd.fileoff, // Offset to the data for this section in 1666 // the file 1667 load_cmd.filesize, // Size in bytes of this section as found 1668 // in the file 1669 0, // Segments have no alignment information 1670 load_cmd.flags); // Flags for this section 1671 1672 segment_sp->SetIsEncrypted(segment_is_encrypted); 1673 m_sections_up->AddSection(segment_sp); 1674 segment_sp->SetPermissions(segment_permissions); 1675 if (add_to_unified) 1676 context.UnifiedList.AddSection(segment_sp); 1677 } else if (unified_section_sp) { 1678 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1679 // Check to see if the module was read from memory? 1680 if (module_sp->GetObjectFile()->IsInMemory()) { 1681 // We have a module that is in memory and needs to have its file 1682 // address adjusted. We need to do this because when we load a file 1683 // from memory, its addresses will be slid already, yet the addresses 1684 // in the new symbol file will still be unslid. Since everything is 1685 // stored as section offset, this shouldn't cause any problems. 1686 1687 // Make sure we've parsed the symbol table from the ObjectFile before 1688 // we go around changing its Sections. 1689 module_sp->GetObjectFile()->GetSymtab(); 1690 // eh_frame would present the same problems but we parse that on a per- 1691 // function basis as-needed so it's more difficult to remove its use of 1692 // the Sections. Realistically, the environments where this code path 1693 // will be taken will not have eh_frame sections. 1694 1695 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1696 1697 // Notify the module that the section addresses have been changed once 1698 // we're done so any file-address caches can be updated. 1699 context.FileAddressesChanged = true; 1700 } 1701 } 1702 m_sections_up->AddSection(unified_section_sp); 1703 } 1704 1705 llvm::MachO::section_64 sect64; 1706 ::memset(§64, 0, sizeof(sect64)); 1707 // Push a section into our mach sections for the section at index zero 1708 // (NO_SECT) if we don't have any mach sections yet... 1709 if (m_mach_sections.empty()) 1710 m_mach_sections.push_back(sect64); 1711 uint32_t segment_sect_idx; 1712 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1713 1714 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1715 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1716 ++segment_sect_idx) { 1717 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1718 sizeof(sect64.sectname)) == nullptr) 1719 break; 1720 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1721 sizeof(sect64.segname)) == nullptr) 1722 break; 1723 sect64.addr = m_data.GetAddress(&offset); 1724 sect64.size = m_data.GetAddress(&offset); 1725 1726 if (m_data.GetU32(&offset, §64.offset, num_u32s) == nullptr) 1727 break; 1728 1729 if ((m_header.flags & MH_DYLIB_IN_CACHE) && !IsInMemory()) { 1730 sect64.offset = sect64.addr - m_text_address; 1731 } 1732 1733 // Keep a list of mach sections around in case we need to get at data that 1734 // isn't stored in the abstracted Sections. 1735 m_mach_sections.push_back(sect64); 1736 1737 if (add_section) { 1738 ConstString section_name( 1739 sect64.sectname, strnlen(sect64.sectname, sizeof(sect64.sectname))); 1740 if (!const_segname) { 1741 // We have a segment with no name so we need to conjure up segments 1742 // that correspond to the section's segname if there isn't already such 1743 // a section. If there is such a section, we resize the section so that 1744 // it spans all sections. We also mark these sections as fake so 1745 // address matches don't hit if they land in the gaps between the child 1746 // sections. 1747 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1748 sizeof(sect64.segname)); 1749 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1750 if (segment_sp.get()) { 1751 Section *segment = segment_sp.get(); 1752 // Grow the section size as needed. 1753 const lldb::addr_t sect64_min_addr = sect64.addr; 1754 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1755 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1756 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1757 const lldb::addr_t curr_seg_max_addr = 1758 curr_seg_min_addr + curr_seg_byte_size; 1759 if (sect64_min_addr >= curr_seg_min_addr) { 1760 const lldb::addr_t new_seg_byte_size = 1761 sect64_max_addr - curr_seg_min_addr; 1762 // Only grow the section size if needed 1763 if (new_seg_byte_size > curr_seg_byte_size) 1764 segment->SetByteSize(new_seg_byte_size); 1765 } else { 1766 // We need to change the base address of the segment and adjust the 1767 // child section offsets for all existing children. 1768 const lldb::addr_t slide_amount = 1769 sect64_min_addr - curr_seg_min_addr; 1770 segment->Slide(slide_amount, false); 1771 segment->GetChildren().Slide(-slide_amount, false); 1772 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1773 } 1774 1775 // Grow the section size as needed. 1776 if (sect64.offset) { 1777 const lldb::addr_t segment_min_file_offset = 1778 segment->GetFileOffset(); 1779 const lldb::addr_t segment_max_file_offset = 1780 segment_min_file_offset + segment->GetFileSize(); 1781 1782 const lldb::addr_t section_min_file_offset = sect64.offset; 1783 const lldb::addr_t section_max_file_offset = 1784 section_min_file_offset + sect64.size; 1785 const lldb::addr_t new_file_offset = 1786 std::min(section_min_file_offset, segment_min_file_offset); 1787 const lldb::addr_t new_file_size = 1788 std::max(section_max_file_offset, segment_max_file_offset) - 1789 new_file_offset; 1790 segment->SetFileOffset(new_file_offset); 1791 segment->SetFileSize(new_file_size); 1792 } 1793 } else { 1794 // Create a fake section for the section's named segment 1795 segment_sp = std::make_shared<Section>( 1796 segment_sp, // Parent section 1797 module_sp, // Module to which this section belongs 1798 this, // Object file to which this section belongs 1799 ++context.NextSegmentIdx 1800 << 8, // Section ID is the 1 based segment index 1801 // shifted right by 8 bits as not to 1802 // collide with any of the 256 section IDs 1803 // that are possible 1804 const_segname, // Name of this section 1805 eSectionTypeContainer, // This section is a container of 1806 // other sections. 1807 sect64.addr, // File VM address == addresses as they are 1808 // found in the object file 1809 sect64.size, // VM size in bytes of this section 1810 sect64.offset, // Offset to the data for this section in 1811 // the file 1812 sect64.offset ? sect64.size : 0, // Size in bytes of 1813 // this section as 1814 // found in the file 1815 sect64.align, 1816 load_cmd.flags); // Flags for this section 1817 segment_sp->SetIsFake(true); 1818 segment_sp->SetPermissions(segment_permissions); 1819 m_sections_up->AddSection(segment_sp); 1820 if (add_to_unified) 1821 context.UnifiedList.AddSection(segment_sp); 1822 segment_sp->SetIsEncrypted(segment_is_encrypted); 1823 } 1824 } 1825 assert(segment_sp.get()); 1826 1827 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1828 1829 SectionSP section_sp(new Section( 1830 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1831 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1832 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1833 sect64.flags)); 1834 // Set the section to be encrypted to match the segment 1835 1836 bool section_is_encrypted = false; 1837 if (!segment_is_encrypted && load_cmd.filesize != 0) 1838 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1839 sect64.offset) != nullptr; 1840 1841 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1842 section_sp->SetPermissions(segment_permissions); 1843 segment_sp->GetChildren().AddSection(section_sp); 1844 1845 if (segment_sp->IsFake()) { 1846 segment_sp.reset(); 1847 const_segname.Clear(); 1848 } 1849 } 1850 } 1851 if (segment_sp && is_dsym) { 1852 if (first_segment_sectID <= context.NextSectionIdx) { 1853 lldb::user_id_t sect_uid; 1854 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1855 ++sect_uid) { 1856 SectionSP curr_section_sp( 1857 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1858 SectionSP next_section_sp; 1859 if (sect_uid + 1 <= context.NextSectionIdx) 1860 next_section_sp = 1861 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1862 1863 if (curr_section_sp.get()) { 1864 if (curr_section_sp->GetByteSize() == 0) { 1865 if (next_section_sp.get() != nullptr) 1866 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1867 curr_section_sp->GetFileAddress()); 1868 else 1869 curr_section_sp->SetByteSize(load_cmd.vmsize); 1870 } 1871 } 1872 } 1873 } 1874 } 1875 } 1876 1877 void ObjectFileMachO::ProcessDysymtabCommand( 1878 const llvm::MachO::load_command &load_cmd, lldb::offset_t offset) { 1879 m_dysymtab.cmd = load_cmd.cmd; 1880 m_dysymtab.cmdsize = load_cmd.cmdsize; 1881 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1882 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1883 } 1884 1885 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1886 if (m_sections_up) 1887 return; 1888 1889 m_sections_up = std::make_unique<SectionList>(); 1890 1891 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1892 // bool dump_sections = false; 1893 ModuleSP module_sp(GetModule()); 1894 1895 offset = MachHeaderSizeFromMagic(m_header.magic); 1896 1897 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1898 llvm::MachO::load_command load_cmd; 1899 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1900 const lldb::offset_t load_cmd_offset = offset; 1901 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 1902 break; 1903 1904 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1905 ProcessSegmentCommand(load_cmd, offset, i, context); 1906 else if (load_cmd.cmd == LC_DYSYMTAB) 1907 ProcessDysymtabCommand(load_cmd, offset); 1908 1909 offset = load_cmd_offset + load_cmd.cmdsize; 1910 } 1911 1912 if (context.FileAddressesChanged && module_sp) 1913 module_sp->SectionFileAddressesChanged(); 1914 } 1915 1916 class MachSymtabSectionInfo { 1917 public: 1918 MachSymtabSectionInfo(SectionList *section_list) 1919 : m_section_list(section_list), m_section_infos() { 1920 // Get the number of sections down to a depth of 1 to include all segments 1921 // and their sections, but no other sections that may be added for debug 1922 // map or 1923 m_section_infos.resize(section_list->GetNumSections(1)); 1924 } 1925 1926 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1927 if (n_sect == 0) 1928 return SectionSP(); 1929 if (n_sect < m_section_infos.size()) { 1930 if (!m_section_infos[n_sect].section_sp) { 1931 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1932 m_section_infos[n_sect].section_sp = section_sp; 1933 if (section_sp) { 1934 m_section_infos[n_sect].vm_range.SetBaseAddress( 1935 section_sp->GetFileAddress()); 1936 m_section_infos[n_sect].vm_range.SetByteSize( 1937 section_sp->GetByteSize()); 1938 } else { 1939 std::string filename = "<unknown>"; 1940 SectionSP first_section_sp(m_section_list->GetSectionAtIndex(0)); 1941 if (first_section_sp) 1942 filename = first_section_sp->GetObjectFile()->GetFileSpec().GetPath(); 1943 1944 Host::SystemLog(Host::eSystemLogError, 1945 "error: unable to find section %d for a symbol in " 1946 "%s, corrupt file?\n", 1947 n_sect, filename.c_str()); 1948 } 1949 } 1950 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1951 // Symbol is in section. 1952 return m_section_infos[n_sect].section_sp; 1953 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1954 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1955 file_addr) { 1956 // Symbol is in section with zero size, but has the same start address 1957 // as the section. This can happen with linker symbols (symbols that 1958 // start with the letter 'l' or 'L'. 1959 return m_section_infos[n_sect].section_sp; 1960 } 1961 } 1962 return m_section_list->FindSectionContainingFileAddress(file_addr); 1963 } 1964 1965 protected: 1966 struct SectionInfo { 1967 SectionInfo() : vm_range(), section_sp() {} 1968 1969 VMRange vm_range; 1970 SectionSP section_sp; 1971 }; 1972 SectionList *m_section_list; 1973 std::vector<SectionInfo> m_section_infos; 1974 }; 1975 1976 #define TRIE_SYMBOL_IS_THUMB (1ULL << 63) 1977 struct TrieEntry { 1978 void Dump() const { 1979 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1980 static_cast<unsigned long long>(address), 1981 static_cast<unsigned long long>(flags), 1982 static_cast<unsigned long long>(other), name.GetCString()); 1983 if (import_name) 1984 printf(" -> \"%s\"\n", import_name.GetCString()); 1985 else 1986 printf("\n"); 1987 } 1988 ConstString name; 1989 uint64_t address = LLDB_INVALID_ADDRESS; 1990 uint64_t flags = 1991 0; // EXPORT_SYMBOL_FLAGS_REEXPORT, EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER, 1992 // TRIE_SYMBOL_IS_THUMB 1993 uint64_t other = 0; 1994 ConstString import_name; 1995 }; 1996 1997 struct TrieEntryWithOffset { 1998 lldb::offset_t nodeOffset; 1999 TrieEntry entry; 2000 2001 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 2002 2003 void Dump(uint32_t idx) const { 2004 printf("[%3u] 0x%16.16llx: ", idx, 2005 static_cast<unsigned long long>(nodeOffset)); 2006 entry.Dump(); 2007 } 2008 2009 bool operator<(const TrieEntryWithOffset &other) const { 2010 return (nodeOffset < other.nodeOffset); 2011 } 2012 }; 2013 2014 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2015 const bool is_arm, addr_t text_seg_base_addr, 2016 std::vector<llvm::StringRef> &nameSlices, 2017 std::set<lldb::addr_t> &resolver_addresses, 2018 std::vector<TrieEntryWithOffset> &reexports, 2019 std::vector<TrieEntryWithOffset> &ext_symbols) { 2020 if (!data.ValidOffset(offset)) 2021 return true; 2022 2023 // Terminal node -- end of a branch, possibly add this to 2024 // the symbol table or resolver table. 2025 const uint64_t terminalSize = data.GetULEB128(&offset); 2026 lldb::offset_t children_offset = offset + terminalSize; 2027 if (terminalSize != 0) { 2028 TrieEntryWithOffset e(offset); 2029 e.entry.flags = data.GetULEB128(&offset); 2030 const char *import_name = nullptr; 2031 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2032 e.entry.address = 0; 2033 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2034 import_name = data.GetCStr(&offset); 2035 } else { 2036 e.entry.address = data.GetULEB128(&offset); 2037 if (text_seg_base_addr != LLDB_INVALID_ADDRESS) 2038 e.entry.address += text_seg_base_addr; 2039 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2040 e.entry.other = data.GetULEB128(&offset); 2041 uint64_t resolver_addr = e.entry.other; 2042 if (text_seg_base_addr != LLDB_INVALID_ADDRESS) 2043 resolver_addr += text_seg_base_addr; 2044 if (is_arm) 2045 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2046 resolver_addresses.insert(resolver_addr); 2047 } else 2048 e.entry.other = 0; 2049 } 2050 bool add_this_entry = false; 2051 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT) && 2052 import_name && import_name[0]) { 2053 // add symbols that are reexport symbols with a valid import name. 2054 add_this_entry = true; 2055 } else if (e.entry.flags == 0 && 2056 (import_name == nullptr || import_name[0] == '\0')) { 2057 // add externally visible symbols, in case the nlist record has 2058 // been stripped/omitted. 2059 add_this_entry = true; 2060 } 2061 if (add_this_entry) { 2062 std::string name; 2063 if (!nameSlices.empty()) { 2064 for (auto name_slice : nameSlices) 2065 name.append(name_slice.data(), name_slice.size()); 2066 } 2067 if (name.size() > 1) { 2068 // Skip the leading '_' 2069 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2070 } 2071 if (import_name) { 2072 // Skip the leading '_' 2073 e.entry.import_name.SetCString(import_name + 1); 2074 } 2075 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT)) { 2076 reexports.push_back(e); 2077 } else { 2078 if (is_arm && (e.entry.address & 1)) { 2079 e.entry.flags |= TRIE_SYMBOL_IS_THUMB; 2080 e.entry.address &= THUMB_ADDRESS_BIT_MASK; 2081 } 2082 ext_symbols.push_back(e); 2083 } 2084 } 2085 } 2086 2087 const uint8_t childrenCount = data.GetU8(&children_offset); 2088 for (uint8_t i = 0; i < childrenCount; ++i) { 2089 const char *cstr = data.GetCStr(&children_offset); 2090 if (cstr) 2091 nameSlices.push_back(llvm::StringRef(cstr)); 2092 else 2093 return false; // Corrupt data 2094 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2095 if (childNodeOffset) { 2096 if (!ParseTrieEntries(data, childNodeOffset, is_arm, text_seg_base_addr, 2097 nameSlices, resolver_addresses, reexports, 2098 ext_symbols)) { 2099 return false; 2100 } 2101 } 2102 nameSlices.pop_back(); 2103 } 2104 return true; 2105 } 2106 2107 static SymbolType GetSymbolType(const char *&symbol_name, 2108 bool &demangled_is_synthesized, 2109 const SectionSP &text_section_sp, 2110 const SectionSP &data_section_sp, 2111 const SectionSP &data_dirty_section_sp, 2112 const SectionSP &data_const_section_sp, 2113 const SectionSP &symbol_section) { 2114 SymbolType type = eSymbolTypeInvalid; 2115 2116 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2117 if (symbol_section->IsDescendant(text_section_sp.get())) { 2118 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 2119 S_ATTR_SELF_MODIFYING_CODE | 2120 S_ATTR_SOME_INSTRUCTIONS)) 2121 type = eSymbolTypeData; 2122 else 2123 type = eSymbolTypeCode; 2124 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 2125 symbol_section->IsDescendant(data_dirty_section_sp.get()) || 2126 symbol_section->IsDescendant(data_const_section_sp.get())) { 2127 if (symbol_sect_name && 2128 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 2129 type = eSymbolTypeRuntime; 2130 2131 if (symbol_name) { 2132 llvm::StringRef symbol_name_ref(symbol_name); 2133 if (symbol_name_ref.startswith("OBJC_")) { 2134 static const llvm::StringRef g_objc_v2_prefix_class("OBJC_CLASS_$_"); 2135 static const llvm::StringRef g_objc_v2_prefix_metaclass( 2136 "OBJC_METACLASS_$_"); 2137 static const llvm::StringRef g_objc_v2_prefix_ivar("OBJC_IVAR_$_"); 2138 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 2139 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2140 type = eSymbolTypeObjCClass; 2141 demangled_is_synthesized = true; 2142 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 2143 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2144 type = eSymbolTypeObjCMetaClass; 2145 demangled_is_synthesized = true; 2146 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 2147 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2148 type = eSymbolTypeObjCIVar; 2149 demangled_is_synthesized = true; 2150 } 2151 } 2152 } 2153 } else if (symbol_sect_name && 2154 ::strstr(symbol_sect_name, "__gcc_except_tab") == 2155 symbol_sect_name) { 2156 type = eSymbolTypeException; 2157 } else { 2158 type = eSymbolTypeData; 2159 } 2160 } else if (symbol_sect_name && 2161 ::strstr(symbol_sect_name, "__IMPORT") == symbol_sect_name) { 2162 type = eSymbolTypeTrampoline; 2163 } 2164 return type; 2165 } 2166 2167 // Read the UUID out of a dyld_shared_cache file on-disk. 2168 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2169 const ByteOrder byte_order, 2170 const uint32_t addr_byte_size) { 2171 UUID dsc_uuid; 2172 DataBufferSP DscData = MapFileData( 2173 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2174 if (!DscData) 2175 return dsc_uuid; 2176 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2177 2178 char version_str[7]; 2179 lldb::offset_t offset = 0; 2180 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2181 version_str[6] = '\0'; 2182 if (strcmp(version_str, "dyld_v") == 0) { 2183 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2184 dsc_uuid = UUID::fromOptionalData( 2185 dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t)); 2186 } 2187 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2188 if (log && dsc_uuid.IsValid()) { 2189 LLDB_LOGF(log, "Shared cache %s has UUID %s", 2190 dyld_shared_cache.GetPath().c_str(), 2191 dsc_uuid.GetAsString().c_str()); 2192 } 2193 return dsc_uuid; 2194 } 2195 2196 static llvm::Optional<struct nlist_64> 2197 ParseNList(DataExtractor &nlist_data, lldb::offset_t &nlist_data_offset, 2198 size_t nlist_byte_size) { 2199 struct nlist_64 nlist; 2200 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2201 return {}; 2202 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 2203 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 2204 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 2205 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 2206 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 2207 return nlist; 2208 } 2209 2210 enum { DebugSymbols = true, NonDebugSymbols = false }; 2211 2212 size_t ObjectFileMachO::ParseSymtab() { 2213 LLDB_SCOPED_TIMERF("ObjectFileMachO::ParseSymtab () module = %s", 2214 m_file.GetFilename().AsCString("")); 2215 ModuleSP module_sp(GetModule()); 2216 if (!module_sp) 2217 return 0; 2218 2219 Progress progress(llvm::formatv("Parsing symbol table for {0}", 2220 m_file.GetFilename().AsCString("<Unknown>"))); 2221 2222 llvm::MachO::symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2223 llvm::MachO::linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2224 llvm::MachO::linkedit_data_command exports_trie_load_command = {0, 0, 0, 0}; 2225 llvm::MachO::dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2226 // The data element of type bool indicates that this entry is thumb 2227 // code. 2228 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2229 2230 // Record the address of every function/data that we add to the symtab. 2231 // We add symbols to the table in the order of most information (nlist 2232 // records) to least (function starts), and avoid duplicating symbols 2233 // via this set. 2234 llvm::DenseSet<addr_t> symbols_added; 2235 2236 // We are using a llvm::DenseSet for "symbols_added" so we must be sure we 2237 // do not add the tombstone or empty keys to the set. 2238 auto add_symbol_addr = [&symbols_added](lldb::addr_t file_addr) { 2239 // Don't add the tombstone or empty keys. 2240 if (file_addr == UINT64_MAX || file_addr == UINT64_MAX - 1) 2241 return; 2242 symbols_added.insert(file_addr); 2243 }; 2244 FunctionStarts function_starts; 2245 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2246 uint32_t i; 2247 FileSpecList dylib_files; 2248 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2249 llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2250 llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2251 llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2252 2253 for (i = 0; i < m_header.ncmds; ++i) { 2254 const lldb::offset_t cmd_offset = offset; 2255 // Read in the load command and load command size 2256 llvm::MachO::load_command lc; 2257 if (m_data.GetU32(&offset, &lc, 2) == nullptr) 2258 break; 2259 // Watch for the symbol table load command 2260 switch (lc.cmd) { 2261 case LC_SYMTAB: 2262 symtab_load_command.cmd = lc.cmd; 2263 symtab_load_command.cmdsize = lc.cmdsize; 2264 // Read in the rest of the symtab load command 2265 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2266 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2267 return 0; 2268 break; 2269 2270 case LC_DYLD_INFO: 2271 case LC_DYLD_INFO_ONLY: 2272 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2273 dyld_info.cmd = lc.cmd; 2274 dyld_info.cmdsize = lc.cmdsize; 2275 } else { 2276 memset(&dyld_info, 0, sizeof(dyld_info)); 2277 } 2278 break; 2279 2280 case LC_LOAD_DYLIB: 2281 case LC_LOAD_WEAK_DYLIB: 2282 case LC_REEXPORT_DYLIB: 2283 case LC_LOADFVMLIB: 2284 case LC_LOAD_UPWARD_DYLIB: { 2285 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2286 const char *path = m_data.PeekCStr(name_offset); 2287 if (path) { 2288 FileSpec file_spec(path); 2289 // Strip the path if there is @rpath, @executable, etc so we just use 2290 // the basename 2291 if (path[0] == '@') 2292 file_spec.GetDirectory().Clear(); 2293 2294 if (lc.cmd == LC_REEXPORT_DYLIB) { 2295 m_reexported_dylibs.AppendIfUnique(file_spec); 2296 } 2297 2298 dylib_files.Append(file_spec); 2299 } 2300 } break; 2301 2302 case LC_DYLD_EXPORTS_TRIE: 2303 exports_trie_load_command.cmd = lc.cmd; 2304 exports_trie_load_command.cmdsize = lc.cmdsize; 2305 if (m_data.GetU32(&offset, &exports_trie_load_command.dataoff, 2) == 2306 nullptr) // fill in offset and size fields 2307 memset(&exports_trie_load_command, 0, 2308 sizeof(exports_trie_load_command)); 2309 break; 2310 case LC_FUNCTION_STARTS: 2311 function_starts_load_command.cmd = lc.cmd; 2312 function_starts_load_command.cmdsize = lc.cmdsize; 2313 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2314 nullptr) // fill in data offset and size fields 2315 memset(&function_starts_load_command, 0, 2316 sizeof(function_starts_load_command)); 2317 break; 2318 2319 default: 2320 break; 2321 } 2322 offset = cmd_offset + lc.cmdsize; 2323 } 2324 2325 if (!symtab_load_command.cmd) 2326 return 0; 2327 2328 Symtab *symtab = m_symtab_up.get(); 2329 SectionList *section_list = GetSectionList(); 2330 if (section_list == nullptr) 2331 return 0; 2332 2333 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2334 const ByteOrder byte_order = m_data.GetByteOrder(); 2335 bool bit_width_32 = addr_byte_size == 4; 2336 const size_t nlist_byte_size = 2337 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2338 2339 DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size); 2340 DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size); 2341 DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size); 2342 DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order, 2343 addr_byte_size); 2344 DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size); 2345 2346 const addr_t nlist_data_byte_size = 2347 symtab_load_command.nsyms * nlist_byte_size; 2348 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2349 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2350 2351 ProcessSP process_sp(m_process_wp.lock()); 2352 Process *process = process_sp.get(); 2353 2354 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2355 bool is_shared_cache_image = m_header.flags & MH_DYLIB_IN_CACHE; 2356 bool is_local_shared_cache_image = is_shared_cache_image && !IsInMemory(); 2357 SectionSP linkedit_section_sp( 2358 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2359 2360 if (process && m_header.filetype != llvm::MachO::MH_OBJECT && 2361 !is_local_shared_cache_image) { 2362 Target &target = process->GetTarget(); 2363 2364 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2365 2366 // Reading mach file from memory in a process or core file... 2367 2368 if (linkedit_section_sp) { 2369 addr_t linkedit_load_addr = 2370 linkedit_section_sp->GetLoadBaseAddress(&target); 2371 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2372 // We might be trying to access the symbol table before the 2373 // __LINKEDIT's load address has been set in the target. We can't 2374 // fail to read the symbol table, so calculate the right address 2375 // manually 2376 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2377 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2378 } 2379 2380 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2381 const addr_t symoff_addr = linkedit_load_addr + 2382 symtab_load_command.symoff - 2383 linkedit_file_offset; 2384 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2385 linkedit_file_offset; 2386 2387 // Always load dyld - the dynamic linker - from memory if we didn't 2388 // find a binary anywhere else. lldb will not register 2389 // dylib/framework/bundle loads/unloads if we don't have the dyld 2390 // symbols, we force dyld to load from memory despite the user's 2391 // target.memory-module-load-level setting. 2392 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2393 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2394 DataBufferSP nlist_data_sp( 2395 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2396 if (nlist_data_sp) 2397 nlist_data.SetData(nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2398 if (m_dysymtab.nindirectsyms != 0) { 2399 const addr_t indirect_syms_addr = linkedit_load_addr + 2400 m_dysymtab.indirectsymoff - 2401 linkedit_file_offset; 2402 DataBufferSP indirect_syms_data_sp(ReadMemory( 2403 process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 2404 if (indirect_syms_data_sp) 2405 indirect_symbol_index_data.SetData( 2406 indirect_syms_data_sp, 0, 2407 indirect_syms_data_sp->GetByteSize()); 2408 // If this binary is outside the shared cache, 2409 // cache the string table. 2410 // Binaries in the shared cache all share a giant string table, 2411 // and we can't share the string tables across multiple 2412 // ObjectFileMachO's, so we'd end up re-reading this mega-strtab 2413 // for every binary in the shared cache - it would be a big perf 2414 // problem. For binaries outside the shared cache, it's faster to 2415 // read the entire strtab at once instead of piece-by-piece as we 2416 // process the nlist records. 2417 if (!is_shared_cache_image) { 2418 DataBufferSP strtab_data_sp( 2419 ReadMemory(process_sp, strtab_addr, strtab_data_byte_size)); 2420 if (strtab_data_sp) { 2421 strtab_data.SetData(strtab_data_sp, 0, 2422 strtab_data_sp->GetByteSize()); 2423 } 2424 } 2425 } 2426 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) { 2427 if (function_starts_load_command.cmd) { 2428 const addr_t func_start_addr = 2429 linkedit_load_addr + function_starts_load_command.dataoff - 2430 linkedit_file_offset; 2431 DataBufferSP func_start_data_sp( 2432 ReadMemory(process_sp, func_start_addr, 2433 function_starts_load_command.datasize)); 2434 if (func_start_data_sp) 2435 function_starts_data.SetData(func_start_data_sp, 0, 2436 func_start_data_sp->GetByteSize()); 2437 } 2438 } 2439 } 2440 } 2441 } else { 2442 if (is_local_shared_cache_image) { 2443 // The load commands in shared cache images are relative to the 2444 // beginning of the shared cache, not the library image. The 2445 // data we get handed when creating the ObjectFileMachO starts 2446 // at the beginning of a specific library and spans to the end 2447 // of the cache to be able to reach the shared LINKEDIT 2448 // segments. We need to convert the load command offsets to be 2449 // relative to the beginning of our specific image. 2450 lldb::addr_t linkedit_offset = linkedit_section_sp->GetFileOffset(); 2451 lldb::offset_t linkedit_slide = 2452 linkedit_offset - m_linkedit_original_offset; 2453 symtab_load_command.symoff += linkedit_slide; 2454 symtab_load_command.stroff += linkedit_slide; 2455 dyld_info.export_off += linkedit_slide; 2456 m_dysymtab.indirectsymoff += linkedit_slide; 2457 function_starts_load_command.dataoff += linkedit_slide; 2458 exports_trie_load_command.dataoff += linkedit_slide; 2459 } 2460 2461 nlist_data.SetData(m_data, symtab_load_command.symoff, 2462 nlist_data_byte_size); 2463 strtab_data.SetData(m_data, symtab_load_command.stroff, 2464 strtab_data_byte_size); 2465 2466 // We shouldn't have exports data from both the LC_DYLD_INFO command 2467 // AND the LC_DYLD_EXPORTS_TRIE command in the same binary: 2468 lldbassert(!((dyld_info.export_size > 0) 2469 && (exports_trie_load_command.datasize > 0))); 2470 if (dyld_info.export_size > 0) { 2471 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2472 dyld_info.export_size); 2473 } else if (exports_trie_load_command.datasize > 0) { 2474 dyld_trie_data.SetData(m_data, exports_trie_load_command.dataoff, 2475 exports_trie_load_command.datasize); 2476 } 2477 2478 if (m_dysymtab.nindirectsyms != 0) { 2479 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2480 m_dysymtab.nindirectsyms * 4); 2481 } 2482 if (function_starts_load_command.cmd) { 2483 function_starts_data.SetData(m_data, function_starts_load_command.dataoff, 2484 function_starts_load_command.datasize); 2485 } 2486 } 2487 2488 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2489 2490 ConstString g_segment_name_TEXT = GetSegmentNameTEXT(); 2491 ConstString g_segment_name_DATA = GetSegmentNameDATA(); 2492 ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2493 ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2494 ConstString g_segment_name_OBJC = GetSegmentNameOBJC(); 2495 ConstString g_section_name_eh_frame = GetSectionNameEHFrame(); 2496 SectionSP text_section_sp( 2497 section_list->FindSectionByName(g_segment_name_TEXT)); 2498 SectionSP data_section_sp( 2499 section_list->FindSectionByName(g_segment_name_DATA)); 2500 SectionSP data_dirty_section_sp( 2501 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2502 SectionSP data_const_section_sp( 2503 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2504 SectionSP objc_section_sp( 2505 section_list->FindSectionByName(g_segment_name_OBJC)); 2506 SectionSP eh_frame_section_sp; 2507 if (text_section_sp.get()) 2508 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2509 g_section_name_eh_frame); 2510 else 2511 eh_frame_section_sp = 2512 section_list->FindSectionByName(g_section_name_eh_frame); 2513 2514 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2515 const bool always_thumb = GetArchitecture().IsAlwaysThumbInstructions(); 2516 2517 // lldb works best if it knows the start address of all functions in a 2518 // module. Linker symbols or debug info are normally the best source of 2519 // information for start addr / size but they may be stripped in a released 2520 // binary. Two additional sources of information exist in Mach-O binaries: 2521 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2522 // function's start address in the 2523 // binary, relative to the text section. 2524 // eh_frame - the eh_frame FDEs have the start addr & size of 2525 // each function 2526 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2527 // all modern binaries. 2528 // Binaries built to run on older releases may need to use eh_frame 2529 // information. 2530 2531 if (text_section_sp && function_starts_data.GetByteSize()) { 2532 FunctionStarts::Entry function_start_entry; 2533 function_start_entry.data = false; 2534 lldb::offset_t function_start_offset = 0; 2535 function_start_entry.addr = text_section_sp->GetFileAddress(); 2536 uint64_t delta; 2537 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2538 0) { 2539 // Now append the current entry 2540 function_start_entry.addr += delta; 2541 if (is_arm) { 2542 if (function_start_entry.addr & 1) { 2543 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2544 function_start_entry.data = true; 2545 } else if (always_thumb) { 2546 function_start_entry.data = true; 2547 } 2548 } 2549 function_starts.Append(function_start_entry); 2550 } 2551 } else { 2552 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2553 // load command claiming an eh_frame but it doesn't actually have the 2554 // eh_frame content. And if we have a dSYM, we don't need to do any of 2555 // this fill-in-the-missing-symbols works anyway - the debug info should 2556 // give us all the functions in the module. 2557 if (text_section_sp.get() && eh_frame_section_sp.get() && 2558 m_type != eTypeDebugInfo) { 2559 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2560 DWARFCallFrameInfo::EH); 2561 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2562 eh_frame.GetFunctionAddressAndSizeVector(functions); 2563 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2564 size_t count = functions.GetSize(); 2565 for (size_t i = 0; i < count; ++i) { 2566 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2567 functions.GetEntryAtIndex(i); 2568 if (func) { 2569 FunctionStarts::Entry function_start_entry; 2570 function_start_entry.addr = func->base - text_base_addr; 2571 if (is_arm) { 2572 if (function_start_entry.addr & 1) { 2573 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2574 function_start_entry.data = true; 2575 } else if (always_thumb) { 2576 function_start_entry.data = true; 2577 } 2578 } 2579 function_starts.Append(function_start_entry); 2580 } 2581 } 2582 } 2583 } 2584 2585 const size_t function_starts_count = function_starts.GetSize(); 2586 2587 // For user process binaries (executables, dylibs, frameworks, bundles), if 2588 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2589 // going to assume the binary has been stripped. Don't allow assembly 2590 // language instruction emulation because we don't know proper function 2591 // start boundaries. 2592 // 2593 // For all other types of binaries (kernels, stand-alone bare board 2594 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2595 // sections - we should not make any assumptions about them based on that. 2596 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2597 m_allow_assembly_emulation_unwind_plans = false; 2598 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2599 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2600 2601 if (unwind_or_symbol_log) 2602 module_sp->LogMessage( 2603 unwind_or_symbol_log, 2604 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2605 } 2606 2607 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() 2608 ? eh_frame_section_sp->GetID() 2609 : static_cast<user_id_t>(NO_SECT); 2610 2611 lldb::offset_t nlist_data_offset = 0; 2612 2613 uint32_t N_SO_index = UINT32_MAX; 2614 2615 MachSymtabSectionInfo section_info(section_list); 2616 std::vector<uint32_t> N_FUN_indexes; 2617 std::vector<uint32_t> N_NSYM_indexes; 2618 std::vector<uint32_t> N_INCL_indexes; 2619 std::vector<uint32_t> N_BRAC_indexes; 2620 std::vector<uint32_t> N_COMM_indexes; 2621 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2622 typedef llvm::DenseMap<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2623 typedef llvm::DenseMap<const char *, uint32_t> ConstNameToSymbolIndexMap; 2624 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2625 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2626 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2627 // Any symbols that get merged into another will get an entry in this map 2628 // so we know 2629 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2630 uint32_t nlist_idx = 0; 2631 Symbol *symbol_ptr = nullptr; 2632 2633 uint32_t sym_idx = 0; 2634 Symbol *sym = nullptr; 2635 size_t num_syms = 0; 2636 std::string memory_symbol_name; 2637 uint32_t unmapped_local_symbols_found = 0; 2638 2639 std::vector<TrieEntryWithOffset> reexport_trie_entries; 2640 std::vector<TrieEntryWithOffset> external_sym_trie_entries; 2641 std::set<lldb::addr_t> resolver_addresses; 2642 2643 if (dyld_trie_data.GetByteSize() > 0) { 2644 ConstString text_segment_name("__TEXT"); 2645 SectionSP text_segment_sp = 2646 GetSectionList()->FindSectionByName(text_segment_name); 2647 lldb::addr_t text_segment_file_addr = LLDB_INVALID_ADDRESS; 2648 if (text_segment_sp) 2649 text_segment_file_addr = text_segment_sp->GetFileAddress(); 2650 std::vector<llvm::StringRef> nameSlices; 2651 ParseTrieEntries(dyld_trie_data, 0, is_arm, text_segment_file_addr, 2652 nameSlices, resolver_addresses, reexport_trie_entries, 2653 external_sym_trie_entries); 2654 } 2655 2656 typedef std::set<ConstString> IndirectSymbols; 2657 IndirectSymbols indirect_symbol_names; 2658 2659 #if TARGET_OS_IPHONE 2660 2661 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2662 // optimized by moving LOCAL symbols out of the memory mapped portion of 2663 // the DSC. The symbol information has all been retained, but it isn't 2664 // available in the normal nlist data. However, there *are* duplicate 2665 // entries of *some* 2666 // LOCAL symbols in the normal nlist data. To handle this situation 2667 // correctly, we must first attempt 2668 // to parse any DSC unmapped symbol information. If we find any, we set a 2669 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2670 2671 if (m_header.flags & MH_DYLIB_IN_CACHE) { 2672 // Before we can start mapping the DSC, we need to make certain the 2673 // target process is actually using the cache we can find. 2674 2675 // Next we need to determine the correct path for the dyld shared cache. 2676 2677 ArchSpec header_arch = GetArchitecture(); 2678 char dsc_path[PATH_MAX]; 2679 char dsc_path_development[PATH_MAX]; 2680 2681 snprintf( 2682 dsc_path, sizeof(dsc_path), "%s%s%s", 2683 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2684 */ 2685 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2686 header_arch.GetArchitectureName()); 2687 2688 snprintf( 2689 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2690 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2691 */ 2692 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2693 header_arch.GetArchitectureName(), ".development"); 2694 2695 FileSpec dsc_nondevelopment_filespec(dsc_path); 2696 FileSpec dsc_development_filespec(dsc_path_development); 2697 FileSpec dsc_filespec; 2698 2699 UUID dsc_uuid; 2700 UUID process_shared_cache_uuid; 2701 addr_t process_shared_cache_base_addr; 2702 2703 if (process) { 2704 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, 2705 process_shared_cache_uuid); 2706 } 2707 2708 // First see if we can find an exact match for the inferior process 2709 // shared cache UUID in the development or non-development shared caches 2710 // on disk. 2711 if (process_shared_cache_uuid.IsValid()) { 2712 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2713 UUID dsc_development_uuid = GetSharedCacheUUID( 2714 dsc_development_filespec, byte_order, addr_byte_size); 2715 if (dsc_development_uuid.IsValid() && 2716 dsc_development_uuid == process_shared_cache_uuid) { 2717 dsc_filespec = dsc_development_filespec; 2718 dsc_uuid = dsc_development_uuid; 2719 } 2720 } 2721 if (!dsc_uuid.IsValid() && 2722 FileSystem::Instance().Exists(dsc_nondevelopment_filespec)) { 2723 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2724 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2725 if (dsc_nondevelopment_uuid.IsValid() && 2726 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2727 dsc_filespec = dsc_nondevelopment_filespec; 2728 dsc_uuid = dsc_nondevelopment_uuid; 2729 } 2730 } 2731 } 2732 2733 // Failing a UUID match, prefer the development dyld_shared cache if both 2734 // are present. 2735 if (!FileSystem::Instance().Exists(dsc_filespec)) { 2736 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2737 dsc_filespec = dsc_development_filespec; 2738 } else { 2739 dsc_filespec = dsc_nondevelopment_filespec; 2740 } 2741 } 2742 2743 /* The dyld_cache_header has a pointer to the 2744 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2745 The dyld_cache_local_symbols_info structure gives us three things: 2746 1. The start and count of the nlist records in the dyld_shared_cache 2747 file 2748 2. The start and size of the strings for these nlist records 2749 3. The start and count of dyld_cache_local_symbols_entry entries 2750 2751 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2752 dyld shared cache. 2753 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2754 the dyld shared cache. 2755 The dyld_cache_local_symbols_entry also lists the start of this 2756 dylib/framework's nlist records 2757 and the count of how many nlist records there are for this 2758 dylib/framework. 2759 */ 2760 2761 // Process the dyld shared cache header to find the unmapped symbols 2762 2763 DataBufferSP dsc_data_sp = MapFileData( 2764 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2765 if (!dsc_uuid.IsValid()) { 2766 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2767 } 2768 if (dsc_data_sp) { 2769 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2770 2771 bool uuid_match = true; 2772 if (dsc_uuid.IsValid() && process) { 2773 if (process_shared_cache_uuid.IsValid() && 2774 dsc_uuid != process_shared_cache_uuid) { 2775 // The on-disk dyld_shared_cache file is not the same as the one in 2776 // this process' memory, don't use it. 2777 uuid_match = false; 2778 ModuleSP module_sp(GetModule()); 2779 if (module_sp) 2780 module_sp->ReportWarning("process shared cache does not match " 2781 "on-disk dyld_shared_cache file, some " 2782 "symbol names will be missing."); 2783 } 2784 } 2785 2786 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2787 2788 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2789 2790 // If the mappingOffset points to a location inside the header, we've 2791 // opened an old dyld shared cache, and should not proceed further. 2792 if (uuid_match && 2793 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2794 2795 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2796 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2797 mappingOffset); 2798 2799 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2800 byte_order, addr_byte_size); 2801 offset = 0; 2802 2803 // The File addresses (from the in-memory Mach-O load commands) for 2804 // the shared libraries in the shared library cache need to be 2805 // adjusted by an offset to match up with the dylibOffset identifying 2806 // field in the dyld_cache_local_symbol_entry's. This offset is 2807 // recorded in mapping_offset_value. 2808 const uint64_t mapping_offset_value = 2809 dsc_mapping_info_data.GetU64(&offset); 2810 2811 offset = 2812 offsetof(struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2813 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2814 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2815 2816 if (localSymbolsOffset && localSymbolsSize) { 2817 // Map the local symbols 2818 DataBufferSP dsc_local_symbols_data_sp = 2819 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2820 2821 if (dsc_local_symbols_data_sp) { 2822 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2823 byte_order, addr_byte_size); 2824 2825 offset = 0; 2826 2827 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 2828 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 2829 UndefinedNameToDescMap undefined_name_to_desc; 2830 SymbolIndexToName reexport_shlib_needs_fixup; 2831 2832 // Read the local_symbols_infos struct in one shot 2833 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2834 dsc_local_symbols_data.GetU32(&offset, 2835 &local_symbols_info.nlistOffset, 6); 2836 2837 SectionSP text_section_sp( 2838 section_list->FindSectionByName(GetSegmentNameTEXT())); 2839 2840 uint32_t header_file_offset = 2841 (text_section_sp->GetFileAddress() - mapping_offset_value); 2842 2843 offset = local_symbols_info.entriesOffset; 2844 for (uint32_t entry_index = 0; 2845 entry_index < local_symbols_info.entriesCount; entry_index++) { 2846 struct lldb_copy_dyld_cache_local_symbols_entry 2847 local_symbols_entry; 2848 local_symbols_entry.dylibOffset = 2849 dsc_local_symbols_data.GetU32(&offset); 2850 local_symbols_entry.nlistStartIndex = 2851 dsc_local_symbols_data.GetU32(&offset); 2852 local_symbols_entry.nlistCount = 2853 dsc_local_symbols_data.GetU32(&offset); 2854 2855 if (header_file_offset == local_symbols_entry.dylibOffset) { 2856 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2857 2858 // The normal nlist code cannot correctly size the Symbols 2859 // array, we need to allocate it here. 2860 sym = symtab->Resize( 2861 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2862 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2863 num_syms = symtab->GetNumSymbols(); 2864 2865 nlist_data_offset = 2866 local_symbols_info.nlistOffset + 2867 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2868 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2869 2870 for (uint32_t nlist_index = 0; 2871 nlist_index < local_symbols_entry.nlistCount; 2872 nlist_index++) { 2873 ///////////////////////////// 2874 { 2875 llvm::Optional<struct nlist_64> nlist_maybe = 2876 ParseNList(dsc_local_symbols_data, nlist_data_offset, 2877 nlist_byte_size); 2878 if (!nlist_maybe) 2879 break; 2880 struct nlist_64 nlist = *nlist_maybe; 2881 2882 SymbolType type = eSymbolTypeInvalid; 2883 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2884 string_table_offset + nlist.n_strx); 2885 2886 if (symbol_name == NULL) { 2887 // No symbol should be NULL, even the symbols with no 2888 // string values should have an offset zero which 2889 // points to an empty C-string 2890 Host::SystemLog( 2891 Host::eSystemLogError, 2892 "error: DSC unmapped local symbol[%u] has invalid " 2893 "string table offset 0x%x in %s, ignoring symbol\n", 2894 entry_index, nlist.n_strx, 2895 module_sp->GetFileSpec().GetPath().c_str()); 2896 continue; 2897 } 2898 if (symbol_name[0] == '\0') 2899 symbol_name = NULL; 2900 2901 const char *symbol_name_non_abi_mangled = NULL; 2902 2903 SectionSP symbol_section; 2904 uint32_t symbol_byte_size = 0; 2905 bool add_nlist = true; 2906 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2907 bool demangled_is_synthesized = false; 2908 bool is_gsym = false; 2909 bool set_value = true; 2910 2911 assert(sym_idx < num_syms); 2912 2913 sym[sym_idx].SetDebug(is_debug); 2914 2915 if (is_debug) { 2916 switch (nlist.n_type) { 2917 case N_GSYM: 2918 // global symbol: name,,NO_SECT,type,0 2919 // Sometimes the N_GSYM value contains the address. 2920 2921 // FIXME: In the .o files, we have a GSYM and a debug 2922 // symbol for all the ObjC data. They 2923 // have the same address, but we want to ensure that 2924 // we always find only the real symbol, 'cause we 2925 // don't currently correctly attribute the 2926 // GSYM one to the ObjCClass/Ivar/MetaClass 2927 // symbol type. This is a temporary hack to make 2928 // sure the ObjectiveC symbols get treated correctly. 2929 // To do this right, we should coalesce all the GSYM 2930 // & global symbols that have the same address. 2931 2932 is_gsym = true; 2933 sym[sym_idx].SetExternal(true); 2934 2935 if (symbol_name && symbol_name[0] == '_' && 2936 symbol_name[1] == 'O') { 2937 llvm::StringRef symbol_name_ref(symbol_name); 2938 if (symbol_name_ref.startswith( 2939 g_objc_v2_prefix_class)) { 2940 symbol_name_non_abi_mangled = symbol_name + 1; 2941 symbol_name = 2942 symbol_name + g_objc_v2_prefix_class.size(); 2943 type = eSymbolTypeObjCClass; 2944 demangled_is_synthesized = true; 2945 2946 } else if (symbol_name_ref.startswith( 2947 g_objc_v2_prefix_metaclass)) { 2948 symbol_name_non_abi_mangled = symbol_name + 1; 2949 symbol_name = 2950 symbol_name + g_objc_v2_prefix_metaclass.size(); 2951 type = eSymbolTypeObjCMetaClass; 2952 demangled_is_synthesized = true; 2953 } else if (symbol_name_ref.startswith( 2954 g_objc_v2_prefix_ivar)) { 2955 symbol_name_non_abi_mangled = symbol_name + 1; 2956 symbol_name = 2957 symbol_name + g_objc_v2_prefix_ivar.size(); 2958 type = eSymbolTypeObjCIVar; 2959 demangled_is_synthesized = true; 2960 } 2961 } else { 2962 if (nlist.n_value != 0) 2963 symbol_section = section_info.GetSection( 2964 nlist.n_sect, nlist.n_value); 2965 type = eSymbolTypeData; 2966 } 2967 break; 2968 2969 case N_FNAME: 2970 // procedure name (f77 kludge): name,,NO_SECT,0,0 2971 type = eSymbolTypeCompiler; 2972 break; 2973 2974 case N_FUN: 2975 // procedure: name,,n_sect,linenumber,address 2976 if (symbol_name) { 2977 type = eSymbolTypeCode; 2978 symbol_section = section_info.GetSection( 2979 nlist.n_sect, nlist.n_value); 2980 2981 N_FUN_addr_to_sym_idx.insert( 2982 std::make_pair(nlist.n_value, sym_idx)); 2983 // We use the current number of symbols in the 2984 // symbol table in lieu of using nlist_idx in case 2985 // we ever start trimming entries out 2986 N_FUN_indexes.push_back(sym_idx); 2987 } else { 2988 type = eSymbolTypeCompiler; 2989 2990 if (!N_FUN_indexes.empty()) { 2991 // Copy the size of the function into the 2992 // original 2993 // STAB entry so we don't have 2994 // to hunt for it later 2995 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2996 ->SetByteSize(nlist.n_value); 2997 N_FUN_indexes.pop_back(); 2998 // We don't really need the end function STAB as 2999 // it contains the size which we already placed 3000 // with the original symbol, so don't add it if 3001 // we want a minimal symbol table 3002 add_nlist = false; 3003 } 3004 } 3005 break; 3006 3007 case N_STSYM: 3008 // static symbol: name,,n_sect,type,address 3009 N_STSYM_addr_to_sym_idx.insert( 3010 std::make_pair(nlist.n_value, sym_idx)); 3011 symbol_section = section_info.GetSection(nlist.n_sect, 3012 nlist.n_value); 3013 if (symbol_name && symbol_name[0]) { 3014 type = ObjectFile::GetSymbolTypeFromName( 3015 symbol_name + 1, eSymbolTypeData); 3016 } 3017 break; 3018 3019 case N_LCSYM: 3020 // .lcomm symbol: name,,n_sect,type,address 3021 symbol_section = section_info.GetSection(nlist.n_sect, 3022 nlist.n_value); 3023 type = eSymbolTypeCommonBlock; 3024 break; 3025 3026 case N_BNSYM: 3027 // We use the current number of symbols in the symbol 3028 // table in lieu of using nlist_idx in case we ever 3029 // start trimming entries out Skip these if we want 3030 // minimal symbol tables 3031 add_nlist = false; 3032 break; 3033 3034 case N_ENSYM: 3035 // Set the size of the N_BNSYM to the terminating 3036 // index of this N_ENSYM so that we can always skip 3037 // the entire symbol if we need to navigate more 3038 // quickly at the source level when parsing STABS 3039 // Skip these if we want minimal symbol tables 3040 add_nlist = false; 3041 break; 3042 3043 case N_OPT: 3044 // emitted with gcc2_compiled and in gcc source 3045 type = eSymbolTypeCompiler; 3046 break; 3047 3048 case N_RSYM: 3049 // register sym: name,,NO_SECT,type,register 3050 type = eSymbolTypeVariable; 3051 break; 3052 3053 case N_SLINE: 3054 // src line: 0,,n_sect,linenumber,address 3055 symbol_section = section_info.GetSection(nlist.n_sect, 3056 nlist.n_value); 3057 type = eSymbolTypeLineEntry; 3058 break; 3059 3060 case N_SSYM: 3061 // structure elt: name,,NO_SECT,type,struct_offset 3062 type = eSymbolTypeVariableType; 3063 break; 3064 3065 case N_SO: 3066 // source file name 3067 type = eSymbolTypeSourceFile; 3068 if (symbol_name == NULL) { 3069 add_nlist = false; 3070 if (N_SO_index != UINT32_MAX) { 3071 // Set the size of the N_SO to the terminating 3072 // index of this N_SO so that we can always skip 3073 // the entire N_SO if we need to navigate more 3074 // quickly at the source level when parsing STABS 3075 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3076 symbol_ptr->SetByteSize(sym_idx); 3077 symbol_ptr->SetSizeIsSibling(true); 3078 } 3079 N_NSYM_indexes.clear(); 3080 N_INCL_indexes.clear(); 3081 N_BRAC_indexes.clear(); 3082 N_COMM_indexes.clear(); 3083 N_FUN_indexes.clear(); 3084 N_SO_index = UINT32_MAX; 3085 } else { 3086 // We use the current number of symbols in the 3087 // symbol table in lieu of using nlist_idx in case 3088 // we ever start trimming entries out 3089 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3090 if (N_SO_has_full_path) { 3091 if ((N_SO_index == sym_idx - 1) && 3092 ((sym_idx - 1) < num_syms)) { 3093 // We have two consecutive N_SO entries where 3094 // the first contains a directory and the 3095 // second contains a full path. 3096 sym[sym_idx - 1].GetMangled().SetValue( 3097 ConstString(symbol_name), false); 3098 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3099 add_nlist = false; 3100 } else { 3101 // This is the first entry in a N_SO that 3102 // contains a directory or 3103 // a full path to the source file 3104 N_SO_index = sym_idx; 3105 } 3106 } else if ((N_SO_index == sym_idx - 1) && 3107 ((sym_idx - 1) < num_syms)) { 3108 // This is usually the second N_SO entry that 3109 // contains just the filename, so here we combine 3110 // it with the first one if we are minimizing the 3111 // symbol table 3112 const char *so_path = sym[sym_idx - 1] 3113 .GetMangled() 3114 .GetDemangledName() 3115 .AsCString(); 3116 if (so_path && so_path[0]) { 3117 std::string full_so_path(so_path); 3118 const size_t double_slash_pos = 3119 full_so_path.find("//"); 3120 if (double_slash_pos != std::string::npos) { 3121 // The linker has been generating bad N_SO 3122 // entries with doubled up paths 3123 // in the format "%s%s" where the first 3124 // string in the DW_AT_comp_dir, and the 3125 // second is the directory for the source 3126 // file so you end up with a path that looks 3127 // like "/tmp/src//tmp/src/" 3128 FileSpec so_dir(so_path); 3129 if (!FileSystem::Instance().Exists(so_dir)) { 3130 so_dir.SetFile( 3131 &full_so_path[double_slash_pos + 1], 3132 FileSpec::Style::native); 3133 if (FileSystem::Instance().Exists(so_dir)) { 3134 // Trim off the incorrect path 3135 full_so_path.erase(0, double_slash_pos + 1); 3136 } 3137 } 3138 } 3139 if (*full_so_path.rbegin() != '/') 3140 full_so_path += '/'; 3141 full_so_path += symbol_name; 3142 sym[sym_idx - 1].GetMangled().SetValue( 3143 ConstString(full_so_path.c_str()), false); 3144 add_nlist = false; 3145 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3146 } 3147 } else { 3148 // This could be a relative path to a N_SO 3149 N_SO_index = sym_idx; 3150 } 3151 } 3152 break; 3153 3154 case N_OSO: 3155 // object file name: name,,0,0,st_mtime 3156 type = eSymbolTypeObjectFile; 3157 break; 3158 3159 case N_LSYM: 3160 // local sym: name,,NO_SECT,type,offset 3161 type = eSymbolTypeLocal; 3162 break; 3163 3164 // INCL scopes 3165 case N_BINCL: 3166 // include file beginning: name,,NO_SECT,0,sum We use 3167 // the current number of symbols in the symbol table 3168 // in lieu of using nlist_idx in case we ever start 3169 // trimming entries out 3170 N_INCL_indexes.push_back(sym_idx); 3171 type = eSymbolTypeScopeBegin; 3172 break; 3173 3174 case N_EINCL: 3175 // include file end: name,,NO_SECT,0,0 3176 // Set the size of the N_BINCL to the terminating 3177 // index of this N_EINCL so that we can always skip 3178 // the entire symbol if we need to navigate more 3179 // quickly at the source level when parsing STABS 3180 if (!N_INCL_indexes.empty()) { 3181 symbol_ptr = 3182 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3183 symbol_ptr->SetByteSize(sym_idx + 1); 3184 symbol_ptr->SetSizeIsSibling(true); 3185 N_INCL_indexes.pop_back(); 3186 } 3187 type = eSymbolTypeScopeEnd; 3188 break; 3189 3190 case N_SOL: 3191 // #included file name: name,,n_sect,0,address 3192 type = eSymbolTypeHeaderFile; 3193 3194 // We currently don't use the header files on darwin 3195 add_nlist = false; 3196 break; 3197 3198 case N_PARAMS: 3199 // compiler parameters: name,,NO_SECT,0,0 3200 type = eSymbolTypeCompiler; 3201 break; 3202 3203 case N_VERSION: 3204 // compiler version: name,,NO_SECT,0,0 3205 type = eSymbolTypeCompiler; 3206 break; 3207 3208 case N_OLEVEL: 3209 // compiler -O level: name,,NO_SECT,0,0 3210 type = eSymbolTypeCompiler; 3211 break; 3212 3213 case N_PSYM: 3214 // parameter: name,,NO_SECT,type,offset 3215 type = eSymbolTypeVariable; 3216 break; 3217 3218 case N_ENTRY: 3219 // alternate entry: name,,n_sect,linenumber,address 3220 symbol_section = section_info.GetSection(nlist.n_sect, 3221 nlist.n_value); 3222 type = eSymbolTypeLineEntry; 3223 break; 3224 3225 // Left and Right Braces 3226 case N_LBRAC: 3227 // left bracket: 0,,NO_SECT,nesting level,address We 3228 // use the current number of symbols in the symbol 3229 // table in lieu of using nlist_idx in case we ever 3230 // start trimming entries out 3231 symbol_section = section_info.GetSection(nlist.n_sect, 3232 nlist.n_value); 3233 N_BRAC_indexes.push_back(sym_idx); 3234 type = eSymbolTypeScopeBegin; 3235 break; 3236 3237 case N_RBRAC: 3238 // right bracket: 0,,NO_SECT,nesting level,address 3239 // Set the size of the N_LBRAC to the terminating 3240 // index of this N_RBRAC so that we can always skip 3241 // the entire symbol if we need to navigate more 3242 // quickly at the source level when parsing STABS 3243 symbol_section = section_info.GetSection(nlist.n_sect, 3244 nlist.n_value); 3245 if (!N_BRAC_indexes.empty()) { 3246 symbol_ptr = 3247 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3248 symbol_ptr->SetByteSize(sym_idx + 1); 3249 symbol_ptr->SetSizeIsSibling(true); 3250 N_BRAC_indexes.pop_back(); 3251 } 3252 type = eSymbolTypeScopeEnd; 3253 break; 3254 3255 case N_EXCL: 3256 // deleted include file: name,,NO_SECT,0,sum 3257 type = eSymbolTypeHeaderFile; 3258 break; 3259 3260 // COMM scopes 3261 case N_BCOMM: 3262 // begin common: name,,NO_SECT,0,0 3263 // We use the current number of symbols in the symbol 3264 // table in lieu of using nlist_idx in case we ever 3265 // start trimming entries out 3266 type = eSymbolTypeScopeBegin; 3267 N_COMM_indexes.push_back(sym_idx); 3268 break; 3269 3270 case N_ECOML: 3271 // end common (local name): 0,,n_sect,0,address 3272 symbol_section = section_info.GetSection(nlist.n_sect, 3273 nlist.n_value); 3274 // Fall through 3275 3276 case N_ECOMM: 3277 // end common: name,,n_sect,0,0 3278 // Set the size of the N_BCOMM to the terminating 3279 // index of this N_ECOMM/N_ECOML so that we can 3280 // always skip the entire symbol if we need to 3281 // navigate more quickly at the source level when 3282 // parsing STABS 3283 if (!N_COMM_indexes.empty()) { 3284 symbol_ptr = 3285 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3286 symbol_ptr->SetByteSize(sym_idx + 1); 3287 symbol_ptr->SetSizeIsSibling(true); 3288 N_COMM_indexes.pop_back(); 3289 } 3290 type = eSymbolTypeScopeEnd; 3291 break; 3292 3293 case N_LENG: 3294 // second stab entry with length information 3295 type = eSymbolTypeAdditional; 3296 break; 3297 3298 default: 3299 break; 3300 } 3301 } else { 3302 // uint8_t n_pext = N_PEXT & nlist.n_type; 3303 uint8_t n_type = N_TYPE & nlist.n_type; 3304 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3305 3306 switch (n_type) { 3307 case N_INDR: { 3308 const char *reexport_name_cstr = 3309 strtab_data.PeekCStr(nlist.n_value); 3310 if (reexport_name_cstr && reexport_name_cstr[0]) { 3311 type = eSymbolTypeReExported; 3312 ConstString reexport_name( 3313 reexport_name_cstr + 3314 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3315 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3316 set_value = false; 3317 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3318 indirect_symbol_names.insert(ConstString( 3319 symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3320 } else 3321 type = eSymbolTypeUndefined; 3322 } break; 3323 3324 case N_UNDF: 3325 if (symbol_name && symbol_name[0]) { 3326 ConstString undefined_name( 3327 symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3328 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3329 } 3330 // Fall through 3331 case N_PBUD: 3332 type = eSymbolTypeUndefined; 3333 break; 3334 3335 case N_ABS: 3336 type = eSymbolTypeAbsolute; 3337 break; 3338 3339 case N_SECT: { 3340 symbol_section = section_info.GetSection(nlist.n_sect, 3341 nlist.n_value); 3342 3343 if (symbol_section == NULL) { 3344 // TODO: warn about this? 3345 add_nlist = false; 3346 break; 3347 } 3348 3349 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3350 type = eSymbolTypeException; 3351 } else { 3352 uint32_t section_type = 3353 symbol_section->Get() & SECTION_TYPE; 3354 3355 switch (section_type) { 3356 case S_CSTRING_LITERALS: 3357 type = eSymbolTypeData; 3358 break; // section with only literal C strings 3359 case S_4BYTE_LITERALS: 3360 type = eSymbolTypeData; 3361 break; // section with only 4 byte literals 3362 case S_8BYTE_LITERALS: 3363 type = eSymbolTypeData; 3364 break; // section with only 8 byte literals 3365 case S_LITERAL_POINTERS: 3366 type = eSymbolTypeTrampoline; 3367 break; // section with only pointers to literals 3368 case S_NON_LAZY_SYMBOL_POINTERS: 3369 type = eSymbolTypeTrampoline; 3370 break; // section with only non-lazy symbol 3371 // pointers 3372 case S_LAZY_SYMBOL_POINTERS: 3373 type = eSymbolTypeTrampoline; 3374 break; // section with only lazy symbol pointers 3375 case S_SYMBOL_STUBS: 3376 type = eSymbolTypeTrampoline; 3377 break; // section with only symbol stubs, byte 3378 // size of stub in the reserved2 field 3379 case S_MOD_INIT_FUNC_POINTERS: 3380 type = eSymbolTypeCode; 3381 break; // section with only function pointers for 3382 // initialization 3383 case S_MOD_TERM_FUNC_POINTERS: 3384 type = eSymbolTypeCode; 3385 break; // section with only function pointers for 3386 // termination 3387 case S_INTERPOSING: 3388 type = eSymbolTypeTrampoline; 3389 break; // section with only pairs of function 3390 // pointers for interposing 3391 case S_16BYTE_LITERALS: 3392 type = eSymbolTypeData; 3393 break; // section with only 16 byte literals 3394 case S_DTRACE_DOF: 3395 type = eSymbolTypeInstrumentation; 3396 break; 3397 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3398 type = eSymbolTypeTrampoline; 3399 break; 3400 default: 3401 switch (symbol_section->GetType()) { 3402 case lldb::eSectionTypeCode: 3403 type = eSymbolTypeCode; 3404 break; 3405 case eSectionTypeData: 3406 case eSectionTypeDataCString: // Inlined C string 3407 // data 3408 case eSectionTypeDataCStringPointers: // Pointers 3409 // to C 3410 // string 3411 // data 3412 case eSectionTypeDataSymbolAddress: // Address of 3413 // a symbol in 3414 // the symbol 3415 // table 3416 case eSectionTypeData4: 3417 case eSectionTypeData8: 3418 case eSectionTypeData16: 3419 type = eSymbolTypeData; 3420 break; 3421 default: 3422 break; 3423 } 3424 break; 3425 } 3426 3427 if (type == eSymbolTypeInvalid) { 3428 const char *symbol_sect_name = 3429 symbol_section->GetName().AsCString(); 3430 if (symbol_section->IsDescendant( 3431 text_section_sp.get())) { 3432 if (symbol_section->IsClear( 3433 S_ATTR_PURE_INSTRUCTIONS | 3434 S_ATTR_SELF_MODIFYING_CODE | 3435 S_ATTR_SOME_INSTRUCTIONS)) 3436 type = eSymbolTypeData; 3437 else 3438 type = eSymbolTypeCode; 3439 } else if (symbol_section->IsDescendant( 3440 data_section_sp.get()) || 3441 symbol_section->IsDescendant( 3442 data_dirty_section_sp.get()) || 3443 symbol_section->IsDescendant( 3444 data_const_section_sp.get())) { 3445 if (symbol_sect_name && 3446 ::strstr(symbol_sect_name, "__objc") == 3447 symbol_sect_name) { 3448 type = eSymbolTypeRuntime; 3449 3450 if (symbol_name) { 3451 llvm::StringRef symbol_name_ref(symbol_name); 3452 if (symbol_name_ref.startswith("_OBJC_")) { 3453 llvm::StringRef 3454 g_objc_v2_prefix_class( 3455 "_OBJC_CLASS_$_"); 3456 llvm::StringRef 3457 g_objc_v2_prefix_metaclass( 3458 "_OBJC_METACLASS_$_"); 3459 llvm::StringRef 3460 g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 3461 if (symbol_name_ref.startswith( 3462 g_objc_v2_prefix_class)) { 3463 symbol_name_non_abi_mangled = 3464 symbol_name + 1; 3465 symbol_name = 3466 symbol_name + 3467 g_objc_v2_prefix_class.size(); 3468 type = eSymbolTypeObjCClass; 3469 demangled_is_synthesized = true; 3470 } else if ( 3471 symbol_name_ref.startswith( 3472 g_objc_v2_prefix_metaclass)) { 3473 symbol_name_non_abi_mangled = 3474 symbol_name + 1; 3475 symbol_name = 3476 symbol_name + 3477 g_objc_v2_prefix_metaclass.size(); 3478 type = eSymbolTypeObjCMetaClass; 3479 demangled_is_synthesized = true; 3480 } else if (symbol_name_ref.startswith( 3481 g_objc_v2_prefix_ivar)) { 3482 symbol_name_non_abi_mangled = 3483 symbol_name + 1; 3484 symbol_name = 3485 symbol_name + 3486 g_objc_v2_prefix_ivar.size(); 3487 type = eSymbolTypeObjCIVar; 3488 demangled_is_synthesized = true; 3489 } 3490 } 3491 } 3492 } else if (symbol_sect_name && 3493 ::strstr(symbol_sect_name, 3494 "__gcc_except_tab") == 3495 symbol_sect_name) { 3496 type = eSymbolTypeException; 3497 } else { 3498 type = eSymbolTypeData; 3499 } 3500 } else if (symbol_sect_name && 3501 ::strstr(symbol_sect_name, "__IMPORT") == 3502 symbol_sect_name) { 3503 type = eSymbolTypeTrampoline; 3504 } else if (symbol_section->IsDescendant( 3505 objc_section_sp.get())) { 3506 type = eSymbolTypeRuntime; 3507 if (symbol_name && symbol_name[0] == '.') { 3508 llvm::StringRef symbol_name_ref(symbol_name); 3509 llvm::StringRef 3510 g_objc_v1_prefix_class(".objc_class_name_"); 3511 if (symbol_name_ref.startswith( 3512 g_objc_v1_prefix_class)) { 3513 symbol_name_non_abi_mangled = symbol_name; 3514 symbol_name = symbol_name + 3515 g_objc_v1_prefix_class.size(); 3516 type = eSymbolTypeObjCClass; 3517 demangled_is_synthesized = true; 3518 } 3519 } 3520 } 3521 } 3522 } 3523 } break; 3524 } 3525 } 3526 3527 if (add_nlist) { 3528 uint64_t symbol_value = nlist.n_value; 3529 if (symbol_name_non_abi_mangled) { 3530 sym[sym_idx].GetMangled().SetMangledName( 3531 ConstString(symbol_name_non_abi_mangled)); 3532 sym[sym_idx].GetMangled().SetDemangledName( 3533 ConstString(symbol_name)); 3534 } else { 3535 bool symbol_name_is_mangled = false; 3536 3537 if (symbol_name && symbol_name[0] == '_') { 3538 symbol_name_is_mangled = symbol_name[1] == '_'; 3539 symbol_name++; // Skip the leading underscore 3540 } 3541 3542 if (symbol_name) { 3543 ConstString const_symbol_name(symbol_name); 3544 sym[sym_idx].GetMangled().SetValue( 3545 const_symbol_name, symbol_name_is_mangled); 3546 if (is_gsym && is_debug) { 3547 const char *gsym_name = 3548 sym[sym_idx] 3549 .GetMangled() 3550 .GetName(Mangled::ePreferMangled) 3551 .GetCString(); 3552 if (gsym_name) 3553 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3554 } 3555 } 3556 } 3557 if (symbol_section) { 3558 const addr_t section_file_addr = 3559 symbol_section->GetFileAddress(); 3560 if (symbol_byte_size == 0 && 3561 function_starts_count > 0) { 3562 addr_t symbol_lookup_file_addr = nlist.n_value; 3563 // Do an exact address match for non-ARM addresses, 3564 // else get the closest since the symbol might be a 3565 // thumb symbol which has an address with bit zero 3566 // set 3567 FunctionStarts::Entry *func_start_entry = 3568 function_starts.FindEntry(symbol_lookup_file_addr, 3569 !is_arm); 3570 if (is_arm && func_start_entry) { 3571 // Verify that the function start address is the 3572 // symbol address (ARM) or the symbol address + 1 3573 // (thumb) 3574 if (func_start_entry->addr != 3575 symbol_lookup_file_addr && 3576 func_start_entry->addr != 3577 (symbol_lookup_file_addr + 1)) { 3578 // Not the right entry, NULL it out... 3579 func_start_entry = NULL; 3580 } 3581 } 3582 if (func_start_entry) { 3583 func_start_entry->data = true; 3584 3585 addr_t symbol_file_addr = func_start_entry->addr; 3586 uint32_t symbol_flags = 0; 3587 if (is_arm) { 3588 if (symbol_file_addr & 1) 3589 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3590 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3591 } 3592 3593 const FunctionStarts::Entry *next_func_start_entry = 3594 function_starts.FindNextEntry(func_start_entry); 3595 const addr_t section_end_file_addr = 3596 section_file_addr + 3597 symbol_section->GetByteSize(); 3598 if (next_func_start_entry) { 3599 addr_t next_symbol_file_addr = 3600 next_func_start_entry->addr; 3601 // Be sure the clear the Thumb address bit when 3602 // we calculate the size from the current and 3603 // next address 3604 if (is_arm) 3605 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3606 symbol_byte_size = std::min<lldb::addr_t>( 3607 next_symbol_file_addr - symbol_file_addr, 3608 section_end_file_addr - symbol_file_addr); 3609 } else { 3610 symbol_byte_size = 3611 section_end_file_addr - symbol_file_addr; 3612 } 3613 } 3614 } 3615 symbol_value -= section_file_addr; 3616 } 3617 3618 if (is_debug == false) { 3619 if (type == eSymbolTypeCode) { 3620 // See if we can find a N_FUN entry for any code 3621 // symbols. If we do find a match, and the name 3622 // matches, then we can merge the two into just the 3623 // function symbol to avoid duplicate entries in 3624 // the symbol table 3625 auto range = 3626 N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3627 if (range.first != range.second) { 3628 bool found_it = false; 3629 for (auto pos = range.first; pos != range.second; 3630 ++pos) { 3631 if (sym[sym_idx].GetMangled().GetName( 3632 Mangled::ePreferMangled) == 3633 sym[pos->second].GetMangled().GetName( 3634 Mangled::ePreferMangled)) { 3635 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3636 // We just need the flags from the linker 3637 // symbol, so put these flags 3638 // into the N_FUN flags to avoid duplicate 3639 // symbols in the symbol table 3640 sym[pos->second].SetExternal( 3641 sym[sym_idx].IsExternal()); 3642 sym[pos->second].SetFlags(nlist.n_type << 16 | 3643 nlist.n_desc); 3644 if (resolver_addresses.find(nlist.n_value) != 3645 resolver_addresses.end()) 3646 sym[pos->second].SetType(eSymbolTypeResolver); 3647 sym[sym_idx].Clear(); 3648 found_it = true; 3649 break; 3650 } 3651 } 3652 if (found_it) 3653 continue; 3654 } else { 3655 if (resolver_addresses.find(nlist.n_value) != 3656 resolver_addresses.end()) 3657 type = eSymbolTypeResolver; 3658 } 3659 } else if (type == eSymbolTypeData || 3660 type == eSymbolTypeObjCClass || 3661 type == eSymbolTypeObjCMetaClass || 3662 type == eSymbolTypeObjCIVar) { 3663 // See if we can find a N_STSYM entry for any data 3664 // symbols. If we do find a match, and the name 3665 // matches, then we can merge the two into just the 3666 // Static symbol to avoid duplicate entries in the 3667 // symbol table 3668 auto range = N_STSYM_addr_to_sym_idx.equal_range( 3669 nlist.n_value); 3670 if (range.first != range.second) { 3671 bool found_it = false; 3672 for (auto pos = range.first; pos != range.second; 3673 ++pos) { 3674 if (sym[sym_idx].GetMangled().GetName( 3675 Mangled::ePreferMangled) == 3676 sym[pos->second].GetMangled().GetName( 3677 Mangled::ePreferMangled)) { 3678 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3679 // We just need the flags from the linker 3680 // symbol, so put these flags 3681 // into the N_STSYM flags to avoid duplicate 3682 // symbols in the symbol table 3683 sym[pos->second].SetExternal( 3684 sym[sym_idx].IsExternal()); 3685 sym[pos->second].SetFlags(nlist.n_type << 16 | 3686 nlist.n_desc); 3687 sym[sym_idx].Clear(); 3688 found_it = true; 3689 break; 3690 } 3691 } 3692 if (found_it) 3693 continue; 3694 } else { 3695 const char *gsym_name = 3696 sym[sym_idx] 3697 .GetMangled() 3698 .GetName(Mangled::ePreferMangled) 3699 .GetCString(); 3700 if (gsym_name) { 3701 // Combine N_GSYM stab entries with the non 3702 // stab symbol 3703 ConstNameToSymbolIndexMap::const_iterator pos = 3704 N_GSYM_name_to_sym_idx.find(gsym_name); 3705 if (pos != N_GSYM_name_to_sym_idx.end()) { 3706 const uint32_t GSYM_sym_idx = pos->second; 3707 m_nlist_idx_to_sym_idx[nlist_idx] = 3708 GSYM_sym_idx; 3709 // Copy the address, because often the N_GSYM 3710 // address has an invalid address of zero 3711 // when the global is a common symbol 3712 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3713 symbol_section); 3714 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3715 symbol_value); 3716 add_symbol_addr(sym[GSYM_sym_idx] 3717 .GetAddress() 3718 .GetFileAddress()); 3719 // We just need the flags from the linker 3720 // symbol, so put these flags 3721 // into the N_GSYM flags to avoid duplicate 3722 // symbols in the symbol table 3723 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 3724 nlist.n_desc); 3725 sym[sym_idx].Clear(); 3726 continue; 3727 } 3728 } 3729 } 3730 } 3731 } 3732 3733 sym[sym_idx].SetID(nlist_idx); 3734 sym[sym_idx].SetType(type); 3735 if (set_value) { 3736 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 3737 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3738 add_symbol_addr( 3739 sym[sym_idx].GetAddress().GetFileAddress()); 3740 } 3741 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 3742 3743 if (symbol_byte_size > 0) 3744 sym[sym_idx].SetByteSize(symbol_byte_size); 3745 3746 if (demangled_is_synthesized) 3747 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3748 ++sym_idx; 3749 } else { 3750 sym[sym_idx].Clear(); 3751 } 3752 } 3753 ///////////////////////////// 3754 } 3755 break; // No more entries to consider 3756 } 3757 } 3758 3759 for (const auto &pos : reexport_shlib_needs_fixup) { 3760 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3761 if (undef_pos != undefined_name_to_desc.end()) { 3762 const uint8_t dylib_ordinal = 3763 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3764 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3765 sym[pos.first].SetReExportedSymbolSharedLibrary( 3766 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3767 } 3768 } 3769 } 3770 } 3771 } 3772 } 3773 } 3774 3775 // Must reset this in case it was mutated above! 3776 nlist_data_offset = 0; 3777 #endif 3778 3779 if (nlist_data.GetByteSize() > 0) { 3780 3781 // If the sym array was not created while parsing the DSC unmapped 3782 // symbols, create it now. 3783 if (sym == nullptr) { 3784 sym = 3785 symtab->Resize(symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3786 num_syms = symtab->GetNumSymbols(); 3787 } 3788 3789 if (unmapped_local_symbols_found) { 3790 assert(m_dysymtab.ilocalsym == 0); 3791 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3792 nlist_idx = m_dysymtab.nlocalsym; 3793 } else { 3794 nlist_idx = 0; 3795 } 3796 3797 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 3798 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 3799 UndefinedNameToDescMap undefined_name_to_desc; 3800 SymbolIndexToName reexport_shlib_needs_fixup; 3801 3802 // Symtab parsing is a huge mess. Everything is entangled and the code 3803 // requires access to a ridiculous amount of variables. LLDB depends 3804 // heavily on the proper merging of symbols and to get that right we need 3805 // to make sure we have parsed all the debug symbols first. Therefore we 3806 // invoke the lambda twice, once to parse only the debug symbols and then 3807 // once more to parse the remaining symbols. 3808 auto ParseSymbolLambda = [&](struct nlist_64 &nlist, uint32_t nlist_idx, 3809 bool debug_only) { 3810 const bool is_debug = ((nlist.n_type & N_STAB) != 0); 3811 if (is_debug != debug_only) 3812 return true; 3813 3814 const char *symbol_name_non_abi_mangled = nullptr; 3815 const char *symbol_name = nullptr; 3816 3817 if (have_strtab_data) { 3818 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3819 3820 if (symbol_name == nullptr) { 3821 // No symbol should be NULL, even the symbols with no string values 3822 // should have an offset zero which points to an empty C-string 3823 Host::SystemLog(Host::eSystemLogError, 3824 "error: symbol[%u] has invalid string table offset " 3825 "0x%x in %s, ignoring symbol\n", 3826 nlist_idx, nlist.n_strx, 3827 module_sp->GetFileSpec().GetPath().c_str()); 3828 return true; 3829 } 3830 if (symbol_name[0] == '\0') 3831 symbol_name = nullptr; 3832 } else { 3833 const addr_t str_addr = strtab_addr + nlist.n_strx; 3834 Status str_error; 3835 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3836 str_error)) 3837 symbol_name = memory_symbol_name.c_str(); 3838 } 3839 3840 SymbolType type = eSymbolTypeInvalid; 3841 SectionSP symbol_section; 3842 lldb::addr_t symbol_byte_size = 0; 3843 bool add_nlist = true; 3844 bool is_gsym = false; 3845 bool demangled_is_synthesized = false; 3846 bool set_value = true; 3847 3848 assert(sym_idx < num_syms); 3849 sym[sym_idx].SetDebug(is_debug); 3850 3851 if (is_debug) { 3852 switch (nlist.n_type) { 3853 case N_GSYM: 3854 // global symbol: name,,NO_SECT,type,0 3855 // Sometimes the N_GSYM value contains the address. 3856 3857 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3858 // the ObjC data. They 3859 // have the same address, but we want to ensure that we always find 3860 // only the real symbol, 'cause we don't currently correctly 3861 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3862 // type. This is a temporary hack to make sure the ObjectiveC 3863 // symbols get treated correctly. To do this right, we should 3864 // coalesce all the GSYM & global symbols that have the same 3865 // address. 3866 is_gsym = true; 3867 sym[sym_idx].SetExternal(true); 3868 3869 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3870 llvm::StringRef symbol_name_ref(symbol_name); 3871 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3872 symbol_name_non_abi_mangled = symbol_name + 1; 3873 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3874 type = eSymbolTypeObjCClass; 3875 demangled_is_synthesized = true; 3876 3877 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 3878 symbol_name_non_abi_mangled = symbol_name + 1; 3879 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3880 type = eSymbolTypeObjCMetaClass; 3881 demangled_is_synthesized = true; 3882 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3883 symbol_name_non_abi_mangled = symbol_name + 1; 3884 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3885 type = eSymbolTypeObjCIVar; 3886 demangled_is_synthesized = true; 3887 } 3888 } else { 3889 if (nlist.n_value != 0) 3890 symbol_section = 3891 section_info.GetSection(nlist.n_sect, nlist.n_value); 3892 type = eSymbolTypeData; 3893 } 3894 break; 3895 3896 case N_FNAME: 3897 // procedure name (f77 kludge): name,,NO_SECT,0,0 3898 type = eSymbolTypeCompiler; 3899 break; 3900 3901 case N_FUN: 3902 // procedure: name,,n_sect,linenumber,address 3903 if (symbol_name) { 3904 type = eSymbolTypeCode; 3905 symbol_section = 3906 section_info.GetSection(nlist.n_sect, nlist.n_value); 3907 3908 N_FUN_addr_to_sym_idx.insert( 3909 std::make_pair(nlist.n_value, sym_idx)); 3910 // We use the current number of symbols in the symbol table in 3911 // lieu of using nlist_idx in case we ever start trimming entries 3912 // out 3913 N_FUN_indexes.push_back(sym_idx); 3914 } else { 3915 type = eSymbolTypeCompiler; 3916 3917 if (!N_FUN_indexes.empty()) { 3918 // Copy the size of the function into the original STAB entry 3919 // so we don't have to hunt for it later 3920 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3921 ->SetByteSize(nlist.n_value); 3922 N_FUN_indexes.pop_back(); 3923 // We don't really need the end function STAB as it contains 3924 // the size which we already placed with the original symbol, 3925 // so don't add it if we want a minimal symbol table 3926 add_nlist = false; 3927 } 3928 } 3929 break; 3930 3931 case N_STSYM: 3932 // static symbol: name,,n_sect,type,address 3933 N_STSYM_addr_to_sym_idx.insert( 3934 std::make_pair(nlist.n_value, sym_idx)); 3935 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3936 if (symbol_name && symbol_name[0]) { 3937 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3938 eSymbolTypeData); 3939 } 3940 break; 3941 3942 case N_LCSYM: 3943 // .lcomm symbol: name,,n_sect,type,address 3944 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3945 type = eSymbolTypeCommonBlock; 3946 break; 3947 3948 case N_BNSYM: 3949 // We use the current number of symbols in the symbol table in lieu 3950 // of using nlist_idx in case we ever start trimming entries out 3951 // Skip these if we want minimal symbol tables 3952 add_nlist = false; 3953 break; 3954 3955 case N_ENSYM: 3956 // Set the size of the N_BNSYM to the terminating index of this 3957 // N_ENSYM so that we can always skip the entire symbol if we need 3958 // to navigate more quickly at the source level when parsing STABS 3959 // Skip these if we want minimal symbol tables 3960 add_nlist = false; 3961 break; 3962 3963 case N_OPT: 3964 // emitted with gcc2_compiled and in gcc source 3965 type = eSymbolTypeCompiler; 3966 break; 3967 3968 case N_RSYM: 3969 // register sym: name,,NO_SECT,type,register 3970 type = eSymbolTypeVariable; 3971 break; 3972 3973 case N_SLINE: 3974 // src line: 0,,n_sect,linenumber,address 3975 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3976 type = eSymbolTypeLineEntry; 3977 break; 3978 3979 case N_SSYM: 3980 // structure elt: name,,NO_SECT,type,struct_offset 3981 type = eSymbolTypeVariableType; 3982 break; 3983 3984 case N_SO: 3985 // source file name 3986 type = eSymbolTypeSourceFile; 3987 if (symbol_name == nullptr) { 3988 add_nlist = false; 3989 if (N_SO_index != UINT32_MAX) { 3990 // Set the size of the N_SO to the terminating index of this 3991 // N_SO so that we can always skip the entire N_SO if we need 3992 // to navigate more quickly at the source level when parsing 3993 // STABS 3994 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3995 symbol_ptr->SetByteSize(sym_idx); 3996 symbol_ptr->SetSizeIsSibling(true); 3997 } 3998 N_NSYM_indexes.clear(); 3999 N_INCL_indexes.clear(); 4000 N_BRAC_indexes.clear(); 4001 N_COMM_indexes.clear(); 4002 N_FUN_indexes.clear(); 4003 N_SO_index = UINT32_MAX; 4004 } else { 4005 // We use the current number of symbols in the symbol table in 4006 // lieu of using nlist_idx in case we ever start trimming entries 4007 // out 4008 const bool N_SO_has_full_path = symbol_name[0] == '/'; 4009 if (N_SO_has_full_path) { 4010 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 4011 // We have two consecutive N_SO entries where the first 4012 // contains a directory and the second contains a full path. 4013 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), 4014 false); 4015 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4016 add_nlist = false; 4017 } else { 4018 // This is the first entry in a N_SO that contains a 4019 // directory or a full path to the source file 4020 N_SO_index = sym_idx; 4021 } 4022 } else if ((N_SO_index == sym_idx - 1) && 4023 ((sym_idx - 1) < num_syms)) { 4024 // This is usually the second N_SO entry that contains just the 4025 // filename, so here we combine it with the first one if we are 4026 // minimizing the symbol table 4027 const char *so_path = 4028 sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 4029 if (so_path && so_path[0]) { 4030 std::string full_so_path(so_path); 4031 const size_t double_slash_pos = full_so_path.find("//"); 4032 if (double_slash_pos != std::string::npos) { 4033 // The linker has been generating bad N_SO entries with 4034 // doubled up paths in the format "%s%s" where the first 4035 // string in the DW_AT_comp_dir, and the second is the 4036 // directory for the source file so you end up with a path 4037 // that looks like "/tmp/src//tmp/src/" 4038 FileSpec so_dir(so_path); 4039 if (!FileSystem::Instance().Exists(so_dir)) { 4040 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 4041 FileSpec::Style::native); 4042 if (FileSystem::Instance().Exists(so_dir)) { 4043 // Trim off the incorrect path 4044 full_so_path.erase(0, double_slash_pos + 1); 4045 } 4046 } 4047 } 4048 if (*full_so_path.rbegin() != '/') 4049 full_so_path += '/'; 4050 full_so_path += symbol_name; 4051 sym[sym_idx - 1].GetMangled().SetValue( 4052 ConstString(full_so_path.c_str()), false); 4053 add_nlist = false; 4054 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4055 } 4056 } else { 4057 // This could be a relative path to a N_SO 4058 N_SO_index = sym_idx; 4059 } 4060 } 4061 break; 4062 4063 case N_OSO: 4064 // object file name: name,,0,0,st_mtime 4065 type = eSymbolTypeObjectFile; 4066 break; 4067 4068 case N_LSYM: 4069 // local sym: name,,NO_SECT,type,offset 4070 type = eSymbolTypeLocal; 4071 break; 4072 4073 // INCL scopes 4074 case N_BINCL: 4075 // include file beginning: name,,NO_SECT,0,sum We use the current 4076 // number of symbols in the symbol table in lieu of using nlist_idx 4077 // in case we ever start trimming entries out 4078 N_INCL_indexes.push_back(sym_idx); 4079 type = eSymbolTypeScopeBegin; 4080 break; 4081 4082 case N_EINCL: 4083 // include file end: name,,NO_SECT,0,0 4084 // Set the size of the N_BINCL to the terminating index of this 4085 // N_EINCL so that we can always skip the entire symbol if we need 4086 // to navigate more quickly at the source level when parsing STABS 4087 if (!N_INCL_indexes.empty()) { 4088 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4089 symbol_ptr->SetByteSize(sym_idx + 1); 4090 symbol_ptr->SetSizeIsSibling(true); 4091 N_INCL_indexes.pop_back(); 4092 } 4093 type = eSymbolTypeScopeEnd; 4094 break; 4095 4096 case N_SOL: 4097 // #included file name: name,,n_sect,0,address 4098 type = eSymbolTypeHeaderFile; 4099 4100 // We currently don't use the header files on darwin 4101 add_nlist = false; 4102 break; 4103 4104 case N_PARAMS: 4105 // compiler parameters: name,,NO_SECT,0,0 4106 type = eSymbolTypeCompiler; 4107 break; 4108 4109 case N_VERSION: 4110 // compiler version: name,,NO_SECT,0,0 4111 type = eSymbolTypeCompiler; 4112 break; 4113 4114 case N_OLEVEL: 4115 // compiler -O level: name,,NO_SECT,0,0 4116 type = eSymbolTypeCompiler; 4117 break; 4118 4119 case N_PSYM: 4120 // parameter: name,,NO_SECT,type,offset 4121 type = eSymbolTypeVariable; 4122 break; 4123 4124 case N_ENTRY: 4125 // alternate entry: name,,n_sect,linenumber,address 4126 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4127 type = eSymbolTypeLineEntry; 4128 break; 4129 4130 // Left and Right Braces 4131 case N_LBRAC: 4132 // left bracket: 0,,NO_SECT,nesting level,address We use the 4133 // current number of symbols in the symbol table in lieu of using 4134 // nlist_idx in case we ever start trimming entries out 4135 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4136 N_BRAC_indexes.push_back(sym_idx); 4137 type = eSymbolTypeScopeBegin; 4138 break; 4139 4140 case N_RBRAC: 4141 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4142 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4143 // can always skip the entire symbol if we need to navigate more 4144 // quickly at the source level when parsing STABS 4145 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4146 if (!N_BRAC_indexes.empty()) { 4147 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4148 symbol_ptr->SetByteSize(sym_idx + 1); 4149 symbol_ptr->SetSizeIsSibling(true); 4150 N_BRAC_indexes.pop_back(); 4151 } 4152 type = eSymbolTypeScopeEnd; 4153 break; 4154 4155 case N_EXCL: 4156 // deleted include file: name,,NO_SECT,0,sum 4157 type = eSymbolTypeHeaderFile; 4158 break; 4159 4160 // COMM scopes 4161 case N_BCOMM: 4162 // begin common: name,,NO_SECT,0,0 4163 // We use the current number of symbols in the symbol table in lieu 4164 // of using nlist_idx in case we ever start trimming entries out 4165 type = eSymbolTypeScopeBegin; 4166 N_COMM_indexes.push_back(sym_idx); 4167 break; 4168 4169 case N_ECOML: 4170 // end common (local name): 0,,n_sect,0,address 4171 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4172 LLVM_FALLTHROUGH; 4173 4174 case N_ECOMM: 4175 // end common: name,,n_sect,0,0 4176 // Set the size of the N_BCOMM to the terminating index of this 4177 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4178 // we need to navigate more quickly at the source level when 4179 // parsing STABS 4180 if (!N_COMM_indexes.empty()) { 4181 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4182 symbol_ptr->SetByteSize(sym_idx + 1); 4183 symbol_ptr->SetSizeIsSibling(true); 4184 N_COMM_indexes.pop_back(); 4185 } 4186 type = eSymbolTypeScopeEnd; 4187 break; 4188 4189 case N_LENG: 4190 // second stab entry with length information 4191 type = eSymbolTypeAdditional; 4192 break; 4193 4194 default: 4195 break; 4196 } 4197 } else { 4198 uint8_t n_type = N_TYPE & nlist.n_type; 4199 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4200 4201 switch (n_type) { 4202 case N_INDR: { 4203 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4204 if (reexport_name_cstr && reexport_name_cstr[0]) { 4205 type = eSymbolTypeReExported; 4206 ConstString reexport_name(reexport_name_cstr + 4207 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4208 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4209 set_value = false; 4210 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4211 indirect_symbol_names.insert( 4212 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4213 } else 4214 type = eSymbolTypeUndefined; 4215 } break; 4216 4217 case N_UNDF: 4218 if (symbol_name && symbol_name[0]) { 4219 ConstString undefined_name(symbol_name + 4220 ((symbol_name[0] == '_') ? 1 : 0)); 4221 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4222 } 4223 LLVM_FALLTHROUGH; 4224 4225 case N_PBUD: 4226 type = eSymbolTypeUndefined; 4227 break; 4228 4229 case N_ABS: 4230 type = eSymbolTypeAbsolute; 4231 break; 4232 4233 case N_SECT: { 4234 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4235 4236 if (!symbol_section) { 4237 // TODO: warn about this? 4238 add_nlist = false; 4239 break; 4240 } 4241 4242 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4243 type = eSymbolTypeException; 4244 } else { 4245 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4246 4247 switch (section_type) { 4248 case S_CSTRING_LITERALS: 4249 type = eSymbolTypeData; 4250 break; // section with only literal C strings 4251 case S_4BYTE_LITERALS: 4252 type = eSymbolTypeData; 4253 break; // section with only 4 byte literals 4254 case S_8BYTE_LITERALS: 4255 type = eSymbolTypeData; 4256 break; // section with only 8 byte literals 4257 case S_LITERAL_POINTERS: 4258 type = eSymbolTypeTrampoline; 4259 break; // section with only pointers to literals 4260 case S_NON_LAZY_SYMBOL_POINTERS: 4261 type = eSymbolTypeTrampoline; 4262 break; // section with only non-lazy symbol pointers 4263 case S_LAZY_SYMBOL_POINTERS: 4264 type = eSymbolTypeTrampoline; 4265 break; // section with only lazy symbol pointers 4266 case S_SYMBOL_STUBS: 4267 type = eSymbolTypeTrampoline; 4268 break; // section with only symbol stubs, byte size of stub in 4269 // the reserved2 field 4270 case S_MOD_INIT_FUNC_POINTERS: 4271 type = eSymbolTypeCode; 4272 break; // section with only function pointers for initialization 4273 case S_MOD_TERM_FUNC_POINTERS: 4274 type = eSymbolTypeCode; 4275 break; // section with only function pointers for termination 4276 case S_INTERPOSING: 4277 type = eSymbolTypeTrampoline; 4278 break; // section with only pairs of function pointers for 4279 // interposing 4280 case S_16BYTE_LITERALS: 4281 type = eSymbolTypeData; 4282 break; // section with only 16 byte literals 4283 case S_DTRACE_DOF: 4284 type = eSymbolTypeInstrumentation; 4285 break; 4286 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4287 type = eSymbolTypeTrampoline; 4288 break; 4289 default: 4290 switch (symbol_section->GetType()) { 4291 case lldb::eSectionTypeCode: 4292 type = eSymbolTypeCode; 4293 break; 4294 case eSectionTypeData: 4295 case eSectionTypeDataCString: // Inlined C string data 4296 case eSectionTypeDataCStringPointers: // Pointers to C string 4297 // data 4298 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4299 // the symbol table 4300 case eSectionTypeData4: 4301 case eSectionTypeData8: 4302 case eSectionTypeData16: 4303 type = eSymbolTypeData; 4304 break; 4305 default: 4306 break; 4307 } 4308 break; 4309 } 4310 4311 if (type == eSymbolTypeInvalid) { 4312 const char *symbol_sect_name = 4313 symbol_section->GetName().AsCString(); 4314 if (symbol_section->IsDescendant(text_section_sp.get())) { 4315 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4316 S_ATTR_SELF_MODIFYING_CODE | 4317 S_ATTR_SOME_INSTRUCTIONS)) 4318 type = eSymbolTypeData; 4319 else 4320 type = eSymbolTypeCode; 4321 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 4322 symbol_section->IsDescendant( 4323 data_dirty_section_sp.get()) || 4324 symbol_section->IsDescendant( 4325 data_const_section_sp.get())) { 4326 if (symbol_sect_name && 4327 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 4328 type = eSymbolTypeRuntime; 4329 4330 if (symbol_name) { 4331 llvm::StringRef symbol_name_ref(symbol_name); 4332 if (symbol_name_ref.startswith("_OBJC_")) { 4333 llvm::StringRef g_objc_v2_prefix_class( 4334 "_OBJC_CLASS_$_"); 4335 llvm::StringRef g_objc_v2_prefix_metaclass( 4336 "_OBJC_METACLASS_$_"); 4337 llvm::StringRef g_objc_v2_prefix_ivar( 4338 "_OBJC_IVAR_$_"); 4339 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 4340 symbol_name_non_abi_mangled = symbol_name + 1; 4341 symbol_name = 4342 symbol_name + g_objc_v2_prefix_class.size(); 4343 type = eSymbolTypeObjCClass; 4344 demangled_is_synthesized = true; 4345 } else if (symbol_name_ref.startswith( 4346 g_objc_v2_prefix_metaclass)) { 4347 symbol_name_non_abi_mangled = symbol_name + 1; 4348 symbol_name = 4349 symbol_name + g_objc_v2_prefix_metaclass.size(); 4350 type = eSymbolTypeObjCMetaClass; 4351 demangled_is_synthesized = true; 4352 } else if (symbol_name_ref.startswith( 4353 g_objc_v2_prefix_ivar)) { 4354 symbol_name_non_abi_mangled = symbol_name + 1; 4355 symbol_name = 4356 symbol_name + g_objc_v2_prefix_ivar.size(); 4357 type = eSymbolTypeObjCIVar; 4358 demangled_is_synthesized = true; 4359 } 4360 } 4361 } 4362 } else if (symbol_sect_name && 4363 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4364 symbol_sect_name) { 4365 type = eSymbolTypeException; 4366 } else { 4367 type = eSymbolTypeData; 4368 } 4369 } else if (symbol_sect_name && 4370 ::strstr(symbol_sect_name, "__IMPORT") == 4371 symbol_sect_name) { 4372 type = eSymbolTypeTrampoline; 4373 } else if (symbol_section->IsDescendant(objc_section_sp.get())) { 4374 type = eSymbolTypeRuntime; 4375 if (symbol_name && symbol_name[0] == '.') { 4376 llvm::StringRef symbol_name_ref(symbol_name); 4377 llvm::StringRef g_objc_v1_prefix_class( 4378 ".objc_class_name_"); 4379 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4380 symbol_name_non_abi_mangled = symbol_name; 4381 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4382 type = eSymbolTypeObjCClass; 4383 demangled_is_synthesized = true; 4384 } 4385 } 4386 } 4387 } 4388 } 4389 } break; 4390 } 4391 } 4392 4393 if (!add_nlist) { 4394 sym[sym_idx].Clear(); 4395 return true; 4396 } 4397 4398 uint64_t symbol_value = nlist.n_value; 4399 4400 if (symbol_name_non_abi_mangled) { 4401 sym[sym_idx].GetMangled().SetMangledName( 4402 ConstString(symbol_name_non_abi_mangled)); 4403 sym[sym_idx].GetMangled().SetDemangledName(ConstString(symbol_name)); 4404 } else { 4405 bool symbol_name_is_mangled = false; 4406 4407 if (symbol_name && symbol_name[0] == '_') { 4408 symbol_name_is_mangled = symbol_name[1] == '_'; 4409 symbol_name++; // Skip the leading underscore 4410 } 4411 4412 if (symbol_name) { 4413 ConstString const_symbol_name(symbol_name); 4414 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4415 symbol_name_is_mangled); 4416 } 4417 } 4418 4419 if (is_gsym) { 4420 const char *gsym_name = sym[sym_idx] 4421 .GetMangled() 4422 .GetName(Mangled::ePreferMangled) 4423 .GetCString(); 4424 if (gsym_name) 4425 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4426 } 4427 4428 if (symbol_section) { 4429 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4430 if (symbol_byte_size == 0 && function_starts_count > 0) { 4431 addr_t symbol_lookup_file_addr = nlist.n_value; 4432 // Do an exact address match for non-ARM addresses, else get the 4433 // closest since the symbol might be a thumb symbol which has an 4434 // address with bit zero set. 4435 FunctionStarts::Entry *func_start_entry = 4436 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4437 if (is_arm && func_start_entry) { 4438 // Verify that the function start address is the symbol address 4439 // (ARM) or the symbol address + 1 (thumb). 4440 if (func_start_entry->addr != symbol_lookup_file_addr && 4441 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4442 // Not the right entry, NULL it out... 4443 func_start_entry = nullptr; 4444 } 4445 } 4446 if (func_start_entry) { 4447 func_start_entry->data = true; 4448 4449 addr_t symbol_file_addr = func_start_entry->addr; 4450 if (is_arm) 4451 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4452 4453 const FunctionStarts::Entry *next_func_start_entry = 4454 function_starts.FindNextEntry(func_start_entry); 4455 const addr_t section_end_file_addr = 4456 section_file_addr + symbol_section->GetByteSize(); 4457 if (next_func_start_entry) { 4458 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4459 // Be sure the clear the Thumb address bit when we calculate the 4460 // size from the current and next address 4461 if (is_arm) 4462 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4463 symbol_byte_size = std::min<lldb::addr_t>( 4464 next_symbol_file_addr - symbol_file_addr, 4465 section_end_file_addr - symbol_file_addr); 4466 } else { 4467 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4468 } 4469 } 4470 } 4471 symbol_value -= section_file_addr; 4472 } 4473 4474 if (!is_debug) { 4475 if (type == eSymbolTypeCode) { 4476 // See if we can find a N_FUN entry for any code symbols. If we do 4477 // find a match, and the name matches, then we can merge the two into 4478 // just the function symbol to avoid duplicate entries in the symbol 4479 // table. 4480 std::pair<ValueToSymbolIndexMap::const_iterator, 4481 ValueToSymbolIndexMap::const_iterator> 4482 range; 4483 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4484 if (range.first != range.second) { 4485 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4486 pos != range.second; ++pos) { 4487 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4488 sym[pos->second].GetMangled().GetName( 4489 Mangled::ePreferMangled)) { 4490 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4491 // We just need the flags from the linker symbol, so put these 4492 // flags into the N_FUN flags to avoid duplicate symbols in the 4493 // symbol table. 4494 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4495 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4496 if (resolver_addresses.find(nlist.n_value) != 4497 resolver_addresses.end()) 4498 sym[pos->second].SetType(eSymbolTypeResolver); 4499 sym[sym_idx].Clear(); 4500 return true; 4501 } 4502 } 4503 } else { 4504 if (resolver_addresses.find(nlist.n_value) != 4505 resolver_addresses.end()) 4506 type = eSymbolTypeResolver; 4507 } 4508 } else if (type == eSymbolTypeData || type == eSymbolTypeObjCClass || 4509 type == eSymbolTypeObjCMetaClass || 4510 type == eSymbolTypeObjCIVar) { 4511 // See if we can find a N_STSYM entry for any data symbols. If we do 4512 // find a match, and the name matches, then we can merge the two into 4513 // just the Static symbol to avoid duplicate entries in the symbol 4514 // table. 4515 std::pair<ValueToSymbolIndexMap::const_iterator, 4516 ValueToSymbolIndexMap::const_iterator> 4517 range; 4518 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4519 if (range.first != range.second) { 4520 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4521 pos != range.second; ++pos) { 4522 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4523 sym[pos->second].GetMangled().GetName( 4524 Mangled::ePreferMangled)) { 4525 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4526 // We just need the flags from the linker symbol, so put these 4527 // flags into the N_STSYM flags to avoid duplicate symbols in 4528 // the symbol table. 4529 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4530 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4531 sym[sym_idx].Clear(); 4532 return true; 4533 } 4534 } 4535 } else { 4536 // Combine N_GSYM stab entries with the non stab symbol. 4537 const char *gsym_name = sym[sym_idx] 4538 .GetMangled() 4539 .GetName(Mangled::ePreferMangled) 4540 .GetCString(); 4541 if (gsym_name) { 4542 ConstNameToSymbolIndexMap::const_iterator pos = 4543 N_GSYM_name_to_sym_idx.find(gsym_name); 4544 if (pos != N_GSYM_name_to_sym_idx.end()) { 4545 const uint32_t GSYM_sym_idx = pos->second; 4546 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4547 // Copy the address, because often the N_GSYM address has an 4548 // invalid address of zero when the global is a common symbol. 4549 sym[GSYM_sym_idx].GetAddressRef().SetSection(symbol_section); 4550 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4551 add_symbol_addr( 4552 sym[GSYM_sym_idx].GetAddress().GetFileAddress()); 4553 // We just need the flags from the linker symbol, so put these 4554 // flags into the N_GSYM flags to avoid duplicate symbols in 4555 // the symbol table. 4556 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4557 sym[sym_idx].Clear(); 4558 return true; 4559 } 4560 } 4561 } 4562 } 4563 } 4564 4565 sym[sym_idx].SetID(nlist_idx); 4566 sym[sym_idx].SetType(type); 4567 if (set_value) { 4568 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4569 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4570 if (symbol_section) 4571 add_symbol_addr(sym[sym_idx].GetAddress().GetFileAddress()); 4572 } 4573 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4574 if (nlist.n_desc & N_WEAK_REF) 4575 sym[sym_idx].SetIsWeak(true); 4576 4577 if (symbol_byte_size > 0) 4578 sym[sym_idx].SetByteSize(symbol_byte_size); 4579 4580 if (demangled_is_synthesized) 4581 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4582 4583 ++sym_idx; 4584 return true; 4585 }; 4586 4587 // First parse all the nlists but don't process them yet. See the next 4588 // comment for an explanation why. 4589 std::vector<struct nlist_64> nlists; 4590 nlists.reserve(symtab_load_command.nsyms); 4591 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 4592 if (auto nlist = 4593 ParseNList(nlist_data, nlist_data_offset, nlist_byte_size)) 4594 nlists.push_back(*nlist); 4595 else 4596 break; 4597 } 4598 4599 // Now parse all the debug symbols. This is needed to merge non-debug 4600 // symbols in the next step. Non-debug symbols are always coalesced into 4601 // the debug symbol. Doing this in one step would mean that some symbols 4602 // won't be merged. 4603 nlist_idx = 0; 4604 for (auto &nlist : nlists) { 4605 if (!ParseSymbolLambda(nlist, nlist_idx++, DebugSymbols)) 4606 break; 4607 } 4608 4609 // Finally parse all the non debug symbols. 4610 nlist_idx = 0; 4611 for (auto &nlist : nlists) { 4612 if (!ParseSymbolLambda(nlist, nlist_idx++, NonDebugSymbols)) 4613 break; 4614 } 4615 4616 for (const auto &pos : reexport_shlib_needs_fixup) { 4617 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4618 if (undef_pos != undefined_name_to_desc.end()) { 4619 const uint8_t dylib_ordinal = 4620 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4621 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4622 sym[pos.first].SetReExportedSymbolSharedLibrary( 4623 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4624 } 4625 } 4626 } 4627 4628 // Count how many trie symbols we'll add to the symbol table 4629 int trie_symbol_table_augment_count = 0; 4630 for (auto &e : external_sym_trie_entries) { 4631 if (symbols_added.find(e.entry.address) == symbols_added.end()) 4632 trie_symbol_table_augment_count++; 4633 } 4634 4635 if (num_syms < sym_idx + trie_symbol_table_augment_count) { 4636 num_syms = sym_idx + trie_symbol_table_augment_count; 4637 sym = symtab->Resize(num_syms); 4638 } 4639 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4640 4641 // Add symbols from the trie to the symbol table. 4642 for (auto &e : external_sym_trie_entries) { 4643 if (symbols_added.find(e.entry.address) != symbols_added.end()) 4644 continue; 4645 4646 // Find the section that this trie address is in, use that to annotate 4647 // symbol type as we add the trie address and name to the symbol table. 4648 Address symbol_addr; 4649 if (module_sp->ResolveFileAddress(e.entry.address, symbol_addr)) { 4650 SectionSP symbol_section(symbol_addr.GetSection()); 4651 const char *symbol_name = e.entry.name.GetCString(); 4652 bool demangled_is_synthesized = false; 4653 SymbolType type = 4654 GetSymbolType(symbol_name, demangled_is_synthesized, text_section_sp, 4655 data_section_sp, data_dirty_section_sp, 4656 data_const_section_sp, symbol_section); 4657 4658 sym[sym_idx].SetType(type); 4659 if (symbol_section) { 4660 sym[sym_idx].SetID(synthetic_sym_id++); 4661 sym[sym_idx].GetMangled().SetMangledName(ConstString(symbol_name)); 4662 if (demangled_is_synthesized) 4663 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4664 sym[sym_idx].SetIsSynthetic(true); 4665 sym[sym_idx].SetExternal(true); 4666 sym[sym_idx].GetAddressRef() = symbol_addr; 4667 add_symbol_addr(symbol_addr.GetFileAddress()); 4668 if (e.entry.flags & TRIE_SYMBOL_IS_THUMB) 4669 sym[sym_idx].SetFlags(MACHO_NLIST_ARM_SYMBOL_IS_THUMB); 4670 ++sym_idx; 4671 } 4672 } 4673 } 4674 4675 if (function_starts_count > 0) { 4676 uint32_t num_synthetic_function_symbols = 0; 4677 for (i = 0; i < function_starts_count; ++i) { 4678 if (symbols_added.find(function_starts.GetEntryRef(i).addr) == 4679 symbols_added.end()) 4680 ++num_synthetic_function_symbols; 4681 } 4682 4683 if (num_synthetic_function_symbols > 0) { 4684 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4685 num_syms = sym_idx + num_synthetic_function_symbols; 4686 sym = symtab->Resize(num_syms); 4687 } 4688 for (i = 0; i < function_starts_count; ++i) { 4689 const FunctionStarts::Entry *func_start_entry = 4690 function_starts.GetEntryAtIndex(i); 4691 if (symbols_added.find(func_start_entry->addr) == symbols_added.end()) { 4692 addr_t symbol_file_addr = func_start_entry->addr; 4693 uint32_t symbol_flags = 0; 4694 if (func_start_entry->data) 4695 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4696 Address symbol_addr; 4697 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4698 SectionSP symbol_section(symbol_addr.GetSection()); 4699 uint32_t symbol_byte_size = 0; 4700 if (symbol_section) { 4701 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4702 const FunctionStarts::Entry *next_func_start_entry = 4703 function_starts.FindNextEntry(func_start_entry); 4704 const addr_t section_end_file_addr = 4705 section_file_addr + symbol_section->GetByteSize(); 4706 if (next_func_start_entry) { 4707 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4708 if (is_arm) 4709 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4710 symbol_byte_size = std::min<lldb::addr_t>( 4711 next_symbol_file_addr - symbol_file_addr, 4712 section_end_file_addr - symbol_file_addr); 4713 } else { 4714 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4715 } 4716 sym[sym_idx].SetID(synthetic_sym_id++); 4717 // Don't set the name for any synthetic symbols, the Symbol 4718 // object will generate one if needed when the name is accessed 4719 // via accessors. 4720 sym[sym_idx].GetMangled().SetDemangledName(ConstString()); 4721 sym[sym_idx].SetType(eSymbolTypeCode); 4722 sym[sym_idx].SetIsSynthetic(true); 4723 sym[sym_idx].GetAddressRef() = symbol_addr; 4724 add_symbol_addr(symbol_addr.GetFileAddress()); 4725 if (symbol_flags) 4726 sym[sym_idx].SetFlags(symbol_flags); 4727 if (symbol_byte_size) 4728 sym[sym_idx].SetByteSize(symbol_byte_size); 4729 ++sym_idx; 4730 } 4731 } 4732 } 4733 } 4734 } 4735 } 4736 4737 // Trim our symbols down to just what we ended up with after removing any 4738 // symbols. 4739 if (sym_idx < num_syms) { 4740 num_syms = sym_idx; 4741 sym = symtab->Resize(num_syms); 4742 } 4743 4744 // Now synthesize indirect symbols 4745 if (m_dysymtab.nindirectsyms != 0) { 4746 if (indirect_symbol_index_data.GetByteSize()) { 4747 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4748 m_nlist_idx_to_sym_idx.end(); 4749 4750 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4751 ++sect_idx) { 4752 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4753 S_SYMBOL_STUBS) { 4754 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4755 if (symbol_stub_byte_size == 0) 4756 continue; 4757 4758 const uint32_t num_symbol_stubs = 4759 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4760 4761 if (num_symbol_stubs == 0) 4762 continue; 4763 4764 const uint32_t symbol_stub_index_offset = 4765 m_mach_sections[sect_idx].reserved1; 4766 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) { 4767 const uint32_t symbol_stub_index = 4768 symbol_stub_index_offset + stub_idx; 4769 const lldb::addr_t symbol_stub_addr = 4770 m_mach_sections[sect_idx].addr + 4771 (stub_idx * symbol_stub_byte_size); 4772 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4773 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4774 symbol_stub_offset, 4)) { 4775 const uint32_t stub_sym_id = 4776 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4777 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4778 continue; 4779 4780 NListIndexToSymbolIndexMap::const_iterator index_pos = 4781 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4782 Symbol *stub_symbol = nullptr; 4783 if (index_pos != end_index_pos) { 4784 // We have a remapping from the original nlist index to a 4785 // current symbol index, so just look this up by index 4786 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4787 } else { 4788 // We need to lookup a symbol using the original nlist symbol 4789 // index since this index is coming from the S_SYMBOL_STUBS 4790 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4791 } 4792 4793 if (stub_symbol) { 4794 Address so_addr(symbol_stub_addr, section_list); 4795 4796 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4797 // Change the external symbol into a trampoline that makes 4798 // sense These symbols were N_UNDF N_EXT, and are useless 4799 // to us, so we can re-use them so we don't have to make up 4800 // a synthetic symbol for no good reason. 4801 if (resolver_addresses.find(symbol_stub_addr) == 4802 resolver_addresses.end()) 4803 stub_symbol->SetType(eSymbolTypeTrampoline); 4804 else 4805 stub_symbol->SetType(eSymbolTypeResolver); 4806 stub_symbol->SetExternal(false); 4807 stub_symbol->GetAddressRef() = so_addr; 4808 stub_symbol->SetByteSize(symbol_stub_byte_size); 4809 } else { 4810 // Make a synthetic symbol to describe the trampoline stub 4811 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4812 if (sym_idx >= num_syms) { 4813 sym = symtab->Resize(++num_syms); 4814 stub_symbol = nullptr; // this pointer no longer valid 4815 } 4816 sym[sym_idx].SetID(synthetic_sym_id++); 4817 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4818 if (resolver_addresses.find(symbol_stub_addr) == 4819 resolver_addresses.end()) 4820 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4821 else 4822 sym[sym_idx].SetType(eSymbolTypeResolver); 4823 sym[sym_idx].SetIsSynthetic(true); 4824 sym[sym_idx].GetAddressRef() = so_addr; 4825 add_symbol_addr(so_addr.GetFileAddress()); 4826 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4827 ++sym_idx; 4828 } 4829 } else { 4830 if (log) 4831 log->Warning("symbol stub referencing symbol table symbol " 4832 "%u that isn't in our minimal symbol table, " 4833 "fix this!!!", 4834 stub_sym_id); 4835 } 4836 } 4837 } 4838 } 4839 } 4840 } 4841 } 4842 4843 if (!reexport_trie_entries.empty()) { 4844 for (const auto &e : reexport_trie_entries) { 4845 if (e.entry.import_name) { 4846 // Only add indirect symbols from the Trie entries if we didn't have 4847 // a N_INDR nlist entry for this already 4848 if (indirect_symbol_names.find(e.entry.name) == 4849 indirect_symbol_names.end()) { 4850 // Make a synthetic symbol to describe re-exported symbol. 4851 if (sym_idx >= num_syms) 4852 sym = symtab->Resize(++num_syms); 4853 sym[sym_idx].SetID(synthetic_sym_id++); 4854 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4855 sym[sym_idx].SetType(eSymbolTypeReExported); 4856 sym[sym_idx].SetIsSynthetic(true); 4857 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4858 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4859 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4860 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4861 } 4862 ++sym_idx; 4863 } 4864 } 4865 } 4866 } 4867 4868 // StreamFile s(stdout, false); 4869 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4870 // symtab->Dump(&s, NULL, eSortOrderNone); 4871 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4872 // sizes 4873 symtab->CalculateSymbolSizes(); 4874 4875 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4876 // symtab->Dump(&s, NULL, eSortOrderNone); 4877 4878 return symtab->GetNumSymbols(); 4879 } 4880 4881 void ObjectFileMachO::Dump(Stream *s) { 4882 ModuleSP module_sp(GetModule()); 4883 if (module_sp) { 4884 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4885 s->Printf("%p: ", static_cast<void *>(this)); 4886 s->Indent(); 4887 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4888 s->PutCString("ObjectFileMachO64"); 4889 else 4890 s->PutCString("ObjectFileMachO32"); 4891 4892 *s << ", file = '" << m_file; 4893 ModuleSpecList all_specs; 4894 ModuleSpec base_spec; 4895 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 4896 base_spec, all_specs); 4897 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 4898 *s << "', triple"; 4899 if (e) 4900 s->Printf("[%d]", i); 4901 *s << " = "; 4902 *s << all_specs.GetModuleSpecRefAtIndex(i) 4903 .GetArchitecture() 4904 .GetTriple() 4905 .getTriple(); 4906 } 4907 *s << "\n"; 4908 SectionList *sections = GetSectionList(); 4909 if (sections) 4910 sections->Dump(s->AsRawOstream(), s->GetIndentLevel(), nullptr, true, 4911 UINT32_MAX); 4912 4913 if (m_symtab_up) 4914 m_symtab_up->Dump(s, nullptr, eSortOrderNone); 4915 } 4916 } 4917 4918 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4919 const lldb_private::DataExtractor &data, 4920 lldb::offset_t lc_offset) { 4921 uint32_t i; 4922 llvm::MachO::uuid_command load_cmd; 4923 4924 lldb::offset_t offset = lc_offset; 4925 for (i = 0; i < header.ncmds; ++i) { 4926 const lldb::offset_t cmd_offset = offset; 4927 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4928 break; 4929 4930 if (load_cmd.cmd == LC_UUID) { 4931 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4932 4933 if (uuid_bytes) { 4934 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4935 // We pretend these object files have no UUID to prevent crashing. 4936 4937 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4938 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4939 0xbb, 0x14, 0xf0, 0x0d}; 4940 4941 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4942 return UUID(); 4943 4944 return UUID::fromOptionalData(uuid_bytes, 16); 4945 } 4946 return UUID(); 4947 } 4948 offset = cmd_offset + load_cmd.cmdsize; 4949 } 4950 return UUID(); 4951 } 4952 4953 static llvm::StringRef GetOSName(uint32_t cmd) { 4954 switch (cmd) { 4955 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4956 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4957 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4958 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4959 case llvm::MachO::LC_VERSION_MIN_TVOS: 4960 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4961 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4962 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4963 default: 4964 llvm_unreachable("unexpected LC_VERSION load command"); 4965 } 4966 } 4967 4968 namespace { 4969 struct OSEnv { 4970 llvm::StringRef os_type; 4971 llvm::StringRef environment; 4972 OSEnv(uint32_t cmd) { 4973 switch (cmd) { 4974 case llvm::MachO::PLATFORM_MACOS: 4975 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4976 return; 4977 case llvm::MachO::PLATFORM_IOS: 4978 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4979 return; 4980 case llvm::MachO::PLATFORM_TVOS: 4981 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4982 return; 4983 case llvm::MachO::PLATFORM_WATCHOS: 4984 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4985 return; 4986 // NEED_BRIDGEOS_TRIPLE case llvm::MachO::PLATFORM_BRIDGEOS: 4987 // NEED_BRIDGEOS_TRIPLE os_type = 4988 // llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4989 // NEED_BRIDGEOS_TRIPLE return; 4990 case llvm::MachO::PLATFORM_MACCATALYST: 4991 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4992 environment = llvm::Triple::getEnvironmentTypeName(llvm::Triple::MacABI); 4993 return; 4994 case llvm::MachO::PLATFORM_IOSSIMULATOR: 4995 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4996 environment = 4997 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4998 return; 4999 case llvm::MachO::PLATFORM_TVOSSIMULATOR: 5000 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 5001 environment = 5002 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 5003 return; 5004 case llvm::MachO::PLATFORM_WATCHOSSIMULATOR: 5005 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 5006 environment = 5007 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 5008 return; 5009 default: { 5010 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5011 LIBLLDB_LOG_PROCESS)); 5012 LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION"); 5013 } 5014 } 5015 } 5016 }; 5017 5018 struct MinOS { 5019 uint32_t major_version, minor_version, patch_version; 5020 MinOS(uint32_t version) 5021 : major_version(version >> 16), minor_version((version >> 8) & 0xffu), 5022 patch_version(version & 0xffu) {} 5023 }; 5024 } // namespace 5025 5026 void ObjectFileMachO::GetAllArchSpecs(const llvm::MachO::mach_header &header, 5027 const lldb_private::DataExtractor &data, 5028 lldb::offset_t lc_offset, 5029 ModuleSpec &base_spec, 5030 lldb_private::ModuleSpecList &all_specs) { 5031 auto &base_arch = base_spec.GetArchitecture(); 5032 base_arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 5033 if (!base_arch.IsValid()) 5034 return; 5035 5036 bool found_any = false; 5037 auto add_triple = [&](const llvm::Triple &triple) { 5038 auto spec = base_spec; 5039 spec.GetArchitecture().GetTriple() = triple; 5040 if (spec.GetArchitecture().IsValid()) { 5041 spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset); 5042 all_specs.Append(spec); 5043 found_any = true; 5044 } 5045 }; 5046 5047 // Set OS to an unspecified unknown or a "*" so it can match any OS 5048 llvm::Triple base_triple = base_arch.GetTriple(); 5049 base_triple.setOS(llvm::Triple::UnknownOS); 5050 base_triple.setOSName(llvm::StringRef()); 5051 5052 if (header.filetype == MH_PRELOAD) { 5053 if (header.cputype == CPU_TYPE_ARM) { 5054 // If this is a 32-bit arm binary, and it's a standalone binary, force 5055 // the Vendor to Apple so we don't accidentally pick up the generic 5056 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 5057 // frame pointer register; most other armv7 ABIs use a combination of 5058 // r7 and r11. 5059 base_triple.setVendor(llvm::Triple::Apple); 5060 } else { 5061 // Set vendor to an unspecified unknown or a "*" so it can match any 5062 // vendor This is required for correct behavior of EFI debugging on 5063 // x86_64 5064 base_triple.setVendor(llvm::Triple::UnknownVendor); 5065 base_triple.setVendorName(llvm::StringRef()); 5066 } 5067 return add_triple(base_triple); 5068 } 5069 5070 llvm::MachO::load_command load_cmd; 5071 5072 // See if there is an LC_VERSION_MIN_* load command that can give 5073 // us the OS type. 5074 lldb::offset_t offset = lc_offset; 5075 for (uint32_t i = 0; i < header.ncmds; ++i) { 5076 const lldb::offset_t cmd_offset = offset; 5077 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 5078 break; 5079 5080 llvm::MachO::version_min_command version_min; 5081 switch (load_cmd.cmd) { 5082 case llvm::MachO::LC_VERSION_MIN_MACOSX: 5083 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 5084 case llvm::MachO::LC_VERSION_MIN_TVOS: 5085 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 5086 if (load_cmd.cmdsize != sizeof(version_min)) 5087 break; 5088 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 5089 data.GetByteOrder(), &version_min) == 0) 5090 break; 5091 MinOS min_os(version_min.version); 5092 llvm::SmallString<32> os_name; 5093 llvm::raw_svector_ostream os(os_name); 5094 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 5095 << min_os.minor_version << '.' << min_os.patch_version; 5096 5097 auto triple = base_triple; 5098 triple.setOSName(os.str()); 5099 5100 // Disambiguate legacy simulator platforms. 5101 if (load_cmd.cmd != llvm::MachO::LC_VERSION_MIN_MACOSX && 5102 (base_triple.getArch() == llvm::Triple::x86_64 || 5103 base_triple.getArch() == llvm::Triple::x86)) { 5104 // The combination of legacy LC_VERSION_MIN load command and 5105 // x86 architecture always indicates a simulator environment. 5106 // The combination of LC_VERSION_MIN and arm architecture only 5107 // appears for native binaries. Back-deploying simulator 5108 // binaries on Apple Silicon Macs use the modern unambigous 5109 // LC_BUILD_VERSION load commands; no special handling required. 5110 triple.setEnvironment(llvm::Triple::Simulator); 5111 } 5112 add_triple(triple); 5113 break; 5114 } 5115 default: 5116 break; 5117 } 5118 5119 offset = cmd_offset + load_cmd.cmdsize; 5120 } 5121 5122 // See if there are LC_BUILD_VERSION load commands that can give 5123 // us the OS type. 5124 offset = lc_offset; 5125 for (uint32_t i = 0; i < header.ncmds; ++i) { 5126 const lldb::offset_t cmd_offset = offset; 5127 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 5128 break; 5129 5130 do { 5131 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 5132 llvm::MachO::build_version_command build_version; 5133 if (load_cmd.cmdsize < sizeof(build_version)) { 5134 // Malformed load command. 5135 break; 5136 } 5137 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 5138 data.GetByteOrder(), &build_version) == 0) 5139 break; 5140 MinOS min_os(build_version.minos); 5141 OSEnv os_env(build_version.platform); 5142 llvm::SmallString<16> os_name; 5143 llvm::raw_svector_ostream os(os_name); 5144 os << os_env.os_type << min_os.major_version << '.' 5145 << min_os.minor_version << '.' << min_os.patch_version; 5146 auto triple = base_triple; 5147 triple.setOSName(os.str()); 5148 os_name.clear(); 5149 if (!os_env.environment.empty()) 5150 triple.setEnvironmentName(os_env.environment); 5151 add_triple(triple); 5152 } 5153 } while (0); 5154 offset = cmd_offset + load_cmd.cmdsize; 5155 } 5156 5157 if (!found_any) { 5158 if (header.filetype == MH_KEXT_BUNDLE) { 5159 base_triple.setVendor(llvm::Triple::Apple); 5160 add_triple(base_triple); 5161 } else { 5162 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 5163 // so lets not say our Vendor is Apple, leave it as an unspecified 5164 // unknown. 5165 base_triple.setVendor(llvm::Triple::UnknownVendor); 5166 base_triple.setVendorName(llvm::StringRef()); 5167 add_triple(base_triple); 5168 } 5169 } 5170 } 5171 5172 ArchSpec ObjectFileMachO::GetArchitecture( 5173 ModuleSP module_sp, const llvm::MachO::mach_header &header, 5174 const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) { 5175 ModuleSpecList all_specs; 5176 ModuleSpec base_spec; 5177 GetAllArchSpecs(header, data, MachHeaderSizeFromMagic(header.magic), 5178 base_spec, all_specs); 5179 5180 // If the object file offers multiple alternative load commands, 5181 // pick the one that matches the module. 5182 if (module_sp) { 5183 const ArchSpec &module_arch = module_sp->GetArchitecture(); 5184 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 5185 ArchSpec mach_arch = 5186 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 5187 if (module_arch.IsCompatibleMatch(mach_arch)) 5188 return mach_arch; 5189 } 5190 } 5191 5192 // Return the first arch we found. 5193 if (all_specs.GetSize() == 0) 5194 return {}; 5195 return all_specs.GetModuleSpecRefAtIndex(0).GetArchitecture(); 5196 } 5197 5198 UUID ObjectFileMachO::GetUUID() { 5199 ModuleSP module_sp(GetModule()); 5200 if (module_sp) { 5201 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5202 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5203 return GetUUID(m_header, m_data, offset); 5204 } 5205 return UUID(); 5206 } 5207 5208 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5209 uint32_t count = 0; 5210 ModuleSP module_sp(GetModule()); 5211 if (module_sp) { 5212 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5213 llvm::MachO::load_command load_cmd; 5214 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5215 std::vector<std::string> rpath_paths; 5216 std::vector<std::string> rpath_relative_paths; 5217 std::vector<std::string> at_exec_relative_paths; 5218 uint32_t i; 5219 for (i = 0; i < m_header.ncmds; ++i) { 5220 const uint32_t cmd_offset = offset; 5221 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5222 break; 5223 5224 switch (load_cmd.cmd) { 5225 case LC_RPATH: 5226 case LC_LOAD_DYLIB: 5227 case LC_LOAD_WEAK_DYLIB: 5228 case LC_REEXPORT_DYLIB: 5229 case LC_LOAD_DYLINKER: 5230 case LC_LOADFVMLIB: 5231 case LC_LOAD_UPWARD_DYLIB: { 5232 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5233 const char *path = m_data.PeekCStr(name_offset); 5234 if (path) { 5235 if (load_cmd.cmd == LC_RPATH) 5236 rpath_paths.push_back(path); 5237 else { 5238 if (path[0] == '@') { 5239 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5240 rpath_relative_paths.push_back(path + strlen("@rpath")); 5241 else if (strncmp(path, "@executable_path", 5242 strlen("@executable_path")) == 0) 5243 at_exec_relative_paths.push_back(path + 5244 strlen("@executable_path")); 5245 } else { 5246 FileSpec file_spec(path); 5247 if (files.AppendIfUnique(file_spec)) 5248 count++; 5249 } 5250 } 5251 } 5252 } break; 5253 5254 default: 5255 break; 5256 } 5257 offset = cmd_offset + load_cmd.cmdsize; 5258 } 5259 5260 FileSpec this_file_spec(m_file); 5261 FileSystem::Instance().Resolve(this_file_spec); 5262 5263 if (!rpath_paths.empty()) { 5264 // Fixup all LC_RPATH values to be absolute paths 5265 std::string loader_path("@loader_path"); 5266 std::string executable_path("@executable_path"); 5267 for (auto &rpath : rpath_paths) { 5268 if (llvm::StringRef(rpath).startswith(loader_path)) { 5269 rpath.erase(0, loader_path.size()); 5270 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5271 } else if (llvm::StringRef(rpath).startswith(executable_path)) { 5272 rpath.erase(0, executable_path.size()); 5273 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5274 } 5275 } 5276 5277 for (const auto &rpath_relative_path : rpath_relative_paths) { 5278 for (const auto &rpath : rpath_paths) { 5279 std::string path = rpath; 5280 path += rpath_relative_path; 5281 // It is OK to resolve this path because we must find a file on disk 5282 // for us to accept it anyway if it is rpath relative. 5283 FileSpec file_spec(path); 5284 FileSystem::Instance().Resolve(file_spec); 5285 if (FileSystem::Instance().Exists(file_spec) && 5286 files.AppendIfUnique(file_spec)) { 5287 count++; 5288 break; 5289 } 5290 } 5291 } 5292 } 5293 5294 // We may have @executable_paths but no RPATHS. Figure those out here. 5295 // Only do this if this object file is the executable. We have no way to 5296 // get back to the actual executable otherwise, so we won't get the right 5297 // path. 5298 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5299 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5300 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5301 FileSpec file_spec = 5302 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5303 if (FileSystem::Instance().Exists(file_spec) && 5304 files.AppendIfUnique(file_spec)) 5305 count++; 5306 } 5307 } 5308 } 5309 return count; 5310 } 5311 5312 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5313 // If the object file is not an executable it can't hold the entry point. 5314 // m_entry_point_address is initialized to an invalid address, so we can just 5315 // return that. If m_entry_point_address is valid it means we've found it 5316 // already, so return the cached value. 5317 5318 if ((!IsExecutable() && !IsDynamicLoader()) || 5319 m_entry_point_address.IsValid()) { 5320 return m_entry_point_address; 5321 } 5322 5323 // Otherwise, look for the UnixThread or Thread command. The data for the 5324 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5325 // 5326 // uint32_t flavor - this is the flavor argument you would pass to 5327 // thread_get_state 5328 // uint32_t count - this is the count of longs in the thread state data 5329 // struct XXX_thread_state state - this is the structure from 5330 // <machine/thread_status.h> corresponding to the flavor. 5331 // <repeat this trio> 5332 // 5333 // So we just keep reading the various register flavors till we find the GPR 5334 // one, then read the PC out of there. 5335 // FIXME: We will need to have a "RegisterContext data provider" class at some 5336 // point that can get all the registers 5337 // out of data in this form & attach them to a given thread. That should 5338 // underlie the MacOS X User process plugin, and we'll also need it for the 5339 // MacOS X Core File process plugin. When we have that we can also use it 5340 // here. 5341 // 5342 // For now we hard-code the offsets and flavors we need: 5343 // 5344 // 5345 5346 ModuleSP module_sp(GetModule()); 5347 if (module_sp) { 5348 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5349 llvm::MachO::load_command load_cmd; 5350 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5351 uint32_t i; 5352 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5353 bool done = false; 5354 5355 for (i = 0; i < m_header.ncmds; ++i) { 5356 const lldb::offset_t cmd_offset = offset; 5357 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5358 break; 5359 5360 switch (load_cmd.cmd) { 5361 case LC_UNIXTHREAD: 5362 case LC_THREAD: { 5363 while (offset < cmd_offset + load_cmd.cmdsize) { 5364 uint32_t flavor = m_data.GetU32(&offset); 5365 uint32_t count = m_data.GetU32(&offset); 5366 if (count == 0) { 5367 // We've gotten off somehow, log and exit; 5368 return m_entry_point_address; 5369 } 5370 5371 switch (m_header.cputype) { 5372 case llvm::MachO::CPU_TYPE_ARM: 5373 if (flavor == 1 || 5374 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 5375 // from mach/arm/thread_status.h 5376 { 5377 offset += 60; // This is the offset of pc in the GPR thread state 5378 // data structure. 5379 start_address = m_data.GetU32(&offset); 5380 done = true; 5381 } 5382 break; 5383 case llvm::MachO::CPU_TYPE_ARM64: 5384 case llvm::MachO::CPU_TYPE_ARM64_32: 5385 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5386 { 5387 offset += 256; // This is the offset of pc in the GPR thread state 5388 // data structure. 5389 start_address = m_data.GetU64(&offset); 5390 done = true; 5391 } 5392 break; 5393 case llvm::MachO::CPU_TYPE_I386: 5394 if (flavor == 5395 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5396 { 5397 offset += 40; // This is the offset of eip in the GPR thread state 5398 // data structure. 5399 start_address = m_data.GetU32(&offset); 5400 done = true; 5401 } 5402 break; 5403 case llvm::MachO::CPU_TYPE_X86_64: 5404 if (flavor == 5405 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5406 { 5407 offset += 16 * 8; // This is the offset of rip in the GPR thread 5408 // state data structure. 5409 start_address = m_data.GetU64(&offset); 5410 done = true; 5411 } 5412 break; 5413 default: 5414 return m_entry_point_address; 5415 } 5416 // Haven't found the GPR flavor yet, skip over the data for this 5417 // flavor: 5418 if (done) 5419 break; 5420 offset += count * 4; 5421 } 5422 } break; 5423 case LC_MAIN: { 5424 ConstString text_segment_name("__TEXT"); 5425 uint64_t entryoffset = m_data.GetU64(&offset); 5426 SectionSP text_segment_sp = 5427 GetSectionList()->FindSectionByName(text_segment_name); 5428 if (text_segment_sp) { 5429 done = true; 5430 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5431 } 5432 } break; 5433 5434 default: 5435 break; 5436 } 5437 if (done) 5438 break; 5439 5440 // Go to the next load command: 5441 offset = cmd_offset + load_cmd.cmdsize; 5442 } 5443 5444 if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) { 5445 if (GetSymtab()) { 5446 Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType( 5447 ConstString("_dyld_start"), SymbolType::eSymbolTypeCode, 5448 Symtab::eDebugAny, Symtab::eVisibilityAny); 5449 if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) { 5450 start_address = dyld_start_sym->GetAddress().GetFileAddress(); 5451 } 5452 } 5453 } 5454 5455 if (start_address != LLDB_INVALID_ADDRESS) { 5456 // We got the start address from the load commands, so now resolve that 5457 // address in the sections of this ObjectFile: 5458 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5459 start_address, GetSectionList())) { 5460 m_entry_point_address.Clear(); 5461 } 5462 } else { 5463 // We couldn't read the UnixThread load command - maybe it wasn't there. 5464 // As a fallback look for the "start" symbol in the main executable. 5465 5466 ModuleSP module_sp(GetModule()); 5467 5468 if (module_sp) { 5469 SymbolContextList contexts; 5470 SymbolContext context; 5471 module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5472 eSymbolTypeCode, contexts); 5473 if (contexts.GetSize()) { 5474 if (contexts.GetContextAtIndex(0, context)) 5475 m_entry_point_address = context.symbol->GetAddress(); 5476 } 5477 } 5478 } 5479 } 5480 5481 return m_entry_point_address; 5482 } 5483 5484 lldb_private::Address ObjectFileMachO::GetBaseAddress() { 5485 lldb_private::Address header_addr; 5486 SectionList *section_list = GetSectionList(); 5487 if (section_list) { 5488 SectionSP text_segment_sp( 5489 section_list->FindSectionByName(GetSegmentNameTEXT())); 5490 if (text_segment_sp) { 5491 header_addr.SetSection(text_segment_sp); 5492 header_addr.SetOffset(0); 5493 } 5494 } 5495 return header_addr; 5496 } 5497 5498 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5499 ModuleSP module_sp(GetModule()); 5500 if (module_sp) { 5501 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5502 if (!m_thread_context_offsets_valid) { 5503 m_thread_context_offsets_valid = true; 5504 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5505 FileRangeArray::Entry file_range; 5506 llvm::MachO::thread_command thread_cmd; 5507 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5508 const uint32_t cmd_offset = offset; 5509 if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr) 5510 break; 5511 5512 if (thread_cmd.cmd == LC_THREAD) { 5513 file_range.SetRangeBase(offset); 5514 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5515 m_thread_context_offsets.Append(file_range); 5516 } 5517 offset = cmd_offset + thread_cmd.cmdsize; 5518 } 5519 } 5520 } 5521 return m_thread_context_offsets.GetSize(); 5522 } 5523 5524 std::string ObjectFileMachO::GetIdentifierString() { 5525 std::string result; 5526 ModuleSP module_sp(GetModule()); 5527 if (module_sp) { 5528 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5529 5530 // First, look over the load commands for an LC_NOTE load command with 5531 // data_owner string "kern ver str" & use that if found. 5532 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5533 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5534 const uint32_t cmd_offset = offset; 5535 llvm::MachO::load_command lc; 5536 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5537 break; 5538 if (lc.cmd == LC_NOTE) { 5539 char data_owner[17]; 5540 m_data.CopyData(offset, 16, data_owner); 5541 data_owner[16] = '\0'; 5542 offset += 16; 5543 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5544 uint64_t size = m_data.GetU64_unchecked(&offset); 5545 5546 // "kern ver str" has a uint32_t version and then a nul terminated 5547 // c-string. 5548 if (strcmp("kern ver str", data_owner) == 0) { 5549 offset = fileoff; 5550 uint32_t version; 5551 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5552 if (version == 1) { 5553 uint32_t strsize = size - sizeof(uint32_t); 5554 char *buf = (char *)malloc(strsize); 5555 if (buf) { 5556 m_data.CopyData(offset, strsize, buf); 5557 buf[strsize - 1] = '\0'; 5558 result = buf; 5559 if (buf) 5560 free(buf); 5561 return result; 5562 } 5563 } 5564 } 5565 } 5566 } 5567 offset = cmd_offset + lc.cmdsize; 5568 } 5569 5570 // Second, make a pass over the load commands looking for an obsolete 5571 // LC_IDENT load command. 5572 offset = MachHeaderSizeFromMagic(m_header.magic); 5573 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5574 const uint32_t cmd_offset = offset; 5575 llvm::MachO::ident_command ident_command; 5576 if (m_data.GetU32(&offset, &ident_command, 2) == nullptr) 5577 break; 5578 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5579 char *buf = (char *)malloc(ident_command.cmdsize); 5580 if (buf != nullptr && m_data.CopyData(offset, ident_command.cmdsize, 5581 buf) == ident_command.cmdsize) { 5582 buf[ident_command.cmdsize - 1] = '\0'; 5583 result = buf; 5584 } 5585 if (buf) 5586 free(buf); 5587 } 5588 offset = cmd_offset + ident_command.cmdsize; 5589 } 5590 } 5591 return result; 5592 } 5593 5594 addr_t ObjectFileMachO::GetAddressMask() { 5595 addr_t mask = 0; 5596 ModuleSP module_sp(GetModule()); 5597 if (module_sp) { 5598 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5599 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5600 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5601 const uint32_t cmd_offset = offset; 5602 llvm::MachO::load_command lc; 5603 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5604 break; 5605 if (lc.cmd == LC_NOTE) { 5606 char data_owner[17]; 5607 m_data.CopyData(offset, 16, data_owner); 5608 data_owner[16] = '\0'; 5609 offset += 16; 5610 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5611 5612 // "addrable bits" has a uint32_t version and a uint32_t 5613 // number of bits used in addressing. 5614 if (strcmp("addrable bits", data_owner) == 0) { 5615 offset = fileoff; 5616 uint32_t version; 5617 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5618 if (version == 3) { 5619 uint32_t num_addr_bits = m_data.GetU32_unchecked(&offset); 5620 if (num_addr_bits != 0) { 5621 mask = ~((1ULL << num_addr_bits) - 1); 5622 } 5623 break; 5624 } 5625 } 5626 } 5627 } 5628 offset = cmd_offset + lc.cmdsize; 5629 } 5630 } 5631 return mask; 5632 } 5633 5634 bool ObjectFileMachO::GetCorefileMainBinaryInfo(addr_t &address, UUID &uuid, 5635 ObjectFile::BinaryType &type) { 5636 address = LLDB_INVALID_ADDRESS; 5637 uuid.Clear(); 5638 ModuleSP module_sp(GetModule()); 5639 if (module_sp) { 5640 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5641 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5642 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5643 const uint32_t cmd_offset = offset; 5644 llvm::MachO::load_command lc; 5645 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5646 break; 5647 if (lc.cmd == LC_NOTE) { 5648 char data_owner[17]; 5649 memset(data_owner, 0, sizeof(data_owner)); 5650 m_data.CopyData(offset, 16, data_owner); 5651 offset += 16; 5652 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5653 uint64_t size = m_data.GetU64_unchecked(&offset); 5654 5655 // "main bin spec" (main binary specification) data payload is 5656 // formatted: 5657 // uint32_t version [currently 1] 5658 // uint32_t type [0 == unspecified, 1 == kernel, 5659 // 2 == user process, 3 == firmware ] 5660 // uint64_t address [ UINT64_MAX if address not specified ] 5661 // uuid_t uuid [ all zero's if uuid not specified ] 5662 // uint32_t log2_pagesize [ process page size in log base 5663 // 2, e.g. 4k pages are 12. 5664 // 0 for unspecified ] 5665 // uint32_t unused [ for alignment ] 5666 5667 if (strcmp("main bin spec", data_owner) == 0 && size >= 32) { 5668 offset = fileoff; 5669 uint32_t version; 5670 if (m_data.GetU32(&offset, &version, 1) != nullptr && version == 1) { 5671 uint32_t binspec_type = 0; 5672 uuid_t raw_uuid; 5673 memset(raw_uuid, 0, sizeof(uuid_t)); 5674 5675 if (m_data.GetU32(&offset, &binspec_type, 1) && 5676 m_data.GetU64(&offset, &address, 1) && 5677 m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5678 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5679 // convert the "main bin spec" type into our 5680 // ObjectFile::BinaryType enum 5681 switch (binspec_type) { 5682 case 0: 5683 type = eBinaryTypeUnknown; 5684 break; 5685 case 1: 5686 type = eBinaryTypeKernel; 5687 break; 5688 case 2: 5689 type = eBinaryTypeUser; 5690 break; 5691 case 3: 5692 type = eBinaryTypeStandalone; 5693 break; 5694 } 5695 return true; 5696 } 5697 } 5698 } 5699 } 5700 offset = cmd_offset + lc.cmdsize; 5701 } 5702 } 5703 return false; 5704 } 5705 5706 lldb::RegisterContextSP 5707 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5708 lldb_private::Thread &thread) { 5709 lldb::RegisterContextSP reg_ctx_sp; 5710 5711 ModuleSP module_sp(GetModule()); 5712 if (module_sp) { 5713 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5714 if (!m_thread_context_offsets_valid) 5715 GetNumThreadContexts(); 5716 5717 const FileRangeArray::Entry *thread_context_file_range = 5718 m_thread_context_offsets.GetEntryAtIndex(idx); 5719 if (thread_context_file_range) { 5720 5721 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5722 thread_context_file_range->GetByteSize()); 5723 5724 switch (m_header.cputype) { 5725 case llvm::MachO::CPU_TYPE_ARM64: 5726 case llvm::MachO::CPU_TYPE_ARM64_32: 5727 reg_ctx_sp = 5728 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data); 5729 break; 5730 5731 case llvm::MachO::CPU_TYPE_ARM: 5732 reg_ctx_sp = 5733 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data); 5734 break; 5735 5736 case llvm::MachO::CPU_TYPE_I386: 5737 reg_ctx_sp = 5738 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data); 5739 break; 5740 5741 case llvm::MachO::CPU_TYPE_X86_64: 5742 reg_ctx_sp = 5743 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data); 5744 break; 5745 } 5746 } 5747 } 5748 return reg_ctx_sp; 5749 } 5750 5751 ObjectFile::Type ObjectFileMachO::CalculateType() { 5752 switch (m_header.filetype) { 5753 case MH_OBJECT: // 0x1u 5754 if (GetAddressByteSize() == 4) { 5755 // 32 bit kexts are just object files, but they do have a valid 5756 // UUID load command. 5757 if (GetUUID()) { 5758 // this checking for the UUID load command is not enough we could 5759 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5760 // this is required of kexts 5761 if (m_strata == eStrataInvalid) 5762 m_strata = eStrataKernel; 5763 return eTypeSharedLibrary; 5764 } 5765 } 5766 return eTypeObjectFile; 5767 5768 case MH_EXECUTE: 5769 return eTypeExecutable; // 0x2u 5770 case MH_FVMLIB: 5771 return eTypeSharedLibrary; // 0x3u 5772 case MH_CORE: 5773 return eTypeCoreFile; // 0x4u 5774 case MH_PRELOAD: 5775 return eTypeSharedLibrary; // 0x5u 5776 case MH_DYLIB: 5777 return eTypeSharedLibrary; // 0x6u 5778 case MH_DYLINKER: 5779 return eTypeDynamicLinker; // 0x7u 5780 case MH_BUNDLE: 5781 return eTypeSharedLibrary; // 0x8u 5782 case MH_DYLIB_STUB: 5783 return eTypeStubLibrary; // 0x9u 5784 case MH_DSYM: 5785 return eTypeDebugInfo; // 0xAu 5786 case MH_KEXT_BUNDLE: 5787 return eTypeSharedLibrary; // 0xBu 5788 default: 5789 break; 5790 } 5791 return eTypeUnknown; 5792 } 5793 5794 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5795 switch (m_header.filetype) { 5796 case MH_OBJECT: // 0x1u 5797 { 5798 // 32 bit kexts are just object files, but they do have a valid 5799 // UUID load command. 5800 if (GetUUID()) { 5801 // this checking for the UUID load command is not enough we could 5802 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5803 // this is required of kexts 5804 if (m_type == eTypeInvalid) 5805 m_type = eTypeSharedLibrary; 5806 5807 return eStrataKernel; 5808 } 5809 } 5810 return eStrataUnknown; 5811 5812 case MH_EXECUTE: // 0x2u 5813 // Check for the MH_DYLDLINK bit in the flags 5814 if (m_header.flags & MH_DYLDLINK) { 5815 return eStrataUser; 5816 } else { 5817 SectionList *section_list = GetSectionList(); 5818 if (section_list) { 5819 static ConstString g_kld_section_name("__KLD"); 5820 if (section_list->FindSectionByName(g_kld_section_name)) 5821 return eStrataKernel; 5822 } 5823 } 5824 return eStrataRawImage; 5825 5826 case MH_FVMLIB: 5827 return eStrataUser; // 0x3u 5828 case MH_CORE: 5829 return eStrataUnknown; // 0x4u 5830 case MH_PRELOAD: 5831 return eStrataRawImage; // 0x5u 5832 case MH_DYLIB: 5833 return eStrataUser; // 0x6u 5834 case MH_DYLINKER: 5835 return eStrataUser; // 0x7u 5836 case MH_BUNDLE: 5837 return eStrataUser; // 0x8u 5838 case MH_DYLIB_STUB: 5839 return eStrataUser; // 0x9u 5840 case MH_DSYM: 5841 return eStrataUnknown; // 0xAu 5842 case MH_KEXT_BUNDLE: 5843 return eStrataKernel; // 0xBu 5844 default: 5845 break; 5846 } 5847 return eStrataUnknown; 5848 } 5849 5850 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5851 ModuleSP module_sp(GetModule()); 5852 if (module_sp) { 5853 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5854 llvm::MachO::dylib_command load_cmd; 5855 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5856 uint32_t version_cmd = 0; 5857 uint64_t version = 0; 5858 uint32_t i; 5859 for (i = 0; i < m_header.ncmds; ++i) { 5860 const lldb::offset_t cmd_offset = offset; 5861 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5862 break; 5863 5864 if (load_cmd.cmd == LC_ID_DYLIB) { 5865 if (version_cmd == 0) { 5866 version_cmd = load_cmd.cmd; 5867 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr) 5868 break; 5869 version = load_cmd.dylib.current_version; 5870 } 5871 break; // Break for now unless there is another more complete version 5872 // number load command in the future. 5873 } 5874 offset = cmd_offset + load_cmd.cmdsize; 5875 } 5876 5877 if (version_cmd == LC_ID_DYLIB) { 5878 unsigned major = (version & 0xFFFF0000ull) >> 16; 5879 unsigned minor = (version & 0x0000FF00ull) >> 8; 5880 unsigned subminor = (version & 0x000000FFull); 5881 return llvm::VersionTuple(major, minor, subminor); 5882 } 5883 } 5884 return llvm::VersionTuple(); 5885 } 5886 5887 ArchSpec ObjectFileMachO::GetArchitecture() { 5888 ModuleSP module_sp(GetModule()); 5889 ArchSpec arch; 5890 if (module_sp) { 5891 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5892 5893 return GetArchitecture(module_sp, m_header, m_data, 5894 MachHeaderSizeFromMagic(m_header.magic)); 5895 } 5896 return arch; 5897 } 5898 5899 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, 5900 addr_t &base_addr, UUID &uuid) { 5901 uuid.Clear(); 5902 base_addr = LLDB_INVALID_ADDRESS; 5903 if (process && process->GetDynamicLoader()) { 5904 DynamicLoader *dl = process->GetDynamicLoader(); 5905 LazyBool using_shared_cache; 5906 LazyBool private_shared_cache; 5907 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5908 private_shared_cache); 5909 } 5910 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5911 LIBLLDB_LOG_PROCESS)); 5912 LLDB_LOGF( 5913 log, 5914 "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64, 5915 uuid.GetAsString().c_str(), base_addr); 5916 } 5917 5918 // From dyld SPI header dyld_process_info.h 5919 typedef void *dyld_process_info; 5920 struct lldb_copy__dyld_process_cache_info { 5921 uuid_t cacheUUID; // UUID of cache used by process 5922 uint64_t cacheBaseAddress; // load address of dyld shared cache 5923 bool noCache; // process is running without a dyld cache 5924 bool privateCache; // process is using a private copy of its dyld cache 5925 }; 5926 5927 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with 5928 // llvm enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile 5929 // errors. So we need to use the actual underlying types of task_t and 5930 // kern_return_t below. 5931 extern "C" unsigned int /*task_t*/ mach_task_self(); 5932 5933 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5934 uuid.Clear(); 5935 base_addr = LLDB_INVALID_ADDRESS; 5936 5937 #if defined(__APPLE__) 5938 uint8_t *(*dyld_get_all_image_infos)(void); 5939 dyld_get_all_image_infos = 5940 (uint8_t * (*)()) dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5941 if (dyld_get_all_image_infos) { 5942 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5943 if (dyld_all_image_infos_address) { 5944 uint32_t *version = (uint32_t *) 5945 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5946 if (*version >= 13) { 5947 uuid_t *sharedCacheUUID_address = 0; 5948 int wordsize = sizeof(uint8_t *); 5949 if (wordsize == 8) { 5950 sharedCacheUUID_address = 5951 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5952 160); // sharedCacheUUID <mach-o/dyld_images.h> 5953 if (*version >= 15) 5954 base_addr = 5955 *(uint64_t 5956 *)((uint8_t *)dyld_all_image_infos_address + 5957 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5958 } else { 5959 sharedCacheUUID_address = 5960 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5961 84); // sharedCacheUUID <mach-o/dyld_images.h> 5962 if (*version >= 15) { 5963 base_addr = 0; 5964 base_addr = 5965 *(uint32_t 5966 *)((uint8_t *)dyld_all_image_infos_address + 5967 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5968 } 5969 } 5970 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5971 } 5972 } 5973 } else { 5974 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5975 dyld_process_info (*dyld_process_info_create)( 5976 unsigned int /* task_t */ task, uint64_t timestamp, 5977 unsigned int /*kern_return_t*/ *kernelError); 5978 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5979 void (*dyld_process_info_release)(dyld_process_info info); 5980 5981 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, 5982 unsigned int /*kern_return_t*/ *)) 5983 dlsym(RTLD_DEFAULT, "_dyld_process_info_create"); 5984 dyld_process_info_get_cache = (void (*)(void *, void *))dlsym( 5985 RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5986 dyld_process_info_release = 5987 (void (*)(void *))dlsym(RTLD_DEFAULT, "_dyld_process_info_release"); 5988 5989 if (dyld_process_info_create && dyld_process_info_get_cache) { 5990 unsigned int /*kern_return_t */ kern_ret; 5991 dyld_process_info process_info = 5992 dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5993 if (process_info) { 5994 struct lldb_copy__dyld_process_cache_info sc_info; 5995 memset(&sc_info, 0, sizeof(struct lldb_copy__dyld_process_cache_info)); 5996 dyld_process_info_get_cache(process_info, &sc_info); 5997 if (sc_info.cacheBaseAddress != 0) { 5998 base_addr = sc_info.cacheBaseAddress; 5999 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 6000 } 6001 dyld_process_info_release(process_info); 6002 } 6003 } 6004 } 6005 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 6006 LIBLLDB_LOG_PROCESS)); 6007 if (log && uuid.IsValid()) 6008 LLDB_LOGF(log, 6009 "lldb's in-memory shared cache has a UUID of %s base address of " 6010 "0x%" PRIx64, 6011 uuid.GetAsString().c_str(), base_addr); 6012 #endif 6013 } 6014 6015 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 6016 if (!m_min_os_version) { 6017 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 6018 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6019 const lldb::offset_t load_cmd_offset = offset; 6020 6021 llvm::MachO::version_min_command lc; 6022 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6023 break; 6024 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 6025 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 6026 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 6027 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 6028 if (m_data.GetU32(&offset, &lc.version, 6029 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 6030 const uint32_t xxxx = lc.version >> 16; 6031 const uint32_t yy = (lc.version >> 8) & 0xffu; 6032 const uint32_t zz = lc.version & 0xffu; 6033 if (xxxx) { 6034 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 6035 break; 6036 } 6037 } 6038 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 6039 // struct build_version_command { 6040 // uint32_t cmd; /* LC_BUILD_VERSION */ 6041 // uint32_t cmdsize; /* sizeof(struct 6042 // build_version_command) plus */ 6043 // /* ntools * sizeof(struct 6044 // build_tool_version) */ 6045 // uint32_t platform; /* platform */ 6046 // uint32_t minos; /* X.Y.Z is encoded in nibbles 6047 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded in 6048 // nibbles xxxx.yy.zz */ uint32_t ntools; /* number of 6049 // tool entries following this */ 6050 // }; 6051 6052 offset += 4; // skip platform 6053 uint32_t minos = m_data.GetU32(&offset); 6054 6055 const uint32_t xxxx = minos >> 16; 6056 const uint32_t yy = (minos >> 8) & 0xffu; 6057 const uint32_t zz = minos & 0xffu; 6058 if (xxxx) { 6059 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 6060 break; 6061 } 6062 } 6063 6064 offset = load_cmd_offset + lc.cmdsize; 6065 } 6066 6067 if (!m_min_os_version) { 6068 // Set version to an empty value so we don't keep trying to 6069 m_min_os_version = llvm::VersionTuple(); 6070 } 6071 } 6072 6073 return *m_min_os_version; 6074 } 6075 6076 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { 6077 if (!m_sdk_versions.hasValue()) { 6078 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 6079 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6080 const lldb::offset_t load_cmd_offset = offset; 6081 6082 llvm::MachO::version_min_command lc; 6083 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6084 break; 6085 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 6086 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 6087 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 6088 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 6089 if (m_data.GetU32(&offset, &lc.version, 6090 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 6091 const uint32_t xxxx = lc.sdk >> 16; 6092 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 6093 const uint32_t zz = lc.sdk & 0xffu; 6094 if (xxxx) { 6095 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 6096 break; 6097 } else { 6098 GetModule()->ReportWarning("minimum OS version load command with " 6099 "invalid (0) version found."); 6100 } 6101 } 6102 } 6103 offset = load_cmd_offset + lc.cmdsize; 6104 } 6105 6106 if (!m_sdk_versions.hasValue()) { 6107 offset = MachHeaderSizeFromMagic(m_header.magic); 6108 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6109 const lldb::offset_t load_cmd_offset = offset; 6110 6111 llvm::MachO::version_min_command lc; 6112 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6113 break; 6114 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 6115 // struct build_version_command { 6116 // uint32_t cmd; /* LC_BUILD_VERSION */ 6117 // uint32_t cmdsize; /* sizeof(struct 6118 // build_version_command) plus */ 6119 // /* ntools * sizeof(struct 6120 // build_tool_version) */ 6121 // uint32_t platform; /* platform */ 6122 // uint32_t minos; /* X.Y.Z is encoded in nibbles 6123 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded 6124 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number 6125 // of tool entries following this */ 6126 // }; 6127 6128 offset += 4; // skip platform 6129 uint32_t minos = m_data.GetU32(&offset); 6130 6131 const uint32_t xxxx = minos >> 16; 6132 const uint32_t yy = (minos >> 8) & 0xffu; 6133 const uint32_t zz = minos & 0xffu; 6134 if (xxxx) { 6135 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 6136 break; 6137 } 6138 } 6139 offset = load_cmd_offset + lc.cmdsize; 6140 } 6141 } 6142 6143 if (!m_sdk_versions.hasValue()) 6144 m_sdk_versions = llvm::VersionTuple(); 6145 } 6146 6147 return m_sdk_versions.getValue(); 6148 } 6149 6150 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 6151 return m_header.filetype == llvm::MachO::MH_DYLINKER; 6152 } 6153 6154 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 6155 return m_allow_assembly_emulation_unwind_plans; 6156 } 6157 6158 // PluginInterface protocol 6159 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 6160 return GetPluginNameStatic(); 6161 } 6162 6163 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 6164 6165 Section *ObjectFileMachO::GetMachHeaderSection() { 6166 // Find the first address of the mach header which is the first non-zero file 6167 // sized section whose file offset is zero. This is the base file address of 6168 // the mach-o file which can be subtracted from the vmaddr of the other 6169 // segments found in memory and added to the load address 6170 ModuleSP module_sp = GetModule(); 6171 if (!module_sp) 6172 return nullptr; 6173 SectionList *section_list = GetSectionList(); 6174 if (!section_list) 6175 return nullptr; 6176 const size_t num_sections = section_list->GetSize(); 6177 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6178 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 6179 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 6180 return section; 6181 } 6182 return nullptr; 6183 } 6184 6185 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 6186 if (!section) 6187 return false; 6188 const bool is_dsym = (m_header.filetype == MH_DSYM); 6189 if (section->GetFileSize() == 0 && !is_dsym) 6190 return false; 6191 if (section->IsThreadSpecific()) 6192 return false; 6193 if (GetModule().get() != section->GetModule().get()) 6194 return false; 6195 // Be careful with __LINKEDIT and __DWARF segments 6196 if (section->GetName() == GetSegmentNameLINKEDIT() || 6197 section->GetName() == GetSegmentNameDWARF()) { 6198 // Only map __LINKEDIT and __DWARF if we have an in memory image and 6199 // this isn't a kernel binary like a kext or mach_kernel. 6200 const bool is_memory_image = (bool)m_process_wp.lock(); 6201 const Strata strata = GetStrata(); 6202 if (is_memory_image == false || strata == eStrataKernel) 6203 return false; 6204 } 6205 return true; 6206 } 6207 6208 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 6209 lldb::addr_t header_load_address, const Section *header_section, 6210 const Section *section) { 6211 ModuleSP module_sp = GetModule(); 6212 if (module_sp && header_section && section && 6213 header_load_address != LLDB_INVALID_ADDRESS) { 6214 lldb::addr_t file_addr = header_section->GetFileAddress(); 6215 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 6216 return section->GetFileAddress() - file_addr + header_load_address; 6217 } 6218 return LLDB_INVALID_ADDRESS; 6219 } 6220 6221 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 6222 bool value_is_offset) { 6223 ModuleSP module_sp = GetModule(); 6224 if (!module_sp) 6225 return false; 6226 6227 SectionList *section_list = GetSectionList(); 6228 if (!section_list) 6229 return false; 6230 6231 size_t num_loaded_sections = 0; 6232 const size_t num_sections = section_list->GetSize(); 6233 6234 if (value_is_offset) { 6235 // "value" is an offset to apply to each top level segment 6236 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6237 // Iterate through the object file sections to find all of the 6238 // sections that size on disk (to avoid __PAGEZERO) and load them 6239 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6240 if (SectionIsLoadable(section_sp.get())) 6241 if (target.GetSectionLoadList().SetSectionLoadAddress( 6242 section_sp, section_sp->GetFileAddress() + value)) 6243 ++num_loaded_sections; 6244 } 6245 } else { 6246 // "value" is the new base address of the mach_header, adjust each 6247 // section accordingly 6248 6249 Section *mach_header_section = GetMachHeaderSection(); 6250 if (mach_header_section) { 6251 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6252 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6253 6254 lldb::addr_t section_load_addr = 6255 CalculateSectionLoadAddressForMemoryImage( 6256 value, mach_header_section, section_sp.get()); 6257 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6258 if (target.GetSectionLoadList().SetSectionLoadAddress( 6259 section_sp, section_load_addr)) 6260 ++num_loaded_sections; 6261 } 6262 } 6263 } 6264 } 6265 return num_loaded_sections > 0; 6266 } 6267 6268 struct all_image_infos_header { 6269 uint32_t version; // currently 1 6270 uint32_t imgcount; // number of binary images 6271 uint64_t entries_fileoff; // file offset in the corefile of where the array of 6272 // struct entry's begin. 6273 uint32_t entries_size; // size of 'struct entry'. 6274 uint32_t unused; 6275 }; 6276 6277 struct image_entry { 6278 uint64_t filepath_offset; // offset in corefile to c-string of the file path, 6279 // UINT64_MAX if unavailable. 6280 uuid_t uuid; // uint8_t[16]. should be set to all zeroes if 6281 // uuid is unknown. 6282 uint64_t load_address; // UINT64_MAX if unknown. 6283 uint64_t seg_addrs_offset; // offset to the array of struct segment_vmaddr's. 6284 uint32_t segment_count; // The number of segments for this binary. 6285 uint32_t unused; 6286 6287 image_entry() { 6288 filepath_offset = UINT64_MAX; 6289 memset(&uuid, 0, sizeof(uuid_t)); 6290 segment_count = 0; 6291 load_address = UINT64_MAX; 6292 seg_addrs_offset = UINT64_MAX; 6293 unused = 0; 6294 } 6295 image_entry(const image_entry &rhs) { 6296 filepath_offset = rhs.filepath_offset; 6297 memcpy(&uuid, &rhs.uuid, sizeof(uuid_t)); 6298 segment_count = rhs.segment_count; 6299 seg_addrs_offset = rhs.seg_addrs_offset; 6300 load_address = rhs.load_address; 6301 unused = rhs.unused; 6302 } 6303 }; 6304 6305 struct segment_vmaddr { 6306 char segname[16]; 6307 uint64_t vmaddr; 6308 uint64_t unused; 6309 6310 segment_vmaddr() { 6311 memset(&segname, 0, 16); 6312 vmaddr = UINT64_MAX; 6313 unused = 0; 6314 } 6315 segment_vmaddr(const segment_vmaddr &rhs) { 6316 memcpy(&segname, &rhs.segname, 16); 6317 vmaddr = rhs.vmaddr; 6318 unused = rhs.unused; 6319 } 6320 }; 6321 6322 // Write the payload for the "all image infos" LC_NOTE into 6323 // the supplied all_image_infos_payload, assuming that this 6324 // will be written into the corefile starting at 6325 // initial_file_offset. 6326 // 6327 // The placement of this payload is a little tricky. We're 6328 // laying this out as 6329 // 6330 // 1. header (struct all_image_info_header) 6331 // 2. Array of fixed-size (struct image_entry)'s, one 6332 // per binary image present in the process. 6333 // 3. Arrays of (struct segment_vmaddr)'s, a varying number 6334 // for each binary image. 6335 // 4. Variable length c-strings of binary image filepaths, 6336 // one per binary. 6337 // 6338 // To compute where everything will be laid out in the 6339 // payload, we need to iterate over the images and calculate 6340 // how many segment_vmaddr structures each image will need, 6341 // and how long each image's filepath c-string is. There 6342 // are some multiple passes over the image list while calculating 6343 // everything. 6344 6345 static offset_t CreateAllImageInfosPayload( 6346 const lldb::ProcessSP &process_sp, offset_t initial_file_offset, 6347 StreamString &all_image_infos_payload, SaveCoreStyle core_style) { 6348 Target &target = process_sp->GetTarget(); 6349 ModuleList modules = target.GetImages(); 6350 6351 // stack-only corefiles have no reason to include binaries that 6352 // are not executing; we're trying to make the smallest corefile 6353 // we can, so leave the rest out. 6354 if (core_style == SaveCoreStyle::eSaveCoreStackOnly) 6355 modules.Clear(); 6356 6357 std::set<std::string> executing_uuids; 6358 ThreadList &thread_list(process_sp->GetThreadList()); 6359 for (uint32_t i = 0; i < thread_list.GetSize(); i++) { 6360 ThreadSP thread_sp = thread_list.GetThreadAtIndex(i); 6361 uint32_t stack_frame_count = thread_sp->GetStackFrameCount(); 6362 for (uint32_t j = 0; j < stack_frame_count; j++) { 6363 StackFrameSP stack_frame_sp = thread_sp->GetStackFrameAtIndex(j); 6364 Address pc = stack_frame_sp->GetFrameCodeAddress(); 6365 ModuleSP module_sp = pc.GetModule(); 6366 if (module_sp) { 6367 UUID uuid = module_sp->GetUUID(); 6368 if (uuid.IsValid()) { 6369 executing_uuids.insert(uuid.GetAsString()); 6370 modules.AppendIfNeeded(module_sp); 6371 } 6372 } 6373 } 6374 } 6375 size_t modules_count = modules.GetSize(); 6376 6377 struct all_image_infos_header infos; 6378 infos.version = 1; 6379 infos.imgcount = modules_count; 6380 infos.entries_size = sizeof(image_entry); 6381 infos.entries_fileoff = initial_file_offset + sizeof(all_image_infos_header); 6382 infos.unused = 0; 6383 6384 all_image_infos_payload.PutHex32(infos.version); 6385 all_image_infos_payload.PutHex32(infos.imgcount); 6386 all_image_infos_payload.PutHex64(infos.entries_fileoff); 6387 all_image_infos_payload.PutHex32(infos.entries_size); 6388 all_image_infos_payload.PutHex32(infos.unused); 6389 6390 // First create the structures for all of the segment name+vmaddr vectors 6391 // for each module, so we will know the size of them as we add the 6392 // module entries. 6393 std::vector<std::vector<segment_vmaddr>> modules_segment_vmaddrs; 6394 for (size_t i = 0; i < modules_count; i++) { 6395 ModuleSP module = modules.GetModuleAtIndex(i); 6396 6397 SectionList *sections = module->GetSectionList(); 6398 size_t sections_count = sections->GetSize(); 6399 std::vector<segment_vmaddr> segment_vmaddrs; 6400 for (size_t j = 0; j < sections_count; j++) { 6401 SectionSP section = sections->GetSectionAtIndex(j); 6402 if (!section->GetParent().get()) { 6403 addr_t vmaddr = section->GetLoadBaseAddress(&target); 6404 if (vmaddr == LLDB_INVALID_ADDRESS) 6405 continue; 6406 ConstString name = section->GetName(); 6407 segment_vmaddr seg_vmaddr; 6408 strncpy(seg_vmaddr.segname, name.AsCString(), 6409 sizeof(seg_vmaddr.segname)); 6410 seg_vmaddr.vmaddr = vmaddr; 6411 seg_vmaddr.unused = 0; 6412 segment_vmaddrs.push_back(seg_vmaddr); 6413 } 6414 } 6415 modules_segment_vmaddrs.push_back(segment_vmaddrs); 6416 } 6417 6418 offset_t size_of_vmaddr_structs = 0; 6419 for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) { 6420 size_of_vmaddr_structs += 6421 modules_segment_vmaddrs[i].size() * sizeof(segment_vmaddr); 6422 } 6423 6424 offset_t size_of_filepath_cstrings = 0; 6425 for (size_t i = 0; i < modules_count; i++) { 6426 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6427 size_of_filepath_cstrings += module_sp->GetFileSpec().GetPath().size() + 1; 6428 } 6429 6430 // Calculate the file offsets of our "all image infos" payload in the 6431 // corefile. initial_file_offset the original value passed in to this method. 6432 6433 offset_t start_of_entries = 6434 initial_file_offset + sizeof(all_image_infos_header); 6435 offset_t start_of_seg_vmaddrs = 6436 start_of_entries + sizeof(image_entry) * modules_count; 6437 offset_t start_of_filenames = start_of_seg_vmaddrs + size_of_vmaddr_structs; 6438 6439 offset_t final_file_offset = start_of_filenames + size_of_filepath_cstrings; 6440 6441 // Now write the one-per-module 'struct image_entry' into the 6442 // StringStream; keep track of where the struct segment_vmaddr 6443 // entries for each module will end up in the corefile. 6444 6445 offset_t current_string_offset = start_of_filenames; 6446 offset_t current_segaddrs_offset = start_of_seg_vmaddrs; 6447 std::vector<struct image_entry> image_entries; 6448 for (size_t i = 0; i < modules_count; i++) { 6449 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6450 6451 struct image_entry ent; 6452 memcpy(&ent.uuid, module_sp->GetUUID().GetBytes().data(), sizeof(ent.uuid)); 6453 if (modules_segment_vmaddrs[i].size() > 0) { 6454 ent.segment_count = modules_segment_vmaddrs[i].size(); 6455 ent.seg_addrs_offset = current_segaddrs_offset; 6456 } 6457 ent.filepath_offset = current_string_offset; 6458 ObjectFile *objfile = module_sp->GetObjectFile(); 6459 if (objfile) { 6460 Address base_addr(objfile->GetBaseAddress()); 6461 if (base_addr.IsValid()) { 6462 ent.load_address = base_addr.GetLoadAddress(&target); 6463 } 6464 } 6465 6466 all_image_infos_payload.PutHex64(ent.filepath_offset); 6467 all_image_infos_payload.PutRawBytes(ent.uuid, sizeof(ent.uuid)); 6468 all_image_infos_payload.PutHex64(ent.load_address); 6469 all_image_infos_payload.PutHex64(ent.seg_addrs_offset); 6470 all_image_infos_payload.PutHex32(ent.segment_count); 6471 6472 if (executing_uuids.find(module_sp->GetUUID().GetAsString()) != 6473 executing_uuids.end()) 6474 all_image_infos_payload.PutHex32(1); 6475 else 6476 all_image_infos_payload.PutHex32(0); 6477 6478 current_segaddrs_offset += ent.segment_count * sizeof(segment_vmaddr); 6479 current_string_offset += module_sp->GetFileSpec().GetPath().size() + 1; 6480 } 6481 6482 // Now write the struct segment_vmaddr entries into the StringStream. 6483 6484 for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) { 6485 if (modules_segment_vmaddrs[i].size() == 0) 6486 continue; 6487 for (struct segment_vmaddr segvm : modules_segment_vmaddrs[i]) { 6488 all_image_infos_payload.PutRawBytes(segvm.segname, sizeof(segvm.segname)); 6489 all_image_infos_payload.PutHex64(segvm.vmaddr); 6490 all_image_infos_payload.PutHex64(segvm.unused); 6491 } 6492 } 6493 6494 for (size_t i = 0; i < modules_count; i++) { 6495 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6496 std::string filepath = module_sp->GetFileSpec().GetPath(); 6497 all_image_infos_payload.PutRawBytes(filepath.data(), filepath.size() + 1); 6498 } 6499 6500 return final_file_offset; 6501 } 6502 6503 // Temp struct used to combine contiguous memory regions with 6504 // identical permissions. 6505 struct page_object { 6506 addr_t addr; 6507 addr_t size; 6508 uint32_t prot; 6509 }; 6510 6511 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6512 const FileSpec &outfile, 6513 lldb::SaveCoreStyle &core_style, Status &error) { 6514 if (!process_sp) 6515 return false; 6516 6517 // Default on macOS is to create a dirty-memory-only corefile. 6518 if (core_style == SaveCoreStyle::eSaveCoreUnspecified) { 6519 core_style = SaveCoreStyle::eSaveCoreDirtyOnly; 6520 } 6521 6522 Target &target = process_sp->GetTarget(); 6523 const ArchSpec target_arch = target.GetArchitecture(); 6524 const llvm::Triple &target_triple = target_arch.GetTriple(); 6525 if (target_triple.getVendor() == llvm::Triple::Apple && 6526 (target_triple.getOS() == llvm::Triple::MacOSX || 6527 target_triple.getOS() == llvm::Triple::IOS || 6528 target_triple.getOS() == llvm::Triple::WatchOS || 6529 target_triple.getOS() == llvm::Triple::TvOS)) { 6530 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) 6531 // { 6532 bool make_core = false; 6533 switch (target_arch.GetMachine()) { 6534 case llvm::Triple::aarch64: 6535 case llvm::Triple::aarch64_32: 6536 case llvm::Triple::arm: 6537 case llvm::Triple::thumb: 6538 case llvm::Triple::x86: 6539 case llvm::Triple::x86_64: 6540 make_core = true; 6541 break; 6542 default: 6543 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6544 target_triple.str().c_str()); 6545 break; 6546 } 6547 6548 if (make_core) { 6549 std::vector<llvm::MachO::segment_command_64> segment_load_commands; 6550 // uint32_t range_info_idx = 0; 6551 MemoryRegionInfo range_info; 6552 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6553 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6554 const ByteOrder byte_order = target_arch.GetByteOrder(); 6555 std::vector<page_object> pages_to_copy; 6556 6557 if (range_error.Success()) { 6558 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6559 // Calculate correct protections 6560 uint32_t prot = 0; 6561 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6562 prot |= VM_PROT_READ; 6563 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6564 prot |= VM_PROT_WRITE; 6565 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6566 prot |= VM_PROT_EXECUTE; 6567 6568 const addr_t addr = range_info.GetRange().GetRangeBase(); 6569 const addr_t size = range_info.GetRange().GetByteSize(); 6570 6571 if (size == 0) 6572 break; 6573 6574 bool include_this_region = true; 6575 bool dirty_pages_only = false; 6576 if (core_style == SaveCoreStyle::eSaveCoreStackOnly) { 6577 dirty_pages_only = true; 6578 if (range_info.IsStackMemory() != MemoryRegionInfo::eYes) { 6579 include_this_region = false; 6580 } 6581 } 6582 if (core_style == SaveCoreStyle::eSaveCoreDirtyOnly) { 6583 dirty_pages_only = true; 6584 } 6585 6586 if (prot != 0 && include_this_region) { 6587 addr_t pagesize = range_info.GetPageSize(); 6588 const llvm::Optional<std::vector<addr_t>> &dirty_page_list = 6589 range_info.GetDirtyPageList(); 6590 if (dirty_pages_only && dirty_page_list.hasValue()) { 6591 for (addr_t dirtypage : dirty_page_list.getValue()) { 6592 page_object obj; 6593 obj.addr = dirtypage; 6594 obj.size = pagesize; 6595 obj.prot = prot; 6596 pages_to_copy.push_back(obj); 6597 } 6598 } else { 6599 page_object obj; 6600 obj.addr = addr; 6601 obj.size = size; 6602 obj.prot = prot; 6603 pages_to_copy.push_back(obj); 6604 } 6605 } 6606 6607 range_error = process_sp->GetMemoryRegionInfo( 6608 range_info.GetRange().GetRangeEnd(), range_info); 6609 if (range_error.Fail()) 6610 break; 6611 } 6612 6613 // Combine contiguous entries that have the same 6614 // protections so we don't have an excess of 6615 // load commands. 6616 std::vector<page_object> combined_page_objects; 6617 page_object last_obj; 6618 last_obj.addr = LLDB_INVALID_ADDRESS; 6619 last_obj.size = 0; 6620 for (page_object obj : pages_to_copy) { 6621 if (last_obj.addr == LLDB_INVALID_ADDRESS) { 6622 last_obj = obj; 6623 continue; 6624 } 6625 if (last_obj.addr + last_obj.size == obj.addr && 6626 last_obj.prot == obj.prot) { 6627 last_obj.size += obj.size; 6628 continue; 6629 } 6630 combined_page_objects.push_back(last_obj); 6631 last_obj = obj; 6632 } 6633 // Add the last entry we were looking to combine 6634 // on to the array. 6635 if (last_obj.addr != LLDB_INVALID_ADDRESS && last_obj.size != 0) 6636 combined_page_objects.push_back(last_obj); 6637 6638 for (page_object obj : combined_page_objects) { 6639 uint32_t cmd_type = LC_SEGMENT_64; 6640 uint32_t segment_size = sizeof(llvm::MachO::segment_command_64); 6641 if (addr_byte_size == 4) { 6642 cmd_type = LC_SEGMENT; 6643 segment_size = sizeof(llvm::MachO::segment_command); 6644 } 6645 llvm::MachO::segment_command_64 segment = { 6646 cmd_type, // uint32_t cmd; 6647 segment_size, // uint32_t cmdsize; 6648 {0}, // char segname[16]; 6649 obj.addr, // uint64_t vmaddr; // uint32_t for 32-bit 6650 // Mach-O 6651 obj.size, // uint64_t vmsize; // uint32_t for 32-bit 6652 // Mach-O 6653 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6654 obj.size, // uint64_t filesize; // uint32_t for 32-bit 6655 // Mach-O 6656 obj.prot, // uint32_t maxprot; 6657 obj.prot, // uint32_t initprot; 6658 0, // uint32_t nsects; 6659 0}; // uint32_t flags; 6660 segment_load_commands.push_back(segment); 6661 } 6662 6663 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6664 6665 llvm::MachO::mach_header_64 mach_header; 6666 if (addr_byte_size == 8) { 6667 mach_header.magic = MH_MAGIC_64; 6668 } else { 6669 mach_header.magic = MH_MAGIC; 6670 } 6671 mach_header.cputype = target_arch.GetMachOCPUType(); 6672 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6673 mach_header.filetype = MH_CORE; 6674 mach_header.ncmds = segment_load_commands.size(); 6675 mach_header.flags = 0; 6676 mach_header.reserved = 0; 6677 ThreadList &thread_list = process_sp->GetThreadList(); 6678 const uint32_t num_threads = thread_list.GetSize(); 6679 6680 // Make an array of LC_THREAD data items. Each one contains the 6681 // contents of the LC_THREAD load command. The data doesn't contain 6682 // the load command + load command size, we will add the load command 6683 // and load command size as we emit the data. 6684 std::vector<StreamString> LC_THREAD_datas(num_threads); 6685 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6686 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6687 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6688 LC_THREAD_data.SetByteOrder(byte_order); 6689 } 6690 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) { 6691 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6692 if (thread_sp) { 6693 switch (mach_header.cputype) { 6694 case llvm::MachO::CPU_TYPE_ARM64: 6695 case llvm::MachO::CPU_TYPE_ARM64_32: 6696 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6697 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6698 break; 6699 6700 case llvm::MachO::CPU_TYPE_ARM: 6701 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6702 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6703 break; 6704 6705 case llvm::MachO::CPU_TYPE_I386: 6706 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6707 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6708 break; 6709 6710 case llvm::MachO::CPU_TYPE_X86_64: 6711 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6712 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6713 break; 6714 } 6715 } 6716 } 6717 6718 // The size of the load command is the size of the segments... 6719 if (addr_byte_size == 8) { 6720 mach_header.sizeofcmds = segment_load_commands.size() * 6721 sizeof(llvm::MachO::segment_command_64); 6722 } else { 6723 mach_header.sizeofcmds = segment_load_commands.size() * 6724 sizeof(llvm::MachO::segment_command); 6725 } 6726 6727 // and the size of all LC_THREAD load command 6728 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6729 ++mach_header.ncmds; 6730 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6731 } 6732 6733 // Bits will be set to indicate which bits are NOT used in 6734 // addressing in this process or 0 for unknown. 6735 uint64_t address_mask = process_sp->GetCodeAddressMask(); 6736 if (address_mask != 0) { 6737 // LC_NOTE "addrable bits" 6738 mach_header.ncmds++; 6739 mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); 6740 } 6741 6742 // LC_NOTE "all image infos" 6743 mach_header.ncmds++; 6744 mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); 6745 6746 // Write the mach header 6747 buffer.PutHex32(mach_header.magic); 6748 buffer.PutHex32(mach_header.cputype); 6749 buffer.PutHex32(mach_header.cpusubtype); 6750 buffer.PutHex32(mach_header.filetype); 6751 buffer.PutHex32(mach_header.ncmds); 6752 buffer.PutHex32(mach_header.sizeofcmds); 6753 buffer.PutHex32(mach_header.flags); 6754 if (addr_byte_size == 8) { 6755 buffer.PutHex32(mach_header.reserved); 6756 } 6757 6758 // Skip the mach header and all load commands and align to the next 6759 // 0x1000 byte boundary 6760 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6761 6762 file_offset = llvm::alignTo(file_offset, 16); 6763 std::vector<std::unique_ptr<LCNoteEntry>> lc_notes; 6764 6765 // Add "addrable bits" LC_NOTE when an address mask is available 6766 if (address_mask != 0) { 6767 std::unique_ptr<LCNoteEntry> addrable_bits_lcnote_up( 6768 new LCNoteEntry(addr_byte_size, byte_order)); 6769 addrable_bits_lcnote_up->name = "addrable bits"; 6770 addrable_bits_lcnote_up->payload_file_offset = file_offset; 6771 int bits = std::bitset<64>(~address_mask).count(); 6772 addrable_bits_lcnote_up->payload.PutHex32(3); // version 6773 addrable_bits_lcnote_up->payload.PutHex32( 6774 bits); // # of bits used for addressing 6775 addrable_bits_lcnote_up->payload.PutHex64(0); // unused 6776 6777 file_offset += addrable_bits_lcnote_up->payload.GetSize(); 6778 6779 lc_notes.push_back(std::move(addrable_bits_lcnote_up)); 6780 } 6781 6782 // Add "all image infos" LC_NOTE 6783 std::unique_ptr<LCNoteEntry> all_image_infos_lcnote_up( 6784 new LCNoteEntry(addr_byte_size, byte_order)); 6785 all_image_infos_lcnote_up->name = "all image infos"; 6786 all_image_infos_lcnote_up->payload_file_offset = file_offset; 6787 file_offset = CreateAllImageInfosPayload( 6788 process_sp, file_offset, all_image_infos_lcnote_up->payload, 6789 core_style); 6790 lc_notes.push_back(std::move(all_image_infos_lcnote_up)); 6791 6792 // Add LC_NOTE load commands 6793 for (auto &lcnote : lc_notes) { 6794 // Add the LC_NOTE load command to the file. 6795 buffer.PutHex32(LC_NOTE); 6796 buffer.PutHex32(sizeof(llvm::MachO::note_command)); 6797 char namebuf[16]; 6798 memset(namebuf, 0, sizeof(namebuf)); 6799 // this is the uncommon case where strncpy is exactly 6800 // the right one, doesn't need to be nul terminated. 6801 strncpy(namebuf, lcnote->name.c_str(), sizeof(namebuf)); 6802 buffer.PutRawBytes(namebuf, sizeof(namebuf)); 6803 buffer.PutHex64(lcnote->payload_file_offset); 6804 buffer.PutHex64(lcnote->payload.GetSize()); 6805 } 6806 6807 // Align to 4096-byte page boundary for the LC_SEGMENTs. 6808 file_offset = llvm::alignTo(file_offset, 4096); 6809 6810 for (auto &segment : segment_load_commands) { 6811 segment.fileoff = file_offset; 6812 file_offset += segment.filesize; 6813 } 6814 6815 // Write out all of the LC_THREAD load commands 6816 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6817 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6818 buffer.PutHex32(LC_THREAD); 6819 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6820 buffer.Write(LC_THREAD_data.GetString().data(), LC_THREAD_data_size); 6821 } 6822 6823 // Write out all of the segment load commands 6824 for (const auto &segment : segment_load_commands) { 6825 buffer.PutHex32(segment.cmd); 6826 buffer.PutHex32(segment.cmdsize); 6827 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6828 if (addr_byte_size == 8) { 6829 buffer.PutHex64(segment.vmaddr); 6830 buffer.PutHex64(segment.vmsize); 6831 buffer.PutHex64(segment.fileoff); 6832 buffer.PutHex64(segment.filesize); 6833 } else { 6834 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6835 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6836 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6837 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6838 } 6839 buffer.PutHex32(segment.maxprot); 6840 buffer.PutHex32(segment.initprot); 6841 buffer.PutHex32(segment.nsects); 6842 buffer.PutHex32(segment.flags); 6843 } 6844 6845 std::string core_file_path(outfile.GetPath()); 6846 auto core_file = FileSystem::Instance().Open( 6847 outfile, File::eOpenOptionWriteOnly | File::eOpenOptionTruncate | 6848 File::eOpenOptionCanCreate); 6849 if (!core_file) { 6850 error = core_file.takeError(); 6851 } else { 6852 // Read 1 page at a time 6853 uint8_t bytes[0x1000]; 6854 // Write the mach header and load commands out to the core file 6855 size_t bytes_written = buffer.GetString().size(); 6856 error = 6857 core_file.get()->Write(buffer.GetString().data(), bytes_written); 6858 if (error.Success()) { 6859 6860 for (auto &lcnote : lc_notes) { 6861 if (core_file.get()->SeekFromStart(lcnote->payload_file_offset) == 6862 -1) { 6863 error.SetErrorStringWithFormat("Unable to seek to corefile pos " 6864 "to write '%s' LC_NOTE payload", 6865 lcnote->name.c_str()); 6866 return false; 6867 } 6868 bytes_written = lcnote->payload.GetSize(); 6869 error = core_file.get()->Write(lcnote->payload.GetData(), 6870 bytes_written); 6871 if (!error.Success()) 6872 return false; 6873 } 6874 6875 // Now write the file data for all memory segments in the process 6876 for (const auto &segment : segment_load_commands) { 6877 if (core_file.get()->SeekFromStart(segment.fileoff) == -1) { 6878 error.SetErrorStringWithFormat( 6879 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6880 segment.fileoff, core_file_path.c_str()); 6881 break; 6882 } 6883 6884 target.GetDebugger().GetAsyncOutputStream()->Printf( 6885 "Saving %" PRId64 6886 " bytes of data for memory region at 0x%" PRIx64 "\n", 6887 segment.vmsize, segment.vmaddr); 6888 addr_t bytes_left = segment.vmsize; 6889 addr_t addr = segment.vmaddr; 6890 Status memory_read_error; 6891 while (bytes_left > 0 && error.Success()) { 6892 const size_t bytes_to_read = 6893 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6894 6895 // In a savecore setting, we don't really care about caching, 6896 // as the data is dumped and very likely never read again, 6897 // so we call ReadMemoryFromInferior to bypass it. 6898 const size_t bytes_read = process_sp->ReadMemoryFromInferior( 6899 addr, bytes, bytes_to_read, memory_read_error); 6900 6901 if (bytes_read == bytes_to_read) { 6902 size_t bytes_written = bytes_read; 6903 error = core_file.get()->Write(bytes, bytes_written); 6904 bytes_left -= bytes_read; 6905 addr += bytes_read; 6906 } else { 6907 // Some pages within regions are not readable, those should 6908 // be zero filled 6909 memset(bytes, 0, bytes_to_read); 6910 size_t bytes_written = bytes_to_read; 6911 error = core_file.get()->Write(bytes, bytes_written); 6912 bytes_left -= bytes_to_read; 6913 addr += bytes_to_read; 6914 } 6915 } 6916 } 6917 } 6918 } 6919 } else { 6920 error.SetErrorString( 6921 "process doesn't support getting memory region info"); 6922 } 6923 } 6924 return true; // This is the right plug to handle saving core files for 6925 // this process 6926 } 6927 return false; 6928 } 6929 6930 ObjectFileMachO::MachOCorefileAllImageInfos 6931 ObjectFileMachO::GetCorefileAllImageInfos() { 6932 MachOCorefileAllImageInfos image_infos; 6933 6934 // Look for an "all image infos" LC_NOTE. 6935 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 6936 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6937 const uint32_t cmd_offset = offset; 6938 llvm::MachO::load_command lc; 6939 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6940 break; 6941 if (lc.cmd == LC_NOTE) { 6942 char data_owner[17]; 6943 m_data.CopyData(offset, 16, data_owner); 6944 data_owner[16] = '\0'; 6945 offset += 16; 6946 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 6947 offset += 4; /* size unused */ 6948 6949 if (strcmp("all image infos", data_owner) == 0) { 6950 offset = fileoff; 6951 // Read the struct all_image_infos_header. 6952 uint32_t version = m_data.GetU32(&offset); 6953 if (version != 1) { 6954 return image_infos; 6955 } 6956 uint32_t imgcount = m_data.GetU32(&offset); 6957 uint64_t entries_fileoff = m_data.GetU64(&offset); 6958 offset += 4; // uint32_t entries_size; 6959 offset += 4; // uint32_t unused; 6960 6961 offset = entries_fileoff; 6962 for (uint32_t i = 0; i < imgcount; i++) { 6963 // Read the struct image_entry. 6964 offset_t filepath_offset = m_data.GetU64(&offset); 6965 uuid_t uuid; 6966 memcpy(&uuid, m_data.GetData(&offset, sizeof(uuid_t)), 6967 sizeof(uuid_t)); 6968 uint64_t load_address = m_data.GetU64(&offset); 6969 offset_t seg_addrs_offset = m_data.GetU64(&offset); 6970 uint32_t segment_count = m_data.GetU32(&offset); 6971 uint32_t currently_executing = m_data.GetU32(&offset); 6972 6973 MachOCorefileImageEntry image_entry; 6974 image_entry.filename = (const char *)m_data.GetCStr(&filepath_offset); 6975 image_entry.uuid = UUID::fromData(uuid, sizeof(uuid_t)); 6976 image_entry.load_address = load_address; 6977 image_entry.currently_executing = currently_executing; 6978 6979 offset_t seg_vmaddrs_offset = seg_addrs_offset; 6980 for (uint32_t j = 0; j < segment_count; j++) { 6981 char segname[17]; 6982 m_data.CopyData(seg_vmaddrs_offset, 16, segname); 6983 segname[16] = '\0'; 6984 seg_vmaddrs_offset += 16; 6985 uint64_t vmaddr = m_data.GetU64(&seg_vmaddrs_offset); 6986 seg_vmaddrs_offset += 8; /* unused */ 6987 6988 std::tuple<ConstString, addr_t> new_seg{ConstString(segname), 6989 vmaddr}; 6990 image_entry.segment_load_addresses.push_back(new_seg); 6991 } 6992 image_infos.all_image_infos.push_back(image_entry); 6993 } 6994 } 6995 } 6996 offset = cmd_offset + lc.cmdsize; 6997 } 6998 6999 return image_infos; 7000 } 7001 7002 bool ObjectFileMachO::LoadCoreFileImages(lldb_private::Process &process) { 7003 MachOCorefileAllImageInfos image_infos = GetCorefileAllImageInfos(); 7004 bool added_images = false; 7005 if (image_infos.IsValid()) { 7006 for (const MachOCorefileImageEntry &image : image_infos.all_image_infos) { 7007 ModuleSpec module_spec; 7008 module_spec.GetUUID() = image.uuid; 7009 module_spec.GetFileSpec() = FileSpec(image.filename.c_str()); 7010 if (image.currently_executing) { 7011 Symbols::DownloadObjectAndSymbolFile(module_spec, true); 7012 if (FileSystem::Instance().Exists(module_spec.GetFileSpec())) { 7013 process.GetTarget().GetOrCreateModule(module_spec, false); 7014 } 7015 } 7016 Status error; 7017 ModuleSP module_sp = 7018 process.GetTarget().GetOrCreateModule(module_spec, false, &error); 7019 if (!module_sp.get() || !module_sp->GetObjectFile()) { 7020 if (image.load_address != LLDB_INVALID_ADDRESS) { 7021 module_sp = process.ReadModuleFromMemory(module_spec.GetFileSpec(), 7022 image.load_address); 7023 } 7024 } 7025 if (module_sp.get() && module_sp->GetObjectFile()) { 7026 added_images = true; 7027 if (module_sp->GetObjectFile()->GetType() == 7028 ObjectFile::eTypeExecutable) { 7029 process.GetTarget().SetExecutableModule(module_sp, eLoadDependentsNo); 7030 } 7031 for (auto name_vmaddr_tuple : image.segment_load_addresses) { 7032 SectionList *sectlist = module_sp->GetObjectFile()->GetSectionList(); 7033 if (sectlist) { 7034 SectionSP sect_sp = 7035 sectlist->FindSectionByName(std::get<0>(name_vmaddr_tuple)); 7036 if (sect_sp) { 7037 process.GetTarget().SetSectionLoadAddress( 7038 sect_sp, std::get<1>(name_vmaddr_tuple)); 7039 } 7040 } 7041 } 7042 } 7043 } 7044 } 7045 return added_images; 7046 } 7047