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/ScopeExit.h" 10 #include "llvm/ADT/StringRef.h" 11 12 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 13 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 14 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 15 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 16 #include "lldb/Core/Debugger.h" 17 #include "lldb/Core/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/Host/Host.h" 23 #include "lldb/Symbol/DWARFCallFrameInfo.h" 24 #include "lldb/Symbol/ObjectFile.h" 25 #include "lldb/Target/DynamicLoader.h" 26 #include "lldb/Target/MemoryRegionInfo.h" 27 #include "lldb/Target/Platform.h" 28 #include "lldb/Target/Process.h" 29 #include "lldb/Target/SectionLoadList.h" 30 #include "lldb/Target/Target.h" 31 #include "lldb/Target/Thread.h" 32 #include "lldb/Target/ThreadList.h" 33 #include "lldb/Utility/ArchSpec.h" 34 #include "lldb/Utility/DataBuffer.h" 35 #include "lldb/Utility/FileSpec.h" 36 #include "lldb/Utility/FileSpecList.h" 37 #include "lldb/Utility/LLDBLog.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 #include <mach/mach_init.h> 59 #include <mach/vm_map.h> 60 #include <lldb/Host/SafeMachO.h> 61 #endif 62 63 #ifndef __APPLE__ 64 #include "lldb/Utility/AppleUuidCompatibility.h" 65 #else 66 #include <uuid/uuid.h> 67 #endif 68 69 #include <bitset> 70 #include <memory> 71 #include <optional> 72 73 // Unfortunately the signpost header pulls in the system MachO header, too. 74 #ifdef CPU_TYPE_ARM 75 #undef CPU_TYPE_ARM 76 #endif 77 #ifdef CPU_TYPE_ARM64 78 #undef CPU_TYPE_ARM64 79 #endif 80 #ifdef CPU_TYPE_ARM64_32 81 #undef CPU_TYPE_ARM64_32 82 #endif 83 #ifdef CPU_TYPE_I386 84 #undef CPU_TYPE_I386 85 #endif 86 #ifdef CPU_TYPE_X86_64 87 #undef CPU_TYPE_X86_64 88 #endif 89 #ifdef MH_DYLINKER 90 #undef MH_DYLINKER 91 #endif 92 #ifdef MH_OBJECT 93 #undef MH_OBJECT 94 #endif 95 #ifdef LC_VERSION_MIN_MACOSX 96 #undef LC_VERSION_MIN_MACOSX 97 #endif 98 #ifdef LC_VERSION_MIN_IPHONEOS 99 #undef LC_VERSION_MIN_IPHONEOS 100 #endif 101 #ifdef LC_VERSION_MIN_TVOS 102 #undef LC_VERSION_MIN_TVOS 103 #endif 104 #ifdef LC_VERSION_MIN_WATCHOS 105 #undef LC_VERSION_MIN_WATCHOS 106 #endif 107 #ifdef LC_BUILD_VERSION 108 #undef LC_BUILD_VERSION 109 #endif 110 #ifdef PLATFORM_MACOS 111 #undef PLATFORM_MACOS 112 #endif 113 #ifdef PLATFORM_MACCATALYST 114 #undef PLATFORM_MACCATALYST 115 #endif 116 #ifdef PLATFORM_IOS 117 #undef PLATFORM_IOS 118 #endif 119 #ifdef PLATFORM_IOSSIMULATOR 120 #undef PLATFORM_IOSSIMULATOR 121 #endif 122 #ifdef PLATFORM_TVOS 123 #undef PLATFORM_TVOS 124 #endif 125 #ifdef PLATFORM_TVOSSIMULATOR 126 #undef PLATFORM_TVOSSIMULATOR 127 #endif 128 #ifdef PLATFORM_WATCHOS 129 #undef PLATFORM_WATCHOS 130 #endif 131 #ifdef PLATFORM_WATCHOSSIMULATOR 132 #undef PLATFORM_WATCHOSSIMULATOR 133 #endif 134 135 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 136 using namespace lldb; 137 using namespace lldb_private; 138 using namespace llvm::MachO; 139 140 static constexpr llvm::StringLiteral g_loader_path = "@loader_path"; 141 static constexpr llvm::StringLiteral g_executable_path = "@executable_path"; 142 143 LLDB_PLUGIN_DEFINE(ObjectFileMachO) 144 145 static void PrintRegisterValue(RegisterContext *reg_ctx, const char *name, 146 const char *alt_name, size_t reg_byte_size, 147 Stream &data) { 148 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 149 if (reg_info == nullptr) 150 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 151 if (reg_info) { 152 lldb_private::RegisterValue reg_value; 153 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 154 if (reg_info->byte_size >= reg_byte_size) 155 data.Write(reg_value.GetBytes(), reg_byte_size); 156 else { 157 data.Write(reg_value.GetBytes(), reg_info->byte_size); 158 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; ++i) 159 data.PutChar(0); 160 } 161 return; 162 } 163 } 164 // Just write zeros if all else fails 165 for (size_t i = 0; i < reg_byte_size; ++i) 166 data.PutChar(0); 167 } 168 169 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 170 public: 171 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 172 const DataExtractor &data) 173 : RegisterContextDarwin_x86_64(thread, 0) { 174 SetRegisterDataFrom_LC_THREAD(data); 175 } 176 177 void InvalidateAllRegisters() override { 178 // Do nothing... registers are always valid... 179 } 180 181 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 182 lldb::offset_t offset = 0; 183 SetError(GPRRegSet, Read, -1); 184 SetError(FPURegSet, Read, -1); 185 SetError(EXCRegSet, Read, -1); 186 bool done = false; 187 188 while (!done) { 189 int flavor = data.GetU32(&offset); 190 if (flavor == 0) 191 done = true; 192 else { 193 uint32_t i; 194 uint32_t count = data.GetU32(&offset); 195 switch (flavor) { 196 case GPRRegSet: 197 for (i = 0; i < count; ++i) 198 (&gpr.rax)[i] = data.GetU64(&offset); 199 SetError(GPRRegSet, Read, 0); 200 done = true; 201 202 break; 203 case FPURegSet: 204 // TODO: fill in FPU regs.... 205 // SetError (FPURegSet, Read, -1); 206 done = true; 207 208 break; 209 case EXCRegSet: 210 exc.trapno = data.GetU32(&offset); 211 exc.err = data.GetU32(&offset); 212 exc.faultvaddr = data.GetU64(&offset); 213 SetError(EXCRegSet, Read, 0); 214 done = true; 215 break; 216 case 7: 217 case 8: 218 case 9: 219 // fancy flavors that encapsulate of the above flavors... 220 break; 221 222 default: 223 done = true; 224 break; 225 } 226 } 227 } 228 } 229 230 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 231 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 232 if (reg_ctx_sp) { 233 RegisterContext *reg_ctx = reg_ctx_sp.get(); 234 235 data.PutHex32(GPRRegSet); // Flavor 236 data.PutHex32(GPRWordCount); 237 PrintRegisterValue(reg_ctx, "rax", nullptr, 8, data); 238 PrintRegisterValue(reg_ctx, "rbx", nullptr, 8, data); 239 PrintRegisterValue(reg_ctx, "rcx", nullptr, 8, data); 240 PrintRegisterValue(reg_ctx, "rdx", nullptr, 8, data); 241 PrintRegisterValue(reg_ctx, "rdi", nullptr, 8, data); 242 PrintRegisterValue(reg_ctx, "rsi", nullptr, 8, data); 243 PrintRegisterValue(reg_ctx, "rbp", nullptr, 8, data); 244 PrintRegisterValue(reg_ctx, "rsp", nullptr, 8, data); 245 PrintRegisterValue(reg_ctx, "r8", nullptr, 8, data); 246 PrintRegisterValue(reg_ctx, "r9", nullptr, 8, data); 247 PrintRegisterValue(reg_ctx, "r10", nullptr, 8, data); 248 PrintRegisterValue(reg_ctx, "r11", nullptr, 8, data); 249 PrintRegisterValue(reg_ctx, "r12", nullptr, 8, data); 250 PrintRegisterValue(reg_ctx, "r13", nullptr, 8, data); 251 PrintRegisterValue(reg_ctx, "r14", nullptr, 8, data); 252 PrintRegisterValue(reg_ctx, "r15", nullptr, 8, data); 253 PrintRegisterValue(reg_ctx, "rip", nullptr, 8, data); 254 PrintRegisterValue(reg_ctx, "rflags", nullptr, 8, data); 255 PrintRegisterValue(reg_ctx, "cs", nullptr, 8, data); 256 PrintRegisterValue(reg_ctx, "fs", nullptr, 8, data); 257 PrintRegisterValue(reg_ctx, "gs", nullptr, 8, data); 258 259 // // Write out the FPU registers 260 // const size_t fpu_byte_size = sizeof(FPU); 261 // size_t bytes_written = 0; 262 // data.PutHex32 (FPURegSet); 263 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 264 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 265 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 266 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 267 // data); // uint16_t fcw; // "fctrl" 268 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 269 // data); // uint16_t fsw; // "fstat" 270 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 271 // data); // uint8_t ftw; // "ftag" 272 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 273 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 274 // data); // uint16_t fop; // "fop" 275 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 276 // data); // uint32_t ip; // "fioff" 277 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 278 // data); // uint16_t cs; // "fiseg" 279 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 280 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 281 // data); // uint32_t dp; // "fooff" 282 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 283 // data); // uint16_t ds; // "foseg" 284 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 285 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 286 // data); // uint32_t mxcsr; 287 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 288 // 4, data);// uint32_t mxcsrmask; 289 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 290 // sizeof(MMSReg), data); 291 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 292 // sizeof(MMSReg), data); 293 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 294 // sizeof(MMSReg), data); 295 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 296 // sizeof(MMSReg), data); 297 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 298 // sizeof(MMSReg), data); 299 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 300 // sizeof(MMSReg), data); 301 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 302 // sizeof(MMSReg), data); 303 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 304 // sizeof(MMSReg), data); 305 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 306 // sizeof(XMMReg), data); 307 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 308 // sizeof(XMMReg), data); 309 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 310 // sizeof(XMMReg), data); 311 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 312 // sizeof(XMMReg), data); 313 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 314 // sizeof(XMMReg), data); 315 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 316 // sizeof(XMMReg), data); 317 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 318 // sizeof(XMMReg), data); 319 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 320 // sizeof(XMMReg), data); 321 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 322 // sizeof(XMMReg), data); 323 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 324 // sizeof(XMMReg), data); 325 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 326 // sizeof(XMMReg), data); 327 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 328 // sizeof(XMMReg), data); 329 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 330 // sizeof(XMMReg), data); 331 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 332 // sizeof(XMMReg), data); 333 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 334 // sizeof(XMMReg), data); 335 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 336 // sizeof(XMMReg), data); 337 // 338 // // Fill rest with zeros 339 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 340 // i) 341 // data.PutChar(0); 342 343 // Write out the EXC registers 344 data.PutHex32(EXCRegSet); 345 data.PutHex32(EXCWordCount); 346 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 347 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 348 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 8, data); 349 return true; 350 } 351 return false; 352 } 353 354 protected: 355 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 356 357 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 358 359 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 360 361 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 362 return 0; 363 } 364 365 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 366 return 0; 367 } 368 369 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 370 return 0; 371 } 372 }; 373 374 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 375 public: 376 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 377 const DataExtractor &data) 378 : RegisterContextDarwin_i386(thread, 0) { 379 SetRegisterDataFrom_LC_THREAD(data); 380 } 381 382 void InvalidateAllRegisters() override { 383 // Do nothing... registers are always valid... 384 } 385 386 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 387 lldb::offset_t offset = 0; 388 SetError(GPRRegSet, Read, -1); 389 SetError(FPURegSet, Read, -1); 390 SetError(EXCRegSet, Read, -1); 391 bool done = false; 392 393 while (!done) { 394 int flavor = data.GetU32(&offset); 395 if (flavor == 0) 396 done = true; 397 else { 398 uint32_t i; 399 uint32_t count = data.GetU32(&offset); 400 switch (flavor) { 401 case GPRRegSet: 402 for (i = 0; i < count; ++i) 403 (&gpr.eax)[i] = data.GetU32(&offset); 404 SetError(GPRRegSet, Read, 0); 405 done = true; 406 407 break; 408 case FPURegSet: 409 // TODO: fill in FPU regs.... 410 // SetError (FPURegSet, Read, -1); 411 done = true; 412 413 break; 414 case EXCRegSet: 415 exc.trapno = data.GetU32(&offset); 416 exc.err = data.GetU32(&offset); 417 exc.faultvaddr = data.GetU32(&offset); 418 SetError(EXCRegSet, Read, 0); 419 done = true; 420 break; 421 case 7: 422 case 8: 423 case 9: 424 // fancy flavors that encapsulate of the above flavors... 425 break; 426 427 default: 428 done = true; 429 break; 430 } 431 } 432 } 433 } 434 435 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 436 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 437 if (reg_ctx_sp) { 438 RegisterContext *reg_ctx = reg_ctx_sp.get(); 439 440 data.PutHex32(GPRRegSet); // Flavor 441 data.PutHex32(GPRWordCount); 442 PrintRegisterValue(reg_ctx, "eax", nullptr, 4, data); 443 PrintRegisterValue(reg_ctx, "ebx", nullptr, 4, data); 444 PrintRegisterValue(reg_ctx, "ecx", nullptr, 4, data); 445 PrintRegisterValue(reg_ctx, "edx", nullptr, 4, data); 446 PrintRegisterValue(reg_ctx, "edi", nullptr, 4, data); 447 PrintRegisterValue(reg_ctx, "esi", nullptr, 4, data); 448 PrintRegisterValue(reg_ctx, "ebp", nullptr, 4, data); 449 PrintRegisterValue(reg_ctx, "esp", nullptr, 4, data); 450 PrintRegisterValue(reg_ctx, "ss", nullptr, 4, data); 451 PrintRegisterValue(reg_ctx, "eflags", nullptr, 4, data); 452 PrintRegisterValue(reg_ctx, "eip", nullptr, 4, data); 453 PrintRegisterValue(reg_ctx, "cs", nullptr, 4, data); 454 PrintRegisterValue(reg_ctx, "ds", nullptr, 4, data); 455 PrintRegisterValue(reg_ctx, "es", nullptr, 4, data); 456 PrintRegisterValue(reg_ctx, "fs", nullptr, 4, data); 457 PrintRegisterValue(reg_ctx, "gs", nullptr, 4, data); 458 459 // Write out the EXC registers 460 data.PutHex32(EXCRegSet); 461 data.PutHex32(EXCWordCount); 462 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 463 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 464 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 4, data); 465 return true; 466 } 467 return false; 468 } 469 470 protected: 471 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 472 473 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 474 475 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 476 477 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 478 return 0; 479 } 480 481 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 482 return 0; 483 } 484 485 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 486 return 0; 487 } 488 }; 489 490 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 491 public: 492 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 493 const DataExtractor &data) 494 : RegisterContextDarwin_arm(thread, 0) { 495 SetRegisterDataFrom_LC_THREAD(data); 496 } 497 498 void InvalidateAllRegisters() override { 499 // Do nothing... registers are always valid... 500 } 501 502 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 503 lldb::offset_t offset = 0; 504 SetError(GPRRegSet, Read, -1); 505 SetError(FPURegSet, Read, -1); 506 SetError(EXCRegSet, Read, -1); 507 bool done = false; 508 509 while (!done) { 510 int flavor = data.GetU32(&offset); 511 uint32_t count = data.GetU32(&offset); 512 lldb::offset_t next_thread_state = offset + (count * 4); 513 switch (flavor) { 514 case GPRAltRegSet: 515 case GPRRegSet: { 516 // r0-r15, plus CPSR 517 uint32_t gpr_buf_count = (sizeof(gpr.r) / sizeof(gpr.r[0])) + 1; 518 if (count == gpr_buf_count) { 519 for (uint32_t i = 0; i < (count - 1); ++i) { 520 gpr.r[i] = data.GetU32(&offset); 521 } 522 gpr.cpsr = data.GetU32(&offset); 523 524 SetError(GPRRegSet, Read, 0); 525 } 526 } 527 offset = next_thread_state; 528 break; 529 530 case FPURegSet: { 531 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats; 532 const int fpu_reg_buf_size = sizeof(fpu.floats); 533 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 534 fpu_reg_buf) == fpu_reg_buf_size) { 535 offset += fpu_reg_buf_size; 536 fpu.fpscr = data.GetU32(&offset); 537 SetError(FPURegSet, Read, 0); 538 } else { 539 done = true; 540 } 541 } 542 offset = next_thread_state; 543 break; 544 545 case EXCRegSet: 546 if (count == 3) { 547 exc.exception = data.GetU32(&offset); 548 exc.fsr = data.GetU32(&offset); 549 exc.far = data.GetU32(&offset); 550 SetError(EXCRegSet, Read, 0); 551 } 552 done = true; 553 offset = next_thread_state; 554 break; 555 556 // Unknown register set flavor, stop trying to parse. 557 default: 558 done = true; 559 } 560 } 561 } 562 563 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 564 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 565 if (reg_ctx_sp) { 566 RegisterContext *reg_ctx = reg_ctx_sp.get(); 567 568 data.PutHex32(GPRRegSet); // Flavor 569 data.PutHex32(GPRWordCount); 570 PrintRegisterValue(reg_ctx, "r0", nullptr, 4, data); 571 PrintRegisterValue(reg_ctx, "r1", nullptr, 4, data); 572 PrintRegisterValue(reg_ctx, "r2", nullptr, 4, data); 573 PrintRegisterValue(reg_ctx, "r3", nullptr, 4, data); 574 PrintRegisterValue(reg_ctx, "r4", nullptr, 4, data); 575 PrintRegisterValue(reg_ctx, "r5", nullptr, 4, data); 576 PrintRegisterValue(reg_ctx, "r6", nullptr, 4, data); 577 PrintRegisterValue(reg_ctx, "r7", nullptr, 4, data); 578 PrintRegisterValue(reg_ctx, "r8", nullptr, 4, data); 579 PrintRegisterValue(reg_ctx, "r9", nullptr, 4, data); 580 PrintRegisterValue(reg_ctx, "r10", nullptr, 4, data); 581 PrintRegisterValue(reg_ctx, "r11", nullptr, 4, data); 582 PrintRegisterValue(reg_ctx, "r12", nullptr, 4, data); 583 PrintRegisterValue(reg_ctx, "sp", nullptr, 4, data); 584 PrintRegisterValue(reg_ctx, "lr", nullptr, 4, data); 585 PrintRegisterValue(reg_ctx, "pc", nullptr, 4, data); 586 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 587 588 // Write out the EXC registers 589 // data.PutHex32 (EXCRegSet); 590 // data.PutHex32 (EXCWordCount); 591 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 592 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 593 // WriteRegister (reg_ctx, "far", NULL, 4, data); 594 return true; 595 } 596 return false; 597 } 598 599 protected: 600 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 601 602 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 603 604 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 605 606 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 607 608 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 609 return 0; 610 } 611 612 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 613 return 0; 614 } 615 616 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 617 return 0; 618 } 619 620 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 621 return -1; 622 } 623 }; 624 625 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 626 public: 627 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 628 const DataExtractor &data) 629 : RegisterContextDarwin_arm64(thread, 0) { 630 SetRegisterDataFrom_LC_THREAD(data); 631 } 632 633 void InvalidateAllRegisters() override { 634 // Do nothing... registers are always valid... 635 } 636 637 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 638 lldb::offset_t offset = 0; 639 SetError(GPRRegSet, Read, -1); 640 SetError(FPURegSet, Read, -1); 641 SetError(EXCRegSet, Read, -1); 642 bool done = false; 643 while (!done) { 644 int flavor = data.GetU32(&offset); 645 uint32_t count = data.GetU32(&offset); 646 lldb::offset_t next_thread_state = offset + (count * 4); 647 switch (flavor) { 648 case GPRRegSet: 649 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 650 // 32-bit register) 651 if (count >= (33 * 2) + 1) { 652 for (uint32_t i = 0; i < 29; ++i) 653 gpr.x[i] = data.GetU64(&offset); 654 gpr.fp = data.GetU64(&offset); 655 gpr.lr = data.GetU64(&offset); 656 gpr.sp = data.GetU64(&offset); 657 gpr.pc = data.GetU64(&offset); 658 gpr.cpsr = data.GetU32(&offset); 659 SetError(GPRRegSet, Read, 0); 660 } 661 offset = next_thread_state; 662 break; 663 case FPURegSet: { 664 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 665 const int fpu_reg_buf_size = sizeof(fpu); 666 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 667 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 668 fpu_reg_buf) == fpu_reg_buf_size) { 669 SetError(FPURegSet, Read, 0); 670 } else { 671 done = true; 672 } 673 } 674 offset = next_thread_state; 675 break; 676 case EXCRegSet: 677 if (count == 4) { 678 exc.far = data.GetU64(&offset); 679 exc.esr = data.GetU32(&offset); 680 exc.exception = data.GetU32(&offset); 681 SetError(EXCRegSet, Read, 0); 682 } 683 offset = next_thread_state; 684 break; 685 default: 686 done = true; 687 break; 688 } 689 } 690 } 691 692 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 693 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 694 if (reg_ctx_sp) { 695 RegisterContext *reg_ctx = reg_ctx_sp.get(); 696 697 data.PutHex32(GPRRegSet); // Flavor 698 data.PutHex32(GPRWordCount); 699 PrintRegisterValue(reg_ctx, "x0", nullptr, 8, data); 700 PrintRegisterValue(reg_ctx, "x1", nullptr, 8, data); 701 PrintRegisterValue(reg_ctx, "x2", nullptr, 8, data); 702 PrintRegisterValue(reg_ctx, "x3", nullptr, 8, data); 703 PrintRegisterValue(reg_ctx, "x4", nullptr, 8, data); 704 PrintRegisterValue(reg_ctx, "x5", nullptr, 8, data); 705 PrintRegisterValue(reg_ctx, "x6", nullptr, 8, data); 706 PrintRegisterValue(reg_ctx, "x7", nullptr, 8, data); 707 PrintRegisterValue(reg_ctx, "x8", nullptr, 8, data); 708 PrintRegisterValue(reg_ctx, "x9", nullptr, 8, data); 709 PrintRegisterValue(reg_ctx, "x10", nullptr, 8, data); 710 PrintRegisterValue(reg_ctx, "x11", nullptr, 8, data); 711 PrintRegisterValue(reg_ctx, "x12", nullptr, 8, data); 712 PrintRegisterValue(reg_ctx, "x13", nullptr, 8, data); 713 PrintRegisterValue(reg_ctx, "x14", nullptr, 8, data); 714 PrintRegisterValue(reg_ctx, "x15", nullptr, 8, data); 715 PrintRegisterValue(reg_ctx, "x16", nullptr, 8, data); 716 PrintRegisterValue(reg_ctx, "x17", nullptr, 8, data); 717 PrintRegisterValue(reg_ctx, "x18", nullptr, 8, data); 718 PrintRegisterValue(reg_ctx, "x19", nullptr, 8, data); 719 PrintRegisterValue(reg_ctx, "x20", nullptr, 8, data); 720 PrintRegisterValue(reg_ctx, "x21", nullptr, 8, data); 721 PrintRegisterValue(reg_ctx, "x22", nullptr, 8, data); 722 PrintRegisterValue(reg_ctx, "x23", nullptr, 8, data); 723 PrintRegisterValue(reg_ctx, "x24", nullptr, 8, data); 724 PrintRegisterValue(reg_ctx, "x25", nullptr, 8, data); 725 PrintRegisterValue(reg_ctx, "x26", nullptr, 8, data); 726 PrintRegisterValue(reg_ctx, "x27", nullptr, 8, data); 727 PrintRegisterValue(reg_ctx, "x28", nullptr, 8, data); 728 PrintRegisterValue(reg_ctx, "fp", nullptr, 8, data); 729 PrintRegisterValue(reg_ctx, "lr", nullptr, 8, data); 730 PrintRegisterValue(reg_ctx, "sp", nullptr, 8, data); 731 PrintRegisterValue(reg_ctx, "pc", nullptr, 8, data); 732 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 733 data.PutHex32(0); // uint32_t pad at the end 734 735 // Write out the EXC registers 736 data.PutHex32(EXCRegSet); 737 data.PutHex32(EXCWordCount); 738 PrintRegisterValue(reg_ctx, "far", nullptr, 8, data); 739 PrintRegisterValue(reg_ctx, "esr", nullptr, 4, data); 740 PrintRegisterValue(reg_ctx, "exception", nullptr, 4, data); 741 return true; 742 } 743 return false; 744 } 745 746 protected: 747 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 748 749 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 750 751 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 752 753 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 754 755 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 756 return 0; 757 } 758 759 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 760 return 0; 761 } 762 763 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 764 return 0; 765 } 766 767 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 768 return -1; 769 } 770 }; 771 772 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 773 switch (magic) { 774 case MH_MAGIC: 775 case MH_CIGAM: 776 return sizeof(struct llvm::MachO::mach_header); 777 778 case MH_MAGIC_64: 779 case MH_CIGAM_64: 780 return sizeof(struct llvm::MachO::mach_header_64); 781 break; 782 783 default: 784 break; 785 } 786 return 0; 787 } 788 789 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 790 791 char ObjectFileMachO::ID; 792 793 void ObjectFileMachO::Initialize() { 794 PluginManager::RegisterPlugin( 795 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 796 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 797 } 798 799 void ObjectFileMachO::Terminate() { 800 PluginManager::UnregisterPlugin(CreateInstance); 801 } 802 803 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 804 DataBufferSP data_sp, 805 lldb::offset_t data_offset, 806 const FileSpec *file, 807 lldb::offset_t file_offset, 808 lldb::offset_t length) { 809 if (!data_sp) { 810 data_sp = MapFileData(*file, length, file_offset); 811 if (!data_sp) 812 return nullptr; 813 data_offset = 0; 814 } 815 816 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 817 return nullptr; 818 819 // Update the data to contain the entire file if it doesn't already 820 if (data_sp->GetByteSize() < length) { 821 data_sp = MapFileData(*file, length, file_offset); 822 if (!data_sp) 823 return nullptr; 824 data_offset = 0; 825 } 826 auto objfile_up = std::make_unique<ObjectFileMachO>( 827 module_sp, data_sp, data_offset, file, file_offset, length); 828 if (!objfile_up || !objfile_up->ParseHeader()) 829 return nullptr; 830 831 return objfile_up.release(); 832 } 833 834 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 835 const lldb::ModuleSP &module_sp, WritableDataBufferSP data_sp, 836 const ProcessSP &process_sp, lldb::addr_t header_addr) { 837 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 838 std::unique_ptr<ObjectFile> objfile_up( 839 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 840 if (objfile_up.get() && objfile_up->ParseHeader()) 841 return objfile_up.release(); 842 } 843 return nullptr; 844 } 845 846 size_t ObjectFileMachO::GetModuleSpecifications( 847 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 848 lldb::offset_t data_offset, lldb::offset_t file_offset, 849 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 850 const size_t initial_count = specs.GetSize(); 851 852 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 853 DataExtractor data; 854 data.SetData(data_sp); 855 llvm::MachO::mach_header header; 856 if (ParseHeader(data, &data_offset, header)) { 857 size_t header_and_load_cmds = 858 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 859 if (header_and_load_cmds >= data_sp->GetByteSize()) { 860 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 861 data.SetData(data_sp); 862 data_offset = MachHeaderSizeFromMagic(header.magic); 863 } 864 if (data_sp) { 865 ModuleSpec base_spec; 866 base_spec.GetFileSpec() = file; 867 base_spec.SetObjectOffset(file_offset); 868 base_spec.SetObjectSize(length); 869 GetAllArchSpecs(header, data, data_offset, base_spec, specs); 870 } 871 } 872 } 873 return specs.GetSize() - initial_count; 874 } 875 876 ConstString ObjectFileMachO::GetSegmentNameTEXT() { 877 static ConstString g_segment_name_TEXT("__TEXT"); 878 return g_segment_name_TEXT; 879 } 880 881 ConstString ObjectFileMachO::GetSegmentNameDATA() { 882 static ConstString g_segment_name_DATA("__DATA"); 883 return g_segment_name_DATA; 884 } 885 886 ConstString ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 887 static ConstString g_segment_name("__DATA_DIRTY"); 888 return g_segment_name; 889 } 890 891 ConstString ObjectFileMachO::GetSegmentNameDATA_CONST() { 892 static ConstString g_segment_name("__DATA_CONST"); 893 return g_segment_name; 894 } 895 896 ConstString ObjectFileMachO::GetSegmentNameOBJC() { 897 static ConstString g_segment_name_OBJC("__OBJC"); 898 return g_segment_name_OBJC; 899 } 900 901 ConstString ObjectFileMachO::GetSegmentNameLINKEDIT() { 902 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 903 return g_section_name_LINKEDIT; 904 } 905 906 ConstString ObjectFileMachO::GetSegmentNameDWARF() { 907 static ConstString g_section_name("__DWARF"); 908 return g_section_name; 909 } 910 911 ConstString ObjectFileMachO::GetSegmentNameLLVM_COV() { 912 static ConstString g_section_name("__LLVM_COV"); 913 return g_section_name; 914 } 915 916 ConstString ObjectFileMachO::GetSectionNameEHFrame() { 917 static ConstString g_section_name_eh_frame("__eh_frame"); 918 return g_section_name_eh_frame; 919 } 920 921 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP data_sp, 922 lldb::addr_t data_offset, 923 lldb::addr_t data_length) { 924 DataExtractor data; 925 data.SetData(data_sp, data_offset, data_length); 926 lldb::offset_t offset = 0; 927 uint32_t magic = data.GetU32(&offset); 928 929 offset += 4; // cputype 930 offset += 4; // cpusubtype 931 uint32_t filetype = data.GetU32(&offset); 932 933 // A fileset has a Mach-O header but is not an 934 // individual file and must be handled via an 935 // ObjectContainer plugin. 936 if (filetype == llvm::MachO::MH_FILESET) 937 return false; 938 939 return MachHeaderSizeFromMagic(magic) != 0; 940 } 941 942 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 943 DataBufferSP data_sp, 944 lldb::offset_t data_offset, 945 const FileSpec *file, 946 lldb::offset_t file_offset, 947 lldb::offset_t length) 948 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 949 m_mach_sections(), m_entry_point_address(), m_thread_context_offsets(), 950 m_thread_context_offsets_valid(false), m_reexported_dylibs(), 951 m_allow_assembly_emulation_unwind_plans(true) { 952 ::memset(&m_header, 0, sizeof(m_header)); 953 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 954 } 955 956 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 957 lldb::WritableDataBufferSP header_data_sp, 958 const lldb::ProcessSP &process_sp, 959 lldb::addr_t header_addr) 960 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 961 m_mach_sections(), m_entry_point_address(), m_thread_context_offsets(), 962 m_thread_context_offsets_valid(false), m_reexported_dylibs(), 963 m_allow_assembly_emulation_unwind_plans(true) { 964 ::memset(&m_header, 0, sizeof(m_header)); 965 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 966 } 967 968 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 969 lldb::offset_t *data_offset_ptr, 970 llvm::MachO::mach_header &header) { 971 data.SetByteOrder(endian::InlHostByteOrder()); 972 // Leave magic in the original byte order 973 header.magic = data.GetU32(data_offset_ptr); 974 bool can_parse = false; 975 bool is_64_bit = false; 976 switch (header.magic) { 977 case MH_MAGIC: 978 data.SetByteOrder(endian::InlHostByteOrder()); 979 data.SetAddressByteSize(4); 980 can_parse = true; 981 break; 982 983 case MH_MAGIC_64: 984 data.SetByteOrder(endian::InlHostByteOrder()); 985 data.SetAddressByteSize(8); 986 can_parse = true; 987 is_64_bit = true; 988 break; 989 990 case MH_CIGAM: 991 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 992 ? eByteOrderLittle 993 : eByteOrderBig); 994 data.SetAddressByteSize(4); 995 can_parse = true; 996 break; 997 998 case MH_CIGAM_64: 999 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1000 ? eByteOrderLittle 1001 : eByteOrderBig); 1002 data.SetAddressByteSize(8); 1003 is_64_bit = true; 1004 can_parse = true; 1005 break; 1006 1007 default: 1008 break; 1009 } 1010 1011 if (can_parse) { 1012 data.GetU32(data_offset_ptr, &header.cputype, 6); 1013 if (is_64_bit) 1014 *data_offset_ptr += 4; 1015 return true; 1016 } else { 1017 memset(&header, 0, sizeof(header)); 1018 } 1019 return false; 1020 } 1021 1022 bool ObjectFileMachO::ParseHeader() { 1023 ModuleSP module_sp(GetModule()); 1024 if (!module_sp) 1025 return false; 1026 1027 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1028 bool can_parse = false; 1029 lldb::offset_t offset = 0; 1030 m_data.SetByteOrder(endian::InlHostByteOrder()); 1031 // Leave magic in the original byte order 1032 m_header.magic = m_data.GetU32(&offset); 1033 switch (m_header.magic) { 1034 case MH_MAGIC: 1035 m_data.SetByteOrder(endian::InlHostByteOrder()); 1036 m_data.SetAddressByteSize(4); 1037 can_parse = true; 1038 break; 1039 1040 case MH_MAGIC_64: 1041 m_data.SetByteOrder(endian::InlHostByteOrder()); 1042 m_data.SetAddressByteSize(8); 1043 can_parse = true; 1044 break; 1045 1046 case MH_CIGAM: 1047 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1048 ? eByteOrderLittle 1049 : eByteOrderBig); 1050 m_data.SetAddressByteSize(4); 1051 can_parse = true; 1052 break; 1053 1054 case MH_CIGAM_64: 1055 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1056 ? eByteOrderLittle 1057 : eByteOrderBig); 1058 m_data.SetAddressByteSize(8); 1059 can_parse = true; 1060 break; 1061 1062 default: 1063 break; 1064 } 1065 1066 if (can_parse) { 1067 m_data.GetU32(&offset, &m_header.cputype, 6); 1068 1069 ModuleSpecList all_specs; 1070 ModuleSpec base_spec; 1071 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 1072 base_spec, all_specs); 1073 1074 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 1075 ArchSpec mach_arch = 1076 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 1077 1078 // Check if the module has a required architecture 1079 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1080 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1081 continue; 1082 1083 if (SetModulesArchitecture(mach_arch)) { 1084 const size_t header_and_lc_size = 1085 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1086 if (m_data.GetByteSize() < header_and_lc_size) { 1087 DataBufferSP data_sp; 1088 ProcessSP process_sp(m_process_wp.lock()); 1089 if (process_sp) { 1090 data_sp = ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1091 } else { 1092 // Read in all only the load command data from the file on disk 1093 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset); 1094 if (data_sp->GetByteSize() != header_and_lc_size) 1095 continue; 1096 } 1097 if (data_sp) 1098 m_data.SetData(data_sp); 1099 } 1100 } 1101 return true; 1102 } 1103 // None found. 1104 return false; 1105 } else { 1106 memset(&m_header, 0, sizeof(struct llvm::MachO::mach_header)); 1107 } 1108 return false; 1109 } 1110 1111 ByteOrder ObjectFileMachO::GetByteOrder() const { 1112 return m_data.GetByteOrder(); 1113 } 1114 1115 bool ObjectFileMachO::IsExecutable() const { 1116 return m_header.filetype == MH_EXECUTE; 1117 } 1118 1119 bool ObjectFileMachO::IsDynamicLoader() const { 1120 return m_header.filetype == MH_DYLINKER; 1121 } 1122 1123 bool ObjectFileMachO::IsSharedCacheBinary() const { 1124 return m_header.flags & MH_DYLIB_IN_CACHE; 1125 } 1126 1127 bool ObjectFileMachO::IsKext() const { 1128 return m_header.filetype == MH_KEXT_BUNDLE; 1129 } 1130 1131 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1132 return m_data.GetAddressByteSize(); 1133 } 1134 1135 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1136 Symtab *symtab = GetSymtab(); 1137 if (!symtab) 1138 return AddressClass::eUnknown; 1139 1140 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1141 if (symbol) { 1142 if (symbol->ValueIsAddress()) { 1143 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1144 if (section_sp) { 1145 const lldb::SectionType section_type = section_sp->GetType(); 1146 switch (section_type) { 1147 case eSectionTypeInvalid: 1148 return AddressClass::eUnknown; 1149 1150 case eSectionTypeCode: 1151 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1152 // For ARM we have a bit in the n_desc field of the symbol that 1153 // tells us ARM/Thumb which is bit 0x0008. 1154 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1155 return AddressClass::eCodeAlternateISA; 1156 } 1157 return AddressClass::eCode; 1158 1159 case eSectionTypeContainer: 1160 return AddressClass::eUnknown; 1161 1162 case eSectionTypeData: 1163 case eSectionTypeDataCString: 1164 case eSectionTypeDataCStringPointers: 1165 case eSectionTypeDataSymbolAddress: 1166 case eSectionTypeData4: 1167 case eSectionTypeData8: 1168 case eSectionTypeData16: 1169 case eSectionTypeDataPointers: 1170 case eSectionTypeZeroFill: 1171 case eSectionTypeDataObjCMessageRefs: 1172 case eSectionTypeDataObjCCFStrings: 1173 case eSectionTypeGoSymtab: 1174 return AddressClass::eData; 1175 1176 case eSectionTypeDebug: 1177 case eSectionTypeDWARFDebugAbbrev: 1178 case eSectionTypeDWARFDebugAbbrevDwo: 1179 case eSectionTypeDWARFDebugAddr: 1180 case eSectionTypeDWARFDebugAranges: 1181 case eSectionTypeDWARFDebugCuIndex: 1182 case eSectionTypeDWARFDebugFrame: 1183 case eSectionTypeDWARFDebugInfo: 1184 case eSectionTypeDWARFDebugInfoDwo: 1185 case eSectionTypeDWARFDebugLine: 1186 case eSectionTypeDWARFDebugLineStr: 1187 case eSectionTypeDWARFDebugLoc: 1188 case eSectionTypeDWARFDebugLocDwo: 1189 case eSectionTypeDWARFDebugLocLists: 1190 case eSectionTypeDWARFDebugLocListsDwo: 1191 case eSectionTypeDWARFDebugMacInfo: 1192 case eSectionTypeDWARFDebugMacro: 1193 case eSectionTypeDWARFDebugNames: 1194 case eSectionTypeDWARFDebugPubNames: 1195 case eSectionTypeDWARFDebugPubTypes: 1196 case eSectionTypeDWARFDebugRanges: 1197 case eSectionTypeDWARFDebugRngLists: 1198 case eSectionTypeDWARFDebugRngListsDwo: 1199 case eSectionTypeDWARFDebugStr: 1200 case eSectionTypeDWARFDebugStrDwo: 1201 case eSectionTypeDWARFDebugStrOffsets: 1202 case eSectionTypeDWARFDebugStrOffsetsDwo: 1203 case eSectionTypeDWARFDebugTuIndex: 1204 case eSectionTypeDWARFDebugTypes: 1205 case eSectionTypeDWARFDebugTypesDwo: 1206 case eSectionTypeDWARFAppleNames: 1207 case eSectionTypeDWARFAppleTypes: 1208 case eSectionTypeDWARFAppleNamespaces: 1209 case eSectionTypeDWARFAppleObjC: 1210 case eSectionTypeDWARFGNUDebugAltLink: 1211 case eSectionTypeCTF: 1212 case eSectionTypeLLDBTypeSummaries: 1213 case eSectionTypeSwiftModules: 1214 return AddressClass::eDebug; 1215 1216 case eSectionTypeEHFrame: 1217 case eSectionTypeARMexidx: 1218 case eSectionTypeARMextab: 1219 case eSectionTypeCompactUnwind: 1220 return AddressClass::eRuntime; 1221 1222 case eSectionTypeAbsoluteAddress: 1223 case eSectionTypeELFSymbolTable: 1224 case eSectionTypeELFDynamicSymbols: 1225 case eSectionTypeELFRelocationEntries: 1226 case eSectionTypeELFDynamicLinkInfo: 1227 case eSectionTypeOther: 1228 return AddressClass::eUnknown; 1229 } 1230 } 1231 } 1232 1233 const SymbolType symbol_type = symbol->GetType(); 1234 switch (symbol_type) { 1235 case eSymbolTypeAny: 1236 return AddressClass::eUnknown; 1237 case eSymbolTypeAbsolute: 1238 return AddressClass::eUnknown; 1239 1240 case eSymbolTypeCode: 1241 case eSymbolTypeTrampoline: 1242 case eSymbolTypeResolver: 1243 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1244 // For ARM we have a bit in the n_desc field of the symbol that tells 1245 // us ARM/Thumb which is bit 0x0008. 1246 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1247 return AddressClass::eCodeAlternateISA; 1248 } 1249 return AddressClass::eCode; 1250 1251 case eSymbolTypeData: 1252 return AddressClass::eData; 1253 case eSymbolTypeRuntime: 1254 return AddressClass::eRuntime; 1255 case eSymbolTypeException: 1256 return AddressClass::eRuntime; 1257 case eSymbolTypeSourceFile: 1258 return AddressClass::eDebug; 1259 case eSymbolTypeHeaderFile: 1260 return AddressClass::eDebug; 1261 case eSymbolTypeObjectFile: 1262 return AddressClass::eDebug; 1263 case eSymbolTypeCommonBlock: 1264 return AddressClass::eDebug; 1265 case eSymbolTypeBlock: 1266 return AddressClass::eDebug; 1267 case eSymbolTypeLocal: 1268 return AddressClass::eData; 1269 case eSymbolTypeParam: 1270 return AddressClass::eData; 1271 case eSymbolTypeVariable: 1272 return AddressClass::eData; 1273 case eSymbolTypeVariableType: 1274 return AddressClass::eDebug; 1275 case eSymbolTypeLineEntry: 1276 return AddressClass::eDebug; 1277 case eSymbolTypeLineHeader: 1278 return AddressClass::eDebug; 1279 case eSymbolTypeScopeBegin: 1280 return AddressClass::eDebug; 1281 case eSymbolTypeScopeEnd: 1282 return AddressClass::eDebug; 1283 case eSymbolTypeAdditional: 1284 return AddressClass::eUnknown; 1285 case eSymbolTypeCompiler: 1286 return AddressClass::eDebug; 1287 case eSymbolTypeInstrumentation: 1288 return AddressClass::eDebug; 1289 case eSymbolTypeUndefined: 1290 return AddressClass::eUnknown; 1291 case eSymbolTypeObjCClass: 1292 return AddressClass::eRuntime; 1293 case eSymbolTypeObjCMetaClass: 1294 return AddressClass::eRuntime; 1295 case eSymbolTypeObjCIVar: 1296 return AddressClass::eRuntime; 1297 case eSymbolTypeReExported: 1298 return AddressClass::eRuntime; 1299 } 1300 } 1301 return AddressClass::eUnknown; 1302 } 1303 1304 bool ObjectFileMachO::IsStripped() { 1305 if (m_dysymtab.cmd == 0) { 1306 ModuleSP module_sp(GetModule()); 1307 if (module_sp) { 1308 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1309 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1310 const lldb::offset_t load_cmd_offset = offset; 1311 1312 llvm::MachO::load_command lc = {}; 1313 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 1314 break; 1315 if (lc.cmd == LC_DYSYMTAB) { 1316 m_dysymtab.cmd = lc.cmd; 1317 m_dysymtab.cmdsize = lc.cmdsize; 1318 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1319 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1320 nullptr) { 1321 // Clear m_dysymtab if we were unable to read all items from the 1322 // load command 1323 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1324 } 1325 } 1326 offset = load_cmd_offset + lc.cmdsize; 1327 } 1328 } 1329 } 1330 if (m_dysymtab.cmd) 1331 return m_dysymtab.nlocalsym <= 1; 1332 return false; 1333 } 1334 1335 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1336 EncryptedFileRanges result; 1337 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1338 1339 llvm::MachO::encryption_info_command encryption_cmd; 1340 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1341 const lldb::offset_t load_cmd_offset = offset; 1342 if (m_data.GetU32(&offset, &encryption_cmd, 2) == nullptr) 1343 break; 1344 1345 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the 1346 // 3 fields we care about, so treat them the same. 1347 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1348 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1349 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1350 if (encryption_cmd.cryptid != 0) { 1351 EncryptedFileRanges::Entry entry; 1352 entry.SetRangeBase(encryption_cmd.cryptoff); 1353 entry.SetByteSize(encryption_cmd.cryptsize); 1354 result.Append(entry); 1355 } 1356 } 1357 } 1358 offset = load_cmd_offset + encryption_cmd.cmdsize; 1359 } 1360 1361 return result; 1362 } 1363 1364 void ObjectFileMachO::SanitizeSegmentCommand( 1365 llvm::MachO::segment_command_64 &seg_cmd, uint32_t cmd_idx) { 1366 if (m_length == 0 || seg_cmd.filesize == 0) 1367 return; 1368 1369 if (IsSharedCacheBinary() && !IsInMemory()) { 1370 // In shared cache images, the load commands are relative to the 1371 // shared cache file, and not the specific image we are 1372 // examining. Let's fix this up so that it looks like a normal 1373 // image. 1374 if (strncmp(seg_cmd.segname, GetSegmentNameTEXT().GetCString(), 1375 sizeof(seg_cmd.segname)) == 0) 1376 m_text_address = seg_cmd.vmaddr; 1377 if (strncmp(seg_cmd.segname, GetSegmentNameLINKEDIT().GetCString(), 1378 sizeof(seg_cmd.segname)) == 0) 1379 m_linkedit_original_offset = seg_cmd.fileoff; 1380 1381 seg_cmd.fileoff = seg_cmd.vmaddr - m_text_address; 1382 } 1383 1384 if (seg_cmd.fileoff > m_length) { 1385 // We have a load command that says it extends past the end of the file. 1386 // This is likely a corrupt file. We don't have any way to return an error 1387 // condition here (this method was likely invoked from something like 1388 // ObjectFile::GetSectionList()), so we just null out the section contents, 1389 // and dump a message to stdout. The most common case here is core file 1390 // debugging with a truncated file. 1391 const char *lc_segment_name = 1392 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1393 GetModule()->ReportWarning( 1394 "load command {0} {1} has a fileoff ({2:x16}) that extends beyond " 1395 "the end of the file ({3:x16}), ignoring this section", 1396 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1397 1398 seg_cmd.fileoff = 0; 1399 seg_cmd.filesize = 0; 1400 } 1401 1402 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1403 // We have a load command that says it extends past the end of the file. 1404 // This is likely a corrupt file. We don't have any way to return an error 1405 // condition here (this method was likely invoked from something like 1406 // ObjectFile::GetSectionList()), so we just null out the section contents, 1407 // and dump a message to stdout. The most common case here is core file 1408 // debugging with a truncated file. 1409 const char *lc_segment_name = 1410 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1411 GetModule()->ReportWarning( 1412 "load command {0} {1} has a fileoff + filesize ({2:x16}) that " 1413 "extends beyond the end of the file ({3:x16}), the segment will be " 1414 "truncated to match", 1415 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1416 1417 // Truncate the length 1418 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1419 } 1420 } 1421 1422 static uint32_t 1423 GetSegmentPermissions(const llvm::MachO::segment_command_64 &seg_cmd) { 1424 uint32_t result = 0; 1425 if (seg_cmd.initprot & VM_PROT_READ) 1426 result |= ePermissionsReadable; 1427 if (seg_cmd.initprot & VM_PROT_WRITE) 1428 result |= ePermissionsWritable; 1429 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1430 result |= ePermissionsExecutable; 1431 return result; 1432 } 1433 1434 static lldb::SectionType GetSectionType(uint32_t flags, 1435 ConstString section_name) { 1436 1437 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1438 return eSectionTypeCode; 1439 1440 uint32_t mach_sect_type = flags & SECTION_TYPE; 1441 static ConstString g_sect_name_objc_data("__objc_data"); 1442 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1443 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1444 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1445 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1446 static ConstString g_sect_name_objc_const("__objc_const"); 1447 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1448 static ConstString g_sect_name_cfstring("__cfstring"); 1449 1450 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1451 static ConstString g_sect_name_dwarf_debug_abbrev_dwo("__debug_abbrev.dwo"); 1452 static ConstString g_sect_name_dwarf_debug_addr("__debug_addr"); 1453 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1454 static ConstString g_sect_name_dwarf_debug_cu_index("__debug_cu_index"); 1455 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1456 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1457 static ConstString g_sect_name_dwarf_debug_info_dwo("__debug_info.dwo"); 1458 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1459 static ConstString g_sect_name_dwarf_debug_line_dwo("__debug_line.dwo"); 1460 static ConstString g_sect_name_dwarf_debug_line_str("__debug_line_str"); 1461 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1462 static ConstString g_sect_name_dwarf_debug_loclists("__debug_loclists"); 1463 static ConstString g_sect_name_dwarf_debug_loclists_dwo("__debug_loclists.dwo"); 1464 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1465 static ConstString g_sect_name_dwarf_debug_macro("__debug_macro"); 1466 static ConstString g_sect_name_dwarf_debug_macro_dwo("__debug_macro.dwo"); 1467 static ConstString g_sect_name_dwarf_debug_names("__debug_names"); 1468 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1469 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1470 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1471 static ConstString g_sect_name_dwarf_debug_rnglists("__debug_rnglists"); 1472 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1473 static ConstString g_sect_name_dwarf_debug_str_dwo("__debug_str.dwo"); 1474 static ConstString g_sect_name_dwarf_debug_str_offs("__debug_str_offs"); 1475 static ConstString g_sect_name_dwarf_debug_str_offs_dwo("__debug_str_offs.dwo"); 1476 static ConstString g_sect_name_dwarf_debug_tu_index("__debug_tu_index"); 1477 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1478 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1479 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1480 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1481 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1482 static ConstString g_sect_name_eh_frame("__eh_frame"); 1483 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1484 static ConstString g_sect_name_text("__text"); 1485 static ConstString g_sect_name_data("__data"); 1486 static ConstString g_sect_name_go_symtab("__gosymtab"); 1487 static ConstString g_sect_name_ctf("__ctf"); 1488 static ConstString g_sect_name_lldb_summaries("__lldbsummaries"); 1489 static ConstString g_sect_name_swift_ast("__swift_ast"); 1490 1491 if (section_name == g_sect_name_dwarf_debug_abbrev) 1492 return eSectionTypeDWARFDebugAbbrev; 1493 if (section_name == g_sect_name_dwarf_debug_abbrev_dwo) 1494 return eSectionTypeDWARFDebugAbbrevDwo; 1495 if (section_name == g_sect_name_dwarf_debug_addr) 1496 return eSectionTypeDWARFDebugAddr; 1497 if (section_name == g_sect_name_dwarf_debug_aranges) 1498 return eSectionTypeDWARFDebugAranges; 1499 if (section_name == g_sect_name_dwarf_debug_cu_index) 1500 return eSectionTypeDWARFDebugCuIndex; 1501 if (section_name == g_sect_name_dwarf_debug_frame) 1502 return eSectionTypeDWARFDebugFrame; 1503 if (section_name == g_sect_name_dwarf_debug_info) 1504 return eSectionTypeDWARFDebugInfo; 1505 if (section_name == g_sect_name_dwarf_debug_info_dwo) 1506 return eSectionTypeDWARFDebugInfoDwo; 1507 if (section_name == g_sect_name_dwarf_debug_line) 1508 return eSectionTypeDWARFDebugLine; 1509 if (section_name == g_sect_name_dwarf_debug_line_dwo) 1510 return eSectionTypeDWARFDebugLine; // Same as debug_line. 1511 if (section_name == g_sect_name_dwarf_debug_line_str) 1512 return eSectionTypeDWARFDebugLineStr; 1513 if (section_name == g_sect_name_dwarf_debug_loc) 1514 return eSectionTypeDWARFDebugLoc; 1515 if (section_name == g_sect_name_dwarf_debug_loclists) 1516 return eSectionTypeDWARFDebugLocLists; 1517 if (section_name == g_sect_name_dwarf_debug_loclists_dwo) 1518 return eSectionTypeDWARFDebugLocListsDwo; 1519 if (section_name == g_sect_name_dwarf_debug_macinfo) 1520 return eSectionTypeDWARFDebugMacInfo; 1521 if (section_name == g_sect_name_dwarf_debug_macro) 1522 return eSectionTypeDWARFDebugMacro; 1523 if (section_name == g_sect_name_dwarf_debug_macro_dwo) 1524 return eSectionTypeDWARFDebugMacInfo; // Same as debug_macro. 1525 if (section_name == g_sect_name_dwarf_debug_names) 1526 return eSectionTypeDWARFDebugNames; 1527 if (section_name == g_sect_name_dwarf_debug_pubnames) 1528 return eSectionTypeDWARFDebugPubNames; 1529 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1530 return eSectionTypeDWARFDebugPubTypes; 1531 if (section_name == g_sect_name_dwarf_debug_ranges) 1532 return eSectionTypeDWARFDebugRanges; 1533 if (section_name == g_sect_name_dwarf_debug_rnglists) 1534 return eSectionTypeDWARFDebugRngLists; 1535 if (section_name == g_sect_name_dwarf_debug_str) 1536 return eSectionTypeDWARFDebugStr; 1537 if (section_name == g_sect_name_dwarf_debug_str_dwo) 1538 return eSectionTypeDWARFDebugStrDwo; 1539 if (section_name == g_sect_name_dwarf_debug_str_offs) 1540 return eSectionTypeDWARFDebugStrOffsets; 1541 if (section_name == g_sect_name_dwarf_debug_str_offs_dwo) 1542 return eSectionTypeDWARFDebugStrOffsetsDwo; 1543 if (section_name == g_sect_name_dwarf_debug_tu_index) 1544 return eSectionTypeDWARFDebugTuIndex; 1545 if (section_name == g_sect_name_dwarf_debug_types) 1546 return eSectionTypeDWARFDebugTypes; 1547 if (section_name == g_sect_name_dwarf_apple_names) 1548 return eSectionTypeDWARFAppleNames; 1549 if (section_name == g_sect_name_dwarf_apple_types) 1550 return eSectionTypeDWARFAppleTypes; 1551 if (section_name == g_sect_name_dwarf_apple_namespaces) 1552 return eSectionTypeDWARFAppleNamespaces; 1553 if (section_name == g_sect_name_dwarf_apple_objc) 1554 return eSectionTypeDWARFAppleObjC; 1555 if (section_name == g_sect_name_objc_selrefs) 1556 return eSectionTypeDataCStringPointers; 1557 if (section_name == g_sect_name_objc_msgrefs) 1558 return eSectionTypeDataObjCMessageRefs; 1559 if (section_name == g_sect_name_eh_frame) 1560 return eSectionTypeEHFrame; 1561 if (section_name == g_sect_name_compact_unwind) 1562 return eSectionTypeCompactUnwind; 1563 if (section_name == g_sect_name_cfstring) 1564 return eSectionTypeDataObjCCFStrings; 1565 if (section_name == g_sect_name_go_symtab) 1566 return eSectionTypeGoSymtab; 1567 if (section_name == g_sect_name_ctf) 1568 return eSectionTypeCTF; 1569 if (section_name == g_sect_name_lldb_summaries) 1570 return lldb::eSectionTypeLLDBTypeSummaries; 1571 if (section_name == g_sect_name_swift_ast) 1572 return eSectionTypeSwiftModules; 1573 if (section_name == g_sect_name_objc_data || 1574 section_name == g_sect_name_objc_classrefs || 1575 section_name == g_sect_name_objc_superrefs || 1576 section_name == g_sect_name_objc_const || 1577 section_name == g_sect_name_objc_classlist) { 1578 return eSectionTypeDataPointers; 1579 } 1580 1581 switch (mach_sect_type) { 1582 // TODO: categorize sections by other flags for regular sections 1583 case S_REGULAR: 1584 if (section_name == g_sect_name_text) 1585 return eSectionTypeCode; 1586 if (section_name == g_sect_name_data) 1587 return eSectionTypeData; 1588 return eSectionTypeOther; 1589 case S_ZEROFILL: 1590 return eSectionTypeZeroFill; 1591 case S_CSTRING_LITERALS: // section with only literal C strings 1592 return eSectionTypeDataCString; 1593 case S_4BYTE_LITERALS: // section with only 4 byte literals 1594 return eSectionTypeData4; 1595 case S_8BYTE_LITERALS: // section with only 8 byte literals 1596 return eSectionTypeData8; 1597 case S_LITERAL_POINTERS: // section with only pointers to literals 1598 return eSectionTypeDataPointers; 1599 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1600 return eSectionTypeDataPointers; 1601 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1602 return eSectionTypeDataPointers; 1603 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1604 // the reserved2 field 1605 return eSectionTypeCode; 1606 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1607 // initialization 1608 return eSectionTypeDataPointers; 1609 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1610 // termination 1611 return eSectionTypeDataPointers; 1612 case S_COALESCED: 1613 return eSectionTypeOther; 1614 case S_GB_ZEROFILL: 1615 return eSectionTypeZeroFill; 1616 case S_INTERPOSING: // section with only pairs of function pointers for 1617 // interposing 1618 return eSectionTypeCode; 1619 case S_16BYTE_LITERALS: // section with only 16 byte literals 1620 return eSectionTypeData16; 1621 case S_DTRACE_DOF: 1622 return eSectionTypeDebug; 1623 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1624 return eSectionTypeDataPointers; 1625 default: 1626 return eSectionTypeOther; 1627 } 1628 } 1629 1630 struct ObjectFileMachO::SegmentParsingContext { 1631 const EncryptedFileRanges EncryptedRanges; 1632 lldb_private::SectionList &UnifiedList; 1633 uint32_t NextSegmentIdx = 0; 1634 uint32_t NextSectionIdx = 0; 1635 bool FileAddressesChanged = false; 1636 1637 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1638 lldb_private::SectionList &UnifiedList) 1639 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1640 }; 1641 1642 void ObjectFileMachO::ProcessSegmentCommand( 1643 const llvm::MachO::load_command &load_cmd_, lldb::offset_t offset, 1644 uint32_t cmd_idx, SegmentParsingContext &context) { 1645 llvm::MachO::segment_command_64 load_cmd; 1646 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1647 1648 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1649 return; 1650 1651 ModuleSP module_sp = GetModule(); 1652 const bool is_core = GetType() == eTypeCoreFile; 1653 const bool is_dsym = (m_header.filetype == MH_DSYM); 1654 bool add_section = true; 1655 bool add_to_unified = true; 1656 ConstString const_segname( 1657 load_cmd.segname, strnlen(load_cmd.segname, sizeof(load_cmd.segname))); 1658 1659 SectionSP unified_section_sp( 1660 context.UnifiedList.FindSectionByName(const_segname)); 1661 if (is_dsym && unified_section_sp) { 1662 if (const_segname == GetSegmentNameLINKEDIT()) { 1663 // We need to keep the __LINKEDIT segment private to this object file 1664 // only 1665 add_to_unified = false; 1666 } else { 1667 // This is the dSYM file and this section has already been created by the 1668 // object file, no need to create it. 1669 add_section = false; 1670 } 1671 } 1672 load_cmd.vmaddr = m_data.GetAddress(&offset); 1673 load_cmd.vmsize = m_data.GetAddress(&offset); 1674 load_cmd.fileoff = m_data.GetAddress(&offset); 1675 load_cmd.filesize = m_data.GetAddress(&offset); 1676 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1677 return; 1678 1679 SanitizeSegmentCommand(load_cmd, cmd_idx); 1680 1681 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1682 const bool segment_is_encrypted = 1683 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1684 1685 // Use a segment ID of the segment index shifted left by 8 so they never 1686 // conflict with any of the sections. 1687 SectionSP segment_sp; 1688 if (add_section && (const_segname || is_core)) { 1689 segment_sp = std::make_shared<Section>( 1690 module_sp, // Module to which this section belongs 1691 this, // Object file to which this sections belongs 1692 ++context.NextSegmentIdx 1693 << 8, // Section ID is the 1 based segment index 1694 // shifted right by 8 bits as not to collide with any of the 256 1695 // section IDs that are possible 1696 const_segname, // Name of this section 1697 eSectionTypeContainer, // This section is a container of other 1698 // sections. 1699 load_cmd.vmaddr, // File VM address == addresses as they are 1700 // found in the object file 1701 load_cmd.vmsize, // VM size in bytes of this section 1702 load_cmd.fileoff, // Offset to the data for this section in 1703 // the file 1704 load_cmd.filesize, // Size in bytes of this section as found 1705 // in the file 1706 0, // Segments have no alignment information 1707 load_cmd.flags); // Flags for this section 1708 1709 segment_sp->SetIsEncrypted(segment_is_encrypted); 1710 m_sections_up->AddSection(segment_sp); 1711 segment_sp->SetPermissions(segment_permissions); 1712 if (add_to_unified) 1713 context.UnifiedList.AddSection(segment_sp); 1714 } else if (unified_section_sp) { 1715 // If this is a dSYM and the file addresses in the dSYM differ from the 1716 // file addresses in the ObjectFile, we must use the file base address for 1717 // the Section from the dSYM for the DWARF to resolve correctly. 1718 // This only happens with binaries in the shared cache in practice; 1719 // normally a mismatch like this would give a binary & dSYM that do not 1720 // match UUIDs. When a binary is included in the shared cache, its 1721 // segments are rearranged to optimize the shared cache, so its file 1722 // addresses will differ from what the ObjectFile had originally, 1723 // and what the dSYM has. 1724 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1725 Log *log = GetLog(LLDBLog::Symbols); 1726 if (log) { 1727 log->Printf( 1728 "Installing dSYM's %s segment file address over ObjectFile's " 1729 "so symbol table/debug info resolves correctly for %s", 1730 const_segname.AsCString(), 1731 module_sp->GetFileSpec().GetFilename().AsCString()); 1732 } 1733 1734 // Make sure we've parsed the symbol table from the ObjectFile before 1735 // we go around changing its Sections. 1736 module_sp->GetObjectFile()->GetSymtab(); 1737 // eh_frame would present the same problems but we parse that on a per- 1738 // function basis as-needed so it's more difficult to remove its use of 1739 // the Sections. Realistically, the environments where this code path 1740 // will be taken will not have eh_frame sections. 1741 1742 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1743 1744 // Notify the module that the section addresses have been changed once 1745 // we're done so any file-address caches can be updated. 1746 context.FileAddressesChanged = true; 1747 } 1748 m_sections_up->AddSection(unified_section_sp); 1749 } 1750 1751 llvm::MachO::section_64 sect64; 1752 ::memset(§64, 0, sizeof(sect64)); 1753 // Push a section into our mach sections for the section at index zero 1754 // (NO_SECT) if we don't have any mach sections yet... 1755 if (m_mach_sections.empty()) 1756 m_mach_sections.push_back(sect64); 1757 uint32_t segment_sect_idx; 1758 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1759 1760 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1761 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1762 ++segment_sect_idx) { 1763 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1764 sizeof(sect64.sectname)) == nullptr) 1765 break; 1766 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1767 sizeof(sect64.segname)) == nullptr) 1768 break; 1769 sect64.addr = m_data.GetAddress(&offset); 1770 sect64.size = m_data.GetAddress(&offset); 1771 1772 if (m_data.GetU32(&offset, §64.offset, num_u32s) == nullptr) 1773 break; 1774 1775 if (IsSharedCacheBinary() && !IsInMemory()) { 1776 sect64.offset = sect64.addr - m_text_address; 1777 } 1778 1779 // Keep a list of mach sections around in case we need to get at data that 1780 // isn't stored in the abstracted Sections. 1781 m_mach_sections.push_back(sect64); 1782 1783 if (add_section) { 1784 ConstString section_name( 1785 sect64.sectname, strnlen(sect64.sectname, sizeof(sect64.sectname))); 1786 if (!const_segname) { 1787 // We have a segment with no name so we need to conjure up segments 1788 // that correspond to the section's segname if there isn't already such 1789 // a section. If there is such a section, we resize the section so that 1790 // it spans all sections. We also mark these sections as fake so 1791 // address matches don't hit if they land in the gaps between the child 1792 // sections. 1793 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1794 sizeof(sect64.segname)); 1795 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1796 if (segment_sp.get()) { 1797 Section *segment = segment_sp.get(); 1798 // Grow the section size as needed. 1799 const lldb::addr_t sect64_min_addr = sect64.addr; 1800 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1801 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1802 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1803 const lldb::addr_t curr_seg_max_addr = 1804 curr_seg_min_addr + curr_seg_byte_size; 1805 if (sect64_min_addr >= curr_seg_min_addr) { 1806 const lldb::addr_t new_seg_byte_size = 1807 sect64_max_addr - curr_seg_min_addr; 1808 // Only grow the section size if needed 1809 if (new_seg_byte_size > curr_seg_byte_size) 1810 segment->SetByteSize(new_seg_byte_size); 1811 } else { 1812 // We need to change the base address of the segment and adjust the 1813 // child section offsets for all existing children. 1814 const lldb::addr_t slide_amount = 1815 sect64_min_addr - curr_seg_min_addr; 1816 segment->Slide(slide_amount, false); 1817 segment->GetChildren().Slide(-slide_amount, false); 1818 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1819 } 1820 1821 // Grow the section size as needed. 1822 if (sect64.offset) { 1823 const lldb::addr_t segment_min_file_offset = 1824 segment->GetFileOffset(); 1825 const lldb::addr_t segment_max_file_offset = 1826 segment_min_file_offset + segment->GetFileSize(); 1827 1828 const lldb::addr_t section_min_file_offset = sect64.offset; 1829 const lldb::addr_t section_max_file_offset = 1830 section_min_file_offset + sect64.size; 1831 const lldb::addr_t new_file_offset = 1832 std::min(section_min_file_offset, segment_min_file_offset); 1833 const lldb::addr_t new_file_size = 1834 std::max(section_max_file_offset, segment_max_file_offset) - 1835 new_file_offset; 1836 segment->SetFileOffset(new_file_offset); 1837 segment->SetFileSize(new_file_size); 1838 } 1839 } else { 1840 // Create a fake section for the section's named segment 1841 segment_sp = std::make_shared<Section>( 1842 segment_sp, // Parent section 1843 module_sp, // Module to which this section belongs 1844 this, // Object file to which this section belongs 1845 ++context.NextSegmentIdx 1846 << 8, // Section ID is the 1 based segment index 1847 // shifted right by 8 bits as not to 1848 // collide with any of the 256 section IDs 1849 // that are possible 1850 const_segname, // Name of this section 1851 eSectionTypeContainer, // This section is a container of 1852 // other sections. 1853 sect64.addr, // File VM address == addresses as they are 1854 // found in the object file 1855 sect64.size, // VM size in bytes of this section 1856 sect64.offset, // Offset to the data for this section in 1857 // the file 1858 sect64.offset ? sect64.size : 0, // Size in bytes of 1859 // this section as 1860 // found in the file 1861 sect64.align, 1862 load_cmd.flags); // Flags for this section 1863 segment_sp->SetIsFake(true); 1864 segment_sp->SetPermissions(segment_permissions); 1865 m_sections_up->AddSection(segment_sp); 1866 if (add_to_unified) 1867 context.UnifiedList.AddSection(segment_sp); 1868 segment_sp->SetIsEncrypted(segment_is_encrypted); 1869 } 1870 } 1871 assert(segment_sp.get()); 1872 1873 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1874 1875 SectionSP section_sp(new Section( 1876 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1877 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1878 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1879 sect64.flags)); 1880 // Set the section to be encrypted to match the segment 1881 1882 bool section_is_encrypted = false; 1883 if (!segment_is_encrypted && load_cmd.filesize != 0) 1884 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1885 sect64.offset) != nullptr; 1886 1887 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1888 section_sp->SetPermissions(segment_permissions); 1889 segment_sp->GetChildren().AddSection(section_sp); 1890 1891 if (segment_sp->IsFake()) { 1892 segment_sp.reset(); 1893 const_segname.Clear(); 1894 } 1895 } 1896 } 1897 if (segment_sp && is_dsym) { 1898 if (first_segment_sectID <= context.NextSectionIdx) { 1899 lldb::user_id_t sect_uid; 1900 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1901 ++sect_uid) { 1902 SectionSP curr_section_sp( 1903 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1904 SectionSP next_section_sp; 1905 if (sect_uid + 1 <= context.NextSectionIdx) 1906 next_section_sp = 1907 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1908 1909 if (curr_section_sp.get()) { 1910 if (curr_section_sp->GetByteSize() == 0) { 1911 if (next_section_sp.get() != nullptr) 1912 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1913 curr_section_sp->GetFileAddress()); 1914 else 1915 curr_section_sp->SetByteSize(load_cmd.vmsize); 1916 } 1917 } 1918 } 1919 } 1920 } 1921 } 1922 1923 void ObjectFileMachO::ProcessDysymtabCommand( 1924 const llvm::MachO::load_command &load_cmd, lldb::offset_t offset) { 1925 m_dysymtab.cmd = load_cmd.cmd; 1926 m_dysymtab.cmdsize = load_cmd.cmdsize; 1927 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1928 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1929 } 1930 1931 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1932 if (m_sections_up) 1933 return; 1934 1935 m_sections_up = std::make_unique<SectionList>(); 1936 1937 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1938 // bool dump_sections = false; 1939 ModuleSP module_sp(GetModule()); 1940 1941 offset = MachHeaderSizeFromMagic(m_header.magic); 1942 1943 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1944 llvm::MachO::load_command load_cmd; 1945 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1946 const lldb::offset_t load_cmd_offset = offset; 1947 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 1948 break; 1949 1950 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1951 ProcessSegmentCommand(load_cmd, offset, i, context); 1952 else if (load_cmd.cmd == LC_DYSYMTAB) 1953 ProcessDysymtabCommand(load_cmd, offset); 1954 1955 offset = load_cmd_offset + load_cmd.cmdsize; 1956 } 1957 1958 if (context.FileAddressesChanged && module_sp) 1959 module_sp->SectionFileAddressesChanged(); 1960 } 1961 1962 class MachSymtabSectionInfo { 1963 public: 1964 MachSymtabSectionInfo(SectionList *section_list) 1965 : m_section_list(section_list), m_section_infos() { 1966 // Get the number of sections down to a depth of 1 to include all segments 1967 // and their sections, but no other sections that may be added for debug 1968 // map or 1969 m_section_infos.resize(section_list->GetNumSections(1)); 1970 } 1971 1972 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1973 if (n_sect == 0) 1974 return SectionSP(); 1975 if (n_sect < m_section_infos.size()) { 1976 if (!m_section_infos[n_sect].section_sp) { 1977 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1978 m_section_infos[n_sect].section_sp = section_sp; 1979 if (section_sp) { 1980 m_section_infos[n_sect].vm_range.SetBaseAddress( 1981 section_sp->GetFileAddress()); 1982 m_section_infos[n_sect].vm_range.SetByteSize( 1983 section_sp->GetByteSize()); 1984 } else { 1985 std::string filename = "<unknown>"; 1986 SectionSP first_section_sp(m_section_list->GetSectionAtIndex(0)); 1987 if (first_section_sp) 1988 filename = first_section_sp->GetObjectFile()->GetFileSpec().GetPath(); 1989 1990 Debugger::ReportError( 1991 llvm::formatv("unable to find section {0} for a symbol in " 1992 "{1}, corrupt file?", 1993 n_sect, filename)); 1994 } 1995 } 1996 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1997 // Symbol is in section. 1998 return m_section_infos[n_sect].section_sp; 1999 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 2000 m_section_infos[n_sect].vm_range.GetBaseAddress() == 2001 file_addr) { 2002 // Symbol is in section with zero size, but has the same start address 2003 // as the section. This can happen with linker symbols (symbols that 2004 // start with the letter 'l' or 'L'. 2005 return m_section_infos[n_sect].section_sp; 2006 } 2007 } 2008 return m_section_list->FindSectionContainingFileAddress(file_addr); 2009 } 2010 2011 protected: 2012 struct SectionInfo { 2013 SectionInfo() : vm_range(), section_sp() {} 2014 2015 VMRange vm_range; 2016 SectionSP section_sp; 2017 }; 2018 SectionList *m_section_list; 2019 std::vector<SectionInfo> m_section_infos; 2020 }; 2021 2022 #define TRIE_SYMBOL_IS_THUMB (1ULL << 63) 2023 struct TrieEntry { 2024 void Dump() const { 2025 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 2026 static_cast<unsigned long long>(address), 2027 static_cast<unsigned long long>(flags), 2028 static_cast<unsigned long long>(other), name.GetCString()); 2029 if (import_name) 2030 printf(" -> \"%s\"\n", import_name.GetCString()); 2031 else 2032 printf("\n"); 2033 } 2034 ConstString name; 2035 uint64_t address = LLDB_INVALID_ADDRESS; 2036 uint64_t flags = 2037 0; // EXPORT_SYMBOL_FLAGS_REEXPORT, EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER, 2038 // TRIE_SYMBOL_IS_THUMB 2039 uint64_t other = 0; 2040 ConstString import_name; 2041 }; 2042 2043 struct TrieEntryWithOffset { 2044 lldb::offset_t nodeOffset; 2045 TrieEntry entry; 2046 2047 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 2048 2049 void Dump(uint32_t idx) const { 2050 printf("[%3u] 0x%16.16llx: ", idx, 2051 static_cast<unsigned long long>(nodeOffset)); 2052 entry.Dump(); 2053 } 2054 2055 bool operator<(const TrieEntryWithOffset &other) const { 2056 return (nodeOffset < other.nodeOffset); 2057 } 2058 }; 2059 2060 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2061 const bool is_arm, addr_t text_seg_base_addr, 2062 std::vector<llvm::StringRef> &nameSlices, 2063 std::set<lldb::addr_t> &resolver_addresses, 2064 std::vector<TrieEntryWithOffset> &reexports, 2065 std::vector<TrieEntryWithOffset> &ext_symbols) { 2066 if (!data.ValidOffset(offset)) 2067 return true; 2068 2069 // Terminal node -- end of a branch, possibly add this to 2070 // the symbol table or resolver table. 2071 const uint64_t terminalSize = data.GetULEB128(&offset); 2072 lldb::offset_t children_offset = offset + terminalSize; 2073 if (terminalSize != 0) { 2074 TrieEntryWithOffset e(offset); 2075 e.entry.flags = data.GetULEB128(&offset); 2076 const char *import_name = nullptr; 2077 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2078 e.entry.address = 0; 2079 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2080 import_name = data.GetCStr(&offset); 2081 } else { 2082 e.entry.address = data.GetULEB128(&offset); 2083 if (text_seg_base_addr != LLDB_INVALID_ADDRESS) 2084 e.entry.address += text_seg_base_addr; 2085 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2086 e.entry.other = data.GetULEB128(&offset); 2087 uint64_t resolver_addr = e.entry.other; 2088 if (text_seg_base_addr != LLDB_INVALID_ADDRESS) 2089 resolver_addr += text_seg_base_addr; 2090 if (is_arm) 2091 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2092 resolver_addresses.insert(resolver_addr); 2093 } else 2094 e.entry.other = 0; 2095 } 2096 bool add_this_entry = false; 2097 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT) && 2098 import_name && import_name[0]) { 2099 // add symbols that are reexport symbols with a valid import name. 2100 add_this_entry = true; 2101 } else if (e.entry.flags == 0 && 2102 (import_name == nullptr || import_name[0] == '\0')) { 2103 // add externally visible symbols, in case the nlist record has 2104 // been stripped/omitted. 2105 add_this_entry = true; 2106 } 2107 if (add_this_entry) { 2108 std::string name; 2109 if (!nameSlices.empty()) { 2110 for (auto name_slice : nameSlices) 2111 name.append(name_slice.data(), name_slice.size()); 2112 } 2113 if (name.size() > 1) { 2114 // Skip the leading '_' 2115 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2116 } 2117 if (import_name) { 2118 // Skip the leading '_' 2119 e.entry.import_name.SetCString(import_name + 1); 2120 } 2121 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT)) { 2122 reexports.push_back(e); 2123 } else { 2124 if (is_arm && (e.entry.address & 1)) { 2125 e.entry.flags |= TRIE_SYMBOL_IS_THUMB; 2126 e.entry.address &= THUMB_ADDRESS_BIT_MASK; 2127 } 2128 ext_symbols.push_back(e); 2129 } 2130 } 2131 } 2132 2133 const uint8_t childrenCount = data.GetU8(&children_offset); 2134 for (uint8_t i = 0; i < childrenCount; ++i) { 2135 const char *cstr = data.GetCStr(&children_offset); 2136 if (cstr) 2137 nameSlices.push_back(llvm::StringRef(cstr)); 2138 else 2139 return false; // Corrupt data 2140 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2141 if (childNodeOffset) { 2142 if (!ParseTrieEntries(data, childNodeOffset, is_arm, text_seg_base_addr, 2143 nameSlices, resolver_addresses, reexports, 2144 ext_symbols)) { 2145 return false; 2146 } 2147 } 2148 nameSlices.pop_back(); 2149 } 2150 return true; 2151 } 2152 2153 static SymbolType GetSymbolType(const char *&symbol_name, 2154 bool &demangled_is_synthesized, 2155 const SectionSP &text_section_sp, 2156 const SectionSP &data_section_sp, 2157 const SectionSP &data_dirty_section_sp, 2158 const SectionSP &data_const_section_sp, 2159 const SectionSP &symbol_section) { 2160 SymbolType type = eSymbolTypeInvalid; 2161 2162 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2163 if (symbol_section->IsDescendant(text_section_sp.get())) { 2164 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 2165 S_ATTR_SELF_MODIFYING_CODE | 2166 S_ATTR_SOME_INSTRUCTIONS)) 2167 type = eSymbolTypeData; 2168 else 2169 type = eSymbolTypeCode; 2170 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 2171 symbol_section->IsDescendant(data_dirty_section_sp.get()) || 2172 symbol_section->IsDescendant(data_const_section_sp.get())) { 2173 if (symbol_sect_name && 2174 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 2175 type = eSymbolTypeRuntime; 2176 2177 if (symbol_name) { 2178 llvm::StringRef symbol_name_ref(symbol_name); 2179 if (symbol_name_ref.starts_with("OBJC_")) { 2180 static const llvm::StringRef g_objc_v2_prefix_class("OBJC_CLASS_$_"); 2181 static const llvm::StringRef g_objc_v2_prefix_metaclass( 2182 "OBJC_METACLASS_$_"); 2183 static const llvm::StringRef g_objc_v2_prefix_ivar("OBJC_IVAR_$_"); 2184 if (symbol_name_ref.starts_with(g_objc_v2_prefix_class)) { 2185 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2186 type = eSymbolTypeObjCClass; 2187 demangled_is_synthesized = true; 2188 } else if (symbol_name_ref.starts_with(g_objc_v2_prefix_metaclass)) { 2189 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2190 type = eSymbolTypeObjCMetaClass; 2191 demangled_is_synthesized = true; 2192 } else if (symbol_name_ref.starts_with(g_objc_v2_prefix_ivar)) { 2193 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2194 type = eSymbolTypeObjCIVar; 2195 demangled_is_synthesized = true; 2196 } 2197 } 2198 } 2199 } else if (symbol_sect_name && 2200 ::strstr(symbol_sect_name, "__gcc_except_tab") == 2201 symbol_sect_name) { 2202 type = eSymbolTypeException; 2203 } else { 2204 type = eSymbolTypeData; 2205 } 2206 } else if (symbol_sect_name && 2207 ::strstr(symbol_sect_name, "__IMPORT") == symbol_sect_name) { 2208 type = eSymbolTypeTrampoline; 2209 } 2210 return type; 2211 } 2212 2213 static std::optional<struct nlist_64> 2214 ParseNList(DataExtractor &nlist_data, lldb::offset_t &nlist_data_offset, 2215 size_t nlist_byte_size) { 2216 struct nlist_64 nlist; 2217 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2218 return {}; 2219 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 2220 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 2221 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 2222 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 2223 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 2224 return nlist; 2225 } 2226 2227 enum { DebugSymbols = true, NonDebugSymbols = false }; 2228 2229 void ObjectFileMachO::ParseSymtab(Symtab &symtab) { 2230 ModuleSP module_sp(GetModule()); 2231 if (!module_sp) 2232 return; 2233 2234 Log *log = GetLog(LLDBLog::Symbols); 2235 2236 const FileSpec &file = m_file ? m_file : module_sp->GetFileSpec(); 2237 const char *file_name = file.GetFilename().AsCString("<Unknown>"); 2238 LLDB_SCOPED_TIMERF("ObjectFileMachO::ParseSymtab () module = %s", file_name); 2239 LLDB_LOG(log, "Parsing symbol table for {0}", file_name); 2240 Progress progress("Parsing symbol table", file_name); 2241 2242 llvm::MachO::linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2243 llvm::MachO::linkedit_data_command exports_trie_load_command = {0, 0, 0, 0}; 2244 llvm::MachO::dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2245 llvm::MachO::dysymtab_command dysymtab = m_dysymtab; 2246 SymtabCommandLargeOffsets symtab_load_command; 2247 // The data element of type bool indicates that this entry is thumb 2248 // code. 2249 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2250 2251 // Record the address of every function/data that we add to the symtab. 2252 // We add symbols to the table in the order of most information (nlist 2253 // records) to least (function starts), and avoid duplicating symbols 2254 // via this set. 2255 llvm::DenseSet<addr_t> symbols_added; 2256 2257 // We are using a llvm::DenseSet for "symbols_added" so we must be sure we 2258 // do not add the tombstone or empty keys to the set. 2259 auto add_symbol_addr = [&symbols_added](lldb::addr_t file_addr) { 2260 // Don't add the tombstone or empty keys. 2261 if (file_addr == UINT64_MAX || file_addr == UINT64_MAX - 1) 2262 return; 2263 symbols_added.insert(file_addr); 2264 }; 2265 FunctionStarts function_starts; 2266 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2267 uint32_t i; 2268 FileSpecList dylib_files; 2269 llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2270 llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2271 llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2272 UUID image_uuid; 2273 2274 for (i = 0; i < m_header.ncmds; ++i) { 2275 const lldb::offset_t cmd_offset = offset; 2276 // Read in the load command and load command size 2277 llvm::MachO::load_command lc; 2278 if (m_data.GetU32(&offset, &lc, 2) == nullptr) 2279 break; 2280 // Watch for the symbol table load command 2281 switch (lc.cmd) { 2282 case LC_SYMTAB: 2283 // struct symtab_command { 2284 // uint32_t cmd; /* LC_SYMTAB */ 2285 // uint32_t cmdsize; /* sizeof(struct symtab_command) */ 2286 // uint32_t symoff; /* symbol table offset */ 2287 // uint32_t nsyms; /* number of symbol table entries */ 2288 // uint32_t stroff; /* string table offset */ 2289 // uint32_t strsize; /* string table size in bytes */ 2290 // }; 2291 symtab_load_command.cmd = lc.cmd; 2292 symtab_load_command.cmdsize = lc.cmdsize; 2293 symtab_load_command.symoff = m_data.GetU32(&offset); 2294 symtab_load_command.nsyms = m_data.GetU32(&offset); 2295 symtab_load_command.stroff = m_data.GetU32(&offset); 2296 symtab_load_command.strsize = m_data.GetU32(&offset); 2297 break; 2298 2299 case LC_DYLD_INFO: 2300 case LC_DYLD_INFO_ONLY: 2301 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2302 dyld_info.cmd = lc.cmd; 2303 dyld_info.cmdsize = lc.cmdsize; 2304 } else { 2305 memset(&dyld_info, 0, sizeof(dyld_info)); 2306 } 2307 break; 2308 2309 case LC_LOAD_DYLIB: 2310 case LC_LOAD_WEAK_DYLIB: 2311 case LC_REEXPORT_DYLIB: 2312 case LC_LOADFVMLIB: 2313 case LC_LOAD_UPWARD_DYLIB: { 2314 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2315 const char *path = m_data.PeekCStr(name_offset); 2316 if (path) { 2317 FileSpec file_spec(path); 2318 // Strip the path if there is @rpath, @executable, etc so we just use 2319 // the basename 2320 if (path[0] == '@') 2321 file_spec.ClearDirectory(); 2322 2323 if (lc.cmd == LC_REEXPORT_DYLIB) { 2324 m_reexported_dylibs.AppendIfUnique(file_spec); 2325 } 2326 2327 dylib_files.Append(file_spec); 2328 } 2329 } break; 2330 2331 case LC_DYLD_EXPORTS_TRIE: 2332 exports_trie_load_command.cmd = lc.cmd; 2333 exports_trie_load_command.cmdsize = lc.cmdsize; 2334 if (m_data.GetU32(&offset, &exports_trie_load_command.dataoff, 2) == 2335 nullptr) // fill in offset and size fields 2336 memset(&exports_trie_load_command, 0, 2337 sizeof(exports_trie_load_command)); 2338 break; 2339 case LC_FUNCTION_STARTS: 2340 function_starts_load_command.cmd = lc.cmd; 2341 function_starts_load_command.cmdsize = lc.cmdsize; 2342 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2343 nullptr) // fill in data offset and size fields 2344 memset(&function_starts_load_command, 0, 2345 sizeof(function_starts_load_command)); 2346 break; 2347 2348 case LC_UUID: { 2349 const uint8_t *uuid_bytes = m_data.PeekData(offset, 16); 2350 2351 if (uuid_bytes) 2352 image_uuid = UUID(uuid_bytes, 16); 2353 break; 2354 } 2355 2356 default: 2357 break; 2358 } 2359 offset = cmd_offset + lc.cmdsize; 2360 } 2361 2362 if (!symtab_load_command.cmd) 2363 return; 2364 2365 SectionList *section_list = GetSectionList(); 2366 if (section_list == nullptr) 2367 return; 2368 2369 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2370 const ByteOrder byte_order = m_data.GetByteOrder(); 2371 bool bit_width_32 = addr_byte_size == 4; 2372 const size_t nlist_byte_size = 2373 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2374 2375 DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size); 2376 DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size); 2377 DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size); 2378 DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order, 2379 addr_byte_size); 2380 DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size); 2381 2382 const addr_t nlist_data_byte_size = 2383 symtab_load_command.nsyms * nlist_byte_size; 2384 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2385 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2386 2387 ProcessSP process_sp(m_process_wp.lock()); 2388 Process *process = process_sp.get(); 2389 2390 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2391 bool is_shared_cache_image = IsSharedCacheBinary(); 2392 bool is_local_shared_cache_image = is_shared_cache_image && !IsInMemory(); 2393 SectionSP linkedit_section_sp( 2394 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2395 2396 if (process && m_header.filetype != llvm::MachO::MH_OBJECT && 2397 !is_local_shared_cache_image) { 2398 Target &target = process->GetTarget(); 2399 2400 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2401 2402 // Reading mach file from memory in a process or core file... 2403 2404 if (linkedit_section_sp) { 2405 addr_t linkedit_load_addr = 2406 linkedit_section_sp->GetLoadBaseAddress(&target); 2407 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2408 // We might be trying to access the symbol table before the 2409 // __LINKEDIT's load address has been set in the target. We can't 2410 // fail to read the symbol table, so calculate the right address 2411 // manually 2412 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2413 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2414 } 2415 2416 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2417 const addr_t symoff_addr = linkedit_load_addr + 2418 symtab_load_command.symoff - 2419 linkedit_file_offset; 2420 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2421 linkedit_file_offset; 2422 2423 // Always load dyld - the dynamic linker - from memory if we didn't 2424 // find a binary anywhere else. lldb will not register 2425 // dylib/framework/bundle loads/unloads if we don't have the dyld 2426 // symbols, we force dyld to load from memory despite the user's 2427 // target.memory-module-load-level setting. 2428 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2429 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2430 DataBufferSP nlist_data_sp( 2431 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2432 if (nlist_data_sp) 2433 nlist_data.SetData(nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2434 if (dysymtab.nindirectsyms != 0) { 2435 const addr_t indirect_syms_addr = linkedit_load_addr + 2436 dysymtab.indirectsymoff - 2437 linkedit_file_offset; 2438 DataBufferSP indirect_syms_data_sp(ReadMemory( 2439 process_sp, indirect_syms_addr, dysymtab.nindirectsyms * 4)); 2440 if (indirect_syms_data_sp) 2441 indirect_symbol_index_data.SetData( 2442 indirect_syms_data_sp, 0, indirect_syms_data_sp->GetByteSize()); 2443 // If this binary is outside the shared cache, 2444 // cache the string table. 2445 // Binaries in the shared cache all share a giant string table, 2446 // and we can't share the string tables across multiple 2447 // ObjectFileMachO's, so we'd end up re-reading this mega-strtab 2448 // for every binary in the shared cache - it would be a big perf 2449 // problem. For binaries outside the shared cache, it's faster to 2450 // read the entire strtab at once instead of piece-by-piece as we 2451 // process the nlist records. 2452 if (!is_shared_cache_image) { 2453 DataBufferSP strtab_data_sp( 2454 ReadMemory(process_sp, strtab_addr, strtab_data_byte_size)); 2455 if (strtab_data_sp) { 2456 strtab_data.SetData(strtab_data_sp, 0, 2457 strtab_data_sp->GetByteSize()); 2458 } 2459 } 2460 } 2461 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) { 2462 if (function_starts_load_command.cmd) { 2463 const addr_t func_start_addr = 2464 linkedit_load_addr + function_starts_load_command.dataoff - 2465 linkedit_file_offset; 2466 DataBufferSP func_start_data_sp( 2467 ReadMemory(process_sp, func_start_addr, 2468 function_starts_load_command.datasize)); 2469 if (func_start_data_sp) 2470 function_starts_data.SetData(func_start_data_sp, 0, 2471 func_start_data_sp->GetByteSize()); 2472 } 2473 } 2474 } 2475 } 2476 } else { 2477 if (is_local_shared_cache_image) { 2478 // The load commands in shared cache images are relative to the 2479 // beginning of the shared cache, not the library image. The 2480 // data we get handed when creating the ObjectFileMachO starts 2481 // at the beginning of a specific library and spans to the end 2482 // of the cache to be able to reach the shared LINKEDIT 2483 // segments. We need to convert the load command offsets to be 2484 // relative to the beginning of our specific image. 2485 lldb::addr_t linkedit_offset = linkedit_section_sp->GetFileOffset(); 2486 lldb::offset_t linkedit_slide = 2487 linkedit_offset - m_linkedit_original_offset; 2488 symtab_load_command.symoff += linkedit_slide; 2489 symtab_load_command.stroff += linkedit_slide; 2490 dyld_info.export_off += linkedit_slide; 2491 dysymtab.indirectsymoff += linkedit_slide; 2492 function_starts_load_command.dataoff += linkedit_slide; 2493 exports_trie_load_command.dataoff += linkedit_slide; 2494 } 2495 2496 nlist_data.SetData(m_data, symtab_load_command.symoff, 2497 nlist_data_byte_size); 2498 strtab_data.SetData(m_data, symtab_load_command.stroff, 2499 strtab_data_byte_size); 2500 2501 // We shouldn't have exports data from both the LC_DYLD_INFO command 2502 // AND the LC_DYLD_EXPORTS_TRIE command in the same binary: 2503 lldbassert(!((dyld_info.export_size > 0) 2504 && (exports_trie_load_command.datasize > 0))); 2505 if (dyld_info.export_size > 0) { 2506 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2507 dyld_info.export_size); 2508 } else if (exports_trie_load_command.datasize > 0) { 2509 dyld_trie_data.SetData(m_data, exports_trie_load_command.dataoff, 2510 exports_trie_load_command.datasize); 2511 } 2512 2513 if (dysymtab.nindirectsyms != 0) { 2514 indirect_symbol_index_data.SetData(m_data, dysymtab.indirectsymoff, 2515 dysymtab.nindirectsyms * 4); 2516 } 2517 if (function_starts_load_command.cmd) { 2518 function_starts_data.SetData(m_data, function_starts_load_command.dataoff, 2519 function_starts_load_command.datasize); 2520 } 2521 } 2522 2523 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2524 2525 ConstString g_segment_name_TEXT = GetSegmentNameTEXT(); 2526 ConstString g_segment_name_DATA = GetSegmentNameDATA(); 2527 ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2528 ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2529 ConstString g_segment_name_OBJC = GetSegmentNameOBJC(); 2530 ConstString g_section_name_eh_frame = GetSectionNameEHFrame(); 2531 SectionSP text_section_sp( 2532 section_list->FindSectionByName(g_segment_name_TEXT)); 2533 SectionSP data_section_sp( 2534 section_list->FindSectionByName(g_segment_name_DATA)); 2535 SectionSP data_dirty_section_sp( 2536 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2537 SectionSP data_const_section_sp( 2538 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2539 SectionSP objc_section_sp( 2540 section_list->FindSectionByName(g_segment_name_OBJC)); 2541 SectionSP eh_frame_section_sp; 2542 if (text_section_sp.get()) 2543 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2544 g_section_name_eh_frame); 2545 else 2546 eh_frame_section_sp = 2547 section_list->FindSectionByName(g_section_name_eh_frame); 2548 2549 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2550 const bool always_thumb = GetArchitecture().IsAlwaysThumbInstructions(); 2551 2552 // lldb works best if it knows the start address of all functions in a 2553 // module. Linker symbols or debug info are normally the best source of 2554 // information for start addr / size but they may be stripped in a released 2555 // binary. Two additional sources of information exist in Mach-O binaries: 2556 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2557 // function's start address in the 2558 // binary, relative to the text section. 2559 // eh_frame - the eh_frame FDEs have the start addr & size of 2560 // each function 2561 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2562 // all modern binaries. 2563 // Binaries built to run on older releases may need to use eh_frame 2564 // information. 2565 2566 if (text_section_sp && function_starts_data.GetByteSize()) { 2567 FunctionStarts::Entry function_start_entry; 2568 function_start_entry.data = false; 2569 lldb::offset_t function_start_offset = 0; 2570 function_start_entry.addr = text_section_sp->GetFileAddress(); 2571 uint64_t delta; 2572 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2573 0) { 2574 // Now append the current entry 2575 function_start_entry.addr += delta; 2576 if (is_arm) { 2577 if (function_start_entry.addr & 1) { 2578 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2579 function_start_entry.data = true; 2580 } else if (always_thumb) { 2581 function_start_entry.data = true; 2582 } 2583 } 2584 function_starts.Append(function_start_entry); 2585 } 2586 } else { 2587 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2588 // load command claiming an eh_frame but it doesn't actually have the 2589 // eh_frame content. And if we have a dSYM, we don't need to do any of 2590 // this fill-in-the-missing-symbols works anyway - the debug info should 2591 // give us all the functions in the module. 2592 if (text_section_sp.get() && eh_frame_section_sp.get() && 2593 m_type != eTypeDebugInfo) { 2594 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2595 DWARFCallFrameInfo::EH); 2596 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2597 eh_frame.GetFunctionAddressAndSizeVector(functions); 2598 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2599 size_t count = functions.GetSize(); 2600 for (size_t i = 0; i < count; ++i) { 2601 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2602 functions.GetEntryAtIndex(i); 2603 if (func) { 2604 FunctionStarts::Entry function_start_entry; 2605 function_start_entry.addr = func->base - text_base_addr; 2606 if (is_arm) { 2607 if (function_start_entry.addr & 1) { 2608 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2609 function_start_entry.data = true; 2610 } else if (always_thumb) { 2611 function_start_entry.data = true; 2612 } 2613 } 2614 function_starts.Append(function_start_entry); 2615 } 2616 } 2617 } 2618 } 2619 2620 const size_t function_starts_count = function_starts.GetSize(); 2621 2622 // For user process binaries (executables, dylibs, frameworks, bundles), if 2623 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2624 // going to assume the binary has been stripped. Don't allow assembly 2625 // language instruction emulation because we don't know proper function 2626 // start boundaries. 2627 // 2628 // For all other types of binaries (kernels, stand-alone bare board 2629 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2630 // sections - we should not make any assumptions about them based on that. 2631 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2632 m_allow_assembly_emulation_unwind_plans = false; 2633 Log *unwind_or_symbol_log(GetLog(LLDBLog::Symbols | LLDBLog::Unwind)); 2634 2635 if (unwind_or_symbol_log) 2636 module_sp->LogMessage( 2637 unwind_or_symbol_log, 2638 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2639 } 2640 2641 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() 2642 ? eh_frame_section_sp->GetID() 2643 : static_cast<user_id_t>(NO_SECT); 2644 2645 uint32_t N_SO_index = UINT32_MAX; 2646 2647 MachSymtabSectionInfo section_info(section_list); 2648 std::vector<uint32_t> N_FUN_indexes; 2649 std::vector<uint32_t> N_NSYM_indexes; 2650 std::vector<uint32_t> N_INCL_indexes; 2651 std::vector<uint32_t> N_BRAC_indexes; 2652 std::vector<uint32_t> N_COMM_indexes; 2653 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2654 typedef llvm::DenseMap<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2655 typedef llvm::DenseMap<const char *, uint32_t> ConstNameToSymbolIndexMap; 2656 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2657 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2658 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2659 // Any symbols that get merged into another will get an entry in this map 2660 // so we know 2661 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2662 uint32_t nlist_idx = 0; 2663 Symbol *symbol_ptr = nullptr; 2664 2665 uint32_t sym_idx = 0; 2666 Symbol *sym = nullptr; 2667 size_t num_syms = 0; 2668 std::string memory_symbol_name; 2669 uint32_t unmapped_local_symbols_found = 0; 2670 2671 std::vector<TrieEntryWithOffset> reexport_trie_entries; 2672 std::vector<TrieEntryWithOffset> external_sym_trie_entries; 2673 std::set<lldb::addr_t> resolver_addresses; 2674 2675 const size_t dyld_trie_data_size = dyld_trie_data.GetByteSize(); 2676 if (dyld_trie_data_size > 0) { 2677 LLDB_LOG(log, "Parsing {0} bytes of dyld trie data", dyld_trie_data_size); 2678 SectionSP text_segment_sp = 2679 GetSectionList()->FindSectionByName(GetSegmentNameTEXT()); 2680 lldb::addr_t text_segment_file_addr = LLDB_INVALID_ADDRESS; 2681 if (text_segment_sp) 2682 text_segment_file_addr = text_segment_sp->GetFileAddress(); 2683 std::vector<llvm::StringRef> nameSlices; 2684 ParseTrieEntries(dyld_trie_data, 0, is_arm, text_segment_file_addr, 2685 nameSlices, resolver_addresses, reexport_trie_entries, 2686 external_sym_trie_entries); 2687 } 2688 2689 typedef std::set<ConstString> IndirectSymbols; 2690 IndirectSymbols indirect_symbol_names; 2691 2692 #if TARGET_OS_IPHONE 2693 2694 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2695 // optimized by moving LOCAL symbols out of the memory mapped portion of 2696 // the DSC. The symbol information has all been retained, but it isn't 2697 // available in the normal nlist data. However, there *are* duplicate 2698 // entries of *some* 2699 // LOCAL symbols in the normal nlist data. To handle this situation 2700 // correctly, we must first attempt 2701 // to parse any DSC unmapped symbol information. If we find any, we set a 2702 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2703 2704 if (IsSharedCacheBinary()) { 2705 // Before we can start mapping the DSC, we need to make certain the 2706 // target process is actually using the cache we can find. 2707 2708 // Next we need to determine the correct path for the dyld shared cache. 2709 2710 ArchSpec header_arch = GetArchitecture(); 2711 2712 UUID dsc_uuid; 2713 UUID process_shared_cache_uuid; 2714 addr_t process_shared_cache_base_addr; 2715 2716 if (process) { 2717 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, 2718 process_shared_cache_uuid); 2719 } 2720 2721 __block bool found_image = false; 2722 __block void *nlist_buffer = nullptr; 2723 __block unsigned nlist_count = 0; 2724 __block char *string_table = nullptr; 2725 __block vm_offset_t vm_nlist_memory = 0; 2726 __block mach_msg_type_number_t vm_nlist_bytes_read = 0; 2727 __block vm_offset_t vm_string_memory = 0; 2728 __block mach_msg_type_number_t vm_string_bytes_read = 0; 2729 2730 auto _ = llvm::make_scope_exit(^{ 2731 if (vm_nlist_memory) 2732 vm_deallocate(mach_task_self(), vm_nlist_memory, vm_nlist_bytes_read); 2733 if (vm_string_memory) 2734 vm_deallocate(mach_task_self(), vm_string_memory, vm_string_bytes_read); 2735 }); 2736 2737 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 2738 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 2739 UndefinedNameToDescMap undefined_name_to_desc; 2740 SymbolIndexToName reexport_shlib_needs_fixup; 2741 2742 dyld_for_each_installed_shared_cache(^(dyld_shared_cache_t shared_cache) { 2743 uuid_t cache_uuid; 2744 dyld_shared_cache_copy_uuid(shared_cache, &cache_uuid); 2745 if (found_image) 2746 return; 2747 2748 if (process_shared_cache_uuid.IsValid() && 2749 process_shared_cache_uuid != UUID::fromData(&cache_uuid, 16)) 2750 return; 2751 2752 dyld_shared_cache_for_each_image(shared_cache, ^(dyld_image_t image) { 2753 uuid_t dsc_image_uuid; 2754 if (found_image) 2755 return; 2756 2757 dyld_image_copy_uuid(image, &dsc_image_uuid); 2758 if (image_uuid != UUID::fromData(dsc_image_uuid, 16)) 2759 return; 2760 2761 found_image = true; 2762 2763 // Compute the size of the string table. We need to ask dyld for a 2764 // new SPI to avoid this step. 2765 dyld_image_local_nlist_content_4Symbolication( 2766 image, ^(const void *nlistStart, uint64_t nlistCount, 2767 const char *stringTable) { 2768 if (!nlistStart || !nlistCount) 2769 return; 2770 2771 // The buffers passed here are valid only inside the block. 2772 // Use vm_read to make a cheap copy of them available for our 2773 // processing later. 2774 kern_return_t ret = 2775 vm_read(mach_task_self(), (vm_address_t)nlistStart, 2776 nlist_byte_size * nlistCount, &vm_nlist_memory, 2777 &vm_nlist_bytes_read); 2778 if (ret != KERN_SUCCESS) 2779 return; 2780 assert(vm_nlist_bytes_read == nlist_byte_size * nlistCount); 2781 2782 // We don't know the size of the string table. It's cheaper 2783 // to map the whole VM region than to determine the size by 2784 // parsing all the nlist entries. 2785 vm_address_t string_address = (vm_address_t)stringTable; 2786 vm_size_t region_size; 2787 mach_msg_type_number_t info_count = VM_REGION_BASIC_INFO_COUNT_64; 2788 vm_region_basic_info_data_t info; 2789 memory_object_name_t object; 2790 ret = vm_region_64(mach_task_self(), &string_address, 2791 ®ion_size, VM_REGION_BASIC_INFO_64, 2792 (vm_region_info_t)&info, &info_count, &object); 2793 if (ret != KERN_SUCCESS) 2794 return; 2795 2796 ret = vm_read(mach_task_self(), (vm_address_t)stringTable, 2797 region_size - 2798 ((vm_address_t)stringTable - string_address), 2799 &vm_string_memory, &vm_string_bytes_read); 2800 if (ret != KERN_SUCCESS) 2801 return; 2802 2803 nlist_buffer = (void *)vm_nlist_memory; 2804 string_table = (char *)vm_string_memory; 2805 nlist_count = nlistCount; 2806 }); 2807 }); 2808 }); 2809 if (nlist_buffer) { 2810 DataExtractor dsc_local_symbols_data(nlist_buffer, 2811 nlist_count * nlist_byte_size, 2812 byte_order, addr_byte_size); 2813 unmapped_local_symbols_found = nlist_count; 2814 2815 // The normal nlist code cannot correctly size the Symbols 2816 // array, we need to allocate it here. 2817 sym = symtab.Resize( 2818 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2819 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2820 num_syms = symtab.GetNumSymbols(); 2821 2822 lldb::offset_t nlist_data_offset = 0; 2823 2824 for (uint32_t nlist_index = 0; 2825 nlist_index < nlist_count; 2826 nlist_index++) { 2827 ///////////////////////////// 2828 { 2829 std::optional<struct nlist_64> nlist_maybe = 2830 ParseNList(dsc_local_symbols_data, nlist_data_offset, 2831 nlist_byte_size); 2832 if (!nlist_maybe) 2833 break; 2834 struct nlist_64 nlist = *nlist_maybe; 2835 2836 SymbolType type = eSymbolTypeInvalid; 2837 const char *symbol_name = string_table + nlist.n_strx; 2838 2839 if (symbol_name == NULL) { 2840 // No symbol should be NULL, even the symbols with no 2841 // string values should have an offset zero which 2842 // points to an empty C-string 2843 Debugger::ReportError(llvm::formatv( 2844 "DSC unmapped local symbol[{0}] has invalid " 2845 "string table offset {1:x} in {2}, ignoring symbol", 2846 nlist_index, nlist.n_strx, 2847 module_sp->GetFileSpec().GetPath()); 2848 continue; 2849 } 2850 if (symbol_name[0] == '\0') 2851 symbol_name = NULL; 2852 2853 const char *symbol_name_non_abi_mangled = NULL; 2854 2855 SectionSP symbol_section; 2856 uint32_t symbol_byte_size = 0; 2857 bool add_nlist = true; 2858 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2859 bool demangled_is_synthesized = false; 2860 bool is_gsym = false; 2861 bool set_value = true; 2862 2863 assert(sym_idx < num_syms); 2864 2865 sym[sym_idx].SetDebug(is_debug); 2866 2867 if (is_debug) { 2868 switch (nlist.n_type) { 2869 case N_GSYM: 2870 // global symbol: name,,NO_SECT,type,0 2871 // Sometimes the N_GSYM value contains the address. 2872 2873 // FIXME: In the .o files, we have a GSYM and a debug 2874 // symbol for all the ObjC data. They 2875 // have the same address, but we want to ensure that 2876 // we always find only the real symbol, 'cause we 2877 // don't currently correctly attribute the 2878 // GSYM one to the ObjCClass/Ivar/MetaClass 2879 // symbol type. This is a temporary hack to make 2880 // sure the ObjectiveC symbols get treated correctly. 2881 // To do this right, we should coalesce all the GSYM 2882 // & global symbols that have the same address. 2883 2884 is_gsym = true; 2885 sym[sym_idx].SetExternal(true); 2886 2887 if (symbol_name && symbol_name[0] == '_' && 2888 symbol_name[1] == 'O') { 2889 llvm::StringRef symbol_name_ref(symbol_name); 2890 if (symbol_name_ref.starts_with( 2891 g_objc_v2_prefix_class)) { 2892 symbol_name_non_abi_mangled = symbol_name + 1; 2893 symbol_name = 2894 symbol_name + g_objc_v2_prefix_class.size(); 2895 type = eSymbolTypeObjCClass; 2896 demangled_is_synthesized = true; 2897 2898 } else if (symbol_name_ref.starts_with( 2899 g_objc_v2_prefix_metaclass)) { 2900 symbol_name_non_abi_mangled = symbol_name + 1; 2901 symbol_name = 2902 symbol_name + g_objc_v2_prefix_metaclass.size(); 2903 type = eSymbolTypeObjCMetaClass; 2904 demangled_is_synthesized = true; 2905 } else if (symbol_name_ref.starts_with( 2906 g_objc_v2_prefix_ivar)) { 2907 symbol_name_non_abi_mangled = symbol_name + 1; 2908 symbol_name = 2909 symbol_name + g_objc_v2_prefix_ivar.size(); 2910 type = eSymbolTypeObjCIVar; 2911 demangled_is_synthesized = true; 2912 } 2913 } else { 2914 if (nlist.n_value != 0) 2915 symbol_section = section_info.GetSection( 2916 nlist.n_sect, nlist.n_value); 2917 type = eSymbolTypeData; 2918 } 2919 break; 2920 2921 case N_FNAME: 2922 // procedure name (f77 kludge): name,,NO_SECT,0,0 2923 type = eSymbolTypeCompiler; 2924 break; 2925 2926 case N_FUN: 2927 // procedure: name,,n_sect,linenumber,address 2928 if (symbol_name) { 2929 type = eSymbolTypeCode; 2930 symbol_section = section_info.GetSection( 2931 nlist.n_sect, nlist.n_value); 2932 2933 N_FUN_addr_to_sym_idx.insert( 2934 std::make_pair(nlist.n_value, sym_idx)); 2935 // We use the current number of symbols in the 2936 // symbol table in lieu of using nlist_idx in case 2937 // we ever start trimming entries out 2938 N_FUN_indexes.push_back(sym_idx); 2939 } else { 2940 type = eSymbolTypeCompiler; 2941 2942 if (!N_FUN_indexes.empty()) { 2943 // Copy the size of the function into the 2944 // original 2945 // STAB entry so we don't have 2946 // to hunt for it later 2947 symtab.SymbolAtIndex(N_FUN_indexes.back()) 2948 ->SetByteSize(nlist.n_value); 2949 N_FUN_indexes.pop_back(); 2950 // We don't really need the end function STAB as 2951 // it contains the size which we already placed 2952 // with the original symbol, so don't add it if 2953 // we want a minimal symbol table 2954 add_nlist = false; 2955 } 2956 } 2957 break; 2958 2959 case N_STSYM: 2960 // static symbol: name,,n_sect,type,address 2961 N_STSYM_addr_to_sym_idx.insert( 2962 std::make_pair(nlist.n_value, sym_idx)); 2963 symbol_section = section_info.GetSection(nlist.n_sect, 2964 nlist.n_value); 2965 if (symbol_name && symbol_name[0]) { 2966 type = ObjectFile::GetSymbolTypeFromName( 2967 symbol_name + 1, eSymbolTypeData); 2968 } 2969 break; 2970 2971 case N_LCSYM: 2972 // .lcomm symbol: name,,n_sect,type,address 2973 symbol_section = section_info.GetSection(nlist.n_sect, 2974 nlist.n_value); 2975 type = eSymbolTypeCommonBlock; 2976 break; 2977 2978 case N_BNSYM: 2979 // We use the current number of symbols in the symbol 2980 // table in lieu of using nlist_idx in case we ever 2981 // start trimming entries out Skip these if we want 2982 // minimal symbol tables 2983 add_nlist = false; 2984 break; 2985 2986 case N_ENSYM: 2987 // Set the size of the N_BNSYM to the terminating 2988 // index of this N_ENSYM so that we can always skip 2989 // the entire symbol if we need to navigate more 2990 // quickly at the source level when parsing STABS 2991 // Skip these if we want minimal symbol tables 2992 add_nlist = false; 2993 break; 2994 2995 case N_OPT: 2996 // emitted with gcc2_compiled and in gcc source 2997 type = eSymbolTypeCompiler; 2998 break; 2999 3000 case N_RSYM: 3001 // register sym: name,,NO_SECT,type,register 3002 type = eSymbolTypeVariable; 3003 break; 3004 3005 case N_SLINE: 3006 // src line: 0,,n_sect,linenumber,address 3007 symbol_section = section_info.GetSection(nlist.n_sect, 3008 nlist.n_value); 3009 type = eSymbolTypeLineEntry; 3010 break; 3011 3012 case N_SSYM: 3013 // structure elt: name,,NO_SECT,type,struct_offset 3014 type = eSymbolTypeVariableType; 3015 break; 3016 3017 case N_SO: 3018 // source file name 3019 type = eSymbolTypeSourceFile; 3020 if (symbol_name == NULL) { 3021 add_nlist = false; 3022 if (N_SO_index != UINT32_MAX) { 3023 // Set the size of the N_SO to the terminating 3024 // index of this N_SO so that we can always skip 3025 // the entire N_SO if we need to navigate more 3026 // quickly at the source level when parsing STABS 3027 symbol_ptr = symtab.SymbolAtIndex(N_SO_index); 3028 symbol_ptr->SetByteSize(sym_idx); 3029 symbol_ptr->SetSizeIsSibling(true); 3030 } 3031 N_NSYM_indexes.clear(); 3032 N_INCL_indexes.clear(); 3033 N_BRAC_indexes.clear(); 3034 N_COMM_indexes.clear(); 3035 N_FUN_indexes.clear(); 3036 N_SO_index = UINT32_MAX; 3037 } else { 3038 // We use the current number of symbols in the 3039 // symbol table in lieu of using nlist_idx in case 3040 // we ever start trimming entries out 3041 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3042 if (N_SO_has_full_path) { 3043 if ((N_SO_index == sym_idx - 1) && 3044 ((sym_idx - 1) < num_syms)) { 3045 // We have two consecutive N_SO entries where 3046 // the first contains a directory and the 3047 // second contains a full path. 3048 sym[sym_idx - 1].GetMangled().SetValue( 3049 ConstString(symbol_name)); 3050 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3051 add_nlist = false; 3052 } else { 3053 // This is the first entry in a N_SO that 3054 // contains a directory or 3055 // a full path to the source file 3056 N_SO_index = sym_idx; 3057 } 3058 } else if ((N_SO_index == sym_idx - 1) && 3059 ((sym_idx - 1) < num_syms)) { 3060 // This is usually the second N_SO entry that 3061 // contains just the filename, so here we combine 3062 // it with the first one if we are minimizing the 3063 // symbol table 3064 const char *so_path = sym[sym_idx - 1] 3065 .GetMangled() 3066 .GetDemangledName() 3067 .AsCString(); 3068 if (so_path && so_path[0]) { 3069 std::string full_so_path(so_path); 3070 const size_t double_slash_pos = 3071 full_so_path.find("//"); 3072 if (double_slash_pos != std::string::npos) { 3073 // The linker has been generating bad N_SO 3074 // entries with doubled up paths 3075 // in the format "%s%s" where the first 3076 // string in the DW_AT_comp_dir, and the 3077 // second is the directory for the source 3078 // file so you end up with a path that looks 3079 // like "/tmp/src//tmp/src/" 3080 FileSpec so_dir(so_path); 3081 if (!FileSystem::Instance().Exists(so_dir)) { 3082 so_dir.SetFile( 3083 &full_so_path[double_slash_pos + 1], 3084 FileSpec::Style::native); 3085 if (FileSystem::Instance().Exists(so_dir)) { 3086 // Trim off the incorrect path 3087 full_so_path.erase(0, double_slash_pos + 1); 3088 } 3089 } 3090 } 3091 if (*full_so_path.rbegin() != '/') 3092 full_so_path += '/'; 3093 full_so_path += symbol_name; 3094 sym[sym_idx - 1].GetMangled().SetValue( 3095 ConstString(full_so_path.c_str())); 3096 add_nlist = false; 3097 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3098 } 3099 } else { 3100 // This could be a relative path to a N_SO 3101 N_SO_index = sym_idx; 3102 } 3103 } 3104 break; 3105 3106 case N_OSO: 3107 // object file name: name,,0,0,st_mtime 3108 type = eSymbolTypeObjectFile; 3109 break; 3110 3111 case N_LSYM: 3112 // local sym: name,,NO_SECT,type,offset 3113 type = eSymbolTypeLocal; 3114 break; 3115 3116 // INCL scopes 3117 case N_BINCL: 3118 // include file beginning: name,,NO_SECT,0,sum We use 3119 // the current number of symbols in the symbol table 3120 // in lieu of using nlist_idx in case we ever start 3121 // trimming entries out 3122 N_INCL_indexes.push_back(sym_idx); 3123 type = eSymbolTypeScopeBegin; 3124 break; 3125 3126 case N_EINCL: 3127 // include file end: name,,NO_SECT,0,0 3128 // Set the size of the N_BINCL to the terminating 3129 // index of this N_EINCL so that we can always skip 3130 // the entire symbol if we need to navigate more 3131 // quickly at the source level when parsing STABS 3132 if (!N_INCL_indexes.empty()) { 3133 symbol_ptr = 3134 symtab.SymbolAtIndex(N_INCL_indexes.back()); 3135 symbol_ptr->SetByteSize(sym_idx + 1); 3136 symbol_ptr->SetSizeIsSibling(true); 3137 N_INCL_indexes.pop_back(); 3138 } 3139 type = eSymbolTypeScopeEnd; 3140 break; 3141 3142 case N_SOL: 3143 // #included file name: name,,n_sect,0,address 3144 type = eSymbolTypeHeaderFile; 3145 3146 // We currently don't use the header files on darwin 3147 add_nlist = false; 3148 break; 3149 3150 case N_PARAMS: 3151 // compiler parameters: name,,NO_SECT,0,0 3152 type = eSymbolTypeCompiler; 3153 break; 3154 3155 case N_VERSION: 3156 // compiler version: name,,NO_SECT,0,0 3157 type = eSymbolTypeCompiler; 3158 break; 3159 3160 case N_OLEVEL: 3161 // compiler -O level: name,,NO_SECT,0,0 3162 type = eSymbolTypeCompiler; 3163 break; 3164 3165 case N_PSYM: 3166 // parameter: name,,NO_SECT,type,offset 3167 type = eSymbolTypeVariable; 3168 break; 3169 3170 case N_ENTRY: 3171 // alternate entry: name,,n_sect,linenumber,address 3172 symbol_section = section_info.GetSection(nlist.n_sect, 3173 nlist.n_value); 3174 type = eSymbolTypeLineEntry; 3175 break; 3176 3177 // Left and Right Braces 3178 case N_LBRAC: 3179 // left bracket: 0,,NO_SECT,nesting level,address We 3180 // use the current number of symbols in the symbol 3181 // table in lieu of using nlist_idx in case we ever 3182 // start trimming entries out 3183 symbol_section = section_info.GetSection(nlist.n_sect, 3184 nlist.n_value); 3185 N_BRAC_indexes.push_back(sym_idx); 3186 type = eSymbolTypeScopeBegin; 3187 break; 3188 3189 case N_RBRAC: 3190 // right bracket: 0,,NO_SECT,nesting level,address 3191 // Set the size of the N_LBRAC to the terminating 3192 // index of this N_RBRAC so that we can always skip 3193 // the entire symbol if we need to navigate more 3194 // quickly at the source level when parsing STABS 3195 symbol_section = section_info.GetSection(nlist.n_sect, 3196 nlist.n_value); 3197 if (!N_BRAC_indexes.empty()) { 3198 symbol_ptr = 3199 symtab.SymbolAtIndex(N_BRAC_indexes.back()); 3200 symbol_ptr->SetByteSize(sym_idx + 1); 3201 symbol_ptr->SetSizeIsSibling(true); 3202 N_BRAC_indexes.pop_back(); 3203 } 3204 type = eSymbolTypeScopeEnd; 3205 break; 3206 3207 case N_EXCL: 3208 // deleted include file: name,,NO_SECT,0,sum 3209 type = eSymbolTypeHeaderFile; 3210 break; 3211 3212 // COMM scopes 3213 case N_BCOMM: 3214 // begin common: name,,NO_SECT,0,0 3215 // We use the current number of symbols in the symbol 3216 // table in lieu of using nlist_idx in case we ever 3217 // start trimming entries out 3218 type = eSymbolTypeScopeBegin; 3219 N_COMM_indexes.push_back(sym_idx); 3220 break; 3221 3222 case N_ECOML: 3223 // end common (local name): 0,,n_sect,0,address 3224 symbol_section = section_info.GetSection(nlist.n_sect, 3225 nlist.n_value); 3226 // Fall through 3227 3228 case N_ECOMM: 3229 // end common: name,,n_sect,0,0 3230 // Set the size of the N_BCOMM to the terminating 3231 // index of this N_ECOMM/N_ECOML so that we can 3232 // always skip the entire symbol if we need to 3233 // navigate more quickly at the source level when 3234 // parsing STABS 3235 if (!N_COMM_indexes.empty()) { 3236 symbol_ptr = 3237 symtab.SymbolAtIndex(N_COMM_indexes.back()); 3238 symbol_ptr->SetByteSize(sym_idx + 1); 3239 symbol_ptr->SetSizeIsSibling(true); 3240 N_COMM_indexes.pop_back(); 3241 } 3242 type = eSymbolTypeScopeEnd; 3243 break; 3244 3245 case N_LENG: 3246 // second stab entry with length information 3247 type = eSymbolTypeAdditional; 3248 break; 3249 3250 default: 3251 break; 3252 } 3253 } else { 3254 // uint8_t n_pext = N_PEXT & nlist.n_type; 3255 uint8_t n_type = N_TYPE & nlist.n_type; 3256 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3257 3258 switch (n_type) { 3259 case N_INDR: { 3260 const char *reexport_name_cstr = 3261 strtab_data.PeekCStr(nlist.n_value); 3262 if (reexport_name_cstr && reexport_name_cstr[0]) { 3263 type = eSymbolTypeReExported; 3264 ConstString reexport_name( 3265 reexport_name_cstr + 3266 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3267 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3268 set_value = false; 3269 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3270 indirect_symbol_names.insert(ConstString( 3271 symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3272 } else 3273 type = eSymbolTypeUndefined; 3274 } break; 3275 3276 case N_UNDF: 3277 if (symbol_name && symbol_name[0]) { 3278 ConstString undefined_name( 3279 symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3280 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3281 } 3282 // Fall through 3283 case N_PBUD: 3284 type = eSymbolTypeUndefined; 3285 break; 3286 3287 case N_ABS: 3288 type = eSymbolTypeAbsolute; 3289 break; 3290 3291 case N_SECT: { 3292 symbol_section = section_info.GetSection(nlist.n_sect, 3293 nlist.n_value); 3294 3295 if (symbol_section == NULL) { 3296 // TODO: warn about this? 3297 add_nlist = false; 3298 break; 3299 } 3300 3301 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3302 type = eSymbolTypeException; 3303 } else { 3304 uint32_t section_type = 3305 symbol_section->Get() & SECTION_TYPE; 3306 3307 switch (section_type) { 3308 case S_CSTRING_LITERALS: 3309 type = eSymbolTypeData; 3310 break; // section with only literal C strings 3311 case S_4BYTE_LITERALS: 3312 type = eSymbolTypeData; 3313 break; // section with only 4 byte literals 3314 case S_8BYTE_LITERALS: 3315 type = eSymbolTypeData; 3316 break; // section with only 8 byte literals 3317 case S_LITERAL_POINTERS: 3318 type = eSymbolTypeTrampoline; 3319 break; // section with only pointers to literals 3320 case S_NON_LAZY_SYMBOL_POINTERS: 3321 type = eSymbolTypeTrampoline; 3322 break; // section with only non-lazy symbol 3323 // pointers 3324 case S_LAZY_SYMBOL_POINTERS: 3325 type = eSymbolTypeTrampoline; 3326 break; // section with only lazy symbol pointers 3327 case S_SYMBOL_STUBS: 3328 type = eSymbolTypeTrampoline; 3329 break; // section with only symbol stubs, byte 3330 // size of stub in the reserved2 field 3331 case S_MOD_INIT_FUNC_POINTERS: 3332 type = eSymbolTypeCode; 3333 break; // section with only function pointers for 3334 // initialization 3335 case S_MOD_TERM_FUNC_POINTERS: 3336 type = eSymbolTypeCode; 3337 break; // section with only function pointers for 3338 // termination 3339 case S_INTERPOSING: 3340 type = eSymbolTypeTrampoline; 3341 break; // section with only pairs of function 3342 // pointers for interposing 3343 case S_16BYTE_LITERALS: 3344 type = eSymbolTypeData; 3345 break; // section with only 16 byte literals 3346 case S_DTRACE_DOF: 3347 type = eSymbolTypeInstrumentation; 3348 break; 3349 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3350 type = eSymbolTypeTrampoline; 3351 break; 3352 default: 3353 switch (symbol_section->GetType()) { 3354 case lldb::eSectionTypeCode: 3355 type = eSymbolTypeCode; 3356 break; 3357 case eSectionTypeData: 3358 case eSectionTypeDataCString: // Inlined C string 3359 // data 3360 case eSectionTypeDataCStringPointers: // Pointers 3361 // to C 3362 // string 3363 // data 3364 case eSectionTypeDataSymbolAddress: // Address of 3365 // a symbol in 3366 // the symbol 3367 // table 3368 case eSectionTypeData4: 3369 case eSectionTypeData8: 3370 case eSectionTypeData16: 3371 type = eSymbolTypeData; 3372 break; 3373 default: 3374 break; 3375 } 3376 break; 3377 } 3378 3379 if (type == eSymbolTypeInvalid) { 3380 const char *symbol_sect_name = 3381 symbol_section->GetName().AsCString(); 3382 if (symbol_section->IsDescendant( 3383 text_section_sp.get())) { 3384 if (symbol_section->IsClear( 3385 S_ATTR_PURE_INSTRUCTIONS | 3386 S_ATTR_SELF_MODIFYING_CODE | 3387 S_ATTR_SOME_INSTRUCTIONS)) 3388 type = eSymbolTypeData; 3389 else 3390 type = eSymbolTypeCode; 3391 } else if (symbol_section->IsDescendant( 3392 data_section_sp.get()) || 3393 symbol_section->IsDescendant( 3394 data_dirty_section_sp.get()) || 3395 symbol_section->IsDescendant( 3396 data_const_section_sp.get())) { 3397 if (symbol_sect_name && 3398 ::strstr(symbol_sect_name, "__objc") == 3399 symbol_sect_name) { 3400 type = eSymbolTypeRuntime; 3401 3402 if (symbol_name) { 3403 llvm::StringRef symbol_name_ref(symbol_name); 3404 if (symbol_name_ref.starts_with("_OBJC_")) { 3405 llvm::StringRef 3406 g_objc_v2_prefix_class( 3407 "_OBJC_CLASS_$_"); 3408 llvm::StringRef 3409 g_objc_v2_prefix_metaclass( 3410 "_OBJC_METACLASS_$_"); 3411 llvm::StringRef 3412 g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 3413 if (symbol_name_ref.starts_with( 3414 g_objc_v2_prefix_class)) { 3415 symbol_name_non_abi_mangled = 3416 symbol_name + 1; 3417 symbol_name = 3418 symbol_name + 3419 g_objc_v2_prefix_class.size(); 3420 type = eSymbolTypeObjCClass; 3421 demangled_is_synthesized = true; 3422 } else if ( 3423 symbol_name_ref.starts_with( 3424 g_objc_v2_prefix_metaclass)) { 3425 symbol_name_non_abi_mangled = 3426 symbol_name + 1; 3427 symbol_name = 3428 symbol_name + 3429 g_objc_v2_prefix_metaclass.size(); 3430 type = eSymbolTypeObjCMetaClass; 3431 demangled_is_synthesized = true; 3432 } else if (symbol_name_ref.starts_with( 3433 g_objc_v2_prefix_ivar)) { 3434 symbol_name_non_abi_mangled = 3435 symbol_name + 1; 3436 symbol_name = 3437 symbol_name + 3438 g_objc_v2_prefix_ivar.size(); 3439 type = eSymbolTypeObjCIVar; 3440 demangled_is_synthesized = true; 3441 } 3442 } 3443 } 3444 } else if (symbol_sect_name && 3445 ::strstr(symbol_sect_name, 3446 "__gcc_except_tab") == 3447 symbol_sect_name) { 3448 type = eSymbolTypeException; 3449 } else { 3450 type = eSymbolTypeData; 3451 } 3452 } else if (symbol_sect_name && 3453 ::strstr(symbol_sect_name, "__IMPORT") == 3454 symbol_sect_name) { 3455 type = eSymbolTypeTrampoline; 3456 } else if (symbol_section->IsDescendant( 3457 objc_section_sp.get())) { 3458 type = eSymbolTypeRuntime; 3459 if (symbol_name && symbol_name[0] == '.') { 3460 llvm::StringRef symbol_name_ref(symbol_name); 3461 llvm::StringRef 3462 g_objc_v1_prefix_class(".objc_class_name_"); 3463 if (symbol_name_ref.starts_with( 3464 g_objc_v1_prefix_class)) { 3465 symbol_name_non_abi_mangled = symbol_name; 3466 symbol_name = symbol_name + 3467 g_objc_v1_prefix_class.size(); 3468 type = eSymbolTypeObjCClass; 3469 demangled_is_synthesized = true; 3470 } 3471 } 3472 } 3473 } 3474 } 3475 } break; 3476 } 3477 } 3478 3479 if (add_nlist) { 3480 uint64_t symbol_value = nlist.n_value; 3481 if (symbol_name_non_abi_mangled) { 3482 sym[sym_idx].GetMangled().SetMangledName( 3483 ConstString(symbol_name_non_abi_mangled)); 3484 sym[sym_idx].GetMangled().SetDemangledName( 3485 ConstString(symbol_name)); 3486 } else { 3487 if (symbol_name && symbol_name[0] == '_') { 3488 symbol_name++; // Skip the leading underscore 3489 } 3490 3491 if (symbol_name) { 3492 ConstString const_symbol_name(symbol_name); 3493 sym[sym_idx].GetMangled().SetValue(const_symbol_name); 3494 if (is_gsym && is_debug) { 3495 const char *gsym_name = 3496 sym[sym_idx] 3497 .GetMangled() 3498 .GetName(Mangled::ePreferMangled) 3499 .GetCString(); 3500 if (gsym_name) 3501 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3502 } 3503 } 3504 } 3505 if (symbol_section) { 3506 const addr_t section_file_addr = 3507 symbol_section->GetFileAddress(); 3508 if (symbol_byte_size == 0 && 3509 function_starts_count > 0) { 3510 addr_t symbol_lookup_file_addr = nlist.n_value; 3511 // Do an exact address match for non-ARM addresses, 3512 // else get the closest since the symbol might be a 3513 // thumb symbol which has an address with bit zero 3514 // set 3515 FunctionStarts::Entry *func_start_entry = 3516 function_starts.FindEntry(symbol_lookup_file_addr, 3517 !is_arm); 3518 if (is_arm && func_start_entry) { 3519 // Verify that the function start address is the 3520 // symbol address (ARM) or the symbol address + 1 3521 // (thumb) 3522 if (func_start_entry->addr != 3523 symbol_lookup_file_addr && 3524 func_start_entry->addr != 3525 (symbol_lookup_file_addr + 1)) { 3526 // Not the right entry, NULL it out... 3527 func_start_entry = NULL; 3528 } 3529 } 3530 if (func_start_entry) { 3531 func_start_entry->data = true; 3532 3533 addr_t symbol_file_addr = func_start_entry->addr; 3534 uint32_t symbol_flags = 0; 3535 if (is_arm) { 3536 if (symbol_file_addr & 1) 3537 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3538 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3539 } 3540 3541 const FunctionStarts::Entry *next_func_start_entry = 3542 function_starts.FindNextEntry(func_start_entry); 3543 const addr_t section_end_file_addr = 3544 section_file_addr + 3545 symbol_section->GetByteSize(); 3546 if (next_func_start_entry) { 3547 addr_t next_symbol_file_addr = 3548 next_func_start_entry->addr; 3549 // Be sure the clear the Thumb address bit when 3550 // we calculate the size from the current and 3551 // next address 3552 if (is_arm) 3553 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3554 symbol_byte_size = std::min<lldb::addr_t>( 3555 next_symbol_file_addr - symbol_file_addr, 3556 section_end_file_addr - symbol_file_addr); 3557 } else { 3558 symbol_byte_size = 3559 section_end_file_addr - symbol_file_addr; 3560 } 3561 } 3562 } 3563 symbol_value -= section_file_addr; 3564 } 3565 3566 if (is_debug == false) { 3567 if (type == eSymbolTypeCode) { 3568 // See if we can find a N_FUN entry for any code 3569 // symbols. If we do find a match, and the name 3570 // matches, then we can merge the two into just the 3571 // function symbol to avoid duplicate entries in 3572 // the symbol table 3573 auto range = 3574 N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3575 if (range.first != range.second) { 3576 bool found_it = false; 3577 for (auto pos = range.first; pos != range.second; 3578 ++pos) { 3579 if (sym[sym_idx].GetMangled().GetName( 3580 Mangled::ePreferMangled) == 3581 sym[pos->second].GetMangled().GetName( 3582 Mangled::ePreferMangled)) { 3583 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3584 // We just need the flags from the linker 3585 // symbol, so put these flags 3586 // into the N_FUN flags to avoid duplicate 3587 // symbols in the symbol table 3588 sym[pos->second].SetExternal( 3589 sym[sym_idx].IsExternal()); 3590 sym[pos->second].SetFlags(nlist.n_type << 16 | 3591 nlist.n_desc); 3592 if (resolver_addresses.find(nlist.n_value) != 3593 resolver_addresses.end()) 3594 sym[pos->second].SetType(eSymbolTypeResolver); 3595 sym[sym_idx].Clear(); 3596 found_it = true; 3597 break; 3598 } 3599 } 3600 if (found_it) 3601 continue; 3602 } else { 3603 if (resolver_addresses.find(nlist.n_value) != 3604 resolver_addresses.end()) 3605 type = eSymbolTypeResolver; 3606 } 3607 } else if (type == eSymbolTypeData || 3608 type == eSymbolTypeObjCClass || 3609 type == eSymbolTypeObjCMetaClass || 3610 type == eSymbolTypeObjCIVar) { 3611 // See if we can find a N_STSYM entry for any data 3612 // symbols. If we do find a match, and the name 3613 // matches, then we can merge the two into just the 3614 // Static symbol to avoid duplicate entries in the 3615 // symbol table 3616 auto range = N_STSYM_addr_to_sym_idx.equal_range( 3617 nlist.n_value); 3618 if (range.first != range.second) { 3619 bool found_it = false; 3620 for (auto pos = range.first; pos != range.second; 3621 ++pos) { 3622 if (sym[sym_idx].GetMangled().GetName( 3623 Mangled::ePreferMangled) == 3624 sym[pos->second].GetMangled().GetName( 3625 Mangled::ePreferMangled)) { 3626 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3627 // We just need the flags from the linker 3628 // symbol, so put these flags 3629 // into the N_STSYM flags to avoid duplicate 3630 // symbols in the symbol table 3631 sym[pos->second].SetExternal( 3632 sym[sym_idx].IsExternal()); 3633 sym[pos->second].SetFlags(nlist.n_type << 16 | 3634 nlist.n_desc); 3635 sym[sym_idx].Clear(); 3636 found_it = true; 3637 break; 3638 } 3639 } 3640 if (found_it) 3641 continue; 3642 } else { 3643 const char *gsym_name = 3644 sym[sym_idx] 3645 .GetMangled() 3646 .GetName(Mangled::ePreferMangled) 3647 .GetCString(); 3648 if (gsym_name) { 3649 // Combine N_GSYM stab entries with the non 3650 // stab symbol 3651 ConstNameToSymbolIndexMap::const_iterator pos = 3652 N_GSYM_name_to_sym_idx.find(gsym_name); 3653 if (pos != N_GSYM_name_to_sym_idx.end()) { 3654 const uint32_t GSYM_sym_idx = pos->second; 3655 m_nlist_idx_to_sym_idx[nlist_idx] = 3656 GSYM_sym_idx; 3657 // Copy the address, because often the N_GSYM 3658 // address has an invalid address of zero 3659 // when the global is a common symbol 3660 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3661 symbol_section); 3662 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3663 symbol_value); 3664 add_symbol_addr(sym[GSYM_sym_idx] 3665 .GetAddress() 3666 .GetFileAddress()); 3667 // We just need the flags from the linker 3668 // symbol, so put these flags 3669 // into the N_GSYM flags to avoid duplicate 3670 // symbols in the symbol table 3671 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 3672 nlist.n_desc); 3673 sym[sym_idx].Clear(); 3674 continue; 3675 } 3676 } 3677 } 3678 } 3679 } 3680 3681 sym[sym_idx].SetID(nlist_idx); 3682 sym[sym_idx].SetType(type); 3683 if (set_value) { 3684 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 3685 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3686 add_symbol_addr( 3687 sym[sym_idx].GetAddress().GetFileAddress()); 3688 } 3689 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 3690 3691 if (symbol_byte_size > 0) 3692 sym[sym_idx].SetByteSize(symbol_byte_size); 3693 3694 if (demangled_is_synthesized) 3695 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3696 ++sym_idx; 3697 } else { 3698 sym[sym_idx].Clear(); 3699 } 3700 } 3701 ///////////////////////////// 3702 } 3703 } 3704 3705 for (const auto &pos : reexport_shlib_needs_fixup) { 3706 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3707 if (undef_pos != undefined_name_to_desc.end()) { 3708 const uint8_t dylib_ordinal = 3709 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3710 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3711 sym[pos.first].SetReExportedSymbolSharedLibrary( 3712 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3713 } 3714 } 3715 } 3716 3717 #endif 3718 lldb::offset_t nlist_data_offset = 0; 3719 3720 if (nlist_data.GetByteSize() > 0) { 3721 3722 // If the sym array was not created while parsing the DSC unmapped 3723 // symbols, create it now. 3724 if (sym == nullptr) { 3725 sym = 3726 symtab.Resize(symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3727 num_syms = symtab.GetNumSymbols(); 3728 } 3729 3730 if (unmapped_local_symbols_found) { 3731 assert(m_dysymtab.ilocalsym == 0); 3732 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3733 nlist_idx = m_dysymtab.nlocalsym; 3734 } else { 3735 nlist_idx = 0; 3736 } 3737 3738 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 3739 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 3740 UndefinedNameToDescMap undefined_name_to_desc; 3741 SymbolIndexToName reexport_shlib_needs_fixup; 3742 3743 // Symtab parsing is a huge mess. Everything is entangled and the code 3744 // requires access to a ridiculous amount of variables. LLDB depends 3745 // heavily on the proper merging of symbols and to get that right we need 3746 // to make sure we have parsed all the debug symbols first. Therefore we 3747 // invoke the lambda twice, once to parse only the debug symbols and then 3748 // once more to parse the remaining symbols. 3749 auto ParseSymbolLambda = [&](struct nlist_64 &nlist, uint32_t nlist_idx, 3750 bool debug_only) { 3751 const bool is_debug = ((nlist.n_type & N_STAB) != 0); 3752 if (is_debug != debug_only) 3753 return true; 3754 3755 const char *symbol_name_non_abi_mangled = nullptr; 3756 const char *symbol_name = nullptr; 3757 3758 if (have_strtab_data) { 3759 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3760 3761 if (symbol_name == nullptr) { 3762 // No symbol should be NULL, even the symbols with no string values 3763 // should have an offset zero which points to an empty C-string 3764 Debugger::ReportError(llvm::formatv( 3765 "symbol[{0}] has invalid string table offset {1:x} in {2}, " 3766 "ignoring symbol", 3767 nlist_idx, nlist.n_strx, module_sp->GetFileSpec().GetPath())); 3768 return true; 3769 } 3770 if (symbol_name[0] == '\0') 3771 symbol_name = nullptr; 3772 } else { 3773 const addr_t str_addr = strtab_addr + nlist.n_strx; 3774 Status str_error; 3775 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3776 str_error)) 3777 symbol_name = memory_symbol_name.c_str(); 3778 } 3779 3780 SymbolType type = eSymbolTypeInvalid; 3781 SectionSP symbol_section; 3782 lldb::addr_t symbol_byte_size = 0; 3783 bool add_nlist = true; 3784 bool is_gsym = false; 3785 bool demangled_is_synthesized = false; 3786 bool set_value = true; 3787 3788 assert(sym_idx < num_syms); 3789 sym[sym_idx].SetDebug(is_debug); 3790 3791 if (is_debug) { 3792 switch (nlist.n_type) { 3793 case N_GSYM: 3794 // global symbol: name,,NO_SECT,type,0 3795 // Sometimes the N_GSYM value contains the address. 3796 3797 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3798 // the ObjC data. They 3799 // have the same address, but we want to ensure that we always find 3800 // only the real symbol, 'cause we don't currently correctly 3801 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3802 // type. This is a temporary hack to make sure the ObjectiveC 3803 // symbols get treated correctly. To do this right, we should 3804 // coalesce all the GSYM & global symbols that have the same 3805 // address. 3806 is_gsym = true; 3807 sym[sym_idx].SetExternal(true); 3808 3809 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3810 llvm::StringRef symbol_name_ref(symbol_name); 3811 if (symbol_name_ref.starts_with(g_objc_v2_prefix_class)) { 3812 symbol_name_non_abi_mangled = symbol_name + 1; 3813 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3814 type = eSymbolTypeObjCClass; 3815 demangled_is_synthesized = true; 3816 3817 } else if (symbol_name_ref.starts_with( 3818 g_objc_v2_prefix_metaclass)) { 3819 symbol_name_non_abi_mangled = symbol_name + 1; 3820 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3821 type = eSymbolTypeObjCMetaClass; 3822 demangled_is_synthesized = true; 3823 } else if (symbol_name_ref.starts_with(g_objc_v2_prefix_ivar)) { 3824 symbol_name_non_abi_mangled = symbol_name + 1; 3825 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3826 type = eSymbolTypeObjCIVar; 3827 demangled_is_synthesized = true; 3828 } 3829 } else { 3830 if (nlist.n_value != 0) 3831 symbol_section = 3832 section_info.GetSection(nlist.n_sect, nlist.n_value); 3833 type = eSymbolTypeData; 3834 } 3835 break; 3836 3837 case N_FNAME: 3838 // procedure name (f77 kludge): name,,NO_SECT,0,0 3839 type = eSymbolTypeCompiler; 3840 break; 3841 3842 case N_FUN: 3843 // procedure: name,,n_sect,linenumber,address 3844 if (symbol_name) { 3845 type = eSymbolTypeCode; 3846 symbol_section = 3847 section_info.GetSection(nlist.n_sect, nlist.n_value); 3848 3849 N_FUN_addr_to_sym_idx.insert( 3850 std::make_pair(nlist.n_value, sym_idx)); 3851 // We use the current number of symbols in the symbol table in 3852 // lieu of using nlist_idx in case we ever start trimming entries 3853 // out 3854 N_FUN_indexes.push_back(sym_idx); 3855 } else { 3856 type = eSymbolTypeCompiler; 3857 3858 if (!N_FUN_indexes.empty()) { 3859 // Copy the size of the function into the original STAB entry 3860 // so we don't have to hunt for it later 3861 symtab.SymbolAtIndex(N_FUN_indexes.back()) 3862 ->SetByteSize(nlist.n_value); 3863 N_FUN_indexes.pop_back(); 3864 // We don't really need the end function STAB as it contains 3865 // the size which we already placed with the original symbol, 3866 // so don't add it if we want a minimal symbol table 3867 add_nlist = false; 3868 } 3869 } 3870 break; 3871 3872 case N_STSYM: 3873 // static symbol: name,,n_sect,type,address 3874 N_STSYM_addr_to_sym_idx.insert( 3875 std::make_pair(nlist.n_value, sym_idx)); 3876 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3877 if (symbol_name && symbol_name[0]) { 3878 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3879 eSymbolTypeData); 3880 } 3881 break; 3882 3883 case N_LCSYM: 3884 // .lcomm symbol: name,,n_sect,type,address 3885 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3886 type = eSymbolTypeCommonBlock; 3887 break; 3888 3889 case N_BNSYM: 3890 // We use the current number of symbols in the symbol table in lieu 3891 // of using nlist_idx in case we ever start trimming entries out 3892 // Skip these if we want minimal symbol tables 3893 add_nlist = false; 3894 break; 3895 3896 case N_ENSYM: 3897 // Set the size of the N_BNSYM to the terminating index of this 3898 // N_ENSYM so that we can always skip the entire symbol if we need 3899 // to navigate more quickly at the source level when parsing STABS 3900 // Skip these if we want minimal symbol tables 3901 add_nlist = false; 3902 break; 3903 3904 case N_OPT: 3905 // emitted with gcc2_compiled and in gcc source 3906 type = eSymbolTypeCompiler; 3907 break; 3908 3909 case N_RSYM: 3910 // register sym: name,,NO_SECT,type,register 3911 type = eSymbolTypeVariable; 3912 break; 3913 3914 case N_SLINE: 3915 // src line: 0,,n_sect,linenumber,address 3916 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3917 type = eSymbolTypeLineEntry; 3918 break; 3919 3920 case N_SSYM: 3921 // structure elt: name,,NO_SECT,type,struct_offset 3922 type = eSymbolTypeVariableType; 3923 break; 3924 3925 case N_SO: 3926 // source file name 3927 type = eSymbolTypeSourceFile; 3928 if (symbol_name == nullptr) { 3929 add_nlist = false; 3930 if (N_SO_index != UINT32_MAX) { 3931 // Set the size of the N_SO to the terminating index of this 3932 // N_SO so that we can always skip the entire N_SO if we need 3933 // to navigate more quickly at the source level when parsing 3934 // STABS 3935 symbol_ptr = symtab.SymbolAtIndex(N_SO_index); 3936 symbol_ptr->SetByteSize(sym_idx); 3937 symbol_ptr->SetSizeIsSibling(true); 3938 } 3939 N_NSYM_indexes.clear(); 3940 N_INCL_indexes.clear(); 3941 N_BRAC_indexes.clear(); 3942 N_COMM_indexes.clear(); 3943 N_FUN_indexes.clear(); 3944 N_SO_index = UINT32_MAX; 3945 } else { 3946 // We use the current number of symbols in the symbol table in 3947 // lieu of using nlist_idx in case we ever start trimming entries 3948 // out 3949 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3950 if (N_SO_has_full_path) { 3951 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3952 // We have two consecutive N_SO entries where the first 3953 // contains a directory and the second contains a full path. 3954 sym[sym_idx - 1].GetMangled().SetValue( 3955 ConstString(symbol_name)); 3956 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3957 add_nlist = false; 3958 } else { 3959 // This is the first entry in a N_SO that contains a 3960 // directory or a full path to the source file 3961 N_SO_index = sym_idx; 3962 } 3963 } else if ((N_SO_index == sym_idx - 1) && 3964 ((sym_idx - 1) < num_syms)) { 3965 // This is usually the second N_SO entry that contains just the 3966 // filename, so here we combine it with the first one if we are 3967 // minimizing the symbol table 3968 const char *so_path = 3969 sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3970 if (so_path && so_path[0]) { 3971 std::string full_so_path(so_path); 3972 const size_t double_slash_pos = full_so_path.find("//"); 3973 if (double_slash_pos != std::string::npos) { 3974 // The linker has been generating bad N_SO entries with 3975 // doubled up paths in the format "%s%s" where the first 3976 // string in the DW_AT_comp_dir, and the second is the 3977 // directory for the source file so you end up with a path 3978 // that looks like "/tmp/src//tmp/src/" 3979 FileSpec so_dir(so_path); 3980 if (!FileSystem::Instance().Exists(so_dir)) { 3981 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 3982 FileSpec::Style::native); 3983 if (FileSystem::Instance().Exists(so_dir)) { 3984 // Trim off the incorrect path 3985 full_so_path.erase(0, double_slash_pos + 1); 3986 } 3987 } 3988 } 3989 if (*full_so_path.rbegin() != '/') 3990 full_so_path += '/'; 3991 full_so_path += symbol_name; 3992 sym[sym_idx - 1].GetMangled().SetValue( 3993 ConstString(full_so_path.c_str())); 3994 add_nlist = false; 3995 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3996 } 3997 } else { 3998 // This could be a relative path to a N_SO 3999 N_SO_index = sym_idx; 4000 } 4001 } 4002 break; 4003 4004 case N_OSO: 4005 // object file name: name,,0,0,st_mtime 4006 type = eSymbolTypeObjectFile; 4007 break; 4008 4009 case N_LSYM: 4010 // local sym: name,,NO_SECT,type,offset 4011 type = eSymbolTypeLocal; 4012 break; 4013 4014 // INCL scopes 4015 case N_BINCL: 4016 // include file beginning: name,,NO_SECT,0,sum We use the current 4017 // number of symbols in the symbol table in lieu of using nlist_idx 4018 // in case we ever start trimming entries out 4019 N_INCL_indexes.push_back(sym_idx); 4020 type = eSymbolTypeScopeBegin; 4021 break; 4022 4023 case N_EINCL: 4024 // include file end: name,,NO_SECT,0,0 4025 // Set the size of the N_BINCL to the terminating index of this 4026 // N_EINCL so that we can always skip the entire symbol if we need 4027 // to navigate more quickly at the source level when parsing STABS 4028 if (!N_INCL_indexes.empty()) { 4029 symbol_ptr = symtab.SymbolAtIndex(N_INCL_indexes.back()); 4030 symbol_ptr->SetByteSize(sym_idx + 1); 4031 symbol_ptr->SetSizeIsSibling(true); 4032 N_INCL_indexes.pop_back(); 4033 } 4034 type = eSymbolTypeScopeEnd; 4035 break; 4036 4037 case N_SOL: 4038 // #included file name: name,,n_sect,0,address 4039 type = eSymbolTypeHeaderFile; 4040 4041 // We currently don't use the header files on darwin 4042 add_nlist = false; 4043 break; 4044 4045 case N_PARAMS: 4046 // compiler parameters: name,,NO_SECT,0,0 4047 type = eSymbolTypeCompiler; 4048 break; 4049 4050 case N_VERSION: 4051 // compiler version: name,,NO_SECT,0,0 4052 type = eSymbolTypeCompiler; 4053 break; 4054 4055 case N_OLEVEL: 4056 // compiler -O level: name,,NO_SECT,0,0 4057 type = eSymbolTypeCompiler; 4058 break; 4059 4060 case N_PSYM: 4061 // parameter: name,,NO_SECT,type,offset 4062 type = eSymbolTypeVariable; 4063 break; 4064 4065 case N_ENTRY: 4066 // alternate entry: name,,n_sect,linenumber,address 4067 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4068 type = eSymbolTypeLineEntry; 4069 break; 4070 4071 // Left and Right Braces 4072 case N_LBRAC: 4073 // left bracket: 0,,NO_SECT,nesting level,address We use the 4074 // current number of symbols in the symbol table in lieu of using 4075 // nlist_idx in case we ever start trimming entries out 4076 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4077 N_BRAC_indexes.push_back(sym_idx); 4078 type = eSymbolTypeScopeBegin; 4079 break; 4080 4081 case N_RBRAC: 4082 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4083 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4084 // can always skip the entire symbol if we need to navigate more 4085 // quickly at the source level when parsing STABS 4086 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4087 if (!N_BRAC_indexes.empty()) { 4088 symbol_ptr = symtab.SymbolAtIndex(N_BRAC_indexes.back()); 4089 symbol_ptr->SetByteSize(sym_idx + 1); 4090 symbol_ptr->SetSizeIsSibling(true); 4091 N_BRAC_indexes.pop_back(); 4092 } 4093 type = eSymbolTypeScopeEnd; 4094 break; 4095 4096 case N_EXCL: 4097 // deleted include file: name,,NO_SECT,0,sum 4098 type = eSymbolTypeHeaderFile; 4099 break; 4100 4101 // COMM scopes 4102 case N_BCOMM: 4103 // begin common: name,,NO_SECT,0,0 4104 // We use the current number of symbols in the symbol table in lieu 4105 // of using nlist_idx in case we ever start trimming entries out 4106 type = eSymbolTypeScopeBegin; 4107 N_COMM_indexes.push_back(sym_idx); 4108 break; 4109 4110 case N_ECOML: 4111 // end common (local name): 0,,n_sect,0,address 4112 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4113 [[fallthrough]]; 4114 4115 case N_ECOMM: 4116 // end common: name,,n_sect,0,0 4117 // Set the size of the N_BCOMM to the terminating index of this 4118 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4119 // we need to navigate more quickly at the source level when 4120 // parsing STABS 4121 if (!N_COMM_indexes.empty()) { 4122 symbol_ptr = symtab.SymbolAtIndex(N_COMM_indexes.back()); 4123 symbol_ptr->SetByteSize(sym_idx + 1); 4124 symbol_ptr->SetSizeIsSibling(true); 4125 N_COMM_indexes.pop_back(); 4126 } 4127 type = eSymbolTypeScopeEnd; 4128 break; 4129 4130 case N_LENG: 4131 // second stab entry with length information 4132 type = eSymbolTypeAdditional; 4133 break; 4134 4135 default: 4136 break; 4137 } 4138 } else { 4139 uint8_t n_type = N_TYPE & nlist.n_type; 4140 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4141 4142 switch (n_type) { 4143 case N_INDR: { 4144 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4145 if (reexport_name_cstr && reexport_name_cstr[0] && symbol_name) { 4146 type = eSymbolTypeReExported; 4147 ConstString reexport_name(reexport_name_cstr + 4148 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4149 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4150 set_value = false; 4151 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4152 indirect_symbol_names.insert( 4153 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4154 } else 4155 type = eSymbolTypeUndefined; 4156 } break; 4157 4158 case N_UNDF: 4159 if (symbol_name && symbol_name[0]) { 4160 ConstString undefined_name(symbol_name + 4161 ((symbol_name[0] == '_') ? 1 : 0)); 4162 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4163 } 4164 [[fallthrough]]; 4165 4166 case N_PBUD: 4167 type = eSymbolTypeUndefined; 4168 break; 4169 4170 case N_ABS: 4171 type = eSymbolTypeAbsolute; 4172 break; 4173 4174 case N_SECT: { 4175 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4176 4177 if (!symbol_section) { 4178 // TODO: warn about this? 4179 add_nlist = false; 4180 break; 4181 } 4182 4183 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4184 type = eSymbolTypeException; 4185 } else { 4186 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4187 4188 switch (section_type) { 4189 case S_CSTRING_LITERALS: 4190 type = eSymbolTypeData; 4191 break; // section with only literal C strings 4192 case S_4BYTE_LITERALS: 4193 type = eSymbolTypeData; 4194 break; // section with only 4 byte literals 4195 case S_8BYTE_LITERALS: 4196 type = eSymbolTypeData; 4197 break; // section with only 8 byte literals 4198 case S_LITERAL_POINTERS: 4199 type = eSymbolTypeTrampoline; 4200 break; // section with only pointers to literals 4201 case S_NON_LAZY_SYMBOL_POINTERS: 4202 type = eSymbolTypeTrampoline; 4203 break; // section with only non-lazy symbol pointers 4204 case S_LAZY_SYMBOL_POINTERS: 4205 type = eSymbolTypeTrampoline; 4206 break; // section with only lazy symbol pointers 4207 case S_SYMBOL_STUBS: 4208 type = eSymbolTypeTrampoline; 4209 break; // section with only symbol stubs, byte size of stub in 4210 // the reserved2 field 4211 case S_MOD_INIT_FUNC_POINTERS: 4212 type = eSymbolTypeCode; 4213 break; // section with only function pointers for initialization 4214 case S_MOD_TERM_FUNC_POINTERS: 4215 type = eSymbolTypeCode; 4216 break; // section with only function pointers for termination 4217 case S_INTERPOSING: 4218 type = eSymbolTypeTrampoline; 4219 break; // section with only pairs of function pointers for 4220 // interposing 4221 case S_16BYTE_LITERALS: 4222 type = eSymbolTypeData; 4223 break; // section with only 16 byte literals 4224 case S_DTRACE_DOF: 4225 type = eSymbolTypeInstrumentation; 4226 break; 4227 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4228 type = eSymbolTypeTrampoline; 4229 break; 4230 default: 4231 switch (symbol_section->GetType()) { 4232 case lldb::eSectionTypeCode: 4233 type = eSymbolTypeCode; 4234 break; 4235 case eSectionTypeData: 4236 case eSectionTypeDataCString: // Inlined C string data 4237 case eSectionTypeDataCStringPointers: // Pointers to C string 4238 // data 4239 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4240 // the symbol table 4241 case eSectionTypeData4: 4242 case eSectionTypeData8: 4243 case eSectionTypeData16: 4244 type = eSymbolTypeData; 4245 break; 4246 default: 4247 break; 4248 } 4249 break; 4250 } 4251 4252 if (type == eSymbolTypeInvalid) { 4253 const char *symbol_sect_name = 4254 symbol_section->GetName().AsCString(); 4255 if (symbol_section->IsDescendant(text_section_sp.get())) { 4256 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4257 S_ATTR_SELF_MODIFYING_CODE | 4258 S_ATTR_SOME_INSTRUCTIONS)) 4259 type = eSymbolTypeData; 4260 else 4261 type = eSymbolTypeCode; 4262 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 4263 symbol_section->IsDescendant( 4264 data_dirty_section_sp.get()) || 4265 symbol_section->IsDescendant( 4266 data_const_section_sp.get())) { 4267 if (symbol_sect_name && 4268 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 4269 type = eSymbolTypeRuntime; 4270 4271 if (symbol_name) { 4272 llvm::StringRef symbol_name_ref(symbol_name); 4273 if (symbol_name_ref.starts_with("_OBJC_")) { 4274 llvm::StringRef g_objc_v2_prefix_class( 4275 "_OBJC_CLASS_$_"); 4276 llvm::StringRef g_objc_v2_prefix_metaclass( 4277 "_OBJC_METACLASS_$_"); 4278 llvm::StringRef g_objc_v2_prefix_ivar( 4279 "_OBJC_IVAR_$_"); 4280 if (symbol_name_ref.starts_with(g_objc_v2_prefix_class)) { 4281 symbol_name_non_abi_mangled = symbol_name + 1; 4282 symbol_name = 4283 symbol_name + g_objc_v2_prefix_class.size(); 4284 type = eSymbolTypeObjCClass; 4285 demangled_is_synthesized = true; 4286 } else if (symbol_name_ref.starts_with( 4287 g_objc_v2_prefix_metaclass)) { 4288 symbol_name_non_abi_mangled = symbol_name + 1; 4289 symbol_name = 4290 symbol_name + g_objc_v2_prefix_metaclass.size(); 4291 type = eSymbolTypeObjCMetaClass; 4292 demangled_is_synthesized = true; 4293 } else if (symbol_name_ref.starts_with( 4294 g_objc_v2_prefix_ivar)) { 4295 symbol_name_non_abi_mangled = symbol_name + 1; 4296 symbol_name = 4297 symbol_name + g_objc_v2_prefix_ivar.size(); 4298 type = eSymbolTypeObjCIVar; 4299 demangled_is_synthesized = true; 4300 } 4301 } 4302 } 4303 } else if (symbol_sect_name && 4304 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4305 symbol_sect_name) { 4306 type = eSymbolTypeException; 4307 } else { 4308 type = eSymbolTypeData; 4309 } 4310 } else if (symbol_sect_name && 4311 ::strstr(symbol_sect_name, "__IMPORT") == 4312 symbol_sect_name) { 4313 type = eSymbolTypeTrampoline; 4314 } else if (symbol_section->IsDescendant(objc_section_sp.get())) { 4315 type = eSymbolTypeRuntime; 4316 if (symbol_name && symbol_name[0] == '.') { 4317 llvm::StringRef symbol_name_ref(symbol_name); 4318 llvm::StringRef g_objc_v1_prefix_class( 4319 ".objc_class_name_"); 4320 if (symbol_name_ref.starts_with(g_objc_v1_prefix_class)) { 4321 symbol_name_non_abi_mangled = symbol_name; 4322 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4323 type = eSymbolTypeObjCClass; 4324 demangled_is_synthesized = true; 4325 } 4326 } 4327 } 4328 } 4329 } 4330 } break; 4331 } 4332 } 4333 4334 if (!add_nlist) { 4335 sym[sym_idx].Clear(); 4336 return true; 4337 } 4338 4339 uint64_t symbol_value = nlist.n_value; 4340 4341 if (symbol_name_non_abi_mangled) { 4342 sym[sym_idx].GetMangled().SetMangledName( 4343 ConstString(symbol_name_non_abi_mangled)); 4344 sym[sym_idx].GetMangled().SetDemangledName(ConstString(symbol_name)); 4345 } else { 4346 4347 if (symbol_name && symbol_name[0] == '_') { 4348 symbol_name++; // Skip the leading underscore 4349 } 4350 4351 if (symbol_name) { 4352 ConstString const_symbol_name(symbol_name); 4353 sym[sym_idx].GetMangled().SetValue(const_symbol_name); 4354 } 4355 } 4356 4357 if (is_gsym) { 4358 const char *gsym_name = sym[sym_idx] 4359 .GetMangled() 4360 .GetName(Mangled::ePreferMangled) 4361 .GetCString(); 4362 if (gsym_name) 4363 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4364 } 4365 4366 if (symbol_section) { 4367 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4368 if (symbol_byte_size == 0 && function_starts_count > 0) { 4369 addr_t symbol_lookup_file_addr = nlist.n_value; 4370 // Do an exact address match for non-ARM addresses, else get the 4371 // closest since the symbol might be a thumb symbol which has an 4372 // address with bit zero set. 4373 FunctionStarts::Entry *func_start_entry = 4374 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4375 if (is_arm && func_start_entry) { 4376 // Verify that the function start address is the symbol address 4377 // (ARM) or the symbol address + 1 (thumb). 4378 if (func_start_entry->addr != symbol_lookup_file_addr && 4379 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4380 // Not the right entry, NULL it out... 4381 func_start_entry = nullptr; 4382 } 4383 } 4384 if (func_start_entry) { 4385 func_start_entry->data = true; 4386 4387 addr_t symbol_file_addr = func_start_entry->addr; 4388 if (is_arm) 4389 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4390 4391 const FunctionStarts::Entry *next_func_start_entry = 4392 function_starts.FindNextEntry(func_start_entry); 4393 const addr_t section_end_file_addr = 4394 section_file_addr + symbol_section->GetByteSize(); 4395 if (next_func_start_entry) { 4396 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4397 // Be sure the clear the Thumb address bit when we calculate the 4398 // size from the current and next address 4399 if (is_arm) 4400 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4401 symbol_byte_size = std::min<lldb::addr_t>( 4402 next_symbol_file_addr - symbol_file_addr, 4403 section_end_file_addr - symbol_file_addr); 4404 } else { 4405 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4406 } 4407 } 4408 } 4409 symbol_value -= section_file_addr; 4410 } 4411 4412 if (!is_debug) { 4413 if (type == eSymbolTypeCode) { 4414 // See if we can find a N_FUN entry for any code symbols. If we do 4415 // find a match, and the name matches, then we can merge the two into 4416 // just the function symbol to avoid duplicate entries in the symbol 4417 // table. 4418 std::pair<ValueToSymbolIndexMap::const_iterator, 4419 ValueToSymbolIndexMap::const_iterator> 4420 range; 4421 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4422 if (range.first != range.second) { 4423 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4424 pos != range.second; ++pos) { 4425 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4426 sym[pos->second].GetMangled().GetName( 4427 Mangled::ePreferMangled)) { 4428 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4429 // We just need the flags from the linker symbol, so put these 4430 // flags into the N_FUN flags to avoid duplicate symbols in the 4431 // symbol table. 4432 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4433 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4434 if (resolver_addresses.find(nlist.n_value) != 4435 resolver_addresses.end()) 4436 sym[pos->second].SetType(eSymbolTypeResolver); 4437 sym[sym_idx].Clear(); 4438 return true; 4439 } 4440 } 4441 } else { 4442 if (resolver_addresses.find(nlist.n_value) != 4443 resolver_addresses.end()) 4444 type = eSymbolTypeResolver; 4445 } 4446 } else if (type == eSymbolTypeData || type == eSymbolTypeObjCClass || 4447 type == eSymbolTypeObjCMetaClass || 4448 type == eSymbolTypeObjCIVar) { 4449 // See if we can find a N_STSYM entry for any data symbols. If we do 4450 // find a match, and the name matches, then we can merge the two into 4451 // just the Static symbol to avoid duplicate entries in the symbol 4452 // table. 4453 std::pair<ValueToSymbolIndexMap::const_iterator, 4454 ValueToSymbolIndexMap::const_iterator> 4455 range; 4456 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4457 if (range.first != range.second) { 4458 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4459 pos != range.second; ++pos) { 4460 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4461 sym[pos->second].GetMangled().GetName( 4462 Mangled::ePreferMangled)) { 4463 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4464 // We just need the flags from the linker symbol, so put these 4465 // flags into the N_STSYM flags to avoid duplicate symbols in 4466 // the symbol table. 4467 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4468 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4469 sym[sym_idx].Clear(); 4470 return true; 4471 } 4472 } 4473 } else { 4474 // Combine N_GSYM stab entries with the non stab symbol. 4475 const char *gsym_name = sym[sym_idx] 4476 .GetMangled() 4477 .GetName(Mangled::ePreferMangled) 4478 .GetCString(); 4479 if (gsym_name) { 4480 ConstNameToSymbolIndexMap::const_iterator pos = 4481 N_GSYM_name_to_sym_idx.find(gsym_name); 4482 if (pos != N_GSYM_name_to_sym_idx.end()) { 4483 const uint32_t GSYM_sym_idx = pos->second; 4484 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4485 // Copy the address, because often the N_GSYM address has an 4486 // invalid address of zero when the global is a common symbol. 4487 sym[GSYM_sym_idx].GetAddressRef().SetSection(symbol_section); 4488 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4489 add_symbol_addr( 4490 sym[GSYM_sym_idx].GetAddress().GetFileAddress()); 4491 // We just need the flags from the linker symbol, so put these 4492 // flags into the N_GSYM flags to avoid duplicate symbols in 4493 // the symbol table. 4494 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4495 sym[sym_idx].Clear(); 4496 return true; 4497 } 4498 } 4499 } 4500 } 4501 } 4502 4503 sym[sym_idx].SetID(nlist_idx); 4504 sym[sym_idx].SetType(type); 4505 if (set_value) { 4506 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4507 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4508 if (symbol_section) 4509 add_symbol_addr(sym[sym_idx].GetAddress().GetFileAddress()); 4510 } 4511 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4512 if (nlist.n_desc & N_WEAK_REF) 4513 sym[sym_idx].SetIsWeak(true); 4514 4515 if (symbol_byte_size > 0) 4516 sym[sym_idx].SetByteSize(symbol_byte_size); 4517 4518 if (demangled_is_synthesized) 4519 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4520 4521 ++sym_idx; 4522 return true; 4523 }; 4524 4525 // First parse all the nlists but don't process them yet. See the next 4526 // comment for an explanation why. 4527 std::vector<struct nlist_64> nlists; 4528 nlists.reserve(symtab_load_command.nsyms); 4529 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 4530 if (auto nlist = 4531 ParseNList(nlist_data, nlist_data_offset, nlist_byte_size)) 4532 nlists.push_back(*nlist); 4533 else 4534 break; 4535 } 4536 4537 // Now parse all the debug symbols. This is needed to merge non-debug 4538 // symbols in the next step. Non-debug symbols are always coalesced into 4539 // the debug symbol. Doing this in one step would mean that some symbols 4540 // won't be merged. 4541 nlist_idx = 0; 4542 for (auto &nlist : nlists) { 4543 if (!ParseSymbolLambda(nlist, nlist_idx++, DebugSymbols)) 4544 break; 4545 } 4546 4547 // Finally parse all the non debug symbols. 4548 nlist_idx = 0; 4549 for (auto &nlist : nlists) { 4550 if (!ParseSymbolLambda(nlist, nlist_idx++, NonDebugSymbols)) 4551 break; 4552 } 4553 4554 for (const auto &pos : reexport_shlib_needs_fixup) { 4555 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4556 if (undef_pos != undefined_name_to_desc.end()) { 4557 const uint8_t dylib_ordinal = 4558 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4559 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4560 sym[pos.first].SetReExportedSymbolSharedLibrary( 4561 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4562 } 4563 } 4564 } 4565 4566 // Count how many trie symbols we'll add to the symbol table 4567 int trie_symbol_table_augment_count = 0; 4568 for (auto &e : external_sym_trie_entries) { 4569 if (!symbols_added.contains(e.entry.address)) 4570 trie_symbol_table_augment_count++; 4571 } 4572 4573 if (num_syms < sym_idx + trie_symbol_table_augment_count) { 4574 num_syms = sym_idx + trie_symbol_table_augment_count; 4575 sym = symtab.Resize(num_syms); 4576 } 4577 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4578 4579 // Add symbols from the trie to the symbol table. 4580 for (auto &e : external_sym_trie_entries) { 4581 if (symbols_added.contains(e.entry.address)) 4582 continue; 4583 4584 // Find the section that this trie address is in, use that to annotate 4585 // symbol type as we add the trie address and name to the symbol table. 4586 Address symbol_addr; 4587 if (module_sp->ResolveFileAddress(e.entry.address, symbol_addr)) { 4588 SectionSP symbol_section(symbol_addr.GetSection()); 4589 const char *symbol_name = e.entry.name.GetCString(); 4590 bool demangled_is_synthesized = false; 4591 SymbolType type = 4592 GetSymbolType(symbol_name, demangled_is_synthesized, text_section_sp, 4593 data_section_sp, data_dirty_section_sp, 4594 data_const_section_sp, symbol_section); 4595 4596 sym[sym_idx].SetType(type); 4597 if (symbol_section) { 4598 sym[sym_idx].SetID(synthetic_sym_id++); 4599 sym[sym_idx].GetMangled().SetMangledName(ConstString(symbol_name)); 4600 if (demangled_is_synthesized) 4601 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4602 sym[sym_idx].SetIsSynthetic(true); 4603 sym[sym_idx].SetExternal(true); 4604 sym[sym_idx].GetAddressRef() = symbol_addr; 4605 add_symbol_addr(symbol_addr.GetFileAddress()); 4606 if (e.entry.flags & TRIE_SYMBOL_IS_THUMB) 4607 sym[sym_idx].SetFlags(MACHO_NLIST_ARM_SYMBOL_IS_THUMB); 4608 ++sym_idx; 4609 } 4610 } 4611 } 4612 4613 if (function_starts_count > 0) { 4614 uint32_t num_synthetic_function_symbols = 0; 4615 for (i = 0; i < function_starts_count; ++i) { 4616 if (!symbols_added.contains(function_starts.GetEntryRef(i).addr)) 4617 ++num_synthetic_function_symbols; 4618 } 4619 4620 if (num_synthetic_function_symbols > 0) { 4621 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4622 num_syms = sym_idx + num_synthetic_function_symbols; 4623 sym = symtab.Resize(num_syms); 4624 } 4625 for (i = 0; i < function_starts_count; ++i) { 4626 const FunctionStarts::Entry *func_start_entry = 4627 function_starts.GetEntryAtIndex(i); 4628 if (!symbols_added.contains(func_start_entry->addr)) { 4629 addr_t symbol_file_addr = func_start_entry->addr; 4630 uint32_t symbol_flags = 0; 4631 if (func_start_entry->data) 4632 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4633 Address symbol_addr; 4634 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4635 SectionSP symbol_section(symbol_addr.GetSection()); 4636 uint32_t symbol_byte_size = 0; 4637 if (symbol_section) { 4638 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4639 const FunctionStarts::Entry *next_func_start_entry = 4640 function_starts.FindNextEntry(func_start_entry); 4641 const addr_t section_end_file_addr = 4642 section_file_addr + symbol_section->GetByteSize(); 4643 if (next_func_start_entry) { 4644 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4645 if (is_arm) 4646 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4647 symbol_byte_size = std::min<lldb::addr_t>( 4648 next_symbol_file_addr - symbol_file_addr, 4649 section_end_file_addr - symbol_file_addr); 4650 } else { 4651 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4652 } 4653 sym[sym_idx].SetID(synthetic_sym_id++); 4654 // Don't set the name for any synthetic symbols, the Symbol 4655 // object will generate one if needed when the name is accessed 4656 // via accessors. 4657 sym[sym_idx].GetMangled().SetDemangledName(ConstString()); 4658 sym[sym_idx].SetType(eSymbolTypeCode); 4659 sym[sym_idx].SetIsSynthetic(true); 4660 sym[sym_idx].GetAddressRef() = symbol_addr; 4661 add_symbol_addr(symbol_addr.GetFileAddress()); 4662 if (symbol_flags) 4663 sym[sym_idx].SetFlags(symbol_flags); 4664 if (symbol_byte_size) 4665 sym[sym_idx].SetByteSize(symbol_byte_size); 4666 ++sym_idx; 4667 } 4668 } 4669 } 4670 } 4671 } 4672 } 4673 4674 // Trim our symbols down to just what we ended up with after removing any 4675 // symbols. 4676 if (sym_idx < num_syms) { 4677 num_syms = sym_idx; 4678 sym = symtab.Resize(num_syms); 4679 } 4680 4681 // Now synthesize indirect symbols 4682 if (m_dysymtab.nindirectsyms != 0) { 4683 if (indirect_symbol_index_data.GetByteSize()) { 4684 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4685 m_nlist_idx_to_sym_idx.end(); 4686 4687 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4688 ++sect_idx) { 4689 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4690 S_SYMBOL_STUBS) { 4691 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4692 if (symbol_stub_byte_size == 0) 4693 continue; 4694 4695 const uint32_t num_symbol_stubs = 4696 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4697 4698 if (num_symbol_stubs == 0) 4699 continue; 4700 4701 const uint32_t symbol_stub_index_offset = 4702 m_mach_sections[sect_idx].reserved1; 4703 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) { 4704 const uint32_t symbol_stub_index = 4705 symbol_stub_index_offset + stub_idx; 4706 const lldb::addr_t symbol_stub_addr = 4707 m_mach_sections[sect_idx].addr + 4708 (stub_idx * symbol_stub_byte_size); 4709 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4710 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4711 symbol_stub_offset, 4)) { 4712 const uint32_t stub_sym_id = 4713 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4714 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4715 continue; 4716 4717 NListIndexToSymbolIndexMap::const_iterator index_pos = 4718 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4719 Symbol *stub_symbol = nullptr; 4720 if (index_pos != end_index_pos) { 4721 // We have a remapping from the original nlist index to a 4722 // current symbol index, so just look this up by index 4723 stub_symbol = symtab.SymbolAtIndex(index_pos->second); 4724 } else { 4725 // We need to lookup a symbol using the original nlist symbol 4726 // index since this index is coming from the S_SYMBOL_STUBS 4727 stub_symbol = symtab.FindSymbolByID(stub_sym_id); 4728 } 4729 4730 if (stub_symbol) { 4731 Address so_addr(symbol_stub_addr, section_list); 4732 4733 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4734 // Change the external symbol into a trampoline that makes 4735 // sense These symbols were N_UNDF N_EXT, and are useless 4736 // to us, so we can re-use them so we don't have to make up 4737 // a synthetic symbol for no good reason. 4738 if (resolver_addresses.find(symbol_stub_addr) == 4739 resolver_addresses.end()) 4740 stub_symbol->SetType(eSymbolTypeTrampoline); 4741 else 4742 stub_symbol->SetType(eSymbolTypeResolver); 4743 stub_symbol->SetExternal(false); 4744 stub_symbol->GetAddressRef() = so_addr; 4745 stub_symbol->SetByteSize(symbol_stub_byte_size); 4746 } else { 4747 // Make a synthetic symbol to describe the trampoline stub 4748 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4749 if (sym_idx >= num_syms) { 4750 sym = symtab.Resize(++num_syms); 4751 stub_symbol = nullptr; // this pointer no longer valid 4752 } 4753 sym[sym_idx].SetID(synthetic_sym_id++); 4754 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4755 if (resolver_addresses.find(symbol_stub_addr) == 4756 resolver_addresses.end()) 4757 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4758 else 4759 sym[sym_idx].SetType(eSymbolTypeResolver); 4760 sym[sym_idx].SetIsSynthetic(true); 4761 sym[sym_idx].GetAddressRef() = so_addr; 4762 add_symbol_addr(so_addr.GetFileAddress()); 4763 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4764 ++sym_idx; 4765 } 4766 } else { 4767 if (log) 4768 log->Warning("symbol stub referencing symbol table symbol " 4769 "%u that isn't in our minimal symbol table, " 4770 "fix this!!!", 4771 stub_sym_id); 4772 } 4773 } 4774 } 4775 } 4776 } 4777 } 4778 } 4779 4780 if (!reexport_trie_entries.empty()) { 4781 for (const auto &e : reexport_trie_entries) { 4782 if (e.entry.import_name) { 4783 // Only add indirect symbols from the Trie entries if we didn't have 4784 // a N_INDR nlist entry for this already 4785 if (indirect_symbol_names.find(e.entry.name) == 4786 indirect_symbol_names.end()) { 4787 // Make a synthetic symbol to describe re-exported symbol. 4788 if (sym_idx >= num_syms) 4789 sym = symtab.Resize(++num_syms); 4790 sym[sym_idx].SetID(synthetic_sym_id++); 4791 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4792 sym[sym_idx].SetType(eSymbolTypeReExported); 4793 sym[sym_idx].SetIsSynthetic(true); 4794 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4795 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4796 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4797 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4798 } 4799 ++sym_idx; 4800 } 4801 } 4802 } 4803 } 4804 } 4805 4806 void ObjectFileMachO::Dump(Stream *s) { 4807 ModuleSP module_sp(GetModule()); 4808 if (module_sp) { 4809 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4810 s->Printf("%p: ", static_cast<void *>(this)); 4811 s->Indent(); 4812 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4813 s->PutCString("ObjectFileMachO64"); 4814 else 4815 s->PutCString("ObjectFileMachO32"); 4816 4817 *s << ", file = '" << m_file; 4818 ModuleSpecList all_specs; 4819 ModuleSpec base_spec; 4820 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 4821 base_spec, all_specs); 4822 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 4823 *s << "', triple"; 4824 if (e) 4825 s->Printf("[%d]", i); 4826 *s << " = "; 4827 *s << all_specs.GetModuleSpecRefAtIndex(i) 4828 .GetArchitecture() 4829 .GetTriple() 4830 .getTriple(); 4831 } 4832 *s << "\n"; 4833 SectionList *sections = GetSectionList(); 4834 if (sections) 4835 sections->Dump(s->AsRawOstream(), s->GetIndentLevel(), nullptr, true, 4836 UINT32_MAX); 4837 4838 if (m_symtab_up) 4839 m_symtab_up->Dump(s, nullptr, eSortOrderNone); 4840 } 4841 } 4842 4843 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4844 const lldb_private::DataExtractor &data, 4845 lldb::offset_t lc_offset) { 4846 uint32_t i; 4847 llvm::MachO::uuid_command load_cmd; 4848 4849 lldb::offset_t offset = lc_offset; 4850 for (i = 0; i < header.ncmds; ++i) { 4851 const lldb::offset_t cmd_offset = offset; 4852 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4853 break; 4854 4855 if (load_cmd.cmd == LC_UUID) { 4856 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4857 4858 if (uuid_bytes) { 4859 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4860 // We pretend these object files have no UUID to prevent crashing. 4861 4862 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4863 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4864 0xbb, 0x14, 0xf0, 0x0d}; 4865 4866 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4867 return UUID(); 4868 4869 return UUID(uuid_bytes, 16); 4870 } 4871 return UUID(); 4872 } 4873 offset = cmd_offset + load_cmd.cmdsize; 4874 } 4875 return UUID(); 4876 } 4877 4878 static llvm::StringRef GetOSName(uint32_t cmd) { 4879 switch (cmd) { 4880 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4881 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4882 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4883 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4884 case llvm::MachO::LC_VERSION_MIN_TVOS: 4885 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4886 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4887 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4888 default: 4889 llvm_unreachable("unexpected LC_VERSION load command"); 4890 } 4891 } 4892 4893 namespace { 4894 struct OSEnv { 4895 llvm::StringRef os_type; 4896 llvm::StringRef environment; 4897 OSEnv(uint32_t cmd) { 4898 switch (cmd) { 4899 case llvm::MachO::PLATFORM_MACOS: 4900 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4901 return; 4902 case llvm::MachO::PLATFORM_IOS: 4903 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4904 return; 4905 case llvm::MachO::PLATFORM_TVOS: 4906 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4907 return; 4908 case llvm::MachO::PLATFORM_WATCHOS: 4909 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4910 return; 4911 case llvm::MachO::PLATFORM_BRIDGEOS: 4912 os_type = llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4913 return; 4914 case llvm::MachO::PLATFORM_DRIVERKIT: 4915 os_type = llvm::Triple::getOSTypeName(llvm::Triple::DriverKit); 4916 return; 4917 case llvm::MachO::PLATFORM_MACCATALYST: 4918 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4919 environment = llvm::Triple::getEnvironmentTypeName(llvm::Triple::MacABI); 4920 return; 4921 case llvm::MachO::PLATFORM_IOSSIMULATOR: 4922 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4923 environment = 4924 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4925 return; 4926 case llvm::MachO::PLATFORM_TVOSSIMULATOR: 4927 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4928 environment = 4929 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4930 return; 4931 case llvm::MachO::PLATFORM_WATCHOSSIMULATOR: 4932 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4933 environment = 4934 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4935 return; 4936 case llvm::MachO::PLATFORM_XROS: 4937 os_type = llvm::Triple::getOSTypeName(llvm::Triple::XROS); 4938 return; 4939 case llvm::MachO::PLATFORM_XROS_SIMULATOR: 4940 os_type = llvm::Triple::getOSTypeName(llvm::Triple::XROS); 4941 environment = 4942 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4943 return; 4944 default: { 4945 Log *log(GetLog(LLDBLog::Symbols | LLDBLog::Process)); 4946 LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION"); 4947 } 4948 } 4949 } 4950 }; 4951 4952 struct MinOS { 4953 uint32_t major_version, minor_version, patch_version; 4954 MinOS(uint32_t version) 4955 : major_version(version >> 16), minor_version((version >> 8) & 0xffu), 4956 patch_version(version & 0xffu) {} 4957 }; 4958 } // namespace 4959 4960 void ObjectFileMachO::GetAllArchSpecs(const llvm::MachO::mach_header &header, 4961 const lldb_private::DataExtractor &data, 4962 lldb::offset_t lc_offset, 4963 ModuleSpec &base_spec, 4964 lldb_private::ModuleSpecList &all_specs) { 4965 auto &base_arch = base_spec.GetArchitecture(); 4966 base_arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4967 if (!base_arch.IsValid()) 4968 return; 4969 4970 bool found_any = false; 4971 auto add_triple = [&](const llvm::Triple &triple) { 4972 auto spec = base_spec; 4973 spec.GetArchitecture().GetTriple() = triple; 4974 if (spec.GetArchitecture().IsValid()) { 4975 spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset); 4976 all_specs.Append(spec); 4977 found_any = true; 4978 } 4979 }; 4980 4981 // Set OS to an unspecified unknown or a "*" so it can match any OS 4982 llvm::Triple base_triple = base_arch.GetTriple(); 4983 base_triple.setOS(llvm::Triple::UnknownOS); 4984 base_triple.setOSName(llvm::StringRef()); 4985 4986 if (header.filetype == MH_PRELOAD) { 4987 if (header.cputype == CPU_TYPE_ARM) { 4988 // If this is a 32-bit arm binary, and it's a standalone binary, force 4989 // the Vendor to Apple so we don't accidentally pick up the generic 4990 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4991 // frame pointer register; most other armv7 ABIs use a combination of 4992 // r7 and r11. 4993 base_triple.setVendor(llvm::Triple::Apple); 4994 } else { 4995 // Set vendor to an unspecified unknown or a "*" so it can match any 4996 // vendor This is required for correct behavior of EFI debugging on 4997 // x86_64 4998 base_triple.setVendor(llvm::Triple::UnknownVendor); 4999 base_triple.setVendorName(llvm::StringRef()); 5000 } 5001 return add_triple(base_triple); 5002 } 5003 5004 llvm::MachO::load_command load_cmd; 5005 5006 // See if there is an LC_VERSION_MIN_* load command that can give 5007 // us the OS type. 5008 lldb::offset_t offset = lc_offset; 5009 for (uint32_t i = 0; i < header.ncmds; ++i) { 5010 const lldb::offset_t cmd_offset = offset; 5011 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 5012 break; 5013 5014 llvm::MachO::version_min_command version_min; 5015 switch (load_cmd.cmd) { 5016 case llvm::MachO::LC_VERSION_MIN_MACOSX: 5017 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 5018 case llvm::MachO::LC_VERSION_MIN_TVOS: 5019 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 5020 if (load_cmd.cmdsize != sizeof(version_min)) 5021 break; 5022 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 5023 data.GetByteOrder(), &version_min) == 0) 5024 break; 5025 MinOS min_os(version_min.version); 5026 llvm::SmallString<32> os_name; 5027 llvm::raw_svector_ostream os(os_name); 5028 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 5029 << min_os.minor_version << '.' << min_os.patch_version; 5030 5031 auto triple = base_triple; 5032 triple.setOSName(os.str()); 5033 5034 // Disambiguate legacy simulator platforms. 5035 if (load_cmd.cmd != llvm::MachO::LC_VERSION_MIN_MACOSX && 5036 (base_triple.getArch() == llvm::Triple::x86_64 || 5037 base_triple.getArch() == llvm::Triple::x86)) { 5038 // The combination of legacy LC_VERSION_MIN load command and 5039 // x86 architecture always indicates a simulator environment. 5040 // The combination of LC_VERSION_MIN and arm architecture only 5041 // appears for native binaries. Back-deploying simulator 5042 // binaries on Apple Silicon Macs use the modern unambigous 5043 // LC_BUILD_VERSION load commands; no special handling required. 5044 triple.setEnvironment(llvm::Triple::Simulator); 5045 } 5046 add_triple(triple); 5047 break; 5048 } 5049 default: 5050 break; 5051 } 5052 5053 offset = cmd_offset + load_cmd.cmdsize; 5054 } 5055 5056 // See if there are LC_BUILD_VERSION load commands that can give 5057 // us the OS type. 5058 offset = lc_offset; 5059 for (uint32_t i = 0; i < header.ncmds; ++i) { 5060 const lldb::offset_t cmd_offset = offset; 5061 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 5062 break; 5063 5064 do { 5065 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 5066 llvm::MachO::build_version_command build_version; 5067 if (load_cmd.cmdsize < sizeof(build_version)) { 5068 // Malformed load command. 5069 break; 5070 } 5071 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 5072 data.GetByteOrder(), &build_version) == 0) 5073 break; 5074 MinOS min_os(build_version.minos); 5075 OSEnv os_env(build_version.platform); 5076 llvm::SmallString<16> os_name; 5077 llvm::raw_svector_ostream os(os_name); 5078 os << os_env.os_type << min_os.major_version << '.' 5079 << min_os.minor_version << '.' << min_os.patch_version; 5080 auto triple = base_triple; 5081 triple.setOSName(os.str()); 5082 os_name.clear(); 5083 if (!os_env.environment.empty()) 5084 triple.setEnvironmentName(os_env.environment); 5085 add_triple(triple); 5086 } 5087 } while (false); 5088 offset = cmd_offset + load_cmd.cmdsize; 5089 } 5090 5091 if (!found_any) { 5092 add_triple(base_triple); 5093 } 5094 } 5095 5096 ArchSpec ObjectFileMachO::GetArchitecture( 5097 ModuleSP module_sp, const llvm::MachO::mach_header &header, 5098 const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) { 5099 ModuleSpecList all_specs; 5100 ModuleSpec base_spec; 5101 GetAllArchSpecs(header, data, MachHeaderSizeFromMagic(header.magic), 5102 base_spec, all_specs); 5103 5104 // If the object file offers multiple alternative load commands, 5105 // pick the one that matches the module. 5106 if (module_sp) { 5107 const ArchSpec &module_arch = module_sp->GetArchitecture(); 5108 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 5109 ArchSpec mach_arch = 5110 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 5111 if (module_arch.IsCompatibleMatch(mach_arch)) 5112 return mach_arch; 5113 } 5114 } 5115 5116 // Return the first arch we found. 5117 if (all_specs.GetSize() == 0) 5118 return {}; 5119 return all_specs.GetModuleSpecRefAtIndex(0).GetArchitecture(); 5120 } 5121 5122 UUID ObjectFileMachO::GetUUID() { 5123 ModuleSP module_sp(GetModule()); 5124 if (module_sp) { 5125 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5126 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5127 return GetUUID(m_header, m_data, offset); 5128 } 5129 return UUID(); 5130 } 5131 5132 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5133 ModuleSP module_sp = GetModule(); 5134 if (!module_sp) 5135 return 0; 5136 5137 uint32_t count = 0; 5138 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5139 llvm::MachO::load_command load_cmd; 5140 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5141 std::vector<std::string> rpath_paths; 5142 std::vector<std::string> rpath_relative_paths; 5143 std::vector<std::string> at_exec_relative_paths; 5144 uint32_t i; 5145 for (i = 0; i < m_header.ncmds; ++i) { 5146 const uint32_t cmd_offset = offset; 5147 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5148 break; 5149 5150 switch (load_cmd.cmd) { 5151 case LC_RPATH: 5152 case LC_LOAD_DYLIB: 5153 case LC_LOAD_WEAK_DYLIB: 5154 case LC_REEXPORT_DYLIB: 5155 case LC_LOAD_DYLINKER: 5156 case LC_LOADFVMLIB: 5157 case LC_LOAD_UPWARD_DYLIB: { 5158 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5159 // For LC_LOAD_DYLIB there is an alternate encoding 5160 // which adds a uint32_t `flags` field for `DYLD_USE_*` 5161 // flags. This can be detected by a timestamp field with 5162 // the `DYLIB_USE_MARKER` constant value. 5163 bool is_delayed_init = false; 5164 uint32_t use_command_marker = m_data.GetU32(&offset); 5165 if (use_command_marker == 0x1a741800 /* DYLIB_USE_MARKER */) { 5166 offset += 4; /* uint32_t current_version */ 5167 offset += 4; /* uint32_t compat_version */ 5168 uint32_t flags = m_data.GetU32(&offset); 5169 // If this LC_LOAD_DYLIB is marked delay-init, 5170 // don't report it as a dependent library -- it 5171 // may be loaded in the process at some point, 5172 // but will most likely not be load at launch. 5173 if (flags & 0x08 /* DYLIB_USE_DELAYED_INIT */) 5174 is_delayed_init = true; 5175 } 5176 const char *path = m_data.PeekCStr(name_offset); 5177 if (path && !is_delayed_init) { 5178 if (load_cmd.cmd == LC_RPATH) 5179 rpath_paths.push_back(path); 5180 else { 5181 if (path[0] == '@') { 5182 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5183 rpath_relative_paths.push_back(path + strlen("@rpath")); 5184 else if (strncmp(path, "@executable_path", 5185 strlen("@executable_path")) == 0) 5186 at_exec_relative_paths.push_back(path + 5187 strlen("@executable_path")); 5188 } else { 5189 FileSpec file_spec(path); 5190 if (files.AppendIfUnique(file_spec)) 5191 count++; 5192 } 5193 } 5194 } 5195 } break; 5196 5197 default: 5198 break; 5199 } 5200 offset = cmd_offset + load_cmd.cmdsize; 5201 } 5202 5203 FileSpec this_file_spec(m_file); 5204 FileSystem::Instance().Resolve(this_file_spec); 5205 5206 if (!rpath_paths.empty()) { 5207 // Fixup all LC_RPATH values to be absolute paths. 5208 const std::string this_directory = 5209 this_file_spec.GetDirectory().GetString(); 5210 for (auto &rpath : rpath_paths) { 5211 if (llvm::StringRef(rpath).starts_with(g_loader_path)) 5212 rpath = this_directory + rpath.substr(g_loader_path.size()); 5213 else if (llvm::StringRef(rpath).starts_with(g_executable_path)) 5214 rpath = this_directory + rpath.substr(g_executable_path.size()); 5215 } 5216 5217 for (const auto &rpath_relative_path : rpath_relative_paths) { 5218 for (const auto &rpath : rpath_paths) { 5219 std::string path = rpath; 5220 path += rpath_relative_path; 5221 // It is OK to resolve this path because we must find a file on disk 5222 // for us to accept it anyway if it is rpath relative. 5223 FileSpec file_spec(path); 5224 FileSystem::Instance().Resolve(file_spec); 5225 if (FileSystem::Instance().Exists(file_spec) && 5226 files.AppendIfUnique(file_spec)) { 5227 count++; 5228 break; 5229 } 5230 } 5231 } 5232 } 5233 5234 // We may have @executable_paths but no RPATHS. Figure those out here. 5235 // Only do this if this object file is the executable. We have no way to 5236 // get back to the actual executable otherwise, so we won't get the right 5237 // path. 5238 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5239 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5240 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5241 FileSpec file_spec = 5242 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5243 if (FileSystem::Instance().Exists(file_spec) && 5244 files.AppendIfUnique(file_spec)) 5245 count++; 5246 } 5247 } 5248 return count; 5249 } 5250 5251 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5252 // If the object file is not an executable it can't hold the entry point. 5253 // m_entry_point_address is initialized to an invalid address, so we can just 5254 // return that. If m_entry_point_address is valid it means we've found it 5255 // already, so return the cached value. 5256 5257 if ((!IsExecutable() && !IsDynamicLoader()) || 5258 m_entry_point_address.IsValid()) { 5259 return m_entry_point_address; 5260 } 5261 5262 // Otherwise, look for the UnixThread or Thread command. The data for the 5263 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5264 // 5265 // uint32_t flavor - this is the flavor argument you would pass to 5266 // thread_get_state 5267 // uint32_t count - this is the count of longs in the thread state data 5268 // struct XXX_thread_state state - this is the structure from 5269 // <machine/thread_status.h> corresponding to the flavor. 5270 // <repeat this trio> 5271 // 5272 // So we just keep reading the various register flavors till we find the GPR 5273 // one, then read the PC out of there. 5274 // FIXME: We will need to have a "RegisterContext data provider" class at some 5275 // point that can get all the registers 5276 // out of data in this form & attach them to a given thread. That should 5277 // underlie the MacOS X User process plugin, and we'll also need it for the 5278 // MacOS X Core File process plugin. When we have that we can also use it 5279 // here. 5280 // 5281 // For now we hard-code the offsets and flavors we need: 5282 // 5283 // 5284 5285 ModuleSP module_sp(GetModule()); 5286 if (module_sp) { 5287 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5288 llvm::MachO::load_command load_cmd; 5289 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5290 uint32_t i; 5291 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5292 bool done = false; 5293 5294 for (i = 0; i < m_header.ncmds; ++i) { 5295 const lldb::offset_t cmd_offset = offset; 5296 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5297 break; 5298 5299 switch (load_cmd.cmd) { 5300 case LC_UNIXTHREAD: 5301 case LC_THREAD: { 5302 while (offset < cmd_offset + load_cmd.cmdsize) { 5303 uint32_t flavor = m_data.GetU32(&offset); 5304 uint32_t count = m_data.GetU32(&offset); 5305 if (count == 0) { 5306 // We've gotten off somehow, log and exit; 5307 return m_entry_point_address; 5308 } 5309 5310 switch (m_header.cputype) { 5311 case llvm::MachO::CPU_TYPE_ARM: 5312 if (flavor == 1 || 5313 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 5314 // from mach/arm/thread_status.h 5315 { 5316 offset += 60; // This is the offset of pc in the GPR thread state 5317 // data structure. 5318 start_address = m_data.GetU32(&offset); 5319 done = true; 5320 } 5321 break; 5322 case llvm::MachO::CPU_TYPE_ARM64: 5323 case llvm::MachO::CPU_TYPE_ARM64_32: 5324 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5325 { 5326 offset += 256; // This is the offset of pc in the GPR thread state 5327 // data structure. 5328 start_address = m_data.GetU64(&offset); 5329 done = true; 5330 } 5331 break; 5332 case llvm::MachO::CPU_TYPE_I386: 5333 if (flavor == 5334 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5335 { 5336 offset += 40; // This is the offset of eip in the GPR thread state 5337 // data structure. 5338 start_address = m_data.GetU32(&offset); 5339 done = true; 5340 } 5341 break; 5342 case llvm::MachO::CPU_TYPE_X86_64: 5343 if (flavor == 5344 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5345 { 5346 offset += 16 * 8; // This is the offset of rip in the GPR thread 5347 // state data structure. 5348 start_address = m_data.GetU64(&offset); 5349 done = true; 5350 } 5351 break; 5352 default: 5353 return m_entry_point_address; 5354 } 5355 // Haven't found the GPR flavor yet, skip over the data for this 5356 // flavor: 5357 if (done) 5358 break; 5359 offset += count * 4; 5360 } 5361 } break; 5362 case LC_MAIN: { 5363 uint64_t entryoffset = m_data.GetU64(&offset); 5364 SectionSP text_segment_sp = 5365 GetSectionList()->FindSectionByName(GetSegmentNameTEXT()); 5366 if (text_segment_sp) { 5367 done = true; 5368 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5369 } 5370 } break; 5371 5372 default: 5373 break; 5374 } 5375 if (done) 5376 break; 5377 5378 // Go to the next load command: 5379 offset = cmd_offset + load_cmd.cmdsize; 5380 } 5381 5382 if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) { 5383 if (GetSymtab()) { 5384 Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType( 5385 ConstString("_dyld_start"), SymbolType::eSymbolTypeCode, 5386 Symtab::eDebugAny, Symtab::eVisibilityAny); 5387 if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) { 5388 start_address = dyld_start_sym->GetAddress().GetFileAddress(); 5389 } 5390 } 5391 } 5392 5393 if (start_address != LLDB_INVALID_ADDRESS) { 5394 // We got the start address from the load commands, so now resolve that 5395 // address in the sections of this ObjectFile: 5396 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5397 start_address, GetSectionList())) { 5398 m_entry_point_address.Clear(); 5399 } 5400 } else { 5401 // We couldn't read the UnixThread load command - maybe it wasn't there. 5402 // As a fallback look for the "start" symbol in the main executable. 5403 5404 ModuleSP module_sp(GetModule()); 5405 5406 if (module_sp) { 5407 SymbolContextList contexts; 5408 SymbolContext context; 5409 module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5410 eSymbolTypeCode, contexts); 5411 if (contexts.GetSize()) { 5412 if (contexts.GetContextAtIndex(0, context)) 5413 m_entry_point_address = context.symbol->GetAddress(); 5414 } 5415 } 5416 } 5417 } 5418 5419 return m_entry_point_address; 5420 } 5421 5422 lldb_private::Address ObjectFileMachO::GetBaseAddress() { 5423 lldb_private::Address header_addr; 5424 SectionList *section_list = GetSectionList(); 5425 if (section_list) { 5426 SectionSP text_segment_sp( 5427 section_list->FindSectionByName(GetSegmentNameTEXT())); 5428 if (text_segment_sp) { 5429 header_addr.SetSection(text_segment_sp); 5430 header_addr.SetOffset(0); 5431 } 5432 } 5433 return header_addr; 5434 } 5435 5436 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5437 ModuleSP module_sp(GetModule()); 5438 if (module_sp) { 5439 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5440 if (!m_thread_context_offsets_valid) { 5441 m_thread_context_offsets_valid = true; 5442 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5443 FileRangeArray::Entry file_range; 5444 llvm::MachO::thread_command thread_cmd; 5445 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5446 const uint32_t cmd_offset = offset; 5447 if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr) 5448 break; 5449 5450 if (thread_cmd.cmd == LC_THREAD) { 5451 file_range.SetRangeBase(offset); 5452 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5453 m_thread_context_offsets.Append(file_range); 5454 } 5455 offset = cmd_offset + thread_cmd.cmdsize; 5456 } 5457 } 5458 } 5459 return m_thread_context_offsets.GetSize(); 5460 } 5461 5462 std::vector<std::tuple<offset_t, offset_t>> 5463 ObjectFileMachO::FindLC_NOTEByName(std::string name) { 5464 std::vector<std::tuple<offset_t, offset_t>> results; 5465 ModuleSP module_sp(GetModule()); 5466 if (module_sp) { 5467 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5468 5469 offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5470 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5471 const uint32_t cmd_offset = offset; 5472 llvm::MachO::load_command lc = {}; 5473 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5474 break; 5475 if (lc.cmd == LC_NOTE) { 5476 char data_owner[17]; 5477 m_data.CopyData(offset, 16, data_owner); 5478 data_owner[16] = '\0'; 5479 offset += 16; 5480 5481 if (name == data_owner) { 5482 offset_t payload_offset = m_data.GetU64_unchecked(&offset); 5483 offset_t payload_size = m_data.GetU64_unchecked(&offset); 5484 results.push_back({payload_offset, payload_size}); 5485 } 5486 } 5487 offset = cmd_offset + lc.cmdsize; 5488 } 5489 } 5490 return results; 5491 } 5492 5493 std::string ObjectFileMachO::GetIdentifierString() { 5494 Log *log( 5495 GetLog(LLDBLog::Symbols | LLDBLog::Process | LLDBLog::DynamicLoader)); 5496 ModuleSP module_sp(GetModule()); 5497 if (module_sp) { 5498 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5499 5500 auto lc_notes = FindLC_NOTEByName("kern ver str"); 5501 for (auto lc_note : lc_notes) { 5502 offset_t payload_offset = std::get<0>(lc_note); 5503 offset_t payload_size = std::get<1>(lc_note); 5504 uint32_t version; 5505 if (m_data.GetU32(&payload_offset, &version, 1) != nullptr) { 5506 if (version == 1) { 5507 uint32_t strsize = payload_size - sizeof(uint32_t); 5508 std::string result(strsize, '\0'); 5509 m_data.CopyData(payload_offset, strsize, result.data()); 5510 LLDB_LOGF(log, "LC_NOTE 'kern ver str' found with text '%s'", 5511 result.c_str()); 5512 return result; 5513 } 5514 } 5515 } 5516 5517 // Second, make a pass over the load commands looking for an obsolete 5518 // LC_IDENT load command. 5519 offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5520 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5521 const uint32_t cmd_offset = offset; 5522 llvm::MachO::ident_command ident_command; 5523 if (m_data.GetU32(&offset, &ident_command, 2) == nullptr) 5524 break; 5525 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5526 std::string result(ident_command.cmdsize, '\0'); 5527 if (m_data.CopyData(offset, ident_command.cmdsize, result.data()) == 5528 ident_command.cmdsize) { 5529 LLDB_LOGF(log, "LC_IDENT found with text '%s'", result.c_str()); 5530 return result; 5531 } 5532 } 5533 offset = cmd_offset + ident_command.cmdsize; 5534 } 5535 } 5536 return {}; 5537 } 5538 5539 AddressableBits ObjectFileMachO::GetAddressableBits() { 5540 AddressableBits addressable_bits; 5541 5542 Log *log(GetLog(LLDBLog::Process)); 5543 ModuleSP module_sp(GetModule()); 5544 if (module_sp) { 5545 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5546 auto lc_notes = FindLC_NOTEByName("addrable bits"); 5547 for (auto lc_note : lc_notes) { 5548 offset_t payload_offset = std::get<0>(lc_note); 5549 uint32_t version; 5550 if (m_data.GetU32(&payload_offset, &version, 1) != nullptr) { 5551 if (version == 3) { 5552 uint32_t num_addr_bits = m_data.GetU32_unchecked(&payload_offset); 5553 addressable_bits.SetAddressableBits(num_addr_bits); 5554 LLDB_LOGF(log, 5555 "LC_NOTE 'addrable bits' v3 found, value %d " 5556 "bits", 5557 num_addr_bits); 5558 } 5559 if (version == 4) { 5560 uint32_t lo_addr_bits = m_data.GetU32_unchecked(&payload_offset); 5561 uint32_t hi_addr_bits = m_data.GetU32_unchecked(&payload_offset); 5562 5563 if (lo_addr_bits == hi_addr_bits) 5564 addressable_bits.SetAddressableBits(lo_addr_bits); 5565 else 5566 addressable_bits.SetAddressableBits(lo_addr_bits, hi_addr_bits); 5567 LLDB_LOGF(log, "LC_NOTE 'addrable bits' v4 found, value %d & %d bits", 5568 lo_addr_bits, hi_addr_bits); 5569 } 5570 } 5571 } 5572 } 5573 return addressable_bits; 5574 } 5575 5576 bool ObjectFileMachO::GetCorefileMainBinaryInfo(addr_t &value, 5577 bool &value_is_offset, 5578 UUID &uuid, 5579 ObjectFile::BinaryType &type) { 5580 Log *log( 5581 GetLog(LLDBLog::Symbols | LLDBLog::Process | LLDBLog::DynamicLoader)); 5582 value = LLDB_INVALID_ADDRESS; 5583 value_is_offset = false; 5584 uuid.Clear(); 5585 uint32_t log2_pagesize = 0; // not currently passed up to caller 5586 uint32_t platform = 0; // not currently passed up to caller 5587 ModuleSP module_sp(GetModule()); 5588 if (module_sp) { 5589 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5590 5591 auto lc_notes = FindLC_NOTEByName("main bin spec"); 5592 for (auto lc_note : lc_notes) { 5593 offset_t payload_offset = std::get<0>(lc_note); 5594 5595 // struct main_bin_spec 5596 // { 5597 // uint32_t version; // currently 2 5598 // uint32_t type; // 0 == unspecified, 1 == kernel, 5599 // // 2 == user process, 5600 // // 3 == standalone binary 5601 // uint64_t address; // UINT64_MAX if address not specified 5602 // uint64_t slide; // slide, UINT64_MAX if unspecified 5603 // // 0 if no slide needs to be applied to 5604 // // file address 5605 // uuid_t uuid; // all zero's if uuid not specified 5606 // uint32_t log2_pagesize; // process page size in log base 2, 5607 // // e.g. 4k pages are 12. 5608 // // 0 for unspecified 5609 // uint32_t platform; // The Mach-O platform for this corefile. 5610 // // 0 for unspecified. 5611 // // The values are defined in 5612 // // <mach-o/loader.h>, PLATFORM_*. 5613 // } __attribute((packed)); 5614 5615 // "main bin spec" (main binary specification) data payload is 5616 // formatted: 5617 // uint32_t version [currently 1] 5618 // uint32_t type [0 == unspecified, 1 == kernel, 5619 // 2 == user process, 3 == firmware ] 5620 // uint64_t address [ UINT64_MAX if address not specified ] 5621 // uuid_t uuid [ all zero's if uuid not specified ] 5622 // uint32_t log2_pagesize [ process page size in log base 5623 // 2, e.g. 4k pages are 12. 5624 // 0 for unspecified ] 5625 // uint32_t unused [ for alignment ] 5626 5627 uint32_t version; 5628 if (m_data.GetU32(&payload_offset, &version, 1) != nullptr && 5629 version <= 2) { 5630 uint32_t binspec_type = 0; 5631 uuid_t raw_uuid; 5632 memset(raw_uuid, 0, sizeof(uuid_t)); 5633 5634 if (!m_data.GetU32(&payload_offset, &binspec_type, 1)) 5635 return false; 5636 if (!m_data.GetU64(&payload_offset, &value, 1)) 5637 return false; 5638 uint64_t slide = LLDB_INVALID_ADDRESS; 5639 if (version > 1 && !m_data.GetU64(&payload_offset, &slide, 1)) 5640 return false; 5641 if (value == LLDB_INVALID_ADDRESS && slide != LLDB_INVALID_ADDRESS) { 5642 value = slide; 5643 value_is_offset = true; 5644 } 5645 5646 if (m_data.CopyData(payload_offset, sizeof(uuid_t), raw_uuid) != 0) { 5647 uuid = UUID(raw_uuid, sizeof(uuid_t)); 5648 // convert the "main bin spec" type into our 5649 // ObjectFile::BinaryType enum 5650 const char *typestr = "unrecognized type"; 5651 switch (binspec_type) { 5652 case 0: 5653 type = eBinaryTypeUnknown; 5654 typestr = "uknown"; 5655 break; 5656 case 1: 5657 type = eBinaryTypeKernel; 5658 typestr = "xnu kernel"; 5659 break; 5660 case 2: 5661 type = eBinaryTypeUser; 5662 typestr = "userland dyld"; 5663 break; 5664 case 3: 5665 type = eBinaryTypeStandalone; 5666 typestr = "standalone"; 5667 break; 5668 } 5669 LLDB_LOGF(log, 5670 "LC_NOTE 'main bin spec' found, version %d type %d " 5671 "(%s), value 0x%" PRIx64 " value-is-slide==%s uuid %s", 5672 version, type, typestr, value, 5673 value_is_offset ? "true" : "false", 5674 uuid.GetAsString().c_str()); 5675 if (!m_data.GetU32(&payload_offset, &log2_pagesize, 1)) 5676 return false; 5677 if (version > 1 && !m_data.GetU32(&payload_offset, &platform, 1)) 5678 return false; 5679 return true; 5680 } 5681 } 5682 } 5683 } 5684 return false; 5685 } 5686 5687 bool ObjectFileMachO::GetCorefileThreadExtraInfos( 5688 std::vector<lldb::tid_t> &tids) { 5689 tids.clear(); 5690 ModuleSP module_sp(GetModule()); 5691 if (module_sp) { 5692 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5693 5694 Log *log(GetLog(LLDBLog::Object | LLDBLog::Process | LLDBLog::Thread)); 5695 auto lc_notes = FindLC_NOTEByName("process metadata"); 5696 for (auto lc_note : lc_notes) { 5697 offset_t payload_offset = std::get<0>(lc_note); 5698 offset_t strsize = std::get<1>(lc_note); 5699 std::string buf(strsize, '\0'); 5700 if (m_data.CopyData(payload_offset, strsize, buf.data()) != strsize) { 5701 LLDB_LOGF(log, 5702 "Unable to read %" PRIu64 5703 " bytes of 'process metadata' LC_NOTE JSON contents", 5704 strsize); 5705 return false; 5706 } 5707 while (buf.back() == '\0') 5708 buf.resize(buf.size() - 1); 5709 StructuredData::ObjectSP object_sp = StructuredData::ParseJSON(buf); 5710 StructuredData::Dictionary *dict = object_sp->GetAsDictionary(); 5711 if (!dict) { 5712 LLDB_LOGF(log, "Unable to read 'process metadata' LC_NOTE, did not " 5713 "get a dictionary."); 5714 return false; 5715 } 5716 StructuredData::Array *threads; 5717 if (!dict->GetValueForKeyAsArray("threads", threads) || !threads) { 5718 LLDB_LOGF(log, 5719 "'process metadata' LC_NOTE does not have a 'threads' key"); 5720 return false; 5721 } 5722 if (threads->GetSize() != GetNumThreadContexts()) { 5723 LLDB_LOGF(log, "Unable to read 'process metadata' LC_NOTE, number of " 5724 "threads does not match number of LC_THREADS."); 5725 return false; 5726 } 5727 const size_t num_threads = threads->GetSize(); 5728 for (size_t i = 0; i < num_threads; i++) { 5729 std::optional<StructuredData::Dictionary *> maybe_thread = 5730 threads->GetItemAtIndexAsDictionary(i); 5731 if (!maybe_thread) { 5732 LLDB_LOGF(log, 5733 "Unable to read 'process metadata' LC_NOTE, threads " 5734 "array does not have a dictionary at index %zu.", 5735 i); 5736 return false; 5737 } 5738 StructuredData::Dictionary *thread = *maybe_thread; 5739 lldb::tid_t tid = LLDB_INVALID_THREAD_ID; 5740 if (thread->GetValueForKeyAsInteger<lldb::tid_t>("thread_id", tid)) 5741 if (tid == 0) 5742 tid = LLDB_INVALID_THREAD_ID; 5743 tids.push_back(tid); 5744 } 5745 5746 if (log) { 5747 StreamString logmsg; 5748 logmsg.Printf("LC_NOTE 'process metadata' found: "); 5749 dict->Dump(logmsg, /* pretty_print */ false); 5750 LLDB_LOGF(log, "%s", logmsg.GetData()); 5751 } 5752 return true; 5753 } 5754 } 5755 return false; 5756 } 5757 5758 lldb::RegisterContextSP 5759 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5760 lldb_private::Thread &thread) { 5761 lldb::RegisterContextSP reg_ctx_sp; 5762 5763 ModuleSP module_sp(GetModule()); 5764 if (module_sp) { 5765 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5766 if (!m_thread_context_offsets_valid) 5767 GetNumThreadContexts(); 5768 5769 const FileRangeArray::Entry *thread_context_file_range = 5770 m_thread_context_offsets.GetEntryAtIndex(idx); 5771 if (thread_context_file_range) { 5772 5773 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5774 thread_context_file_range->GetByteSize()); 5775 5776 switch (m_header.cputype) { 5777 case llvm::MachO::CPU_TYPE_ARM64: 5778 case llvm::MachO::CPU_TYPE_ARM64_32: 5779 reg_ctx_sp = 5780 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data); 5781 break; 5782 5783 case llvm::MachO::CPU_TYPE_ARM: 5784 reg_ctx_sp = 5785 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data); 5786 break; 5787 5788 case llvm::MachO::CPU_TYPE_I386: 5789 reg_ctx_sp = 5790 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data); 5791 break; 5792 5793 case llvm::MachO::CPU_TYPE_X86_64: 5794 reg_ctx_sp = 5795 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data); 5796 break; 5797 } 5798 } 5799 } 5800 return reg_ctx_sp; 5801 } 5802 5803 ObjectFile::Type ObjectFileMachO::CalculateType() { 5804 switch (m_header.filetype) { 5805 case MH_OBJECT: // 0x1u 5806 if (GetAddressByteSize() == 4) { 5807 // 32 bit kexts are just object files, but they do have a valid 5808 // UUID load command. 5809 if (GetUUID()) { 5810 // this checking for the UUID load command is not enough we could 5811 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5812 // this is required of kexts 5813 if (m_strata == eStrataInvalid) 5814 m_strata = eStrataKernel; 5815 return eTypeSharedLibrary; 5816 } 5817 } 5818 return eTypeObjectFile; 5819 5820 case MH_EXECUTE: 5821 return eTypeExecutable; // 0x2u 5822 case MH_FVMLIB: 5823 return eTypeSharedLibrary; // 0x3u 5824 case MH_CORE: 5825 return eTypeCoreFile; // 0x4u 5826 case MH_PRELOAD: 5827 return eTypeSharedLibrary; // 0x5u 5828 case MH_DYLIB: 5829 return eTypeSharedLibrary; // 0x6u 5830 case MH_DYLINKER: 5831 return eTypeDynamicLinker; // 0x7u 5832 case MH_BUNDLE: 5833 return eTypeSharedLibrary; // 0x8u 5834 case MH_DYLIB_STUB: 5835 return eTypeStubLibrary; // 0x9u 5836 case MH_DSYM: 5837 return eTypeDebugInfo; // 0xAu 5838 case MH_KEXT_BUNDLE: 5839 return eTypeSharedLibrary; // 0xBu 5840 default: 5841 break; 5842 } 5843 return eTypeUnknown; 5844 } 5845 5846 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5847 switch (m_header.filetype) { 5848 case MH_OBJECT: // 0x1u 5849 { 5850 // 32 bit kexts are just object files, but they do have a valid 5851 // UUID load command. 5852 if (GetUUID()) { 5853 // this checking for the UUID load command is not enough we could 5854 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5855 // this is required of kexts 5856 if (m_type == eTypeInvalid) 5857 m_type = eTypeSharedLibrary; 5858 5859 return eStrataKernel; 5860 } 5861 } 5862 return eStrataUnknown; 5863 5864 case MH_EXECUTE: // 0x2u 5865 // Check for the MH_DYLDLINK bit in the flags 5866 if (m_header.flags & MH_DYLDLINK) { 5867 return eStrataUser; 5868 } else { 5869 SectionList *section_list = GetSectionList(); 5870 if (section_list) { 5871 static ConstString g_kld_section_name("__KLD"); 5872 if (section_list->FindSectionByName(g_kld_section_name)) 5873 return eStrataKernel; 5874 } 5875 } 5876 return eStrataRawImage; 5877 5878 case MH_FVMLIB: 5879 return eStrataUser; // 0x3u 5880 case MH_CORE: 5881 return eStrataUnknown; // 0x4u 5882 case MH_PRELOAD: 5883 return eStrataRawImage; // 0x5u 5884 case MH_DYLIB: 5885 return eStrataUser; // 0x6u 5886 case MH_DYLINKER: 5887 return eStrataUser; // 0x7u 5888 case MH_BUNDLE: 5889 return eStrataUser; // 0x8u 5890 case MH_DYLIB_STUB: 5891 return eStrataUser; // 0x9u 5892 case MH_DSYM: 5893 return eStrataUnknown; // 0xAu 5894 case MH_KEXT_BUNDLE: 5895 return eStrataKernel; // 0xBu 5896 default: 5897 break; 5898 } 5899 return eStrataUnknown; 5900 } 5901 5902 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5903 ModuleSP module_sp(GetModule()); 5904 if (module_sp) { 5905 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5906 llvm::MachO::dylib_command load_cmd; 5907 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5908 uint32_t version_cmd = 0; 5909 uint64_t version = 0; 5910 uint32_t i; 5911 for (i = 0; i < m_header.ncmds; ++i) { 5912 const lldb::offset_t cmd_offset = offset; 5913 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5914 break; 5915 5916 if (load_cmd.cmd == LC_ID_DYLIB) { 5917 if (version_cmd == 0) { 5918 version_cmd = load_cmd.cmd; 5919 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr) 5920 break; 5921 version = load_cmd.dylib.current_version; 5922 } 5923 break; // Break for now unless there is another more complete version 5924 // number load command in the future. 5925 } 5926 offset = cmd_offset + load_cmd.cmdsize; 5927 } 5928 5929 if (version_cmd == LC_ID_DYLIB) { 5930 unsigned major = (version & 0xFFFF0000ull) >> 16; 5931 unsigned minor = (version & 0x0000FF00ull) >> 8; 5932 unsigned subminor = (version & 0x000000FFull); 5933 return llvm::VersionTuple(major, minor, subminor); 5934 } 5935 } 5936 return llvm::VersionTuple(); 5937 } 5938 5939 ArchSpec ObjectFileMachO::GetArchitecture() { 5940 ModuleSP module_sp(GetModule()); 5941 ArchSpec arch; 5942 if (module_sp) { 5943 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5944 5945 return GetArchitecture(module_sp, m_header, m_data, 5946 MachHeaderSizeFromMagic(m_header.magic)); 5947 } 5948 return arch; 5949 } 5950 5951 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, 5952 addr_t &base_addr, UUID &uuid) { 5953 uuid.Clear(); 5954 base_addr = LLDB_INVALID_ADDRESS; 5955 if (process && process->GetDynamicLoader()) { 5956 DynamicLoader *dl = process->GetDynamicLoader(); 5957 LazyBool using_shared_cache; 5958 LazyBool private_shared_cache; 5959 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5960 private_shared_cache); 5961 } 5962 Log *log(GetLog(LLDBLog::Symbols | LLDBLog::Process)); 5963 LLDB_LOGF( 5964 log, 5965 "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64, 5966 uuid.GetAsString().c_str(), base_addr); 5967 } 5968 5969 // From dyld SPI header dyld_process_info.h 5970 typedef void *dyld_process_info; 5971 struct lldb_copy__dyld_process_cache_info { 5972 uuid_t cacheUUID; // UUID of cache used by process 5973 uint64_t cacheBaseAddress; // load address of dyld shared cache 5974 bool noCache; // process is running without a dyld cache 5975 bool privateCache; // process is using a private copy of its dyld cache 5976 }; 5977 5978 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with 5979 // llvm enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile 5980 // errors. So we need to use the actual underlying types of task_t and 5981 // kern_return_t below. 5982 extern "C" unsigned int /*task_t*/ mach_task_self(); 5983 5984 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5985 uuid.Clear(); 5986 base_addr = LLDB_INVALID_ADDRESS; 5987 5988 #if defined(__APPLE__) 5989 uint8_t *(*dyld_get_all_image_infos)(void); 5990 dyld_get_all_image_infos = 5991 (uint8_t * (*)()) dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5992 if (dyld_get_all_image_infos) { 5993 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5994 if (dyld_all_image_infos_address) { 5995 uint32_t *version = (uint32_t *) 5996 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5997 if (*version >= 13) { 5998 uuid_t *sharedCacheUUID_address = 0; 5999 int wordsize = sizeof(uint8_t *); 6000 if (wordsize == 8) { 6001 sharedCacheUUID_address = 6002 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 6003 160); // sharedCacheUUID <mach-o/dyld_images.h> 6004 if (*version >= 15) 6005 base_addr = 6006 *(uint64_t 6007 *)((uint8_t *)dyld_all_image_infos_address + 6008 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 6009 } else { 6010 sharedCacheUUID_address = 6011 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 6012 84); // sharedCacheUUID <mach-o/dyld_images.h> 6013 if (*version >= 15) { 6014 base_addr = 0; 6015 base_addr = 6016 *(uint32_t 6017 *)((uint8_t *)dyld_all_image_infos_address + 6018 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 6019 } 6020 } 6021 uuid = UUID(sharedCacheUUID_address, sizeof(uuid_t)); 6022 } 6023 } 6024 } else { 6025 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 6026 dyld_process_info (*dyld_process_info_create)( 6027 unsigned int /* task_t */ task, uint64_t timestamp, 6028 unsigned int /*kern_return_t*/ *kernelError); 6029 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 6030 void (*dyld_process_info_release)(dyld_process_info info); 6031 6032 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, 6033 unsigned int /*kern_return_t*/ *)) 6034 dlsym(RTLD_DEFAULT, "_dyld_process_info_create"); 6035 dyld_process_info_get_cache = (void (*)(void *, void *))dlsym( 6036 RTLD_DEFAULT, "_dyld_process_info_get_cache"); 6037 dyld_process_info_release = 6038 (void (*)(void *))dlsym(RTLD_DEFAULT, "_dyld_process_info_release"); 6039 6040 if (dyld_process_info_create && dyld_process_info_get_cache) { 6041 unsigned int /*kern_return_t */ kern_ret; 6042 dyld_process_info process_info = 6043 dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 6044 if (process_info) { 6045 struct lldb_copy__dyld_process_cache_info sc_info; 6046 memset(&sc_info, 0, sizeof(struct lldb_copy__dyld_process_cache_info)); 6047 dyld_process_info_get_cache(process_info, &sc_info); 6048 if (sc_info.cacheBaseAddress != 0) { 6049 base_addr = sc_info.cacheBaseAddress; 6050 uuid = UUID(sc_info.cacheUUID, sizeof(uuid_t)); 6051 } 6052 dyld_process_info_release(process_info); 6053 } 6054 } 6055 } 6056 Log *log(GetLog(LLDBLog::Symbols | LLDBLog::Process)); 6057 if (log && uuid.IsValid()) 6058 LLDB_LOGF(log, 6059 "lldb's in-memory shared cache has a UUID of %s base address of " 6060 "0x%" PRIx64, 6061 uuid.GetAsString().c_str(), base_addr); 6062 #endif 6063 } 6064 6065 static llvm::VersionTuple FindMinimumVersionInfo(DataExtractor &data, 6066 lldb::offset_t offset, 6067 size_t ncmds) { 6068 for (size_t i = 0; i < ncmds; i++) { 6069 const lldb::offset_t load_cmd_offset = offset; 6070 llvm::MachO::load_command lc = {}; 6071 if (data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6072 break; 6073 6074 uint32_t version = 0; 6075 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 6076 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 6077 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 6078 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 6079 // struct version_min_command { 6080 // uint32_t cmd; // LC_VERSION_MIN_* 6081 // uint32_t cmdsize; 6082 // uint32_t version; // X.Y.Z encoded in nibbles xxxx.yy.zz 6083 // uint32_t sdk; 6084 // }; 6085 // We want to read version. 6086 version = data.GetU32(&offset); 6087 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 6088 // struct build_version_command { 6089 // uint32_t cmd; // LC_BUILD_VERSION 6090 // uint32_t cmdsize; 6091 // uint32_t platform; 6092 // uint32_t minos; // X.Y.Z encoded in nibbles xxxx.yy.zz 6093 // uint32_t sdk; 6094 // uint32_t ntools; 6095 // }; 6096 // We want to read minos. 6097 offset += sizeof(uint32_t); // Skip over platform 6098 version = data.GetU32(&offset); // Extract minos 6099 } 6100 6101 if (version) { 6102 const uint32_t xxxx = version >> 16; 6103 const uint32_t yy = (version >> 8) & 0xffu; 6104 const uint32_t zz = version & 0xffu; 6105 if (xxxx) 6106 return llvm::VersionTuple(xxxx, yy, zz); 6107 } 6108 offset = load_cmd_offset + lc.cmdsize; 6109 } 6110 return llvm::VersionTuple(); 6111 } 6112 6113 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 6114 if (!m_min_os_version) 6115 m_min_os_version = FindMinimumVersionInfo( 6116 m_data, MachHeaderSizeFromMagic(m_header.magic), m_header.ncmds); 6117 return *m_min_os_version; 6118 } 6119 6120 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { 6121 if (!m_sdk_versions) 6122 m_sdk_versions = FindMinimumVersionInfo( 6123 m_data, MachHeaderSizeFromMagic(m_header.magic), m_header.ncmds); 6124 return *m_sdk_versions; 6125 } 6126 6127 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 6128 return m_header.filetype == llvm::MachO::MH_DYLINKER; 6129 } 6130 6131 bool ObjectFileMachO::CanTrustAddressRanges() { 6132 // Dsymutil guarantees that the .debug_aranges accelerator is complete and can 6133 // be trusted by LLDB. 6134 return m_header.filetype == llvm::MachO::MH_DSYM; 6135 } 6136 6137 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 6138 return m_allow_assembly_emulation_unwind_plans; 6139 } 6140 6141 Section *ObjectFileMachO::GetMachHeaderSection() { 6142 // Find the first address of the mach header which is the first non-zero file 6143 // sized section whose file offset is zero. This is the base file address of 6144 // the mach-o file which can be subtracted from the vmaddr of the other 6145 // segments found in memory and added to the load address 6146 ModuleSP module_sp = GetModule(); 6147 if (!module_sp) 6148 return nullptr; 6149 SectionList *section_list = GetSectionList(); 6150 if (!section_list) 6151 return nullptr; 6152 6153 // Some binaries can have a TEXT segment with a non-zero file offset. 6154 // Binaries in the shared cache are one example. Some hand-generated 6155 // binaries may not be laid out in the normal TEXT,DATA,LC_SYMTAB order 6156 // in the file, even though they're laid out correctly in vmaddr terms. 6157 SectionSP text_segment_sp = 6158 section_list->FindSectionByName(GetSegmentNameTEXT()); 6159 if (text_segment_sp.get() && SectionIsLoadable(text_segment_sp.get())) 6160 return text_segment_sp.get(); 6161 6162 const size_t num_sections = section_list->GetSize(); 6163 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6164 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 6165 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 6166 return section; 6167 } 6168 6169 return nullptr; 6170 } 6171 6172 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 6173 if (!section) 6174 return false; 6175 if (section->IsThreadSpecific()) 6176 return false; 6177 if (GetModule().get() != section->GetModule().get()) 6178 return false; 6179 // firmware style binaries with llvm gcov segment do 6180 // not have that segment mapped into memory. 6181 if (section->GetName() == GetSegmentNameLLVM_COV()) { 6182 const Strata strata = GetStrata(); 6183 if (strata == eStrataKernel || strata == eStrataRawImage) 6184 return false; 6185 } 6186 // Be careful with __LINKEDIT and __DWARF segments 6187 if (section->GetName() == GetSegmentNameLINKEDIT() || 6188 section->GetName() == GetSegmentNameDWARF()) { 6189 // Only map __LINKEDIT and __DWARF if we have an in memory image and 6190 // this isn't a kernel binary like a kext or mach_kernel. 6191 const bool is_memory_image = (bool)m_process_wp.lock(); 6192 const Strata strata = GetStrata(); 6193 if (is_memory_image == false || strata == eStrataKernel) 6194 return false; 6195 } 6196 return true; 6197 } 6198 6199 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 6200 lldb::addr_t header_load_address, const Section *header_section, 6201 const Section *section) { 6202 ModuleSP module_sp = GetModule(); 6203 if (module_sp && header_section && section && 6204 header_load_address != LLDB_INVALID_ADDRESS) { 6205 lldb::addr_t file_addr = header_section->GetFileAddress(); 6206 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 6207 return section->GetFileAddress() - file_addr + header_load_address; 6208 } 6209 return LLDB_INVALID_ADDRESS; 6210 } 6211 6212 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 6213 bool value_is_offset) { 6214 Log *log(GetLog(LLDBLog::DynamicLoader)); 6215 ModuleSP module_sp = GetModule(); 6216 if (!module_sp) 6217 return false; 6218 6219 SectionList *section_list = GetSectionList(); 6220 if (!section_list) 6221 return false; 6222 6223 size_t num_loaded_sections = 0; 6224 const size_t num_sections = section_list->GetSize(); 6225 6226 // Warn if some top-level segments map to the same address. The binary may be 6227 // malformed. 6228 const bool warn_multiple = true; 6229 6230 if (log) { 6231 StreamString logmsg; 6232 logmsg << "ObjectFileMachO::SetLoadAddress "; 6233 if (GetFileSpec()) 6234 logmsg << "path='" << GetFileSpec().GetPath() << "' "; 6235 if (GetUUID()) { 6236 logmsg << "uuid=" << GetUUID().GetAsString(); 6237 } 6238 LLDB_LOGF(log, "%s", logmsg.GetData()); 6239 } 6240 if (value_is_offset) { 6241 // "value" is an offset to apply to each top level segment 6242 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6243 // Iterate through the object file sections to find all of the 6244 // sections that size on disk (to avoid __PAGEZERO) and load them 6245 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6246 if (SectionIsLoadable(section_sp.get())) { 6247 LLDB_LOGF(log, 6248 "ObjectFileMachO::SetLoadAddress segment '%s' load addr is " 6249 "0x%" PRIx64, 6250 section_sp->GetName().AsCString(), 6251 section_sp->GetFileAddress() + value); 6252 if (target.GetSectionLoadList().SetSectionLoadAddress( 6253 section_sp, section_sp->GetFileAddress() + value, 6254 warn_multiple)) 6255 ++num_loaded_sections; 6256 } 6257 } 6258 } else { 6259 // "value" is the new base address of the mach_header, adjust each 6260 // section accordingly 6261 6262 Section *mach_header_section = GetMachHeaderSection(); 6263 if (mach_header_section) { 6264 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6265 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6266 6267 lldb::addr_t section_load_addr = 6268 CalculateSectionLoadAddressForMemoryImage( 6269 value, mach_header_section, section_sp.get()); 6270 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6271 LLDB_LOGF(log, 6272 "ObjectFileMachO::SetLoadAddress segment '%s' load addr is " 6273 "0x%" PRIx64, 6274 section_sp->GetName().AsCString(), section_load_addr); 6275 if (target.GetSectionLoadList().SetSectionLoadAddress( 6276 section_sp, section_load_addr, warn_multiple)) 6277 ++num_loaded_sections; 6278 } 6279 } 6280 } 6281 } 6282 return num_loaded_sections > 0; 6283 } 6284 6285 struct all_image_infos_header { 6286 uint32_t version; // currently 1 6287 uint32_t imgcount; // number of binary images 6288 uint64_t entries_fileoff; // file offset in the corefile of where the array of 6289 // struct entry's begin. 6290 uint32_t entries_size; // size of 'struct entry'. 6291 uint32_t unused; 6292 }; 6293 6294 struct image_entry { 6295 uint64_t filepath_offset; // offset in corefile to c-string of the file path, 6296 // UINT64_MAX if unavailable. 6297 uuid_t uuid; // uint8_t[16]. should be set to all zeroes if 6298 // uuid is unknown. 6299 uint64_t load_address; // UINT64_MAX if unknown. 6300 uint64_t seg_addrs_offset; // offset to the array of struct segment_vmaddr's. 6301 uint32_t segment_count; // The number of segments for this binary. 6302 uint32_t unused; 6303 6304 image_entry() { 6305 filepath_offset = UINT64_MAX; 6306 memset(&uuid, 0, sizeof(uuid_t)); 6307 segment_count = 0; 6308 load_address = UINT64_MAX; 6309 seg_addrs_offset = UINT64_MAX; 6310 unused = 0; 6311 } 6312 image_entry(const image_entry &rhs) { 6313 filepath_offset = rhs.filepath_offset; 6314 memcpy(&uuid, &rhs.uuid, sizeof(uuid_t)); 6315 segment_count = rhs.segment_count; 6316 seg_addrs_offset = rhs.seg_addrs_offset; 6317 load_address = rhs.load_address; 6318 unused = rhs.unused; 6319 } 6320 }; 6321 6322 struct segment_vmaddr { 6323 char segname[16]; 6324 uint64_t vmaddr; 6325 uint64_t unused; 6326 6327 segment_vmaddr() { 6328 memset(&segname, 0, 16); 6329 vmaddr = UINT64_MAX; 6330 unused = 0; 6331 } 6332 segment_vmaddr(const segment_vmaddr &rhs) { 6333 memcpy(&segname, &rhs.segname, 16); 6334 vmaddr = rhs.vmaddr; 6335 unused = rhs.unused; 6336 } 6337 }; 6338 6339 // Write the payload for the "all image infos" LC_NOTE into 6340 // the supplied all_image_infos_payload, assuming that this 6341 // will be written into the corefile starting at 6342 // initial_file_offset. 6343 // 6344 // The placement of this payload is a little tricky. We're 6345 // laying this out as 6346 // 6347 // 1. header (struct all_image_info_header) 6348 // 2. Array of fixed-size (struct image_entry)'s, one 6349 // per binary image present in the process. 6350 // 3. Arrays of (struct segment_vmaddr)'s, a varying number 6351 // for each binary image. 6352 // 4. Variable length c-strings of binary image filepaths, 6353 // one per binary. 6354 // 6355 // To compute where everything will be laid out in the 6356 // payload, we need to iterate over the images and calculate 6357 // how many segment_vmaddr structures each image will need, 6358 // and how long each image's filepath c-string is. There 6359 // are some multiple passes over the image list while calculating 6360 // everything. 6361 6362 static offset_t 6363 CreateAllImageInfosPayload(const lldb::ProcessSP &process_sp, 6364 offset_t initial_file_offset, 6365 StreamString &all_image_infos_payload, 6366 lldb_private::SaveCoreOptions &options) { 6367 Target &target = process_sp->GetTarget(); 6368 ModuleList modules = target.GetImages(); 6369 6370 // stack-only corefiles have no reason to include binaries that 6371 // are not executing; we're trying to make the smallest corefile 6372 // we can, so leave the rest out. 6373 if (options.GetStyle() == SaveCoreStyle::eSaveCoreStackOnly) 6374 modules.Clear(); 6375 6376 std::set<std::string> executing_uuids; 6377 std::vector<ThreadSP> thread_list = 6378 process_sp->CalculateCoreFileThreadList(options); 6379 for (const ThreadSP &thread_sp : thread_list) { 6380 uint32_t stack_frame_count = thread_sp->GetStackFrameCount(); 6381 for (uint32_t j = 0; j < stack_frame_count; j++) { 6382 StackFrameSP stack_frame_sp = thread_sp->GetStackFrameAtIndex(j); 6383 Address pc = stack_frame_sp->GetFrameCodeAddress(); 6384 ModuleSP module_sp = pc.GetModule(); 6385 if (module_sp) { 6386 UUID uuid = module_sp->GetUUID(); 6387 if (uuid.IsValid()) { 6388 executing_uuids.insert(uuid.GetAsString()); 6389 modules.AppendIfNeeded(module_sp); 6390 } 6391 } 6392 } 6393 } 6394 size_t modules_count = modules.GetSize(); 6395 6396 struct all_image_infos_header infos; 6397 infos.version = 1; 6398 infos.imgcount = modules_count; 6399 infos.entries_size = sizeof(image_entry); 6400 infos.entries_fileoff = initial_file_offset + sizeof(all_image_infos_header); 6401 infos.unused = 0; 6402 6403 all_image_infos_payload.PutHex32(infos.version); 6404 all_image_infos_payload.PutHex32(infos.imgcount); 6405 all_image_infos_payload.PutHex64(infos.entries_fileoff); 6406 all_image_infos_payload.PutHex32(infos.entries_size); 6407 all_image_infos_payload.PutHex32(infos.unused); 6408 6409 // First create the structures for all of the segment name+vmaddr vectors 6410 // for each module, so we will know the size of them as we add the 6411 // module entries. 6412 std::vector<std::vector<segment_vmaddr>> modules_segment_vmaddrs; 6413 for (size_t i = 0; i < modules_count; i++) { 6414 ModuleSP module = modules.GetModuleAtIndex(i); 6415 6416 SectionList *sections = module->GetSectionList(); 6417 size_t sections_count = sections->GetSize(); 6418 std::vector<segment_vmaddr> segment_vmaddrs; 6419 for (size_t j = 0; j < sections_count; j++) { 6420 SectionSP section = sections->GetSectionAtIndex(j); 6421 if (!section->GetParent().get()) { 6422 addr_t vmaddr = section->GetLoadBaseAddress(&target); 6423 if (vmaddr == LLDB_INVALID_ADDRESS) 6424 continue; 6425 ConstString name = section->GetName(); 6426 segment_vmaddr seg_vmaddr; 6427 // This is the uncommon case where strncpy is exactly 6428 // the right one, doesn't need to be nul terminated. 6429 // The segment name in a Mach-O LC_SEGMENT/LC_SEGMENT_64 is char[16] and 6430 // is not guaranteed to be nul-terminated if all 16 characters are 6431 // used. 6432 // coverity[buffer_size_warning] 6433 strncpy(seg_vmaddr.segname, name.AsCString(), 6434 sizeof(seg_vmaddr.segname)); 6435 seg_vmaddr.vmaddr = vmaddr; 6436 seg_vmaddr.unused = 0; 6437 segment_vmaddrs.push_back(seg_vmaddr); 6438 } 6439 } 6440 modules_segment_vmaddrs.push_back(segment_vmaddrs); 6441 } 6442 6443 offset_t size_of_vmaddr_structs = 0; 6444 for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) { 6445 size_of_vmaddr_structs += 6446 modules_segment_vmaddrs[i].size() * sizeof(segment_vmaddr); 6447 } 6448 6449 offset_t size_of_filepath_cstrings = 0; 6450 for (size_t i = 0; i < modules_count; i++) { 6451 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6452 size_of_filepath_cstrings += module_sp->GetFileSpec().GetPath().size() + 1; 6453 } 6454 6455 // Calculate the file offsets of our "all image infos" payload in the 6456 // corefile. initial_file_offset the original value passed in to this method. 6457 6458 offset_t start_of_entries = 6459 initial_file_offset + sizeof(all_image_infos_header); 6460 offset_t start_of_seg_vmaddrs = 6461 start_of_entries + sizeof(image_entry) * modules_count; 6462 offset_t start_of_filenames = start_of_seg_vmaddrs + size_of_vmaddr_structs; 6463 6464 offset_t final_file_offset = start_of_filenames + size_of_filepath_cstrings; 6465 6466 // Now write the one-per-module 'struct image_entry' into the 6467 // StringStream; keep track of where the struct segment_vmaddr 6468 // entries for each module will end up in the corefile. 6469 6470 offset_t current_string_offset = start_of_filenames; 6471 offset_t current_segaddrs_offset = start_of_seg_vmaddrs; 6472 std::vector<struct image_entry> image_entries; 6473 for (size_t i = 0; i < modules_count; i++) { 6474 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6475 6476 struct image_entry ent; 6477 memcpy(&ent.uuid, module_sp->GetUUID().GetBytes().data(), sizeof(ent.uuid)); 6478 if (modules_segment_vmaddrs[i].size() > 0) { 6479 ent.segment_count = modules_segment_vmaddrs[i].size(); 6480 ent.seg_addrs_offset = current_segaddrs_offset; 6481 } 6482 ent.filepath_offset = current_string_offset; 6483 ObjectFile *objfile = module_sp->GetObjectFile(); 6484 if (objfile) { 6485 Address base_addr(objfile->GetBaseAddress()); 6486 if (base_addr.IsValid()) { 6487 ent.load_address = base_addr.GetLoadAddress(&target); 6488 } 6489 } 6490 6491 all_image_infos_payload.PutHex64(ent.filepath_offset); 6492 all_image_infos_payload.PutRawBytes(ent.uuid, sizeof(ent.uuid)); 6493 all_image_infos_payload.PutHex64(ent.load_address); 6494 all_image_infos_payload.PutHex64(ent.seg_addrs_offset); 6495 all_image_infos_payload.PutHex32(ent.segment_count); 6496 6497 if (executing_uuids.find(module_sp->GetUUID().GetAsString()) != 6498 executing_uuids.end()) 6499 all_image_infos_payload.PutHex32(1); 6500 else 6501 all_image_infos_payload.PutHex32(0); 6502 6503 current_segaddrs_offset += ent.segment_count * sizeof(segment_vmaddr); 6504 current_string_offset += module_sp->GetFileSpec().GetPath().size() + 1; 6505 } 6506 6507 // Now write the struct segment_vmaddr entries into the StringStream. 6508 6509 for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) { 6510 if (modules_segment_vmaddrs[i].size() == 0) 6511 continue; 6512 for (struct segment_vmaddr segvm : modules_segment_vmaddrs[i]) { 6513 all_image_infos_payload.PutRawBytes(segvm.segname, sizeof(segvm.segname)); 6514 all_image_infos_payload.PutHex64(segvm.vmaddr); 6515 all_image_infos_payload.PutHex64(segvm.unused); 6516 } 6517 } 6518 6519 for (size_t i = 0; i < modules_count; i++) { 6520 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6521 std::string filepath = module_sp->GetFileSpec().GetPath(); 6522 all_image_infos_payload.PutRawBytes(filepath.data(), filepath.size() + 1); 6523 } 6524 6525 return final_file_offset; 6526 } 6527 6528 // Temp struct used to combine contiguous memory regions with 6529 // identical permissions. 6530 struct page_object { 6531 addr_t addr; 6532 addr_t size; 6533 uint32_t prot; 6534 }; 6535 6536 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6537 lldb_private::SaveCoreOptions &options, 6538 Status &error) { 6539 // The FileSpec and Process are already checked in PluginManager::SaveCore. 6540 assert(options.GetOutputFile().has_value()); 6541 assert(process_sp); 6542 const FileSpec outfile = options.GetOutputFile().value(); 6543 6544 // MachO defaults to dirty pages 6545 if (options.GetStyle() == SaveCoreStyle::eSaveCoreUnspecified) 6546 options.SetStyle(eSaveCoreDirtyOnly); 6547 6548 Target &target = process_sp->GetTarget(); 6549 const ArchSpec target_arch = target.GetArchitecture(); 6550 const llvm::Triple &target_triple = target_arch.GetTriple(); 6551 if (target_triple.getVendor() == llvm::Triple::Apple && 6552 (target_triple.getOS() == llvm::Triple::MacOSX || 6553 target_triple.getOS() == llvm::Triple::IOS || 6554 target_triple.getOS() == llvm::Triple::WatchOS || 6555 target_triple.getOS() == llvm::Triple::TvOS || 6556 target_triple.getOS() == llvm::Triple::XROS)) { 6557 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) 6558 // { 6559 bool make_core = false; 6560 switch (target_arch.GetMachine()) { 6561 case llvm::Triple::aarch64: 6562 case llvm::Triple::aarch64_32: 6563 case llvm::Triple::arm: 6564 case llvm::Triple::thumb: 6565 case llvm::Triple::x86: 6566 case llvm::Triple::x86_64: 6567 make_core = true; 6568 break; 6569 default: 6570 error = Status::FromErrorStringWithFormat( 6571 "unsupported core architecture: %s", target_triple.str().c_str()); 6572 break; 6573 } 6574 6575 if (make_core) { 6576 CoreFileMemoryRanges core_ranges; 6577 error = process_sp->CalculateCoreFileSaveRanges(options, core_ranges); 6578 if (error.Success()) { 6579 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6580 const ByteOrder byte_order = target_arch.GetByteOrder(); 6581 std::vector<llvm::MachO::segment_command_64> segment_load_commands; 6582 for (const auto &core_range_info : core_ranges) { 6583 // TODO: Refactor RangeDataVector to have a data iterator. 6584 const auto &core_range = core_range_info.data; 6585 uint32_t cmd_type = LC_SEGMENT_64; 6586 uint32_t segment_size = sizeof(llvm::MachO::segment_command_64); 6587 if (addr_byte_size == 4) { 6588 cmd_type = LC_SEGMENT; 6589 segment_size = sizeof(llvm::MachO::segment_command); 6590 } 6591 // Skip any ranges with no read/write/execute permissions and empty 6592 // ranges. 6593 if (core_range.lldb_permissions == 0 || core_range.range.size() == 0) 6594 continue; 6595 uint32_t vm_prot = 0; 6596 if (core_range.lldb_permissions & ePermissionsReadable) 6597 vm_prot |= VM_PROT_READ; 6598 if (core_range.lldb_permissions & ePermissionsWritable) 6599 vm_prot |= VM_PROT_WRITE; 6600 if (core_range.lldb_permissions & ePermissionsExecutable) 6601 vm_prot |= VM_PROT_EXECUTE; 6602 const addr_t vm_addr = core_range.range.start(); 6603 const addr_t vm_size = core_range.range.size(); 6604 llvm::MachO::segment_command_64 segment = { 6605 cmd_type, // uint32_t cmd; 6606 segment_size, // uint32_t cmdsize; 6607 {0}, // char segname[16]; 6608 vm_addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6609 vm_size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6610 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6611 vm_size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6612 vm_prot, // uint32_t maxprot; 6613 vm_prot, // uint32_t initprot; 6614 0, // uint32_t nsects; 6615 0}; // uint32_t flags; 6616 segment_load_commands.push_back(segment); 6617 } 6618 6619 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6620 6621 llvm::MachO::mach_header_64 mach_header; 6622 mach_header.magic = addr_byte_size == 8 ? MH_MAGIC_64 : MH_MAGIC; 6623 mach_header.cputype = target_arch.GetMachOCPUType(); 6624 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6625 mach_header.filetype = MH_CORE; 6626 mach_header.ncmds = segment_load_commands.size(); 6627 mach_header.flags = 0; 6628 mach_header.reserved = 0; 6629 ThreadList &thread_list = process_sp->GetThreadList(); 6630 const uint32_t num_threads = thread_list.GetSize(); 6631 6632 // Make an array of LC_THREAD data items. Each one contains the 6633 // contents of the LC_THREAD load command. The data doesn't contain 6634 // the load command + load command size, we will add the load command 6635 // and load command size as we emit the data. 6636 std::vector<StreamString> LC_THREAD_datas(num_threads); 6637 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6638 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6639 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6640 LC_THREAD_data.SetByteOrder(byte_order); 6641 } 6642 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) { 6643 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6644 if (thread_sp) { 6645 switch (mach_header.cputype) { 6646 case llvm::MachO::CPU_TYPE_ARM64: 6647 case llvm::MachO::CPU_TYPE_ARM64_32: 6648 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6649 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6650 break; 6651 6652 case llvm::MachO::CPU_TYPE_ARM: 6653 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6654 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6655 break; 6656 6657 case llvm::MachO::CPU_TYPE_I386: 6658 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6659 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6660 break; 6661 6662 case llvm::MachO::CPU_TYPE_X86_64: 6663 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6664 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6665 break; 6666 } 6667 } 6668 } 6669 6670 // The size of the load command is the size of the segments... 6671 if (addr_byte_size == 8) { 6672 mach_header.sizeofcmds = segment_load_commands.size() * 6673 sizeof(llvm::MachO::segment_command_64); 6674 } else { 6675 mach_header.sizeofcmds = segment_load_commands.size() * 6676 sizeof(llvm::MachO::segment_command); 6677 } 6678 6679 // and the size of all LC_THREAD load command 6680 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6681 ++mach_header.ncmds; 6682 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6683 } 6684 6685 // Bits will be set to indicate which bits are NOT used in 6686 // addressing in this process or 0 for unknown. 6687 uint64_t address_mask = process_sp->GetCodeAddressMask(); 6688 if (address_mask != LLDB_INVALID_ADDRESS_MASK) { 6689 // LC_NOTE "addrable bits" 6690 mach_header.ncmds++; 6691 mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); 6692 } 6693 6694 // LC_NOTE "process metadata" 6695 mach_header.ncmds++; 6696 mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); 6697 6698 // LC_NOTE "all image infos" 6699 mach_header.ncmds++; 6700 mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); 6701 6702 // Write the mach header 6703 buffer.PutHex32(mach_header.magic); 6704 buffer.PutHex32(mach_header.cputype); 6705 buffer.PutHex32(mach_header.cpusubtype); 6706 buffer.PutHex32(mach_header.filetype); 6707 buffer.PutHex32(mach_header.ncmds); 6708 buffer.PutHex32(mach_header.sizeofcmds); 6709 buffer.PutHex32(mach_header.flags); 6710 if (addr_byte_size == 8) { 6711 buffer.PutHex32(mach_header.reserved); 6712 } 6713 6714 // Skip the mach header and all load commands and align to the next 6715 // 0x1000 byte boundary 6716 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6717 6718 file_offset = llvm::alignTo(file_offset, 16); 6719 std::vector<std::unique_ptr<LCNoteEntry>> lc_notes; 6720 6721 // Add "addrable bits" LC_NOTE when an address mask is available 6722 if (address_mask != LLDB_INVALID_ADDRESS_MASK) { 6723 std::unique_ptr<LCNoteEntry> addrable_bits_lcnote_up( 6724 new LCNoteEntry(addr_byte_size, byte_order)); 6725 addrable_bits_lcnote_up->name = "addrable bits"; 6726 addrable_bits_lcnote_up->payload_file_offset = file_offset; 6727 int bits = std::bitset<64>(~address_mask).count(); 6728 addrable_bits_lcnote_up->payload.PutHex32(4); // version 6729 addrable_bits_lcnote_up->payload.PutHex32( 6730 bits); // # of bits used for low addresses 6731 addrable_bits_lcnote_up->payload.PutHex32( 6732 bits); // # of bits used for high addresses 6733 addrable_bits_lcnote_up->payload.PutHex32(0); // reserved 6734 6735 file_offset += addrable_bits_lcnote_up->payload.GetSize(); 6736 6737 lc_notes.push_back(std::move(addrable_bits_lcnote_up)); 6738 } 6739 6740 // Add "process metadata" LC_NOTE 6741 std::unique_ptr<LCNoteEntry> thread_extrainfo_lcnote_up( 6742 new LCNoteEntry(addr_byte_size, byte_order)); 6743 thread_extrainfo_lcnote_up->name = "process metadata"; 6744 thread_extrainfo_lcnote_up->payload_file_offset = file_offset; 6745 6746 StructuredData::DictionarySP dict( 6747 std::make_shared<StructuredData::Dictionary>()); 6748 StructuredData::ArraySP threads( 6749 std::make_shared<StructuredData::Array>()); 6750 for (const ThreadSP &thread_sp : 6751 process_sp->CalculateCoreFileThreadList(options)) { 6752 StructuredData::DictionarySP thread( 6753 std::make_shared<StructuredData::Dictionary>()); 6754 thread->AddIntegerItem("thread_id", thread_sp->GetID()); 6755 threads->AddItem(thread); 6756 } 6757 dict->AddItem("threads", threads); 6758 StreamString strm; 6759 dict->Dump(strm, /* pretty */ false); 6760 thread_extrainfo_lcnote_up->payload.PutRawBytes(strm.GetData(), 6761 strm.GetSize()); 6762 6763 file_offset += thread_extrainfo_lcnote_up->payload.GetSize(); 6764 file_offset = llvm::alignTo(file_offset, 16); 6765 lc_notes.push_back(std::move(thread_extrainfo_lcnote_up)); 6766 6767 // Add "all image infos" LC_NOTE 6768 std::unique_ptr<LCNoteEntry> all_image_infos_lcnote_up( 6769 new LCNoteEntry(addr_byte_size, byte_order)); 6770 all_image_infos_lcnote_up->name = "all image infos"; 6771 all_image_infos_lcnote_up->payload_file_offset = file_offset; 6772 file_offset = CreateAllImageInfosPayload( 6773 process_sp, file_offset, all_image_infos_lcnote_up->payload, 6774 options); 6775 lc_notes.push_back(std::move(all_image_infos_lcnote_up)); 6776 6777 // Add LC_NOTE load commands 6778 for (auto &lcnote : lc_notes) { 6779 // Add the LC_NOTE load command to the file. 6780 buffer.PutHex32(LC_NOTE); 6781 buffer.PutHex32(sizeof(llvm::MachO::note_command)); 6782 char namebuf[16]; 6783 memset(namebuf, 0, sizeof(namebuf)); 6784 // This is the uncommon case where strncpy is exactly 6785 // the right one, doesn't need to be nul terminated. 6786 // LC_NOTE name field is char[16] and is not guaranteed to be 6787 // nul-terminated. 6788 // coverity[buffer_size_warning] 6789 strncpy(namebuf, lcnote->name.c_str(), sizeof(namebuf)); 6790 buffer.PutRawBytes(namebuf, sizeof(namebuf)); 6791 buffer.PutHex64(lcnote->payload_file_offset); 6792 buffer.PutHex64(lcnote->payload.GetSize()); 6793 } 6794 6795 // Align to 4096-byte page boundary for the LC_SEGMENTs. 6796 file_offset = llvm::alignTo(file_offset, 4096); 6797 6798 for (auto &segment : segment_load_commands) { 6799 segment.fileoff = file_offset; 6800 file_offset += segment.filesize; 6801 } 6802 6803 // Write out all of the LC_THREAD load commands 6804 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6805 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6806 buffer.PutHex32(LC_THREAD); 6807 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6808 buffer.Write(LC_THREAD_data.GetString().data(), LC_THREAD_data_size); 6809 } 6810 6811 // Write out all of the segment load commands 6812 for (const auto &segment : segment_load_commands) { 6813 buffer.PutHex32(segment.cmd); 6814 buffer.PutHex32(segment.cmdsize); 6815 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6816 if (addr_byte_size == 8) { 6817 buffer.PutHex64(segment.vmaddr); 6818 buffer.PutHex64(segment.vmsize); 6819 buffer.PutHex64(segment.fileoff); 6820 buffer.PutHex64(segment.filesize); 6821 } else { 6822 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6823 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6824 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6825 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6826 } 6827 buffer.PutHex32(segment.maxprot); 6828 buffer.PutHex32(segment.initprot); 6829 buffer.PutHex32(segment.nsects); 6830 buffer.PutHex32(segment.flags); 6831 } 6832 6833 std::string core_file_path(outfile.GetPath()); 6834 auto core_file = FileSystem::Instance().Open( 6835 outfile, File::eOpenOptionWriteOnly | File::eOpenOptionTruncate | 6836 File::eOpenOptionCanCreate); 6837 if (!core_file) { 6838 error = Status::FromError(core_file.takeError()); 6839 } else { 6840 // Read 1 page at a time 6841 uint8_t bytes[0x1000]; 6842 // Write the mach header and load commands out to the core file 6843 size_t bytes_written = buffer.GetString().size(); 6844 error = 6845 core_file.get()->Write(buffer.GetString().data(), bytes_written); 6846 if (error.Success()) { 6847 6848 for (auto &lcnote : lc_notes) { 6849 if (core_file.get()->SeekFromStart(lcnote->payload_file_offset) == 6850 -1) { 6851 error = Status::FromErrorStringWithFormat( 6852 "Unable to seek to corefile pos " 6853 "to write '%s' LC_NOTE payload", 6854 lcnote->name.c_str()); 6855 return false; 6856 } 6857 bytes_written = lcnote->payload.GetSize(); 6858 error = core_file.get()->Write(lcnote->payload.GetData(), 6859 bytes_written); 6860 if (!error.Success()) 6861 return false; 6862 } 6863 6864 // Now write the file data for all memory segments in the process 6865 for (const auto &segment : segment_load_commands) { 6866 if (core_file.get()->SeekFromStart(segment.fileoff) == -1) { 6867 error = Status::FromErrorStringWithFormat( 6868 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6869 segment.fileoff, core_file_path.c_str()); 6870 break; 6871 } 6872 6873 target.GetDebugger().GetAsyncOutputStream()->Printf( 6874 "Saving %" PRId64 6875 " bytes of data for memory region at 0x%" PRIx64 "\n", 6876 segment.vmsize, segment.vmaddr); 6877 addr_t bytes_left = segment.vmsize; 6878 addr_t addr = segment.vmaddr; 6879 Status memory_read_error; 6880 while (bytes_left > 0 && error.Success()) { 6881 const size_t bytes_to_read = 6882 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6883 6884 // In a savecore setting, we don't really care about caching, 6885 // as the data is dumped and very likely never read again, 6886 // so we call ReadMemoryFromInferior to bypass it. 6887 const size_t bytes_read = process_sp->ReadMemoryFromInferior( 6888 addr, bytes, bytes_to_read, memory_read_error); 6889 6890 if (bytes_read == bytes_to_read) { 6891 size_t bytes_written = bytes_read; 6892 error = core_file.get()->Write(bytes, bytes_written); 6893 bytes_left -= bytes_read; 6894 addr += bytes_read; 6895 } else { 6896 // Some pages within regions are not readable, those should 6897 // be zero filled 6898 memset(bytes, 0, bytes_to_read); 6899 size_t bytes_written = bytes_to_read; 6900 error = core_file.get()->Write(bytes, bytes_written); 6901 bytes_left -= bytes_to_read; 6902 addr += bytes_to_read; 6903 } 6904 } 6905 } 6906 } 6907 } 6908 } 6909 } 6910 return true; // This is the right plug to handle saving core files for 6911 // this process 6912 } 6913 return false; 6914 } 6915 6916 ObjectFileMachO::MachOCorefileAllImageInfos 6917 ObjectFileMachO::GetCorefileAllImageInfos() { 6918 MachOCorefileAllImageInfos image_infos; 6919 Log *log(GetLog(LLDBLog::Object | LLDBLog::Symbols | LLDBLog::Process | 6920 LLDBLog::DynamicLoader)); 6921 6922 auto lc_notes = FindLC_NOTEByName("all image infos"); 6923 for (auto lc_note : lc_notes) { 6924 offset_t payload_offset = std::get<0>(lc_note); 6925 // Read the struct all_image_infos_header. 6926 uint32_t version = m_data.GetU32(&payload_offset); 6927 if (version != 1) { 6928 return image_infos; 6929 } 6930 uint32_t imgcount = m_data.GetU32(&payload_offset); 6931 uint64_t entries_fileoff = m_data.GetU64(&payload_offset); 6932 // 'entries_size' is not used, nor is the 'unused' entry. 6933 // offset += 4; // uint32_t entries_size; 6934 // offset += 4; // uint32_t unused; 6935 6936 LLDB_LOGF(log, "LC_NOTE 'all image infos' found version %d with %d images", 6937 version, imgcount); 6938 payload_offset = entries_fileoff; 6939 for (uint32_t i = 0; i < imgcount; i++) { 6940 // Read the struct image_entry. 6941 offset_t filepath_offset = m_data.GetU64(&payload_offset); 6942 uuid_t uuid; 6943 memcpy(&uuid, m_data.GetData(&payload_offset, sizeof(uuid_t)), 6944 sizeof(uuid_t)); 6945 uint64_t load_address = m_data.GetU64(&payload_offset); 6946 offset_t seg_addrs_offset = m_data.GetU64(&payload_offset); 6947 uint32_t segment_count = m_data.GetU32(&payload_offset); 6948 uint32_t currently_executing = m_data.GetU32(&payload_offset); 6949 6950 MachOCorefileImageEntry image_entry; 6951 image_entry.filename = (const char *)m_data.GetCStr(&filepath_offset); 6952 image_entry.uuid = UUID(uuid, sizeof(uuid_t)); 6953 image_entry.load_address = load_address; 6954 image_entry.currently_executing = currently_executing; 6955 6956 offset_t seg_vmaddrs_offset = seg_addrs_offset; 6957 for (uint32_t j = 0; j < segment_count; j++) { 6958 char segname[17]; 6959 m_data.CopyData(seg_vmaddrs_offset, 16, segname); 6960 segname[16] = '\0'; 6961 seg_vmaddrs_offset += 16; 6962 uint64_t vmaddr = m_data.GetU64(&seg_vmaddrs_offset); 6963 seg_vmaddrs_offset += 8; /* unused */ 6964 6965 std::tuple<ConstString, addr_t> new_seg{ConstString(segname), vmaddr}; 6966 image_entry.segment_load_addresses.push_back(new_seg); 6967 } 6968 LLDB_LOGF(log, " image entry: %s %s 0x%" PRIx64 " %s", 6969 image_entry.filename.c_str(), 6970 image_entry.uuid.GetAsString().c_str(), 6971 image_entry.load_address, 6972 image_entry.currently_executing ? "currently executing" 6973 : "not currently executing"); 6974 image_infos.all_image_infos.push_back(image_entry); 6975 } 6976 } 6977 6978 lc_notes = FindLC_NOTEByName("load binary"); 6979 for (auto lc_note : lc_notes) { 6980 offset_t payload_offset = std::get<0>(lc_note); 6981 uint32_t version = m_data.GetU32(&payload_offset); 6982 if (version == 1) { 6983 uuid_t uuid; 6984 memcpy(&uuid, m_data.GetData(&payload_offset, sizeof(uuid_t)), 6985 sizeof(uuid_t)); 6986 uint64_t load_address = m_data.GetU64(&payload_offset); 6987 uint64_t slide = m_data.GetU64(&payload_offset); 6988 std::string filename = m_data.GetCStr(&payload_offset); 6989 6990 MachOCorefileImageEntry image_entry; 6991 image_entry.filename = filename; 6992 image_entry.uuid = UUID(uuid, sizeof(uuid_t)); 6993 image_entry.load_address = load_address; 6994 image_entry.slide = slide; 6995 image_entry.currently_executing = true; 6996 image_infos.all_image_infos.push_back(image_entry); 6997 LLDB_LOGF(log, 6998 "LC_NOTE 'load binary' found, filename %s uuid %s load " 6999 "address 0x%" PRIx64 " slide 0x%" PRIx64, 7000 filename.c_str(), 7001 image_entry.uuid.IsValid() 7002 ? image_entry.uuid.GetAsString().c_str() 7003 : "00000000-0000-0000-0000-000000000000", 7004 load_address, slide); 7005 } 7006 } 7007 7008 return image_infos; 7009 } 7010 7011 bool ObjectFileMachO::LoadCoreFileImages(lldb_private::Process &process) { 7012 MachOCorefileAllImageInfos image_infos = GetCorefileAllImageInfos(); 7013 Log *log = GetLog(LLDBLog::Object | LLDBLog::DynamicLoader); 7014 Status error; 7015 7016 bool found_platform_binary = false; 7017 ModuleList added_modules; 7018 for (MachOCorefileImageEntry &image : image_infos.all_image_infos) { 7019 ModuleSP module_sp, local_filesystem_module_sp; 7020 7021 // If this is a platform binary, it has been loaded (or registered with 7022 // the DynamicLoader to be loaded), we don't need to do any further 7023 // processing. We're not going to call ModulesDidLoad on this in this 7024 // method, so notify==true. 7025 if (process.GetTarget() 7026 .GetDebugger() 7027 .GetPlatformList() 7028 .LoadPlatformBinaryAndSetup(&process, image.load_address, 7029 true /* notify */)) { 7030 LLDB_LOGF(log, 7031 "ObjectFileMachO::%s binary at 0x%" PRIx64 7032 " is a platform binary, has been handled by a Platform plugin.", 7033 __FUNCTION__, image.load_address); 7034 continue; 7035 } 7036 7037 bool value_is_offset = image.load_address == LLDB_INVALID_ADDRESS; 7038 uint64_t value = value_is_offset ? image.slide : image.load_address; 7039 if (value_is_offset && value == LLDB_INVALID_ADDRESS) { 7040 // We have neither address nor slide; so we will find the binary 7041 // by UUID and load it at slide/offset 0. 7042 value = 0; 7043 } 7044 7045 // We have either a UUID, or we have a load address which 7046 // and can try to read load commands and find a UUID. 7047 if (image.uuid.IsValid() || 7048 (!value_is_offset && value != LLDB_INVALID_ADDRESS)) { 7049 const bool set_load_address = image.segment_load_addresses.size() == 0; 7050 const bool notify = false; 7051 // Userland Darwin binaries will have segment load addresses via 7052 // the `all image infos` LC_NOTE. 7053 const bool allow_memory_image_last_resort = 7054 image.segment_load_addresses.size(); 7055 module_sp = DynamicLoader::LoadBinaryWithUUIDAndAddress( 7056 &process, image.filename, image.uuid, value, value_is_offset, 7057 image.currently_executing, notify, set_load_address, 7058 allow_memory_image_last_resort); 7059 } 7060 7061 // We have a ModuleSP to load in the Target. Load it at the 7062 // correct address/slide and notify/load scripting resources. 7063 if (module_sp) { 7064 added_modules.Append(module_sp, false /* notify */); 7065 7066 // We have a list of segment load address 7067 if (image.segment_load_addresses.size() > 0) { 7068 if (log) { 7069 std::string uuidstr = image.uuid.GetAsString(); 7070 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 7071 "UUID %s with section load addresses", 7072 module_sp->GetFileSpec().GetPath().c_str(), 7073 uuidstr.c_str()); 7074 } 7075 for (auto name_vmaddr_tuple : image.segment_load_addresses) { 7076 SectionList *sectlist = module_sp->GetObjectFile()->GetSectionList(); 7077 if (sectlist) { 7078 SectionSP sect_sp = 7079 sectlist->FindSectionByName(std::get<0>(name_vmaddr_tuple)); 7080 if (sect_sp) { 7081 process.GetTarget().SetSectionLoadAddress( 7082 sect_sp, std::get<1>(name_vmaddr_tuple)); 7083 } 7084 } 7085 } 7086 } else { 7087 if (log) { 7088 std::string uuidstr = image.uuid.GetAsString(); 7089 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 7090 "UUID %s with %s 0x%" PRIx64, 7091 module_sp->GetFileSpec().GetPath().c_str(), 7092 uuidstr.c_str(), 7093 value_is_offset ? "slide" : "load address", value); 7094 } 7095 bool changed; 7096 module_sp->SetLoadAddress(process.GetTarget(), value, value_is_offset, 7097 changed); 7098 } 7099 } 7100 } 7101 if (added_modules.GetSize() > 0) { 7102 process.GetTarget().ModulesDidLoad(added_modules); 7103 process.Flush(); 7104 return true; 7105 } 7106 // Return true if the only binary we found was the platform binary, 7107 // and it was loaded outside the scope of this method. 7108 if (found_platform_binary) 7109 return true; 7110 7111 // No binaries. 7112 return false; 7113 } 7114