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