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/Symbol/DWARFCallFrameInfo.h" 25 #include "lldb/Symbol/LocateSymbolFile.h" 26 #include "lldb/Symbol/ObjectFile.h" 27 #include "lldb/Target/DynamicLoader.h" 28 #include "lldb/Target/MemoryRegionInfo.h" 29 #include "lldb/Target/Platform.h" 30 #include "lldb/Target/Process.h" 31 #include "lldb/Target/SectionLoadList.h" 32 #include "lldb/Target/Target.h" 33 #include "lldb/Target/Thread.h" 34 #include "lldb/Target/ThreadList.h" 35 #include "lldb/Utility/ArchSpec.h" 36 #include "lldb/Utility/DataBuffer.h" 37 #include "lldb/Utility/FileSpec.h" 38 #include "lldb/Utility/Log.h" 39 #include "lldb/Utility/RangeMap.h" 40 #include "lldb/Utility/RegisterValue.h" 41 #include "lldb/Utility/Status.h" 42 #include "lldb/Utility/StreamString.h" 43 #include "lldb/Utility/Timer.h" 44 #include "lldb/Utility/UUID.h" 45 46 #include "lldb/Host/SafeMachO.h" 47 48 #include "llvm/ADT/DenseSet.h" 49 #include "llvm/Support/FormatVariadic.h" 50 #include "llvm/Support/MemoryBuffer.h" 51 52 #include "ObjectFileMachO.h" 53 54 #if defined(__APPLE__) 55 #include <TargetConditionals.h> 56 // GetLLDBSharedCacheUUID() needs to call dlsym() 57 #include <dlfcn.h> 58 #endif 59 60 #ifndef __APPLE__ 61 #include "Utility/UuidCompatibility.h" 62 #else 63 #include <uuid/uuid.h> 64 #endif 65 66 #include <bitset> 67 #include <memory> 68 69 // Unfortunately the signpost header pulls in the system MachO header, too. 70 #ifdef CPU_TYPE_ARM 71 #undef CPU_TYPE_ARM 72 #endif 73 #ifdef CPU_TYPE_ARM64 74 #undef CPU_TYPE_ARM64 75 #endif 76 #ifdef CPU_TYPE_ARM64_32 77 #undef CPU_TYPE_ARM64_32 78 #endif 79 #ifdef CPU_TYPE_I386 80 #undef CPU_TYPE_I386 81 #endif 82 #ifdef CPU_TYPE_X86_64 83 #undef CPU_TYPE_X86_64 84 #endif 85 #ifdef MH_DYLINKER 86 #undef MH_DYLINKER 87 #endif 88 #ifdef MH_OBJECT 89 #undef MH_OBJECT 90 #endif 91 #ifdef LC_VERSION_MIN_MACOSX 92 #undef LC_VERSION_MIN_MACOSX 93 #endif 94 #ifdef LC_VERSION_MIN_IPHONEOS 95 #undef LC_VERSION_MIN_IPHONEOS 96 #endif 97 #ifdef LC_VERSION_MIN_TVOS 98 #undef LC_VERSION_MIN_TVOS 99 #endif 100 #ifdef LC_VERSION_MIN_WATCHOS 101 #undef LC_VERSION_MIN_WATCHOS 102 #endif 103 #ifdef LC_BUILD_VERSION 104 #undef LC_BUILD_VERSION 105 #endif 106 #ifdef PLATFORM_MACOS 107 #undef PLATFORM_MACOS 108 #endif 109 #ifdef PLATFORM_MACCATALYST 110 #undef PLATFORM_MACCATALYST 111 #endif 112 #ifdef PLATFORM_IOS 113 #undef PLATFORM_IOS 114 #endif 115 #ifdef PLATFORM_IOSSIMULATOR 116 #undef PLATFORM_IOSSIMULATOR 117 #endif 118 #ifdef PLATFORM_TVOS 119 #undef PLATFORM_TVOS 120 #endif 121 #ifdef PLATFORM_TVOSSIMULATOR 122 #undef PLATFORM_TVOSSIMULATOR 123 #endif 124 #ifdef PLATFORM_WATCHOS 125 #undef PLATFORM_WATCHOS 126 #endif 127 #ifdef PLATFORM_WATCHOSSIMULATOR 128 #undef PLATFORM_WATCHOSSIMULATOR 129 #endif 130 131 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 132 using namespace lldb; 133 using namespace lldb_private; 134 using namespace llvm::MachO; 135 136 LLDB_PLUGIN_DEFINE(ObjectFileMachO) 137 138 // Some structure definitions needed for parsing the dyld shared cache files 139 // found on iOS devices. 140 141 struct lldb_copy_dyld_cache_header_v1 { 142 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 143 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 144 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 145 uint32_t imagesOffset; 146 uint32_t imagesCount; 147 uint64_t dyldBaseAddress; 148 uint64_t codeSignatureOffset; 149 uint64_t codeSignatureSize; 150 uint64_t slideInfoOffset; 151 uint64_t slideInfoSize; 152 uint64_t localSymbolsOffset; 153 uint64_t localSymbolsSize; 154 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 155 // and later 156 }; 157 158 struct lldb_copy_dyld_cache_mapping_info { 159 uint64_t address; 160 uint64_t size; 161 uint64_t fileOffset; 162 uint32_t maxProt; 163 uint32_t initProt; 164 }; 165 166 struct lldb_copy_dyld_cache_local_symbols_info { 167 uint32_t nlistOffset; 168 uint32_t nlistCount; 169 uint32_t stringsOffset; 170 uint32_t stringsSize; 171 uint32_t entriesOffset; 172 uint32_t entriesCount; 173 }; 174 struct lldb_copy_dyld_cache_local_symbols_entry { 175 uint32_t dylibOffset; 176 uint32_t nlistStartIndex; 177 uint32_t nlistCount; 178 }; 179 180 static void PrintRegisterValue(RegisterContext *reg_ctx, const char *name, 181 const char *alt_name, size_t reg_byte_size, 182 Stream &data) { 183 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 184 if (reg_info == nullptr) 185 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 186 if (reg_info) { 187 lldb_private::RegisterValue reg_value; 188 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 189 if (reg_info->byte_size >= reg_byte_size) 190 data.Write(reg_value.GetBytes(), reg_byte_size); 191 else { 192 data.Write(reg_value.GetBytes(), reg_info->byte_size); 193 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; ++i) 194 data.PutChar(0); 195 } 196 return; 197 } 198 } 199 // Just write zeros if all else fails 200 for (size_t i = 0; i < reg_byte_size; ++i) 201 data.PutChar(0); 202 } 203 204 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 205 public: 206 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 207 const DataExtractor &data) 208 : RegisterContextDarwin_x86_64(thread, 0) { 209 SetRegisterDataFrom_LC_THREAD(data); 210 } 211 212 void InvalidateAllRegisters() override { 213 // Do nothing... registers are always valid... 214 } 215 216 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 217 lldb::offset_t offset = 0; 218 SetError(GPRRegSet, Read, -1); 219 SetError(FPURegSet, Read, -1); 220 SetError(EXCRegSet, Read, -1); 221 bool done = false; 222 223 while (!done) { 224 int flavor = data.GetU32(&offset); 225 if (flavor == 0) 226 done = true; 227 else { 228 uint32_t i; 229 uint32_t count = data.GetU32(&offset); 230 switch (flavor) { 231 case GPRRegSet: 232 for (i = 0; i < count; ++i) 233 (&gpr.rax)[i] = data.GetU64(&offset); 234 SetError(GPRRegSet, Read, 0); 235 done = true; 236 237 break; 238 case FPURegSet: 239 // TODO: fill in FPU regs.... 240 // SetError (FPURegSet, Read, -1); 241 done = true; 242 243 break; 244 case EXCRegSet: 245 exc.trapno = data.GetU32(&offset); 246 exc.err = data.GetU32(&offset); 247 exc.faultvaddr = data.GetU64(&offset); 248 SetError(EXCRegSet, Read, 0); 249 done = true; 250 break; 251 case 7: 252 case 8: 253 case 9: 254 // fancy flavors that encapsulate of the above flavors... 255 break; 256 257 default: 258 done = true; 259 break; 260 } 261 } 262 } 263 } 264 265 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 266 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 267 if (reg_ctx_sp) { 268 RegisterContext *reg_ctx = reg_ctx_sp.get(); 269 270 data.PutHex32(GPRRegSet); // Flavor 271 data.PutHex32(GPRWordCount); 272 PrintRegisterValue(reg_ctx, "rax", nullptr, 8, data); 273 PrintRegisterValue(reg_ctx, "rbx", nullptr, 8, data); 274 PrintRegisterValue(reg_ctx, "rcx", nullptr, 8, data); 275 PrintRegisterValue(reg_ctx, "rdx", nullptr, 8, data); 276 PrintRegisterValue(reg_ctx, "rdi", nullptr, 8, data); 277 PrintRegisterValue(reg_ctx, "rsi", nullptr, 8, data); 278 PrintRegisterValue(reg_ctx, "rbp", nullptr, 8, data); 279 PrintRegisterValue(reg_ctx, "rsp", nullptr, 8, data); 280 PrintRegisterValue(reg_ctx, "r8", nullptr, 8, data); 281 PrintRegisterValue(reg_ctx, "r9", nullptr, 8, data); 282 PrintRegisterValue(reg_ctx, "r10", nullptr, 8, data); 283 PrintRegisterValue(reg_ctx, "r11", nullptr, 8, data); 284 PrintRegisterValue(reg_ctx, "r12", nullptr, 8, data); 285 PrintRegisterValue(reg_ctx, "r13", nullptr, 8, data); 286 PrintRegisterValue(reg_ctx, "r14", nullptr, 8, data); 287 PrintRegisterValue(reg_ctx, "r15", nullptr, 8, data); 288 PrintRegisterValue(reg_ctx, "rip", nullptr, 8, data); 289 PrintRegisterValue(reg_ctx, "rflags", nullptr, 8, data); 290 PrintRegisterValue(reg_ctx, "cs", nullptr, 8, data); 291 PrintRegisterValue(reg_ctx, "fs", nullptr, 8, data); 292 PrintRegisterValue(reg_ctx, "gs", nullptr, 8, data); 293 294 // // Write out the FPU registers 295 // const size_t fpu_byte_size = sizeof(FPU); 296 // size_t bytes_written = 0; 297 // data.PutHex32 (FPURegSet); 298 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 299 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 300 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 301 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 302 // data); // uint16_t fcw; // "fctrl" 303 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 304 // data); // uint16_t fsw; // "fstat" 305 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 306 // data); // uint8_t ftw; // "ftag" 307 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 308 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 309 // data); // uint16_t fop; // "fop" 310 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 311 // data); // uint32_t ip; // "fioff" 312 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 313 // data); // uint16_t cs; // "fiseg" 314 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 315 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 316 // data); // uint32_t dp; // "fooff" 317 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 318 // data); // uint16_t ds; // "foseg" 319 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 320 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 321 // data); // uint32_t mxcsr; 322 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 323 // 4, data);// uint32_t mxcsrmask; 324 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 325 // sizeof(MMSReg), data); 326 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 327 // sizeof(MMSReg), data); 328 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 329 // sizeof(MMSReg), data); 330 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 331 // sizeof(MMSReg), data); 332 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 333 // sizeof(MMSReg), data); 334 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 335 // sizeof(MMSReg), data); 336 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 337 // sizeof(MMSReg), data); 338 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 339 // sizeof(MMSReg), data); 340 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 341 // sizeof(XMMReg), data); 342 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 343 // sizeof(XMMReg), data); 344 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 345 // sizeof(XMMReg), data); 346 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 347 // sizeof(XMMReg), data); 348 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 349 // sizeof(XMMReg), data); 350 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 351 // sizeof(XMMReg), data); 352 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 353 // sizeof(XMMReg), data); 354 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 355 // sizeof(XMMReg), data); 356 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 357 // sizeof(XMMReg), data); 358 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 359 // sizeof(XMMReg), data); 360 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 361 // sizeof(XMMReg), data); 362 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 363 // sizeof(XMMReg), data); 364 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 365 // sizeof(XMMReg), data); 366 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 367 // sizeof(XMMReg), data); 368 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 369 // sizeof(XMMReg), data); 370 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 371 // sizeof(XMMReg), data); 372 // 373 // // Fill rest with zeros 374 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 375 // i) 376 // data.PutChar(0); 377 378 // Write out the EXC registers 379 data.PutHex32(EXCRegSet); 380 data.PutHex32(EXCWordCount); 381 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 382 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 383 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 8, data); 384 return true; 385 } 386 return false; 387 } 388 389 protected: 390 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 391 392 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 393 394 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 395 396 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 397 return 0; 398 } 399 400 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 401 return 0; 402 } 403 404 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 405 return 0; 406 } 407 }; 408 409 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 410 public: 411 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 412 const DataExtractor &data) 413 : RegisterContextDarwin_i386(thread, 0) { 414 SetRegisterDataFrom_LC_THREAD(data); 415 } 416 417 void InvalidateAllRegisters() override { 418 // Do nothing... registers are always valid... 419 } 420 421 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 422 lldb::offset_t offset = 0; 423 SetError(GPRRegSet, Read, -1); 424 SetError(FPURegSet, Read, -1); 425 SetError(EXCRegSet, Read, -1); 426 bool done = false; 427 428 while (!done) { 429 int flavor = data.GetU32(&offset); 430 if (flavor == 0) 431 done = true; 432 else { 433 uint32_t i; 434 uint32_t count = data.GetU32(&offset); 435 switch (flavor) { 436 case GPRRegSet: 437 for (i = 0; i < count; ++i) 438 (&gpr.eax)[i] = data.GetU32(&offset); 439 SetError(GPRRegSet, Read, 0); 440 done = true; 441 442 break; 443 case FPURegSet: 444 // TODO: fill in FPU regs.... 445 // SetError (FPURegSet, Read, -1); 446 done = true; 447 448 break; 449 case EXCRegSet: 450 exc.trapno = data.GetU32(&offset); 451 exc.err = data.GetU32(&offset); 452 exc.faultvaddr = data.GetU32(&offset); 453 SetError(EXCRegSet, Read, 0); 454 done = true; 455 break; 456 case 7: 457 case 8: 458 case 9: 459 // fancy flavors that encapsulate of the above flavors... 460 break; 461 462 default: 463 done = true; 464 break; 465 } 466 } 467 } 468 } 469 470 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 471 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 472 if (reg_ctx_sp) { 473 RegisterContext *reg_ctx = reg_ctx_sp.get(); 474 475 data.PutHex32(GPRRegSet); // Flavor 476 data.PutHex32(GPRWordCount); 477 PrintRegisterValue(reg_ctx, "eax", nullptr, 4, data); 478 PrintRegisterValue(reg_ctx, "ebx", nullptr, 4, data); 479 PrintRegisterValue(reg_ctx, "ecx", nullptr, 4, data); 480 PrintRegisterValue(reg_ctx, "edx", nullptr, 4, data); 481 PrintRegisterValue(reg_ctx, "edi", nullptr, 4, data); 482 PrintRegisterValue(reg_ctx, "esi", nullptr, 4, data); 483 PrintRegisterValue(reg_ctx, "ebp", nullptr, 4, data); 484 PrintRegisterValue(reg_ctx, "esp", nullptr, 4, data); 485 PrintRegisterValue(reg_ctx, "ss", nullptr, 4, data); 486 PrintRegisterValue(reg_ctx, "eflags", nullptr, 4, data); 487 PrintRegisterValue(reg_ctx, "eip", nullptr, 4, data); 488 PrintRegisterValue(reg_ctx, "cs", nullptr, 4, data); 489 PrintRegisterValue(reg_ctx, "ds", nullptr, 4, data); 490 PrintRegisterValue(reg_ctx, "es", nullptr, 4, data); 491 PrintRegisterValue(reg_ctx, "fs", nullptr, 4, data); 492 PrintRegisterValue(reg_ctx, "gs", nullptr, 4, data); 493 494 // Write out the EXC registers 495 data.PutHex32(EXCRegSet); 496 data.PutHex32(EXCWordCount); 497 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 498 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 499 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 4, data); 500 return true; 501 } 502 return false; 503 } 504 505 protected: 506 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 507 508 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 509 510 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 511 512 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 513 return 0; 514 } 515 516 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 517 return 0; 518 } 519 520 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 521 return 0; 522 } 523 }; 524 525 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 526 public: 527 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 528 const DataExtractor &data) 529 : RegisterContextDarwin_arm(thread, 0) { 530 SetRegisterDataFrom_LC_THREAD(data); 531 } 532 533 void InvalidateAllRegisters() override { 534 // Do nothing... registers are always valid... 535 } 536 537 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 538 lldb::offset_t offset = 0; 539 SetError(GPRRegSet, Read, -1); 540 SetError(FPURegSet, Read, -1); 541 SetError(EXCRegSet, Read, -1); 542 bool done = false; 543 544 while (!done) { 545 int flavor = data.GetU32(&offset); 546 uint32_t count = data.GetU32(&offset); 547 lldb::offset_t next_thread_state = offset + (count * 4); 548 switch (flavor) { 549 case GPRAltRegSet: 550 case GPRRegSet: 551 // On ARM, the CPSR register is also included in the count but it is 552 // not included in gpr.r so loop until (count-1). 553 for (uint32_t i = 0; i < (count - 1); ++i) { 554 gpr.r[i] = data.GetU32(&offset); 555 } 556 // Save cpsr explicitly. 557 gpr.cpsr = data.GetU32(&offset); 558 559 SetError(GPRRegSet, Read, 0); 560 offset = next_thread_state; 561 break; 562 563 case FPURegSet: { 564 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 565 const int fpu_reg_buf_size = sizeof(fpu.floats); 566 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 567 fpu_reg_buf) == fpu_reg_buf_size) { 568 offset += fpu_reg_buf_size; 569 fpu.fpscr = data.GetU32(&offset); 570 SetError(FPURegSet, Read, 0); 571 } else { 572 done = true; 573 } 574 } 575 offset = next_thread_state; 576 break; 577 578 case EXCRegSet: 579 if (count == 3) { 580 exc.exception = data.GetU32(&offset); 581 exc.fsr = data.GetU32(&offset); 582 exc.far = data.GetU32(&offset); 583 SetError(EXCRegSet, Read, 0); 584 } 585 done = true; 586 offset = next_thread_state; 587 break; 588 589 // Unknown register set flavor, stop trying to parse. 590 default: 591 done = true; 592 } 593 } 594 } 595 596 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 597 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 598 if (reg_ctx_sp) { 599 RegisterContext *reg_ctx = reg_ctx_sp.get(); 600 601 data.PutHex32(GPRRegSet); // Flavor 602 data.PutHex32(GPRWordCount); 603 PrintRegisterValue(reg_ctx, "r0", nullptr, 4, data); 604 PrintRegisterValue(reg_ctx, "r1", nullptr, 4, data); 605 PrintRegisterValue(reg_ctx, "r2", nullptr, 4, data); 606 PrintRegisterValue(reg_ctx, "r3", nullptr, 4, data); 607 PrintRegisterValue(reg_ctx, "r4", nullptr, 4, data); 608 PrintRegisterValue(reg_ctx, "r5", nullptr, 4, data); 609 PrintRegisterValue(reg_ctx, "r6", nullptr, 4, data); 610 PrintRegisterValue(reg_ctx, "r7", nullptr, 4, data); 611 PrintRegisterValue(reg_ctx, "r8", nullptr, 4, data); 612 PrintRegisterValue(reg_ctx, "r9", nullptr, 4, data); 613 PrintRegisterValue(reg_ctx, "r10", nullptr, 4, data); 614 PrintRegisterValue(reg_ctx, "r11", nullptr, 4, data); 615 PrintRegisterValue(reg_ctx, "r12", nullptr, 4, data); 616 PrintRegisterValue(reg_ctx, "sp", nullptr, 4, data); 617 PrintRegisterValue(reg_ctx, "lr", nullptr, 4, data); 618 PrintRegisterValue(reg_ctx, "pc", nullptr, 4, data); 619 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 620 621 // Write out the EXC registers 622 // data.PutHex32 (EXCRegSet); 623 // data.PutHex32 (EXCWordCount); 624 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 625 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 626 // WriteRegister (reg_ctx, "far", NULL, 4, data); 627 return true; 628 } 629 return false; 630 } 631 632 protected: 633 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 634 635 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 636 637 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 638 639 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 640 641 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 642 return 0; 643 } 644 645 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 646 return 0; 647 } 648 649 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 650 return 0; 651 } 652 653 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 654 return -1; 655 } 656 }; 657 658 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 659 public: 660 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 661 const DataExtractor &data) 662 : RegisterContextDarwin_arm64(thread, 0) { 663 SetRegisterDataFrom_LC_THREAD(data); 664 } 665 666 void InvalidateAllRegisters() override { 667 // Do nothing... registers are always valid... 668 } 669 670 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 671 lldb::offset_t offset = 0; 672 SetError(GPRRegSet, Read, -1); 673 SetError(FPURegSet, Read, -1); 674 SetError(EXCRegSet, Read, -1); 675 bool done = false; 676 while (!done) { 677 int flavor = data.GetU32(&offset); 678 uint32_t count = data.GetU32(&offset); 679 lldb::offset_t next_thread_state = offset + (count * 4); 680 switch (flavor) { 681 case GPRRegSet: 682 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 683 // 32-bit register) 684 if (count >= (33 * 2) + 1) { 685 for (uint32_t i = 0; i < 29; ++i) 686 gpr.x[i] = data.GetU64(&offset); 687 gpr.fp = data.GetU64(&offset); 688 gpr.lr = data.GetU64(&offset); 689 gpr.sp = data.GetU64(&offset); 690 gpr.pc = data.GetU64(&offset); 691 gpr.cpsr = data.GetU32(&offset); 692 SetError(GPRRegSet, Read, 0); 693 } 694 offset = next_thread_state; 695 break; 696 case FPURegSet: { 697 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 698 const int fpu_reg_buf_size = sizeof(fpu); 699 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 700 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 701 fpu_reg_buf) == fpu_reg_buf_size) { 702 SetError(FPURegSet, Read, 0); 703 } else { 704 done = true; 705 } 706 } 707 offset = next_thread_state; 708 break; 709 case EXCRegSet: 710 if (count == 4) { 711 exc.far = data.GetU64(&offset); 712 exc.esr = data.GetU32(&offset); 713 exc.exception = data.GetU32(&offset); 714 SetError(EXCRegSet, Read, 0); 715 } 716 offset = next_thread_state; 717 break; 718 default: 719 done = true; 720 break; 721 } 722 } 723 } 724 725 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 726 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 727 if (reg_ctx_sp) { 728 RegisterContext *reg_ctx = reg_ctx_sp.get(); 729 730 data.PutHex32(GPRRegSet); // Flavor 731 data.PutHex32(GPRWordCount); 732 PrintRegisterValue(reg_ctx, "x0", nullptr, 8, data); 733 PrintRegisterValue(reg_ctx, "x1", nullptr, 8, data); 734 PrintRegisterValue(reg_ctx, "x2", nullptr, 8, data); 735 PrintRegisterValue(reg_ctx, "x3", nullptr, 8, data); 736 PrintRegisterValue(reg_ctx, "x4", nullptr, 8, data); 737 PrintRegisterValue(reg_ctx, "x5", nullptr, 8, data); 738 PrintRegisterValue(reg_ctx, "x6", nullptr, 8, data); 739 PrintRegisterValue(reg_ctx, "x7", nullptr, 8, data); 740 PrintRegisterValue(reg_ctx, "x8", nullptr, 8, data); 741 PrintRegisterValue(reg_ctx, "x9", nullptr, 8, data); 742 PrintRegisterValue(reg_ctx, "x10", nullptr, 8, data); 743 PrintRegisterValue(reg_ctx, "x11", nullptr, 8, data); 744 PrintRegisterValue(reg_ctx, "x12", nullptr, 8, data); 745 PrintRegisterValue(reg_ctx, "x13", nullptr, 8, data); 746 PrintRegisterValue(reg_ctx, "x14", nullptr, 8, data); 747 PrintRegisterValue(reg_ctx, "x15", nullptr, 8, data); 748 PrintRegisterValue(reg_ctx, "x16", nullptr, 8, data); 749 PrintRegisterValue(reg_ctx, "x17", nullptr, 8, data); 750 PrintRegisterValue(reg_ctx, "x18", nullptr, 8, data); 751 PrintRegisterValue(reg_ctx, "x19", nullptr, 8, data); 752 PrintRegisterValue(reg_ctx, "x20", nullptr, 8, data); 753 PrintRegisterValue(reg_ctx, "x21", nullptr, 8, data); 754 PrintRegisterValue(reg_ctx, "x22", nullptr, 8, data); 755 PrintRegisterValue(reg_ctx, "x23", nullptr, 8, data); 756 PrintRegisterValue(reg_ctx, "x24", nullptr, 8, data); 757 PrintRegisterValue(reg_ctx, "x25", nullptr, 8, data); 758 PrintRegisterValue(reg_ctx, "x26", nullptr, 8, data); 759 PrintRegisterValue(reg_ctx, "x27", nullptr, 8, data); 760 PrintRegisterValue(reg_ctx, "x28", nullptr, 8, data); 761 PrintRegisterValue(reg_ctx, "fp", nullptr, 8, data); 762 PrintRegisterValue(reg_ctx, "lr", nullptr, 8, data); 763 PrintRegisterValue(reg_ctx, "sp", nullptr, 8, data); 764 PrintRegisterValue(reg_ctx, "pc", nullptr, 8, data); 765 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 766 data.PutHex32(0); // uint32_t pad at the end 767 768 // Write out the EXC registers 769 data.PutHex32(EXCRegSet); 770 data.PutHex32(EXCWordCount); 771 PrintRegisterValue(reg_ctx, "far", NULL, 8, data); 772 PrintRegisterValue(reg_ctx, "esr", NULL, 4, data); 773 PrintRegisterValue(reg_ctx, "exception", NULL, 4, data); 774 return true; 775 } 776 return false; 777 } 778 779 protected: 780 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 781 782 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 783 784 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 785 786 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 787 788 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 789 return 0; 790 } 791 792 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 793 return 0; 794 } 795 796 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 797 return 0; 798 } 799 800 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 801 return -1; 802 } 803 }; 804 805 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 806 switch (magic) { 807 case MH_MAGIC: 808 case MH_CIGAM: 809 return sizeof(struct llvm::MachO::mach_header); 810 811 case MH_MAGIC_64: 812 case MH_CIGAM_64: 813 return sizeof(struct llvm::MachO::mach_header_64); 814 break; 815 816 default: 817 break; 818 } 819 return 0; 820 } 821 822 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 823 824 char ObjectFileMachO::ID; 825 826 void ObjectFileMachO::Initialize() { 827 PluginManager::RegisterPlugin( 828 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 829 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 830 } 831 832 void ObjectFileMachO::Terminate() { 833 PluginManager::UnregisterPlugin(CreateInstance); 834 } 835 836 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 837 static ConstString g_name("mach-o"); 838 return g_name; 839 } 840 841 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 842 return "Mach-o object file reader (32 and 64 bit)"; 843 } 844 845 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 846 DataBufferSP &data_sp, 847 lldb::offset_t data_offset, 848 const FileSpec *file, 849 lldb::offset_t file_offset, 850 lldb::offset_t length) { 851 if (!data_sp) { 852 data_sp = MapFileData(*file, length, file_offset); 853 if (!data_sp) 854 return nullptr; 855 data_offset = 0; 856 } 857 858 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 859 return nullptr; 860 861 // Update the data to contain the entire file if it doesn't already 862 if (data_sp->GetByteSize() < length) { 863 data_sp = MapFileData(*file, length, file_offset); 864 if (!data_sp) 865 return nullptr; 866 data_offset = 0; 867 } 868 auto objfile_up = std::make_unique<ObjectFileMachO>( 869 module_sp, data_sp, data_offset, file, file_offset, length); 870 if (!objfile_up || !objfile_up->ParseHeader()) 871 return nullptr; 872 873 return objfile_up.release(); 874 } 875 876 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 877 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 878 const ProcessSP &process_sp, lldb::addr_t header_addr) { 879 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 880 std::unique_ptr<ObjectFile> objfile_up( 881 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 882 if (objfile_up.get() && objfile_up->ParseHeader()) 883 return objfile_up.release(); 884 } 885 return nullptr; 886 } 887 888 size_t ObjectFileMachO::GetModuleSpecifications( 889 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 890 lldb::offset_t data_offset, lldb::offset_t file_offset, 891 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 892 const size_t initial_count = specs.GetSize(); 893 894 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 895 DataExtractor data; 896 data.SetData(data_sp); 897 llvm::MachO::mach_header header; 898 if (ParseHeader(data, &data_offset, header)) { 899 size_t header_and_load_cmds = 900 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 901 if (header_and_load_cmds >= data_sp->GetByteSize()) { 902 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 903 data.SetData(data_sp); 904 data_offset = MachHeaderSizeFromMagic(header.magic); 905 } 906 if (data_sp) { 907 ModuleSpec base_spec; 908 base_spec.GetFileSpec() = file; 909 base_spec.SetObjectOffset(file_offset); 910 base_spec.SetObjectSize(length); 911 GetAllArchSpecs(header, data, data_offset, base_spec, specs); 912 } 913 } 914 } 915 return specs.GetSize() - initial_count; 916 } 917 918 ConstString ObjectFileMachO::GetSegmentNameTEXT() { 919 static ConstString g_segment_name_TEXT("__TEXT"); 920 return g_segment_name_TEXT; 921 } 922 923 ConstString ObjectFileMachO::GetSegmentNameDATA() { 924 static ConstString g_segment_name_DATA("__DATA"); 925 return g_segment_name_DATA; 926 } 927 928 ConstString ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 929 static ConstString g_segment_name("__DATA_DIRTY"); 930 return g_segment_name; 931 } 932 933 ConstString ObjectFileMachO::GetSegmentNameDATA_CONST() { 934 static ConstString g_segment_name("__DATA_CONST"); 935 return g_segment_name; 936 } 937 938 ConstString ObjectFileMachO::GetSegmentNameOBJC() { 939 static ConstString g_segment_name_OBJC("__OBJC"); 940 return g_segment_name_OBJC; 941 } 942 943 ConstString ObjectFileMachO::GetSegmentNameLINKEDIT() { 944 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 945 return g_section_name_LINKEDIT; 946 } 947 948 ConstString ObjectFileMachO::GetSegmentNameDWARF() { 949 static ConstString g_section_name("__DWARF"); 950 return g_section_name; 951 } 952 953 ConstString ObjectFileMachO::GetSectionNameEHFrame() { 954 static ConstString g_section_name_eh_frame("__eh_frame"); 955 return g_section_name_eh_frame; 956 } 957 958 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 959 lldb::addr_t data_offset, 960 lldb::addr_t data_length) { 961 DataExtractor data; 962 data.SetData(data_sp, data_offset, data_length); 963 lldb::offset_t offset = 0; 964 uint32_t magic = data.GetU32(&offset); 965 return MachHeaderSizeFromMagic(magic) != 0; 966 } 967 968 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 969 DataBufferSP &data_sp, 970 lldb::offset_t data_offset, 971 const FileSpec *file, 972 lldb::offset_t file_offset, 973 lldb::offset_t length) 974 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 975 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 976 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 977 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 978 ::memset(&m_header, 0, sizeof(m_header)); 979 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 980 } 981 982 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 983 lldb::DataBufferSP &header_data_sp, 984 const lldb::ProcessSP &process_sp, 985 lldb::addr_t header_addr) 986 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 987 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 988 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 989 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 990 ::memset(&m_header, 0, sizeof(m_header)); 991 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 992 } 993 994 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 995 lldb::offset_t *data_offset_ptr, 996 llvm::MachO::mach_header &header) { 997 data.SetByteOrder(endian::InlHostByteOrder()); 998 // Leave magic in the original byte order 999 header.magic = data.GetU32(data_offset_ptr); 1000 bool can_parse = false; 1001 bool is_64_bit = false; 1002 switch (header.magic) { 1003 case MH_MAGIC: 1004 data.SetByteOrder(endian::InlHostByteOrder()); 1005 data.SetAddressByteSize(4); 1006 can_parse = true; 1007 break; 1008 1009 case MH_MAGIC_64: 1010 data.SetByteOrder(endian::InlHostByteOrder()); 1011 data.SetAddressByteSize(8); 1012 can_parse = true; 1013 is_64_bit = true; 1014 break; 1015 1016 case MH_CIGAM: 1017 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1018 ? eByteOrderLittle 1019 : eByteOrderBig); 1020 data.SetAddressByteSize(4); 1021 can_parse = true; 1022 break; 1023 1024 case MH_CIGAM_64: 1025 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1026 ? eByteOrderLittle 1027 : eByteOrderBig); 1028 data.SetAddressByteSize(8); 1029 is_64_bit = true; 1030 can_parse = true; 1031 break; 1032 1033 default: 1034 break; 1035 } 1036 1037 if (can_parse) { 1038 data.GetU32(data_offset_ptr, &header.cputype, 6); 1039 if (is_64_bit) 1040 *data_offset_ptr += 4; 1041 return true; 1042 } else { 1043 memset(&header, 0, sizeof(header)); 1044 } 1045 return false; 1046 } 1047 1048 bool ObjectFileMachO::ParseHeader() { 1049 ModuleSP module_sp(GetModule()); 1050 if (!module_sp) 1051 return false; 1052 1053 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1054 bool can_parse = false; 1055 lldb::offset_t offset = 0; 1056 m_data.SetByteOrder(endian::InlHostByteOrder()); 1057 // Leave magic in the original byte order 1058 m_header.magic = m_data.GetU32(&offset); 1059 switch (m_header.magic) { 1060 case MH_MAGIC: 1061 m_data.SetByteOrder(endian::InlHostByteOrder()); 1062 m_data.SetAddressByteSize(4); 1063 can_parse = true; 1064 break; 1065 1066 case MH_MAGIC_64: 1067 m_data.SetByteOrder(endian::InlHostByteOrder()); 1068 m_data.SetAddressByteSize(8); 1069 can_parse = true; 1070 break; 1071 1072 case MH_CIGAM: 1073 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1074 ? eByteOrderLittle 1075 : eByteOrderBig); 1076 m_data.SetAddressByteSize(4); 1077 can_parse = true; 1078 break; 1079 1080 case MH_CIGAM_64: 1081 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1082 ? eByteOrderLittle 1083 : eByteOrderBig); 1084 m_data.SetAddressByteSize(8); 1085 can_parse = true; 1086 break; 1087 1088 default: 1089 break; 1090 } 1091 1092 if (can_parse) { 1093 m_data.GetU32(&offset, &m_header.cputype, 6); 1094 1095 ModuleSpecList all_specs; 1096 ModuleSpec base_spec; 1097 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 1098 base_spec, all_specs); 1099 1100 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 1101 ArchSpec mach_arch = 1102 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 1103 1104 // Check if the module has a required architecture 1105 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1106 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1107 continue; 1108 1109 if (SetModulesArchitecture(mach_arch)) { 1110 const size_t header_and_lc_size = 1111 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1112 if (m_data.GetByteSize() < header_and_lc_size) { 1113 DataBufferSP data_sp; 1114 ProcessSP process_sp(m_process_wp.lock()); 1115 if (process_sp) { 1116 data_sp = ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1117 } else { 1118 // Read in all only the load command data from the file on disk 1119 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset); 1120 if (data_sp->GetByteSize() != header_and_lc_size) 1121 continue; 1122 } 1123 if (data_sp) 1124 m_data.SetData(data_sp); 1125 } 1126 } 1127 return true; 1128 } 1129 // None found. 1130 return false; 1131 } else { 1132 memset(&m_header, 0, sizeof(struct llvm::MachO::mach_header)); 1133 } 1134 return false; 1135 } 1136 1137 ByteOrder ObjectFileMachO::GetByteOrder() const { 1138 return m_data.GetByteOrder(); 1139 } 1140 1141 bool ObjectFileMachO::IsExecutable() const { 1142 return m_header.filetype == MH_EXECUTE; 1143 } 1144 1145 bool ObjectFileMachO::IsDynamicLoader() const { 1146 return m_header.filetype == MH_DYLINKER; 1147 } 1148 1149 bool ObjectFileMachO::IsSharedCacheBinary() const { 1150 return m_header.flags & MH_DYLIB_IN_CACHE; 1151 } 1152 1153 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1154 return m_data.GetAddressByteSize(); 1155 } 1156 1157 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1158 Symtab *symtab = GetSymtab(); 1159 if (!symtab) 1160 return AddressClass::eUnknown; 1161 1162 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1163 if (symbol) { 1164 if (symbol->ValueIsAddress()) { 1165 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1166 if (section_sp) { 1167 const lldb::SectionType section_type = section_sp->GetType(); 1168 switch (section_type) { 1169 case eSectionTypeInvalid: 1170 return AddressClass::eUnknown; 1171 1172 case eSectionTypeCode: 1173 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1174 // For ARM we have a bit in the n_desc field of the symbol that 1175 // tells us ARM/Thumb which is bit 0x0008. 1176 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1177 return AddressClass::eCodeAlternateISA; 1178 } 1179 return AddressClass::eCode; 1180 1181 case eSectionTypeContainer: 1182 return AddressClass::eUnknown; 1183 1184 case eSectionTypeData: 1185 case eSectionTypeDataCString: 1186 case eSectionTypeDataCStringPointers: 1187 case eSectionTypeDataSymbolAddress: 1188 case eSectionTypeData4: 1189 case eSectionTypeData8: 1190 case eSectionTypeData16: 1191 case eSectionTypeDataPointers: 1192 case eSectionTypeZeroFill: 1193 case eSectionTypeDataObjCMessageRefs: 1194 case eSectionTypeDataObjCCFStrings: 1195 case eSectionTypeGoSymtab: 1196 return AddressClass::eData; 1197 1198 case eSectionTypeDebug: 1199 case eSectionTypeDWARFDebugAbbrev: 1200 case eSectionTypeDWARFDebugAbbrevDwo: 1201 case eSectionTypeDWARFDebugAddr: 1202 case eSectionTypeDWARFDebugAranges: 1203 case eSectionTypeDWARFDebugCuIndex: 1204 case eSectionTypeDWARFDebugFrame: 1205 case eSectionTypeDWARFDebugInfo: 1206 case eSectionTypeDWARFDebugInfoDwo: 1207 case eSectionTypeDWARFDebugLine: 1208 case eSectionTypeDWARFDebugLineStr: 1209 case eSectionTypeDWARFDebugLoc: 1210 case eSectionTypeDWARFDebugLocDwo: 1211 case eSectionTypeDWARFDebugLocLists: 1212 case eSectionTypeDWARFDebugLocListsDwo: 1213 case eSectionTypeDWARFDebugMacInfo: 1214 case eSectionTypeDWARFDebugMacro: 1215 case eSectionTypeDWARFDebugNames: 1216 case eSectionTypeDWARFDebugPubNames: 1217 case eSectionTypeDWARFDebugPubTypes: 1218 case eSectionTypeDWARFDebugRanges: 1219 case eSectionTypeDWARFDebugRngLists: 1220 case eSectionTypeDWARFDebugRngListsDwo: 1221 case eSectionTypeDWARFDebugStr: 1222 case eSectionTypeDWARFDebugStrDwo: 1223 case eSectionTypeDWARFDebugStrOffsets: 1224 case eSectionTypeDWARFDebugStrOffsetsDwo: 1225 case eSectionTypeDWARFDebugTuIndex: 1226 case eSectionTypeDWARFDebugTypes: 1227 case eSectionTypeDWARFDebugTypesDwo: 1228 case eSectionTypeDWARFAppleNames: 1229 case eSectionTypeDWARFAppleTypes: 1230 case eSectionTypeDWARFAppleNamespaces: 1231 case eSectionTypeDWARFAppleObjC: 1232 case eSectionTypeDWARFGNUDebugAltLink: 1233 return AddressClass::eDebug; 1234 1235 case eSectionTypeEHFrame: 1236 case eSectionTypeARMexidx: 1237 case eSectionTypeARMextab: 1238 case eSectionTypeCompactUnwind: 1239 return AddressClass::eRuntime; 1240 1241 case eSectionTypeAbsoluteAddress: 1242 case eSectionTypeELFSymbolTable: 1243 case eSectionTypeELFDynamicSymbols: 1244 case eSectionTypeELFRelocationEntries: 1245 case eSectionTypeELFDynamicLinkInfo: 1246 case eSectionTypeOther: 1247 return AddressClass::eUnknown; 1248 } 1249 } 1250 } 1251 1252 const SymbolType symbol_type = symbol->GetType(); 1253 switch (symbol_type) { 1254 case eSymbolTypeAny: 1255 return AddressClass::eUnknown; 1256 case eSymbolTypeAbsolute: 1257 return AddressClass::eUnknown; 1258 1259 case eSymbolTypeCode: 1260 case eSymbolTypeTrampoline: 1261 case eSymbolTypeResolver: 1262 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1263 // For ARM we have a bit in the n_desc field of the symbol that tells 1264 // us ARM/Thumb which is bit 0x0008. 1265 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1266 return AddressClass::eCodeAlternateISA; 1267 } 1268 return AddressClass::eCode; 1269 1270 case eSymbolTypeData: 1271 return AddressClass::eData; 1272 case eSymbolTypeRuntime: 1273 return AddressClass::eRuntime; 1274 case eSymbolTypeException: 1275 return AddressClass::eRuntime; 1276 case eSymbolTypeSourceFile: 1277 return AddressClass::eDebug; 1278 case eSymbolTypeHeaderFile: 1279 return AddressClass::eDebug; 1280 case eSymbolTypeObjectFile: 1281 return AddressClass::eDebug; 1282 case eSymbolTypeCommonBlock: 1283 return AddressClass::eDebug; 1284 case eSymbolTypeBlock: 1285 return AddressClass::eDebug; 1286 case eSymbolTypeLocal: 1287 return AddressClass::eData; 1288 case eSymbolTypeParam: 1289 return AddressClass::eData; 1290 case eSymbolTypeVariable: 1291 return AddressClass::eData; 1292 case eSymbolTypeVariableType: 1293 return AddressClass::eDebug; 1294 case eSymbolTypeLineEntry: 1295 return AddressClass::eDebug; 1296 case eSymbolTypeLineHeader: 1297 return AddressClass::eDebug; 1298 case eSymbolTypeScopeBegin: 1299 return AddressClass::eDebug; 1300 case eSymbolTypeScopeEnd: 1301 return AddressClass::eDebug; 1302 case eSymbolTypeAdditional: 1303 return AddressClass::eUnknown; 1304 case eSymbolTypeCompiler: 1305 return AddressClass::eDebug; 1306 case eSymbolTypeInstrumentation: 1307 return AddressClass::eDebug; 1308 case eSymbolTypeUndefined: 1309 return AddressClass::eUnknown; 1310 case eSymbolTypeObjCClass: 1311 return AddressClass::eRuntime; 1312 case eSymbolTypeObjCMetaClass: 1313 return AddressClass::eRuntime; 1314 case eSymbolTypeObjCIVar: 1315 return AddressClass::eRuntime; 1316 case eSymbolTypeReExported: 1317 return AddressClass::eRuntime; 1318 } 1319 } 1320 return AddressClass::eUnknown; 1321 } 1322 1323 Symtab *ObjectFileMachO::GetSymtab() { 1324 ModuleSP module_sp(GetModule()); 1325 if (module_sp) { 1326 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1327 if (m_symtab_up == nullptr) { 1328 m_symtab_up = std::make_unique<Symtab>(this); 1329 std::lock_guard<std::recursive_mutex> symtab_guard( 1330 m_symtab_up->GetMutex()); 1331 ParseSymtab(); 1332 m_symtab_up->Finalize(); 1333 } 1334 } 1335 return m_symtab_up.get(); 1336 } 1337 1338 bool ObjectFileMachO::IsStripped() { 1339 if (m_dysymtab.cmd == 0) { 1340 ModuleSP module_sp(GetModule()); 1341 if (module_sp) { 1342 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1343 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1344 const lldb::offset_t load_cmd_offset = offset; 1345 1346 llvm::MachO::load_command lc; 1347 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 1348 break; 1349 if (lc.cmd == LC_DYSYMTAB) { 1350 m_dysymtab.cmd = lc.cmd; 1351 m_dysymtab.cmdsize = lc.cmdsize; 1352 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1353 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1354 nullptr) { 1355 // Clear m_dysymtab if we were unable to read all items from the 1356 // load command 1357 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1358 } 1359 } 1360 offset = load_cmd_offset + lc.cmdsize; 1361 } 1362 } 1363 } 1364 if (m_dysymtab.cmd) 1365 return m_dysymtab.nlocalsym <= 1; 1366 return false; 1367 } 1368 1369 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1370 EncryptedFileRanges result; 1371 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1372 1373 llvm::MachO::encryption_info_command encryption_cmd; 1374 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1375 const lldb::offset_t load_cmd_offset = offset; 1376 if (m_data.GetU32(&offset, &encryption_cmd, 2) == nullptr) 1377 break; 1378 1379 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the 1380 // 3 fields we care about, so treat them the same. 1381 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1382 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1383 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1384 if (encryption_cmd.cryptid != 0) { 1385 EncryptedFileRanges::Entry entry; 1386 entry.SetRangeBase(encryption_cmd.cryptoff); 1387 entry.SetByteSize(encryption_cmd.cryptsize); 1388 result.Append(entry); 1389 } 1390 } 1391 } 1392 offset = load_cmd_offset + encryption_cmd.cmdsize; 1393 } 1394 1395 return result; 1396 } 1397 1398 void ObjectFileMachO::SanitizeSegmentCommand( 1399 llvm::MachO::segment_command_64 &seg_cmd, uint32_t cmd_idx) { 1400 if (m_length == 0 || seg_cmd.filesize == 0) 1401 return; 1402 1403 if (IsSharedCacheBinary() && !IsInMemory()) { 1404 // In shared cache images, the load commands are relative to the 1405 // shared cache file, and not the specific image we are 1406 // examining. Let's fix this up so that it looks like a normal 1407 // image. 1408 if (strncmp(seg_cmd.segname, "__TEXT", sizeof(seg_cmd.segname)) == 0) 1409 m_text_address = seg_cmd.vmaddr; 1410 if (strncmp(seg_cmd.segname, "__LINKEDIT", sizeof(seg_cmd.segname)) == 0) 1411 m_linkedit_original_offset = seg_cmd.fileoff; 1412 1413 seg_cmd.fileoff = seg_cmd.vmaddr - m_text_address; 1414 } 1415 1416 if (seg_cmd.fileoff > m_length) { 1417 // We have a load command that says it extends past the end of the file. 1418 // This is likely a corrupt file. We don't have any way to return an error 1419 // condition here (this method was likely invoked from something like 1420 // ObjectFile::GetSectionList()), so we just null out the section contents, 1421 // and dump a message to stdout. The most common case here is core file 1422 // debugging with a truncated file. 1423 const char *lc_segment_name = 1424 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1425 GetModule()->ReportWarning( 1426 "load command %u %s has a fileoff (0x%" PRIx64 1427 ") that extends beyond the end of the file (0x%" PRIx64 1428 "), ignoring this section", 1429 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1430 1431 seg_cmd.fileoff = 0; 1432 seg_cmd.filesize = 0; 1433 } 1434 1435 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1436 // We have a load command that says it extends past the end of the file. 1437 // This is likely a corrupt file. We don't have any way to return an error 1438 // condition here (this method was likely invoked from something like 1439 // ObjectFile::GetSectionList()), so we just null out the section contents, 1440 // and dump a message to stdout. The most common case here is core file 1441 // debugging with a truncated file. 1442 const char *lc_segment_name = 1443 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1444 GetModule()->ReportWarning( 1445 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1446 ") that extends beyond the end of the file (0x%" PRIx64 1447 "), the segment will be truncated to match", 1448 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1449 1450 // Truncate the length 1451 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1452 } 1453 } 1454 1455 static uint32_t 1456 GetSegmentPermissions(const llvm::MachO::segment_command_64 &seg_cmd) { 1457 uint32_t result = 0; 1458 if (seg_cmd.initprot & VM_PROT_READ) 1459 result |= ePermissionsReadable; 1460 if (seg_cmd.initprot & VM_PROT_WRITE) 1461 result |= ePermissionsWritable; 1462 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1463 result |= ePermissionsExecutable; 1464 return result; 1465 } 1466 1467 static lldb::SectionType GetSectionType(uint32_t flags, 1468 ConstString section_name) { 1469 1470 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1471 return eSectionTypeCode; 1472 1473 uint32_t mach_sect_type = flags & SECTION_TYPE; 1474 static ConstString g_sect_name_objc_data("__objc_data"); 1475 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1476 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1477 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1478 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1479 static ConstString g_sect_name_objc_const("__objc_const"); 1480 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1481 static ConstString g_sect_name_cfstring("__cfstring"); 1482 1483 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1484 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1485 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1486 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1487 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1488 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1489 static ConstString g_sect_name_dwarf_debug_loclists("__debug_loclists"); 1490 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1491 static ConstString g_sect_name_dwarf_debug_names("__debug_names"); 1492 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1493 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1494 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1495 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1496 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1497 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1498 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1499 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1500 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1501 static ConstString g_sect_name_eh_frame("__eh_frame"); 1502 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1503 static ConstString g_sect_name_text("__text"); 1504 static ConstString g_sect_name_data("__data"); 1505 static ConstString g_sect_name_go_symtab("__gosymtab"); 1506 1507 if (section_name == g_sect_name_dwarf_debug_abbrev) 1508 return eSectionTypeDWARFDebugAbbrev; 1509 if (section_name == g_sect_name_dwarf_debug_aranges) 1510 return eSectionTypeDWARFDebugAranges; 1511 if (section_name == g_sect_name_dwarf_debug_frame) 1512 return eSectionTypeDWARFDebugFrame; 1513 if (section_name == g_sect_name_dwarf_debug_info) 1514 return eSectionTypeDWARFDebugInfo; 1515 if (section_name == g_sect_name_dwarf_debug_line) 1516 return eSectionTypeDWARFDebugLine; 1517 if (section_name == g_sect_name_dwarf_debug_loc) 1518 return eSectionTypeDWARFDebugLoc; 1519 if (section_name == g_sect_name_dwarf_debug_loclists) 1520 return eSectionTypeDWARFDebugLocLists; 1521 if (section_name == g_sect_name_dwarf_debug_macinfo) 1522 return eSectionTypeDWARFDebugMacInfo; 1523 if (section_name == g_sect_name_dwarf_debug_names) 1524 return eSectionTypeDWARFDebugNames; 1525 if (section_name == g_sect_name_dwarf_debug_pubnames) 1526 return eSectionTypeDWARFDebugPubNames; 1527 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1528 return eSectionTypeDWARFDebugPubTypes; 1529 if (section_name == g_sect_name_dwarf_debug_ranges) 1530 return eSectionTypeDWARFDebugRanges; 1531 if (section_name == g_sect_name_dwarf_debug_str) 1532 return eSectionTypeDWARFDebugStr; 1533 if (section_name == g_sect_name_dwarf_debug_types) 1534 return eSectionTypeDWARFDebugTypes; 1535 if (section_name == g_sect_name_dwarf_apple_names) 1536 return eSectionTypeDWARFAppleNames; 1537 if (section_name == g_sect_name_dwarf_apple_types) 1538 return eSectionTypeDWARFAppleTypes; 1539 if (section_name == g_sect_name_dwarf_apple_namespaces) 1540 return eSectionTypeDWARFAppleNamespaces; 1541 if (section_name == g_sect_name_dwarf_apple_objc) 1542 return eSectionTypeDWARFAppleObjC; 1543 if (section_name == g_sect_name_objc_selrefs) 1544 return eSectionTypeDataCStringPointers; 1545 if (section_name == g_sect_name_objc_msgrefs) 1546 return eSectionTypeDataObjCMessageRefs; 1547 if (section_name == g_sect_name_eh_frame) 1548 return eSectionTypeEHFrame; 1549 if (section_name == g_sect_name_compact_unwind) 1550 return eSectionTypeCompactUnwind; 1551 if (section_name == g_sect_name_cfstring) 1552 return eSectionTypeDataObjCCFStrings; 1553 if (section_name == g_sect_name_go_symtab) 1554 return eSectionTypeGoSymtab; 1555 if (section_name == g_sect_name_objc_data || 1556 section_name == g_sect_name_objc_classrefs || 1557 section_name == g_sect_name_objc_superrefs || 1558 section_name == g_sect_name_objc_const || 1559 section_name == g_sect_name_objc_classlist) { 1560 return eSectionTypeDataPointers; 1561 } 1562 1563 switch (mach_sect_type) { 1564 // TODO: categorize sections by other flags for regular sections 1565 case S_REGULAR: 1566 if (section_name == g_sect_name_text) 1567 return eSectionTypeCode; 1568 if (section_name == g_sect_name_data) 1569 return eSectionTypeData; 1570 return eSectionTypeOther; 1571 case S_ZEROFILL: 1572 return eSectionTypeZeroFill; 1573 case S_CSTRING_LITERALS: // section with only literal C strings 1574 return eSectionTypeDataCString; 1575 case S_4BYTE_LITERALS: // section with only 4 byte literals 1576 return eSectionTypeData4; 1577 case S_8BYTE_LITERALS: // section with only 8 byte literals 1578 return eSectionTypeData8; 1579 case S_LITERAL_POINTERS: // section with only pointers to literals 1580 return eSectionTypeDataPointers; 1581 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1582 return eSectionTypeDataPointers; 1583 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1584 return eSectionTypeDataPointers; 1585 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1586 // the reserved2 field 1587 return eSectionTypeCode; 1588 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1589 // initialization 1590 return eSectionTypeDataPointers; 1591 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1592 // termination 1593 return eSectionTypeDataPointers; 1594 case S_COALESCED: 1595 return eSectionTypeOther; 1596 case S_GB_ZEROFILL: 1597 return eSectionTypeZeroFill; 1598 case S_INTERPOSING: // section with only pairs of function pointers for 1599 // interposing 1600 return eSectionTypeCode; 1601 case S_16BYTE_LITERALS: // section with only 16 byte literals 1602 return eSectionTypeData16; 1603 case S_DTRACE_DOF: 1604 return eSectionTypeDebug; 1605 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1606 return eSectionTypeDataPointers; 1607 default: 1608 return eSectionTypeOther; 1609 } 1610 } 1611 1612 struct ObjectFileMachO::SegmentParsingContext { 1613 const EncryptedFileRanges EncryptedRanges; 1614 lldb_private::SectionList &UnifiedList; 1615 uint32_t NextSegmentIdx = 0; 1616 uint32_t NextSectionIdx = 0; 1617 bool FileAddressesChanged = false; 1618 1619 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1620 lldb_private::SectionList &UnifiedList) 1621 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1622 }; 1623 1624 void ObjectFileMachO::ProcessSegmentCommand( 1625 const llvm::MachO::load_command &load_cmd_, lldb::offset_t offset, 1626 uint32_t cmd_idx, SegmentParsingContext &context) { 1627 llvm::MachO::segment_command_64 load_cmd; 1628 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1629 1630 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1631 return; 1632 1633 ModuleSP module_sp = GetModule(); 1634 const bool is_core = GetType() == eTypeCoreFile; 1635 const bool is_dsym = (m_header.filetype == MH_DSYM); 1636 bool add_section = true; 1637 bool add_to_unified = true; 1638 ConstString const_segname( 1639 load_cmd.segname, strnlen(load_cmd.segname, sizeof(load_cmd.segname))); 1640 1641 SectionSP unified_section_sp( 1642 context.UnifiedList.FindSectionByName(const_segname)); 1643 if (is_dsym && unified_section_sp) { 1644 if (const_segname == GetSegmentNameLINKEDIT()) { 1645 // We need to keep the __LINKEDIT segment private to this object file 1646 // only 1647 add_to_unified = false; 1648 } else { 1649 // This is the dSYM file and this section has already been created by the 1650 // object file, no need to create it. 1651 add_section = false; 1652 } 1653 } 1654 load_cmd.vmaddr = m_data.GetAddress(&offset); 1655 load_cmd.vmsize = m_data.GetAddress(&offset); 1656 load_cmd.fileoff = m_data.GetAddress(&offset); 1657 load_cmd.filesize = m_data.GetAddress(&offset); 1658 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1659 return; 1660 1661 SanitizeSegmentCommand(load_cmd, cmd_idx); 1662 1663 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1664 const bool segment_is_encrypted = 1665 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1666 1667 // Keep a list of mach segments around in case we need to get at data that 1668 // isn't stored in the abstracted Sections. 1669 m_mach_segments.push_back(load_cmd); 1670 1671 // Use a segment ID of the segment index shifted left by 8 so they never 1672 // conflict with any of the sections. 1673 SectionSP segment_sp; 1674 if (add_section && (const_segname || is_core)) { 1675 segment_sp = std::make_shared<Section>( 1676 module_sp, // Module to which this section belongs 1677 this, // Object file to which this sections belongs 1678 ++context.NextSegmentIdx 1679 << 8, // Section ID is the 1 based segment index 1680 // shifted right by 8 bits as not to collide with any of the 256 1681 // section IDs that are possible 1682 const_segname, // Name of this section 1683 eSectionTypeContainer, // This section is a container of other 1684 // sections. 1685 load_cmd.vmaddr, // File VM address == addresses as they are 1686 // found in the object file 1687 load_cmd.vmsize, // VM size in bytes of this section 1688 load_cmd.fileoff, // Offset to the data for this section in 1689 // the file 1690 load_cmd.filesize, // Size in bytes of this section as found 1691 // in the file 1692 0, // Segments have no alignment information 1693 load_cmd.flags); // Flags for this section 1694 1695 segment_sp->SetIsEncrypted(segment_is_encrypted); 1696 m_sections_up->AddSection(segment_sp); 1697 segment_sp->SetPermissions(segment_permissions); 1698 if (add_to_unified) 1699 context.UnifiedList.AddSection(segment_sp); 1700 } else if (unified_section_sp) { 1701 // If this is a dSYM and the file addresses in the dSYM differ from the 1702 // file addresses in the ObjectFile, we must use the file base address for 1703 // the Section from the dSYM for the DWARF to resolve correctly. 1704 // This only happens with binaries in the shared cache in practice; 1705 // normally a mismatch like this would give a binary & dSYM that do not 1706 // match UUIDs. When a binary is included in the shared cache, its 1707 // segments are rearranged to optimize the shared cache, so its file 1708 // addresses will differ from what the ObjectFile had originally, 1709 // and what the dSYM has. 1710 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1711 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 1712 if (log) { 1713 log->Printf( 1714 "Installing dSYM's %s segment file address over ObjectFile's " 1715 "so symbol table/debug info resolves correctly for %s", 1716 const_segname.AsCString(), 1717 module_sp->GetFileSpec().GetFilename().AsCString()); 1718 } 1719 1720 // Make sure we've parsed the symbol table from the ObjectFile before 1721 // we go around changing its Sections. 1722 module_sp->GetObjectFile()->GetSymtab(); 1723 // eh_frame would present the same problems but we parse that on a per- 1724 // function basis as-needed so it's more difficult to remove its use of 1725 // the Sections. Realistically, the environments where this code path 1726 // will be taken will not have eh_frame sections. 1727 1728 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1729 1730 // Notify the module that the section addresses have been changed once 1731 // we're done so any file-address caches can be updated. 1732 context.FileAddressesChanged = true; 1733 } 1734 m_sections_up->AddSection(unified_section_sp); 1735 } 1736 1737 llvm::MachO::section_64 sect64; 1738 ::memset(§64, 0, sizeof(sect64)); 1739 // Push a section into our mach sections for the section at index zero 1740 // (NO_SECT) if we don't have any mach sections yet... 1741 if (m_mach_sections.empty()) 1742 m_mach_sections.push_back(sect64); 1743 uint32_t segment_sect_idx; 1744 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1745 1746 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1747 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1748 ++segment_sect_idx) { 1749 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1750 sizeof(sect64.sectname)) == nullptr) 1751 break; 1752 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1753 sizeof(sect64.segname)) == nullptr) 1754 break; 1755 sect64.addr = m_data.GetAddress(&offset); 1756 sect64.size = m_data.GetAddress(&offset); 1757 1758 if (m_data.GetU32(&offset, §64.offset, num_u32s) == nullptr) 1759 break; 1760 1761 if (IsSharedCacheBinary() && !IsInMemory()) { 1762 sect64.offset = sect64.addr - m_text_address; 1763 } 1764 1765 // Keep a list of mach sections around in case we need to get at data that 1766 // isn't stored in the abstracted Sections. 1767 m_mach_sections.push_back(sect64); 1768 1769 if (add_section) { 1770 ConstString section_name( 1771 sect64.sectname, strnlen(sect64.sectname, sizeof(sect64.sectname))); 1772 if (!const_segname) { 1773 // We have a segment with no name so we need to conjure up segments 1774 // that correspond to the section's segname if there isn't already such 1775 // a section. If there is such a section, we resize the section so that 1776 // it spans all sections. We also mark these sections as fake so 1777 // address matches don't hit if they land in the gaps between the child 1778 // sections. 1779 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1780 sizeof(sect64.segname)); 1781 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1782 if (segment_sp.get()) { 1783 Section *segment = segment_sp.get(); 1784 // Grow the section size as needed. 1785 const lldb::addr_t sect64_min_addr = sect64.addr; 1786 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1787 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1788 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1789 const lldb::addr_t curr_seg_max_addr = 1790 curr_seg_min_addr + curr_seg_byte_size; 1791 if (sect64_min_addr >= curr_seg_min_addr) { 1792 const lldb::addr_t new_seg_byte_size = 1793 sect64_max_addr - curr_seg_min_addr; 1794 // Only grow the section size if needed 1795 if (new_seg_byte_size > curr_seg_byte_size) 1796 segment->SetByteSize(new_seg_byte_size); 1797 } else { 1798 // We need to change the base address of the segment and adjust the 1799 // child section offsets for all existing children. 1800 const lldb::addr_t slide_amount = 1801 sect64_min_addr - curr_seg_min_addr; 1802 segment->Slide(slide_amount, false); 1803 segment->GetChildren().Slide(-slide_amount, false); 1804 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1805 } 1806 1807 // Grow the section size as needed. 1808 if (sect64.offset) { 1809 const lldb::addr_t segment_min_file_offset = 1810 segment->GetFileOffset(); 1811 const lldb::addr_t segment_max_file_offset = 1812 segment_min_file_offset + segment->GetFileSize(); 1813 1814 const lldb::addr_t section_min_file_offset = sect64.offset; 1815 const lldb::addr_t section_max_file_offset = 1816 section_min_file_offset + sect64.size; 1817 const lldb::addr_t new_file_offset = 1818 std::min(section_min_file_offset, segment_min_file_offset); 1819 const lldb::addr_t new_file_size = 1820 std::max(section_max_file_offset, segment_max_file_offset) - 1821 new_file_offset; 1822 segment->SetFileOffset(new_file_offset); 1823 segment->SetFileSize(new_file_size); 1824 } 1825 } else { 1826 // Create a fake section for the section's named segment 1827 segment_sp = std::make_shared<Section>( 1828 segment_sp, // Parent section 1829 module_sp, // Module to which this section belongs 1830 this, // Object file to which this section belongs 1831 ++context.NextSegmentIdx 1832 << 8, // Section ID is the 1 based segment index 1833 // shifted right by 8 bits as not to 1834 // collide with any of the 256 section IDs 1835 // that are possible 1836 const_segname, // Name of this section 1837 eSectionTypeContainer, // This section is a container of 1838 // other sections. 1839 sect64.addr, // File VM address == addresses as they are 1840 // found in the object file 1841 sect64.size, // VM size in bytes of this section 1842 sect64.offset, // Offset to the data for this section in 1843 // the file 1844 sect64.offset ? sect64.size : 0, // Size in bytes of 1845 // this section as 1846 // found in the file 1847 sect64.align, 1848 load_cmd.flags); // Flags for this section 1849 segment_sp->SetIsFake(true); 1850 segment_sp->SetPermissions(segment_permissions); 1851 m_sections_up->AddSection(segment_sp); 1852 if (add_to_unified) 1853 context.UnifiedList.AddSection(segment_sp); 1854 segment_sp->SetIsEncrypted(segment_is_encrypted); 1855 } 1856 } 1857 assert(segment_sp.get()); 1858 1859 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1860 1861 SectionSP section_sp(new Section( 1862 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1863 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1864 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1865 sect64.flags)); 1866 // Set the section to be encrypted to match the segment 1867 1868 bool section_is_encrypted = false; 1869 if (!segment_is_encrypted && load_cmd.filesize != 0) 1870 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1871 sect64.offset) != nullptr; 1872 1873 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1874 section_sp->SetPermissions(segment_permissions); 1875 segment_sp->GetChildren().AddSection(section_sp); 1876 1877 if (segment_sp->IsFake()) { 1878 segment_sp.reset(); 1879 const_segname.Clear(); 1880 } 1881 } 1882 } 1883 if (segment_sp && is_dsym) { 1884 if (first_segment_sectID <= context.NextSectionIdx) { 1885 lldb::user_id_t sect_uid; 1886 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1887 ++sect_uid) { 1888 SectionSP curr_section_sp( 1889 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1890 SectionSP next_section_sp; 1891 if (sect_uid + 1 <= context.NextSectionIdx) 1892 next_section_sp = 1893 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1894 1895 if (curr_section_sp.get()) { 1896 if (curr_section_sp->GetByteSize() == 0) { 1897 if (next_section_sp.get() != nullptr) 1898 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1899 curr_section_sp->GetFileAddress()); 1900 else 1901 curr_section_sp->SetByteSize(load_cmd.vmsize); 1902 } 1903 } 1904 } 1905 } 1906 } 1907 } 1908 1909 void ObjectFileMachO::ProcessDysymtabCommand( 1910 const llvm::MachO::load_command &load_cmd, lldb::offset_t offset) { 1911 m_dysymtab.cmd = load_cmd.cmd; 1912 m_dysymtab.cmdsize = load_cmd.cmdsize; 1913 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1914 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1915 } 1916 1917 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1918 if (m_sections_up) 1919 return; 1920 1921 m_sections_up = std::make_unique<SectionList>(); 1922 1923 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1924 // bool dump_sections = false; 1925 ModuleSP module_sp(GetModule()); 1926 1927 offset = MachHeaderSizeFromMagic(m_header.magic); 1928 1929 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1930 llvm::MachO::load_command load_cmd; 1931 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1932 const lldb::offset_t load_cmd_offset = offset; 1933 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 1934 break; 1935 1936 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1937 ProcessSegmentCommand(load_cmd, offset, i, context); 1938 else if (load_cmd.cmd == LC_DYSYMTAB) 1939 ProcessDysymtabCommand(load_cmd, offset); 1940 1941 offset = load_cmd_offset + load_cmd.cmdsize; 1942 } 1943 1944 if (context.FileAddressesChanged && module_sp) 1945 module_sp->SectionFileAddressesChanged(); 1946 } 1947 1948 class MachSymtabSectionInfo { 1949 public: 1950 MachSymtabSectionInfo(SectionList *section_list) 1951 : m_section_list(section_list), m_section_infos() { 1952 // Get the number of sections down to a depth of 1 to include all segments 1953 // and their sections, but no other sections that may be added for debug 1954 // map or 1955 m_section_infos.resize(section_list->GetNumSections(1)); 1956 } 1957 1958 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1959 if (n_sect == 0) 1960 return SectionSP(); 1961 if (n_sect < m_section_infos.size()) { 1962 if (!m_section_infos[n_sect].section_sp) { 1963 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1964 m_section_infos[n_sect].section_sp = section_sp; 1965 if (section_sp) { 1966 m_section_infos[n_sect].vm_range.SetBaseAddress( 1967 section_sp->GetFileAddress()); 1968 m_section_infos[n_sect].vm_range.SetByteSize( 1969 section_sp->GetByteSize()); 1970 } else { 1971 std::string filename = "<unknown>"; 1972 SectionSP first_section_sp(m_section_list->GetSectionAtIndex(0)); 1973 if (first_section_sp) 1974 filename = first_section_sp->GetObjectFile()->GetFileSpec().GetPath(); 1975 1976 Host::SystemLog(Host::eSystemLogError, 1977 "error: unable to find section %d for a symbol in " 1978 "%s, corrupt file?\n", 1979 n_sect, filename.c_str()); 1980 } 1981 } 1982 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1983 // Symbol is in section. 1984 return m_section_infos[n_sect].section_sp; 1985 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1986 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1987 file_addr) { 1988 // Symbol is in section with zero size, but has the same start address 1989 // as the section. This can happen with linker symbols (symbols that 1990 // start with the letter 'l' or 'L'. 1991 return m_section_infos[n_sect].section_sp; 1992 } 1993 } 1994 return m_section_list->FindSectionContainingFileAddress(file_addr); 1995 } 1996 1997 protected: 1998 struct SectionInfo { 1999 SectionInfo() : vm_range(), section_sp() {} 2000 2001 VMRange vm_range; 2002 SectionSP section_sp; 2003 }; 2004 SectionList *m_section_list; 2005 std::vector<SectionInfo> m_section_infos; 2006 }; 2007 2008 #define TRIE_SYMBOL_IS_THUMB (1ULL << 63) 2009 struct TrieEntry { 2010 void Dump() const { 2011 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 2012 static_cast<unsigned long long>(address), 2013 static_cast<unsigned long long>(flags), 2014 static_cast<unsigned long long>(other), name.GetCString()); 2015 if (import_name) 2016 printf(" -> \"%s\"\n", import_name.GetCString()); 2017 else 2018 printf("\n"); 2019 } 2020 ConstString name; 2021 uint64_t address = LLDB_INVALID_ADDRESS; 2022 uint64_t flags = 2023 0; // EXPORT_SYMBOL_FLAGS_REEXPORT, EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER, 2024 // TRIE_SYMBOL_IS_THUMB 2025 uint64_t other = 0; 2026 ConstString import_name; 2027 }; 2028 2029 struct TrieEntryWithOffset { 2030 lldb::offset_t nodeOffset; 2031 TrieEntry entry; 2032 2033 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 2034 2035 void Dump(uint32_t idx) const { 2036 printf("[%3u] 0x%16.16llx: ", idx, 2037 static_cast<unsigned long long>(nodeOffset)); 2038 entry.Dump(); 2039 } 2040 2041 bool operator<(const TrieEntryWithOffset &other) const { 2042 return (nodeOffset < other.nodeOffset); 2043 } 2044 }; 2045 2046 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2047 const bool is_arm, addr_t text_seg_base_addr, 2048 std::vector<llvm::StringRef> &nameSlices, 2049 std::set<lldb::addr_t> &resolver_addresses, 2050 std::vector<TrieEntryWithOffset> &reexports, 2051 std::vector<TrieEntryWithOffset> &ext_symbols) { 2052 if (!data.ValidOffset(offset)) 2053 return true; 2054 2055 // Terminal node -- end of a branch, possibly add this to 2056 // the symbol table or resolver table. 2057 const uint64_t terminalSize = data.GetULEB128(&offset); 2058 lldb::offset_t children_offset = offset + terminalSize; 2059 if (terminalSize != 0) { 2060 TrieEntryWithOffset e(offset); 2061 e.entry.flags = data.GetULEB128(&offset); 2062 const char *import_name = nullptr; 2063 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2064 e.entry.address = 0; 2065 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2066 import_name = data.GetCStr(&offset); 2067 } else { 2068 e.entry.address = data.GetULEB128(&offset); 2069 if (text_seg_base_addr != LLDB_INVALID_ADDRESS) 2070 e.entry.address += text_seg_base_addr; 2071 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2072 e.entry.other = data.GetULEB128(&offset); 2073 uint64_t resolver_addr = e.entry.other; 2074 if (text_seg_base_addr != LLDB_INVALID_ADDRESS) 2075 resolver_addr += text_seg_base_addr; 2076 if (is_arm) 2077 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2078 resolver_addresses.insert(resolver_addr); 2079 } else 2080 e.entry.other = 0; 2081 } 2082 bool add_this_entry = false; 2083 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT) && 2084 import_name && import_name[0]) { 2085 // add symbols that are reexport symbols with a valid import name. 2086 add_this_entry = true; 2087 } else if (e.entry.flags == 0 && 2088 (import_name == nullptr || import_name[0] == '\0')) { 2089 // add externally visible symbols, in case the nlist record has 2090 // been stripped/omitted. 2091 add_this_entry = true; 2092 } 2093 if (add_this_entry) { 2094 std::string name; 2095 if (!nameSlices.empty()) { 2096 for (auto name_slice : nameSlices) 2097 name.append(name_slice.data(), name_slice.size()); 2098 } 2099 if (name.size() > 1) { 2100 // Skip the leading '_' 2101 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2102 } 2103 if (import_name) { 2104 // Skip the leading '_' 2105 e.entry.import_name.SetCString(import_name + 1); 2106 } 2107 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT)) { 2108 reexports.push_back(e); 2109 } else { 2110 if (is_arm && (e.entry.address & 1)) { 2111 e.entry.flags |= TRIE_SYMBOL_IS_THUMB; 2112 e.entry.address &= THUMB_ADDRESS_BIT_MASK; 2113 } 2114 ext_symbols.push_back(e); 2115 } 2116 } 2117 } 2118 2119 const uint8_t childrenCount = data.GetU8(&children_offset); 2120 for (uint8_t i = 0; i < childrenCount; ++i) { 2121 const char *cstr = data.GetCStr(&children_offset); 2122 if (cstr) 2123 nameSlices.push_back(llvm::StringRef(cstr)); 2124 else 2125 return false; // Corrupt data 2126 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2127 if (childNodeOffset) { 2128 if (!ParseTrieEntries(data, childNodeOffset, is_arm, text_seg_base_addr, 2129 nameSlices, resolver_addresses, reexports, 2130 ext_symbols)) { 2131 return false; 2132 } 2133 } 2134 nameSlices.pop_back(); 2135 } 2136 return true; 2137 } 2138 2139 static SymbolType GetSymbolType(const char *&symbol_name, 2140 bool &demangled_is_synthesized, 2141 const SectionSP &text_section_sp, 2142 const SectionSP &data_section_sp, 2143 const SectionSP &data_dirty_section_sp, 2144 const SectionSP &data_const_section_sp, 2145 const SectionSP &symbol_section) { 2146 SymbolType type = eSymbolTypeInvalid; 2147 2148 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2149 if (symbol_section->IsDescendant(text_section_sp.get())) { 2150 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 2151 S_ATTR_SELF_MODIFYING_CODE | 2152 S_ATTR_SOME_INSTRUCTIONS)) 2153 type = eSymbolTypeData; 2154 else 2155 type = eSymbolTypeCode; 2156 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 2157 symbol_section->IsDescendant(data_dirty_section_sp.get()) || 2158 symbol_section->IsDescendant(data_const_section_sp.get())) { 2159 if (symbol_sect_name && 2160 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 2161 type = eSymbolTypeRuntime; 2162 2163 if (symbol_name) { 2164 llvm::StringRef symbol_name_ref(symbol_name); 2165 if (symbol_name_ref.startswith("OBJC_")) { 2166 static const llvm::StringRef g_objc_v2_prefix_class("OBJC_CLASS_$_"); 2167 static const llvm::StringRef g_objc_v2_prefix_metaclass( 2168 "OBJC_METACLASS_$_"); 2169 static const llvm::StringRef g_objc_v2_prefix_ivar("OBJC_IVAR_$_"); 2170 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 2171 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2172 type = eSymbolTypeObjCClass; 2173 demangled_is_synthesized = true; 2174 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 2175 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2176 type = eSymbolTypeObjCMetaClass; 2177 demangled_is_synthesized = true; 2178 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 2179 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2180 type = eSymbolTypeObjCIVar; 2181 demangled_is_synthesized = true; 2182 } 2183 } 2184 } 2185 } else if (symbol_sect_name && 2186 ::strstr(symbol_sect_name, "__gcc_except_tab") == 2187 symbol_sect_name) { 2188 type = eSymbolTypeException; 2189 } else { 2190 type = eSymbolTypeData; 2191 } 2192 } else if (symbol_sect_name && 2193 ::strstr(symbol_sect_name, "__IMPORT") == symbol_sect_name) { 2194 type = eSymbolTypeTrampoline; 2195 } 2196 return type; 2197 } 2198 2199 // Read the UUID out of a dyld_shared_cache file on-disk. 2200 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2201 const ByteOrder byte_order, 2202 const uint32_t addr_byte_size) { 2203 UUID dsc_uuid; 2204 DataBufferSP DscData = MapFileData( 2205 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2206 if (!DscData) 2207 return dsc_uuid; 2208 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2209 2210 char version_str[7]; 2211 lldb::offset_t offset = 0; 2212 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2213 version_str[6] = '\0'; 2214 if (strcmp(version_str, "dyld_v") == 0) { 2215 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2216 dsc_uuid = UUID::fromOptionalData( 2217 dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t)); 2218 } 2219 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2220 if (log && dsc_uuid.IsValid()) { 2221 LLDB_LOGF(log, "Shared cache %s has UUID %s", 2222 dyld_shared_cache.GetPath().c_str(), 2223 dsc_uuid.GetAsString().c_str()); 2224 } 2225 return dsc_uuid; 2226 } 2227 2228 static llvm::Optional<struct nlist_64> 2229 ParseNList(DataExtractor &nlist_data, lldb::offset_t &nlist_data_offset, 2230 size_t nlist_byte_size) { 2231 struct nlist_64 nlist; 2232 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2233 return {}; 2234 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 2235 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 2236 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 2237 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 2238 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 2239 return nlist; 2240 } 2241 2242 enum { DebugSymbols = true, NonDebugSymbols = false }; 2243 2244 size_t ObjectFileMachO::ParseSymtab() { 2245 LLDB_SCOPED_TIMERF("ObjectFileMachO::ParseSymtab () module = %s", 2246 m_file.GetFilename().AsCString("")); 2247 ModuleSP module_sp(GetModule()); 2248 if (!module_sp) 2249 return 0; 2250 2251 Progress progress(llvm::formatv("Parsing symbol table for {0}", 2252 m_file.GetFilename().AsCString("<Unknown>"))); 2253 2254 llvm::MachO::symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2255 llvm::MachO::linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2256 llvm::MachO::linkedit_data_command exports_trie_load_command = {0, 0, 0, 0}; 2257 llvm::MachO::dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2258 // The data element of type bool indicates that this entry is thumb 2259 // code. 2260 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2261 2262 // Record the address of every function/data that we add to the symtab. 2263 // We add symbols to the table in the order of most information (nlist 2264 // records) to least (function starts), and avoid duplicating symbols 2265 // via this set. 2266 llvm::DenseSet<addr_t> symbols_added; 2267 2268 // We are using a llvm::DenseSet for "symbols_added" so we must be sure we 2269 // do not add the tombstone or empty keys to the set. 2270 auto add_symbol_addr = [&symbols_added](lldb::addr_t file_addr) { 2271 // Don't add the tombstone or empty keys. 2272 if (file_addr == UINT64_MAX || file_addr == UINT64_MAX - 1) 2273 return; 2274 symbols_added.insert(file_addr); 2275 }; 2276 FunctionStarts function_starts; 2277 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2278 uint32_t i; 2279 FileSpecList dylib_files; 2280 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2281 llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2282 llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2283 llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2284 2285 for (i = 0; i < m_header.ncmds; ++i) { 2286 const lldb::offset_t cmd_offset = offset; 2287 // Read in the load command and load command size 2288 llvm::MachO::load_command lc; 2289 if (m_data.GetU32(&offset, &lc, 2) == nullptr) 2290 break; 2291 // Watch for the symbol table load command 2292 switch (lc.cmd) { 2293 case LC_SYMTAB: 2294 symtab_load_command.cmd = lc.cmd; 2295 symtab_load_command.cmdsize = lc.cmdsize; 2296 // Read in the rest of the symtab load command 2297 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2298 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2299 return 0; 2300 break; 2301 2302 case LC_DYLD_INFO: 2303 case LC_DYLD_INFO_ONLY: 2304 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2305 dyld_info.cmd = lc.cmd; 2306 dyld_info.cmdsize = lc.cmdsize; 2307 } else { 2308 memset(&dyld_info, 0, sizeof(dyld_info)); 2309 } 2310 break; 2311 2312 case LC_LOAD_DYLIB: 2313 case LC_LOAD_WEAK_DYLIB: 2314 case LC_REEXPORT_DYLIB: 2315 case LC_LOADFVMLIB: 2316 case LC_LOAD_UPWARD_DYLIB: { 2317 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2318 const char *path = m_data.PeekCStr(name_offset); 2319 if (path) { 2320 FileSpec file_spec(path); 2321 // Strip the path if there is @rpath, @executable, etc so we just use 2322 // the basename 2323 if (path[0] == '@') 2324 file_spec.GetDirectory().Clear(); 2325 2326 if (lc.cmd == LC_REEXPORT_DYLIB) { 2327 m_reexported_dylibs.AppendIfUnique(file_spec); 2328 } 2329 2330 dylib_files.Append(file_spec); 2331 } 2332 } break; 2333 2334 case LC_DYLD_EXPORTS_TRIE: 2335 exports_trie_load_command.cmd = lc.cmd; 2336 exports_trie_load_command.cmdsize = lc.cmdsize; 2337 if (m_data.GetU32(&offset, &exports_trie_load_command.dataoff, 2) == 2338 nullptr) // fill in offset and size fields 2339 memset(&exports_trie_load_command, 0, 2340 sizeof(exports_trie_load_command)); 2341 break; 2342 case LC_FUNCTION_STARTS: 2343 function_starts_load_command.cmd = lc.cmd; 2344 function_starts_load_command.cmdsize = lc.cmdsize; 2345 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2346 nullptr) // fill in data offset and size fields 2347 memset(&function_starts_load_command, 0, 2348 sizeof(function_starts_load_command)); 2349 break; 2350 2351 default: 2352 break; 2353 } 2354 offset = cmd_offset + lc.cmdsize; 2355 } 2356 2357 if (!symtab_load_command.cmd) 2358 return 0; 2359 2360 Symtab *symtab = m_symtab_up.get(); 2361 SectionList *section_list = GetSectionList(); 2362 if (section_list == nullptr) 2363 return 0; 2364 2365 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2366 const ByteOrder byte_order = m_data.GetByteOrder(); 2367 bool bit_width_32 = addr_byte_size == 4; 2368 const size_t nlist_byte_size = 2369 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2370 2371 DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size); 2372 DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size); 2373 DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size); 2374 DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order, 2375 addr_byte_size); 2376 DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size); 2377 2378 const addr_t nlist_data_byte_size = 2379 symtab_load_command.nsyms * nlist_byte_size; 2380 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2381 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2382 2383 ProcessSP process_sp(m_process_wp.lock()); 2384 Process *process = process_sp.get(); 2385 2386 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2387 bool is_shared_cache_image = IsSharedCacheBinary(); 2388 bool is_local_shared_cache_image = is_shared_cache_image && !IsInMemory(); 2389 SectionSP linkedit_section_sp( 2390 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2391 2392 if (process && m_header.filetype != llvm::MachO::MH_OBJECT && 2393 !is_local_shared_cache_image) { 2394 Target &target = process->GetTarget(); 2395 2396 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2397 2398 // Reading mach file from memory in a process or core file... 2399 2400 if (linkedit_section_sp) { 2401 addr_t linkedit_load_addr = 2402 linkedit_section_sp->GetLoadBaseAddress(&target); 2403 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2404 // We might be trying to access the symbol table before the 2405 // __LINKEDIT's load address has been set in the target. We can't 2406 // fail to read the symbol table, so calculate the right address 2407 // manually 2408 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2409 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2410 } 2411 2412 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2413 const addr_t symoff_addr = linkedit_load_addr + 2414 symtab_load_command.symoff - 2415 linkedit_file_offset; 2416 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2417 linkedit_file_offset; 2418 2419 // Always load dyld - the dynamic linker - from memory if we didn't 2420 // find a binary anywhere else. lldb will not register 2421 // dylib/framework/bundle loads/unloads if we don't have the dyld 2422 // symbols, we force dyld to load from memory despite the user's 2423 // target.memory-module-load-level setting. 2424 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2425 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2426 DataBufferSP nlist_data_sp( 2427 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2428 if (nlist_data_sp) 2429 nlist_data.SetData(nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2430 if (m_dysymtab.nindirectsyms != 0) { 2431 const addr_t indirect_syms_addr = linkedit_load_addr + 2432 m_dysymtab.indirectsymoff - 2433 linkedit_file_offset; 2434 DataBufferSP indirect_syms_data_sp(ReadMemory( 2435 process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 2436 if (indirect_syms_data_sp) 2437 indirect_symbol_index_data.SetData( 2438 indirect_syms_data_sp, 0, 2439 indirect_syms_data_sp->GetByteSize()); 2440 // If this binary is outside the shared cache, 2441 // cache the string table. 2442 // Binaries in the shared cache all share a giant string table, 2443 // and we can't share the string tables across multiple 2444 // ObjectFileMachO's, so we'd end up re-reading this mega-strtab 2445 // for every binary in the shared cache - it would be a big perf 2446 // problem. For binaries outside the shared cache, it's faster to 2447 // read the entire strtab at once instead of piece-by-piece as we 2448 // process the nlist records. 2449 if (!is_shared_cache_image) { 2450 DataBufferSP strtab_data_sp( 2451 ReadMemory(process_sp, strtab_addr, strtab_data_byte_size)); 2452 if (strtab_data_sp) { 2453 strtab_data.SetData(strtab_data_sp, 0, 2454 strtab_data_sp->GetByteSize()); 2455 } 2456 } 2457 } 2458 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) { 2459 if (function_starts_load_command.cmd) { 2460 const addr_t func_start_addr = 2461 linkedit_load_addr + function_starts_load_command.dataoff - 2462 linkedit_file_offset; 2463 DataBufferSP func_start_data_sp( 2464 ReadMemory(process_sp, func_start_addr, 2465 function_starts_load_command.datasize)); 2466 if (func_start_data_sp) 2467 function_starts_data.SetData(func_start_data_sp, 0, 2468 func_start_data_sp->GetByteSize()); 2469 } 2470 } 2471 } 2472 } 2473 } else { 2474 if (is_local_shared_cache_image) { 2475 // The load commands in shared cache images are relative to the 2476 // beginning of the shared cache, not the library image. The 2477 // data we get handed when creating the ObjectFileMachO starts 2478 // at the beginning of a specific library and spans to the end 2479 // of the cache to be able to reach the shared LINKEDIT 2480 // segments. We need to convert the load command offsets to be 2481 // relative to the beginning of our specific image. 2482 lldb::addr_t linkedit_offset = linkedit_section_sp->GetFileOffset(); 2483 lldb::offset_t linkedit_slide = 2484 linkedit_offset - m_linkedit_original_offset; 2485 symtab_load_command.symoff += linkedit_slide; 2486 symtab_load_command.stroff += linkedit_slide; 2487 dyld_info.export_off += linkedit_slide; 2488 m_dysymtab.indirectsymoff += linkedit_slide; 2489 function_starts_load_command.dataoff += linkedit_slide; 2490 exports_trie_load_command.dataoff += linkedit_slide; 2491 } 2492 2493 nlist_data.SetData(m_data, symtab_load_command.symoff, 2494 nlist_data_byte_size); 2495 strtab_data.SetData(m_data, symtab_load_command.stroff, 2496 strtab_data_byte_size); 2497 2498 // We shouldn't have exports data from both the LC_DYLD_INFO command 2499 // AND the LC_DYLD_EXPORTS_TRIE command in the same binary: 2500 lldbassert(!((dyld_info.export_size > 0) 2501 && (exports_trie_load_command.datasize > 0))); 2502 if (dyld_info.export_size > 0) { 2503 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2504 dyld_info.export_size); 2505 } else if (exports_trie_load_command.datasize > 0) { 2506 dyld_trie_data.SetData(m_data, exports_trie_load_command.dataoff, 2507 exports_trie_load_command.datasize); 2508 } 2509 2510 if (m_dysymtab.nindirectsyms != 0) { 2511 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2512 m_dysymtab.nindirectsyms * 4); 2513 } 2514 if (function_starts_load_command.cmd) { 2515 function_starts_data.SetData(m_data, function_starts_load_command.dataoff, 2516 function_starts_load_command.datasize); 2517 } 2518 } 2519 2520 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2521 2522 ConstString g_segment_name_TEXT = GetSegmentNameTEXT(); 2523 ConstString g_segment_name_DATA = GetSegmentNameDATA(); 2524 ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2525 ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2526 ConstString g_segment_name_OBJC = GetSegmentNameOBJC(); 2527 ConstString g_section_name_eh_frame = GetSectionNameEHFrame(); 2528 SectionSP text_section_sp( 2529 section_list->FindSectionByName(g_segment_name_TEXT)); 2530 SectionSP data_section_sp( 2531 section_list->FindSectionByName(g_segment_name_DATA)); 2532 SectionSP data_dirty_section_sp( 2533 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2534 SectionSP data_const_section_sp( 2535 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2536 SectionSP objc_section_sp( 2537 section_list->FindSectionByName(g_segment_name_OBJC)); 2538 SectionSP eh_frame_section_sp; 2539 if (text_section_sp.get()) 2540 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2541 g_section_name_eh_frame); 2542 else 2543 eh_frame_section_sp = 2544 section_list->FindSectionByName(g_section_name_eh_frame); 2545 2546 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2547 const bool always_thumb = GetArchitecture().IsAlwaysThumbInstructions(); 2548 2549 // lldb works best if it knows the start address of all functions in a 2550 // module. Linker symbols or debug info are normally the best source of 2551 // information for start addr / size but they may be stripped in a released 2552 // binary. Two additional sources of information exist in Mach-O binaries: 2553 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2554 // function's start address in the 2555 // binary, relative to the text section. 2556 // eh_frame - the eh_frame FDEs have the start addr & size of 2557 // each function 2558 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2559 // all modern binaries. 2560 // Binaries built to run on older releases may need to use eh_frame 2561 // information. 2562 2563 if (text_section_sp && function_starts_data.GetByteSize()) { 2564 FunctionStarts::Entry function_start_entry; 2565 function_start_entry.data = false; 2566 lldb::offset_t function_start_offset = 0; 2567 function_start_entry.addr = text_section_sp->GetFileAddress(); 2568 uint64_t delta; 2569 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2570 0) { 2571 // Now append the current entry 2572 function_start_entry.addr += delta; 2573 if (is_arm) { 2574 if (function_start_entry.addr & 1) { 2575 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2576 function_start_entry.data = true; 2577 } else if (always_thumb) { 2578 function_start_entry.data = true; 2579 } 2580 } 2581 function_starts.Append(function_start_entry); 2582 } 2583 } else { 2584 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2585 // load command claiming an eh_frame but it doesn't actually have the 2586 // eh_frame content. And if we have a dSYM, we don't need to do any of 2587 // this fill-in-the-missing-symbols works anyway - the debug info should 2588 // give us all the functions in the module. 2589 if (text_section_sp.get() && eh_frame_section_sp.get() && 2590 m_type != eTypeDebugInfo) { 2591 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2592 DWARFCallFrameInfo::EH); 2593 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2594 eh_frame.GetFunctionAddressAndSizeVector(functions); 2595 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2596 size_t count = functions.GetSize(); 2597 for (size_t i = 0; i < count; ++i) { 2598 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2599 functions.GetEntryAtIndex(i); 2600 if (func) { 2601 FunctionStarts::Entry function_start_entry; 2602 function_start_entry.addr = func->base - text_base_addr; 2603 if (is_arm) { 2604 if (function_start_entry.addr & 1) { 2605 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2606 function_start_entry.data = true; 2607 } else if (always_thumb) { 2608 function_start_entry.data = true; 2609 } 2610 } 2611 function_starts.Append(function_start_entry); 2612 } 2613 } 2614 } 2615 } 2616 2617 const size_t function_starts_count = function_starts.GetSize(); 2618 2619 // For user process binaries (executables, dylibs, frameworks, bundles), if 2620 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2621 // going to assume the binary has been stripped. Don't allow assembly 2622 // language instruction emulation because we don't know proper function 2623 // start boundaries. 2624 // 2625 // For all other types of binaries (kernels, stand-alone bare board 2626 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2627 // sections - we should not make any assumptions about them based on that. 2628 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2629 m_allow_assembly_emulation_unwind_plans = false; 2630 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2631 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2632 2633 if (unwind_or_symbol_log) 2634 module_sp->LogMessage( 2635 unwind_or_symbol_log, 2636 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2637 } 2638 2639 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() 2640 ? eh_frame_section_sp->GetID() 2641 : static_cast<user_id_t>(NO_SECT); 2642 2643 lldb::offset_t nlist_data_offset = 0; 2644 2645 uint32_t N_SO_index = UINT32_MAX; 2646 2647 MachSymtabSectionInfo section_info(section_list); 2648 std::vector<uint32_t> N_FUN_indexes; 2649 std::vector<uint32_t> N_NSYM_indexes; 2650 std::vector<uint32_t> N_INCL_indexes; 2651 std::vector<uint32_t> N_BRAC_indexes; 2652 std::vector<uint32_t> N_COMM_indexes; 2653 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2654 typedef llvm::DenseMap<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2655 typedef llvm::DenseMap<const char *, uint32_t> ConstNameToSymbolIndexMap; 2656 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2657 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2658 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2659 // Any symbols that get merged into another will get an entry in this map 2660 // so we know 2661 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2662 uint32_t nlist_idx = 0; 2663 Symbol *symbol_ptr = nullptr; 2664 2665 uint32_t sym_idx = 0; 2666 Symbol *sym = nullptr; 2667 size_t num_syms = 0; 2668 std::string memory_symbol_name; 2669 uint32_t unmapped_local_symbols_found = 0; 2670 2671 std::vector<TrieEntryWithOffset> reexport_trie_entries; 2672 std::vector<TrieEntryWithOffset> external_sym_trie_entries; 2673 std::set<lldb::addr_t> resolver_addresses; 2674 2675 if (dyld_trie_data.GetByteSize() > 0) { 2676 ConstString text_segment_name("__TEXT"); 2677 SectionSP text_segment_sp = 2678 GetSectionList()->FindSectionByName(text_segment_name); 2679 lldb::addr_t text_segment_file_addr = LLDB_INVALID_ADDRESS; 2680 if (text_segment_sp) 2681 text_segment_file_addr = text_segment_sp->GetFileAddress(); 2682 std::vector<llvm::StringRef> nameSlices; 2683 ParseTrieEntries(dyld_trie_data, 0, is_arm, text_segment_file_addr, 2684 nameSlices, resolver_addresses, reexport_trie_entries, 2685 external_sym_trie_entries); 2686 } 2687 2688 typedef std::set<ConstString> IndirectSymbols; 2689 IndirectSymbols indirect_symbol_names; 2690 2691 #if TARGET_OS_IPHONE 2692 2693 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2694 // optimized by moving LOCAL symbols out of the memory mapped portion of 2695 // the DSC. The symbol information has all been retained, but it isn't 2696 // available in the normal nlist data. However, there *are* duplicate 2697 // entries of *some* 2698 // LOCAL symbols in the normal nlist data. To handle this situation 2699 // correctly, we must first attempt 2700 // to parse any DSC unmapped symbol information. If we find any, we set a 2701 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2702 2703 if (IsSharedCacheBinary()) { 2704 // Before we can start mapping the DSC, we need to make certain the 2705 // target process is actually using the cache we can find. 2706 2707 // Next we need to determine the correct path for the dyld shared cache. 2708 2709 ArchSpec header_arch = GetArchitecture(); 2710 char dsc_path[PATH_MAX]; 2711 char dsc_path_development[PATH_MAX]; 2712 2713 snprintf( 2714 dsc_path, sizeof(dsc_path), "%s%s%s", 2715 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2716 */ 2717 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2718 header_arch.GetArchitectureName()); 2719 2720 snprintf( 2721 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2722 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2723 */ 2724 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2725 header_arch.GetArchitectureName(), ".development"); 2726 2727 FileSpec dsc_nondevelopment_filespec(dsc_path); 2728 FileSpec dsc_development_filespec(dsc_path_development); 2729 FileSpec dsc_filespec; 2730 2731 UUID dsc_uuid; 2732 UUID process_shared_cache_uuid; 2733 addr_t process_shared_cache_base_addr; 2734 2735 if (process) { 2736 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, 2737 process_shared_cache_uuid); 2738 } 2739 2740 // First see if we can find an exact match for the inferior process 2741 // shared cache UUID in the development or non-development shared caches 2742 // on disk. 2743 if (process_shared_cache_uuid.IsValid()) { 2744 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2745 UUID dsc_development_uuid = GetSharedCacheUUID( 2746 dsc_development_filespec, byte_order, addr_byte_size); 2747 if (dsc_development_uuid.IsValid() && 2748 dsc_development_uuid == process_shared_cache_uuid) { 2749 dsc_filespec = dsc_development_filespec; 2750 dsc_uuid = dsc_development_uuid; 2751 } 2752 } 2753 if (!dsc_uuid.IsValid() && 2754 FileSystem::Instance().Exists(dsc_nondevelopment_filespec)) { 2755 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2756 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2757 if (dsc_nondevelopment_uuid.IsValid() && 2758 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2759 dsc_filespec = dsc_nondevelopment_filespec; 2760 dsc_uuid = dsc_nondevelopment_uuid; 2761 } 2762 } 2763 } 2764 2765 // Failing a UUID match, prefer the development dyld_shared cache if both 2766 // are present. 2767 if (!FileSystem::Instance().Exists(dsc_filespec)) { 2768 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2769 dsc_filespec = dsc_development_filespec; 2770 } else { 2771 dsc_filespec = dsc_nondevelopment_filespec; 2772 } 2773 } 2774 2775 /* The dyld_cache_header has a pointer to the 2776 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2777 The dyld_cache_local_symbols_info structure gives us three things: 2778 1. The start and count of the nlist records in the dyld_shared_cache 2779 file 2780 2. The start and size of the strings for these nlist records 2781 3. The start and count of dyld_cache_local_symbols_entry entries 2782 2783 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2784 dyld shared cache. 2785 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2786 the dyld shared cache. 2787 The dyld_cache_local_symbols_entry also lists the start of this 2788 dylib/framework's nlist records 2789 and the count of how many nlist records there are for this 2790 dylib/framework. 2791 */ 2792 2793 // Process the dyld shared cache header to find the unmapped symbols 2794 2795 DataBufferSP dsc_data_sp = MapFileData( 2796 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2797 if (!dsc_uuid.IsValid()) { 2798 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2799 } 2800 if (dsc_data_sp) { 2801 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2802 2803 bool uuid_match = true; 2804 if (dsc_uuid.IsValid() && process) { 2805 if (process_shared_cache_uuid.IsValid() && 2806 dsc_uuid != process_shared_cache_uuid) { 2807 // The on-disk dyld_shared_cache file is not the same as the one in 2808 // this process' memory, don't use it. 2809 uuid_match = false; 2810 ModuleSP module_sp(GetModule()); 2811 if (module_sp) 2812 module_sp->ReportWarning("process shared cache does not match " 2813 "on-disk dyld_shared_cache file, some " 2814 "symbol names will be missing."); 2815 } 2816 } 2817 2818 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2819 2820 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2821 2822 // If the mappingOffset points to a location inside the header, we've 2823 // opened an old dyld shared cache, and should not proceed further. 2824 if (uuid_match && 2825 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2826 2827 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2828 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2829 mappingOffset); 2830 2831 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2832 byte_order, addr_byte_size); 2833 offset = 0; 2834 2835 // The File addresses (from the in-memory Mach-O load commands) for 2836 // the shared libraries in the shared library cache need to be 2837 // adjusted by an offset to match up with the dylibOffset identifying 2838 // field in the dyld_cache_local_symbol_entry's. This offset is 2839 // recorded in mapping_offset_value. 2840 const uint64_t mapping_offset_value = 2841 dsc_mapping_info_data.GetU64(&offset); 2842 2843 offset = 2844 offsetof(struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2845 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2846 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2847 2848 if (localSymbolsOffset && localSymbolsSize) { 2849 // Map the local symbols 2850 DataBufferSP dsc_local_symbols_data_sp = 2851 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2852 2853 if (dsc_local_symbols_data_sp) { 2854 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2855 byte_order, addr_byte_size); 2856 2857 offset = 0; 2858 2859 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 2860 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 2861 UndefinedNameToDescMap undefined_name_to_desc; 2862 SymbolIndexToName reexport_shlib_needs_fixup; 2863 2864 // Read the local_symbols_infos struct in one shot 2865 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2866 dsc_local_symbols_data.GetU32(&offset, 2867 &local_symbols_info.nlistOffset, 6); 2868 2869 SectionSP text_section_sp( 2870 section_list->FindSectionByName(GetSegmentNameTEXT())); 2871 2872 uint32_t header_file_offset = 2873 (text_section_sp->GetFileAddress() - mapping_offset_value); 2874 2875 offset = local_symbols_info.entriesOffset; 2876 for (uint32_t entry_index = 0; 2877 entry_index < local_symbols_info.entriesCount; entry_index++) { 2878 struct lldb_copy_dyld_cache_local_symbols_entry 2879 local_symbols_entry; 2880 local_symbols_entry.dylibOffset = 2881 dsc_local_symbols_data.GetU32(&offset); 2882 local_symbols_entry.nlistStartIndex = 2883 dsc_local_symbols_data.GetU32(&offset); 2884 local_symbols_entry.nlistCount = 2885 dsc_local_symbols_data.GetU32(&offset); 2886 2887 if (header_file_offset == local_symbols_entry.dylibOffset) { 2888 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2889 2890 // The normal nlist code cannot correctly size the Symbols 2891 // array, we need to allocate it here. 2892 sym = symtab->Resize( 2893 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2894 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2895 num_syms = symtab->GetNumSymbols(); 2896 2897 nlist_data_offset = 2898 local_symbols_info.nlistOffset + 2899 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2900 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2901 2902 for (uint32_t nlist_index = 0; 2903 nlist_index < local_symbols_entry.nlistCount; 2904 nlist_index++) { 2905 ///////////////////////////// 2906 { 2907 llvm::Optional<struct nlist_64> nlist_maybe = 2908 ParseNList(dsc_local_symbols_data, nlist_data_offset, 2909 nlist_byte_size); 2910 if (!nlist_maybe) 2911 break; 2912 struct nlist_64 nlist = *nlist_maybe; 2913 2914 SymbolType type = eSymbolTypeInvalid; 2915 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2916 string_table_offset + nlist.n_strx); 2917 2918 if (symbol_name == NULL) { 2919 // No symbol should be NULL, even the symbols with no 2920 // string values should have an offset zero which 2921 // points to an empty C-string 2922 Host::SystemLog( 2923 Host::eSystemLogError, 2924 "error: DSC unmapped local symbol[%u] has invalid " 2925 "string table offset 0x%x in %s, ignoring symbol\n", 2926 entry_index, nlist.n_strx, 2927 module_sp->GetFileSpec().GetPath().c_str()); 2928 continue; 2929 } 2930 if (symbol_name[0] == '\0') 2931 symbol_name = NULL; 2932 2933 const char *symbol_name_non_abi_mangled = NULL; 2934 2935 SectionSP symbol_section; 2936 uint32_t symbol_byte_size = 0; 2937 bool add_nlist = true; 2938 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2939 bool demangled_is_synthesized = false; 2940 bool is_gsym = false; 2941 bool set_value = true; 2942 2943 assert(sym_idx < num_syms); 2944 2945 sym[sym_idx].SetDebug(is_debug); 2946 2947 if (is_debug) { 2948 switch (nlist.n_type) { 2949 case N_GSYM: 2950 // global symbol: name,,NO_SECT,type,0 2951 // Sometimes the N_GSYM value contains the address. 2952 2953 // FIXME: In the .o files, we have a GSYM and a debug 2954 // symbol for all the ObjC data. They 2955 // have the same address, but we want to ensure that 2956 // we always find only the real symbol, 'cause we 2957 // don't currently correctly attribute the 2958 // GSYM one to the ObjCClass/Ivar/MetaClass 2959 // symbol type. This is a temporary hack to make 2960 // sure the ObjectiveC symbols get treated correctly. 2961 // To do this right, we should coalesce all the GSYM 2962 // & global symbols that have the same address. 2963 2964 is_gsym = true; 2965 sym[sym_idx].SetExternal(true); 2966 2967 if (symbol_name && symbol_name[0] == '_' && 2968 symbol_name[1] == 'O') { 2969 llvm::StringRef symbol_name_ref(symbol_name); 2970 if (symbol_name_ref.startswith( 2971 g_objc_v2_prefix_class)) { 2972 symbol_name_non_abi_mangled = symbol_name + 1; 2973 symbol_name = 2974 symbol_name + g_objc_v2_prefix_class.size(); 2975 type = eSymbolTypeObjCClass; 2976 demangled_is_synthesized = true; 2977 2978 } else if (symbol_name_ref.startswith( 2979 g_objc_v2_prefix_metaclass)) { 2980 symbol_name_non_abi_mangled = symbol_name + 1; 2981 symbol_name = 2982 symbol_name + g_objc_v2_prefix_metaclass.size(); 2983 type = eSymbolTypeObjCMetaClass; 2984 demangled_is_synthesized = true; 2985 } else if (symbol_name_ref.startswith( 2986 g_objc_v2_prefix_ivar)) { 2987 symbol_name_non_abi_mangled = symbol_name + 1; 2988 symbol_name = 2989 symbol_name + g_objc_v2_prefix_ivar.size(); 2990 type = eSymbolTypeObjCIVar; 2991 demangled_is_synthesized = true; 2992 } 2993 } else { 2994 if (nlist.n_value != 0) 2995 symbol_section = section_info.GetSection( 2996 nlist.n_sect, nlist.n_value); 2997 type = eSymbolTypeData; 2998 } 2999 break; 3000 3001 case N_FNAME: 3002 // procedure name (f77 kludge): name,,NO_SECT,0,0 3003 type = eSymbolTypeCompiler; 3004 break; 3005 3006 case N_FUN: 3007 // procedure: name,,n_sect,linenumber,address 3008 if (symbol_name) { 3009 type = eSymbolTypeCode; 3010 symbol_section = section_info.GetSection( 3011 nlist.n_sect, nlist.n_value); 3012 3013 N_FUN_addr_to_sym_idx.insert( 3014 std::make_pair(nlist.n_value, sym_idx)); 3015 // We use the current number of symbols in the 3016 // symbol table in lieu of using nlist_idx in case 3017 // we ever start trimming entries out 3018 N_FUN_indexes.push_back(sym_idx); 3019 } else { 3020 type = eSymbolTypeCompiler; 3021 3022 if (!N_FUN_indexes.empty()) { 3023 // Copy the size of the function into the 3024 // original 3025 // STAB entry so we don't have 3026 // to hunt for it later 3027 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3028 ->SetByteSize(nlist.n_value); 3029 N_FUN_indexes.pop_back(); 3030 // We don't really need the end function STAB as 3031 // it contains the size which we already placed 3032 // with the original symbol, so don't add it if 3033 // we want a minimal symbol table 3034 add_nlist = false; 3035 } 3036 } 3037 break; 3038 3039 case N_STSYM: 3040 // static symbol: name,,n_sect,type,address 3041 N_STSYM_addr_to_sym_idx.insert( 3042 std::make_pair(nlist.n_value, sym_idx)); 3043 symbol_section = section_info.GetSection(nlist.n_sect, 3044 nlist.n_value); 3045 if (symbol_name && symbol_name[0]) { 3046 type = ObjectFile::GetSymbolTypeFromName( 3047 symbol_name + 1, eSymbolTypeData); 3048 } 3049 break; 3050 3051 case N_LCSYM: 3052 // .lcomm symbol: name,,n_sect,type,address 3053 symbol_section = section_info.GetSection(nlist.n_sect, 3054 nlist.n_value); 3055 type = eSymbolTypeCommonBlock; 3056 break; 3057 3058 case N_BNSYM: 3059 // We use the current number of symbols in the symbol 3060 // table in lieu of using nlist_idx in case we ever 3061 // start trimming entries out Skip these if we want 3062 // minimal symbol tables 3063 add_nlist = false; 3064 break; 3065 3066 case N_ENSYM: 3067 // Set the size of the N_BNSYM to the terminating 3068 // index of this N_ENSYM so that we can always skip 3069 // the entire symbol if we need to navigate more 3070 // quickly at the source level when parsing STABS 3071 // Skip these if we want minimal symbol tables 3072 add_nlist = false; 3073 break; 3074 3075 case N_OPT: 3076 // emitted with gcc2_compiled and in gcc source 3077 type = eSymbolTypeCompiler; 3078 break; 3079 3080 case N_RSYM: 3081 // register sym: name,,NO_SECT,type,register 3082 type = eSymbolTypeVariable; 3083 break; 3084 3085 case N_SLINE: 3086 // src line: 0,,n_sect,linenumber,address 3087 symbol_section = section_info.GetSection(nlist.n_sect, 3088 nlist.n_value); 3089 type = eSymbolTypeLineEntry; 3090 break; 3091 3092 case N_SSYM: 3093 // structure elt: name,,NO_SECT,type,struct_offset 3094 type = eSymbolTypeVariableType; 3095 break; 3096 3097 case N_SO: 3098 // source file name 3099 type = eSymbolTypeSourceFile; 3100 if (symbol_name == NULL) { 3101 add_nlist = false; 3102 if (N_SO_index != UINT32_MAX) { 3103 // Set the size of the N_SO to the terminating 3104 // index of this N_SO so that we can always skip 3105 // the entire N_SO if we need to navigate more 3106 // quickly at the source level when parsing STABS 3107 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3108 symbol_ptr->SetByteSize(sym_idx); 3109 symbol_ptr->SetSizeIsSibling(true); 3110 } 3111 N_NSYM_indexes.clear(); 3112 N_INCL_indexes.clear(); 3113 N_BRAC_indexes.clear(); 3114 N_COMM_indexes.clear(); 3115 N_FUN_indexes.clear(); 3116 N_SO_index = UINT32_MAX; 3117 } else { 3118 // We use the current number of symbols in the 3119 // symbol table in lieu of using nlist_idx in case 3120 // we ever start trimming entries out 3121 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3122 if (N_SO_has_full_path) { 3123 if ((N_SO_index == sym_idx - 1) && 3124 ((sym_idx - 1) < num_syms)) { 3125 // We have two consecutive N_SO entries where 3126 // the first contains a directory and the 3127 // second contains a full path. 3128 sym[sym_idx - 1].GetMangled().SetValue( 3129 ConstString(symbol_name), false); 3130 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3131 add_nlist = false; 3132 } else { 3133 // This is the first entry in a N_SO that 3134 // contains a directory or 3135 // a full path to the source file 3136 N_SO_index = sym_idx; 3137 } 3138 } else if ((N_SO_index == sym_idx - 1) && 3139 ((sym_idx - 1) < num_syms)) { 3140 // This is usually the second N_SO entry that 3141 // contains just the filename, so here we combine 3142 // it with the first one if we are minimizing the 3143 // symbol table 3144 const char *so_path = sym[sym_idx - 1] 3145 .GetMangled() 3146 .GetDemangledName() 3147 .AsCString(); 3148 if (so_path && so_path[0]) { 3149 std::string full_so_path(so_path); 3150 const size_t double_slash_pos = 3151 full_so_path.find("//"); 3152 if (double_slash_pos != std::string::npos) { 3153 // The linker has been generating bad N_SO 3154 // entries with doubled up paths 3155 // in the format "%s%s" where the first 3156 // string in the DW_AT_comp_dir, and the 3157 // second is the directory for the source 3158 // file so you end up with a path that looks 3159 // like "/tmp/src//tmp/src/" 3160 FileSpec so_dir(so_path); 3161 if (!FileSystem::Instance().Exists(so_dir)) { 3162 so_dir.SetFile( 3163 &full_so_path[double_slash_pos + 1], 3164 FileSpec::Style::native); 3165 if (FileSystem::Instance().Exists(so_dir)) { 3166 // Trim off the incorrect path 3167 full_so_path.erase(0, double_slash_pos + 1); 3168 } 3169 } 3170 } 3171 if (*full_so_path.rbegin() != '/') 3172 full_so_path += '/'; 3173 full_so_path += symbol_name; 3174 sym[sym_idx - 1].GetMangled().SetValue( 3175 ConstString(full_so_path.c_str()), false); 3176 add_nlist = false; 3177 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3178 } 3179 } else { 3180 // This could be a relative path to a N_SO 3181 N_SO_index = sym_idx; 3182 } 3183 } 3184 break; 3185 3186 case N_OSO: 3187 // object file name: name,,0,0,st_mtime 3188 type = eSymbolTypeObjectFile; 3189 break; 3190 3191 case N_LSYM: 3192 // local sym: name,,NO_SECT,type,offset 3193 type = eSymbolTypeLocal; 3194 break; 3195 3196 // INCL scopes 3197 case N_BINCL: 3198 // include file beginning: name,,NO_SECT,0,sum We use 3199 // the current number of symbols in the symbol table 3200 // in lieu of using nlist_idx in case we ever start 3201 // trimming entries out 3202 N_INCL_indexes.push_back(sym_idx); 3203 type = eSymbolTypeScopeBegin; 3204 break; 3205 3206 case N_EINCL: 3207 // include file end: name,,NO_SECT,0,0 3208 // Set the size of the N_BINCL to the terminating 3209 // index of this N_EINCL so that we can always skip 3210 // the entire symbol if we need to navigate more 3211 // quickly at the source level when parsing STABS 3212 if (!N_INCL_indexes.empty()) { 3213 symbol_ptr = 3214 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3215 symbol_ptr->SetByteSize(sym_idx + 1); 3216 symbol_ptr->SetSizeIsSibling(true); 3217 N_INCL_indexes.pop_back(); 3218 } 3219 type = eSymbolTypeScopeEnd; 3220 break; 3221 3222 case N_SOL: 3223 // #included file name: name,,n_sect,0,address 3224 type = eSymbolTypeHeaderFile; 3225 3226 // We currently don't use the header files on darwin 3227 add_nlist = false; 3228 break; 3229 3230 case N_PARAMS: 3231 // compiler parameters: name,,NO_SECT,0,0 3232 type = eSymbolTypeCompiler; 3233 break; 3234 3235 case N_VERSION: 3236 // compiler version: name,,NO_SECT,0,0 3237 type = eSymbolTypeCompiler; 3238 break; 3239 3240 case N_OLEVEL: 3241 // compiler -O level: name,,NO_SECT,0,0 3242 type = eSymbolTypeCompiler; 3243 break; 3244 3245 case N_PSYM: 3246 // parameter: name,,NO_SECT,type,offset 3247 type = eSymbolTypeVariable; 3248 break; 3249 3250 case N_ENTRY: 3251 // alternate entry: name,,n_sect,linenumber,address 3252 symbol_section = section_info.GetSection(nlist.n_sect, 3253 nlist.n_value); 3254 type = eSymbolTypeLineEntry; 3255 break; 3256 3257 // Left and Right Braces 3258 case N_LBRAC: 3259 // left bracket: 0,,NO_SECT,nesting level,address We 3260 // use the current number of symbols in the symbol 3261 // table in lieu of using nlist_idx in case we ever 3262 // start trimming entries out 3263 symbol_section = section_info.GetSection(nlist.n_sect, 3264 nlist.n_value); 3265 N_BRAC_indexes.push_back(sym_idx); 3266 type = eSymbolTypeScopeBegin; 3267 break; 3268 3269 case N_RBRAC: 3270 // right bracket: 0,,NO_SECT,nesting level,address 3271 // Set the size of the N_LBRAC to the terminating 3272 // index of this N_RBRAC so that we can always skip 3273 // the entire symbol if we need to navigate more 3274 // quickly at the source level when parsing STABS 3275 symbol_section = section_info.GetSection(nlist.n_sect, 3276 nlist.n_value); 3277 if (!N_BRAC_indexes.empty()) { 3278 symbol_ptr = 3279 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3280 symbol_ptr->SetByteSize(sym_idx + 1); 3281 symbol_ptr->SetSizeIsSibling(true); 3282 N_BRAC_indexes.pop_back(); 3283 } 3284 type = eSymbolTypeScopeEnd; 3285 break; 3286 3287 case N_EXCL: 3288 // deleted include file: name,,NO_SECT,0,sum 3289 type = eSymbolTypeHeaderFile; 3290 break; 3291 3292 // COMM scopes 3293 case N_BCOMM: 3294 // begin common: name,,NO_SECT,0,0 3295 // We use the current number of symbols in the symbol 3296 // table in lieu of using nlist_idx in case we ever 3297 // start trimming entries out 3298 type = eSymbolTypeScopeBegin; 3299 N_COMM_indexes.push_back(sym_idx); 3300 break; 3301 3302 case N_ECOML: 3303 // end common (local name): 0,,n_sect,0,address 3304 symbol_section = section_info.GetSection(nlist.n_sect, 3305 nlist.n_value); 3306 // Fall through 3307 3308 case N_ECOMM: 3309 // end common: name,,n_sect,0,0 3310 // Set the size of the N_BCOMM to the terminating 3311 // index of this N_ECOMM/N_ECOML so that we can 3312 // always skip the entire symbol if we need to 3313 // navigate more quickly at the source level when 3314 // parsing STABS 3315 if (!N_COMM_indexes.empty()) { 3316 symbol_ptr = 3317 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3318 symbol_ptr->SetByteSize(sym_idx + 1); 3319 symbol_ptr->SetSizeIsSibling(true); 3320 N_COMM_indexes.pop_back(); 3321 } 3322 type = eSymbolTypeScopeEnd; 3323 break; 3324 3325 case N_LENG: 3326 // second stab entry with length information 3327 type = eSymbolTypeAdditional; 3328 break; 3329 3330 default: 3331 break; 3332 } 3333 } else { 3334 // uint8_t n_pext = N_PEXT & nlist.n_type; 3335 uint8_t n_type = N_TYPE & nlist.n_type; 3336 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3337 3338 switch (n_type) { 3339 case N_INDR: { 3340 const char *reexport_name_cstr = 3341 strtab_data.PeekCStr(nlist.n_value); 3342 if (reexport_name_cstr && reexport_name_cstr[0]) { 3343 type = eSymbolTypeReExported; 3344 ConstString reexport_name( 3345 reexport_name_cstr + 3346 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3347 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3348 set_value = false; 3349 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3350 indirect_symbol_names.insert(ConstString( 3351 symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3352 } else 3353 type = eSymbolTypeUndefined; 3354 } break; 3355 3356 case N_UNDF: 3357 if (symbol_name && symbol_name[0]) { 3358 ConstString undefined_name( 3359 symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3360 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3361 } 3362 // Fall through 3363 case N_PBUD: 3364 type = eSymbolTypeUndefined; 3365 break; 3366 3367 case N_ABS: 3368 type = eSymbolTypeAbsolute; 3369 break; 3370 3371 case N_SECT: { 3372 symbol_section = section_info.GetSection(nlist.n_sect, 3373 nlist.n_value); 3374 3375 if (symbol_section == NULL) { 3376 // TODO: warn about this? 3377 add_nlist = false; 3378 break; 3379 } 3380 3381 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3382 type = eSymbolTypeException; 3383 } else { 3384 uint32_t section_type = 3385 symbol_section->Get() & SECTION_TYPE; 3386 3387 switch (section_type) { 3388 case S_CSTRING_LITERALS: 3389 type = eSymbolTypeData; 3390 break; // section with only literal C strings 3391 case S_4BYTE_LITERALS: 3392 type = eSymbolTypeData; 3393 break; // section with only 4 byte literals 3394 case S_8BYTE_LITERALS: 3395 type = eSymbolTypeData; 3396 break; // section with only 8 byte literals 3397 case S_LITERAL_POINTERS: 3398 type = eSymbolTypeTrampoline; 3399 break; // section with only pointers to literals 3400 case S_NON_LAZY_SYMBOL_POINTERS: 3401 type = eSymbolTypeTrampoline; 3402 break; // section with only non-lazy symbol 3403 // pointers 3404 case S_LAZY_SYMBOL_POINTERS: 3405 type = eSymbolTypeTrampoline; 3406 break; // section with only lazy symbol pointers 3407 case S_SYMBOL_STUBS: 3408 type = eSymbolTypeTrampoline; 3409 break; // section with only symbol stubs, byte 3410 // size of stub in the reserved2 field 3411 case S_MOD_INIT_FUNC_POINTERS: 3412 type = eSymbolTypeCode; 3413 break; // section with only function pointers for 3414 // initialization 3415 case S_MOD_TERM_FUNC_POINTERS: 3416 type = eSymbolTypeCode; 3417 break; // section with only function pointers for 3418 // termination 3419 case S_INTERPOSING: 3420 type = eSymbolTypeTrampoline; 3421 break; // section with only pairs of function 3422 // pointers for interposing 3423 case S_16BYTE_LITERALS: 3424 type = eSymbolTypeData; 3425 break; // section with only 16 byte literals 3426 case S_DTRACE_DOF: 3427 type = eSymbolTypeInstrumentation; 3428 break; 3429 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3430 type = eSymbolTypeTrampoline; 3431 break; 3432 default: 3433 switch (symbol_section->GetType()) { 3434 case lldb::eSectionTypeCode: 3435 type = eSymbolTypeCode; 3436 break; 3437 case eSectionTypeData: 3438 case eSectionTypeDataCString: // Inlined C string 3439 // data 3440 case eSectionTypeDataCStringPointers: // Pointers 3441 // to C 3442 // string 3443 // data 3444 case eSectionTypeDataSymbolAddress: // Address of 3445 // a symbol in 3446 // the symbol 3447 // table 3448 case eSectionTypeData4: 3449 case eSectionTypeData8: 3450 case eSectionTypeData16: 3451 type = eSymbolTypeData; 3452 break; 3453 default: 3454 break; 3455 } 3456 break; 3457 } 3458 3459 if (type == eSymbolTypeInvalid) { 3460 const char *symbol_sect_name = 3461 symbol_section->GetName().AsCString(); 3462 if (symbol_section->IsDescendant( 3463 text_section_sp.get())) { 3464 if (symbol_section->IsClear( 3465 S_ATTR_PURE_INSTRUCTIONS | 3466 S_ATTR_SELF_MODIFYING_CODE | 3467 S_ATTR_SOME_INSTRUCTIONS)) 3468 type = eSymbolTypeData; 3469 else 3470 type = eSymbolTypeCode; 3471 } else if (symbol_section->IsDescendant( 3472 data_section_sp.get()) || 3473 symbol_section->IsDescendant( 3474 data_dirty_section_sp.get()) || 3475 symbol_section->IsDescendant( 3476 data_const_section_sp.get())) { 3477 if (symbol_sect_name && 3478 ::strstr(symbol_sect_name, "__objc") == 3479 symbol_sect_name) { 3480 type = eSymbolTypeRuntime; 3481 3482 if (symbol_name) { 3483 llvm::StringRef symbol_name_ref(symbol_name); 3484 if (symbol_name_ref.startswith("_OBJC_")) { 3485 llvm::StringRef 3486 g_objc_v2_prefix_class( 3487 "_OBJC_CLASS_$_"); 3488 llvm::StringRef 3489 g_objc_v2_prefix_metaclass( 3490 "_OBJC_METACLASS_$_"); 3491 llvm::StringRef 3492 g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 3493 if (symbol_name_ref.startswith( 3494 g_objc_v2_prefix_class)) { 3495 symbol_name_non_abi_mangled = 3496 symbol_name + 1; 3497 symbol_name = 3498 symbol_name + 3499 g_objc_v2_prefix_class.size(); 3500 type = eSymbolTypeObjCClass; 3501 demangled_is_synthesized = true; 3502 } else if ( 3503 symbol_name_ref.startswith( 3504 g_objc_v2_prefix_metaclass)) { 3505 symbol_name_non_abi_mangled = 3506 symbol_name + 1; 3507 symbol_name = 3508 symbol_name + 3509 g_objc_v2_prefix_metaclass.size(); 3510 type = eSymbolTypeObjCMetaClass; 3511 demangled_is_synthesized = true; 3512 } else if (symbol_name_ref.startswith( 3513 g_objc_v2_prefix_ivar)) { 3514 symbol_name_non_abi_mangled = 3515 symbol_name + 1; 3516 symbol_name = 3517 symbol_name + 3518 g_objc_v2_prefix_ivar.size(); 3519 type = eSymbolTypeObjCIVar; 3520 demangled_is_synthesized = true; 3521 } 3522 } 3523 } 3524 } else if (symbol_sect_name && 3525 ::strstr(symbol_sect_name, 3526 "__gcc_except_tab") == 3527 symbol_sect_name) { 3528 type = eSymbolTypeException; 3529 } else { 3530 type = eSymbolTypeData; 3531 } 3532 } else if (symbol_sect_name && 3533 ::strstr(symbol_sect_name, "__IMPORT") == 3534 symbol_sect_name) { 3535 type = eSymbolTypeTrampoline; 3536 } else if (symbol_section->IsDescendant( 3537 objc_section_sp.get())) { 3538 type = eSymbolTypeRuntime; 3539 if (symbol_name && symbol_name[0] == '.') { 3540 llvm::StringRef symbol_name_ref(symbol_name); 3541 llvm::StringRef 3542 g_objc_v1_prefix_class(".objc_class_name_"); 3543 if (symbol_name_ref.startswith( 3544 g_objc_v1_prefix_class)) { 3545 symbol_name_non_abi_mangled = symbol_name; 3546 symbol_name = symbol_name + 3547 g_objc_v1_prefix_class.size(); 3548 type = eSymbolTypeObjCClass; 3549 demangled_is_synthesized = true; 3550 } 3551 } 3552 } 3553 } 3554 } 3555 } break; 3556 } 3557 } 3558 3559 if (add_nlist) { 3560 uint64_t symbol_value = nlist.n_value; 3561 if (symbol_name_non_abi_mangled) { 3562 sym[sym_idx].GetMangled().SetMangledName( 3563 ConstString(symbol_name_non_abi_mangled)); 3564 sym[sym_idx].GetMangled().SetDemangledName( 3565 ConstString(symbol_name)); 3566 } else { 3567 bool symbol_name_is_mangled = false; 3568 3569 if (symbol_name && symbol_name[0] == '_') { 3570 symbol_name_is_mangled = symbol_name[1] == '_'; 3571 symbol_name++; // Skip the leading underscore 3572 } 3573 3574 if (symbol_name) { 3575 ConstString const_symbol_name(symbol_name); 3576 sym[sym_idx].GetMangled().SetValue( 3577 const_symbol_name, symbol_name_is_mangled); 3578 if (is_gsym && is_debug) { 3579 const char *gsym_name = 3580 sym[sym_idx] 3581 .GetMangled() 3582 .GetName(Mangled::ePreferMangled) 3583 .GetCString(); 3584 if (gsym_name) 3585 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3586 } 3587 } 3588 } 3589 if (symbol_section) { 3590 const addr_t section_file_addr = 3591 symbol_section->GetFileAddress(); 3592 if (symbol_byte_size == 0 && 3593 function_starts_count > 0) { 3594 addr_t symbol_lookup_file_addr = nlist.n_value; 3595 // Do an exact address match for non-ARM addresses, 3596 // else get the closest since the symbol might be a 3597 // thumb symbol which has an address with bit zero 3598 // set 3599 FunctionStarts::Entry *func_start_entry = 3600 function_starts.FindEntry(symbol_lookup_file_addr, 3601 !is_arm); 3602 if (is_arm && func_start_entry) { 3603 // Verify that the function start address is the 3604 // symbol address (ARM) or the symbol address + 1 3605 // (thumb) 3606 if (func_start_entry->addr != 3607 symbol_lookup_file_addr && 3608 func_start_entry->addr != 3609 (symbol_lookup_file_addr + 1)) { 3610 // Not the right entry, NULL it out... 3611 func_start_entry = NULL; 3612 } 3613 } 3614 if (func_start_entry) { 3615 func_start_entry->data = true; 3616 3617 addr_t symbol_file_addr = func_start_entry->addr; 3618 uint32_t symbol_flags = 0; 3619 if (is_arm) { 3620 if (symbol_file_addr & 1) 3621 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3622 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3623 } 3624 3625 const FunctionStarts::Entry *next_func_start_entry = 3626 function_starts.FindNextEntry(func_start_entry); 3627 const addr_t section_end_file_addr = 3628 section_file_addr + 3629 symbol_section->GetByteSize(); 3630 if (next_func_start_entry) { 3631 addr_t next_symbol_file_addr = 3632 next_func_start_entry->addr; 3633 // Be sure the clear the Thumb address bit when 3634 // we calculate the size from the current and 3635 // next address 3636 if (is_arm) 3637 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3638 symbol_byte_size = std::min<lldb::addr_t>( 3639 next_symbol_file_addr - symbol_file_addr, 3640 section_end_file_addr - symbol_file_addr); 3641 } else { 3642 symbol_byte_size = 3643 section_end_file_addr - symbol_file_addr; 3644 } 3645 } 3646 } 3647 symbol_value -= section_file_addr; 3648 } 3649 3650 if (is_debug == false) { 3651 if (type == eSymbolTypeCode) { 3652 // See if we can find a N_FUN entry for any code 3653 // symbols. If we do find a match, and the name 3654 // matches, then we can merge the two into just the 3655 // function symbol to avoid duplicate entries in 3656 // the symbol table 3657 auto range = 3658 N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3659 if (range.first != range.second) { 3660 bool found_it = false; 3661 for (auto pos = range.first; pos != range.second; 3662 ++pos) { 3663 if (sym[sym_idx].GetMangled().GetName( 3664 Mangled::ePreferMangled) == 3665 sym[pos->second].GetMangled().GetName( 3666 Mangled::ePreferMangled)) { 3667 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3668 // We just need the flags from the linker 3669 // symbol, so put these flags 3670 // into the N_FUN flags to avoid duplicate 3671 // symbols in the symbol table 3672 sym[pos->second].SetExternal( 3673 sym[sym_idx].IsExternal()); 3674 sym[pos->second].SetFlags(nlist.n_type << 16 | 3675 nlist.n_desc); 3676 if (resolver_addresses.find(nlist.n_value) != 3677 resolver_addresses.end()) 3678 sym[pos->second].SetType(eSymbolTypeResolver); 3679 sym[sym_idx].Clear(); 3680 found_it = true; 3681 break; 3682 } 3683 } 3684 if (found_it) 3685 continue; 3686 } else { 3687 if (resolver_addresses.find(nlist.n_value) != 3688 resolver_addresses.end()) 3689 type = eSymbolTypeResolver; 3690 } 3691 } else if (type == eSymbolTypeData || 3692 type == eSymbolTypeObjCClass || 3693 type == eSymbolTypeObjCMetaClass || 3694 type == eSymbolTypeObjCIVar) { 3695 // See if we can find a N_STSYM entry for any data 3696 // symbols. If we do find a match, and the name 3697 // matches, then we can merge the two into just the 3698 // Static symbol to avoid duplicate entries in the 3699 // symbol table 3700 auto range = N_STSYM_addr_to_sym_idx.equal_range( 3701 nlist.n_value); 3702 if (range.first != range.second) { 3703 bool found_it = false; 3704 for (auto pos = range.first; pos != range.second; 3705 ++pos) { 3706 if (sym[sym_idx].GetMangled().GetName( 3707 Mangled::ePreferMangled) == 3708 sym[pos->second].GetMangled().GetName( 3709 Mangled::ePreferMangled)) { 3710 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3711 // We just need the flags from the linker 3712 // symbol, so put these flags 3713 // into the N_STSYM flags to avoid duplicate 3714 // symbols in the symbol table 3715 sym[pos->second].SetExternal( 3716 sym[sym_idx].IsExternal()); 3717 sym[pos->second].SetFlags(nlist.n_type << 16 | 3718 nlist.n_desc); 3719 sym[sym_idx].Clear(); 3720 found_it = true; 3721 break; 3722 } 3723 } 3724 if (found_it) 3725 continue; 3726 } else { 3727 const char *gsym_name = 3728 sym[sym_idx] 3729 .GetMangled() 3730 .GetName(Mangled::ePreferMangled) 3731 .GetCString(); 3732 if (gsym_name) { 3733 // Combine N_GSYM stab entries with the non 3734 // stab symbol 3735 ConstNameToSymbolIndexMap::const_iterator pos = 3736 N_GSYM_name_to_sym_idx.find(gsym_name); 3737 if (pos != N_GSYM_name_to_sym_idx.end()) { 3738 const uint32_t GSYM_sym_idx = pos->second; 3739 m_nlist_idx_to_sym_idx[nlist_idx] = 3740 GSYM_sym_idx; 3741 // Copy the address, because often the N_GSYM 3742 // address has an invalid address of zero 3743 // when the global is a common symbol 3744 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3745 symbol_section); 3746 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3747 symbol_value); 3748 add_symbol_addr(sym[GSYM_sym_idx] 3749 .GetAddress() 3750 .GetFileAddress()); 3751 // We just need the flags from the linker 3752 // symbol, so put these flags 3753 // into the N_GSYM flags to avoid duplicate 3754 // symbols in the symbol table 3755 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 3756 nlist.n_desc); 3757 sym[sym_idx].Clear(); 3758 continue; 3759 } 3760 } 3761 } 3762 } 3763 } 3764 3765 sym[sym_idx].SetID(nlist_idx); 3766 sym[sym_idx].SetType(type); 3767 if (set_value) { 3768 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 3769 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3770 add_symbol_addr( 3771 sym[sym_idx].GetAddress().GetFileAddress()); 3772 } 3773 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 3774 3775 if (symbol_byte_size > 0) 3776 sym[sym_idx].SetByteSize(symbol_byte_size); 3777 3778 if (demangled_is_synthesized) 3779 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3780 ++sym_idx; 3781 } else { 3782 sym[sym_idx].Clear(); 3783 } 3784 } 3785 ///////////////////////////// 3786 } 3787 break; // No more entries to consider 3788 } 3789 } 3790 3791 for (const auto &pos : reexport_shlib_needs_fixup) { 3792 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3793 if (undef_pos != undefined_name_to_desc.end()) { 3794 const uint8_t dylib_ordinal = 3795 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3796 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3797 sym[pos.first].SetReExportedSymbolSharedLibrary( 3798 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3799 } 3800 } 3801 } 3802 } 3803 } 3804 } 3805 } 3806 3807 // Must reset this in case it was mutated above! 3808 nlist_data_offset = 0; 3809 #endif 3810 3811 if (nlist_data.GetByteSize() > 0) { 3812 3813 // If the sym array was not created while parsing the DSC unmapped 3814 // symbols, create it now. 3815 if (sym == nullptr) { 3816 sym = 3817 symtab->Resize(symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3818 num_syms = symtab->GetNumSymbols(); 3819 } 3820 3821 if (unmapped_local_symbols_found) { 3822 assert(m_dysymtab.ilocalsym == 0); 3823 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3824 nlist_idx = m_dysymtab.nlocalsym; 3825 } else { 3826 nlist_idx = 0; 3827 } 3828 3829 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 3830 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 3831 UndefinedNameToDescMap undefined_name_to_desc; 3832 SymbolIndexToName reexport_shlib_needs_fixup; 3833 3834 // Symtab parsing is a huge mess. Everything is entangled and the code 3835 // requires access to a ridiculous amount of variables. LLDB depends 3836 // heavily on the proper merging of symbols and to get that right we need 3837 // to make sure we have parsed all the debug symbols first. Therefore we 3838 // invoke the lambda twice, once to parse only the debug symbols and then 3839 // once more to parse the remaining symbols. 3840 auto ParseSymbolLambda = [&](struct nlist_64 &nlist, uint32_t nlist_idx, 3841 bool debug_only) { 3842 const bool is_debug = ((nlist.n_type & N_STAB) != 0); 3843 if (is_debug != debug_only) 3844 return true; 3845 3846 const char *symbol_name_non_abi_mangled = nullptr; 3847 const char *symbol_name = nullptr; 3848 3849 if (have_strtab_data) { 3850 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3851 3852 if (symbol_name == nullptr) { 3853 // No symbol should be NULL, even the symbols with no string values 3854 // should have an offset zero which points to an empty C-string 3855 Host::SystemLog(Host::eSystemLogError, 3856 "error: symbol[%u] has invalid string table offset " 3857 "0x%x in %s, ignoring symbol\n", 3858 nlist_idx, nlist.n_strx, 3859 module_sp->GetFileSpec().GetPath().c_str()); 3860 return true; 3861 } 3862 if (symbol_name[0] == '\0') 3863 symbol_name = nullptr; 3864 } else { 3865 const addr_t str_addr = strtab_addr + nlist.n_strx; 3866 Status str_error; 3867 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3868 str_error)) 3869 symbol_name = memory_symbol_name.c_str(); 3870 } 3871 3872 SymbolType type = eSymbolTypeInvalid; 3873 SectionSP symbol_section; 3874 lldb::addr_t symbol_byte_size = 0; 3875 bool add_nlist = true; 3876 bool is_gsym = false; 3877 bool demangled_is_synthesized = false; 3878 bool set_value = true; 3879 3880 assert(sym_idx < num_syms); 3881 sym[sym_idx].SetDebug(is_debug); 3882 3883 if (is_debug) { 3884 switch (nlist.n_type) { 3885 case N_GSYM: 3886 // global symbol: name,,NO_SECT,type,0 3887 // Sometimes the N_GSYM value contains the address. 3888 3889 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3890 // the ObjC data. They 3891 // have the same address, but we want to ensure that we always find 3892 // only the real symbol, 'cause we don't currently correctly 3893 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3894 // type. This is a temporary hack to make sure the ObjectiveC 3895 // symbols get treated correctly. To do this right, we should 3896 // coalesce all the GSYM & global symbols that have the same 3897 // address. 3898 is_gsym = true; 3899 sym[sym_idx].SetExternal(true); 3900 3901 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3902 llvm::StringRef symbol_name_ref(symbol_name); 3903 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3904 symbol_name_non_abi_mangled = symbol_name + 1; 3905 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3906 type = eSymbolTypeObjCClass; 3907 demangled_is_synthesized = true; 3908 3909 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 3910 symbol_name_non_abi_mangled = symbol_name + 1; 3911 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3912 type = eSymbolTypeObjCMetaClass; 3913 demangled_is_synthesized = true; 3914 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3915 symbol_name_non_abi_mangled = symbol_name + 1; 3916 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3917 type = eSymbolTypeObjCIVar; 3918 demangled_is_synthesized = true; 3919 } 3920 } else { 3921 if (nlist.n_value != 0) 3922 symbol_section = 3923 section_info.GetSection(nlist.n_sect, nlist.n_value); 3924 type = eSymbolTypeData; 3925 } 3926 break; 3927 3928 case N_FNAME: 3929 // procedure name (f77 kludge): name,,NO_SECT,0,0 3930 type = eSymbolTypeCompiler; 3931 break; 3932 3933 case N_FUN: 3934 // procedure: name,,n_sect,linenumber,address 3935 if (symbol_name) { 3936 type = eSymbolTypeCode; 3937 symbol_section = 3938 section_info.GetSection(nlist.n_sect, nlist.n_value); 3939 3940 N_FUN_addr_to_sym_idx.insert( 3941 std::make_pair(nlist.n_value, sym_idx)); 3942 // We use the current number of symbols in the symbol table in 3943 // lieu of using nlist_idx in case we ever start trimming entries 3944 // out 3945 N_FUN_indexes.push_back(sym_idx); 3946 } else { 3947 type = eSymbolTypeCompiler; 3948 3949 if (!N_FUN_indexes.empty()) { 3950 // Copy the size of the function into the original STAB entry 3951 // so we don't have to hunt for it later 3952 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3953 ->SetByteSize(nlist.n_value); 3954 N_FUN_indexes.pop_back(); 3955 // We don't really need the end function STAB as it contains 3956 // the size which we already placed with the original symbol, 3957 // so don't add it if we want a minimal symbol table 3958 add_nlist = false; 3959 } 3960 } 3961 break; 3962 3963 case N_STSYM: 3964 // static symbol: name,,n_sect,type,address 3965 N_STSYM_addr_to_sym_idx.insert( 3966 std::make_pair(nlist.n_value, sym_idx)); 3967 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3968 if (symbol_name && symbol_name[0]) { 3969 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3970 eSymbolTypeData); 3971 } 3972 break; 3973 3974 case N_LCSYM: 3975 // .lcomm symbol: name,,n_sect,type,address 3976 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3977 type = eSymbolTypeCommonBlock; 3978 break; 3979 3980 case N_BNSYM: 3981 // We use the current number of symbols in the symbol table in lieu 3982 // of using nlist_idx in case we ever start trimming entries out 3983 // Skip these if we want minimal symbol tables 3984 add_nlist = false; 3985 break; 3986 3987 case N_ENSYM: 3988 // Set the size of the N_BNSYM to the terminating index of this 3989 // N_ENSYM so that we can always skip the entire symbol if we need 3990 // to navigate more quickly at the source level when parsing STABS 3991 // Skip these if we want minimal symbol tables 3992 add_nlist = false; 3993 break; 3994 3995 case N_OPT: 3996 // emitted with gcc2_compiled and in gcc source 3997 type = eSymbolTypeCompiler; 3998 break; 3999 4000 case N_RSYM: 4001 // register sym: name,,NO_SECT,type,register 4002 type = eSymbolTypeVariable; 4003 break; 4004 4005 case N_SLINE: 4006 // src line: 0,,n_sect,linenumber,address 4007 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4008 type = eSymbolTypeLineEntry; 4009 break; 4010 4011 case N_SSYM: 4012 // structure elt: name,,NO_SECT,type,struct_offset 4013 type = eSymbolTypeVariableType; 4014 break; 4015 4016 case N_SO: 4017 // source file name 4018 type = eSymbolTypeSourceFile; 4019 if (symbol_name == nullptr) { 4020 add_nlist = false; 4021 if (N_SO_index != UINT32_MAX) { 4022 // Set the size of the N_SO to the terminating index of this 4023 // N_SO so that we can always skip the entire N_SO if we need 4024 // to navigate more quickly at the source level when parsing 4025 // STABS 4026 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 4027 symbol_ptr->SetByteSize(sym_idx); 4028 symbol_ptr->SetSizeIsSibling(true); 4029 } 4030 N_NSYM_indexes.clear(); 4031 N_INCL_indexes.clear(); 4032 N_BRAC_indexes.clear(); 4033 N_COMM_indexes.clear(); 4034 N_FUN_indexes.clear(); 4035 N_SO_index = UINT32_MAX; 4036 } else { 4037 // We use the current number of symbols in the symbol table in 4038 // lieu of using nlist_idx in case we ever start trimming entries 4039 // out 4040 const bool N_SO_has_full_path = symbol_name[0] == '/'; 4041 if (N_SO_has_full_path) { 4042 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 4043 // We have two consecutive N_SO entries where the first 4044 // contains a directory and the second contains a full path. 4045 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), 4046 false); 4047 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4048 add_nlist = false; 4049 } else { 4050 // This is the first entry in a N_SO that contains a 4051 // directory or a full path to the source file 4052 N_SO_index = sym_idx; 4053 } 4054 } else if ((N_SO_index == sym_idx - 1) && 4055 ((sym_idx - 1) < num_syms)) { 4056 // This is usually the second N_SO entry that contains just the 4057 // filename, so here we combine it with the first one if we are 4058 // minimizing the symbol table 4059 const char *so_path = 4060 sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 4061 if (so_path && so_path[0]) { 4062 std::string full_so_path(so_path); 4063 const size_t double_slash_pos = full_so_path.find("//"); 4064 if (double_slash_pos != std::string::npos) { 4065 // The linker has been generating bad N_SO entries with 4066 // doubled up paths in the format "%s%s" where the first 4067 // string in the DW_AT_comp_dir, and the second is the 4068 // directory for the source file so you end up with a path 4069 // that looks like "/tmp/src//tmp/src/" 4070 FileSpec so_dir(so_path); 4071 if (!FileSystem::Instance().Exists(so_dir)) { 4072 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 4073 FileSpec::Style::native); 4074 if (FileSystem::Instance().Exists(so_dir)) { 4075 // Trim off the incorrect path 4076 full_so_path.erase(0, double_slash_pos + 1); 4077 } 4078 } 4079 } 4080 if (*full_so_path.rbegin() != '/') 4081 full_so_path += '/'; 4082 full_so_path += symbol_name; 4083 sym[sym_idx - 1].GetMangled().SetValue( 4084 ConstString(full_so_path.c_str()), false); 4085 add_nlist = false; 4086 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4087 } 4088 } else { 4089 // This could be a relative path to a N_SO 4090 N_SO_index = sym_idx; 4091 } 4092 } 4093 break; 4094 4095 case N_OSO: 4096 // object file name: name,,0,0,st_mtime 4097 type = eSymbolTypeObjectFile; 4098 break; 4099 4100 case N_LSYM: 4101 // local sym: name,,NO_SECT,type,offset 4102 type = eSymbolTypeLocal; 4103 break; 4104 4105 // INCL scopes 4106 case N_BINCL: 4107 // include file beginning: name,,NO_SECT,0,sum We use the current 4108 // number of symbols in the symbol table in lieu of using nlist_idx 4109 // in case we ever start trimming entries out 4110 N_INCL_indexes.push_back(sym_idx); 4111 type = eSymbolTypeScopeBegin; 4112 break; 4113 4114 case N_EINCL: 4115 // include file end: name,,NO_SECT,0,0 4116 // Set the size of the N_BINCL to the terminating index of this 4117 // N_EINCL so that we can always skip the entire symbol if we need 4118 // to navigate more quickly at the source level when parsing STABS 4119 if (!N_INCL_indexes.empty()) { 4120 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4121 symbol_ptr->SetByteSize(sym_idx + 1); 4122 symbol_ptr->SetSizeIsSibling(true); 4123 N_INCL_indexes.pop_back(); 4124 } 4125 type = eSymbolTypeScopeEnd; 4126 break; 4127 4128 case N_SOL: 4129 // #included file name: name,,n_sect,0,address 4130 type = eSymbolTypeHeaderFile; 4131 4132 // We currently don't use the header files on darwin 4133 add_nlist = false; 4134 break; 4135 4136 case N_PARAMS: 4137 // compiler parameters: name,,NO_SECT,0,0 4138 type = eSymbolTypeCompiler; 4139 break; 4140 4141 case N_VERSION: 4142 // compiler version: name,,NO_SECT,0,0 4143 type = eSymbolTypeCompiler; 4144 break; 4145 4146 case N_OLEVEL: 4147 // compiler -O level: name,,NO_SECT,0,0 4148 type = eSymbolTypeCompiler; 4149 break; 4150 4151 case N_PSYM: 4152 // parameter: name,,NO_SECT,type,offset 4153 type = eSymbolTypeVariable; 4154 break; 4155 4156 case N_ENTRY: 4157 // alternate entry: name,,n_sect,linenumber,address 4158 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4159 type = eSymbolTypeLineEntry; 4160 break; 4161 4162 // Left and Right Braces 4163 case N_LBRAC: 4164 // left bracket: 0,,NO_SECT,nesting level,address We use the 4165 // current number of symbols in the symbol table in lieu of using 4166 // nlist_idx in case we ever start trimming entries out 4167 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4168 N_BRAC_indexes.push_back(sym_idx); 4169 type = eSymbolTypeScopeBegin; 4170 break; 4171 4172 case N_RBRAC: 4173 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4174 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4175 // can always skip the entire symbol if we need to navigate more 4176 // quickly at the source level when parsing STABS 4177 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4178 if (!N_BRAC_indexes.empty()) { 4179 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4180 symbol_ptr->SetByteSize(sym_idx + 1); 4181 symbol_ptr->SetSizeIsSibling(true); 4182 N_BRAC_indexes.pop_back(); 4183 } 4184 type = eSymbolTypeScopeEnd; 4185 break; 4186 4187 case N_EXCL: 4188 // deleted include file: name,,NO_SECT,0,sum 4189 type = eSymbolTypeHeaderFile; 4190 break; 4191 4192 // COMM scopes 4193 case N_BCOMM: 4194 // begin common: name,,NO_SECT,0,0 4195 // We use the current number of symbols in the symbol table in lieu 4196 // of using nlist_idx in case we ever start trimming entries out 4197 type = eSymbolTypeScopeBegin; 4198 N_COMM_indexes.push_back(sym_idx); 4199 break; 4200 4201 case N_ECOML: 4202 // end common (local name): 0,,n_sect,0,address 4203 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4204 LLVM_FALLTHROUGH; 4205 4206 case N_ECOMM: 4207 // end common: name,,n_sect,0,0 4208 // Set the size of the N_BCOMM to the terminating index of this 4209 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4210 // we need to navigate more quickly at the source level when 4211 // parsing STABS 4212 if (!N_COMM_indexes.empty()) { 4213 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4214 symbol_ptr->SetByteSize(sym_idx + 1); 4215 symbol_ptr->SetSizeIsSibling(true); 4216 N_COMM_indexes.pop_back(); 4217 } 4218 type = eSymbolTypeScopeEnd; 4219 break; 4220 4221 case N_LENG: 4222 // second stab entry with length information 4223 type = eSymbolTypeAdditional; 4224 break; 4225 4226 default: 4227 break; 4228 } 4229 } else { 4230 uint8_t n_type = N_TYPE & nlist.n_type; 4231 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4232 4233 switch (n_type) { 4234 case N_INDR: { 4235 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4236 if (reexport_name_cstr && reexport_name_cstr[0]) { 4237 type = eSymbolTypeReExported; 4238 ConstString reexport_name(reexport_name_cstr + 4239 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4240 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4241 set_value = false; 4242 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4243 indirect_symbol_names.insert( 4244 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4245 } else 4246 type = eSymbolTypeUndefined; 4247 } break; 4248 4249 case N_UNDF: 4250 if (symbol_name && symbol_name[0]) { 4251 ConstString undefined_name(symbol_name + 4252 ((symbol_name[0] == '_') ? 1 : 0)); 4253 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4254 } 4255 LLVM_FALLTHROUGH; 4256 4257 case N_PBUD: 4258 type = eSymbolTypeUndefined; 4259 break; 4260 4261 case N_ABS: 4262 type = eSymbolTypeAbsolute; 4263 break; 4264 4265 case N_SECT: { 4266 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4267 4268 if (!symbol_section) { 4269 // TODO: warn about this? 4270 add_nlist = false; 4271 break; 4272 } 4273 4274 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4275 type = eSymbolTypeException; 4276 } else { 4277 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4278 4279 switch (section_type) { 4280 case S_CSTRING_LITERALS: 4281 type = eSymbolTypeData; 4282 break; // section with only literal C strings 4283 case S_4BYTE_LITERALS: 4284 type = eSymbolTypeData; 4285 break; // section with only 4 byte literals 4286 case S_8BYTE_LITERALS: 4287 type = eSymbolTypeData; 4288 break; // section with only 8 byte literals 4289 case S_LITERAL_POINTERS: 4290 type = eSymbolTypeTrampoline; 4291 break; // section with only pointers to literals 4292 case S_NON_LAZY_SYMBOL_POINTERS: 4293 type = eSymbolTypeTrampoline; 4294 break; // section with only non-lazy symbol pointers 4295 case S_LAZY_SYMBOL_POINTERS: 4296 type = eSymbolTypeTrampoline; 4297 break; // section with only lazy symbol pointers 4298 case S_SYMBOL_STUBS: 4299 type = eSymbolTypeTrampoline; 4300 break; // section with only symbol stubs, byte size of stub in 4301 // the reserved2 field 4302 case S_MOD_INIT_FUNC_POINTERS: 4303 type = eSymbolTypeCode; 4304 break; // section with only function pointers for initialization 4305 case S_MOD_TERM_FUNC_POINTERS: 4306 type = eSymbolTypeCode; 4307 break; // section with only function pointers for termination 4308 case S_INTERPOSING: 4309 type = eSymbolTypeTrampoline; 4310 break; // section with only pairs of function pointers for 4311 // interposing 4312 case S_16BYTE_LITERALS: 4313 type = eSymbolTypeData; 4314 break; // section with only 16 byte literals 4315 case S_DTRACE_DOF: 4316 type = eSymbolTypeInstrumentation; 4317 break; 4318 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4319 type = eSymbolTypeTrampoline; 4320 break; 4321 default: 4322 switch (symbol_section->GetType()) { 4323 case lldb::eSectionTypeCode: 4324 type = eSymbolTypeCode; 4325 break; 4326 case eSectionTypeData: 4327 case eSectionTypeDataCString: // Inlined C string data 4328 case eSectionTypeDataCStringPointers: // Pointers to C string 4329 // data 4330 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4331 // the symbol table 4332 case eSectionTypeData4: 4333 case eSectionTypeData8: 4334 case eSectionTypeData16: 4335 type = eSymbolTypeData; 4336 break; 4337 default: 4338 break; 4339 } 4340 break; 4341 } 4342 4343 if (type == eSymbolTypeInvalid) { 4344 const char *symbol_sect_name = 4345 symbol_section->GetName().AsCString(); 4346 if (symbol_section->IsDescendant(text_section_sp.get())) { 4347 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4348 S_ATTR_SELF_MODIFYING_CODE | 4349 S_ATTR_SOME_INSTRUCTIONS)) 4350 type = eSymbolTypeData; 4351 else 4352 type = eSymbolTypeCode; 4353 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 4354 symbol_section->IsDescendant( 4355 data_dirty_section_sp.get()) || 4356 symbol_section->IsDescendant( 4357 data_const_section_sp.get())) { 4358 if (symbol_sect_name && 4359 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 4360 type = eSymbolTypeRuntime; 4361 4362 if (symbol_name) { 4363 llvm::StringRef symbol_name_ref(symbol_name); 4364 if (symbol_name_ref.startswith("_OBJC_")) { 4365 llvm::StringRef g_objc_v2_prefix_class( 4366 "_OBJC_CLASS_$_"); 4367 llvm::StringRef g_objc_v2_prefix_metaclass( 4368 "_OBJC_METACLASS_$_"); 4369 llvm::StringRef g_objc_v2_prefix_ivar( 4370 "_OBJC_IVAR_$_"); 4371 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 4372 symbol_name_non_abi_mangled = symbol_name + 1; 4373 symbol_name = 4374 symbol_name + g_objc_v2_prefix_class.size(); 4375 type = eSymbolTypeObjCClass; 4376 demangled_is_synthesized = true; 4377 } else if (symbol_name_ref.startswith( 4378 g_objc_v2_prefix_metaclass)) { 4379 symbol_name_non_abi_mangled = symbol_name + 1; 4380 symbol_name = 4381 symbol_name + g_objc_v2_prefix_metaclass.size(); 4382 type = eSymbolTypeObjCMetaClass; 4383 demangled_is_synthesized = true; 4384 } else if (symbol_name_ref.startswith( 4385 g_objc_v2_prefix_ivar)) { 4386 symbol_name_non_abi_mangled = symbol_name + 1; 4387 symbol_name = 4388 symbol_name + g_objc_v2_prefix_ivar.size(); 4389 type = eSymbolTypeObjCIVar; 4390 demangled_is_synthesized = true; 4391 } 4392 } 4393 } 4394 } else if (symbol_sect_name && 4395 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4396 symbol_sect_name) { 4397 type = eSymbolTypeException; 4398 } else { 4399 type = eSymbolTypeData; 4400 } 4401 } else if (symbol_sect_name && 4402 ::strstr(symbol_sect_name, "__IMPORT") == 4403 symbol_sect_name) { 4404 type = eSymbolTypeTrampoline; 4405 } else if (symbol_section->IsDescendant(objc_section_sp.get())) { 4406 type = eSymbolTypeRuntime; 4407 if (symbol_name && symbol_name[0] == '.') { 4408 llvm::StringRef symbol_name_ref(symbol_name); 4409 llvm::StringRef g_objc_v1_prefix_class( 4410 ".objc_class_name_"); 4411 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4412 symbol_name_non_abi_mangled = symbol_name; 4413 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4414 type = eSymbolTypeObjCClass; 4415 demangled_is_synthesized = true; 4416 } 4417 } 4418 } 4419 } 4420 } 4421 } break; 4422 } 4423 } 4424 4425 if (!add_nlist) { 4426 sym[sym_idx].Clear(); 4427 return true; 4428 } 4429 4430 uint64_t symbol_value = nlist.n_value; 4431 4432 if (symbol_name_non_abi_mangled) { 4433 sym[sym_idx].GetMangled().SetMangledName( 4434 ConstString(symbol_name_non_abi_mangled)); 4435 sym[sym_idx].GetMangled().SetDemangledName(ConstString(symbol_name)); 4436 } else { 4437 bool symbol_name_is_mangled = false; 4438 4439 if (symbol_name && symbol_name[0] == '_') { 4440 symbol_name_is_mangled = symbol_name[1] == '_'; 4441 symbol_name++; // Skip the leading underscore 4442 } 4443 4444 if (symbol_name) { 4445 ConstString const_symbol_name(symbol_name); 4446 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4447 symbol_name_is_mangled); 4448 } 4449 } 4450 4451 if (is_gsym) { 4452 const char *gsym_name = sym[sym_idx] 4453 .GetMangled() 4454 .GetName(Mangled::ePreferMangled) 4455 .GetCString(); 4456 if (gsym_name) 4457 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4458 } 4459 4460 if (symbol_section) { 4461 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4462 if (symbol_byte_size == 0 && function_starts_count > 0) { 4463 addr_t symbol_lookup_file_addr = nlist.n_value; 4464 // Do an exact address match for non-ARM addresses, else get the 4465 // closest since the symbol might be a thumb symbol which has an 4466 // address with bit zero set. 4467 FunctionStarts::Entry *func_start_entry = 4468 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4469 if (is_arm && func_start_entry) { 4470 // Verify that the function start address is the symbol address 4471 // (ARM) or the symbol address + 1 (thumb). 4472 if (func_start_entry->addr != symbol_lookup_file_addr && 4473 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4474 // Not the right entry, NULL it out... 4475 func_start_entry = nullptr; 4476 } 4477 } 4478 if (func_start_entry) { 4479 func_start_entry->data = true; 4480 4481 addr_t symbol_file_addr = func_start_entry->addr; 4482 if (is_arm) 4483 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4484 4485 const FunctionStarts::Entry *next_func_start_entry = 4486 function_starts.FindNextEntry(func_start_entry); 4487 const addr_t section_end_file_addr = 4488 section_file_addr + symbol_section->GetByteSize(); 4489 if (next_func_start_entry) { 4490 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4491 // Be sure the clear the Thumb address bit when we calculate the 4492 // size from the current and next address 4493 if (is_arm) 4494 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4495 symbol_byte_size = std::min<lldb::addr_t>( 4496 next_symbol_file_addr - symbol_file_addr, 4497 section_end_file_addr - symbol_file_addr); 4498 } else { 4499 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4500 } 4501 } 4502 } 4503 symbol_value -= section_file_addr; 4504 } 4505 4506 if (!is_debug) { 4507 if (type == eSymbolTypeCode) { 4508 // See if we can find a N_FUN entry for any code symbols. If we do 4509 // find a match, and the name matches, then we can merge the two into 4510 // just the function symbol to avoid duplicate entries in the symbol 4511 // table. 4512 std::pair<ValueToSymbolIndexMap::const_iterator, 4513 ValueToSymbolIndexMap::const_iterator> 4514 range; 4515 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4516 if (range.first != range.second) { 4517 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4518 pos != range.second; ++pos) { 4519 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4520 sym[pos->second].GetMangled().GetName( 4521 Mangled::ePreferMangled)) { 4522 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4523 // We just need the flags from the linker symbol, so put these 4524 // flags into the N_FUN flags to avoid duplicate symbols in the 4525 // symbol table. 4526 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4527 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4528 if (resolver_addresses.find(nlist.n_value) != 4529 resolver_addresses.end()) 4530 sym[pos->second].SetType(eSymbolTypeResolver); 4531 sym[sym_idx].Clear(); 4532 return true; 4533 } 4534 } 4535 } else { 4536 if (resolver_addresses.find(nlist.n_value) != 4537 resolver_addresses.end()) 4538 type = eSymbolTypeResolver; 4539 } 4540 } else if (type == eSymbolTypeData || type == eSymbolTypeObjCClass || 4541 type == eSymbolTypeObjCMetaClass || 4542 type == eSymbolTypeObjCIVar) { 4543 // See if we can find a N_STSYM entry for any data symbols. If we do 4544 // find a match, and the name matches, then we can merge the two into 4545 // just the Static symbol to avoid duplicate entries in the symbol 4546 // table. 4547 std::pair<ValueToSymbolIndexMap::const_iterator, 4548 ValueToSymbolIndexMap::const_iterator> 4549 range; 4550 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4551 if (range.first != range.second) { 4552 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4553 pos != range.second; ++pos) { 4554 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4555 sym[pos->second].GetMangled().GetName( 4556 Mangled::ePreferMangled)) { 4557 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4558 // We just need the flags from the linker symbol, so put these 4559 // flags into the N_STSYM flags to avoid duplicate symbols in 4560 // the symbol table. 4561 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4562 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4563 sym[sym_idx].Clear(); 4564 return true; 4565 } 4566 } 4567 } else { 4568 // Combine N_GSYM stab entries with the non stab symbol. 4569 const char *gsym_name = sym[sym_idx] 4570 .GetMangled() 4571 .GetName(Mangled::ePreferMangled) 4572 .GetCString(); 4573 if (gsym_name) { 4574 ConstNameToSymbolIndexMap::const_iterator pos = 4575 N_GSYM_name_to_sym_idx.find(gsym_name); 4576 if (pos != N_GSYM_name_to_sym_idx.end()) { 4577 const uint32_t GSYM_sym_idx = pos->second; 4578 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4579 // Copy the address, because often the N_GSYM address has an 4580 // invalid address of zero when the global is a common symbol. 4581 sym[GSYM_sym_idx].GetAddressRef().SetSection(symbol_section); 4582 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4583 add_symbol_addr( 4584 sym[GSYM_sym_idx].GetAddress().GetFileAddress()); 4585 // We just need the flags from the linker symbol, so put these 4586 // flags into the N_GSYM flags to avoid duplicate symbols in 4587 // the symbol table. 4588 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4589 sym[sym_idx].Clear(); 4590 return true; 4591 } 4592 } 4593 } 4594 } 4595 } 4596 4597 sym[sym_idx].SetID(nlist_idx); 4598 sym[sym_idx].SetType(type); 4599 if (set_value) { 4600 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4601 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4602 if (symbol_section) 4603 add_symbol_addr(sym[sym_idx].GetAddress().GetFileAddress()); 4604 } 4605 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4606 if (nlist.n_desc & N_WEAK_REF) 4607 sym[sym_idx].SetIsWeak(true); 4608 4609 if (symbol_byte_size > 0) 4610 sym[sym_idx].SetByteSize(symbol_byte_size); 4611 4612 if (demangled_is_synthesized) 4613 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4614 4615 ++sym_idx; 4616 return true; 4617 }; 4618 4619 // First parse all the nlists but don't process them yet. See the next 4620 // comment for an explanation why. 4621 std::vector<struct nlist_64> nlists; 4622 nlists.reserve(symtab_load_command.nsyms); 4623 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 4624 if (auto nlist = 4625 ParseNList(nlist_data, nlist_data_offset, nlist_byte_size)) 4626 nlists.push_back(*nlist); 4627 else 4628 break; 4629 } 4630 4631 // Now parse all the debug symbols. This is needed to merge non-debug 4632 // symbols in the next step. Non-debug symbols are always coalesced into 4633 // the debug symbol. Doing this in one step would mean that some symbols 4634 // won't be merged. 4635 nlist_idx = 0; 4636 for (auto &nlist : nlists) { 4637 if (!ParseSymbolLambda(nlist, nlist_idx++, DebugSymbols)) 4638 break; 4639 } 4640 4641 // Finally parse all the non debug symbols. 4642 nlist_idx = 0; 4643 for (auto &nlist : nlists) { 4644 if (!ParseSymbolLambda(nlist, nlist_idx++, NonDebugSymbols)) 4645 break; 4646 } 4647 4648 for (const auto &pos : reexport_shlib_needs_fixup) { 4649 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4650 if (undef_pos != undefined_name_to_desc.end()) { 4651 const uint8_t dylib_ordinal = 4652 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4653 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4654 sym[pos.first].SetReExportedSymbolSharedLibrary( 4655 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4656 } 4657 } 4658 } 4659 4660 // Count how many trie symbols we'll add to the symbol table 4661 int trie_symbol_table_augment_count = 0; 4662 for (auto &e : external_sym_trie_entries) { 4663 if (symbols_added.find(e.entry.address) == symbols_added.end()) 4664 trie_symbol_table_augment_count++; 4665 } 4666 4667 if (num_syms < sym_idx + trie_symbol_table_augment_count) { 4668 num_syms = sym_idx + trie_symbol_table_augment_count; 4669 sym = symtab->Resize(num_syms); 4670 } 4671 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4672 4673 // Add symbols from the trie to the symbol table. 4674 for (auto &e : external_sym_trie_entries) { 4675 if (symbols_added.find(e.entry.address) != symbols_added.end()) 4676 continue; 4677 4678 // Find the section that this trie address is in, use that to annotate 4679 // symbol type as we add the trie address and name to the symbol table. 4680 Address symbol_addr; 4681 if (module_sp->ResolveFileAddress(e.entry.address, symbol_addr)) { 4682 SectionSP symbol_section(symbol_addr.GetSection()); 4683 const char *symbol_name = e.entry.name.GetCString(); 4684 bool demangled_is_synthesized = false; 4685 SymbolType type = 4686 GetSymbolType(symbol_name, demangled_is_synthesized, text_section_sp, 4687 data_section_sp, data_dirty_section_sp, 4688 data_const_section_sp, symbol_section); 4689 4690 sym[sym_idx].SetType(type); 4691 if (symbol_section) { 4692 sym[sym_idx].SetID(synthetic_sym_id++); 4693 sym[sym_idx].GetMangled().SetMangledName(ConstString(symbol_name)); 4694 if (demangled_is_synthesized) 4695 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4696 sym[sym_idx].SetIsSynthetic(true); 4697 sym[sym_idx].SetExternal(true); 4698 sym[sym_idx].GetAddressRef() = symbol_addr; 4699 add_symbol_addr(symbol_addr.GetFileAddress()); 4700 if (e.entry.flags & TRIE_SYMBOL_IS_THUMB) 4701 sym[sym_idx].SetFlags(MACHO_NLIST_ARM_SYMBOL_IS_THUMB); 4702 ++sym_idx; 4703 } 4704 } 4705 } 4706 4707 if (function_starts_count > 0) { 4708 uint32_t num_synthetic_function_symbols = 0; 4709 for (i = 0; i < function_starts_count; ++i) { 4710 if (symbols_added.find(function_starts.GetEntryRef(i).addr) == 4711 symbols_added.end()) 4712 ++num_synthetic_function_symbols; 4713 } 4714 4715 if (num_synthetic_function_symbols > 0) { 4716 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4717 num_syms = sym_idx + num_synthetic_function_symbols; 4718 sym = symtab->Resize(num_syms); 4719 } 4720 for (i = 0; i < function_starts_count; ++i) { 4721 const FunctionStarts::Entry *func_start_entry = 4722 function_starts.GetEntryAtIndex(i); 4723 if (symbols_added.find(func_start_entry->addr) == symbols_added.end()) { 4724 addr_t symbol_file_addr = func_start_entry->addr; 4725 uint32_t symbol_flags = 0; 4726 if (func_start_entry->data) 4727 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4728 Address symbol_addr; 4729 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4730 SectionSP symbol_section(symbol_addr.GetSection()); 4731 uint32_t symbol_byte_size = 0; 4732 if (symbol_section) { 4733 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4734 const FunctionStarts::Entry *next_func_start_entry = 4735 function_starts.FindNextEntry(func_start_entry); 4736 const addr_t section_end_file_addr = 4737 section_file_addr + symbol_section->GetByteSize(); 4738 if (next_func_start_entry) { 4739 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4740 if (is_arm) 4741 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4742 symbol_byte_size = std::min<lldb::addr_t>( 4743 next_symbol_file_addr - symbol_file_addr, 4744 section_end_file_addr - symbol_file_addr); 4745 } else { 4746 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4747 } 4748 sym[sym_idx].SetID(synthetic_sym_id++); 4749 // Don't set the name for any synthetic symbols, the Symbol 4750 // object will generate one if needed when the name is accessed 4751 // via accessors. 4752 sym[sym_idx].GetMangled().SetDemangledName(ConstString()); 4753 sym[sym_idx].SetType(eSymbolTypeCode); 4754 sym[sym_idx].SetIsSynthetic(true); 4755 sym[sym_idx].GetAddressRef() = symbol_addr; 4756 add_symbol_addr(symbol_addr.GetFileAddress()); 4757 if (symbol_flags) 4758 sym[sym_idx].SetFlags(symbol_flags); 4759 if (symbol_byte_size) 4760 sym[sym_idx].SetByteSize(symbol_byte_size); 4761 ++sym_idx; 4762 } 4763 } 4764 } 4765 } 4766 } 4767 } 4768 4769 // Trim our symbols down to just what we ended up with after removing any 4770 // symbols. 4771 if (sym_idx < num_syms) { 4772 num_syms = sym_idx; 4773 sym = symtab->Resize(num_syms); 4774 } 4775 4776 // Now synthesize indirect symbols 4777 if (m_dysymtab.nindirectsyms != 0) { 4778 if (indirect_symbol_index_data.GetByteSize()) { 4779 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4780 m_nlist_idx_to_sym_idx.end(); 4781 4782 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4783 ++sect_idx) { 4784 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4785 S_SYMBOL_STUBS) { 4786 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4787 if (symbol_stub_byte_size == 0) 4788 continue; 4789 4790 const uint32_t num_symbol_stubs = 4791 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4792 4793 if (num_symbol_stubs == 0) 4794 continue; 4795 4796 const uint32_t symbol_stub_index_offset = 4797 m_mach_sections[sect_idx].reserved1; 4798 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) { 4799 const uint32_t symbol_stub_index = 4800 symbol_stub_index_offset + stub_idx; 4801 const lldb::addr_t symbol_stub_addr = 4802 m_mach_sections[sect_idx].addr + 4803 (stub_idx * symbol_stub_byte_size); 4804 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4805 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4806 symbol_stub_offset, 4)) { 4807 const uint32_t stub_sym_id = 4808 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4809 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4810 continue; 4811 4812 NListIndexToSymbolIndexMap::const_iterator index_pos = 4813 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4814 Symbol *stub_symbol = nullptr; 4815 if (index_pos != end_index_pos) { 4816 // We have a remapping from the original nlist index to a 4817 // current symbol index, so just look this up by index 4818 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4819 } else { 4820 // We need to lookup a symbol using the original nlist symbol 4821 // index since this index is coming from the S_SYMBOL_STUBS 4822 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4823 } 4824 4825 if (stub_symbol) { 4826 Address so_addr(symbol_stub_addr, section_list); 4827 4828 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4829 // Change the external symbol into a trampoline that makes 4830 // sense These symbols were N_UNDF N_EXT, and are useless 4831 // to us, so we can re-use them so we don't have to make up 4832 // a synthetic symbol for no good reason. 4833 if (resolver_addresses.find(symbol_stub_addr) == 4834 resolver_addresses.end()) 4835 stub_symbol->SetType(eSymbolTypeTrampoline); 4836 else 4837 stub_symbol->SetType(eSymbolTypeResolver); 4838 stub_symbol->SetExternal(false); 4839 stub_symbol->GetAddressRef() = so_addr; 4840 stub_symbol->SetByteSize(symbol_stub_byte_size); 4841 } else { 4842 // Make a synthetic symbol to describe the trampoline stub 4843 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4844 if (sym_idx >= num_syms) { 4845 sym = symtab->Resize(++num_syms); 4846 stub_symbol = nullptr; // this pointer no longer valid 4847 } 4848 sym[sym_idx].SetID(synthetic_sym_id++); 4849 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4850 if (resolver_addresses.find(symbol_stub_addr) == 4851 resolver_addresses.end()) 4852 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4853 else 4854 sym[sym_idx].SetType(eSymbolTypeResolver); 4855 sym[sym_idx].SetIsSynthetic(true); 4856 sym[sym_idx].GetAddressRef() = so_addr; 4857 add_symbol_addr(so_addr.GetFileAddress()); 4858 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4859 ++sym_idx; 4860 } 4861 } else { 4862 if (log) 4863 log->Warning("symbol stub referencing symbol table symbol " 4864 "%u that isn't in our minimal symbol table, " 4865 "fix this!!!", 4866 stub_sym_id); 4867 } 4868 } 4869 } 4870 } 4871 } 4872 } 4873 } 4874 4875 if (!reexport_trie_entries.empty()) { 4876 for (const auto &e : reexport_trie_entries) { 4877 if (e.entry.import_name) { 4878 // Only add indirect symbols from the Trie entries if we didn't have 4879 // a N_INDR nlist entry for this already 4880 if (indirect_symbol_names.find(e.entry.name) == 4881 indirect_symbol_names.end()) { 4882 // Make a synthetic symbol to describe re-exported symbol. 4883 if (sym_idx >= num_syms) 4884 sym = symtab->Resize(++num_syms); 4885 sym[sym_idx].SetID(synthetic_sym_id++); 4886 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4887 sym[sym_idx].SetType(eSymbolTypeReExported); 4888 sym[sym_idx].SetIsSynthetic(true); 4889 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4890 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4891 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4892 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4893 } 4894 ++sym_idx; 4895 } 4896 } 4897 } 4898 } 4899 4900 // StreamFile s(stdout, false); 4901 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4902 // symtab->Dump(&s, NULL, eSortOrderNone); 4903 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4904 // sizes 4905 symtab->CalculateSymbolSizes(); 4906 4907 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4908 // symtab->Dump(&s, NULL, eSortOrderNone); 4909 4910 return symtab->GetNumSymbols(); 4911 } 4912 4913 void ObjectFileMachO::Dump(Stream *s) { 4914 ModuleSP module_sp(GetModule()); 4915 if (module_sp) { 4916 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4917 s->Printf("%p: ", static_cast<void *>(this)); 4918 s->Indent(); 4919 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4920 s->PutCString("ObjectFileMachO64"); 4921 else 4922 s->PutCString("ObjectFileMachO32"); 4923 4924 *s << ", file = '" << m_file; 4925 ModuleSpecList all_specs; 4926 ModuleSpec base_spec; 4927 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 4928 base_spec, all_specs); 4929 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 4930 *s << "', triple"; 4931 if (e) 4932 s->Printf("[%d]", i); 4933 *s << " = "; 4934 *s << all_specs.GetModuleSpecRefAtIndex(i) 4935 .GetArchitecture() 4936 .GetTriple() 4937 .getTriple(); 4938 } 4939 *s << "\n"; 4940 SectionList *sections = GetSectionList(); 4941 if (sections) 4942 sections->Dump(s->AsRawOstream(), s->GetIndentLevel(), nullptr, true, 4943 UINT32_MAX); 4944 4945 if (m_symtab_up) 4946 m_symtab_up->Dump(s, nullptr, eSortOrderNone); 4947 } 4948 } 4949 4950 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4951 const lldb_private::DataExtractor &data, 4952 lldb::offset_t lc_offset) { 4953 uint32_t i; 4954 llvm::MachO::uuid_command load_cmd; 4955 4956 lldb::offset_t offset = lc_offset; 4957 for (i = 0; i < header.ncmds; ++i) { 4958 const lldb::offset_t cmd_offset = offset; 4959 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4960 break; 4961 4962 if (load_cmd.cmd == LC_UUID) { 4963 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4964 4965 if (uuid_bytes) { 4966 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4967 // We pretend these object files have no UUID to prevent crashing. 4968 4969 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4970 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4971 0xbb, 0x14, 0xf0, 0x0d}; 4972 4973 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4974 return UUID(); 4975 4976 return UUID::fromOptionalData(uuid_bytes, 16); 4977 } 4978 return UUID(); 4979 } 4980 offset = cmd_offset + load_cmd.cmdsize; 4981 } 4982 return UUID(); 4983 } 4984 4985 static llvm::StringRef GetOSName(uint32_t cmd) { 4986 switch (cmd) { 4987 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4988 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4989 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4990 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4991 case llvm::MachO::LC_VERSION_MIN_TVOS: 4992 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4993 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4994 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4995 default: 4996 llvm_unreachable("unexpected LC_VERSION load command"); 4997 } 4998 } 4999 5000 namespace { 5001 struct OSEnv { 5002 llvm::StringRef os_type; 5003 llvm::StringRef environment; 5004 OSEnv(uint32_t cmd) { 5005 switch (cmd) { 5006 case llvm::MachO::PLATFORM_MACOS: 5007 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 5008 return; 5009 case llvm::MachO::PLATFORM_IOS: 5010 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 5011 return; 5012 case llvm::MachO::PLATFORM_TVOS: 5013 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 5014 return; 5015 case llvm::MachO::PLATFORM_WATCHOS: 5016 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 5017 return; 5018 // TODO: add BridgeOS & DriverKit once in llvm/lib/Support/Triple.cpp 5019 // NEED_BRIDGEOS_TRIPLE 5020 // case llvm::MachO::PLATFORM_BRIDGEOS: 5021 // os_type = llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 5022 // return; 5023 // case llvm::MachO::PLATFORM_DRIVERKIT: 5024 // os_type = llvm::Triple::getOSTypeName(llvm::Triple::DriverKit); 5025 // return; 5026 case llvm::MachO::PLATFORM_MACCATALYST: 5027 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 5028 environment = llvm::Triple::getEnvironmentTypeName(llvm::Triple::MacABI); 5029 return; 5030 case llvm::MachO::PLATFORM_IOSSIMULATOR: 5031 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 5032 environment = 5033 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 5034 return; 5035 case llvm::MachO::PLATFORM_TVOSSIMULATOR: 5036 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 5037 environment = 5038 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 5039 return; 5040 case llvm::MachO::PLATFORM_WATCHOSSIMULATOR: 5041 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 5042 environment = 5043 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 5044 return; 5045 default: { 5046 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5047 LIBLLDB_LOG_PROCESS)); 5048 LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION"); 5049 } 5050 } 5051 } 5052 }; 5053 5054 struct MinOS { 5055 uint32_t major_version, minor_version, patch_version; 5056 MinOS(uint32_t version) 5057 : major_version(version >> 16), minor_version((version >> 8) & 0xffu), 5058 patch_version(version & 0xffu) {} 5059 }; 5060 } // namespace 5061 5062 void ObjectFileMachO::GetAllArchSpecs(const llvm::MachO::mach_header &header, 5063 const lldb_private::DataExtractor &data, 5064 lldb::offset_t lc_offset, 5065 ModuleSpec &base_spec, 5066 lldb_private::ModuleSpecList &all_specs) { 5067 auto &base_arch = base_spec.GetArchitecture(); 5068 base_arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 5069 if (!base_arch.IsValid()) 5070 return; 5071 5072 bool found_any = false; 5073 auto add_triple = [&](const llvm::Triple &triple) { 5074 auto spec = base_spec; 5075 spec.GetArchitecture().GetTriple() = triple; 5076 if (spec.GetArchitecture().IsValid()) { 5077 spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset); 5078 all_specs.Append(spec); 5079 found_any = true; 5080 } 5081 }; 5082 5083 // Set OS to an unspecified unknown or a "*" so it can match any OS 5084 llvm::Triple base_triple = base_arch.GetTriple(); 5085 base_triple.setOS(llvm::Triple::UnknownOS); 5086 base_triple.setOSName(llvm::StringRef()); 5087 5088 if (header.filetype == MH_PRELOAD) { 5089 if (header.cputype == CPU_TYPE_ARM) { 5090 // If this is a 32-bit arm binary, and it's a standalone binary, force 5091 // the Vendor to Apple so we don't accidentally pick up the generic 5092 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 5093 // frame pointer register; most other armv7 ABIs use a combination of 5094 // r7 and r11. 5095 base_triple.setVendor(llvm::Triple::Apple); 5096 } else { 5097 // Set vendor to an unspecified unknown or a "*" so it can match any 5098 // vendor This is required for correct behavior of EFI debugging on 5099 // x86_64 5100 base_triple.setVendor(llvm::Triple::UnknownVendor); 5101 base_triple.setVendorName(llvm::StringRef()); 5102 } 5103 return add_triple(base_triple); 5104 } 5105 5106 llvm::MachO::load_command load_cmd; 5107 5108 // See if there is an LC_VERSION_MIN_* load command that can give 5109 // us the OS type. 5110 lldb::offset_t offset = lc_offset; 5111 for (uint32_t i = 0; i < header.ncmds; ++i) { 5112 const lldb::offset_t cmd_offset = offset; 5113 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 5114 break; 5115 5116 llvm::MachO::version_min_command version_min; 5117 switch (load_cmd.cmd) { 5118 case llvm::MachO::LC_VERSION_MIN_MACOSX: 5119 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 5120 case llvm::MachO::LC_VERSION_MIN_TVOS: 5121 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 5122 if (load_cmd.cmdsize != sizeof(version_min)) 5123 break; 5124 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 5125 data.GetByteOrder(), &version_min) == 0) 5126 break; 5127 MinOS min_os(version_min.version); 5128 llvm::SmallString<32> os_name; 5129 llvm::raw_svector_ostream os(os_name); 5130 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 5131 << min_os.minor_version << '.' << min_os.patch_version; 5132 5133 auto triple = base_triple; 5134 triple.setOSName(os.str()); 5135 5136 // Disambiguate legacy simulator platforms. 5137 if (load_cmd.cmd != llvm::MachO::LC_VERSION_MIN_MACOSX && 5138 (base_triple.getArch() == llvm::Triple::x86_64 || 5139 base_triple.getArch() == llvm::Triple::x86)) { 5140 // The combination of legacy LC_VERSION_MIN load command and 5141 // x86 architecture always indicates a simulator environment. 5142 // The combination of LC_VERSION_MIN and arm architecture only 5143 // appears for native binaries. Back-deploying simulator 5144 // binaries on Apple Silicon Macs use the modern unambigous 5145 // LC_BUILD_VERSION load commands; no special handling required. 5146 triple.setEnvironment(llvm::Triple::Simulator); 5147 } 5148 add_triple(triple); 5149 break; 5150 } 5151 default: 5152 break; 5153 } 5154 5155 offset = cmd_offset + load_cmd.cmdsize; 5156 } 5157 5158 // See if there are LC_BUILD_VERSION load commands that can give 5159 // us the OS type. 5160 offset = lc_offset; 5161 for (uint32_t i = 0; i < header.ncmds; ++i) { 5162 const lldb::offset_t cmd_offset = offset; 5163 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 5164 break; 5165 5166 do { 5167 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 5168 llvm::MachO::build_version_command build_version; 5169 if (load_cmd.cmdsize < sizeof(build_version)) { 5170 // Malformed load command. 5171 break; 5172 } 5173 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 5174 data.GetByteOrder(), &build_version) == 0) 5175 break; 5176 MinOS min_os(build_version.minos); 5177 OSEnv os_env(build_version.platform); 5178 llvm::SmallString<16> os_name; 5179 llvm::raw_svector_ostream os(os_name); 5180 os << os_env.os_type << min_os.major_version << '.' 5181 << min_os.minor_version << '.' << min_os.patch_version; 5182 auto triple = base_triple; 5183 triple.setOSName(os.str()); 5184 os_name.clear(); 5185 if (!os_env.environment.empty()) 5186 triple.setEnvironmentName(os_env.environment); 5187 add_triple(triple); 5188 } 5189 } while (0); 5190 offset = cmd_offset + load_cmd.cmdsize; 5191 } 5192 5193 if (!found_any) { 5194 if (header.filetype == MH_KEXT_BUNDLE) { 5195 base_triple.setVendor(llvm::Triple::Apple); 5196 add_triple(base_triple); 5197 } else { 5198 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 5199 // so lets not say our Vendor is Apple, leave it as an unspecified 5200 // unknown. 5201 base_triple.setVendor(llvm::Triple::UnknownVendor); 5202 base_triple.setVendorName(llvm::StringRef()); 5203 add_triple(base_triple); 5204 } 5205 } 5206 } 5207 5208 ArchSpec ObjectFileMachO::GetArchitecture( 5209 ModuleSP module_sp, const llvm::MachO::mach_header &header, 5210 const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) { 5211 ModuleSpecList all_specs; 5212 ModuleSpec base_spec; 5213 GetAllArchSpecs(header, data, MachHeaderSizeFromMagic(header.magic), 5214 base_spec, all_specs); 5215 5216 // If the object file offers multiple alternative load commands, 5217 // pick the one that matches the module. 5218 if (module_sp) { 5219 const ArchSpec &module_arch = module_sp->GetArchitecture(); 5220 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 5221 ArchSpec mach_arch = 5222 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 5223 if (module_arch.IsCompatibleMatch(mach_arch)) 5224 return mach_arch; 5225 } 5226 } 5227 5228 // Return the first arch we found. 5229 if (all_specs.GetSize() == 0) 5230 return {}; 5231 return all_specs.GetModuleSpecRefAtIndex(0).GetArchitecture(); 5232 } 5233 5234 UUID ObjectFileMachO::GetUUID() { 5235 ModuleSP module_sp(GetModule()); 5236 if (module_sp) { 5237 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5238 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5239 return GetUUID(m_header, m_data, offset); 5240 } 5241 return UUID(); 5242 } 5243 5244 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5245 uint32_t count = 0; 5246 ModuleSP module_sp(GetModule()); 5247 if (module_sp) { 5248 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5249 llvm::MachO::load_command load_cmd; 5250 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5251 std::vector<std::string> rpath_paths; 5252 std::vector<std::string> rpath_relative_paths; 5253 std::vector<std::string> at_exec_relative_paths; 5254 uint32_t i; 5255 for (i = 0; i < m_header.ncmds; ++i) { 5256 const uint32_t cmd_offset = offset; 5257 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5258 break; 5259 5260 switch (load_cmd.cmd) { 5261 case LC_RPATH: 5262 case LC_LOAD_DYLIB: 5263 case LC_LOAD_WEAK_DYLIB: 5264 case LC_REEXPORT_DYLIB: 5265 case LC_LOAD_DYLINKER: 5266 case LC_LOADFVMLIB: 5267 case LC_LOAD_UPWARD_DYLIB: { 5268 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5269 const char *path = m_data.PeekCStr(name_offset); 5270 if (path) { 5271 if (load_cmd.cmd == LC_RPATH) 5272 rpath_paths.push_back(path); 5273 else { 5274 if (path[0] == '@') { 5275 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5276 rpath_relative_paths.push_back(path + strlen("@rpath")); 5277 else if (strncmp(path, "@executable_path", 5278 strlen("@executable_path")) == 0) 5279 at_exec_relative_paths.push_back(path + 5280 strlen("@executable_path")); 5281 } else { 5282 FileSpec file_spec(path); 5283 if (files.AppendIfUnique(file_spec)) 5284 count++; 5285 } 5286 } 5287 } 5288 } break; 5289 5290 default: 5291 break; 5292 } 5293 offset = cmd_offset + load_cmd.cmdsize; 5294 } 5295 5296 FileSpec this_file_spec(m_file); 5297 FileSystem::Instance().Resolve(this_file_spec); 5298 5299 if (!rpath_paths.empty()) { 5300 // Fixup all LC_RPATH values to be absolute paths 5301 std::string loader_path("@loader_path"); 5302 std::string executable_path("@executable_path"); 5303 for (auto &rpath : rpath_paths) { 5304 if (llvm::StringRef(rpath).startswith(loader_path)) { 5305 rpath.erase(0, loader_path.size()); 5306 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5307 } else if (llvm::StringRef(rpath).startswith(executable_path)) { 5308 rpath.erase(0, executable_path.size()); 5309 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5310 } 5311 } 5312 5313 for (const auto &rpath_relative_path : rpath_relative_paths) { 5314 for (const auto &rpath : rpath_paths) { 5315 std::string path = rpath; 5316 path += rpath_relative_path; 5317 // It is OK to resolve this path because we must find a file on disk 5318 // for us to accept it anyway if it is rpath relative. 5319 FileSpec file_spec(path); 5320 FileSystem::Instance().Resolve(file_spec); 5321 if (FileSystem::Instance().Exists(file_spec) && 5322 files.AppendIfUnique(file_spec)) { 5323 count++; 5324 break; 5325 } 5326 } 5327 } 5328 } 5329 5330 // We may have @executable_paths but no RPATHS. Figure those out here. 5331 // Only do this if this object file is the executable. We have no way to 5332 // get back to the actual executable otherwise, so we won't get the right 5333 // path. 5334 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5335 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5336 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5337 FileSpec file_spec = 5338 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5339 if (FileSystem::Instance().Exists(file_spec) && 5340 files.AppendIfUnique(file_spec)) 5341 count++; 5342 } 5343 } 5344 } 5345 return count; 5346 } 5347 5348 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5349 // If the object file is not an executable it can't hold the entry point. 5350 // m_entry_point_address is initialized to an invalid address, so we can just 5351 // return that. If m_entry_point_address is valid it means we've found it 5352 // already, so return the cached value. 5353 5354 if ((!IsExecutable() && !IsDynamicLoader()) || 5355 m_entry_point_address.IsValid()) { 5356 return m_entry_point_address; 5357 } 5358 5359 // Otherwise, look for the UnixThread or Thread command. The data for the 5360 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5361 // 5362 // uint32_t flavor - this is the flavor argument you would pass to 5363 // thread_get_state 5364 // uint32_t count - this is the count of longs in the thread state data 5365 // struct XXX_thread_state state - this is the structure from 5366 // <machine/thread_status.h> corresponding to the flavor. 5367 // <repeat this trio> 5368 // 5369 // So we just keep reading the various register flavors till we find the GPR 5370 // one, then read the PC out of there. 5371 // FIXME: We will need to have a "RegisterContext data provider" class at some 5372 // point that can get all the registers 5373 // out of data in this form & attach them to a given thread. That should 5374 // underlie the MacOS X User process plugin, and we'll also need it for the 5375 // MacOS X Core File process plugin. When we have that we can also use it 5376 // here. 5377 // 5378 // For now we hard-code the offsets and flavors we need: 5379 // 5380 // 5381 5382 ModuleSP module_sp(GetModule()); 5383 if (module_sp) { 5384 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5385 llvm::MachO::load_command load_cmd; 5386 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5387 uint32_t i; 5388 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5389 bool done = false; 5390 5391 for (i = 0; i < m_header.ncmds; ++i) { 5392 const lldb::offset_t cmd_offset = offset; 5393 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5394 break; 5395 5396 switch (load_cmd.cmd) { 5397 case LC_UNIXTHREAD: 5398 case LC_THREAD: { 5399 while (offset < cmd_offset + load_cmd.cmdsize) { 5400 uint32_t flavor = m_data.GetU32(&offset); 5401 uint32_t count = m_data.GetU32(&offset); 5402 if (count == 0) { 5403 // We've gotten off somehow, log and exit; 5404 return m_entry_point_address; 5405 } 5406 5407 switch (m_header.cputype) { 5408 case llvm::MachO::CPU_TYPE_ARM: 5409 if (flavor == 1 || 5410 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 5411 // from mach/arm/thread_status.h 5412 { 5413 offset += 60; // This is the offset of pc in the GPR thread state 5414 // data structure. 5415 start_address = m_data.GetU32(&offset); 5416 done = true; 5417 } 5418 break; 5419 case llvm::MachO::CPU_TYPE_ARM64: 5420 case llvm::MachO::CPU_TYPE_ARM64_32: 5421 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5422 { 5423 offset += 256; // This is the offset of pc in the GPR thread state 5424 // data structure. 5425 start_address = m_data.GetU64(&offset); 5426 done = true; 5427 } 5428 break; 5429 case llvm::MachO::CPU_TYPE_I386: 5430 if (flavor == 5431 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5432 { 5433 offset += 40; // This is the offset of eip in the GPR thread state 5434 // data structure. 5435 start_address = m_data.GetU32(&offset); 5436 done = true; 5437 } 5438 break; 5439 case llvm::MachO::CPU_TYPE_X86_64: 5440 if (flavor == 5441 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5442 { 5443 offset += 16 * 8; // This is the offset of rip in the GPR thread 5444 // state data structure. 5445 start_address = m_data.GetU64(&offset); 5446 done = true; 5447 } 5448 break; 5449 default: 5450 return m_entry_point_address; 5451 } 5452 // Haven't found the GPR flavor yet, skip over the data for this 5453 // flavor: 5454 if (done) 5455 break; 5456 offset += count * 4; 5457 } 5458 } break; 5459 case LC_MAIN: { 5460 ConstString text_segment_name("__TEXT"); 5461 uint64_t entryoffset = m_data.GetU64(&offset); 5462 SectionSP text_segment_sp = 5463 GetSectionList()->FindSectionByName(text_segment_name); 5464 if (text_segment_sp) { 5465 done = true; 5466 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5467 } 5468 } break; 5469 5470 default: 5471 break; 5472 } 5473 if (done) 5474 break; 5475 5476 // Go to the next load command: 5477 offset = cmd_offset + load_cmd.cmdsize; 5478 } 5479 5480 if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) { 5481 if (GetSymtab()) { 5482 Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType( 5483 ConstString("_dyld_start"), SymbolType::eSymbolTypeCode, 5484 Symtab::eDebugAny, Symtab::eVisibilityAny); 5485 if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) { 5486 start_address = dyld_start_sym->GetAddress().GetFileAddress(); 5487 } 5488 } 5489 } 5490 5491 if (start_address != LLDB_INVALID_ADDRESS) { 5492 // We got the start address from the load commands, so now resolve that 5493 // address in the sections of this ObjectFile: 5494 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5495 start_address, GetSectionList())) { 5496 m_entry_point_address.Clear(); 5497 } 5498 } else { 5499 // We couldn't read the UnixThread load command - maybe it wasn't there. 5500 // As a fallback look for the "start" symbol in the main executable. 5501 5502 ModuleSP module_sp(GetModule()); 5503 5504 if (module_sp) { 5505 SymbolContextList contexts; 5506 SymbolContext context; 5507 module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5508 eSymbolTypeCode, contexts); 5509 if (contexts.GetSize()) { 5510 if (contexts.GetContextAtIndex(0, context)) 5511 m_entry_point_address = context.symbol->GetAddress(); 5512 } 5513 } 5514 } 5515 } 5516 5517 return m_entry_point_address; 5518 } 5519 5520 lldb_private::Address ObjectFileMachO::GetBaseAddress() { 5521 lldb_private::Address header_addr; 5522 SectionList *section_list = GetSectionList(); 5523 if (section_list) { 5524 SectionSP text_segment_sp( 5525 section_list->FindSectionByName(GetSegmentNameTEXT())); 5526 if (text_segment_sp) { 5527 header_addr.SetSection(text_segment_sp); 5528 header_addr.SetOffset(0); 5529 } 5530 } 5531 return header_addr; 5532 } 5533 5534 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5535 ModuleSP module_sp(GetModule()); 5536 if (module_sp) { 5537 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5538 if (!m_thread_context_offsets_valid) { 5539 m_thread_context_offsets_valid = true; 5540 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5541 FileRangeArray::Entry file_range; 5542 llvm::MachO::thread_command thread_cmd; 5543 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5544 const uint32_t cmd_offset = offset; 5545 if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr) 5546 break; 5547 5548 if (thread_cmd.cmd == LC_THREAD) { 5549 file_range.SetRangeBase(offset); 5550 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5551 m_thread_context_offsets.Append(file_range); 5552 } 5553 offset = cmd_offset + thread_cmd.cmdsize; 5554 } 5555 } 5556 } 5557 return m_thread_context_offsets.GetSize(); 5558 } 5559 5560 std::string ObjectFileMachO::GetIdentifierString() { 5561 std::string result; 5562 ModuleSP module_sp(GetModule()); 5563 if (module_sp) { 5564 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5565 5566 // First, look over the load commands for an LC_NOTE load command with 5567 // data_owner string "kern ver str" & use that if found. 5568 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5569 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5570 const uint32_t cmd_offset = offset; 5571 llvm::MachO::load_command lc; 5572 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5573 break; 5574 if (lc.cmd == LC_NOTE) { 5575 char data_owner[17]; 5576 m_data.CopyData(offset, 16, data_owner); 5577 data_owner[16] = '\0'; 5578 offset += 16; 5579 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5580 uint64_t size = m_data.GetU64_unchecked(&offset); 5581 5582 // "kern ver str" has a uint32_t version and then a nul terminated 5583 // c-string. 5584 if (strcmp("kern ver str", data_owner) == 0) { 5585 offset = fileoff; 5586 uint32_t version; 5587 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5588 if (version == 1) { 5589 uint32_t strsize = size - sizeof(uint32_t); 5590 char *buf = (char *)malloc(strsize); 5591 if (buf) { 5592 m_data.CopyData(offset, strsize, buf); 5593 buf[strsize - 1] = '\0'; 5594 result = buf; 5595 if (buf) 5596 free(buf); 5597 return result; 5598 } 5599 } 5600 } 5601 } 5602 } 5603 offset = cmd_offset + lc.cmdsize; 5604 } 5605 5606 // Second, make a pass over the load commands looking for an obsolete 5607 // LC_IDENT load command. 5608 offset = MachHeaderSizeFromMagic(m_header.magic); 5609 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5610 const uint32_t cmd_offset = offset; 5611 llvm::MachO::ident_command ident_command; 5612 if (m_data.GetU32(&offset, &ident_command, 2) == nullptr) 5613 break; 5614 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5615 char *buf = (char *)malloc(ident_command.cmdsize); 5616 if (buf != nullptr && m_data.CopyData(offset, ident_command.cmdsize, 5617 buf) == ident_command.cmdsize) { 5618 buf[ident_command.cmdsize - 1] = '\0'; 5619 result = buf; 5620 } 5621 if (buf) 5622 free(buf); 5623 } 5624 offset = cmd_offset + ident_command.cmdsize; 5625 } 5626 } 5627 return result; 5628 } 5629 5630 addr_t ObjectFileMachO::GetAddressMask() { 5631 addr_t mask = 0; 5632 ModuleSP module_sp(GetModule()); 5633 if (module_sp) { 5634 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5635 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5636 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5637 const uint32_t cmd_offset = offset; 5638 llvm::MachO::load_command lc; 5639 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5640 break; 5641 if (lc.cmd == LC_NOTE) { 5642 char data_owner[17]; 5643 m_data.CopyData(offset, 16, data_owner); 5644 data_owner[16] = '\0'; 5645 offset += 16; 5646 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5647 5648 // "addrable bits" has a uint32_t version and a uint32_t 5649 // number of bits used in addressing. 5650 if (strcmp("addrable bits", data_owner) == 0) { 5651 offset = fileoff; 5652 uint32_t version; 5653 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5654 if (version == 3) { 5655 uint32_t num_addr_bits = m_data.GetU32_unchecked(&offset); 5656 if (num_addr_bits != 0) { 5657 mask = ~((1ULL << num_addr_bits) - 1); 5658 } 5659 break; 5660 } 5661 } 5662 } 5663 } 5664 offset = cmd_offset + lc.cmdsize; 5665 } 5666 } 5667 return mask; 5668 } 5669 5670 bool ObjectFileMachO::GetCorefileMainBinaryInfo(addr_t &address, UUID &uuid, 5671 ObjectFile::BinaryType &type) { 5672 address = LLDB_INVALID_ADDRESS; 5673 uuid.Clear(); 5674 ModuleSP module_sp(GetModule()); 5675 if (module_sp) { 5676 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5677 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5678 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5679 const uint32_t cmd_offset = offset; 5680 llvm::MachO::load_command lc; 5681 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5682 break; 5683 if (lc.cmd == LC_NOTE) { 5684 char data_owner[17]; 5685 memset(data_owner, 0, sizeof(data_owner)); 5686 m_data.CopyData(offset, 16, data_owner); 5687 offset += 16; 5688 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5689 uint64_t size = m_data.GetU64_unchecked(&offset); 5690 5691 // "main bin spec" (main binary specification) data payload is 5692 // formatted: 5693 // uint32_t version [currently 1] 5694 // uint32_t type [0 == unspecified, 1 == kernel, 5695 // 2 == user process, 3 == firmware ] 5696 // uint64_t address [ UINT64_MAX if address not specified ] 5697 // uuid_t uuid [ all zero's if uuid not specified ] 5698 // uint32_t log2_pagesize [ process page size in log base 5699 // 2, e.g. 4k pages are 12. 5700 // 0 for unspecified ] 5701 // uint32_t unused [ for alignment ] 5702 5703 if (strcmp("main bin spec", data_owner) == 0 && size >= 32) { 5704 offset = fileoff; 5705 uint32_t version; 5706 if (m_data.GetU32(&offset, &version, 1) != nullptr && version == 1) { 5707 uint32_t binspec_type = 0; 5708 uuid_t raw_uuid; 5709 memset(raw_uuid, 0, sizeof(uuid_t)); 5710 5711 if (m_data.GetU32(&offset, &binspec_type, 1) && 5712 m_data.GetU64(&offset, &address, 1) && 5713 m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5714 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5715 // convert the "main bin spec" type into our 5716 // ObjectFile::BinaryType enum 5717 switch (binspec_type) { 5718 case 0: 5719 type = eBinaryTypeUnknown; 5720 break; 5721 case 1: 5722 type = eBinaryTypeKernel; 5723 break; 5724 case 2: 5725 type = eBinaryTypeUser; 5726 break; 5727 case 3: 5728 type = eBinaryTypeStandalone; 5729 break; 5730 } 5731 return true; 5732 } 5733 } 5734 } 5735 } 5736 offset = cmd_offset + lc.cmdsize; 5737 } 5738 } 5739 return false; 5740 } 5741 5742 lldb::RegisterContextSP 5743 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5744 lldb_private::Thread &thread) { 5745 lldb::RegisterContextSP reg_ctx_sp; 5746 5747 ModuleSP module_sp(GetModule()); 5748 if (module_sp) { 5749 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5750 if (!m_thread_context_offsets_valid) 5751 GetNumThreadContexts(); 5752 5753 const FileRangeArray::Entry *thread_context_file_range = 5754 m_thread_context_offsets.GetEntryAtIndex(idx); 5755 if (thread_context_file_range) { 5756 5757 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5758 thread_context_file_range->GetByteSize()); 5759 5760 switch (m_header.cputype) { 5761 case llvm::MachO::CPU_TYPE_ARM64: 5762 case llvm::MachO::CPU_TYPE_ARM64_32: 5763 reg_ctx_sp = 5764 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data); 5765 break; 5766 5767 case llvm::MachO::CPU_TYPE_ARM: 5768 reg_ctx_sp = 5769 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data); 5770 break; 5771 5772 case llvm::MachO::CPU_TYPE_I386: 5773 reg_ctx_sp = 5774 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data); 5775 break; 5776 5777 case llvm::MachO::CPU_TYPE_X86_64: 5778 reg_ctx_sp = 5779 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data); 5780 break; 5781 } 5782 } 5783 } 5784 return reg_ctx_sp; 5785 } 5786 5787 ObjectFile::Type ObjectFileMachO::CalculateType() { 5788 switch (m_header.filetype) { 5789 case MH_OBJECT: // 0x1u 5790 if (GetAddressByteSize() == 4) { 5791 // 32 bit kexts are just object files, but they do have a valid 5792 // UUID load command. 5793 if (GetUUID()) { 5794 // this checking for the UUID load command is not enough we could 5795 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5796 // this is required of kexts 5797 if (m_strata == eStrataInvalid) 5798 m_strata = eStrataKernel; 5799 return eTypeSharedLibrary; 5800 } 5801 } 5802 return eTypeObjectFile; 5803 5804 case MH_EXECUTE: 5805 return eTypeExecutable; // 0x2u 5806 case MH_FVMLIB: 5807 return eTypeSharedLibrary; // 0x3u 5808 case MH_CORE: 5809 return eTypeCoreFile; // 0x4u 5810 case MH_PRELOAD: 5811 return eTypeSharedLibrary; // 0x5u 5812 case MH_DYLIB: 5813 return eTypeSharedLibrary; // 0x6u 5814 case MH_DYLINKER: 5815 return eTypeDynamicLinker; // 0x7u 5816 case MH_BUNDLE: 5817 return eTypeSharedLibrary; // 0x8u 5818 case MH_DYLIB_STUB: 5819 return eTypeStubLibrary; // 0x9u 5820 case MH_DSYM: 5821 return eTypeDebugInfo; // 0xAu 5822 case MH_KEXT_BUNDLE: 5823 return eTypeSharedLibrary; // 0xBu 5824 default: 5825 break; 5826 } 5827 return eTypeUnknown; 5828 } 5829 5830 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5831 switch (m_header.filetype) { 5832 case MH_OBJECT: // 0x1u 5833 { 5834 // 32 bit kexts are just object files, but they do have a valid 5835 // UUID load command. 5836 if (GetUUID()) { 5837 // this checking for the UUID load command is not enough we could 5838 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5839 // this is required of kexts 5840 if (m_type == eTypeInvalid) 5841 m_type = eTypeSharedLibrary; 5842 5843 return eStrataKernel; 5844 } 5845 } 5846 return eStrataUnknown; 5847 5848 case MH_EXECUTE: // 0x2u 5849 // Check for the MH_DYLDLINK bit in the flags 5850 if (m_header.flags & MH_DYLDLINK) { 5851 return eStrataUser; 5852 } else { 5853 SectionList *section_list = GetSectionList(); 5854 if (section_list) { 5855 static ConstString g_kld_section_name("__KLD"); 5856 if (section_list->FindSectionByName(g_kld_section_name)) 5857 return eStrataKernel; 5858 } 5859 } 5860 return eStrataRawImage; 5861 5862 case MH_FVMLIB: 5863 return eStrataUser; // 0x3u 5864 case MH_CORE: 5865 return eStrataUnknown; // 0x4u 5866 case MH_PRELOAD: 5867 return eStrataRawImage; // 0x5u 5868 case MH_DYLIB: 5869 return eStrataUser; // 0x6u 5870 case MH_DYLINKER: 5871 return eStrataUser; // 0x7u 5872 case MH_BUNDLE: 5873 return eStrataUser; // 0x8u 5874 case MH_DYLIB_STUB: 5875 return eStrataUser; // 0x9u 5876 case MH_DSYM: 5877 return eStrataUnknown; // 0xAu 5878 case MH_KEXT_BUNDLE: 5879 return eStrataKernel; // 0xBu 5880 default: 5881 break; 5882 } 5883 return eStrataUnknown; 5884 } 5885 5886 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5887 ModuleSP module_sp(GetModule()); 5888 if (module_sp) { 5889 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5890 llvm::MachO::dylib_command load_cmd; 5891 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5892 uint32_t version_cmd = 0; 5893 uint64_t version = 0; 5894 uint32_t i; 5895 for (i = 0; i < m_header.ncmds; ++i) { 5896 const lldb::offset_t cmd_offset = offset; 5897 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5898 break; 5899 5900 if (load_cmd.cmd == LC_ID_DYLIB) { 5901 if (version_cmd == 0) { 5902 version_cmd = load_cmd.cmd; 5903 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr) 5904 break; 5905 version = load_cmd.dylib.current_version; 5906 } 5907 break; // Break for now unless there is another more complete version 5908 // number load command in the future. 5909 } 5910 offset = cmd_offset + load_cmd.cmdsize; 5911 } 5912 5913 if (version_cmd == LC_ID_DYLIB) { 5914 unsigned major = (version & 0xFFFF0000ull) >> 16; 5915 unsigned minor = (version & 0x0000FF00ull) >> 8; 5916 unsigned subminor = (version & 0x000000FFull); 5917 return llvm::VersionTuple(major, minor, subminor); 5918 } 5919 } 5920 return llvm::VersionTuple(); 5921 } 5922 5923 ArchSpec ObjectFileMachO::GetArchitecture() { 5924 ModuleSP module_sp(GetModule()); 5925 ArchSpec arch; 5926 if (module_sp) { 5927 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5928 5929 return GetArchitecture(module_sp, m_header, m_data, 5930 MachHeaderSizeFromMagic(m_header.magic)); 5931 } 5932 return arch; 5933 } 5934 5935 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, 5936 addr_t &base_addr, UUID &uuid) { 5937 uuid.Clear(); 5938 base_addr = LLDB_INVALID_ADDRESS; 5939 if (process && process->GetDynamicLoader()) { 5940 DynamicLoader *dl = process->GetDynamicLoader(); 5941 LazyBool using_shared_cache; 5942 LazyBool private_shared_cache; 5943 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5944 private_shared_cache); 5945 } 5946 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5947 LIBLLDB_LOG_PROCESS)); 5948 LLDB_LOGF( 5949 log, 5950 "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64, 5951 uuid.GetAsString().c_str(), base_addr); 5952 } 5953 5954 // From dyld SPI header dyld_process_info.h 5955 typedef void *dyld_process_info; 5956 struct lldb_copy__dyld_process_cache_info { 5957 uuid_t cacheUUID; // UUID of cache used by process 5958 uint64_t cacheBaseAddress; // load address of dyld shared cache 5959 bool noCache; // process is running without a dyld cache 5960 bool privateCache; // process is using a private copy of its dyld cache 5961 }; 5962 5963 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with 5964 // llvm enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile 5965 // errors. So we need to use the actual underlying types of task_t and 5966 // kern_return_t below. 5967 extern "C" unsigned int /*task_t*/ mach_task_self(); 5968 5969 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5970 uuid.Clear(); 5971 base_addr = LLDB_INVALID_ADDRESS; 5972 5973 #if defined(__APPLE__) 5974 uint8_t *(*dyld_get_all_image_infos)(void); 5975 dyld_get_all_image_infos = 5976 (uint8_t * (*)()) dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5977 if (dyld_get_all_image_infos) { 5978 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5979 if (dyld_all_image_infos_address) { 5980 uint32_t *version = (uint32_t *) 5981 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5982 if (*version >= 13) { 5983 uuid_t *sharedCacheUUID_address = 0; 5984 int wordsize = sizeof(uint8_t *); 5985 if (wordsize == 8) { 5986 sharedCacheUUID_address = 5987 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5988 160); // sharedCacheUUID <mach-o/dyld_images.h> 5989 if (*version >= 15) 5990 base_addr = 5991 *(uint64_t 5992 *)((uint8_t *)dyld_all_image_infos_address + 5993 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5994 } else { 5995 sharedCacheUUID_address = 5996 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5997 84); // sharedCacheUUID <mach-o/dyld_images.h> 5998 if (*version >= 15) { 5999 base_addr = 0; 6000 base_addr = 6001 *(uint32_t 6002 *)((uint8_t *)dyld_all_image_infos_address + 6003 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 6004 } 6005 } 6006 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 6007 } 6008 } 6009 } else { 6010 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 6011 dyld_process_info (*dyld_process_info_create)( 6012 unsigned int /* task_t */ task, uint64_t timestamp, 6013 unsigned int /*kern_return_t*/ *kernelError); 6014 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 6015 void (*dyld_process_info_release)(dyld_process_info info); 6016 6017 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, 6018 unsigned int /*kern_return_t*/ *)) 6019 dlsym(RTLD_DEFAULT, "_dyld_process_info_create"); 6020 dyld_process_info_get_cache = (void (*)(void *, void *))dlsym( 6021 RTLD_DEFAULT, "_dyld_process_info_get_cache"); 6022 dyld_process_info_release = 6023 (void (*)(void *))dlsym(RTLD_DEFAULT, "_dyld_process_info_release"); 6024 6025 if (dyld_process_info_create && dyld_process_info_get_cache) { 6026 unsigned int /*kern_return_t */ kern_ret; 6027 dyld_process_info process_info = 6028 dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 6029 if (process_info) { 6030 struct lldb_copy__dyld_process_cache_info sc_info; 6031 memset(&sc_info, 0, sizeof(struct lldb_copy__dyld_process_cache_info)); 6032 dyld_process_info_get_cache(process_info, &sc_info); 6033 if (sc_info.cacheBaseAddress != 0) { 6034 base_addr = sc_info.cacheBaseAddress; 6035 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 6036 } 6037 dyld_process_info_release(process_info); 6038 } 6039 } 6040 } 6041 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 6042 LIBLLDB_LOG_PROCESS)); 6043 if (log && uuid.IsValid()) 6044 LLDB_LOGF(log, 6045 "lldb's in-memory shared cache has a UUID of %s base address of " 6046 "0x%" PRIx64, 6047 uuid.GetAsString().c_str(), base_addr); 6048 #endif 6049 } 6050 6051 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 6052 if (!m_min_os_version) { 6053 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 6054 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6055 const lldb::offset_t load_cmd_offset = offset; 6056 6057 llvm::MachO::version_min_command lc; 6058 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6059 break; 6060 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 6061 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 6062 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 6063 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 6064 if (m_data.GetU32(&offset, &lc.version, 6065 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 6066 const uint32_t xxxx = lc.version >> 16; 6067 const uint32_t yy = (lc.version >> 8) & 0xffu; 6068 const uint32_t zz = lc.version & 0xffu; 6069 if (xxxx) { 6070 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 6071 break; 6072 } 6073 } 6074 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 6075 // struct build_version_command { 6076 // uint32_t cmd; /* LC_BUILD_VERSION */ 6077 // uint32_t cmdsize; /* sizeof(struct 6078 // build_version_command) plus */ 6079 // /* ntools * sizeof(struct 6080 // build_tool_version) */ 6081 // uint32_t platform; /* platform */ 6082 // uint32_t minos; /* X.Y.Z is encoded in nibbles 6083 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded in 6084 // nibbles xxxx.yy.zz */ uint32_t ntools; /* number of 6085 // tool entries following this */ 6086 // }; 6087 6088 offset += 4; // skip platform 6089 uint32_t minos = m_data.GetU32(&offset); 6090 6091 const uint32_t xxxx = minos >> 16; 6092 const uint32_t yy = (minos >> 8) & 0xffu; 6093 const uint32_t zz = minos & 0xffu; 6094 if (xxxx) { 6095 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 6096 break; 6097 } 6098 } 6099 6100 offset = load_cmd_offset + lc.cmdsize; 6101 } 6102 6103 if (!m_min_os_version) { 6104 // Set version to an empty value so we don't keep trying to 6105 m_min_os_version = llvm::VersionTuple(); 6106 } 6107 } 6108 6109 return *m_min_os_version; 6110 } 6111 6112 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { 6113 if (!m_sdk_versions.hasValue()) { 6114 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 6115 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6116 const lldb::offset_t load_cmd_offset = offset; 6117 6118 llvm::MachO::version_min_command lc; 6119 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6120 break; 6121 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 6122 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 6123 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 6124 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 6125 if (m_data.GetU32(&offset, &lc.version, 6126 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 6127 const uint32_t xxxx = lc.sdk >> 16; 6128 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 6129 const uint32_t zz = lc.sdk & 0xffu; 6130 if (xxxx) { 6131 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 6132 break; 6133 } else { 6134 GetModule()->ReportWarning("minimum OS version load command with " 6135 "invalid (0) version found."); 6136 } 6137 } 6138 } 6139 offset = load_cmd_offset + lc.cmdsize; 6140 } 6141 6142 if (!m_sdk_versions.hasValue()) { 6143 offset = MachHeaderSizeFromMagic(m_header.magic); 6144 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6145 const lldb::offset_t load_cmd_offset = offset; 6146 6147 llvm::MachO::version_min_command lc; 6148 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6149 break; 6150 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 6151 // struct build_version_command { 6152 // uint32_t cmd; /* LC_BUILD_VERSION */ 6153 // uint32_t cmdsize; /* sizeof(struct 6154 // build_version_command) plus */ 6155 // /* ntools * sizeof(struct 6156 // build_tool_version) */ 6157 // uint32_t platform; /* platform */ 6158 // uint32_t minos; /* X.Y.Z is encoded in nibbles 6159 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded 6160 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number 6161 // of tool entries following this */ 6162 // }; 6163 6164 offset += 4; // skip platform 6165 uint32_t minos = m_data.GetU32(&offset); 6166 6167 const uint32_t xxxx = minos >> 16; 6168 const uint32_t yy = (minos >> 8) & 0xffu; 6169 const uint32_t zz = minos & 0xffu; 6170 if (xxxx) { 6171 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 6172 break; 6173 } 6174 } 6175 offset = load_cmd_offset + lc.cmdsize; 6176 } 6177 } 6178 6179 if (!m_sdk_versions.hasValue()) 6180 m_sdk_versions = llvm::VersionTuple(); 6181 } 6182 6183 return m_sdk_versions.getValue(); 6184 } 6185 6186 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 6187 return m_header.filetype == llvm::MachO::MH_DYLINKER; 6188 } 6189 6190 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 6191 return m_allow_assembly_emulation_unwind_plans; 6192 } 6193 6194 // PluginInterface protocol 6195 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 6196 return GetPluginNameStatic(); 6197 } 6198 6199 Section *ObjectFileMachO::GetMachHeaderSection() { 6200 // Find the first address of the mach header which is the first non-zero file 6201 // sized section whose file offset is zero. This is the base file address of 6202 // the mach-o file which can be subtracted from the vmaddr of the other 6203 // segments found in memory and added to the load address 6204 ModuleSP module_sp = GetModule(); 6205 if (!module_sp) 6206 return nullptr; 6207 SectionList *section_list = GetSectionList(); 6208 if (!section_list) 6209 return nullptr; 6210 const size_t num_sections = section_list->GetSize(); 6211 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6212 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 6213 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 6214 return section; 6215 } 6216 return nullptr; 6217 } 6218 6219 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 6220 if (!section) 6221 return false; 6222 const bool is_dsym = (m_header.filetype == MH_DSYM); 6223 if (section->GetFileSize() == 0 && !is_dsym) 6224 return false; 6225 if (section->IsThreadSpecific()) 6226 return false; 6227 if (GetModule().get() != section->GetModule().get()) 6228 return false; 6229 // Be careful with __LINKEDIT and __DWARF segments 6230 if (section->GetName() == GetSegmentNameLINKEDIT() || 6231 section->GetName() == GetSegmentNameDWARF()) { 6232 // Only map __LINKEDIT and __DWARF if we have an in memory image and 6233 // this isn't a kernel binary like a kext or mach_kernel. 6234 const bool is_memory_image = (bool)m_process_wp.lock(); 6235 const Strata strata = GetStrata(); 6236 if (is_memory_image == false || strata == eStrataKernel) 6237 return false; 6238 } 6239 return true; 6240 } 6241 6242 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 6243 lldb::addr_t header_load_address, const Section *header_section, 6244 const Section *section) { 6245 ModuleSP module_sp = GetModule(); 6246 if (module_sp && header_section && section && 6247 header_load_address != LLDB_INVALID_ADDRESS) { 6248 lldb::addr_t file_addr = header_section->GetFileAddress(); 6249 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 6250 return section->GetFileAddress() - file_addr + header_load_address; 6251 } 6252 return LLDB_INVALID_ADDRESS; 6253 } 6254 6255 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 6256 bool value_is_offset) { 6257 ModuleSP module_sp = GetModule(); 6258 if (!module_sp) 6259 return false; 6260 6261 SectionList *section_list = GetSectionList(); 6262 if (!section_list) 6263 return false; 6264 6265 size_t num_loaded_sections = 0; 6266 const size_t num_sections = section_list->GetSize(); 6267 6268 if (value_is_offset) { 6269 // "value" is an offset to apply to each top level segment 6270 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6271 // Iterate through the object file sections to find all of the 6272 // sections that size on disk (to avoid __PAGEZERO) and load them 6273 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6274 if (SectionIsLoadable(section_sp.get())) 6275 if (target.GetSectionLoadList().SetSectionLoadAddress( 6276 section_sp, section_sp->GetFileAddress() + value)) 6277 ++num_loaded_sections; 6278 } 6279 } else { 6280 // "value" is the new base address of the mach_header, adjust each 6281 // section accordingly 6282 6283 Section *mach_header_section = GetMachHeaderSection(); 6284 if (mach_header_section) { 6285 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6286 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6287 6288 lldb::addr_t section_load_addr = 6289 CalculateSectionLoadAddressForMemoryImage( 6290 value, mach_header_section, section_sp.get()); 6291 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6292 if (target.GetSectionLoadList().SetSectionLoadAddress( 6293 section_sp, section_load_addr)) 6294 ++num_loaded_sections; 6295 } 6296 } 6297 } 6298 } 6299 return num_loaded_sections > 0; 6300 } 6301 6302 struct all_image_infos_header { 6303 uint32_t version; // currently 1 6304 uint32_t imgcount; // number of binary images 6305 uint64_t entries_fileoff; // file offset in the corefile of where the array of 6306 // struct entry's begin. 6307 uint32_t entries_size; // size of 'struct entry'. 6308 uint32_t unused; 6309 }; 6310 6311 struct image_entry { 6312 uint64_t filepath_offset; // offset in corefile to c-string of the file path, 6313 // UINT64_MAX if unavailable. 6314 uuid_t uuid; // uint8_t[16]. should be set to all zeroes if 6315 // uuid is unknown. 6316 uint64_t load_address; // UINT64_MAX if unknown. 6317 uint64_t seg_addrs_offset; // offset to the array of struct segment_vmaddr's. 6318 uint32_t segment_count; // The number of segments for this binary. 6319 uint32_t unused; 6320 6321 image_entry() { 6322 filepath_offset = UINT64_MAX; 6323 memset(&uuid, 0, sizeof(uuid_t)); 6324 segment_count = 0; 6325 load_address = UINT64_MAX; 6326 seg_addrs_offset = UINT64_MAX; 6327 unused = 0; 6328 } 6329 image_entry(const image_entry &rhs) { 6330 filepath_offset = rhs.filepath_offset; 6331 memcpy(&uuid, &rhs.uuid, sizeof(uuid_t)); 6332 segment_count = rhs.segment_count; 6333 seg_addrs_offset = rhs.seg_addrs_offset; 6334 load_address = rhs.load_address; 6335 unused = rhs.unused; 6336 } 6337 }; 6338 6339 struct segment_vmaddr { 6340 char segname[16]; 6341 uint64_t vmaddr; 6342 uint64_t unused; 6343 6344 segment_vmaddr() { 6345 memset(&segname, 0, 16); 6346 vmaddr = UINT64_MAX; 6347 unused = 0; 6348 } 6349 segment_vmaddr(const segment_vmaddr &rhs) { 6350 memcpy(&segname, &rhs.segname, 16); 6351 vmaddr = rhs.vmaddr; 6352 unused = rhs.unused; 6353 } 6354 }; 6355 6356 // Write the payload for the "all image infos" LC_NOTE into 6357 // the supplied all_image_infos_payload, assuming that this 6358 // will be written into the corefile starting at 6359 // initial_file_offset. 6360 // 6361 // The placement of this payload is a little tricky. We're 6362 // laying this out as 6363 // 6364 // 1. header (struct all_image_info_header) 6365 // 2. Array of fixed-size (struct image_entry)'s, one 6366 // per binary image present in the process. 6367 // 3. Arrays of (struct segment_vmaddr)'s, a varying number 6368 // for each binary image. 6369 // 4. Variable length c-strings of binary image filepaths, 6370 // one per binary. 6371 // 6372 // To compute where everything will be laid out in the 6373 // payload, we need to iterate over the images and calculate 6374 // how many segment_vmaddr structures each image will need, 6375 // and how long each image's filepath c-string is. There 6376 // are some multiple passes over the image list while calculating 6377 // everything. 6378 6379 static offset_t CreateAllImageInfosPayload( 6380 const lldb::ProcessSP &process_sp, offset_t initial_file_offset, 6381 StreamString &all_image_infos_payload, SaveCoreStyle core_style) { 6382 Target &target = process_sp->GetTarget(); 6383 ModuleList modules = target.GetImages(); 6384 6385 // stack-only corefiles have no reason to include binaries that 6386 // are not executing; we're trying to make the smallest corefile 6387 // we can, so leave the rest out. 6388 if (core_style == SaveCoreStyle::eSaveCoreStackOnly) 6389 modules.Clear(); 6390 6391 std::set<std::string> executing_uuids; 6392 ThreadList &thread_list(process_sp->GetThreadList()); 6393 for (uint32_t i = 0; i < thread_list.GetSize(); i++) { 6394 ThreadSP thread_sp = thread_list.GetThreadAtIndex(i); 6395 uint32_t stack_frame_count = thread_sp->GetStackFrameCount(); 6396 for (uint32_t j = 0; j < stack_frame_count; j++) { 6397 StackFrameSP stack_frame_sp = thread_sp->GetStackFrameAtIndex(j); 6398 Address pc = stack_frame_sp->GetFrameCodeAddress(); 6399 ModuleSP module_sp = pc.GetModule(); 6400 if (module_sp) { 6401 UUID uuid = module_sp->GetUUID(); 6402 if (uuid.IsValid()) { 6403 executing_uuids.insert(uuid.GetAsString()); 6404 modules.AppendIfNeeded(module_sp); 6405 } 6406 } 6407 } 6408 } 6409 size_t modules_count = modules.GetSize(); 6410 6411 struct all_image_infos_header infos; 6412 infos.version = 1; 6413 infos.imgcount = modules_count; 6414 infos.entries_size = sizeof(image_entry); 6415 infos.entries_fileoff = initial_file_offset + sizeof(all_image_infos_header); 6416 infos.unused = 0; 6417 6418 all_image_infos_payload.PutHex32(infos.version); 6419 all_image_infos_payload.PutHex32(infos.imgcount); 6420 all_image_infos_payload.PutHex64(infos.entries_fileoff); 6421 all_image_infos_payload.PutHex32(infos.entries_size); 6422 all_image_infos_payload.PutHex32(infos.unused); 6423 6424 // First create the structures for all of the segment name+vmaddr vectors 6425 // for each module, so we will know the size of them as we add the 6426 // module entries. 6427 std::vector<std::vector<segment_vmaddr>> modules_segment_vmaddrs; 6428 for (size_t i = 0; i < modules_count; i++) { 6429 ModuleSP module = modules.GetModuleAtIndex(i); 6430 6431 SectionList *sections = module->GetSectionList(); 6432 size_t sections_count = sections->GetSize(); 6433 std::vector<segment_vmaddr> segment_vmaddrs; 6434 for (size_t j = 0; j < sections_count; j++) { 6435 SectionSP section = sections->GetSectionAtIndex(j); 6436 if (!section->GetParent().get()) { 6437 addr_t vmaddr = section->GetLoadBaseAddress(&target); 6438 if (vmaddr == LLDB_INVALID_ADDRESS) 6439 continue; 6440 ConstString name = section->GetName(); 6441 segment_vmaddr seg_vmaddr; 6442 strncpy(seg_vmaddr.segname, name.AsCString(), 6443 sizeof(seg_vmaddr.segname)); 6444 seg_vmaddr.vmaddr = vmaddr; 6445 seg_vmaddr.unused = 0; 6446 segment_vmaddrs.push_back(seg_vmaddr); 6447 } 6448 } 6449 modules_segment_vmaddrs.push_back(segment_vmaddrs); 6450 } 6451 6452 offset_t size_of_vmaddr_structs = 0; 6453 for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) { 6454 size_of_vmaddr_structs += 6455 modules_segment_vmaddrs[i].size() * sizeof(segment_vmaddr); 6456 } 6457 6458 offset_t size_of_filepath_cstrings = 0; 6459 for (size_t i = 0; i < modules_count; i++) { 6460 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6461 size_of_filepath_cstrings += module_sp->GetFileSpec().GetPath().size() + 1; 6462 } 6463 6464 // Calculate the file offsets of our "all image infos" payload in the 6465 // corefile. initial_file_offset the original value passed in to this method. 6466 6467 offset_t start_of_entries = 6468 initial_file_offset + sizeof(all_image_infos_header); 6469 offset_t start_of_seg_vmaddrs = 6470 start_of_entries + sizeof(image_entry) * modules_count; 6471 offset_t start_of_filenames = start_of_seg_vmaddrs + size_of_vmaddr_structs; 6472 6473 offset_t final_file_offset = start_of_filenames + size_of_filepath_cstrings; 6474 6475 // Now write the one-per-module 'struct image_entry' into the 6476 // StringStream; keep track of where the struct segment_vmaddr 6477 // entries for each module will end up in the corefile. 6478 6479 offset_t current_string_offset = start_of_filenames; 6480 offset_t current_segaddrs_offset = start_of_seg_vmaddrs; 6481 std::vector<struct image_entry> image_entries; 6482 for (size_t i = 0; i < modules_count; i++) { 6483 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6484 6485 struct image_entry ent; 6486 memcpy(&ent.uuid, module_sp->GetUUID().GetBytes().data(), sizeof(ent.uuid)); 6487 if (modules_segment_vmaddrs[i].size() > 0) { 6488 ent.segment_count = modules_segment_vmaddrs[i].size(); 6489 ent.seg_addrs_offset = current_segaddrs_offset; 6490 } 6491 ent.filepath_offset = current_string_offset; 6492 ObjectFile *objfile = module_sp->GetObjectFile(); 6493 if (objfile) { 6494 Address base_addr(objfile->GetBaseAddress()); 6495 if (base_addr.IsValid()) { 6496 ent.load_address = base_addr.GetLoadAddress(&target); 6497 } 6498 } 6499 6500 all_image_infos_payload.PutHex64(ent.filepath_offset); 6501 all_image_infos_payload.PutRawBytes(ent.uuid, sizeof(ent.uuid)); 6502 all_image_infos_payload.PutHex64(ent.load_address); 6503 all_image_infos_payload.PutHex64(ent.seg_addrs_offset); 6504 all_image_infos_payload.PutHex32(ent.segment_count); 6505 6506 if (executing_uuids.find(module_sp->GetUUID().GetAsString()) != 6507 executing_uuids.end()) 6508 all_image_infos_payload.PutHex32(1); 6509 else 6510 all_image_infos_payload.PutHex32(0); 6511 6512 current_segaddrs_offset += ent.segment_count * sizeof(segment_vmaddr); 6513 current_string_offset += module_sp->GetFileSpec().GetPath().size() + 1; 6514 } 6515 6516 // Now write the struct segment_vmaddr entries into the StringStream. 6517 6518 for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) { 6519 if (modules_segment_vmaddrs[i].size() == 0) 6520 continue; 6521 for (struct segment_vmaddr segvm : modules_segment_vmaddrs[i]) { 6522 all_image_infos_payload.PutRawBytes(segvm.segname, sizeof(segvm.segname)); 6523 all_image_infos_payload.PutHex64(segvm.vmaddr); 6524 all_image_infos_payload.PutHex64(segvm.unused); 6525 } 6526 } 6527 6528 for (size_t i = 0; i < modules_count; i++) { 6529 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6530 std::string filepath = module_sp->GetFileSpec().GetPath(); 6531 all_image_infos_payload.PutRawBytes(filepath.data(), filepath.size() + 1); 6532 } 6533 6534 return final_file_offset; 6535 } 6536 6537 // Temp struct used to combine contiguous memory regions with 6538 // identical permissions. 6539 struct page_object { 6540 addr_t addr; 6541 addr_t size; 6542 uint32_t prot; 6543 }; 6544 6545 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6546 const FileSpec &outfile, 6547 lldb::SaveCoreStyle &core_style, Status &error) { 6548 if (!process_sp) 6549 return false; 6550 6551 // Default on macOS is to create a dirty-memory-only corefile. 6552 if (core_style == SaveCoreStyle::eSaveCoreUnspecified) { 6553 core_style = SaveCoreStyle::eSaveCoreDirtyOnly; 6554 } 6555 6556 Target &target = process_sp->GetTarget(); 6557 const ArchSpec target_arch = target.GetArchitecture(); 6558 const llvm::Triple &target_triple = target_arch.GetTriple(); 6559 if (target_triple.getVendor() == llvm::Triple::Apple && 6560 (target_triple.getOS() == llvm::Triple::MacOSX || 6561 target_triple.getOS() == llvm::Triple::IOS || 6562 target_triple.getOS() == llvm::Triple::WatchOS || 6563 target_triple.getOS() == llvm::Triple::TvOS)) { 6564 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) 6565 // { 6566 bool make_core = false; 6567 switch (target_arch.GetMachine()) { 6568 case llvm::Triple::aarch64: 6569 case llvm::Triple::aarch64_32: 6570 case llvm::Triple::arm: 6571 case llvm::Triple::thumb: 6572 case llvm::Triple::x86: 6573 case llvm::Triple::x86_64: 6574 make_core = true; 6575 break; 6576 default: 6577 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6578 target_triple.str().c_str()); 6579 break; 6580 } 6581 6582 if (make_core) { 6583 std::vector<llvm::MachO::segment_command_64> segment_load_commands; 6584 // uint32_t range_info_idx = 0; 6585 MemoryRegionInfo range_info; 6586 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6587 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6588 const ByteOrder byte_order = target_arch.GetByteOrder(); 6589 std::vector<page_object> pages_to_copy; 6590 6591 if (range_error.Success()) { 6592 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6593 // Calculate correct protections 6594 uint32_t prot = 0; 6595 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6596 prot |= VM_PROT_READ; 6597 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6598 prot |= VM_PROT_WRITE; 6599 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6600 prot |= VM_PROT_EXECUTE; 6601 6602 const addr_t addr = range_info.GetRange().GetRangeBase(); 6603 const addr_t size = range_info.GetRange().GetByteSize(); 6604 6605 if (size == 0) 6606 break; 6607 6608 bool include_this_region = true; 6609 bool dirty_pages_only = false; 6610 if (core_style == SaveCoreStyle::eSaveCoreStackOnly) { 6611 dirty_pages_only = true; 6612 if (range_info.IsStackMemory() != MemoryRegionInfo::eYes) { 6613 include_this_region = false; 6614 } 6615 } 6616 if (core_style == SaveCoreStyle::eSaveCoreDirtyOnly) { 6617 dirty_pages_only = true; 6618 } 6619 6620 if (prot != 0 && include_this_region) { 6621 addr_t pagesize = range_info.GetPageSize(); 6622 const llvm::Optional<std::vector<addr_t>> &dirty_page_list = 6623 range_info.GetDirtyPageList(); 6624 if (dirty_pages_only && dirty_page_list.hasValue()) { 6625 for (addr_t dirtypage : dirty_page_list.getValue()) { 6626 page_object obj; 6627 obj.addr = dirtypage; 6628 obj.size = pagesize; 6629 obj.prot = prot; 6630 pages_to_copy.push_back(obj); 6631 } 6632 } else { 6633 page_object obj; 6634 obj.addr = addr; 6635 obj.size = size; 6636 obj.prot = prot; 6637 pages_to_copy.push_back(obj); 6638 } 6639 } 6640 6641 range_error = process_sp->GetMemoryRegionInfo( 6642 range_info.GetRange().GetRangeEnd(), range_info); 6643 if (range_error.Fail()) 6644 break; 6645 } 6646 6647 // Combine contiguous entries that have the same 6648 // protections so we don't have an excess of 6649 // load commands. 6650 std::vector<page_object> combined_page_objects; 6651 page_object last_obj; 6652 last_obj.addr = LLDB_INVALID_ADDRESS; 6653 last_obj.size = 0; 6654 for (page_object obj : pages_to_copy) { 6655 if (last_obj.addr == LLDB_INVALID_ADDRESS) { 6656 last_obj = obj; 6657 continue; 6658 } 6659 if (last_obj.addr + last_obj.size == obj.addr && 6660 last_obj.prot == obj.prot) { 6661 last_obj.size += obj.size; 6662 continue; 6663 } 6664 combined_page_objects.push_back(last_obj); 6665 last_obj = obj; 6666 } 6667 // Add the last entry we were looking to combine 6668 // on to the array. 6669 if (last_obj.addr != LLDB_INVALID_ADDRESS && last_obj.size != 0) 6670 combined_page_objects.push_back(last_obj); 6671 6672 for (page_object obj : combined_page_objects) { 6673 uint32_t cmd_type = LC_SEGMENT_64; 6674 uint32_t segment_size = sizeof(llvm::MachO::segment_command_64); 6675 if (addr_byte_size == 4) { 6676 cmd_type = LC_SEGMENT; 6677 segment_size = sizeof(llvm::MachO::segment_command); 6678 } 6679 llvm::MachO::segment_command_64 segment = { 6680 cmd_type, // uint32_t cmd; 6681 segment_size, // uint32_t cmdsize; 6682 {0}, // char segname[16]; 6683 obj.addr, // uint64_t vmaddr; // uint32_t for 32-bit 6684 // Mach-O 6685 obj.size, // uint64_t vmsize; // uint32_t for 32-bit 6686 // Mach-O 6687 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6688 obj.size, // uint64_t filesize; // uint32_t for 32-bit 6689 // Mach-O 6690 obj.prot, // uint32_t maxprot; 6691 obj.prot, // uint32_t initprot; 6692 0, // uint32_t nsects; 6693 0}; // uint32_t flags; 6694 segment_load_commands.push_back(segment); 6695 } 6696 6697 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6698 6699 llvm::MachO::mach_header_64 mach_header; 6700 if (addr_byte_size == 8) { 6701 mach_header.magic = MH_MAGIC_64; 6702 } else { 6703 mach_header.magic = MH_MAGIC; 6704 } 6705 mach_header.cputype = target_arch.GetMachOCPUType(); 6706 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6707 mach_header.filetype = MH_CORE; 6708 mach_header.ncmds = segment_load_commands.size(); 6709 mach_header.flags = 0; 6710 mach_header.reserved = 0; 6711 ThreadList &thread_list = process_sp->GetThreadList(); 6712 const uint32_t num_threads = thread_list.GetSize(); 6713 6714 // Make an array of LC_THREAD data items. Each one contains the 6715 // contents of the LC_THREAD load command. The data doesn't contain 6716 // the load command + load command size, we will add the load command 6717 // and load command size as we emit the data. 6718 std::vector<StreamString> LC_THREAD_datas(num_threads); 6719 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6720 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6721 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6722 LC_THREAD_data.SetByteOrder(byte_order); 6723 } 6724 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) { 6725 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6726 if (thread_sp) { 6727 switch (mach_header.cputype) { 6728 case llvm::MachO::CPU_TYPE_ARM64: 6729 case llvm::MachO::CPU_TYPE_ARM64_32: 6730 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6731 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6732 break; 6733 6734 case llvm::MachO::CPU_TYPE_ARM: 6735 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6736 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6737 break; 6738 6739 case llvm::MachO::CPU_TYPE_I386: 6740 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6741 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6742 break; 6743 6744 case llvm::MachO::CPU_TYPE_X86_64: 6745 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6746 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6747 break; 6748 } 6749 } 6750 } 6751 6752 // The size of the load command is the size of the segments... 6753 if (addr_byte_size == 8) { 6754 mach_header.sizeofcmds = segment_load_commands.size() * 6755 sizeof(llvm::MachO::segment_command_64); 6756 } else { 6757 mach_header.sizeofcmds = segment_load_commands.size() * 6758 sizeof(llvm::MachO::segment_command); 6759 } 6760 6761 // and the size of all LC_THREAD load command 6762 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6763 ++mach_header.ncmds; 6764 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6765 } 6766 6767 // Bits will be set to indicate which bits are NOT used in 6768 // addressing in this process or 0 for unknown. 6769 uint64_t address_mask = process_sp->GetCodeAddressMask(); 6770 if (address_mask != 0) { 6771 // LC_NOTE "addrable bits" 6772 mach_header.ncmds++; 6773 mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); 6774 } 6775 6776 // LC_NOTE "all image infos" 6777 mach_header.ncmds++; 6778 mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); 6779 6780 // Write the mach header 6781 buffer.PutHex32(mach_header.magic); 6782 buffer.PutHex32(mach_header.cputype); 6783 buffer.PutHex32(mach_header.cpusubtype); 6784 buffer.PutHex32(mach_header.filetype); 6785 buffer.PutHex32(mach_header.ncmds); 6786 buffer.PutHex32(mach_header.sizeofcmds); 6787 buffer.PutHex32(mach_header.flags); 6788 if (addr_byte_size == 8) { 6789 buffer.PutHex32(mach_header.reserved); 6790 } 6791 6792 // Skip the mach header and all load commands and align to the next 6793 // 0x1000 byte boundary 6794 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6795 6796 file_offset = llvm::alignTo(file_offset, 16); 6797 std::vector<std::unique_ptr<LCNoteEntry>> lc_notes; 6798 6799 // Add "addrable bits" LC_NOTE when an address mask is available 6800 if (address_mask != 0) { 6801 std::unique_ptr<LCNoteEntry> addrable_bits_lcnote_up( 6802 new LCNoteEntry(addr_byte_size, byte_order)); 6803 addrable_bits_lcnote_up->name = "addrable bits"; 6804 addrable_bits_lcnote_up->payload_file_offset = file_offset; 6805 int bits = std::bitset<64>(~address_mask).count(); 6806 addrable_bits_lcnote_up->payload.PutHex32(3); // version 6807 addrable_bits_lcnote_up->payload.PutHex32( 6808 bits); // # of bits used for addressing 6809 addrable_bits_lcnote_up->payload.PutHex64(0); // unused 6810 6811 file_offset += addrable_bits_lcnote_up->payload.GetSize(); 6812 6813 lc_notes.push_back(std::move(addrable_bits_lcnote_up)); 6814 } 6815 6816 // Add "all image infos" LC_NOTE 6817 std::unique_ptr<LCNoteEntry> all_image_infos_lcnote_up( 6818 new LCNoteEntry(addr_byte_size, byte_order)); 6819 all_image_infos_lcnote_up->name = "all image infos"; 6820 all_image_infos_lcnote_up->payload_file_offset = file_offset; 6821 file_offset = CreateAllImageInfosPayload( 6822 process_sp, file_offset, all_image_infos_lcnote_up->payload, 6823 core_style); 6824 lc_notes.push_back(std::move(all_image_infos_lcnote_up)); 6825 6826 // Add LC_NOTE load commands 6827 for (auto &lcnote : lc_notes) { 6828 // Add the LC_NOTE load command to the file. 6829 buffer.PutHex32(LC_NOTE); 6830 buffer.PutHex32(sizeof(llvm::MachO::note_command)); 6831 char namebuf[16]; 6832 memset(namebuf, 0, sizeof(namebuf)); 6833 // this is the uncommon case where strncpy is exactly 6834 // the right one, doesn't need to be nul terminated. 6835 strncpy(namebuf, lcnote->name.c_str(), sizeof(namebuf)); 6836 buffer.PutRawBytes(namebuf, sizeof(namebuf)); 6837 buffer.PutHex64(lcnote->payload_file_offset); 6838 buffer.PutHex64(lcnote->payload.GetSize()); 6839 } 6840 6841 // Align to 4096-byte page boundary for the LC_SEGMENTs. 6842 file_offset = llvm::alignTo(file_offset, 4096); 6843 6844 for (auto &segment : segment_load_commands) { 6845 segment.fileoff = file_offset; 6846 file_offset += segment.filesize; 6847 } 6848 6849 // Write out all of the LC_THREAD load commands 6850 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6851 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6852 buffer.PutHex32(LC_THREAD); 6853 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6854 buffer.Write(LC_THREAD_data.GetString().data(), LC_THREAD_data_size); 6855 } 6856 6857 // Write out all of the segment load commands 6858 for (const auto &segment : segment_load_commands) { 6859 buffer.PutHex32(segment.cmd); 6860 buffer.PutHex32(segment.cmdsize); 6861 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6862 if (addr_byte_size == 8) { 6863 buffer.PutHex64(segment.vmaddr); 6864 buffer.PutHex64(segment.vmsize); 6865 buffer.PutHex64(segment.fileoff); 6866 buffer.PutHex64(segment.filesize); 6867 } else { 6868 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6869 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6870 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6871 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6872 } 6873 buffer.PutHex32(segment.maxprot); 6874 buffer.PutHex32(segment.initprot); 6875 buffer.PutHex32(segment.nsects); 6876 buffer.PutHex32(segment.flags); 6877 } 6878 6879 std::string core_file_path(outfile.GetPath()); 6880 auto core_file = FileSystem::Instance().Open( 6881 outfile, File::eOpenOptionWriteOnly | File::eOpenOptionTruncate | 6882 File::eOpenOptionCanCreate); 6883 if (!core_file) { 6884 error = core_file.takeError(); 6885 } else { 6886 // Read 1 page at a time 6887 uint8_t bytes[0x1000]; 6888 // Write the mach header and load commands out to the core file 6889 size_t bytes_written = buffer.GetString().size(); 6890 error = 6891 core_file.get()->Write(buffer.GetString().data(), bytes_written); 6892 if (error.Success()) { 6893 6894 for (auto &lcnote : lc_notes) { 6895 if (core_file.get()->SeekFromStart(lcnote->payload_file_offset) == 6896 -1) { 6897 error.SetErrorStringWithFormat("Unable to seek to corefile pos " 6898 "to write '%s' LC_NOTE payload", 6899 lcnote->name.c_str()); 6900 return false; 6901 } 6902 bytes_written = lcnote->payload.GetSize(); 6903 error = core_file.get()->Write(lcnote->payload.GetData(), 6904 bytes_written); 6905 if (!error.Success()) 6906 return false; 6907 } 6908 6909 // Now write the file data for all memory segments in the process 6910 for (const auto &segment : segment_load_commands) { 6911 if (core_file.get()->SeekFromStart(segment.fileoff) == -1) { 6912 error.SetErrorStringWithFormat( 6913 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6914 segment.fileoff, core_file_path.c_str()); 6915 break; 6916 } 6917 6918 target.GetDebugger().GetAsyncOutputStream()->Printf( 6919 "Saving %" PRId64 6920 " bytes of data for memory region at 0x%" PRIx64 "\n", 6921 segment.vmsize, segment.vmaddr); 6922 addr_t bytes_left = segment.vmsize; 6923 addr_t addr = segment.vmaddr; 6924 Status memory_read_error; 6925 while (bytes_left > 0 && error.Success()) { 6926 const size_t bytes_to_read = 6927 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6928 6929 // In a savecore setting, we don't really care about caching, 6930 // as the data is dumped and very likely never read again, 6931 // so we call ReadMemoryFromInferior to bypass it. 6932 const size_t bytes_read = process_sp->ReadMemoryFromInferior( 6933 addr, bytes, bytes_to_read, memory_read_error); 6934 6935 if (bytes_read == bytes_to_read) { 6936 size_t bytes_written = bytes_read; 6937 error = core_file.get()->Write(bytes, bytes_written); 6938 bytes_left -= bytes_read; 6939 addr += bytes_read; 6940 } else { 6941 // Some pages within regions are not readable, those should 6942 // be zero filled 6943 memset(bytes, 0, bytes_to_read); 6944 size_t bytes_written = bytes_to_read; 6945 error = core_file.get()->Write(bytes, bytes_written); 6946 bytes_left -= bytes_to_read; 6947 addr += bytes_to_read; 6948 } 6949 } 6950 } 6951 } 6952 } 6953 } else { 6954 error.SetErrorString( 6955 "process doesn't support getting memory region info"); 6956 } 6957 } 6958 return true; // This is the right plug to handle saving core files for 6959 // this process 6960 } 6961 return false; 6962 } 6963 6964 ObjectFileMachO::MachOCorefileAllImageInfos 6965 ObjectFileMachO::GetCorefileAllImageInfos() { 6966 MachOCorefileAllImageInfos image_infos; 6967 6968 // Look for an "all image infos" LC_NOTE. 6969 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 6970 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6971 const uint32_t cmd_offset = offset; 6972 llvm::MachO::load_command lc; 6973 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6974 break; 6975 if (lc.cmd == LC_NOTE) { 6976 char data_owner[17]; 6977 m_data.CopyData(offset, 16, data_owner); 6978 data_owner[16] = '\0'; 6979 offset += 16; 6980 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 6981 offset += 4; /* size unused */ 6982 6983 if (strcmp("all image infos", data_owner) == 0) { 6984 offset = fileoff; 6985 // Read the struct all_image_infos_header. 6986 uint32_t version = m_data.GetU32(&offset); 6987 if (version != 1) { 6988 return image_infos; 6989 } 6990 uint32_t imgcount = m_data.GetU32(&offset); 6991 uint64_t entries_fileoff = m_data.GetU64(&offset); 6992 offset += 4; // uint32_t entries_size; 6993 offset += 4; // uint32_t unused; 6994 6995 offset = entries_fileoff; 6996 for (uint32_t i = 0; i < imgcount; i++) { 6997 // Read the struct image_entry. 6998 offset_t filepath_offset = m_data.GetU64(&offset); 6999 uuid_t uuid; 7000 memcpy(&uuid, m_data.GetData(&offset, sizeof(uuid_t)), 7001 sizeof(uuid_t)); 7002 uint64_t load_address = m_data.GetU64(&offset); 7003 offset_t seg_addrs_offset = m_data.GetU64(&offset); 7004 uint32_t segment_count = m_data.GetU32(&offset); 7005 uint32_t currently_executing = m_data.GetU32(&offset); 7006 7007 MachOCorefileImageEntry image_entry; 7008 image_entry.filename = (const char *)m_data.GetCStr(&filepath_offset); 7009 image_entry.uuid = UUID::fromData(uuid, sizeof(uuid_t)); 7010 image_entry.load_address = load_address; 7011 image_entry.currently_executing = currently_executing; 7012 7013 offset_t seg_vmaddrs_offset = seg_addrs_offset; 7014 for (uint32_t j = 0; j < segment_count; j++) { 7015 char segname[17]; 7016 m_data.CopyData(seg_vmaddrs_offset, 16, segname); 7017 segname[16] = '\0'; 7018 seg_vmaddrs_offset += 16; 7019 uint64_t vmaddr = m_data.GetU64(&seg_vmaddrs_offset); 7020 seg_vmaddrs_offset += 8; /* unused */ 7021 7022 std::tuple<ConstString, addr_t> new_seg{ConstString(segname), 7023 vmaddr}; 7024 image_entry.segment_load_addresses.push_back(new_seg); 7025 } 7026 image_infos.all_image_infos.push_back(image_entry); 7027 } 7028 } 7029 } 7030 offset = cmd_offset + lc.cmdsize; 7031 } 7032 7033 return image_infos; 7034 } 7035 7036 bool ObjectFileMachO::LoadCoreFileImages(lldb_private::Process &process) { 7037 MachOCorefileAllImageInfos image_infos = GetCorefileAllImageInfos(); 7038 bool added_images = false; 7039 if (image_infos.IsValid()) { 7040 for (const MachOCorefileImageEntry &image : image_infos.all_image_infos) { 7041 ModuleSpec module_spec; 7042 module_spec.GetUUID() = image.uuid; 7043 module_spec.GetFileSpec() = FileSpec(image.filename.c_str()); 7044 if (image.currently_executing) { 7045 Symbols::DownloadObjectAndSymbolFile(module_spec, true); 7046 if (FileSystem::Instance().Exists(module_spec.GetFileSpec())) { 7047 process.GetTarget().GetOrCreateModule(module_spec, false); 7048 } 7049 } 7050 Status error; 7051 ModuleSP module_sp = 7052 process.GetTarget().GetOrCreateModule(module_spec, false, &error); 7053 if (!module_sp.get() || !module_sp->GetObjectFile()) { 7054 if (image.load_address != LLDB_INVALID_ADDRESS) { 7055 module_sp = process.ReadModuleFromMemory(module_spec.GetFileSpec(), 7056 image.load_address); 7057 } 7058 } 7059 if (module_sp.get()) { 7060 added_images = true; 7061 for (auto name_vmaddr_tuple : image.segment_load_addresses) { 7062 SectionList *sectlist = module_sp->GetObjectFile()->GetSectionList(); 7063 if (sectlist) { 7064 SectionSP sect_sp = 7065 sectlist->FindSectionByName(std::get<0>(name_vmaddr_tuple)); 7066 if (sect_sp) { 7067 process.GetTarget().SetSectionLoadAddress( 7068 sect_sp, std::get<1>(name_vmaddr_tuple)); 7069 } 7070 } 7071 } 7072 } 7073 } 7074 } 7075 return added_images; 7076 } 7077