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/FileSpecList.h" 18 #include "lldb/Core/Module.h" 19 #include "lldb/Core/ModuleSpec.h" 20 #include "lldb/Core/PluginManager.h" 21 #include "lldb/Core/Progress.h" 22 #include "lldb/Core/Section.h" 23 #include "lldb/Core/StreamFile.h" 24 #include "lldb/Host/Host.h" 25 #include "lldb/Symbol/DWARFCallFrameInfo.h" 26 #include "lldb/Symbol/LocateSymbolFile.h" 27 #include "lldb/Symbol/ObjectFile.h" 28 #include "lldb/Target/DynamicLoader.h" 29 #include "lldb/Target/MemoryRegionInfo.h" 30 #include "lldb/Target/Platform.h" 31 #include "lldb/Target/Process.h" 32 #include "lldb/Target/SectionLoadList.h" 33 #include "lldb/Target/Target.h" 34 #include "lldb/Target/Thread.h" 35 #include "lldb/Target/ThreadList.h" 36 #include "lldb/Utility/ArchSpec.h" 37 #include "lldb/Utility/DataBuffer.h" 38 #include "lldb/Utility/FileSpec.h" 39 #include "lldb/Utility/LLDBLog.h" 40 #include "lldb/Utility/Log.h" 41 #include "lldb/Utility/RangeMap.h" 42 #include "lldb/Utility/RegisterValue.h" 43 #include "lldb/Utility/Status.h" 44 #include "lldb/Utility/StreamString.h" 45 #include "lldb/Utility/Timer.h" 46 #include "lldb/Utility/UUID.h" 47 48 #include "lldb/Host/SafeMachO.h" 49 50 #include "llvm/ADT/DenseSet.h" 51 #include "llvm/Support/FormatVariadic.h" 52 #include "llvm/Support/MemoryBuffer.h" 53 54 #include "ObjectFileMachO.h" 55 56 #if defined(__APPLE__) 57 #include <TargetConditionals.h> 58 // GetLLDBSharedCacheUUID() needs to call dlsym() 59 #include <dlfcn.h> 60 #include <mach/mach_init.h> 61 #include <mach/vm_map.h> 62 #include <lldb/Host/SafeMachO.h> 63 #endif 64 65 #ifndef __APPLE__ 66 #include "Utility/UuidCompatibility.h" 67 #else 68 #include <uuid/uuid.h> 69 #endif 70 71 #include <bitset> 72 #include <memory> 73 74 // Unfortunately the signpost header pulls in the system MachO header, too. 75 #ifdef CPU_TYPE_ARM 76 #undef CPU_TYPE_ARM 77 #endif 78 #ifdef CPU_TYPE_ARM64 79 #undef CPU_TYPE_ARM64 80 #endif 81 #ifdef CPU_TYPE_ARM64_32 82 #undef CPU_TYPE_ARM64_32 83 #endif 84 #ifdef CPU_TYPE_I386 85 #undef CPU_TYPE_I386 86 #endif 87 #ifdef CPU_TYPE_X86_64 88 #undef CPU_TYPE_X86_64 89 #endif 90 #ifdef MH_DYLINKER 91 #undef MH_DYLINKER 92 #endif 93 #ifdef MH_OBJECT 94 #undef MH_OBJECT 95 #endif 96 #ifdef LC_VERSION_MIN_MACOSX 97 #undef LC_VERSION_MIN_MACOSX 98 #endif 99 #ifdef LC_VERSION_MIN_IPHONEOS 100 #undef LC_VERSION_MIN_IPHONEOS 101 #endif 102 #ifdef LC_VERSION_MIN_TVOS 103 #undef LC_VERSION_MIN_TVOS 104 #endif 105 #ifdef LC_VERSION_MIN_WATCHOS 106 #undef LC_VERSION_MIN_WATCHOS 107 #endif 108 #ifdef LC_BUILD_VERSION 109 #undef LC_BUILD_VERSION 110 #endif 111 #ifdef PLATFORM_MACOS 112 #undef PLATFORM_MACOS 113 #endif 114 #ifdef PLATFORM_MACCATALYST 115 #undef PLATFORM_MACCATALYST 116 #endif 117 #ifdef PLATFORM_IOS 118 #undef PLATFORM_IOS 119 #endif 120 #ifdef PLATFORM_IOSSIMULATOR 121 #undef PLATFORM_IOSSIMULATOR 122 #endif 123 #ifdef PLATFORM_TVOS 124 #undef PLATFORM_TVOS 125 #endif 126 #ifdef PLATFORM_TVOSSIMULATOR 127 #undef PLATFORM_TVOSSIMULATOR 128 #endif 129 #ifdef PLATFORM_WATCHOS 130 #undef PLATFORM_WATCHOS 131 #endif 132 #ifdef PLATFORM_WATCHOSSIMULATOR 133 #undef PLATFORM_WATCHOSSIMULATOR 134 #endif 135 136 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 137 using namespace lldb; 138 using namespace lldb_private; 139 using namespace llvm::MachO; 140 141 LLDB_PLUGIN_DEFINE(ObjectFileMachO) 142 143 // Some structure definitions needed for parsing the dyld shared cache files 144 // found on iOS devices. 145 146 struct lldb_copy_dyld_cache_header_v1 { 147 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 148 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 149 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 150 uint32_t imagesOffset; 151 uint32_t imagesCount; 152 uint64_t dyldBaseAddress; 153 uint64_t codeSignatureOffset; 154 uint64_t codeSignatureSize; 155 uint64_t slideInfoOffset; 156 uint64_t slideInfoSize; 157 uint64_t localSymbolsOffset; 158 uint64_t localSymbolsSize; 159 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 160 // and later 161 }; 162 163 static void PrintRegisterValue(RegisterContext *reg_ctx, const char *name, 164 const char *alt_name, size_t reg_byte_size, 165 Stream &data) { 166 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 167 if (reg_info == nullptr) 168 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 169 if (reg_info) { 170 lldb_private::RegisterValue reg_value; 171 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 172 if (reg_info->byte_size >= reg_byte_size) 173 data.Write(reg_value.GetBytes(), reg_byte_size); 174 else { 175 data.Write(reg_value.GetBytes(), reg_info->byte_size); 176 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; ++i) 177 data.PutChar(0); 178 } 179 return; 180 } 181 } 182 // Just write zeros if all else fails 183 for (size_t i = 0; i < reg_byte_size; ++i) 184 data.PutChar(0); 185 } 186 187 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 188 public: 189 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 190 const DataExtractor &data) 191 : RegisterContextDarwin_x86_64(thread, 0) { 192 SetRegisterDataFrom_LC_THREAD(data); 193 } 194 195 void InvalidateAllRegisters() override { 196 // Do nothing... registers are always valid... 197 } 198 199 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 200 lldb::offset_t offset = 0; 201 SetError(GPRRegSet, Read, -1); 202 SetError(FPURegSet, Read, -1); 203 SetError(EXCRegSet, Read, -1); 204 bool done = false; 205 206 while (!done) { 207 int flavor = data.GetU32(&offset); 208 if (flavor == 0) 209 done = true; 210 else { 211 uint32_t i; 212 uint32_t count = data.GetU32(&offset); 213 switch (flavor) { 214 case GPRRegSet: 215 for (i = 0; i < count; ++i) 216 (&gpr.rax)[i] = data.GetU64(&offset); 217 SetError(GPRRegSet, Read, 0); 218 done = true; 219 220 break; 221 case FPURegSet: 222 // TODO: fill in FPU regs.... 223 // SetError (FPURegSet, Read, -1); 224 done = true; 225 226 break; 227 case EXCRegSet: 228 exc.trapno = data.GetU32(&offset); 229 exc.err = data.GetU32(&offset); 230 exc.faultvaddr = data.GetU64(&offset); 231 SetError(EXCRegSet, Read, 0); 232 done = true; 233 break; 234 case 7: 235 case 8: 236 case 9: 237 // fancy flavors that encapsulate of the above flavors... 238 break; 239 240 default: 241 done = true; 242 break; 243 } 244 } 245 } 246 } 247 248 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 249 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 250 if (reg_ctx_sp) { 251 RegisterContext *reg_ctx = reg_ctx_sp.get(); 252 253 data.PutHex32(GPRRegSet); // Flavor 254 data.PutHex32(GPRWordCount); 255 PrintRegisterValue(reg_ctx, "rax", nullptr, 8, data); 256 PrintRegisterValue(reg_ctx, "rbx", nullptr, 8, data); 257 PrintRegisterValue(reg_ctx, "rcx", nullptr, 8, data); 258 PrintRegisterValue(reg_ctx, "rdx", nullptr, 8, data); 259 PrintRegisterValue(reg_ctx, "rdi", nullptr, 8, data); 260 PrintRegisterValue(reg_ctx, "rsi", nullptr, 8, data); 261 PrintRegisterValue(reg_ctx, "rbp", nullptr, 8, data); 262 PrintRegisterValue(reg_ctx, "rsp", nullptr, 8, data); 263 PrintRegisterValue(reg_ctx, "r8", nullptr, 8, data); 264 PrintRegisterValue(reg_ctx, "r9", nullptr, 8, data); 265 PrintRegisterValue(reg_ctx, "r10", nullptr, 8, data); 266 PrintRegisterValue(reg_ctx, "r11", nullptr, 8, data); 267 PrintRegisterValue(reg_ctx, "r12", nullptr, 8, data); 268 PrintRegisterValue(reg_ctx, "r13", nullptr, 8, data); 269 PrintRegisterValue(reg_ctx, "r14", nullptr, 8, data); 270 PrintRegisterValue(reg_ctx, "r15", nullptr, 8, data); 271 PrintRegisterValue(reg_ctx, "rip", nullptr, 8, data); 272 PrintRegisterValue(reg_ctx, "rflags", nullptr, 8, data); 273 PrintRegisterValue(reg_ctx, "cs", nullptr, 8, data); 274 PrintRegisterValue(reg_ctx, "fs", nullptr, 8, data); 275 PrintRegisterValue(reg_ctx, "gs", nullptr, 8, data); 276 277 // // Write out the FPU registers 278 // const size_t fpu_byte_size = sizeof(FPU); 279 // size_t bytes_written = 0; 280 // data.PutHex32 (FPURegSet); 281 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 282 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 283 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 284 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 285 // data); // uint16_t fcw; // "fctrl" 286 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 287 // data); // uint16_t fsw; // "fstat" 288 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 289 // data); // uint8_t ftw; // "ftag" 290 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 291 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 292 // data); // uint16_t fop; // "fop" 293 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 294 // data); // uint32_t ip; // "fioff" 295 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 296 // data); // uint16_t cs; // "fiseg" 297 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 298 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 299 // data); // uint32_t dp; // "fooff" 300 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 301 // data); // uint16_t ds; // "foseg" 302 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 303 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 304 // data); // uint32_t mxcsr; 305 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 306 // 4, data);// uint32_t mxcsrmask; 307 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 308 // sizeof(MMSReg), data); 309 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 310 // sizeof(MMSReg), data); 311 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 312 // sizeof(MMSReg), data); 313 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 314 // sizeof(MMSReg), data); 315 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 316 // sizeof(MMSReg), data); 317 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 318 // sizeof(MMSReg), data); 319 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 320 // sizeof(MMSReg), data); 321 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 322 // sizeof(MMSReg), data); 323 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 324 // sizeof(XMMReg), data); 325 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 326 // sizeof(XMMReg), data); 327 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 328 // sizeof(XMMReg), data); 329 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 330 // sizeof(XMMReg), data); 331 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 332 // sizeof(XMMReg), data); 333 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 334 // sizeof(XMMReg), data); 335 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 336 // sizeof(XMMReg), data); 337 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 338 // sizeof(XMMReg), data); 339 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 340 // sizeof(XMMReg), data); 341 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 342 // sizeof(XMMReg), data); 343 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 344 // sizeof(XMMReg), data); 345 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 346 // sizeof(XMMReg), data); 347 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 348 // sizeof(XMMReg), data); 349 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 350 // sizeof(XMMReg), data); 351 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 352 // sizeof(XMMReg), data); 353 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 354 // sizeof(XMMReg), data); 355 // 356 // // Fill rest with zeros 357 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 358 // i) 359 // data.PutChar(0); 360 361 // Write out the EXC registers 362 data.PutHex32(EXCRegSet); 363 data.PutHex32(EXCWordCount); 364 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 365 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 366 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 8, data); 367 return true; 368 } 369 return false; 370 } 371 372 protected: 373 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 374 375 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 376 377 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 378 379 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 380 return 0; 381 } 382 383 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 384 return 0; 385 } 386 387 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 388 return 0; 389 } 390 }; 391 392 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 393 public: 394 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 395 const DataExtractor &data) 396 : RegisterContextDarwin_i386(thread, 0) { 397 SetRegisterDataFrom_LC_THREAD(data); 398 } 399 400 void InvalidateAllRegisters() override { 401 // Do nothing... registers are always valid... 402 } 403 404 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 405 lldb::offset_t offset = 0; 406 SetError(GPRRegSet, Read, -1); 407 SetError(FPURegSet, Read, -1); 408 SetError(EXCRegSet, Read, -1); 409 bool done = false; 410 411 while (!done) { 412 int flavor = data.GetU32(&offset); 413 if (flavor == 0) 414 done = true; 415 else { 416 uint32_t i; 417 uint32_t count = data.GetU32(&offset); 418 switch (flavor) { 419 case GPRRegSet: 420 for (i = 0; i < count; ++i) 421 (&gpr.eax)[i] = data.GetU32(&offset); 422 SetError(GPRRegSet, Read, 0); 423 done = true; 424 425 break; 426 case FPURegSet: 427 // TODO: fill in FPU regs.... 428 // SetError (FPURegSet, Read, -1); 429 done = true; 430 431 break; 432 case EXCRegSet: 433 exc.trapno = data.GetU32(&offset); 434 exc.err = data.GetU32(&offset); 435 exc.faultvaddr = data.GetU32(&offset); 436 SetError(EXCRegSet, Read, 0); 437 done = true; 438 break; 439 case 7: 440 case 8: 441 case 9: 442 // fancy flavors that encapsulate of the above flavors... 443 break; 444 445 default: 446 done = true; 447 break; 448 } 449 } 450 } 451 } 452 453 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 454 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 455 if (reg_ctx_sp) { 456 RegisterContext *reg_ctx = reg_ctx_sp.get(); 457 458 data.PutHex32(GPRRegSet); // Flavor 459 data.PutHex32(GPRWordCount); 460 PrintRegisterValue(reg_ctx, "eax", nullptr, 4, data); 461 PrintRegisterValue(reg_ctx, "ebx", nullptr, 4, data); 462 PrintRegisterValue(reg_ctx, "ecx", nullptr, 4, data); 463 PrintRegisterValue(reg_ctx, "edx", nullptr, 4, data); 464 PrintRegisterValue(reg_ctx, "edi", nullptr, 4, data); 465 PrintRegisterValue(reg_ctx, "esi", nullptr, 4, data); 466 PrintRegisterValue(reg_ctx, "ebp", nullptr, 4, data); 467 PrintRegisterValue(reg_ctx, "esp", nullptr, 4, data); 468 PrintRegisterValue(reg_ctx, "ss", nullptr, 4, data); 469 PrintRegisterValue(reg_ctx, "eflags", nullptr, 4, data); 470 PrintRegisterValue(reg_ctx, "eip", nullptr, 4, data); 471 PrintRegisterValue(reg_ctx, "cs", nullptr, 4, data); 472 PrintRegisterValue(reg_ctx, "ds", nullptr, 4, data); 473 PrintRegisterValue(reg_ctx, "es", nullptr, 4, data); 474 PrintRegisterValue(reg_ctx, "fs", nullptr, 4, data); 475 PrintRegisterValue(reg_ctx, "gs", nullptr, 4, data); 476 477 // Write out the EXC registers 478 data.PutHex32(EXCRegSet); 479 data.PutHex32(EXCWordCount); 480 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 481 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 482 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 4, data); 483 return true; 484 } 485 return false; 486 } 487 488 protected: 489 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 490 491 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 492 493 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 494 495 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 496 return 0; 497 } 498 499 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 500 return 0; 501 } 502 503 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 504 return 0; 505 } 506 }; 507 508 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 509 public: 510 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 511 const DataExtractor &data) 512 : RegisterContextDarwin_arm(thread, 0) { 513 SetRegisterDataFrom_LC_THREAD(data); 514 } 515 516 void InvalidateAllRegisters() override { 517 // Do nothing... registers are always valid... 518 } 519 520 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 521 lldb::offset_t offset = 0; 522 SetError(GPRRegSet, Read, -1); 523 SetError(FPURegSet, Read, -1); 524 SetError(EXCRegSet, Read, -1); 525 bool done = false; 526 527 while (!done) { 528 int flavor = data.GetU32(&offset); 529 uint32_t count = data.GetU32(&offset); 530 lldb::offset_t next_thread_state = offset + (count * 4); 531 switch (flavor) { 532 case GPRAltRegSet: 533 case GPRRegSet: 534 // On ARM, the CPSR register is also included in the count but it is 535 // not included in gpr.r so loop until (count-1). 536 for (uint32_t i = 0; i < (count - 1); ++i) { 537 gpr.r[i] = data.GetU32(&offset); 538 } 539 // Save cpsr explicitly. 540 gpr.cpsr = data.GetU32(&offset); 541 542 SetError(GPRRegSet, Read, 0); 543 offset = next_thread_state; 544 break; 545 546 case FPURegSet: { 547 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 548 const int fpu_reg_buf_size = sizeof(fpu.floats); 549 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 550 fpu_reg_buf) == fpu_reg_buf_size) { 551 offset += fpu_reg_buf_size; 552 fpu.fpscr = data.GetU32(&offset); 553 SetError(FPURegSet, Read, 0); 554 } else { 555 done = true; 556 } 557 } 558 offset = next_thread_state; 559 break; 560 561 case EXCRegSet: 562 if (count == 3) { 563 exc.exception = data.GetU32(&offset); 564 exc.fsr = data.GetU32(&offset); 565 exc.far = data.GetU32(&offset); 566 SetError(EXCRegSet, Read, 0); 567 } 568 done = true; 569 offset = next_thread_state; 570 break; 571 572 // Unknown register set flavor, stop trying to parse. 573 default: 574 done = true; 575 } 576 } 577 } 578 579 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 580 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 581 if (reg_ctx_sp) { 582 RegisterContext *reg_ctx = reg_ctx_sp.get(); 583 584 data.PutHex32(GPRRegSet); // Flavor 585 data.PutHex32(GPRWordCount); 586 PrintRegisterValue(reg_ctx, "r0", nullptr, 4, data); 587 PrintRegisterValue(reg_ctx, "r1", nullptr, 4, data); 588 PrintRegisterValue(reg_ctx, "r2", nullptr, 4, data); 589 PrintRegisterValue(reg_ctx, "r3", nullptr, 4, data); 590 PrintRegisterValue(reg_ctx, "r4", nullptr, 4, data); 591 PrintRegisterValue(reg_ctx, "r5", nullptr, 4, data); 592 PrintRegisterValue(reg_ctx, "r6", nullptr, 4, data); 593 PrintRegisterValue(reg_ctx, "r7", nullptr, 4, data); 594 PrintRegisterValue(reg_ctx, "r8", nullptr, 4, data); 595 PrintRegisterValue(reg_ctx, "r9", nullptr, 4, data); 596 PrintRegisterValue(reg_ctx, "r10", nullptr, 4, data); 597 PrintRegisterValue(reg_ctx, "r11", nullptr, 4, data); 598 PrintRegisterValue(reg_ctx, "r12", nullptr, 4, data); 599 PrintRegisterValue(reg_ctx, "sp", nullptr, 4, data); 600 PrintRegisterValue(reg_ctx, "lr", nullptr, 4, data); 601 PrintRegisterValue(reg_ctx, "pc", nullptr, 4, data); 602 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 603 604 // Write out the EXC registers 605 // data.PutHex32 (EXCRegSet); 606 // data.PutHex32 (EXCWordCount); 607 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 608 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 609 // WriteRegister (reg_ctx, "far", NULL, 4, data); 610 return true; 611 } 612 return false; 613 } 614 615 protected: 616 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 617 618 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 619 620 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 621 622 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 623 624 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 625 return 0; 626 } 627 628 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 629 return 0; 630 } 631 632 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 633 return 0; 634 } 635 636 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 637 return -1; 638 } 639 }; 640 641 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 642 public: 643 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 644 const DataExtractor &data) 645 : RegisterContextDarwin_arm64(thread, 0) { 646 SetRegisterDataFrom_LC_THREAD(data); 647 } 648 649 void InvalidateAllRegisters() override { 650 // Do nothing... registers are always valid... 651 } 652 653 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 654 lldb::offset_t offset = 0; 655 SetError(GPRRegSet, Read, -1); 656 SetError(FPURegSet, Read, -1); 657 SetError(EXCRegSet, Read, -1); 658 bool done = false; 659 while (!done) { 660 int flavor = data.GetU32(&offset); 661 uint32_t count = data.GetU32(&offset); 662 lldb::offset_t next_thread_state = offset + (count * 4); 663 switch (flavor) { 664 case GPRRegSet: 665 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 666 // 32-bit register) 667 if (count >= (33 * 2) + 1) { 668 for (uint32_t i = 0; i < 29; ++i) 669 gpr.x[i] = data.GetU64(&offset); 670 gpr.fp = data.GetU64(&offset); 671 gpr.lr = data.GetU64(&offset); 672 gpr.sp = data.GetU64(&offset); 673 gpr.pc = data.GetU64(&offset); 674 gpr.cpsr = data.GetU32(&offset); 675 SetError(GPRRegSet, Read, 0); 676 } 677 offset = next_thread_state; 678 break; 679 case FPURegSet: { 680 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 681 const int fpu_reg_buf_size = sizeof(fpu); 682 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 683 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 684 fpu_reg_buf) == fpu_reg_buf_size) { 685 SetError(FPURegSet, Read, 0); 686 } else { 687 done = true; 688 } 689 } 690 offset = next_thread_state; 691 break; 692 case EXCRegSet: 693 if (count == 4) { 694 exc.far = data.GetU64(&offset); 695 exc.esr = data.GetU32(&offset); 696 exc.exception = data.GetU32(&offset); 697 SetError(EXCRegSet, Read, 0); 698 } 699 offset = next_thread_state; 700 break; 701 default: 702 done = true; 703 break; 704 } 705 } 706 } 707 708 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 709 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 710 if (reg_ctx_sp) { 711 RegisterContext *reg_ctx = reg_ctx_sp.get(); 712 713 data.PutHex32(GPRRegSet); // Flavor 714 data.PutHex32(GPRWordCount); 715 PrintRegisterValue(reg_ctx, "x0", nullptr, 8, data); 716 PrintRegisterValue(reg_ctx, "x1", nullptr, 8, data); 717 PrintRegisterValue(reg_ctx, "x2", nullptr, 8, data); 718 PrintRegisterValue(reg_ctx, "x3", nullptr, 8, data); 719 PrintRegisterValue(reg_ctx, "x4", nullptr, 8, data); 720 PrintRegisterValue(reg_ctx, "x5", nullptr, 8, data); 721 PrintRegisterValue(reg_ctx, "x6", nullptr, 8, data); 722 PrintRegisterValue(reg_ctx, "x7", nullptr, 8, data); 723 PrintRegisterValue(reg_ctx, "x8", nullptr, 8, data); 724 PrintRegisterValue(reg_ctx, "x9", nullptr, 8, data); 725 PrintRegisterValue(reg_ctx, "x10", nullptr, 8, data); 726 PrintRegisterValue(reg_ctx, "x11", nullptr, 8, data); 727 PrintRegisterValue(reg_ctx, "x12", nullptr, 8, data); 728 PrintRegisterValue(reg_ctx, "x13", nullptr, 8, data); 729 PrintRegisterValue(reg_ctx, "x14", nullptr, 8, data); 730 PrintRegisterValue(reg_ctx, "x15", nullptr, 8, data); 731 PrintRegisterValue(reg_ctx, "x16", nullptr, 8, data); 732 PrintRegisterValue(reg_ctx, "x17", nullptr, 8, data); 733 PrintRegisterValue(reg_ctx, "x18", nullptr, 8, data); 734 PrintRegisterValue(reg_ctx, "x19", nullptr, 8, data); 735 PrintRegisterValue(reg_ctx, "x20", nullptr, 8, data); 736 PrintRegisterValue(reg_ctx, "x21", nullptr, 8, data); 737 PrintRegisterValue(reg_ctx, "x22", nullptr, 8, data); 738 PrintRegisterValue(reg_ctx, "x23", nullptr, 8, data); 739 PrintRegisterValue(reg_ctx, "x24", nullptr, 8, data); 740 PrintRegisterValue(reg_ctx, "x25", nullptr, 8, data); 741 PrintRegisterValue(reg_ctx, "x26", nullptr, 8, data); 742 PrintRegisterValue(reg_ctx, "x27", nullptr, 8, data); 743 PrintRegisterValue(reg_ctx, "x28", nullptr, 8, data); 744 PrintRegisterValue(reg_ctx, "fp", nullptr, 8, data); 745 PrintRegisterValue(reg_ctx, "lr", nullptr, 8, data); 746 PrintRegisterValue(reg_ctx, "sp", nullptr, 8, data); 747 PrintRegisterValue(reg_ctx, "pc", nullptr, 8, data); 748 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 749 data.PutHex32(0); // uint32_t pad at the end 750 751 // Write out the EXC registers 752 data.PutHex32(EXCRegSet); 753 data.PutHex32(EXCWordCount); 754 PrintRegisterValue(reg_ctx, "far", nullptr, 8, data); 755 PrintRegisterValue(reg_ctx, "esr", nullptr, 4, data); 756 PrintRegisterValue(reg_ctx, "exception", nullptr, 4, data); 757 return true; 758 } 759 return false; 760 } 761 762 protected: 763 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 764 765 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 766 767 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 768 769 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 770 771 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 772 return 0; 773 } 774 775 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 776 return 0; 777 } 778 779 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 780 return 0; 781 } 782 783 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 784 return -1; 785 } 786 }; 787 788 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 789 switch (magic) { 790 case MH_MAGIC: 791 case MH_CIGAM: 792 return sizeof(struct llvm::MachO::mach_header); 793 794 case MH_MAGIC_64: 795 case MH_CIGAM_64: 796 return sizeof(struct llvm::MachO::mach_header_64); 797 break; 798 799 default: 800 break; 801 } 802 return 0; 803 } 804 805 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 806 807 char ObjectFileMachO::ID; 808 809 void ObjectFileMachO::Initialize() { 810 PluginManager::RegisterPlugin( 811 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 812 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 813 } 814 815 void ObjectFileMachO::Terminate() { 816 PluginManager::UnregisterPlugin(CreateInstance); 817 } 818 819 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 820 DataBufferSP &data_sp, 821 lldb::offset_t data_offset, 822 const FileSpec *file, 823 lldb::offset_t file_offset, 824 lldb::offset_t length) { 825 if (!data_sp) { 826 data_sp = MapFileData(*file, length, file_offset); 827 if (!data_sp) 828 return nullptr; 829 data_offset = 0; 830 } 831 832 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 833 return nullptr; 834 835 // Update the data to contain the entire file if it doesn't already 836 if (data_sp->GetByteSize() < length) { 837 data_sp = MapFileData(*file, length, file_offset); 838 if (!data_sp) 839 return nullptr; 840 data_offset = 0; 841 } 842 auto objfile_up = std::make_unique<ObjectFileMachO>( 843 module_sp, data_sp, data_offset, file, file_offset, length); 844 if (!objfile_up || !objfile_up->ParseHeader()) 845 return nullptr; 846 847 return objfile_up.release(); 848 } 849 850 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 851 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 852 const ProcessSP &process_sp, lldb::addr_t header_addr) { 853 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 854 std::unique_ptr<ObjectFile> objfile_up( 855 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 856 if (objfile_up.get() && objfile_up->ParseHeader()) 857 return objfile_up.release(); 858 } 859 return nullptr; 860 } 861 862 size_t ObjectFileMachO::GetModuleSpecifications( 863 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 864 lldb::offset_t data_offset, lldb::offset_t file_offset, 865 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 866 const size_t initial_count = specs.GetSize(); 867 868 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 869 DataExtractor data; 870 data.SetData(data_sp); 871 llvm::MachO::mach_header header; 872 if (ParseHeader(data, &data_offset, header)) { 873 size_t header_and_load_cmds = 874 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 875 if (header_and_load_cmds >= data_sp->GetByteSize()) { 876 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 877 data.SetData(data_sp); 878 data_offset = MachHeaderSizeFromMagic(header.magic); 879 } 880 if (data_sp) { 881 ModuleSpec base_spec; 882 base_spec.GetFileSpec() = file; 883 base_spec.SetObjectOffset(file_offset); 884 base_spec.SetObjectSize(length); 885 GetAllArchSpecs(header, data, data_offset, base_spec, specs); 886 } 887 } 888 } 889 return specs.GetSize() - initial_count; 890 } 891 892 ConstString ObjectFileMachO::GetSegmentNameTEXT() { 893 static ConstString g_segment_name_TEXT("__TEXT"); 894 return g_segment_name_TEXT; 895 } 896 897 ConstString ObjectFileMachO::GetSegmentNameDATA() { 898 static ConstString g_segment_name_DATA("__DATA"); 899 return g_segment_name_DATA; 900 } 901 902 ConstString ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 903 static ConstString g_segment_name("__DATA_DIRTY"); 904 return g_segment_name; 905 } 906 907 ConstString ObjectFileMachO::GetSegmentNameDATA_CONST() { 908 static ConstString g_segment_name("__DATA_CONST"); 909 return g_segment_name; 910 } 911 912 ConstString ObjectFileMachO::GetSegmentNameOBJC() { 913 static ConstString g_segment_name_OBJC("__OBJC"); 914 return g_segment_name_OBJC; 915 } 916 917 ConstString ObjectFileMachO::GetSegmentNameLINKEDIT() { 918 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 919 return g_section_name_LINKEDIT; 920 } 921 922 ConstString ObjectFileMachO::GetSegmentNameDWARF() { 923 static ConstString g_section_name("__DWARF"); 924 return g_section_name; 925 } 926 927 ConstString ObjectFileMachO::GetSectionNameEHFrame() { 928 static ConstString g_section_name_eh_frame("__eh_frame"); 929 return g_section_name_eh_frame; 930 } 931 932 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 933 lldb::addr_t data_offset, 934 lldb::addr_t data_length) { 935 DataExtractor data; 936 data.SetData(data_sp, data_offset, data_length); 937 lldb::offset_t offset = 0; 938 uint32_t magic = data.GetU32(&offset); 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_segments(), m_mach_sections(), m_entry_point_address(), 950 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 951 m_reexported_dylibs(), 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::DataBufferSP &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_segments(), m_mach_sections(), m_entry_point_address(), 962 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 963 m_reexported_dylibs(), 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 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1128 return m_data.GetAddressByteSize(); 1129 } 1130 1131 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1132 Symtab *symtab = GetSymtab(); 1133 if (!symtab) 1134 return AddressClass::eUnknown; 1135 1136 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1137 if (symbol) { 1138 if (symbol->ValueIsAddress()) { 1139 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1140 if (section_sp) { 1141 const lldb::SectionType section_type = section_sp->GetType(); 1142 switch (section_type) { 1143 case eSectionTypeInvalid: 1144 return AddressClass::eUnknown; 1145 1146 case eSectionTypeCode: 1147 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1148 // For ARM we have a bit in the n_desc field of the symbol that 1149 // tells us ARM/Thumb which is bit 0x0008. 1150 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1151 return AddressClass::eCodeAlternateISA; 1152 } 1153 return AddressClass::eCode; 1154 1155 case eSectionTypeContainer: 1156 return AddressClass::eUnknown; 1157 1158 case eSectionTypeData: 1159 case eSectionTypeDataCString: 1160 case eSectionTypeDataCStringPointers: 1161 case eSectionTypeDataSymbolAddress: 1162 case eSectionTypeData4: 1163 case eSectionTypeData8: 1164 case eSectionTypeData16: 1165 case eSectionTypeDataPointers: 1166 case eSectionTypeZeroFill: 1167 case eSectionTypeDataObjCMessageRefs: 1168 case eSectionTypeDataObjCCFStrings: 1169 case eSectionTypeGoSymtab: 1170 return AddressClass::eData; 1171 1172 case eSectionTypeDebug: 1173 case eSectionTypeDWARFDebugAbbrev: 1174 case eSectionTypeDWARFDebugAbbrevDwo: 1175 case eSectionTypeDWARFDebugAddr: 1176 case eSectionTypeDWARFDebugAranges: 1177 case eSectionTypeDWARFDebugCuIndex: 1178 case eSectionTypeDWARFDebugFrame: 1179 case eSectionTypeDWARFDebugInfo: 1180 case eSectionTypeDWARFDebugInfoDwo: 1181 case eSectionTypeDWARFDebugLine: 1182 case eSectionTypeDWARFDebugLineStr: 1183 case eSectionTypeDWARFDebugLoc: 1184 case eSectionTypeDWARFDebugLocDwo: 1185 case eSectionTypeDWARFDebugLocLists: 1186 case eSectionTypeDWARFDebugLocListsDwo: 1187 case eSectionTypeDWARFDebugMacInfo: 1188 case eSectionTypeDWARFDebugMacro: 1189 case eSectionTypeDWARFDebugNames: 1190 case eSectionTypeDWARFDebugPubNames: 1191 case eSectionTypeDWARFDebugPubTypes: 1192 case eSectionTypeDWARFDebugRanges: 1193 case eSectionTypeDWARFDebugRngLists: 1194 case eSectionTypeDWARFDebugRngListsDwo: 1195 case eSectionTypeDWARFDebugStr: 1196 case eSectionTypeDWARFDebugStrDwo: 1197 case eSectionTypeDWARFDebugStrOffsets: 1198 case eSectionTypeDWARFDebugStrOffsetsDwo: 1199 case eSectionTypeDWARFDebugTuIndex: 1200 case eSectionTypeDWARFDebugTypes: 1201 case eSectionTypeDWARFDebugTypesDwo: 1202 case eSectionTypeDWARFAppleNames: 1203 case eSectionTypeDWARFAppleTypes: 1204 case eSectionTypeDWARFAppleNamespaces: 1205 case eSectionTypeDWARFAppleObjC: 1206 case eSectionTypeDWARFGNUDebugAltLink: 1207 return AddressClass::eDebug; 1208 1209 case eSectionTypeEHFrame: 1210 case eSectionTypeARMexidx: 1211 case eSectionTypeARMextab: 1212 case eSectionTypeCompactUnwind: 1213 return AddressClass::eRuntime; 1214 1215 case eSectionTypeAbsoluteAddress: 1216 case eSectionTypeELFSymbolTable: 1217 case eSectionTypeELFDynamicSymbols: 1218 case eSectionTypeELFRelocationEntries: 1219 case eSectionTypeELFDynamicLinkInfo: 1220 case eSectionTypeOther: 1221 return AddressClass::eUnknown; 1222 } 1223 } 1224 } 1225 1226 const SymbolType symbol_type = symbol->GetType(); 1227 switch (symbol_type) { 1228 case eSymbolTypeAny: 1229 return AddressClass::eUnknown; 1230 case eSymbolTypeAbsolute: 1231 return AddressClass::eUnknown; 1232 1233 case eSymbolTypeCode: 1234 case eSymbolTypeTrampoline: 1235 case eSymbolTypeResolver: 1236 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1237 // For ARM we have a bit in the n_desc field of the symbol that tells 1238 // us ARM/Thumb which is bit 0x0008. 1239 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1240 return AddressClass::eCodeAlternateISA; 1241 } 1242 return AddressClass::eCode; 1243 1244 case eSymbolTypeData: 1245 return AddressClass::eData; 1246 case eSymbolTypeRuntime: 1247 return AddressClass::eRuntime; 1248 case eSymbolTypeException: 1249 return AddressClass::eRuntime; 1250 case eSymbolTypeSourceFile: 1251 return AddressClass::eDebug; 1252 case eSymbolTypeHeaderFile: 1253 return AddressClass::eDebug; 1254 case eSymbolTypeObjectFile: 1255 return AddressClass::eDebug; 1256 case eSymbolTypeCommonBlock: 1257 return AddressClass::eDebug; 1258 case eSymbolTypeBlock: 1259 return AddressClass::eDebug; 1260 case eSymbolTypeLocal: 1261 return AddressClass::eData; 1262 case eSymbolTypeParam: 1263 return AddressClass::eData; 1264 case eSymbolTypeVariable: 1265 return AddressClass::eData; 1266 case eSymbolTypeVariableType: 1267 return AddressClass::eDebug; 1268 case eSymbolTypeLineEntry: 1269 return AddressClass::eDebug; 1270 case eSymbolTypeLineHeader: 1271 return AddressClass::eDebug; 1272 case eSymbolTypeScopeBegin: 1273 return AddressClass::eDebug; 1274 case eSymbolTypeScopeEnd: 1275 return AddressClass::eDebug; 1276 case eSymbolTypeAdditional: 1277 return AddressClass::eUnknown; 1278 case eSymbolTypeCompiler: 1279 return AddressClass::eDebug; 1280 case eSymbolTypeInstrumentation: 1281 return AddressClass::eDebug; 1282 case eSymbolTypeUndefined: 1283 return AddressClass::eUnknown; 1284 case eSymbolTypeObjCClass: 1285 return AddressClass::eRuntime; 1286 case eSymbolTypeObjCMetaClass: 1287 return AddressClass::eRuntime; 1288 case eSymbolTypeObjCIVar: 1289 return AddressClass::eRuntime; 1290 case eSymbolTypeReExported: 1291 return AddressClass::eRuntime; 1292 } 1293 } 1294 return AddressClass::eUnknown; 1295 } 1296 1297 bool ObjectFileMachO::IsStripped() { 1298 if (m_dysymtab.cmd == 0) { 1299 ModuleSP module_sp(GetModule()); 1300 if (module_sp) { 1301 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1302 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1303 const lldb::offset_t load_cmd_offset = offset; 1304 1305 llvm::MachO::load_command lc; 1306 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 1307 break; 1308 if (lc.cmd == LC_DYSYMTAB) { 1309 m_dysymtab.cmd = lc.cmd; 1310 m_dysymtab.cmdsize = lc.cmdsize; 1311 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1312 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1313 nullptr) { 1314 // Clear m_dysymtab if we were unable to read all items from the 1315 // load command 1316 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1317 } 1318 } 1319 offset = load_cmd_offset + lc.cmdsize; 1320 } 1321 } 1322 } 1323 if (m_dysymtab.cmd) 1324 return m_dysymtab.nlocalsym <= 1; 1325 return false; 1326 } 1327 1328 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1329 EncryptedFileRanges result; 1330 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1331 1332 llvm::MachO::encryption_info_command encryption_cmd; 1333 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1334 const lldb::offset_t load_cmd_offset = offset; 1335 if (m_data.GetU32(&offset, &encryption_cmd, 2) == nullptr) 1336 break; 1337 1338 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the 1339 // 3 fields we care about, so treat them the same. 1340 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1341 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1342 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1343 if (encryption_cmd.cryptid != 0) { 1344 EncryptedFileRanges::Entry entry; 1345 entry.SetRangeBase(encryption_cmd.cryptoff); 1346 entry.SetByteSize(encryption_cmd.cryptsize); 1347 result.Append(entry); 1348 } 1349 } 1350 } 1351 offset = load_cmd_offset + encryption_cmd.cmdsize; 1352 } 1353 1354 return result; 1355 } 1356 1357 void ObjectFileMachO::SanitizeSegmentCommand( 1358 llvm::MachO::segment_command_64 &seg_cmd, uint32_t cmd_idx) { 1359 if (m_length == 0 || seg_cmd.filesize == 0) 1360 return; 1361 1362 if (IsSharedCacheBinary() && !IsInMemory()) { 1363 // In shared cache images, the load commands are relative to the 1364 // shared cache file, and not the specific image we are 1365 // examining. Let's fix this up so that it looks like a normal 1366 // image. 1367 if (strncmp(seg_cmd.segname, "__TEXT", sizeof(seg_cmd.segname)) == 0) 1368 m_text_address = seg_cmd.vmaddr; 1369 if (strncmp(seg_cmd.segname, "__LINKEDIT", sizeof(seg_cmd.segname)) == 0) 1370 m_linkedit_original_offset = seg_cmd.fileoff; 1371 1372 seg_cmd.fileoff = seg_cmd.vmaddr - m_text_address; 1373 } 1374 1375 if (seg_cmd.fileoff > m_length) { 1376 // We have a load command that says it extends past the end of the file. 1377 // This is likely a corrupt file. We don't have any way to return an error 1378 // condition here (this method was likely invoked from something like 1379 // ObjectFile::GetSectionList()), so we just null out the section contents, 1380 // and dump a message to stdout. The most common case here is core file 1381 // debugging with a truncated file. 1382 const char *lc_segment_name = 1383 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1384 GetModule()->ReportWarning( 1385 "load command %u %s has a fileoff (0x%" PRIx64 1386 ") that extends beyond the end of the file (0x%" PRIx64 1387 "), ignoring this section", 1388 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1389 1390 seg_cmd.fileoff = 0; 1391 seg_cmd.filesize = 0; 1392 } 1393 1394 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1395 // We have a load command that says it extends past the end of the file. 1396 // This is likely a corrupt file. We don't have any way to return an error 1397 // condition here (this method was likely invoked from something like 1398 // ObjectFile::GetSectionList()), so we just null out the section contents, 1399 // and dump a message to stdout. The most common case here is core file 1400 // debugging with a truncated file. 1401 const char *lc_segment_name = 1402 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1403 GetModule()->ReportWarning( 1404 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1405 ") that extends beyond the end of the file (0x%" PRIx64 1406 "), the segment will be truncated to match", 1407 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1408 1409 // Truncate the length 1410 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1411 } 1412 } 1413 1414 static uint32_t 1415 GetSegmentPermissions(const llvm::MachO::segment_command_64 &seg_cmd) { 1416 uint32_t result = 0; 1417 if (seg_cmd.initprot & VM_PROT_READ) 1418 result |= ePermissionsReadable; 1419 if (seg_cmd.initprot & VM_PROT_WRITE) 1420 result |= ePermissionsWritable; 1421 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1422 result |= ePermissionsExecutable; 1423 return result; 1424 } 1425 1426 static lldb::SectionType GetSectionType(uint32_t flags, 1427 ConstString section_name) { 1428 1429 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1430 return eSectionTypeCode; 1431 1432 uint32_t mach_sect_type = flags & SECTION_TYPE; 1433 static ConstString g_sect_name_objc_data("__objc_data"); 1434 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1435 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1436 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1437 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1438 static ConstString g_sect_name_objc_const("__objc_const"); 1439 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1440 static ConstString g_sect_name_cfstring("__cfstring"); 1441 1442 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1443 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1444 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1445 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1446 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1447 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1448 static ConstString g_sect_name_dwarf_debug_loclists("__debug_loclists"); 1449 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1450 static ConstString g_sect_name_dwarf_debug_names("__debug_names"); 1451 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1452 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1453 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1454 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1455 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1456 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1457 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1458 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1459 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1460 static ConstString g_sect_name_eh_frame("__eh_frame"); 1461 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1462 static ConstString g_sect_name_text("__text"); 1463 static ConstString g_sect_name_data("__data"); 1464 static ConstString g_sect_name_go_symtab("__gosymtab"); 1465 1466 if (section_name == g_sect_name_dwarf_debug_abbrev) 1467 return eSectionTypeDWARFDebugAbbrev; 1468 if (section_name == g_sect_name_dwarf_debug_aranges) 1469 return eSectionTypeDWARFDebugAranges; 1470 if (section_name == g_sect_name_dwarf_debug_frame) 1471 return eSectionTypeDWARFDebugFrame; 1472 if (section_name == g_sect_name_dwarf_debug_info) 1473 return eSectionTypeDWARFDebugInfo; 1474 if (section_name == g_sect_name_dwarf_debug_line) 1475 return eSectionTypeDWARFDebugLine; 1476 if (section_name == g_sect_name_dwarf_debug_loc) 1477 return eSectionTypeDWARFDebugLoc; 1478 if (section_name == g_sect_name_dwarf_debug_loclists) 1479 return eSectionTypeDWARFDebugLocLists; 1480 if (section_name == g_sect_name_dwarf_debug_macinfo) 1481 return eSectionTypeDWARFDebugMacInfo; 1482 if (section_name == g_sect_name_dwarf_debug_names) 1483 return eSectionTypeDWARFDebugNames; 1484 if (section_name == g_sect_name_dwarf_debug_pubnames) 1485 return eSectionTypeDWARFDebugPubNames; 1486 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1487 return eSectionTypeDWARFDebugPubTypes; 1488 if (section_name == g_sect_name_dwarf_debug_ranges) 1489 return eSectionTypeDWARFDebugRanges; 1490 if (section_name == g_sect_name_dwarf_debug_str) 1491 return eSectionTypeDWARFDebugStr; 1492 if (section_name == g_sect_name_dwarf_debug_types) 1493 return eSectionTypeDWARFDebugTypes; 1494 if (section_name == g_sect_name_dwarf_apple_names) 1495 return eSectionTypeDWARFAppleNames; 1496 if (section_name == g_sect_name_dwarf_apple_types) 1497 return eSectionTypeDWARFAppleTypes; 1498 if (section_name == g_sect_name_dwarf_apple_namespaces) 1499 return eSectionTypeDWARFAppleNamespaces; 1500 if (section_name == g_sect_name_dwarf_apple_objc) 1501 return eSectionTypeDWARFAppleObjC; 1502 if (section_name == g_sect_name_objc_selrefs) 1503 return eSectionTypeDataCStringPointers; 1504 if (section_name == g_sect_name_objc_msgrefs) 1505 return eSectionTypeDataObjCMessageRefs; 1506 if (section_name == g_sect_name_eh_frame) 1507 return eSectionTypeEHFrame; 1508 if (section_name == g_sect_name_compact_unwind) 1509 return eSectionTypeCompactUnwind; 1510 if (section_name == g_sect_name_cfstring) 1511 return eSectionTypeDataObjCCFStrings; 1512 if (section_name == g_sect_name_go_symtab) 1513 return eSectionTypeGoSymtab; 1514 if (section_name == g_sect_name_objc_data || 1515 section_name == g_sect_name_objc_classrefs || 1516 section_name == g_sect_name_objc_superrefs || 1517 section_name == g_sect_name_objc_const || 1518 section_name == g_sect_name_objc_classlist) { 1519 return eSectionTypeDataPointers; 1520 } 1521 1522 switch (mach_sect_type) { 1523 // TODO: categorize sections by other flags for regular sections 1524 case S_REGULAR: 1525 if (section_name == g_sect_name_text) 1526 return eSectionTypeCode; 1527 if (section_name == g_sect_name_data) 1528 return eSectionTypeData; 1529 return eSectionTypeOther; 1530 case S_ZEROFILL: 1531 return eSectionTypeZeroFill; 1532 case S_CSTRING_LITERALS: // section with only literal C strings 1533 return eSectionTypeDataCString; 1534 case S_4BYTE_LITERALS: // section with only 4 byte literals 1535 return eSectionTypeData4; 1536 case S_8BYTE_LITERALS: // section with only 8 byte literals 1537 return eSectionTypeData8; 1538 case S_LITERAL_POINTERS: // section with only pointers to literals 1539 return eSectionTypeDataPointers; 1540 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1541 return eSectionTypeDataPointers; 1542 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1543 return eSectionTypeDataPointers; 1544 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1545 // the reserved2 field 1546 return eSectionTypeCode; 1547 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1548 // initialization 1549 return eSectionTypeDataPointers; 1550 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1551 // termination 1552 return eSectionTypeDataPointers; 1553 case S_COALESCED: 1554 return eSectionTypeOther; 1555 case S_GB_ZEROFILL: 1556 return eSectionTypeZeroFill; 1557 case S_INTERPOSING: // section with only pairs of function pointers for 1558 // interposing 1559 return eSectionTypeCode; 1560 case S_16BYTE_LITERALS: // section with only 16 byte literals 1561 return eSectionTypeData16; 1562 case S_DTRACE_DOF: 1563 return eSectionTypeDebug; 1564 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1565 return eSectionTypeDataPointers; 1566 default: 1567 return eSectionTypeOther; 1568 } 1569 } 1570 1571 struct ObjectFileMachO::SegmentParsingContext { 1572 const EncryptedFileRanges EncryptedRanges; 1573 lldb_private::SectionList &UnifiedList; 1574 uint32_t NextSegmentIdx = 0; 1575 uint32_t NextSectionIdx = 0; 1576 bool FileAddressesChanged = false; 1577 1578 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1579 lldb_private::SectionList &UnifiedList) 1580 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1581 }; 1582 1583 void ObjectFileMachO::ProcessSegmentCommand( 1584 const llvm::MachO::load_command &load_cmd_, lldb::offset_t offset, 1585 uint32_t cmd_idx, SegmentParsingContext &context) { 1586 llvm::MachO::segment_command_64 load_cmd; 1587 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1588 1589 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1590 return; 1591 1592 ModuleSP module_sp = GetModule(); 1593 const bool is_core = GetType() == eTypeCoreFile; 1594 const bool is_dsym = (m_header.filetype == MH_DSYM); 1595 bool add_section = true; 1596 bool add_to_unified = true; 1597 ConstString const_segname( 1598 load_cmd.segname, strnlen(load_cmd.segname, sizeof(load_cmd.segname))); 1599 1600 SectionSP unified_section_sp( 1601 context.UnifiedList.FindSectionByName(const_segname)); 1602 if (is_dsym && unified_section_sp) { 1603 if (const_segname == GetSegmentNameLINKEDIT()) { 1604 // We need to keep the __LINKEDIT segment private to this object file 1605 // only 1606 add_to_unified = false; 1607 } else { 1608 // This is the dSYM file and this section has already been created by the 1609 // object file, no need to create it. 1610 add_section = false; 1611 } 1612 } 1613 load_cmd.vmaddr = m_data.GetAddress(&offset); 1614 load_cmd.vmsize = m_data.GetAddress(&offset); 1615 load_cmd.fileoff = m_data.GetAddress(&offset); 1616 load_cmd.filesize = m_data.GetAddress(&offset); 1617 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1618 return; 1619 1620 SanitizeSegmentCommand(load_cmd, cmd_idx); 1621 1622 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1623 const bool segment_is_encrypted = 1624 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1625 1626 // Keep a list of mach segments around in case we need to get at data that 1627 // isn't stored in the abstracted Sections. 1628 m_mach_segments.push_back(load_cmd); 1629 1630 // Use a segment ID of the segment index shifted left by 8 so they never 1631 // conflict with any of the sections. 1632 SectionSP segment_sp; 1633 if (add_section && (const_segname || is_core)) { 1634 segment_sp = std::make_shared<Section>( 1635 module_sp, // Module to which this section belongs 1636 this, // Object file to which this sections belongs 1637 ++context.NextSegmentIdx 1638 << 8, // Section ID is the 1 based segment index 1639 // shifted right by 8 bits as not to collide with any of the 256 1640 // section IDs that are possible 1641 const_segname, // Name of this section 1642 eSectionTypeContainer, // This section is a container of other 1643 // sections. 1644 load_cmd.vmaddr, // File VM address == addresses as they are 1645 // found in the object file 1646 load_cmd.vmsize, // VM size in bytes of this section 1647 load_cmd.fileoff, // Offset to the data for this section in 1648 // the file 1649 load_cmd.filesize, // Size in bytes of this section as found 1650 // in the file 1651 0, // Segments have no alignment information 1652 load_cmd.flags); // Flags for this section 1653 1654 segment_sp->SetIsEncrypted(segment_is_encrypted); 1655 m_sections_up->AddSection(segment_sp); 1656 segment_sp->SetPermissions(segment_permissions); 1657 if (add_to_unified) 1658 context.UnifiedList.AddSection(segment_sp); 1659 } else if (unified_section_sp) { 1660 // If this is a dSYM and the file addresses in the dSYM differ from the 1661 // file addresses in the ObjectFile, we must use the file base address for 1662 // the Section from the dSYM for the DWARF to resolve correctly. 1663 // This only happens with binaries in the shared cache in practice; 1664 // normally a mismatch like this would give a binary & dSYM that do not 1665 // match UUIDs. When a binary is included in the shared cache, its 1666 // segments are rearranged to optimize the shared cache, so its file 1667 // addresses will differ from what the ObjectFile had originally, 1668 // and what the dSYM has. 1669 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1670 Log *log = GetLog(LLDBLog::Symbols); 1671 if (log) { 1672 log->Printf( 1673 "Installing dSYM's %s segment file address over ObjectFile's " 1674 "so symbol table/debug info resolves correctly for %s", 1675 const_segname.AsCString(), 1676 module_sp->GetFileSpec().GetFilename().AsCString()); 1677 } 1678 1679 // Make sure we've parsed the symbol table from the ObjectFile before 1680 // we go around changing its Sections. 1681 module_sp->GetObjectFile()->GetSymtab(); 1682 // eh_frame would present the same problems but we parse that on a per- 1683 // function basis as-needed so it's more difficult to remove its use of 1684 // the Sections. Realistically, the environments where this code path 1685 // will be taken will not have eh_frame sections. 1686 1687 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1688 1689 // Notify the module that the section addresses have been changed once 1690 // we're done so any file-address caches can be updated. 1691 context.FileAddressesChanged = true; 1692 } 1693 m_sections_up->AddSection(unified_section_sp); 1694 } 1695 1696 llvm::MachO::section_64 sect64; 1697 ::memset(§64, 0, sizeof(sect64)); 1698 // Push a section into our mach sections for the section at index zero 1699 // (NO_SECT) if we don't have any mach sections yet... 1700 if (m_mach_sections.empty()) 1701 m_mach_sections.push_back(sect64); 1702 uint32_t segment_sect_idx; 1703 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1704 1705 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1706 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1707 ++segment_sect_idx) { 1708 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1709 sizeof(sect64.sectname)) == nullptr) 1710 break; 1711 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1712 sizeof(sect64.segname)) == nullptr) 1713 break; 1714 sect64.addr = m_data.GetAddress(&offset); 1715 sect64.size = m_data.GetAddress(&offset); 1716 1717 if (m_data.GetU32(&offset, §64.offset, num_u32s) == nullptr) 1718 break; 1719 1720 if (IsSharedCacheBinary() && !IsInMemory()) { 1721 sect64.offset = sect64.addr - m_text_address; 1722 } 1723 1724 // Keep a list of mach sections around in case we need to get at data that 1725 // isn't stored in the abstracted Sections. 1726 m_mach_sections.push_back(sect64); 1727 1728 if (add_section) { 1729 ConstString section_name( 1730 sect64.sectname, strnlen(sect64.sectname, sizeof(sect64.sectname))); 1731 if (!const_segname) { 1732 // We have a segment with no name so we need to conjure up segments 1733 // that correspond to the section's segname if there isn't already such 1734 // a section. If there is such a section, we resize the section so that 1735 // it spans all sections. We also mark these sections as fake so 1736 // address matches don't hit if they land in the gaps between the child 1737 // sections. 1738 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1739 sizeof(sect64.segname)); 1740 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1741 if (segment_sp.get()) { 1742 Section *segment = segment_sp.get(); 1743 // Grow the section size as needed. 1744 const lldb::addr_t sect64_min_addr = sect64.addr; 1745 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1746 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1747 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1748 const lldb::addr_t curr_seg_max_addr = 1749 curr_seg_min_addr + curr_seg_byte_size; 1750 if (sect64_min_addr >= curr_seg_min_addr) { 1751 const lldb::addr_t new_seg_byte_size = 1752 sect64_max_addr - curr_seg_min_addr; 1753 // Only grow the section size if needed 1754 if (new_seg_byte_size > curr_seg_byte_size) 1755 segment->SetByteSize(new_seg_byte_size); 1756 } else { 1757 // We need to change the base address of the segment and adjust the 1758 // child section offsets for all existing children. 1759 const lldb::addr_t slide_amount = 1760 sect64_min_addr - curr_seg_min_addr; 1761 segment->Slide(slide_amount, false); 1762 segment->GetChildren().Slide(-slide_amount, false); 1763 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1764 } 1765 1766 // Grow the section size as needed. 1767 if (sect64.offset) { 1768 const lldb::addr_t segment_min_file_offset = 1769 segment->GetFileOffset(); 1770 const lldb::addr_t segment_max_file_offset = 1771 segment_min_file_offset + segment->GetFileSize(); 1772 1773 const lldb::addr_t section_min_file_offset = sect64.offset; 1774 const lldb::addr_t section_max_file_offset = 1775 section_min_file_offset + sect64.size; 1776 const lldb::addr_t new_file_offset = 1777 std::min(section_min_file_offset, segment_min_file_offset); 1778 const lldb::addr_t new_file_size = 1779 std::max(section_max_file_offset, segment_max_file_offset) - 1780 new_file_offset; 1781 segment->SetFileOffset(new_file_offset); 1782 segment->SetFileSize(new_file_size); 1783 } 1784 } else { 1785 // Create a fake section for the section's named segment 1786 segment_sp = std::make_shared<Section>( 1787 segment_sp, // Parent section 1788 module_sp, // Module to which this section belongs 1789 this, // Object file to which this section belongs 1790 ++context.NextSegmentIdx 1791 << 8, // Section ID is the 1 based segment index 1792 // shifted right by 8 bits as not to 1793 // collide with any of the 256 section IDs 1794 // that are possible 1795 const_segname, // Name of this section 1796 eSectionTypeContainer, // This section is a container of 1797 // other sections. 1798 sect64.addr, // File VM address == addresses as they are 1799 // found in the object file 1800 sect64.size, // VM size in bytes of this section 1801 sect64.offset, // Offset to the data for this section in 1802 // the file 1803 sect64.offset ? sect64.size : 0, // Size in bytes of 1804 // this section as 1805 // found in the file 1806 sect64.align, 1807 load_cmd.flags); // Flags for this section 1808 segment_sp->SetIsFake(true); 1809 segment_sp->SetPermissions(segment_permissions); 1810 m_sections_up->AddSection(segment_sp); 1811 if (add_to_unified) 1812 context.UnifiedList.AddSection(segment_sp); 1813 segment_sp->SetIsEncrypted(segment_is_encrypted); 1814 } 1815 } 1816 assert(segment_sp.get()); 1817 1818 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1819 1820 SectionSP section_sp(new Section( 1821 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1822 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1823 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1824 sect64.flags)); 1825 // Set the section to be encrypted to match the segment 1826 1827 bool section_is_encrypted = false; 1828 if (!segment_is_encrypted && load_cmd.filesize != 0) 1829 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1830 sect64.offset) != nullptr; 1831 1832 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1833 section_sp->SetPermissions(segment_permissions); 1834 segment_sp->GetChildren().AddSection(section_sp); 1835 1836 if (segment_sp->IsFake()) { 1837 segment_sp.reset(); 1838 const_segname.Clear(); 1839 } 1840 } 1841 } 1842 if (segment_sp && is_dsym) { 1843 if (first_segment_sectID <= context.NextSectionIdx) { 1844 lldb::user_id_t sect_uid; 1845 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1846 ++sect_uid) { 1847 SectionSP curr_section_sp( 1848 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1849 SectionSP next_section_sp; 1850 if (sect_uid + 1 <= context.NextSectionIdx) 1851 next_section_sp = 1852 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1853 1854 if (curr_section_sp.get()) { 1855 if (curr_section_sp->GetByteSize() == 0) { 1856 if (next_section_sp.get() != nullptr) 1857 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1858 curr_section_sp->GetFileAddress()); 1859 else 1860 curr_section_sp->SetByteSize(load_cmd.vmsize); 1861 } 1862 } 1863 } 1864 } 1865 } 1866 } 1867 1868 void ObjectFileMachO::ProcessDysymtabCommand( 1869 const llvm::MachO::load_command &load_cmd, lldb::offset_t offset) { 1870 m_dysymtab.cmd = load_cmd.cmd; 1871 m_dysymtab.cmdsize = load_cmd.cmdsize; 1872 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1873 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1874 } 1875 1876 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1877 if (m_sections_up) 1878 return; 1879 1880 m_sections_up = std::make_unique<SectionList>(); 1881 1882 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1883 // bool dump_sections = false; 1884 ModuleSP module_sp(GetModule()); 1885 1886 offset = MachHeaderSizeFromMagic(m_header.magic); 1887 1888 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1889 llvm::MachO::load_command load_cmd; 1890 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1891 const lldb::offset_t load_cmd_offset = offset; 1892 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 1893 break; 1894 1895 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1896 ProcessSegmentCommand(load_cmd, offset, i, context); 1897 else if (load_cmd.cmd == LC_DYSYMTAB) 1898 ProcessDysymtabCommand(load_cmd, offset); 1899 1900 offset = load_cmd_offset + load_cmd.cmdsize; 1901 } 1902 1903 if (context.FileAddressesChanged && module_sp) 1904 module_sp->SectionFileAddressesChanged(); 1905 } 1906 1907 class MachSymtabSectionInfo { 1908 public: 1909 MachSymtabSectionInfo(SectionList *section_list) 1910 : m_section_list(section_list), m_section_infos() { 1911 // Get the number of sections down to a depth of 1 to include all segments 1912 // and their sections, but no other sections that may be added for debug 1913 // map or 1914 m_section_infos.resize(section_list->GetNumSections(1)); 1915 } 1916 1917 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1918 if (n_sect == 0) 1919 return SectionSP(); 1920 if (n_sect < m_section_infos.size()) { 1921 if (!m_section_infos[n_sect].section_sp) { 1922 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1923 m_section_infos[n_sect].section_sp = section_sp; 1924 if (section_sp) { 1925 m_section_infos[n_sect].vm_range.SetBaseAddress( 1926 section_sp->GetFileAddress()); 1927 m_section_infos[n_sect].vm_range.SetByteSize( 1928 section_sp->GetByteSize()); 1929 } else { 1930 std::string filename = "<unknown>"; 1931 SectionSP first_section_sp(m_section_list->GetSectionAtIndex(0)); 1932 if (first_section_sp) 1933 filename = first_section_sp->GetObjectFile()->GetFileSpec().GetPath(); 1934 1935 Host::SystemLog(Host::eSystemLogError, 1936 "error: unable to find section %d for a symbol in " 1937 "%s, corrupt file?\n", 1938 n_sect, filename.c_str()); 1939 } 1940 } 1941 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1942 // Symbol is in section. 1943 return m_section_infos[n_sect].section_sp; 1944 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1945 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1946 file_addr) { 1947 // Symbol is in section with zero size, but has the same start address 1948 // as the section. This can happen with linker symbols (symbols that 1949 // start with the letter 'l' or 'L'. 1950 return m_section_infos[n_sect].section_sp; 1951 } 1952 } 1953 return m_section_list->FindSectionContainingFileAddress(file_addr); 1954 } 1955 1956 protected: 1957 struct SectionInfo { 1958 SectionInfo() : vm_range(), section_sp() {} 1959 1960 VMRange vm_range; 1961 SectionSP section_sp; 1962 }; 1963 SectionList *m_section_list; 1964 std::vector<SectionInfo> m_section_infos; 1965 }; 1966 1967 #define TRIE_SYMBOL_IS_THUMB (1ULL << 63) 1968 struct TrieEntry { 1969 void Dump() const { 1970 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1971 static_cast<unsigned long long>(address), 1972 static_cast<unsigned long long>(flags), 1973 static_cast<unsigned long long>(other), name.GetCString()); 1974 if (import_name) 1975 printf(" -> \"%s\"\n", import_name.GetCString()); 1976 else 1977 printf("\n"); 1978 } 1979 ConstString name; 1980 uint64_t address = LLDB_INVALID_ADDRESS; 1981 uint64_t flags = 1982 0; // EXPORT_SYMBOL_FLAGS_REEXPORT, EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER, 1983 // TRIE_SYMBOL_IS_THUMB 1984 uint64_t other = 0; 1985 ConstString import_name; 1986 }; 1987 1988 struct TrieEntryWithOffset { 1989 lldb::offset_t nodeOffset; 1990 TrieEntry entry; 1991 1992 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1993 1994 void Dump(uint32_t idx) const { 1995 printf("[%3u] 0x%16.16llx: ", idx, 1996 static_cast<unsigned long long>(nodeOffset)); 1997 entry.Dump(); 1998 } 1999 2000 bool operator<(const TrieEntryWithOffset &other) const { 2001 return (nodeOffset < other.nodeOffset); 2002 } 2003 }; 2004 2005 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2006 const bool is_arm, addr_t text_seg_base_addr, 2007 std::vector<llvm::StringRef> &nameSlices, 2008 std::set<lldb::addr_t> &resolver_addresses, 2009 std::vector<TrieEntryWithOffset> &reexports, 2010 std::vector<TrieEntryWithOffset> &ext_symbols) { 2011 if (!data.ValidOffset(offset)) 2012 return true; 2013 2014 // Terminal node -- end of a branch, possibly add this to 2015 // the symbol table or resolver table. 2016 const uint64_t terminalSize = data.GetULEB128(&offset); 2017 lldb::offset_t children_offset = offset + terminalSize; 2018 if (terminalSize != 0) { 2019 TrieEntryWithOffset e(offset); 2020 e.entry.flags = data.GetULEB128(&offset); 2021 const char *import_name = nullptr; 2022 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2023 e.entry.address = 0; 2024 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2025 import_name = data.GetCStr(&offset); 2026 } else { 2027 e.entry.address = data.GetULEB128(&offset); 2028 if (text_seg_base_addr != LLDB_INVALID_ADDRESS) 2029 e.entry.address += text_seg_base_addr; 2030 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2031 e.entry.other = data.GetULEB128(&offset); 2032 uint64_t resolver_addr = e.entry.other; 2033 if (text_seg_base_addr != LLDB_INVALID_ADDRESS) 2034 resolver_addr += text_seg_base_addr; 2035 if (is_arm) 2036 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2037 resolver_addresses.insert(resolver_addr); 2038 } else 2039 e.entry.other = 0; 2040 } 2041 bool add_this_entry = false; 2042 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT) && 2043 import_name && import_name[0]) { 2044 // add symbols that are reexport symbols with a valid import name. 2045 add_this_entry = true; 2046 } else if (e.entry.flags == 0 && 2047 (import_name == nullptr || import_name[0] == '\0')) { 2048 // add externally visible symbols, in case the nlist record has 2049 // been stripped/omitted. 2050 add_this_entry = true; 2051 } 2052 if (add_this_entry) { 2053 std::string name; 2054 if (!nameSlices.empty()) { 2055 for (auto name_slice : nameSlices) 2056 name.append(name_slice.data(), name_slice.size()); 2057 } 2058 if (name.size() > 1) { 2059 // Skip the leading '_' 2060 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2061 } 2062 if (import_name) { 2063 // Skip the leading '_' 2064 e.entry.import_name.SetCString(import_name + 1); 2065 } 2066 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT)) { 2067 reexports.push_back(e); 2068 } else { 2069 if (is_arm && (e.entry.address & 1)) { 2070 e.entry.flags |= TRIE_SYMBOL_IS_THUMB; 2071 e.entry.address &= THUMB_ADDRESS_BIT_MASK; 2072 } 2073 ext_symbols.push_back(e); 2074 } 2075 } 2076 } 2077 2078 const uint8_t childrenCount = data.GetU8(&children_offset); 2079 for (uint8_t i = 0; i < childrenCount; ++i) { 2080 const char *cstr = data.GetCStr(&children_offset); 2081 if (cstr) 2082 nameSlices.push_back(llvm::StringRef(cstr)); 2083 else 2084 return false; // Corrupt data 2085 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2086 if (childNodeOffset) { 2087 if (!ParseTrieEntries(data, childNodeOffset, is_arm, text_seg_base_addr, 2088 nameSlices, resolver_addresses, reexports, 2089 ext_symbols)) { 2090 return false; 2091 } 2092 } 2093 nameSlices.pop_back(); 2094 } 2095 return true; 2096 } 2097 2098 static SymbolType GetSymbolType(const char *&symbol_name, 2099 bool &demangled_is_synthesized, 2100 const SectionSP &text_section_sp, 2101 const SectionSP &data_section_sp, 2102 const SectionSP &data_dirty_section_sp, 2103 const SectionSP &data_const_section_sp, 2104 const SectionSP &symbol_section) { 2105 SymbolType type = eSymbolTypeInvalid; 2106 2107 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2108 if (symbol_section->IsDescendant(text_section_sp.get())) { 2109 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 2110 S_ATTR_SELF_MODIFYING_CODE | 2111 S_ATTR_SOME_INSTRUCTIONS)) 2112 type = eSymbolTypeData; 2113 else 2114 type = eSymbolTypeCode; 2115 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 2116 symbol_section->IsDescendant(data_dirty_section_sp.get()) || 2117 symbol_section->IsDescendant(data_const_section_sp.get())) { 2118 if (symbol_sect_name && 2119 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 2120 type = eSymbolTypeRuntime; 2121 2122 if (symbol_name) { 2123 llvm::StringRef symbol_name_ref(symbol_name); 2124 if (symbol_name_ref.startswith("OBJC_")) { 2125 static const llvm::StringRef g_objc_v2_prefix_class("OBJC_CLASS_$_"); 2126 static const llvm::StringRef g_objc_v2_prefix_metaclass( 2127 "OBJC_METACLASS_$_"); 2128 static const llvm::StringRef g_objc_v2_prefix_ivar("OBJC_IVAR_$_"); 2129 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 2130 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2131 type = eSymbolTypeObjCClass; 2132 demangled_is_synthesized = true; 2133 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 2134 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2135 type = eSymbolTypeObjCMetaClass; 2136 demangled_is_synthesized = true; 2137 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 2138 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2139 type = eSymbolTypeObjCIVar; 2140 demangled_is_synthesized = true; 2141 } 2142 } 2143 } 2144 } else if (symbol_sect_name && 2145 ::strstr(symbol_sect_name, "__gcc_except_tab") == 2146 symbol_sect_name) { 2147 type = eSymbolTypeException; 2148 } else { 2149 type = eSymbolTypeData; 2150 } 2151 } else if (symbol_sect_name && 2152 ::strstr(symbol_sect_name, "__IMPORT") == symbol_sect_name) { 2153 type = eSymbolTypeTrampoline; 2154 } 2155 return type; 2156 } 2157 2158 // Read the UUID out of a dyld_shared_cache file on-disk. 2159 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2160 const ByteOrder byte_order, 2161 const uint32_t addr_byte_size) { 2162 UUID dsc_uuid; 2163 DataBufferSP DscData = MapFileData( 2164 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2165 if (!DscData) 2166 return dsc_uuid; 2167 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2168 2169 char version_str[7]; 2170 lldb::offset_t offset = 0; 2171 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2172 version_str[6] = '\0'; 2173 if (strcmp(version_str, "dyld_v") == 0) { 2174 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2175 dsc_uuid = UUID::fromOptionalData( 2176 dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t)); 2177 } 2178 Log *log = GetLog(LLDBLog::Symbols); 2179 if (log && dsc_uuid.IsValid()) { 2180 LLDB_LOGF(log, "Shared cache %s has UUID %s", 2181 dyld_shared_cache.GetPath().c_str(), 2182 dsc_uuid.GetAsString().c_str()); 2183 } 2184 return dsc_uuid; 2185 } 2186 2187 static llvm::Optional<struct nlist_64> 2188 ParseNList(DataExtractor &nlist_data, lldb::offset_t &nlist_data_offset, 2189 size_t nlist_byte_size) { 2190 struct nlist_64 nlist; 2191 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2192 return {}; 2193 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 2194 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 2195 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 2196 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 2197 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 2198 return nlist; 2199 } 2200 2201 enum { DebugSymbols = true, NonDebugSymbols = false }; 2202 2203 void ObjectFileMachO::ParseSymtab(Symtab &symtab) { 2204 LLDB_SCOPED_TIMERF("ObjectFileMachO::ParseSymtab () module = %s", 2205 m_file.GetFilename().AsCString("")); 2206 ModuleSP module_sp(GetModule()); 2207 if (!module_sp) 2208 return; 2209 2210 Progress progress(llvm::formatv("Parsing symbol table for {0}", 2211 m_file.GetFilename().AsCString("<Unknown>"))); 2212 2213 llvm::MachO::symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2214 llvm::MachO::linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2215 llvm::MachO::linkedit_data_command exports_trie_load_command = {0, 0, 0, 0}; 2216 llvm::MachO::dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2217 // The data element of type bool indicates that this entry is thumb 2218 // code. 2219 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2220 2221 // Record the address of every function/data that we add to the symtab. 2222 // We add symbols to the table in the order of most information (nlist 2223 // records) to least (function starts), and avoid duplicating symbols 2224 // via this set. 2225 llvm::DenseSet<addr_t> symbols_added; 2226 2227 // We are using a llvm::DenseSet for "symbols_added" so we must be sure we 2228 // do not add the tombstone or empty keys to the set. 2229 auto add_symbol_addr = [&symbols_added](lldb::addr_t file_addr) { 2230 // Don't add the tombstone or empty keys. 2231 if (file_addr == UINT64_MAX || file_addr == UINT64_MAX - 1) 2232 return; 2233 symbols_added.insert(file_addr); 2234 }; 2235 FunctionStarts function_starts; 2236 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2237 uint32_t i; 2238 FileSpecList dylib_files; 2239 Log *log = GetLog(LLDBLog::Symbols); 2240 llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2241 llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2242 llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2243 UUID image_uuid; 2244 2245 for (i = 0; i < m_header.ncmds; ++i) { 2246 const lldb::offset_t cmd_offset = offset; 2247 // Read in the load command and load command size 2248 llvm::MachO::load_command lc; 2249 if (m_data.GetU32(&offset, &lc, 2) == nullptr) 2250 break; 2251 // Watch for the symbol table load command 2252 switch (lc.cmd) { 2253 case LC_SYMTAB: 2254 symtab_load_command.cmd = lc.cmd; 2255 symtab_load_command.cmdsize = lc.cmdsize; 2256 // Read in the rest of the symtab load command 2257 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2258 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2259 return; 2260 break; 2261 2262 case LC_DYLD_INFO: 2263 case LC_DYLD_INFO_ONLY: 2264 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2265 dyld_info.cmd = lc.cmd; 2266 dyld_info.cmdsize = lc.cmdsize; 2267 } else { 2268 memset(&dyld_info, 0, sizeof(dyld_info)); 2269 } 2270 break; 2271 2272 case LC_LOAD_DYLIB: 2273 case LC_LOAD_WEAK_DYLIB: 2274 case LC_REEXPORT_DYLIB: 2275 case LC_LOADFVMLIB: 2276 case LC_LOAD_UPWARD_DYLIB: { 2277 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2278 const char *path = m_data.PeekCStr(name_offset); 2279 if (path) { 2280 FileSpec file_spec(path); 2281 // Strip the path if there is @rpath, @executable, etc so we just use 2282 // the basename 2283 if (path[0] == '@') 2284 file_spec.GetDirectory().Clear(); 2285 2286 if (lc.cmd == LC_REEXPORT_DYLIB) { 2287 m_reexported_dylibs.AppendIfUnique(file_spec); 2288 } 2289 2290 dylib_files.Append(file_spec); 2291 } 2292 } break; 2293 2294 case LC_DYLD_EXPORTS_TRIE: 2295 exports_trie_load_command.cmd = lc.cmd; 2296 exports_trie_load_command.cmdsize = lc.cmdsize; 2297 if (m_data.GetU32(&offset, &exports_trie_load_command.dataoff, 2) == 2298 nullptr) // fill in offset and size fields 2299 memset(&exports_trie_load_command, 0, 2300 sizeof(exports_trie_load_command)); 2301 break; 2302 case LC_FUNCTION_STARTS: 2303 function_starts_load_command.cmd = lc.cmd; 2304 function_starts_load_command.cmdsize = lc.cmdsize; 2305 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2306 nullptr) // fill in data offset and size fields 2307 memset(&function_starts_load_command, 0, 2308 sizeof(function_starts_load_command)); 2309 break; 2310 2311 case LC_UUID: { 2312 const uint8_t *uuid_bytes = m_data.PeekData(offset, 16); 2313 2314 if (uuid_bytes) 2315 image_uuid = UUID::fromOptionalData(uuid_bytes, 16); 2316 break; 2317 } 2318 2319 default: 2320 break; 2321 } 2322 offset = cmd_offset + lc.cmdsize; 2323 } 2324 2325 if (!symtab_load_command.cmd) 2326 return; 2327 2328 SectionList *section_list = GetSectionList(); 2329 if (section_list == nullptr) 2330 return; 2331 2332 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2333 const ByteOrder byte_order = m_data.GetByteOrder(); 2334 bool bit_width_32 = addr_byte_size == 4; 2335 const size_t nlist_byte_size = 2336 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2337 2338 DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size); 2339 DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size); 2340 DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size); 2341 DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order, 2342 addr_byte_size); 2343 DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size); 2344 2345 const addr_t nlist_data_byte_size = 2346 symtab_load_command.nsyms * nlist_byte_size; 2347 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2348 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2349 2350 ProcessSP process_sp(m_process_wp.lock()); 2351 Process *process = process_sp.get(); 2352 2353 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2354 bool is_shared_cache_image = IsSharedCacheBinary(); 2355 bool is_local_shared_cache_image = is_shared_cache_image && !IsInMemory(); 2356 SectionSP linkedit_section_sp( 2357 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2358 2359 if (process && m_header.filetype != llvm::MachO::MH_OBJECT && 2360 !is_local_shared_cache_image) { 2361 Target &target = process->GetTarget(); 2362 2363 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2364 2365 // Reading mach file from memory in a process or core file... 2366 2367 if (linkedit_section_sp) { 2368 addr_t linkedit_load_addr = 2369 linkedit_section_sp->GetLoadBaseAddress(&target); 2370 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2371 // We might be trying to access the symbol table before the 2372 // __LINKEDIT's load address has been set in the target. We can't 2373 // fail to read the symbol table, so calculate the right address 2374 // manually 2375 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2376 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2377 } 2378 2379 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2380 const addr_t symoff_addr = linkedit_load_addr + 2381 symtab_load_command.symoff - 2382 linkedit_file_offset; 2383 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2384 linkedit_file_offset; 2385 2386 // Always load dyld - the dynamic linker - from memory if we didn't 2387 // find a binary anywhere else. lldb will not register 2388 // dylib/framework/bundle loads/unloads if we don't have the dyld 2389 // symbols, we force dyld to load from memory despite the user's 2390 // target.memory-module-load-level setting. 2391 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2392 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2393 DataBufferSP nlist_data_sp( 2394 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2395 if (nlist_data_sp) 2396 nlist_data.SetData(nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2397 if (m_dysymtab.nindirectsyms != 0) { 2398 const addr_t indirect_syms_addr = linkedit_load_addr + 2399 m_dysymtab.indirectsymoff - 2400 linkedit_file_offset; 2401 DataBufferSP indirect_syms_data_sp(ReadMemory( 2402 process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 2403 if (indirect_syms_data_sp) 2404 indirect_symbol_index_data.SetData( 2405 indirect_syms_data_sp, 0, 2406 indirect_syms_data_sp->GetByteSize()); 2407 // If this binary is outside the shared cache, 2408 // cache the string table. 2409 // Binaries in the shared cache all share a giant string table, 2410 // and we can't share the string tables across multiple 2411 // ObjectFileMachO's, so we'd end up re-reading this mega-strtab 2412 // for every binary in the shared cache - it would be a big perf 2413 // problem. For binaries outside the shared cache, it's faster to 2414 // read the entire strtab at once instead of piece-by-piece as we 2415 // process the nlist records. 2416 if (!is_shared_cache_image) { 2417 DataBufferSP strtab_data_sp( 2418 ReadMemory(process_sp, strtab_addr, strtab_data_byte_size)); 2419 if (strtab_data_sp) { 2420 strtab_data.SetData(strtab_data_sp, 0, 2421 strtab_data_sp->GetByteSize()); 2422 } 2423 } 2424 } 2425 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) { 2426 if (function_starts_load_command.cmd) { 2427 const addr_t func_start_addr = 2428 linkedit_load_addr + function_starts_load_command.dataoff - 2429 linkedit_file_offset; 2430 DataBufferSP func_start_data_sp( 2431 ReadMemory(process_sp, func_start_addr, 2432 function_starts_load_command.datasize)); 2433 if (func_start_data_sp) 2434 function_starts_data.SetData(func_start_data_sp, 0, 2435 func_start_data_sp->GetByteSize()); 2436 } 2437 } 2438 } 2439 } 2440 } else { 2441 if (is_local_shared_cache_image) { 2442 // The load commands in shared cache images are relative to the 2443 // beginning of the shared cache, not the library image. The 2444 // data we get handed when creating the ObjectFileMachO starts 2445 // at the beginning of a specific library and spans to the end 2446 // of the cache to be able to reach the shared LINKEDIT 2447 // segments. We need to convert the load command offsets to be 2448 // relative to the beginning of our specific image. 2449 lldb::addr_t linkedit_offset = linkedit_section_sp->GetFileOffset(); 2450 lldb::offset_t linkedit_slide = 2451 linkedit_offset - m_linkedit_original_offset; 2452 symtab_load_command.symoff += linkedit_slide; 2453 symtab_load_command.stroff += linkedit_slide; 2454 dyld_info.export_off += linkedit_slide; 2455 m_dysymtab.indirectsymoff += linkedit_slide; 2456 function_starts_load_command.dataoff += linkedit_slide; 2457 exports_trie_load_command.dataoff += linkedit_slide; 2458 } 2459 2460 nlist_data.SetData(m_data, symtab_load_command.symoff, 2461 nlist_data_byte_size); 2462 strtab_data.SetData(m_data, symtab_load_command.stroff, 2463 strtab_data_byte_size); 2464 2465 // We shouldn't have exports data from both the LC_DYLD_INFO command 2466 // AND the LC_DYLD_EXPORTS_TRIE command in the same binary: 2467 lldbassert(!((dyld_info.export_size > 0) 2468 && (exports_trie_load_command.datasize > 0))); 2469 if (dyld_info.export_size > 0) { 2470 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2471 dyld_info.export_size); 2472 } else if (exports_trie_load_command.datasize > 0) { 2473 dyld_trie_data.SetData(m_data, exports_trie_load_command.dataoff, 2474 exports_trie_load_command.datasize); 2475 } 2476 2477 if (m_dysymtab.nindirectsyms != 0) { 2478 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2479 m_dysymtab.nindirectsyms * 4); 2480 } 2481 if (function_starts_load_command.cmd) { 2482 function_starts_data.SetData(m_data, function_starts_load_command.dataoff, 2483 function_starts_load_command.datasize); 2484 } 2485 } 2486 2487 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2488 2489 ConstString g_segment_name_TEXT = GetSegmentNameTEXT(); 2490 ConstString g_segment_name_DATA = GetSegmentNameDATA(); 2491 ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2492 ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2493 ConstString g_segment_name_OBJC = GetSegmentNameOBJC(); 2494 ConstString g_section_name_eh_frame = GetSectionNameEHFrame(); 2495 SectionSP text_section_sp( 2496 section_list->FindSectionByName(g_segment_name_TEXT)); 2497 SectionSP data_section_sp( 2498 section_list->FindSectionByName(g_segment_name_DATA)); 2499 SectionSP data_dirty_section_sp( 2500 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2501 SectionSP data_const_section_sp( 2502 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2503 SectionSP objc_section_sp( 2504 section_list->FindSectionByName(g_segment_name_OBJC)); 2505 SectionSP eh_frame_section_sp; 2506 if (text_section_sp.get()) 2507 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2508 g_section_name_eh_frame); 2509 else 2510 eh_frame_section_sp = 2511 section_list->FindSectionByName(g_section_name_eh_frame); 2512 2513 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2514 const bool always_thumb = GetArchitecture().IsAlwaysThumbInstructions(); 2515 2516 // lldb works best if it knows the start address of all functions in a 2517 // module. Linker symbols or debug info are normally the best source of 2518 // information for start addr / size but they may be stripped in a released 2519 // binary. Two additional sources of information exist in Mach-O binaries: 2520 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2521 // function's start address in the 2522 // binary, relative to the text section. 2523 // eh_frame - the eh_frame FDEs have the start addr & size of 2524 // each function 2525 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2526 // all modern binaries. 2527 // Binaries built to run on older releases may need to use eh_frame 2528 // information. 2529 2530 if (text_section_sp && function_starts_data.GetByteSize()) { 2531 FunctionStarts::Entry function_start_entry; 2532 function_start_entry.data = false; 2533 lldb::offset_t function_start_offset = 0; 2534 function_start_entry.addr = text_section_sp->GetFileAddress(); 2535 uint64_t delta; 2536 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2537 0) { 2538 // Now append the current entry 2539 function_start_entry.addr += delta; 2540 if (is_arm) { 2541 if (function_start_entry.addr & 1) { 2542 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2543 function_start_entry.data = true; 2544 } else if (always_thumb) { 2545 function_start_entry.data = true; 2546 } 2547 } 2548 function_starts.Append(function_start_entry); 2549 } 2550 } else { 2551 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2552 // load command claiming an eh_frame but it doesn't actually have the 2553 // eh_frame content. And if we have a dSYM, we don't need to do any of 2554 // this fill-in-the-missing-symbols works anyway - the debug info should 2555 // give us all the functions in the module. 2556 if (text_section_sp.get() && eh_frame_section_sp.get() && 2557 m_type != eTypeDebugInfo) { 2558 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2559 DWARFCallFrameInfo::EH); 2560 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2561 eh_frame.GetFunctionAddressAndSizeVector(functions); 2562 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2563 size_t count = functions.GetSize(); 2564 for (size_t i = 0; i < count; ++i) { 2565 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2566 functions.GetEntryAtIndex(i); 2567 if (func) { 2568 FunctionStarts::Entry function_start_entry; 2569 function_start_entry.addr = func->base - text_base_addr; 2570 if (is_arm) { 2571 if (function_start_entry.addr & 1) { 2572 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2573 function_start_entry.data = true; 2574 } else if (always_thumb) { 2575 function_start_entry.data = true; 2576 } 2577 } 2578 function_starts.Append(function_start_entry); 2579 } 2580 } 2581 } 2582 } 2583 2584 const size_t function_starts_count = function_starts.GetSize(); 2585 2586 // For user process binaries (executables, dylibs, frameworks, bundles), if 2587 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2588 // going to assume the binary has been stripped. Don't allow assembly 2589 // language instruction emulation because we don't know proper function 2590 // start boundaries. 2591 // 2592 // For all other types of binaries (kernels, stand-alone bare board 2593 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2594 // sections - we should not make any assumptions about them based on that. 2595 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2596 m_allow_assembly_emulation_unwind_plans = false; 2597 Log *unwind_or_symbol_log(GetLog(LLDBLog::Symbols | LLDBLog::Unwind)); 2598 2599 if (unwind_or_symbol_log) 2600 module_sp->LogMessage( 2601 unwind_or_symbol_log, 2602 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2603 } 2604 2605 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() 2606 ? eh_frame_section_sp->GetID() 2607 : static_cast<user_id_t>(NO_SECT); 2608 2609 uint32_t N_SO_index = UINT32_MAX; 2610 2611 MachSymtabSectionInfo section_info(section_list); 2612 std::vector<uint32_t> N_FUN_indexes; 2613 std::vector<uint32_t> N_NSYM_indexes; 2614 std::vector<uint32_t> N_INCL_indexes; 2615 std::vector<uint32_t> N_BRAC_indexes; 2616 std::vector<uint32_t> N_COMM_indexes; 2617 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2618 typedef llvm::DenseMap<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2619 typedef llvm::DenseMap<const char *, uint32_t> ConstNameToSymbolIndexMap; 2620 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2621 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2622 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2623 // Any symbols that get merged into another will get an entry in this map 2624 // so we know 2625 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2626 uint32_t nlist_idx = 0; 2627 Symbol *symbol_ptr = nullptr; 2628 2629 uint32_t sym_idx = 0; 2630 Symbol *sym = nullptr; 2631 size_t num_syms = 0; 2632 std::string memory_symbol_name; 2633 uint32_t unmapped_local_symbols_found = 0; 2634 2635 std::vector<TrieEntryWithOffset> reexport_trie_entries; 2636 std::vector<TrieEntryWithOffset> external_sym_trie_entries; 2637 std::set<lldb::addr_t> resolver_addresses; 2638 2639 if (dyld_trie_data.GetByteSize() > 0) { 2640 ConstString text_segment_name("__TEXT"); 2641 SectionSP text_segment_sp = 2642 GetSectionList()->FindSectionByName(text_segment_name); 2643 lldb::addr_t text_segment_file_addr = LLDB_INVALID_ADDRESS; 2644 if (text_segment_sp) 2645 text_segment_file_addr = text_segment_sp->GetFileAddress(); 2646 std::vector<llvm::StringRef> nameSlices; 2647 ParseTrieEntries(dyld_trie_data, 0, is_arm, text_segment_file_addr, 2648 nameSlices, resolver_addresses, reexport_trie_entries, 2649 external_sym_trie_entries); 2650 } 2651 2652 typedef std::set<ConstString> IndirectSymbols; 2653 IndirectSymbols indirect_symbol_names; 2654 2655 #if TARGET_OS_IPHONE 2656 2657 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2658 // optimized by moving LOCAL symbols out of the memory mapped portion of 2659 // the DSC. The symbol information has all been retained, but it isn't 2660 // available in the normal nlist data. However, there *are* duplicate 2661 // entries of *some* 2662 // LOCAL symbols in the normal nlist data. To handle this situation 2663 // correctly, we must first attempt 2664 // to parse any DSC unmapped symbol information. If we find any, we set a 2665 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2666 2667 if (IsSharedCacheBinary()) { 2668 // Before we can start mapping the DSC, we need to make certain the 2669 // target process is actually using the cache we can find. 2670 2671 // Next we need to determine the correct path for the dyld shared cache. 2672 2673 ArchSpec header_arch = GetArchitecture(); 2674 2675 UUID dsc_uuid; 2676 UUID process_shared_cache_uuid; 2677 addr_t process_shared_cache_base_addr; 2678 2679 if (process) { 2680 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, 2681 process_shared_cache_uuid); 2682 } 2683 2684 __block bool found_image = false; 2685 __block void *nlist_buffer = nullptr; 2686 __block unsigned nlist_count = 0; 2687 __block char *string_table = nullptr; 2688 __block vm_offset_t vm_nlist_memory = 0; 2689 __block mach_msg_type_number_t vm_nlist_bytes_read = 0; 2690 __block vm_offset_t vm_string_memory = 0; 2691 __block mach_msg_type_number_t vm_string_bytes_read = 0; 2692 2693 auto _ = llvm::make_scope_exit(^{ 2694 if (vm_nlist_memory) 2695 vm_deallocate(mach_task_self(), vm_nlist_memory, vm_nlist_bytes_read); 2696 if (vm_string_memory) 2697 vm_deallocate(mach_task_self(), vm_string_memory, vm_string_bytes_read); 2698 }); 2699 2700 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 2701 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 2702 UndefinedNameToDescMap undefined_name_to_desc; 2703 SymbolIndexToName reexport_shlib_needs_fixup; 2704 2705 dyld_for_each_installed_shared_cache(^(dyld_shared_cache_t shared_cache) { 2706 uuid_t cache_uuid; 2707 dyld_shared_cache_copy_uuid(shared_cache, &cache_uuid); 2708 if (found_image) 2709 return; 2710 2711 if (process_shared_cache_uuid.IsValid() && 2712 process_shared_cache_uuid != UUID::fromOptionalData(&cache_uuid, 16)) 2713 return; 2714 2715 dyld_shared_cache_for_each_image(shared_cache, ^(dyld_image_t image) { 2716 uuid_t dsc_image_uuid; 2717 if (found_image) 2718 return; 2719 2720 dyld_image_copy_uuid(image, &dsc_image_uuid); 2721 if (image_uuid != UUID::fromOptionalData(dsc_image_uuid, 16)) 2722 return; 2723 2724 found_image = true; 2725 2726 // Compute the size of the string table. We need to ask dyld for a 2727 // new SPI to avoid this step. 2728 dyld_image_local_nlist_content_4Symbolication( 2729 image, ^(const void *nlistStart, uint64_t nlistCount, 2730 const char *stringTable) { 2731 if (!nlistStart || !nlistCount) 2732 return; 2733 2734 // The buffers passed here are valid only inside the block. 2735 // Use vm_read to make a cheap copy of them available for our 2736 // processing later. 2737 kern_return_t ret = 2738 vm_read(mach_task_self(), (vm_address_t)nlistStart, 2739 nlist_byte_size * nlistCount, &vm_nlist_memory, 2740 &vm_nlist_bytes_read); 2741 if (ret != KERN_SUCCESS) 2742 return; 2743 assert(vm_nlist_bytes_read == nlist_byte_size * nlistCount); 2744 2745 // We don't know the size of the string table. It's cheaper 2746 // to map the whol VM region than to determine the size by 2747 // parsing all teh nlist entries. 2748 vm_address_t string_address = (vm_address_t)stringTable; 2749 vm_size_t region_size; 2750 mach_msg_type_number_t info_count = VM_REGION_BASIC_INFO_COUNT_64; 2751 vm_region_basic_info_data_t info; 2752 memory_object_name_t object; 2753 ret = vm_region_64(mach_task_self(), &string_address, 2754 ®ion_size, VM_REGION_BASIC_INFO_64, 2755 (vm_region_info_t)&info, &info_count, &object); 2756 if (ret != KERN_SUCCESS) 2757 return; 2758 2759 ret = vm_read(mach_task_self(), (vm_address_t)stringTable, 2760 region_size - 2761 ((vm_address_t)stringTable - string_address), 2762 &vm_string_memory, &vm_string_bytes_read); 2763 if (ret != KERN_SUCCESS) 2764 return; 2765 2766 nlist_buffer = (void *)vm_nlist_memory; 2767 string_table = (char *)vm_string_memory; 2768 nlist_count = nlistCount; 2769 }); 2770 }); 2771 }); 2772 if (nlist_buffer) { 2773 DataExtractor dsc_local_symbols_data(nlist_buffer, 2774 nlist_count * nlist_byte_size, 2775 byte_order, addr_byte_size); 2776 unmapped_local_symbols_found = nlist_count; 2777 2778 // The normal nlist code cannot correctly size the Symbols 2779 // array, we need to allocate it here. 2780 sym = symtab.Resize( 2781 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2782 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2783 num_syms = symtab.GetNumSymbols(); 2784 2785 lldb::offset_t nlist_data_offset = 0; 2786 2787 for (uint32_t nlist_index = 0; 2788 nlist_index < nlist_count; 2789 nlist_index++) { 2790 ///////////////////////////// 2791 { 2792 llvm::Optional<struct nlist_64> nlist_maybe = 2793 ParseNList(dsc_local_symbols_data, nlist_data_offset, 2794 nlist_byte_size); 2795 if (!nlist_maybe) 2796 break; 2797 struct nlist_64 nlist = *nlist_maybe; 2798 2799 SymbolType type = eSymbolTypeInvalid; 2800 const char *symbol_name = string_table + nlist.n_strx; 2801 2802 if (symbol_name == NULL) { 2803 // No symbol should be NULL, even the symbols with no 2804 // string values should have an offset zero which 2805 // points to an empty C-string 2806 Host::SystemLog( 2807 Host::eSystemLogError, 2808 "error: DSC unmapped local symbol[%u] has invalid " 2809 "string table offset 0x%x in %s, ignoring symbol\n", 2810 nlist_index, nlist.n_strx, 2811 module_sp->GetFileSpec().GetPath().c_str()); 2812 continue; 2813 } 2814 if (symbol_name[0] == '\0') 2815 symbol_name = NULL; 2816 2817 const char *symbol_name_non_abi_mangled = NULL; 2818 2819 SectionSP symbol_section; 2820 uint32_t symbol_byte_size = 0; 2821 bool add_nlist = true; 2822 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2823 bool demangled_is_synthesized = false; 2824 bool is_gsym = false; 2825 bool set_value = true; 2826 2827 assert(sym_idx < num_syms); 2828 2829 sym[sym_idx].SetDebug(is_debug); 2830 2831 if (is_debug) { 2832 switch (nlist.n_type) { 2833 case N_GSYM: 2834 // global symbol: name,,NO_SECT,type,0 2835 // Sometimes the N_GSYM value contains the address. 2836 2837 // FIXME: In the .o files, we have a GSYM and a debug 2838 // symbol for all the ObjC data. They 2839 // have the same address, but we want to ensure that 2840 // we always find only the real symbol, 'cause we 2841 // don't currently correctly attribute the 2842 // GSYM one to the ObjCClass/Ivar/MetaClass 2843 // symbol type. This is a temporary hack to make 2844 // sure the ObjectiveC symbols get treated correctly. 2845 // To do this right, we should coalesce all the GSYM 2846 // & global symbols that have the same address. 2847 2848 is_gsym = true; 2849 sym[sym_idx].SetExternal(true); 2850 2851 if (symbol_name && symbol_name[0] == '_' && 2852 symbol_name[1] == 'O') { 2853 llvm::StringRef symbol_name_ref(symbol_name); 2854 if (symbol_name_ref.startswith( 2855 g_objc_v2_prefix_class)) { 2856 symbol_name_non_abi_mangled = symbol_name + 1; 2857 symbol_name = 2858 symbol_name + g_objc_v2_prefix_class.size(); 2859 type = eSymbolTypeObjCClass; 2860 demangled_is_synthesized = true; 2861 2862 } else if (symbol_name_ref.startswith( 2863 g_objc_v2_prefix_metaclass)) { 2864 symbol_name_non_abi_mangled = symbol_name + 1; 2865 symbol_name = 2866 symbol_name + g_objc_v2_prefix_metaclass.size(); 2867 type = eSymbolTypeObjCMetaClass; 2868 demangled_is_synthesized = true; 2869 } else if (symbol_name_ref.startswith( 2870 g_objc_v2_prefix_ivar)) { 2871 symbol_name_non_abi_mangled = symbol_name + 1; 2872 symbol_name = 2873 symbol_name + g_objc_v2_prefix_ivar.size(); 2874 type = eSymbolTypeObjCIVar; 2875 demangled_is_synthesized = true; 2876 } 2877 } else { 2878 if (nlist.n_value != 0) 2879 symbol_section = section_info.GetSection( 2880 nlist.n_sect, nlist.n_value); 2881 type = eSymbolTypeData; 2882 } 2883 break; 2884 2885 case N_FNAME: 2886 // procedure name (f77 kludge): name,,NO_SECT,0,0 2887 type = eSymbolTypeCompiler; 2888 break; 2889 2890 case N_FUN: 2891 // procedure: name,,n_sect,linenumber,address 2892 if (symbol_name) { 2893 type = eSymbolTypeCode; 2894 symbol_section = section_info.GetSection( 2895 nlist.n_sect, nlist.n_value); 2896 2897 N_FUN_addr_to_sym_idx.insert( 2898 std::make_pair(nlist.n_value, sym_idx)); 2899 // We use the current number of symbols in the 2900 // symbol table in lieu of using nlist_idx in case 2901 // we ever start trimming entries out 2902 N_FUN_indexes.push_back(sym_idx); 2903 } else { 2904 type = eSymbolTypeCompiler; 2905 2906 if (!N_FUN_indexes.empty()) { 2907 // Copy the size of the function into the 2908 // original 2909 // STAB entry so we don't have 2910 // to hunt for it later 2911 symtab.SymbolAtIndex(N_FUN_indexes.back()) 2912 ->SetByteSize(nlist.n_value); 2913 N_FUN_indexes.pop_back(); 2914 // We don't really need the end function STAB as 2915 // it contains the size which we already placed 2916 // with the original symbol, so don't add it if 2917 // we want a minimal symbol table 2918 add_nlist = false; 2919 } 2920 } 2921 break; 2922 2923 case N_STSYM: 2924 // static symbol: name,,n_sect,type,address 2925 N_STSYM_addr_to_sym_idx.insert( 2926 std::make_pair(nlist.n_value, sym_idx)); 2927 symbol_section = section_info.GetSection(nlist.n_sect, 2928 nlist.n_value); 2929 if (symbol_name && symbol_name[0]) { 2930 type = ObjectFile::GetSymbolTypeFromName( 2931 symbol_name + 1, eSymbolTypeData); 2932 } 2933 break; 2934 2935 case N_LCSYM: 2936 // .lcomm symbol: name,,n_sect,type,address 2937 symbol_section = section_info.GetSection(nlist.n_sect, 2938 nlist.n_value); 2939 type = eSymbolTypeCommonBlock; 2940 break; 2941 2942 case N_BNSYM: 2943 // We use the current number of symbols in the symbol 2944 // table in lieu of using nlist_idx in case we ever 2945 // start trimming entries out Skip these if we want 2946 // minimal symbol tables 2947 add_nlist = false; 2948 break; 2949 2950 case N_ENSYM: 2951 // Set the size of the N_BNSYM to the terminating 2952 // index of this N_ENSYM so that we can always skip 2953 // the entire symbol if we need to navigate more 2954 // quickly at the source level when parsing STABS 2955 // Skip these if we want minimal symbol tables 2956 add_nlist = false; 2957 break; 2958 2959 case N_OPT: 2960 // emitted with gcc2_compiled and in gcc source 2961 type = eSymbolTypeCompiler; 2962 break; 2963 2964 case N_RSYM: 2965 // register sym: name,,NO_SECT,type,register 2966 type = eSymbolTypeVariable; 2967 break; 2968 2969 case N_SLINE: 2970 // src line: 0,,n_sect,linenumber,address 2971 symbol_section = section_info.GetSection(nlist.n_sect, 2972 nlist.n_value); 2973 type = eSymbolTypeLineEntry; 2974 break; 2975 2976 case N_SSYM: 2977 // structure elt: name,,NO_SECT,type,struct_offset 2978 type = eSymbolTypeVariableType; 2979 break; 2980 2981 case N_SO: 2982 // source file name 2983 type = eSymbolTypeSourceFile; 2984 if (symbol_name == NULL) { 2985 add_nlist = false; 2986 if (N_SO_index != UINT32_MAX) { 2987 // Set the size of the N_SO to the terminating 2988 // index of this N_SO so that we can always skip 2989 // the entire N_SO if we need to navigate more 2990 // quickly at the source level when parsing STABS 2991 symbol_ptr = symtab.SymbolAtIndex(N_SO_index); 2992 symbol_ptr->SetByteSize(sym_idx); 2993 symbol_ptr->SetSizeIsSibling(true); 2994 } 2995 N_NSYM_indexes.clear(); 2996 N_INCL_indexes.clear(); 2997 N_BRAC_indexes.clear(); 2998 N_COMM_indexes.clear(); 2999 N_FUN_indexes.clear(); 3000 N_SO_index = UINT32_MAX; 3001 } else { 3002 // We use the current number of symbols in the 3003 // symbol table in lieu of using nlist_idx in case 3004 // we ever start trimming entries out 3005 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3006 if (N_SO_has_full_path) { 3007 if ((N_SO_index == sym_idx - 1) && 3008 ((sym_idx - 1) < num_syms)) { 3009 // We have two consecutive N_SO entries where 3010 // the first contains a directory and the 3011 // second contains a full path. 3012 sym[sym_idx - 1].GetMangled().SetValue( 3013 ConstString(symbol_name), false); 3014 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3015 add_nlist = false; 3016 } else { 3017 // This is the first entry in a N_SO that 3018 // contains a directory or 3019 // a full path to the source file 3020 N_SO_index = sym_idx; 3021 } 3022 } else if ((N_SO_index == sym_idx - 1) && 3023 ((sym_idx - 1) < num_syms)) { 3024 // This is usually the second N_SO entry that 3025 // contains just the filename, so here we combine 3026 // it with the first one if we are minimizing the 3027 // symbol table 3028 const char *so_path = sym[sym_idx - 1] 3029 .GetMangled() 3030 .GetDemangledName() 3031 .AsCString(); 3032 if (so_path && so_path[0]) { 3033 std::string full_so_path(so_path); 3034 const size_t double_slash_pos = 3035 full_so_path.find("//"); 3036 if (double_slash_pos != std::string::npos) { 3037 // The linker has been generating bad N_SO 3038 // entries with doubled up paths 3039 // in the format "%s%s" where the first 3040 // string in the DW_AT_comp_dir, and the 3041 // second is the directory for the source 3042 // file so you end up with a path that looks 3043 // like "/tmp/src//tmp/src/" 3044 FileSpec so_dir(so_path); 3045 if (!FileSystem::Instance().Exists(so_dir)) { 3046 so_dir.SetFile( 3047 &full_so_path[double_slash_pos + 1], 3048 FileSpec::Style::native); 3049 if (FileSystem::Instance().Exists(so_dir)) { 3050 // Trim off the incorrect path 3051 full_so_path.erase(0, double_slash_pos + 1); 3052 } 3053 } 3054 } 3055 if (*full_so_path.rbegin() != '/') 3056 full_so_path += '/'; 3057 full_so_path += symbol_name; 3058 sym[sym_idx - 1].GetMangled().SetValue( 3059 ConstString(full_so_path.c_str()), false); 3060 add_nlist = false; 3061 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3062 } 3063 } else { 3064 // This could be a relative path to a N_SO 3065 N_SO_index = sym_idx; 3066 } 3067 } 3068 break; 3069 3070 case N_OSO: 3071 // object file name: name,,0,0,st_mtime 3072 type = eSymbolTypeObjectFile; 3073 break; 3074 3075 case N_LSYM: 3076 // local sym: name,,NO_SECT,type,offset 3077 type = eSymbolTypeLocal; 3078 break; 3079 3080 // INCL scopes 3081 case N_BINCL: 3082 // include file beginning: name,,NO_SECT,0,sum We use 3083 // the current number of symbols in the symbol table 3084 // in lieu of using nlist_idx in case we ever start 3085 // trimming entries out 3086 N_INCL_indexes.push_back(sym_idx); 3087 type = eSymbolTypeScopeBegin; 3088 break; 3089 3090 case N_EINCL: 3091 // include file end: name,,NO_SECT,0,0 3092 // Set the size of the N_BINCL to the terminating 3093 // index of this N_EINCL so that we can always skip 3094 // the entire symbol if we need to navigate more 3095 // quickly at the source level when parsing STABS 3096 if (!N_INCL_indexes.empty()) { 3097 symbol_ptr = 3098 symtab.SymbolAtIndex(N_INCL_indexes.back()); 3099 symbol_ptr->SetByteSize(sym_idx + 1); 3100 symbol_ptr->SetSizeIsSibling(true); 3101 N_INCL_indexes.pop_back(); 3102 } 3103 type = eSymbolTypeScopeEnd; 3104 break; 3105 3106 case N_SOL: 3107 // #included file name: name,,n_sect,0,address 3108 type = eSymbolTypeHeaderFile; 3109 3110 // We currently don't use the header files on darwin 3111 add_nlist = false; 3112 break; 3113 3114 case N_PARAMS: 3115 // compiler parameters: name,,NO_SECT,0,0 3116 type = eSymbolTypeCompiler; 3117 break; 3118 3119 case N_VERSION: 3120 // compiler version: name,,NO_SECT,0,0 3121 type = eSymbolTypeCompiler; 3122 break; 3123 3124 case N_OLEVEL: 3125 // compiler -O level: name,,NO_SECT,0,0 3126 type = eSymbolTypeCompiler; 3127 break; 3128 3129 case N_PSYM: 3130 // parameter: name,,NO_SECT,type,offset 3131 type = eSymbolTypeVariable; 3132 break; 3133 3134 case N_ENTRY: 3135 // alternate entry: name,,n_sect,linenumber,address 3136 symbol_section = section_info.GetSection(nlist.n_sect, 3137 nlist.n_value); 3138 type = eSymbolTypeLineEntry; 3139 break; 3140 3141 // Left and Right Braces 3142 case N_LBRAC: 3143 // left bracket: 0,,NO_SECT,nesting level,address We 3144 // use the current number of symbols in the symbol 3145 // table in lieu of using nlist_idx in case we ever 3146 // start trimming entries out 3147 symbol_section = section_info.GetSection(nlist.n_sect, 3148 nlist.n_value); 3149 N_BRAC_indexes.push_back(sym_idx); 3150 type = eSymbolTypeScopeBegin; 3151 break; 3152 3153 case N_RBRAC: 3154 // right bracket: 0,,NO_SECT,nesting level,address 3155 // Set the size of the N_LBRAC to the terminating 3156 // index of this N_RBRAC so that we can always skip 3157 // the entire symbol if we need to navigate more 3158 // quickly at the source level when parsing STABS 3159 symbol_section = section_info.GetSection(nlist.n_sect, 3160 nlist.n_value); 3161 if (!N_BRAC_indexes.empty()) { 3162 symbol_ptr = 3163 symtab.SymbolAtIndex(N_BRAC_indexes.back()); 3164 symbol_ptr->SetByteSize(sym_idx + 1); 3165 symbol_ptr->SetSizeIsSibling(true); 3166 N_BRAC_indexes.pop_back(); 3167 } 3168 type = eSymbolTypeScopeEnd; 3169 break; 3170 3171 case N_EXCL: 3172 // deleted include file: name,,NO_SECT,0,sum 3173 type = eSymbolTypeHeaderFile; 3174 break; 3175 3176 // COMM scopes 3177 case N_BCOMM: 3178 // begin common: name,,NO_SECT,0,0 3179 // We use the current number of symbols in the symbol 3180 // table in lieu of using nlist_idx in case we ever 3181 // start trimming entries out 3182 type = eSymbolTypeScopeBegin; 3183 N_COMM_indexes.push_back(sym_idx); 3184 break; 3185 3186 case N_ECOML: 3187 // end common (local name): 0,,n_sect,0,address 3188 symbol_section = section_info.GetSection(nlist.n_sect, 3189 nlist.n_value); 3190 // Fall through 3191 3192 case N_ECOMM: 3193 // end common: name,,n_sect,0,0 3194 // Set the size of the N_BCOMM to the terminating 3195 // index of this N_ECOMM/N_ECOML so that we can 3196 // always skip the entire symbol if we need to 3197 // navigate more quickly at the source level when 3198 // parsing STABS 3199 if (!N_COMM_indexes.empty()) { 3200 symbol_ptr = 3201 symtab.SymbolAtIndex(N_COMM_indexes.back()); 3202 symbol_ptr->SetByteSize(sym_idx + 1); 3203 symbol_ptr->SetSizeIsSibling(true); 3204 N_COMM_indexes.pop_back(); 3205 } 3206 type = eSymbolTypeScopeEnd; 3207 break; 3208 3209 case N_LENG: 3210 // second stab entry with length information 3211 type = eSymbolTypeAdditional; 3212 break; 3213 3214 default: 3215 break; 3216 } 3217 } else { 3218 // uint8_t n_pext = N_PEXT & nlist.n_type; 3219 uint8_t n_type = N_TYPE & nlist.n_type; 3220 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3221 3222 switch (n_type) { 3223 case N_INDR: { 3224 const char *reexport_name_cstr = 3225 strtab_data.PeekCStr(nlist.n_value); 3226 if (reexport_name_cstr && reexport_name_cstr[0]) { 3227 type = eSymbolTypeReExported; 3228 ConstString reexport_name( 3229 reexport_name_cstr + 3230 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3231 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3232 set_value = false; 3233 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3234 indirect_symbol_names.insert(ConstString( 3235 symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3236 } else 3237 type = eSymbolTypeUndefined; 3238 } break; 3239 3240 case N_UNDF: 3241 if (symbol_name && symbol_name[0]) { 3242 ConstString undefined_name( 3243 symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3244 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3245 } 3246 // Fall through 3247 case N_PBUD: 3248 type = eSymbolTypeUndefined; 3249 break; 3250 3251 case N_ABS: 3252 type = eSymbolTypeAbsolute; 3253 break; 3254 3255 case N_SECT: { 3256 symbol_section = section_info.GetSection(nlist.n_sect, 3257 nlist.n_value); 3258 3259 if (symbol_section == NULL) { 3260 // TODO: warn about this? 3261 add_nlist = false; 3262 break; 3263 } 3264 3265 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3266 type = eSymbolTypeException; 3267 } else { 3268 uint32_t section_type = 3269 symbol_section->Get() & SECTION_TYPE; 3270 3271 switch (section_type) { 3272 case S_CSTRING_LITERALS: 3273 type = eSymbolTypeData; 3274 break; // section with only literal C strings 3275 case S_4BYTE_LITERALS: 3276 type = eSymbolTypeData; 3277 break; // section with only 4 byte literals 3278 case S_8BYTE_LITERALS: 3279 type = eSymbolTypeData; 3280 break; // section with only 8 byte literals 3281 case S_LITERAL_POINTERS: 3282 type = eSymbolTypeTrampoline; 3283 break; // section with only pointers to literals 3284 case S_NON_LAZY_SYMBOL_POINTERS: 3285 type = eSymbolTypeTrampoline; 3286 break; // section with only non-lazy symbol 3287 // pointers 3288 case S_LAZY_SYMBOL_POINTERS: 3289 type = eSymbolTypeTrampoline; 3290 break; // section with only lazy symbol pointers 3291 case S_SYMBOL_STUBS: 3292 type = eSymbolTypeTrampoline; 3293 break; // section with only symbol stubs, byte 3294 // size of stub in the reserved2 field 3295 case S_MOD_INIT_FUNC_POINTERS: 3296 type = eSymbolTypeCode; 3297 break; // section with only function pointers for 3298 // initialization 3299 case S_MOD_TERM_FUNC_POINTERS: 3300 type = eSymbolTypeCode; 3301 break; // section with only function pointers for 3302 // termination 3303 case S_INTERPOSING: 3304 type = eSymbolTypeTrampoline; 3305 break; // section with only pairs of function 3306 // pointers for interposing 3307 case S_16BYTE_LITERALS: 3308 type = eSymbolTypeData; 3309 break; // section with only 16 byte literals 3310 case S_DTRACE_DOF: 3311 type = eSymbolTypeInstrumentation; 3312 break; 3313 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3314 type = eSymbolTypeTrampoline; 3315 break; 3316 default: 3317 switch (symbol_section->GetType()) { 3318 case lldb::eSectionTypeCode: 3319 type = eSymbolTypeCode; 3320 break; 3321 case eSectionTypeData: 3322 case eSectionTypeDataCString: // Inlined C string 3323 // data 3324 case eSectionTypeDataCStringPointers: // Pointers 3325 // to C 3326 // string 3327 // data 3328 case eSectionTypeDataSymbolAddress: // Address of 3329 // a symbol in 3330 // the symbol 3331 // table 3332 case eSectionTypeData4: 3333 case eSectionTypeData8: 3334 case eSectionTypeData16: 3335 type = eSymbolTypeData; 3336 break; 3337 default: 3338 break; 3339 } 3340 break; 3341 } 3342 3343 if (type == eSymbolTypeInvalid) { 3344 const char *symbol_sect_name = 3345 symbol_section->GetName().AsCString(); 3346 if (symbol_section->IsDescendant( 3347 text_section_sp.get())) { 3348 if (symbol_section->IsClear( 3349 S_ATTR_PURE_INSTRUCTIONS | 3350 S_ATTR_SELF_MODIFYING_CODE | 3351 S_ATTR_SOME_INSTRUCTIONS)) 3352 type = eSymbolTypeData; 3353 else 3354 type = eSymbolTypeCode; 3355 } else if (symbol_section->IsDescendant( 3356 data_section_sp.get()) || 3357 symbol_section->IsDescendant( 3358 data_dirty_section_sp.get()) || 3359 symbol_section->IsDescendant( 3360 data_const_section_sp.get())) { 3361 if (symbol_sect_name && 3362 ::strstr(symbol_sect_name, "__objc") == 3363 symbol_sect_name) { 3364 type = eSymbolTypeRuntime; 3365 3366 if (symbol_name) { 3367 llvm::StringRef symbol_name_ref(symbol_name); 3368 if (symbol_name_ref.startswith("_OBJC_")) { 3369 llvm::StringRef 3370 g_objc_v2_prefix_class( 3371 "_OBJC_CLASS_$_"); 3372 llvm::StringRef 3373 g_objc_v2_prefix_metaclass( 3374 "_OBJC_METACLASS_$_"); 3375 llvm::StringRef 3376 g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 3377 if (symbol_name_ref.startswith( 3378 g_objc_v2_prefix_class)) { 3379 symbol_name_non_abi_mangled = 3380 symbol_name + 1; 3381 symbol_name = 3382 symbol_name + 3383 g_objc_v2_prefix_class.size(); 3384 type = eSymbolTypeObjCClass; 3385 demangled_is_synthesized = true; 3386 } else if ( 3387 symbol_name_ref.startswith( 3388 g_objc_v2_prefix_metaclass)) { 3389 symbol_name_non_abi_mangled = 3390 symbol_name + 1; 3391 symbol_name = 3392 symbol_name + 3393 g_objc_v2_prefix_metaclass.size(); 3394 type = eSymbolTypeObjCMetaClass; 3395 demangled_is_synthesized = true; 3396 } else if (symbol_name_ref.startswith( 3397 g_objc_v2_prefix_ivar)) { 3398 symbol_name_non_abi_mangled = 3399 symbol_name + 1; 3400 symbol_name = 3401 symbol_name + 3402 g_objc_v2_prefix_ivar.size(); 3403 type = eSymbolTypeObjCIVar; 3404 demangled_is_synthesized = true; 3405 } 3406 } 3407 } 3408 } else if (symbol_sect_name && 3409 ::strstr(symbol_sect_name, 3410 "__gcc_except_tab") == 3411 symbol_sect_name) { 3412 type = eSymbolTypeException; 3413 } else { 3414 type = eSymbolTypeData; 3415 } 3416 } else if (symbol_sect_name && 3417 ::strstr(symbol_sect_name, "__IMPORT") == 3418 symbol_sect_name) { 3419 type = eSymbolTypeTrampoline; 3420 } else if (symbol_section->IsDescendant( 3421 objc_section_sp.get())) { 3422 type = eSymbolTypeRuntime; 3423 if (symbol_name && symbol_name[0] == '.') { 3424 llvm::StringRef symbol_name_ref(symbol_name); 3425 llvm::StringRef 3426 g_objc_v1_prefix_class(".objc_class_name_"); 3427 if (symbol_name_ref.startswith( 3428 g_objc_v1_prefix_class)) { 3429 symbol_name_non_abi_mangled = symbol_name; 3430 symbol_name = symbol_name + 3431 g_objc_v1_prefix_class.size(); 3432 type = eSymbolTypeObjCClass; 3433 demangled_is_synthesized = true; 3434 } 3435 } 3436 } 3437 } 3438 } 3439 } break; 3440 } 3441 } 3442 3443 if (add_nlist) { 3444 uint64_t symbol_value = nlist.n_value; 3445 if (symbol_name_non_abi_mangled) { 3446 sym[sym_idx].GetMangled().SetMangledName( 3447 ConstString(symbol_name_non_abi_mangled)); 3448 sym[sym_idx].GetMangled().SetDemangledName( 3449 ConstString(symbol_name)); 3450 } else { 3451 bool symbol_name_is_mangled = false; 3452 3453 if (symbol_name && symbol_name[0] == '_') { 3454 symbol_name_is_mangled = symbol_name[1] == '_'; 3455 symbol_name++; // Skip the leading underscore 3456 } 3457 3458 if (symbol_name) { 3459 ConstString const_symbol_name(symbol_name); 3460 sym[sym_idx].GetMangled().SetValue( 3461 const_symbol_name, symbol_name_is_mangled); 3462 if (is_gsym && is_debug) { 3463 const char *gsym_name = 3464 sym[sym_idx] 3465 .GetMangled() 3466 .GetName(Mangled::ePreferMangled) 3467 .GetCString(); 3468 if (gsym_name) 3469 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3470 } 3471 } 3472 } 3473 if (symbol_section) { 3474 const addr_t section_file_addr = 3475 symbol_section->GetFileAddress(); 3476 if (symbol_byte_size == 0 && 3477 function_starts_count > 0) { 3478 addr_t symbol_lookup_file_addr = nlist.n_value; 3479 // Do an exact address match for non-ARM addresses, 3480 // else get the closest since the symbol might be a 3481 // thumb symbol which has an address with bit zero 3482 // set 3483 FunctionStarts::Entry *func_start_entry = 3484 function_starts.FindEntry(symbol_lookup_file_addr, 3485 !is_arm); 3486 if (is_arm && func_start_entry) { 3487 // Verify that the function start address is the 3488 // symbol address (ARM) or the symbol address + 1 3489 // (thumb) 3490 if (func_start_entry->addr != 3491 symbol_lookup_file_addr && 3492 func_start_entry->addr != 3493 (symbol_lookup_file_addr + 1)) { 3494 // Not the right entry, NULL it out... 3495 func_start_entry = NULL; 3496 } 3497 } 3498 if (func_start_entry) { 3499 func_start_entry->data = true; 3500 3501 addr_t symbol_file_addr = func_start_entry->addr; 3502 uint32_t symbol_flags = 0; 3503 if (is_arm) { 3504 if (symbol_file_addr & 1) 3505 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3506 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3507 } 3508 3509 const FunctionStarts::Entry *next_func_start_entry = 3510 function_starts.FindNextEntry(func_start_entry); 3511 const addr_t section_end_file_addr = 3512 section_file_addr + 3513 symbol_section->GetByteSize(); 3514 if (next_func_start_entry) { 3515 addr_t next_symbol_file_addr = 3516 next_func_start_entry->addr; 3517 // Be sure the clear the Thumb address bit when 3518 // we calculate the size from the current and 3519 // next address 3520 if (is_arm) 3521 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3522 symbol_byte_size = std::min<lldb::addr_t>( 3523 next_symbol_file_addr - symbol_file_addr, 3524 section_end_file_addr - symbol_file_addr); 3525 } else { 3526 symbol_byte_size = 3527 section_end_file_addr - symbol_file_addr; 3528 } 3529 } 3530 } 3531 symbol_value -= section_file_addr; 3532 } 3533 3534 if (is_debug == false) { 3535 if (type == eSymbolTypeCode) { 3536 // See if we can find a N_FUN entry for any code 3537 // symbols. If we do find a match, and the name 3538 // matches, then we can merge the two into just the 3539 // function symbol to avoid duplicate entries in 3540 // the symbol table 3541 auto range = 3542 N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3543 if (range.first != range.second) { 3544 bool found_it = false; 3545 for (auto pos = range.first; pos != range.second; 3546 ++pos) { 3547 if (sym[sym_idx].GetMangled().GetName( 3548 Mangled::ePreferMangled) == 3549 sym[pos->second].GetMangled().GetName( 3550 Mangled::ePreferMangled)) { 3551 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3552 // We just need the flags from the linker 3553 // symbol, so put these flags 3554 // into the N_FUN flags to avoid duplicate 3555 // symbols in the symbol table 3556 sym[pos->second].SetExternal( 3557 sym[sym_idx].IsExternal()); 3558 sym[pos->second].SetFlags(nlist.n_type << 16 | 3559 nlist.n_desc); 3560 if (resolver_addresses.find(nlist.n_value) != 3561 resolver_addresses.end()) 3562 sym[pos->second].SetType(eSymbolTypeResolver); 3563 sym[sym_idx].Clear(); 3564 found_it = true; 3565 break; 3566 } 3567 } 3568 if (found_it) 3569 continue; 3570 } else { 3571 if (resolver_addresses.find(nlist.n_value) != 3572 resolver_addresses.end()) 3573 type = eSymbolTypeResolver; 3574 } 3575 } else if (type == eSymbolTypeData || 3576 type == eSymbolTypeObjCClass || 3577 type == eSymbolTypeObjCMetaClass || 3578 type == eSymbolTypeObjCIVar) { 3579 // See if we can find a N_STSYM entry for any data 3580 // symbols. If we do find a match, and the name 3581 // matches, then we can merge the two into just the 3582 // Static symbol to avoid duplicate entries in the 3583 // symbol table 3584 auto range = N_STSYM_addr_to_sym_idx.equal_range( 3585 nlist.n_value); 3586 if (range.first != range.second) { 3587 bool found_it = false; 3588 for (auto pos = range.first; pos != range.second; 3589 ++pos) { 3590 if (sym[sym_idx].GetMangled().GetName( 3591 Mangled::ePreferMangled) == 3592 sym[pos->second].GetMangled().GetName( 3593 Mangled::ePreferMangled)) { 3594 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3595 // We just need the flags from the linker 3596 // symbol, so put these flags 3597 // into the N_STSYM flags to avoid duplicate 3598 // symbols in the symbol table 3599 sym[pos->second].SetExternal( 3600 sym[sym_idx].IsExternal()); 3601 sym[pos->second].SetFlags(nlist.n_type << 16 | 3602 nlist.n_desc); 3603 sym[sym_idx].Clear(); 3604 found_it = true; 3605 break; 3606 } 3607 } 3608 if (found_it) 3609 continue; 3610 } else { 3611 const char *gsym_name = 3612 sym[sym_idx] 3613 .GetMangled() 3614 .GetName(Mangled::ePreferMangled) 3615 .GetCString(); 3616 if (gsym_name) { 3617 // Combine N_GSYM stab entries with the non 3618 // stab symbol 3619 ConstNameToSymbolIndexMap::const_iterator pos = 3620 N_GSYM_name_to_sym_idx.find(gsym_name); 3621 if (pos != N_GSYM_name_to_sym_idx.end()) { 3622 const uint32_t GSYM_sym_idx = pos->second; 3623 m_nlist_idx_to_sym_idx[nlist_idx] = 3624 GSYM_sym_idx; 3625 // Copy the address, because often the N_GSYM 3626 // address has an invalid address of zero 3627 // when the global is a common symbol 3628 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3629 symbol_section); 3630 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3631 symbol_value); 3632 add_symbol_addr(sym[GSYM_sym_idx] 3633 .GetAddress() 3634 .GetFileAddress()); 3635 // We just need the flags from the linker 3636 // symbol, so put these flags 3637 // into the N_GSYM flags to avoid duplicate 3638 // symbols in the symbol table 3639 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 3640 nlist.n_desc); 3641 sym[sym_idx].Clear(); 3642 continue; 3643 } 3644 } 3645 } 3646 } 3647 } 3648 3649 sym[sym_idx].SetID(nlist_idx); 3650 sym[sym_idx].SetType(type); 3651 if (set_value) { 3652 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 3653 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3654 add_symbol_addr( 3655 sym[sym_idx].GetAddress().GetFileAddress()); 3656 } 3657 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 3658 3659 if (symbol_byte_size > 0) 3660 sym[sym_idx].SetByteSize(symbol_byte_size); 3661 3662 if (demangled_is_synthesized) 3663 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3664 ++sym_idx; 3665 } else { 3666 sym[sym_idx].Clear(); 3667 } 3668 } 3669 ///////////////////////////// 3670 } 3671 } 3672 3673 for (const auto &pos : reexport_shlib_needs_fixup) { 3674 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3675 if (undef_pos != undefined_name_to_desc.end()) { 3676 const uint8_t dylib_ordinal = 3677 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3678 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3679 sym[pos.first].SetReExportedSymbolSharedLibrary( 3680 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3681 } 3682 } 3683 } 3684 3685 #endif 3686 lldb::offset_t nlist_data_offset = 0; 3687 3688 if (nlist_data.GetByteSize() > 0) { 3689 3690 // If the sym array was not created while parsing the DSC unmapped 3691 // symbols, create it now. 3692 if (sym == nullptr) { 3693 sym = 3694 symtab.Resize(symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3695 num_syms = symtab.GetNumSymbols(); 3696 } 3697 3698 if (unmapped_local_symbols_found) { 3699 assert(m_dysymtab.ilocalsym == 0); 3700 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3701 nlist_idx = m_dysymtab.nlocalsym; 3702 } else { 3703 nlist_idx = 0; 3704 } 3705 3706 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 3707 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 3708 UndefinedNameToDescMap undefined_name_to_desc; 3709 SymbolIndexToName reexport_shlib_needs_fixup; 3710 3711 // Symtab parsing is a huge mess. Everything is entangled and the code 3712 // requires access to a ridiculous amount of variables. LLDB depends 3713 // heavily on the proper merging of symbols and to get that right we need 3714 // to make sure we have parsed all the debug symbols first. Therefore we 3715 // invoke the lambda twice, once to parse only the debug symbols and then 3716 // once more to parse the remaining symbols. 3717 auto ParseSymbolLambda = [&](struct nlist_64 &nlist, uint32_t nlist_idx, 3718 bool debug_only) { 3719 const bool is_debug = ((nlist.n_type & N_STAB) != 0); 3720 if (is_debug != debug_only) 3721 return true; 3722 3723 const char *symbol_name_non_abi_mangled = nullptr; 3724 const char *symbol_name = nullptr; 3725 3726 if (have_strtab_data) { 3727 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3728 3729 if (symbol_name == nullptr) { 3730 // No symbol should be NULL, even the symbols with no string values 3731 // should have an offset zero which points to an empty C-string 3732 Host::SystemLog(Host::eSystemLogError, 3733 "error: symbol[%u] has invalid string table offset " 3734 "0x%x in %s, ignoring symbol\n", 3735 nlist_idx, nlist.n_strx, 3736 module_sp->GetFileSpec().GetPath().c_str()); 3737 return true; 3738 } 3739 if (symbol_name[0] == '\0') 3740 symbol_name = nullptr; 3741 } else { 3742 const addr_t str_addr = strtab_addr + nlist.n_strx; 3743 Status str_error; 3744 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3745 str_error)) 3746 symbol_name = memory_symbol_name.c_str(); 3747 } 3748 3749 SymbolType type = eSymbolTypeInvalid; 3750 SectionSP symbol_section; 3751 lldb::addr_t symbol_byte_size = 0; 3752 bool add_nlist = true; 3753 bool is_gsym = false; 3754 bool demangled_is_synthesized = false; 3755 bool set_value = true; 3756 3757 assert(sym_idx < num_syms); 3758 sym[sym_idx].SetDebug(is_debug); 3759 3760 if (is_debug) { 3761 switch (nlist.n_type) { 3762 case N_GSYM: 3763 // global symbol: name,,NO_SECT,type,0 3764 // Sometimes the N_GSYM value contains the address. 3765 3766 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3767 // the ObjC data. They 3768 // have the same address, but we want to ensure that we always find 3769 // only the real symbol, 'cause we don't currently correctly 3770 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3771 // type. This is a temporary hack to make sure the ObjectiveC 3772 // symbols get treated correctly. To do this right, we should 3773 // coalesce all the GSYM & global symbols that have the same 3774 // address. 3775 is_gsym = true; 3776 sym[sym_idx].SetExternal(true); 3777 3778 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3779 llvm::StringRef symbol_name_ref(symbol_name); 3780 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3781 symbol_name_non_abi_mangled = symbol_name + 1; 3782 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3783 type = eSymbolTypeObjCClass; 3784 demangled_is_synthesized = true; 3785 3786 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 3787 symbol_name_non_abi_mangled = symbol_name + 1; 3788 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3789 type = eSymbolTypeObjCMetaClass; 3790 demangled_is_synthesized = true; 3791 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3792 symbol_name_non_abi_mangled = symbol_name + 1; 3793 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3794 type = eSymbolTypeObjCIVar; 3795 demangled_is_synthesized = true; 3796 } 3797 } else { 3798 if (nlist.n_value != 0) 3799 symbol_section = 3800 section_info.GetSection(nlist.n_sect, nlist.n_value); 3801 type = eSymbolTypeData; 3802 } 3803 break; 3804 3805 case N_FNAME: 3806 // procedure name (f77 kludge): name,,NO_SECT,0,0 3807 type = eSymbolTypeCompiler; 3808 break; 3809 3810 case N_FUN: 3811 // procedure: name,,n_sect,linenumber,address 3812 if (symbol_name) { 3813 type = eSymbolTypeCode; 3814 symbol_section = 3815 section_info.GetSection(nlist.n_sect, nlist.n_value); 3816 3817 N_FUN_addr_to_sym_idx.insert( 3818 std::make_pair(nlist.n_value, sym_idx)); 3819 // We use the current number of symbols in the symbol table in 3820 // lieu of using nlist_idx in case we ever start trimming entries 3821 // out 3822 N_FUN_indexes.push_back(sym_idx); 3823 } else { 3824 type = eSymbolTypeCompiler; 3825 3826 if (!N_FUN_indexes.empty()) { 3827 // Copy the size of the function into the original STAB entry 3828 // so we don't have to hunt for it later 3829 symtab.SymbolAtIndex(N_FUN_indexes.back()) 3830 ->SetByteSize(nlist.n_value); 3831 N_FUN_indexes.pop_back(); 3832 // We don't really need the end function STAB as it contains 3833 // the size which we already placed with the original symbol, 3834 // so don't add it if we want a minimal symbol table 3835 add_nlist = false; 3836 } 3837 } 3838 break; 3839 3840 case N_STSYM: 3841 // static symbol: name,,n_sect,type,address 3842 N_STSYM_addr_to_sym_idx.insert( 3843 std::make_pair(nlist.n_value, sym_idx)); 3844 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3845 if (symbol_name && symbol_name[0]) { 3846 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3847 eSymbolTypeData); 3848 } 3849 break; 3850 3851 case N_LCSYM: 3852 // .lcomm symbol: name,,n_sect,type,address 3853 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3854 type = eSymbolTypeCommonBlock; 3855 break; 3856 3857 case N_BNSYM: 3858 // We use the current number of symbols in the symbol table in lieu 3859 // of using nlist_idx in case we ever start trimming entries out 3860 // Skip these if we want minimal symbol tables 3861 add_nlist = false; 3862 break; 3863 3864 case N_ENSYM: 3865 // Set the size of the N_BNSYM to the terminating index of this 3866 // N_ENSYM so that we can always skip the entire symbol if we need 3867 // to navigate more quickly at the source level when parsing STABS 3868 // Skip these if we want minimal symbol tables 3869 add_nlist = false; 3870 break; 3871 3872 case N_OPT: 3873 // emitted with gcc2_compiled and in gcc source 3874 type = eSymbolTypeCompiler; 3875 break; 3876 3877 case N_RSYM: 3878 // register sym: name,,NO_SECT,type,register 3879 type = eSymbolTypeVariable; 3880 break; 3881 3882 case N_SLINE: 3883 // src line: 0,,n_sect,linenumber,address 3884 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3885 type = eSymbolTypeLineEntry; 3886 break; 3887 3888 case N_SSYM: 3889 // structure elt: name,,NO_SECT,type,struct_offset 3890 type = eSymbolTypeVariableType; 3891 break; 3892 3893 case N_SO: 3894 // source file name 3895 type = eSymbolTypeSourceFile; 3896 if (symbol_name == nullptr) { 3897 add_nlist = false; 3898 if (N_SO_index != UINT32_MAX) { 3899 // Set the size of the N_SO to the terminating index of this 3900 // N_SO so that we can always skip the entire N_SO if we need 3901 // to navigate more quickly at the source level when parsing 3902 // STABS 3903 symbol_ptr = symtab.SymbolAtIndex(N_SO_index); 3904 symbol_ptr->SetByteSize(sym_idx); 3905 symbol_ptr->SetSizeIsSibling(true); 3906 } 3907 N_NSYM_indexes.clear(); 3908 N_INCL_indexes.clear(); 3909 N_BRAC_indexes.clear(); 3910 N_COMM_indexes.clear(); 3911 N_FUN_indexes.clear(); 3912 N_SO_index = UINT32_MAX; 3913 } else { 3914 // We use the current number of symbols in the symbol table in 3915 // lieu of using nlist_idx in case we ever start trimming entries 3916 // out 3917 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3918 if (N_SO_has_full_path) { 3919 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3920 // We have two consecutive N_SO entries where the first 3921 // contains a directory and the second contains a full path. 3922 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), 3923 false); 3924 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3925 add_nlist = false; 3926 } else { 3927 // This is the first entry in a N_SO that contains a 3928 // directory or a full path to the source file 3929 N_SO_index = sym_idx; 3930 } 3931 } else if ((N_SO_index == sym_idx - 1) && 3932 ((sym_idx - 1) < num_syms)) { 3933 // This is usually the second N_SO entry that contains just the 3934 // filename, so here we combine it with the first one if we are 3935 // minimizing the symbol table 3936 const char *so_path = 3937 sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3938 if (so_path && so_path[0]) { 3939 std::string full_so_path(so_path); 3940 const size_t double_slash_pos = full_so_path.find("//"); 3941 if (double_slash_pos != std::string::npos) { 3942 // The linker has been generating bad N_SO entries with 3943 // doubled up paths in the format "%s%s" where the first 3944 // string in the DW_AT_comp_dir, and the second is the 3945 // directory for the source file so you end up with a path 3946 // that looks like "/tmp/src//tmp/src/" 3947 FileSpec so_dir(so_path); 3948 if (!FileSystem::Instance().Exists(so_dir)) { 3949 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 3950 FileSpec::Style::native); 3951 if (FileSystem::Instance().Exists(so_dir)) { 3952 // Trim off the incorrect path 3953 full_so_path.erase(0, double_slash_pos + 1); 3954 } 3955 } 3956 } 3957 if (*full_so_path.rbegin() != '/') 3958 full_so_path += '/'; 3959 full_so_path += symbol_name; 3960 sym[sym_idx - 1].GetMangled().SetValue( 3961 ConstString(full_so_path.c_str()), false); 3962 add_nlist = false; 3963 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3964 } 3965 } else { 3966 // This could be a relative path to a N_SO 3967 N_SO_index = sym_idx; 3968 } 3969 } 3970 break; 3971 3972 case N_OSO: 3973 // object file name: name,,0,0,st_mtime 3974 type = eSymbolTypeObjectFile; 3975 break; 3976 3977 case N_LSYM: 3978 // local sym: name,,NO_SECT,type,offset 3979 type = eSymbolTypeLocal; 3980 break; 3981 3982 // INCL scopes 3983 case N_BINCL: 3984 // include file beginning: name,,NO_SECT,0,sum We use the current 3985 // number of symbols in the symbol table in lieu of using nlist_idx 3986 // in case we ever start trimming entries out 3987 N_INCL_indexes.push_back(sym_idx); 3988 type = eSymbolTypeScopeBegin; 3989 break; 3990 3991 case N_EINCL: 3992 // include file end: name,,NO_SECT,0,0 3993 // Set the size of the N_BINCL to the terminating index of this 3994 // N_EINCL so that we can always skip the entire symbol if we need 3995 // to navigate more quickly at the source level when parsing STABS 3996 if (!N_INCL_indexes.empty()) { 3997 symbol_ptr = symtab.SymbolAtIndex(N_INCL_indexes.back()); 3998 symbol_ptr->SetByteSize(sym_idx + 1); 3999 symbol_ptr->SetSizeIsSibling(true); 4000 N_INCL_indexes.pop_back(); 4001 } 4002 type = eSymbolTypeScopeEnd; 4003 break; 4004 4005 case N_SOL: 4006 // #included file name: name,,n_sect,0,address 4007 type = eSymbolTypeHeaderFile; 4008 4009 // We currently don't use the header files on darwin 4010 add_nlist = false; 4011 break; 4012 4013 case N_PARAMS: 4014 // compiler parameters: name,,NO_SECT,0,0 4015 type = eSymbolTypeCompiler; 4016 break; 4017 4018 case N_VERSION: 4019 // compiler version: name,,NO_SECT,0,0 4020 type = eSymbolTypeCompiler; 4021 break; 4022 4023 case N_OLEVEL: 4024 // compiler -O level: name,,NO_SECT,0,0 4025 type = eSymbolTypeCompiler; 4026 break; 4027 4028 case N_PSYM: 4029 // parameter: name,,NO_SECT,type,offset 4030 type = eSymbolTypeVariable; 4031 break; 4032 4033 case N_ENTRY: 4034 // alternate entry: name,,n_sect,linenumber,address 4035 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4036 type = eSymbolTypeLineEntry; 4037 break; 4038 4039 // Left and Right Braces 4040 case N_LBRAC: 4041 // left bracket: 0,,NO_SECT,nesting level,address We use the 4042 // current number of symbols in the symbol table in lieu of using 4043 // nlist_idx in case we ever start trimming entries out 4044 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4045 N_BRAC_indexes.push_back(sym_idx); 4046 type = eSymbolTypeScopeBegin; 4047 break; 4048 4049 case N_RBRAC: 4050 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4051 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4052 // can always skip the entire symbol if we need to navigate more 4053 // quickly at the source level when parsing STABS 4054 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4055 if (!N_BRAC_indexes.empty()) { 4056 symbol_ptr = symtab.SymbolAtIndex(N_BRAC_indexes.back()); 4057 symbol_ptr->SetByteSize(sym_idx + 1); 4058 symbol_ptr->SetSizeIsSibling(true); 4059 N_BRAC_indexes.pop_back(); 4060 } 4061 type = eSymbolTypeScopeEnd; 4062 break; 4063 4064 case N_EXCL: 4065 // deleted include file: name,,NO_SECT,0,sum 4066 type = eSymbolTypeHeaderFile; 4067 break; 4068 4069 // COMM scopes 4070 case N_BCOMM: 4071 // begin common: name,,NO_SECT,0,0 4072 // We use the current number of symbols in the symbol table in lieu 4073 // of using nlist_idx in case we ever start trimming entries out 4074 type = eSymbolTypeScopeBegin; 4075 N_COMM_indexes.push_back(sym_idx); 4076 break; 4077 4078 case N_ECOML: 4079 // end common (local name): 0,,n_sect,0,address 4080 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4081 LLVM_FALLTHROUGH; 4082 4083 case N_ECOMM: 4084 // end common: name,,n_sect,0,0 4085 // Set the size of the N_BCOMM to the terminating index of this 4086 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4087 // we need to navigate more quickly at the source level when 4088 // parsing STABS 4089 if (!N_COMM_indexes.empty()) { 4090 symbol_ptr = symtab.SymbolAtIndex(N_COMM_indexes.back()); 4091 symbol_ptr->SetByteSize(sym_idx + 1); 4092 symbol_ptr->SetSizeIsSibling(true); 4093 N_COMM_indexes.pop_back(); 4094 } 4095 type = eSymbolTypeScopeEnd; 4096 break; 4097 4098 case N_LENG: 4099 // second stab entry with length information 4100 type = eSymbolTypeAdditional; 4101 break; 4102 4103 default: 4104 break; 4105 } 4106 } else { 4107 uint8_t n_type = N_TYPE & nlist.n_type; 4108 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4109 4110 switch (n_type) { 4111 case N_INDR: { 4112 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4113 if (reexport_name_cstr && reexport_name_cstr[0]) { 4114 type = eSymbolTypeReExported; 4115 ConstString reexport_name(reexport_name_cstr + 4116 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4117 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4118 set_value = false; 4119 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4120 indirect_symbol_names.insert( 4121 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4122 } else 4123 type = eSymbolTypeUndefined; 4124 } break; 4125 4126 case N_UNDF: 4127 if (symbol_name && symbol_name[0]) { 4128 ConstString undefined_name(symbol_name + 4129 ((symbol_name[0] == '_') ? 1 : 0)); 4130 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4131 } 4132 LLVM_FALLTHROUGH; 4133 4134 case N_PBUD: 4135 type = eSymbolTypeUndefined; 4136 break; 4137 4138 case N_ABS: 4139 type = eSymbolTypeAbsolute; 4140 break; 4141 4142 case N_SECT: { 4143 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4144 4145 if (!symbol_section) { 4146 // TODO: warn about this? 4147 add_nlist = false; 4148 break; 4149 } 4150 4151 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4152 type = eSymbolTypeException; 4153 } else { 4154 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4155 4156 switch (section_type) { 4157 case S_CSTRING_LITERALS: 4158 type = eSymbolTypeData; 4159 break; // section with only literal C strings 4160 case S_4BYTE_LITERALS: 4161 type = eSymbolTypeData; 4162 break; // section with only 4 byte literals 4163 case S_8BYTE_LITERALS: 4164 type = eSymbolTypeData; 4165 break; // section with only 8 byte literals 4166 case S_LITERAL_POINTERS: 4167 type = eSymbolTypeTrampoline; 4168 break; // section with only pointers to literals 4169 case S_NON_LAZY_SYMBOL_POINTERS: 4170 type = eSymbolTypeTrampoline; 4171 break; // section with only non-lazy symbol pointers 4172 case S_LAZY_SYMBOL_POINTERS: 4173 type = eSymbolTypeTrampoline; 4174 break; // section with only lazy symbol pointers 4175 case S_SYMBOL_STUBS: 4176 type = eSymbolTypeTrampoline; 4177 break; // section with only symbol stubs, byte size of stub in 4178 // the reserved2 field 4179 case S_MOD_INIT_FUNC_POINTERS: 4180 type = eSymbolTypeCode; 4181 break; // section with only function pointers for initialization 4182 case S_MOD_TERM_FUNC_POINTERS: 4183 type = eSymbolTypeCode; 4184 break; // section with only function pointers for termination 4185 case S_INTERPOSING: 4186 type = eSymbolTypeTrampoline; 4187 break; // section with only pairs of function pointers for 4188 // interposing 4189 case S_16BYTE_LITERALS: 4190 type = eSymbolTypeData; 4191 break; // section with only 16 byte literals 4192 case S_DTRACE_DOF: 4193 type = eSymbolTypeInstrumentation; 4194 break; 4195 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4196 type = eSymbolTypeTrampoline; 4197 break; 4198 default: 4199 switch (symbol_section->GetType()) { 4200 case lldb::eSectionTypeCode: 4201 type = eSymbolTypeCode; 4202 break; 4203 case eSectionTypeData: 4204 case eSectionTypeDataCString: // Inlined C string data 4205 case eSectionTypeDataCStringPointers: // Pointers to C string 4206 // data 4207 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4208 // the symbol table 4209 case eSectionTypeData4: 4210 case eSectionTypeData8: 4211 case eSectionTypeData16: 4212 type = eSymbolTypeData; 4213 break; 4214 default: 4215 break; 4216 } 4217 break; 4218 } 4219 4220 if (type == eSymbolTypeInvalid) { 4221 const char *symbol_sect_name = 4222 symbol_section->GetName().AsCString(); 4223 if (symbol_section->IsDescendant(text_section_sp.get())) { 4224 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4225 S_ATTR_SELF_MODIFYING_CODE | 4226 S_ATTR_SOME_INSTRUCTIONS)) 4227 type = eSymbolTypeData; 4228 else 4229 type = eSymbolTypeCode; 4230 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 4231 symbol_section->IsDescendant( 4232 data_dirty_section_sp.get()) || 4233 symbol_section->IsDescendant( 4234 data_const_section_sp.get())) { 4235 if (symbol_sect_name && 4236 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 4237 type = eSymbolTypeRuntime; 4238 4239 if (symbol_name) { 4240 llvm::StringRef symbol_name_ref(symbol_name); 4241 if (symbol_name_ref.startswith("_OBJC_")) { 4242 llvm::StringRef g_objc_v2_prefix_class( 4243 "_OBJC_CLASS_$_"); 4244 llvm::StringRef g_objc_v2_prefix_metaclass( 4245 "_OBJC_METACLASS_$_"); 4246 llvm::StringRef g_objc_v2_prefix_ivar( 4247 "_OBJC_IVAR_$_"); 4248 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 4249 symbol_name_non_abi_mangled = symbol_name + 1; 4250 symbol_name = 4251 symbol_name + g_objc_v2_prefix_class.size(); 4252 type = eSymbolTypeObjCClass; 4253 demangled_is_synthesized = true; 4254 } else if (symbol_name_ref.startswith( 4255 g_objc_v2_prefix_metaclass)) { 4256 symbol_name_non_abi_mangled = symbol_name + 1; 4257 symbol_name = 4258 symbol_name + g_objc_v2_prefix_metaclass.size(); 4259 type = eSymbolTypeObjCMetaClass; 4260 demangled_is_synthesized = true; 4261 } else if (symbol_name_ref.startswith( 4262 g_objc_v2_prefix_ivar)) { 4263 symbol_name_non_abi_mangled = symbol_name + 1; 4264 symbol_name = 4265 symbol_name + g_objc_v2_prefix_ivar.size(); 4266 type = eSymbolTypeObjCIVar; 4267 demangled_is_synthesized = true; 4268 } 4269 } 4270 } 4271 } else if (symbol_sect_name && 4272 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4273 symbol_sect_name) { 4274 type = eSymbolTypeException; 4275 } else { 4276 type = eSymbolTypeData; 4277 } 4278 } else if (symbol_sect_name && 4279 ::strstr(symbol_sect_name, "__IMPORT") == 4280 symbol_sect_name) { 4281 type = eSymbolTypeTrampoline; 4282 } else if (symbol_section->IsDescendant(objc_section_sp.get())) { 4283 type = eSymbolTypeRuntime; 4284 if (symbol_name && symbol_name[0] == '.') { 4285 llvm::StringRef symbol_name_ref(symbol_name); 4286 llvm::StringRef g_objc_v1_prefix_class( 4287 ".objc_class_name_"); 4288 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4289 symbol_name_non_abi_mangled = symbol_name; 4290 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4291 type = eSymbolTypeObjCClass; 4292 demangled_is_synthesized = true; 4293 } 4294 } 4295 } 4296 } 4297 } 4298 } break; 4299 } 4300 } 4301 4302 if (!add_nlist) { 4303 sym[sym_idx].Clear(); 4304 return true; 4305 } 4306 4307 uint64_t symbol_value = nlist.n_value; 4308 4309 if (symbol_name_non_abi_mangled) { 4310 sym[sym_idx].GetMangled().SetMangledName( 4311 ConstString(symbol_name_non_abi_mangled)); 4312 sym[sym_idx].GetMangled().SetDemangledName(ConstString(symbol_name)); 4313 } else { 4314 bool symbol_name_is_mangled = false; 4315 4316 if (symbol_name && symbol_name[0] == '_') { 4317 symbol_name_is_mangled = symbol_name[1] == '_'; 4318 symbol_name++; // Skip the leading underscore 4319 } 4320 4321 if (symbol_name) { 4322 ConstString const_symbol_name(symbol_name); 4323 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4324 symbol_name_is_mangled); 4325 } 4326 } 4327 4328 if (is_gsym) { 4329 const char *gsym_name = sym[sym_idx] 4330 .GetMangled() 4331 .GetName(Mangled::ePreferMangled) 4332 .GetCString(); 4333 if (gsym_name) 4334 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4335 } 4336 4337 if (symbol_section) { 4338 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4339 if (symbol_byte_size == 0 && function_starts_count > 0) { 4340 addr_t symbol_lookup_file_addr = nlist.n_value; 4341 // Do an exact address match for non-ARM addresses, else get the 4342 // closest since the symbol might be a thumb symbol which has an 4343 // address with bit zero set. 4344 FunctionStarts::Entry *func_start_entry = 4345 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4346 if (is_arm && func_start_entry) { 4347 // Verify that the function start address is the symbol address 4348 // (ARM) or the symbol address + 1 (thumb). 4349 if (func_start_entry->addr != symbol_lookup_file_addr && 4350 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4351 // Not the right entry, NULL it out... 4352 func_start_entry = nullptr; 4353 } 4354 } 4355 if (func_start_entry) { 4356 func_start_entry->data = true; 4357 4358 addr_t symbol_file_addr = func_start_entry->addr; 4359 if (is_arm) 4360 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4361 4362 const FunctionStarts::Entry *next_func_start_entry = 4363 function_starts.FindNextEntry(func_start_entry); 4364 const addr_t section_end_file_addr = 4365 section_file_addr + symbol_section->GetByteSize(); 4366 if (next_func_start_entry) { 4367 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4368 // Be sure the clear the Thumb address bit when we calculate the 4369 // size from the current and next address 4370 if (is_arm) 4371 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4372 symbol_byte_size = std::min<lldb::addr_t>( 4373 next_symbol_file_addr - symbol_file_addr, 4374 section_end_file_addr - symbol_file_addr); 4375 } else { 4376 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4377 } 4378 } 4379 } 4380 symbol_value -= section_file_addr; 4381 } 4382 4383 if (!is_debug) { 4384 if (type == eSymbolTypeCode) { 4385 // See if we can find a N_FUN entry for any code symbols. If we do 4386 // find a match, and the name matches, then we can merge the two into 4387 // just the function symbol to avoid duplicate entries in the symbol 4388 // table. 4389 std::pair<ValueToSymbolIndexMap::const_iterator, 4390 ValueToSymbolIndexMap::const_iterator> 4391 range; 4392 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4393 if (range.first != range.second) { 4394 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4395 pos != range.second; ++pos) { 4396 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4397 sym[pos->second].GetMangled().GetName( 4398 Mangled::ePreferMangled)) { 4399 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4400 // We just need the flags from the linker symbol, so put these 4401 // flags into the N_FUN flags to avoid duplicate symbols in the 4402 // symbol table. 4403 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4404 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4405 if (resolver_addresses.find(nlist.n_value) != 4406 resolver_addresses.end()) 4407 sym[pos->second].SetType(eSymbolTypeResolver); 4408 sym[sym_idx].Clear(); 4409 return true; 4410 } 4411 } 4412 } else { 4413 if (resolver_addresses.find(nlist.n_value) != 4414 resolver_addresses.end()) 4415 type = eSymbolTypeResolver; 4416 } 4417 } else if (type == eSymbolTypeData || type == eSymbolTypeObjCClass || 4418 type == eSymbolTypeObjCMetaClass || 4419 type == eSymbolTypeObjCIVar) { 4420 // See if we can find a N_STSYM entry for any data symbols. If we do 4421 // find a match, and the name matches, then we can merge the two into 4422 // just the Static symbol to avoid duplicate entries in the symbol 4423 // table. 4424 std::pair<ValueToSymbolIndexMap::const_iterator, 4425 ValueToSymbolIndexMap::const_iterator> 4426 range; 4427 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4428 if (range.first != range.second) { 4429 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4430 pos != range.second; ++pos) { 4431 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4432 sym[pos->second].GetMangled().GetName( 4433 Mangled::ePreferMangled)) { 4434 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4435 // We just need the flags from the linker symbol, so put these 4436 // flags into the N_STSYM flags to avoid duplicate symbols in 4437 // the symbol table. 4438 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4439 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4440 sym[sym_idx].Clear(); 4441 return true; 4442 } 4443 } 4444 } else { 4445 // Combine N_GSYM stab entries with the non stab symbol. 4446 const char *gsym_name = sym[sym_idx] 4447 .GetMangled() 4448 .GetName(Mangled::ePreferMangled) 4449 .GetCString(); 4450 if (gsym_name) { 4451 ConstNameToSymbolIndexMap::const_iterator pos = 4452 N_GSYM_name_to_sym_idx.find(gsym_name); 4453 if (pos != N_GSYM_name_to_sym_idx.end()) { 4454 const uint32_t GSYM_sym_idx = pos->second; 4455 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4456 // Copy the address, because often the N_GSYM address has an 4457 // invalid address of zero when the global is a common symbol. 4458 sym[GSYM_sym_idx].GetAddressRef().SetSection(symbol_section); 4459 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4460 add_symbol_addr( 4461 sym[GSYM_sym_idx].GetAddress().GetFileAddress()); 4462 // We just need the flags from the linker symbol, so put these 4463 // flags into the N_GSYM flags to avoid duplicate symbols in 4464 // the symbol table. 4465 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4466 sym[sym_idx].Clear(); 4467 return true; 4468 } 4469 } 4470 } 4471 } 4472 } 4473 4474 sym[sym_idx].SetID(nlist_idx); 4475 sym[sym_idx].SetType(type); 4476 if (set_value) { 4477 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4478 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4479 if (symbol_section) 4480 add_symbol_addr(sym[sym_idx].GetAddress().GetFileAddress()); 4481 } 4482 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4483 if (nlist.n_desc & N_WEAK_REF) 4484 sym[sym_idx].SetIsWeak(true); 4485 4486 if (symbol_byte_size > 0) 4487 sym[sym_idx].SetByteSize(symbol_byte_size); 4488 4489 if (demangled_is_synthesized) 4490 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4491 4492 ++sym_idx; 4493 return true; 4494 }; 4495 4496 // First parse all the nlists but don't process them yet. See the next 4497 // comment for an explanation why. 4498 std::vector<struct nlist_64> nlists; 4499 nlists.reserve(symtab_load_command.nsyms); 4500 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 4501 if (auto nlist = 4502 ParseNList(nlist_data, nlist_data_offset, nlist_byte_size)) 4503 nlists.push_back(*nlist); 4504 else 4505 break; 4506 } 4507 4508 // Now parse all the debug symbols. This is needed to merge non-debug 4509 // symbols in the next step. Non-debug symbols are always coalesced into 4510 // the debug symbol. Doing this in one step would mean that some symbols 4511 // won't be merged. 4512 nlist_idx = 0; 4513 for (auto &nlist : nlists) { 4514 if (!ParseSymbolLambda(nlist, nlist_idx++, DebugSymbols)) 4515 break; 4516 } 4517 4518 // Finally parse all the non debug symbols. 4519 nlist_idx = 0; 4520 for (auto &nlist : nlists) { 4521 if (!ParseSymbolLambda(nlist, nlist_idx++, NonDebugSymbols)) 4522 break; 4523 } 4524 4525 for (const auto &pos : reexport_shlib_needs_fixup) { 4526 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4527 if (undef_pos != undefined_name_to_desc.end()) { 4528 const uint8_t dylib_ordinal = 4529 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4530 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4531 sym[pos.first].SetReExportedSymbolSharedLibrary( 4532 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4533 } 4534 } 4535 } 4536 4537 // Count how many trie symbols we'll add to the symbol table 4538 int trie_symbol_table_augment_count = 0; 4539 for (auto &e : external_sym_trie_entries) { 4540 if (symbols_added.find(e.entry.address) == symbols_added.end()) 4541 trie_symbol_table_augment_count++; 4542 } 4543 4544 if (num_syms < sym_idx + trie_symbol_table_augment_count) { 4545 num_syms = sym_idx + trie_symbol_table_augment_count; 4546 sym = symtab.Resize(num_syms); 4547 } 4548 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4549 4550 // Add symbols from the trie to the symbol table. 4551 for (auto &e : external_sym_trie_entries) { 4552 if (symbols_added.contains(e.entry.address)) 4553 continue; 4554 4555 // Find the section that this trie address is in, use that to annotate 4556 // symbol type as we add the trie address and name to the symbol table. 4557 Address symbol_addr; 4558 if (module_sp->ResolveFileAddress(e.entry.address, symbol_addr)) { 4559 SectionSP symbol_section(symbol_addr.GetSection()); 4560 const char *symbol_name = e.entry.name.GetCString(); 4561 bool demangled_is_synthesized = false; 4562 SymbolType type = 4563 GetSymbolType(symbol_name, demangled_is_synthesized, text_section_sp, 4564 data_section_sp, data_dirty_section_sp, 4565 data_const_section_sp, symbol_section); 4566 4567 sym[sym_idx].SetType(type); 4568 if (symbol_section) { 4569 sym[sym_idx].SetID(synthetic_sym_id++); 4570 sym[sym_idx].GetMangled().SetMangledName(ConstString(symbol_name)); 4571 if (demangled_is_synthesized) 4572 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4573 sym[sym_idx].SetIsSynthetic(true); 4574 sym[sym_idx].SetExternal(true); 4575 sym[sym_idx].GetAddressRef() = symbol_addr; 4576 add_symbol_addr(symbol_addr.GetFileAddress()); 4577 if (e.entry.flags & TRIE_SYMBOL_IS_THUMB) 4578 sym[sym_idx].SetFlags(MACHO_NLIST_ARM_SYMBOL_IS_THUMB); 4579 ++sym_idx; 4580 } 4581 } 4582 } 4583 4584 if (function_starts_count > 0) { 4585 uint32_t num_synthetic_function_symbols = 0; 4586 for (i = 0; i < function_starts_count; ++i) { 4587 if (symbols_added.find(function_starts.GetEntryRef(i).addr) == 4588 symbols_added.end()) 4589 ++num_synthetic_function_symbols; 4590 } 4591 4592 if (num_synthetic_function_symbols > 0) { 4593 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4594 num_syms = sym_idx + num_synthetic_function_symbols; 4595 sym = symtab.Resize(num_syms); 4596 } 4597 for (i = 0; i < function_starts_count; ++i) { 4598 const FunctionStarts::Entry *func_start_entry = 4599 function_starts.GetEntryAtIndex(i); 4600 if (symbols_added.find(func_start_entry->addr) == symbols_added.end()) { 4601 addr_t symbol_file_addr = func_start_entry->addr; 4602 uint32_t symbol_flags = 0; 4603 if (func_start_entry->data) 4604 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4605 Address symbol_addr; 4606 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4607 SectionSP symbol_section(symbol_addr.GetSection()); 4608 uint32_t symbol_byte_size = 0; 4609 if (symbol_section) { 4610 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4611 const FunctionStarts::Entry *next_func_start_entry = 4612 function_starts.FindNextEntry(func_start_entry); 4613 const addr_t section_end_file_addr = 4614 section_file_addr + symbol_section->GetByteSize(); 4615 if (next_func_start_entry) { 4616 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4617 if (is_arm) 4618 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4619 symbol_byte_size = std::min<lldb::addr_t>( 4620 next_symbol_file_addr - symbol_file_addr, 4621 section_end_file_addr - symbol_file_addr); 4622 } else { 4623 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4624 } 4625 sym[sym_idx].SetID(synthetic_sym_id++); 4626 // Don't set the name for any synthetic symbols, the Symbol 4627 // object will generate one if needed when the name is accessed 4628 // via accessors. 4629 sym[sym_idx].GetMangled().SetDemangledName(ConstString()); 4630 sym[sym_idx].SetType(eSymbolTypeCode); 4631 sym[sym_idx].SetIsSynthetic(true); 4632 sym[sym_idx].GetAddressRef() = symbol_addr; 4633 add_symbol_addr(symbol_addr.GetFileAddress()); 4634 if (symbol_flags) 4635 sym[sym_idx].SetFlags(symbol_flags); 4636 if (symbol_byte_size) 4637 sym[sym_idx].SetByteSize(symbol_byte_size); 4638 ++sym_idx; 4639 } 4640 } 4641 } 4642 } 4643 } 4644 } 4645 4646 // Trim our symbols down to just what we ended up with after removing any 4647 // symbols. 4648 if (sym_idx < num_syms) { 4649 num_syms = sym_idx; 4650 sym = symtab.Resize(num_syms); 4651 } 4652 4653 // Now synthesize indirect symbols 4654 if (m_dysymtab.nindirectsyms != 0) { 4655 if (indirect_symbol_index_data.GetByteSize()) { 4656 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4657 m_nlist_idx_to_sym_idx.end(); 4658 4659 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4660 ++sect_idx) { 4661 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4662 S_SYMBOL_STUBS) { 4663 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4664 if (symbol_stub_byte_size == 0) 4665 continue; 4666 4667 const uint32_t num_symbol_stubs = 4668 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4669 4670 if (num_symbol_stubs == 0) 4671 continue; 4672 4673 const uint32_t symbol_stub_index_offset = 4674 m_mach_sections[sect_idx].reserved1; 4675 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) { 4676 const uint32_t symbol_stub_index = 4677 symbol_stub_index_offset + stub_idx; 4678 const lldb::addr_t symbol_stub_addr = 4679 m_mach_sections[sect_idx].addr + 4680 (stub_idx * symbol_stub_byte_size); 4681 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4682 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4683 symbol_stub_offset, 4)) { 4684 const uint32_t stub_sym_id = 4685 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4686 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4687 continue; 4688 4689 NListIndexToSymbolIndexMap::const_iterator index_pos = 4690 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4691 Symbol *stub_symbol = nullptr; 4692 if (index_pos != end_index_pos) { 4693 // We have a remapping from the original nlist index to a 4694 // current symbol index, so just look this up by index 4695 stub_symbol = symtab.SymbolAtIndex(index_pos->second); 4696 } else { 4697 // We need to lookup a symbol using the original nlist symbol 4698 // index since this index is coming from the S_SYMBOL_STUBS 4699 stub_symbol = symtab.FindSymbolByID(stub_sym_id); 4700 } 4701 4702 if (stub_symbol) { 4703 Address so_addr(symbol_stub_addr, section_list); 4704 4705 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4706 // Change the external symbol into a trampoline that makes 4707 // sense These symbols were N_UNDF N_EXT, and are useless 4708 // to us, so we can re-use them so we don't have to make up 4709 // a synthetic symbol for no good reason. 4710 if (resolver_addresses.find(symbol_stub_addr) == 4711 resolver_addresses.end()) 4712 stub_symbol->SetType(eSymbolTypeTrampoline); 4713 else 4714 stub_symbol->SetType(eSymbolTypeResolver); 4715 stub_symbol->SetExternal(false); 4716 stub_symbol->GetAddressRef() = so_addr; 4717 stub_symbol->SetByteSize(symbol_stub_byte_size); 4718 } else { 4719 // Make a synthetic symbol to describe the trampoline stub 4720 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4721 if (sym_idx >= num_syms) { 4722 sym = symtab.Resize(++num_syms); 4723 stub_symbol = nullptr; // this pointer no longer valid 4724 } 4725 sym[sym_idx].SetID(synthetic_sym_id++); 4726 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4727 if (resolver_addresses.find(symbol_stub_addr) == 4728 resolver_addresses.end()) 4729 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4730 else 4731 sym[sym_idx].SetType(eSymbolTypeResolver); 4732 sym[sym_idx].SetIsSynthetic(true); 4733 sym[sym_idx].GetAddressRef() = so_addr; 4734 add_symbol_addr(so_addr.GetFileAddress()); 4735 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4736 ++sym_idx; 4737 } 4738 } else { 4739 if (log) 4740 log->Warning("symbol stub referencing symbol table symbol " 4741 "%u that isn't in our minimal symbol table, " 4742 "fix this!!!", 4743 stub_sym_id); 4744 } 4745 } 4746 } 4747 } 4748 } 4749 } 4750 } 4751 4752 if (!reexport_trie_entries.empty()) { 4753 for (const auto &e : reexport_trie_entries) { 4754 if (e.entry.import_name) { 4755 // Only add indirect symbols from the Trie entries if we didn't have 4756 // a N_INDR nlist entry for this already 4757 if (indirect_symbol_names.find(e.entry.name) == 4758 indirect_symbol_names.end()) { 4759 // Make a synthetic symbol to describe re-exported symbol. 4760 if (sym_idx >= num_syms) 4761 sym = symtab.Resize(++num_syms); 4762 sym[sym_idx].SetID(synthetic_sym_id++); 4763 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4764 sym[sym_idx].SetType(eSymbolTypeReExported); 4765 sym[sym_idx].SetIsSynthetic(true); 4766 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4767 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4768 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4769 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4770 } 4771 ++sym_idx; 4772 } 4773 } 4774 } 4775 } 4776 } 4777 4778 void ObjectFileMachO::Dump(Stream *s) { 4779 ModuleSP module_sp(GetModule()); 4780 if (module_sp) { 4781 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4782 s->Printf("%p: ", static_cast<void *>(this)); 4783 s->Indent(); 4784 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4785 s->PutCString("ObjectFileMachO64"); 4786 else 4787 s->PutCString("ObjectFileMachO32"); 4788 4789 *s << ", file = '" << m_file; 4790 ModuleSpecList all_specs; 4791 ModuleSpec base_spec; 4792 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 4793 base_spec, all_specs); 4794 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 4795 *s << "', triple"; 4796 if (e) 4797 s->Printf("[%d]", i); 4798 *s << " = "; 4799 *s << all_specs.GetModuleSpecRefAtIndex(i) 4800 .GetArchitecture() 4801 .GetTriple() 4802 .getTriple(); 4803 } 4804 *s << "\n"; 4805 SectionList *sections = GetSectionList(); 4806 if (sections) 4807 sections->Dump(s->AsRawOstream(), s->GetIndentLevel(), nullptr, true, 4808 UINT32_MAX); 4809 4810 if (m_symtab_up) 4811 m_symtab_up->Dump(s, nullptr, eSortOrderNone); 4812 } 4813 } 4814 4815 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4816 const lldb_private::DataExtractor &data, 4817 lldb::offset_t lc_offset) { 4818 uint32_t i; 4819 llvm::MachO::uuid_command load_cmd; 4820 4821 lldb::offset_t offset = lc_offset; 4822 for (i = 0; i < header.ncmds; ++i) { 4823 const lldb::offset_t cmd_offset = offset; 4824 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4825 break; 4826 4827 if (load_cmd.cmd == LC_UUID) { 4828 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4829 4830 if (uuid_bytes) { 4831 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4832 // We pretend these object files have no UUID to prevent crashing. 4833 4834 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4835 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4836 0xbb, 0x14, 0xf0, 0x0d}; 4837 4838 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4839 return UUID(); 4840 4841 return UUID::fromOptionalData(uuid_bytes, 16); 4842 } 4843 return UUID(); 4844 } 4845 offset = cmd_offset + load_cmd.cmdsize; 4846 } 4847 return UUID(); 4848 } 4849 4850 static llvm::StringRef GetOSName(uint32_t cmd) { 4851 switch (cmd) { 4852 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4853 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4854 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4855 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4856 case llvm::MachO::LC_VERSION_MIN_TVOS: 4857 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4858 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4859 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4860 default: 4861 llvm_unreachable("unexpected LC_VERSION load command"); 4862 } 4863 } 4864 4865 namespace { 4866 struct OSEnv { 4867 llvm::StringRef os_type; 4868 llvm::StringRef environment; 4869 OSEnv(uint32_t cmd) { 4870 switch (cmd) { 4871 case llvm::MachO::PLATFORM_MACOS: 4872 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4873 return; 4874 case llvm::MachO::PLATFORM_IOS: 4875 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4876 return; 4877 case llvm::MachO::PLATFORM_TVOS: 4878 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4879 return; 4880 case llvm::MachO::PLATFORM_WATCHOS: 4881 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4882 return; 4883 // TODO: add BridgeOS & DriverKit once in llvm/lib/Support/Triple.cpp 4884 // NEED_BRIDGEOS_TRIPLE 4885 // case llvm::MachO::PLATFORM_BRIDGEOS: 4886 // os_type = llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4887 // return; 4888 // case llvm::MachO::PLATFORM_DRIVERKIT: 4889 // os_type = llvm::Triple::getOSTypeName(llvm::Triple::DriverKit); 4890 // return; 4891 case llvm::MachO::PLATFORM_MACCATALYST: 4892 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4893 environment = llvm::Triple::getEnvironmentTypeName(llvm::Triple::MacABI); 4894 return; 4895 case llvm::MachO::PLATFORM_IOSSIMULATOR: 4896 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4897 environment = 4898 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4899 return; 4900 case llvm::MachO::PLATFORM_TVOSSIMULATOR: 4901 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4902 environment = 4903 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4904 return; 4905 case llvm::MachO::PLATFORM_WATCHOSSIMULATOR: 4906 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4907 environment = 4908 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4909 return; 4910 default: { 4911 Log *log(GetLog(LLDBLog::Symbols | LLDBLog::Process)); 4912 LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION"); 4913 } 4914 } 4915 } 4916 }; 4917 4918 struct MinOS { 4919 uint32_t major_version, minor_version, patch_version; 4920 MinOS(uint32_t version) 4921 : major_version(version >> 16), minor_version((version >> 8) & 0xffu), 4922 patch_version(version & 0xffu) {} 4923 }; 4924 } // namespace 4925 4926 void ObjectFileMachO::GetAllArchSpecs(const llvm::MachO::mach_header &header, 4927 const lldb_private::DataExtractor &data, 4928 lldb::offset_t lc_offset, 4929 ModuleSpec &base_spec, 4930 lldb_private::ModuleSpecList &all_specs) { 4931 auto &base_arch = base_spec.GetArchitecture(); 4932 base_arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4933 if (!base_arch.IsValid()) 4934 return; 4935 4936 bool found_any = false; 4937 auto add_triple = [&](const llvm::Triple &triple) { 4938 auto spec = base_spec; 4939 spec.GetArchitecture().GetTriple() = triple; 4940 if (spec.GetArchitecture().IsValid()) { 4941 spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset); 4942 all_specs.Append(spec); 4943 found_any = true; 4944 } 4945 }; 4946 4947 // Set OS to an unspecified unknown or a "*" so it can match any OS 4948 llvm::Triple base_triple = base_arch.GetTriple(); 4949 base_triple.setOS(llvm::Triple::UnknownOS); 4950 base_triple.setOSName(llvm::StringRef()); 4951 4952 if (header.filetype == MH_PRELOAD) { 4953 if (header.cputype == CPU_TYPE_ARM) { 4954 // If this is a 32-bit arm binary, and it's a standalone binary, force 4955 // the Vendor to Apple so we don't accidentally pick up the generic 4956 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4957 // frame pointer register; most other armv7 ABIs use a combination of 4958 // r7 and r11. 4959 base_triple.setVendor(llvm::Triple::Apple); 4960 } else { 4961 // Set vendor to an unspecified unknown or a "*" so it can match any 4962 // vendor This is required for correct behavior of EFI debugging on 4963 // x86_64 4964 base_triple.setVendor(llvm::Triple::UnknownVendor); 4965 base_triple.setVendorName(llvm::StringRef()); 4966 } 4967 return add_triple(base_triple); 4968 } 4969 4970 llvm::MachO::load_command load_cmd; 4971 4972 // See if there is an LC_VERSION_MIN_* load command that can give 4973 // us the OS type. 4974 lldb::offset_t offset = lc_offset; 4975 for (uint32_t i = 0; i < header.ncmds; ++i) { 4976 const lldb::offset_t cmd_offset = offset; 4977 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4978 break; 4979 4980 llvm::MachO::version_min_command version_min; 4981 switch (load_cmd.cmd) { 4982 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4983 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4984 case llvm::MachO::LC_VERSION_MIN_TVOS: 4985 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 4986 if (load_cmd.cmdsize != sizeof(version_min)) 4987 break; 4988 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 4989 data.GetByteOrder(), &version_min) == 0) 4990 break; 4991 MinOS min_os(version_min.version); 4992 llvm::SmallString<32> os_name; 4993 llvm::raw_svector_ostream os(os_name); 4994 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 4995 << min_os.minor_version << '.' << min_os.patch_version; 4996 4997 auto triple = base_triple; 4998 triple.setOSName(os.str()); 4999 5000 // Disambiguate legacy simulator platforms. 5001 if (load_cmd.cmd != llvm::MachO::LC_VERSION_MIN_MACOSX && 5002 (base_triple.getArch() == llvm::Triple::x86_64 || 5003 base_triple.getArch() == llvm::Triple::x86)) { 5004 // The combination of legacy LC_VERSION_MIN load command and 5005 // x86 architecture always indicates a simulator environment. 5006 // The combination of LC_VERSION_MIN and arm architecture only 5007 // appears for native binaries. Back-deploying simulator 5008 // binaries on Apple Silicon Macs use the modern unambigous 5009 // LC_BUILD_VERSION load commands; no special handling required. 5010 triple.setEnvironment(llvm::Triple::Simulator); 5011 } 5012 add_triple(triple); 5013 break; 5014 } 5015 default: 5016 break; 5017 } 5018 5019 offset = cmd_offset + load_cmd.cmdsize; 5020 } 5021 5022 // See if there are LC_BUILD_VERSION load commands that can give 5023 // us the OS type. 5024 offset = lc_offset; 5025 for (uint32_t i = 0; i < header.ncmds; ++i) { 5026 const lldb::offset_t cmd_offset = offset; 5027 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 5028 break; 5029 5030 do { 5031 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 5032 llvm::MachO::build_version_command build_version; 5033 if (load_cmd.cmdsize < sizeof(build_version)) { 5034 // Malformed load command. 5035 break; 5036 } 5037 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 5038 data.GetByteOrder(), &build_version) == 0) 5039 break; 5040 MinOS min_os(build_version.minos); 5041 OSEnv os_env(build_version.platform); 5042 llvm::SmallString<16> os_name; 5043 llvm::raw_svector_ostream os(os_name); 5044 os << os_env.os_type << min_os.major_version << '.' 5045 << min_os.minor_version << '.' << min_os.patch_version; 5046 auto triple = base_triple; 5047 triple.setOSName(os.str()); 5048 os_name.clear(); 5049 if (!os_env.environment.empty()) 5050 triple.setEnvironmentName(os_env.environment); 5051 add_triple(triple); 5052 } 5053 } while (false); 5054 offset = cmd_offset + load_cmd.cmdsize; 5055 } 5056 5057 if (!found_any) { 5058 add_triple(base_triple); 5059 } 5060 } 5061 5062 ArchSpec ObjectFileMachO::GetArchitecture( 5063 ModuleSP module_sp, const llvm::MachO::mach_header &header, 5064 const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) { 5065 ModuleSpecList all_specs; 5066 ModuleSpec base_spec; 5067 GetAllArchSpecs(header, data, MachHeaderSizeFromMagic(header.magic), 5068 base_spec, all_specs); 5069 5070 // If the object file offers multiple alternative load commands, 5071 // pick the one that matches the module. 5072 if (module_sp) { 5073 const ArchSpec &module_arch = module_sp->GetArchitecture(); 5074 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 5075 ArchSpec mach_arch = 5076 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 5077 if (module_arch.IsCompatibleMatch(mach_arch)) 5078 return mach_arch; 5079 } 5080 } 5081 5082 // Return the first arch we found. 5083 if (all_specs.GetSize() == 0) 5084 return {}; 5085 return all_specs.GetModuleSpecRefAtIndex(0).GetArchitecture(); 5086 } 5087 5088 UUID ObjectFileMachO::GetUUID() { 5089 ModuleSP module_sp(GetModule()); 5090 if (module_sp) { 5091 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5092 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5093 return GetUUID(m_header, m_data, offset); 5094 } 5095 return UUID(); 5096 } 5097 5098 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5099 uint32_t count = 0; 5100 ModuleSP module_sp(GetModule()); 5101 if (module_sp) { 5102 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5103 llvm::MachO::load_command load_cmd; 5104 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5105 std::vector<std::string> rpath_paths; 5106 std::vector<std::string> rpath_relative_paths; 5107 std::vector<std::string> at_exec_relative_paths; 5108 uint32_t i; 5109 for (i = 0; i < m_header.ncmds; ++i) { 5110 const uint32_t cmd_offset = offset; 5111 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5112 break; 5113 5114 switch (load_cmd.cmd) { 5115 case LC_RPATH: 5116 case LC_LOAD_DYLIB: 5117 case LC_LOAD_WEAK_DYLIB: 5118 case LC_REEXPORT_DYLIB: 5119 case LC_LOAD_DYLINKER: 5120 case LC_LOADFVMLIB: 5121 case LC_LOAD_UPWARD_DYLIB: { 5122 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5123 const char *path = m_data.PeekCStr(name_offset); 5124 if (path) { 5125 if (load_cmd.cmd == LC_RPATH) 5126 rpath_paths.push_back(path); 5127 else { 5128 if (path[0] == '@') { 5129 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5130 rpath_relative_paths.push_back(path + strlen("@rpath")); 5131 else if (strncmp(path, "@executable_path", 5132 strlen("@executable_path")) == 0) 5133 at_exec_relative_paths.push_back(path + 5134 strlen("@executable_path")); 5135 } else { 5136 FileSpec file_spec(path); 5137 if (files.AppendIfUnique(file_spec)) 5138 count++; 5139 } 5140 } 5141 } 5142 } break; 5143 5144 default: 5145 break; 5146 } 5147 offset = cmd_offset + load_cmd.cmdsize; 5148 } 5149 5150 FileSpec this_file_spec(m_file); 5151 FileSystem::Instance().Resolve(this_file_spec); 5152 5153 if (!rpath_paths.empty()) { 5154 // Fixup all LC_RPATH values to be absolute paths 5155 std::string loader_path("@loader_path"); 5156 std::string executable_path("@executable_path"); 5157 for (auto &rpath : rpath_paths) { 5158 if (llvm::StringRef(rpath).startswith(loader_path)) { 5159 rpath.erase(0, loader_path.size()); 5160 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5161 } else if (llvm::StringRef(rpath).startswith(executable_path)) { 5162 rpath.erase(0, executable_path.size()); 5163 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5164 } 5165 } 5166 5167 for (const auto &rpath_relative_path : rpath_relative_paths) { 5168 for (const auto &rpath : rpath_paths) { 5169 std::string path = rpath; 5170 path += rpath_relative_path; 5171 // It is OK to resolve this path because we must find a file on disk 5172 // for us to accept it anyway if it is rpath relative. 5173 FileSpec file_spec(path); 5174 FileSystem::Instance().Resolve(file_spec); 5175 if (FileSystem::Instance().Exists(file_spec) && 5176 files.AppendIfUnique(file_spec)) { 5177 count++; 5178 break; 5179 } 5180 } 5181 } 5182 } 5183 5184 // We may have @executable_paths but no RPATHS. Figure those out here. 5185 // Only do this if this object file is the executable. We have no way to 5186 // get back to the actual executable otherwise, so we won't get the right 5187 // path. 5188 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5189 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5190 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5191 FileSpec file_spec = 5192 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5193 if (FileSystem::Instance().Exists(file_spec) && 5194 files.AppendIfUnique(file_spec)) 5195 count++; 5196 } 5197 } 5198 } 5199 return count; 5200 } 5201 5202 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5203 // If the object file is not an executable it can't hold the entry point. 5204 // m_entry_point_address is initialized to an invalid address, so we can just 5205 // return that. If m_entry_point_address is valid it means we've found it 5206 // already, so return the cached value. 5207 5208 if ((!IsExecutable() && !IsDynamicLoader()) || 5209 m_entry_point_address.IsValid()) { 5210 return m_entry_point_address; 5211 } 5212 5213 // Otherwise, look for the UnixThread or Thread command. The data for the 5214 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5215 // 5216 // uint32_t flavor - this is the flavor argument you would pass to 5217 // thread_get_state 5218 // uint32_t count - this is the count of longs in the thread state data 5219 // struct XXX_thread_state state - this is the structure from 5220 // <machine/thread_status.h> corresponding to the flavor. 5221 // <repeat this trio> 5222 // 5223 // So we just keep reading the various register flavors till we find the GPR 5224 // one, then read the PC out of there. 5225 // FIXME: We will need to have a "RegisterContext data provider" class at some 5226 // point that can get all the registers 5227 // out of data in this form & attach them to a given thread. That should 5228 // underlie the MacOS X User process plugin, and we'll also need it for the 5229 // MacOS X Core File process plugin. When we have that we can also use it 5230 // here. 5231 // 5232 // For now we hard-code the offsets and flavors we need: 5233 // 5234 // 5235 5236 ModuleSP module_sp(GetModule()); 5237 if (module_sp) { 5238 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5239 llvm::MachO::load_command load_cmd; 5240 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5241 uint32_t i; 5242 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5243 bool done = false; 5244 5245 for (i = 0; i < m_header.ncmds; ++i) { 5246 const lldb::offset_t cmd_offset = offset; 5247 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5248 break; 5249 5250 switch (load_cmd.cmd) { 5251 case LC_UNIXTHREAD: 5252 case LC_THREAD: { 5253 while (offset < cmd_offset + load_cmd.cmdsize) { 5254 uint32_t flavor = m_data.GetU32(&offset); 5255 uint32_t count = m_data.GetU32(&offset); 5256 if (count == 0) { 5257 // We've gotten off somehow, log and exit; 5258 return m_entry_point_address; 5259 } 5260 5261 switch (m_header.cputype) { 5262 case llvm::MachO::CPU_TYPE_ARM: 5263 if (flavor == 1 || 5264 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 5265 // from mach/arm/thread_status.h 5266 { 5267 offset += 60; // This is the offset of pc in the GPR thread state 5268 // data structure. 5269 start_address = m_data.GetU32(&offset); 5270 done = true; 5271 } 5272 break; 5273 case llvm::MachO::CPU_TYPE_ARM64: 5274 case llvm::MachO::CPU_TYPE_ARM64_32: 5275 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5276 { 5277 offset += 256; // This is the offset of pc in the GPR thread state 5278 // data structure. 5279 start_address = m_data.GetU64(&offset); 5280 done = true; 5281 } 5282 break; 5283 case llvm::MachO::CPU_TYPE_I386: 5284 if (flavor == 5285 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5286 { 5287 offset += 40; // This is the offset of eip in the GPR thread state 5288 // data structure. 5289 start_address = m_data.GetU32(&offset); 5290 done = true; 5291 } 5292 break; 5293 case llvm::MachO::CPU_TYPE_X86_64: 5294 if (flavor == 5295 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5296 { 5297 offset += 16 * 8; // This is the offset of rip in the GPR thread 5298 // state data structure. 5299 start_address = m_data.GetU64(&offset); 5300 done = true; 5301 } 5302 break; 5303 default: 5304 return m_entry_point_address; 5305 } 5306 // Haven't found the GPR flavor yet, skip over the data for this 5307 // flavor: 5308 if (done) 5309 break; 5310 offset += count * 4; 5311 } 5312 } break; 5313 case LC_MAIN: { 5314 ConstString text_segment_name("__TEXT"); 5315 uint64_t entryoffset = m_data.GetU64(&offset); 5316 SectionSP text_segment_sp = 5317 GetSectionList()->FindSectionByName(text_segment_name); 5318 if (text_segment_sp) { 5319 done = true; 5320 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5321 } 5322 } break; 5323 5324 default: 5325 break; 5326 } 5327 if (done) 5328 break; 5329 5330 // Go to the next load command: 5331 offset = cmd_offset + load_cmd.cmdsize; 5332 } 5333 5334 if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) { 5335 if (GetSymtab()) { 5336 Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType( 5337 ConstString("_dyld_start"), SymbolType::eSymbolTypeCode, 5338 Symtab::eDebugAny, Symtab::eVisibilityAny); 5339 if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) { 5340 start_address = dyld_start_sym->GetAddress().GetFileAddress(); 5341 } 5342 } 5343 } 5344 5345 if (start_address != LLDB_INVALID_ADDRESS) { 5346 // We got the start address from the load commands, so now resolve that 5347 // address in the sections of this ObjectFile: 5348 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5349 start_address, GetSectionList())) { 5350 m_entry_point_address.Clear(); 5351 } 5352 } else { 5353 // We couldn't read the UnixThread load command - maybe it wasn't there. 5354 // As a fallback look for the "start" symbol in the main executable. 5355 5356 ModuleSP module_sp(GetModule()); 5357 5358 if (module_sp) { 5359 SymbolContextList contexts; 5360 SymbolContext context; 5361 module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5362 eSymbolTypeCode, contexts); 5363 if (contexts.GetSize()) { 5364 if (contexts.GetContextAtIndex(0, context)) 5365 m_entry_point_address = context.symbol->GetAddress(); 5366 } 5367 } 5368 } 5369 } 5370 5371 return m_entry_point_address; 5372 } 5373 5374 lldb_private::Address ObjectFileMachO::GetBaseAddress() { 5375 lldb_private::Address header_addr; 5376 SectionList *section_list = GetSectionList(); 5377 if (section_list) { 5378 SectionSP text_segment_sp( 5379 section_list->FindSectionByName(GetSegmentNameTEXT())); 5380 if (text_segment_sp) { 5381 header_addr.SetSection(text_segment_sp); 5382 header_addr.SetOffset(0); 5383 } 5384 } 5385 return header_addr; 5386 } 5387 5388 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5389 ModuleSP module_sp(GetModule()); 5390 if (module_sp) { 5391 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5392 if (!m_thread_context_offsets_valid) { 5393 m_thread_context_offsets_valid = true; 5394 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5395 FileRangeArray::Entry file_range; 5396 llvm::MachO::thread_command thread_cmd; 5397 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5398 const uint32_t cmd_offset = offset; 5399 if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr) 5400 break; 5401 5402 if (thread_cmd.cmd == LC_THREAD) { 5403 file_range.SetRangeBase(offset); 5404 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5405 m_thread_context_offsets.Append(file_range); 5406 } 5407 offset = cmd_offset + thread_cmd.cmdsize; 5408 } 5409 } 5410 } 5411 return m_thread_context_offsets.GetSize(); 5412 } 5413 5414 std::string ObjectFileMachO::GetIdentifierString() { 5415 std::string result; 5416 ModuleSP module_sp(GetModule()); 5417 if (module_sp) { 5418 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5419 5420 // First, look over the load commands for an LC_NOTE load command with 5421 // data_owner string "kern ver str" & use that if found. 5422 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5423 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5424 const uint32_t cmd_offset = offset; 5425 llvm::MachO::load_command lc; 5426 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5427 break; 5428 if (lc.cmd == LC_NOTE) { 5429 char data_owner[17]; 5430 m_data.CopyData(offset, 16, data_owner); 5431 data_owner[16] = '\0'; 5432 offset += 16; 5433 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5434 uint64_t size = m_data.GetU64_unchecked(&offset); 5435 5436 // "kern ver str" has a uint32_t version and then a nul terminated 5437 // c-string. 5438 if (strcmp("kern ver str", data_owner) == 0) { 5439 offset = fileoff; 5440 uint32_t version; 5441 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5442 if (version == 1) { 5443 uint32_t strsize = size - sizeof(uint32_t); 5444 char *buf = (char *)malloc(strsize); 5445 if (buf) { 5446 m_data.CopyData(offset, strsize, buf); 5447 buf[strsize - 1] = '\0'; 5448 result = buf; 5449 if (buf) 5450 free(buf); 5451 return result; 5452 } 5453 } 5454 } 5455 } 5456 } 5457 offset = cmd_offset + lc.cmdsize; 5458 } 5459 5460 // Second, make a pass over the load commands looking for an obsolete 5461 // LC_IDENT load command. 5462 offset = MachHeaderSizeFromMagic(m_header.magic); 5463 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5464 const uint32_t cmd_offset = offset; 5465 llvm::MachO::ident_command ident_command; 5466 if (m_data.GetU32(&offset, &ident_command, 2) == nullptr) 5467 break; 5468 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5469 char *buf = (char *)malloc(ident_command.cmdsize); 5470 if (buf != nullptr && m_data.CopyData(offset, ident_command.cmdsize, 5471 buf) == ident_command.cmdsize) { 5472 buf[ident_command.cmdsize - 1] = '\0'; 5473 result = buf; 5474 } 5475 if (buf) 5476 free(buf); 5477 } 5478 offset = cmd_offset + ident_command.cmdsize; 5479 } 5480 } 5481 return result; 5482 } 5483 5484 addr_t ObjectFileMachO::GetAddressMask() { 5485 addr_t mask = 0; 5486 ModuleSP module_sp(GetModule()); 5487 if (module_sp) { 5488 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5489 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5490 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5491 const uint32_t cmd_offset = offset; 5492 llvm::MachO::load_command lc; 5493 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5494 break; 5495 if (lc.cmd == LC_NOTE) { 5496 char data_owner[17]; 5497 m_data.CopyData(offset, 16, data_owner); 5498 data_owner[16] = '\0'; 5499 offset += 16; 5500 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5501 5502 // "addrable bits" has a uint32_t version and a uint32_t 5503 // number of bits used in addressing. 5504 if (strcmp("addrable bits", data_owner) == 0) { 5505 offset = fileoff; 5506 uint32_t version; 5507 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5508 if (version == 3) { 5509 uint32_t num_addr_bits = m_data.GetU32_unchecked(&offset); 5510 if (num_addr_bits != 0) { 5511 mask = ~((1ULL << num_addr_bits) - 1); 5512 } 5513 break; 5514 } 5515 } 5516 } 5517 } 5518 offset = cmd_offset + lc.cmdsize; 5519 } 5520 } 5521 return mask; 5522 } 5523 5524 bool ObjectFileMachO::GetCorefileMainBinaryInfo(addr_t &value, 5525 bool &value_is_offset, 5526 UUID &uuid, 5527 ObjectFile::BinaryType &type) { 5528 value = LLDB_INVALID_ADDRESS; 5529 value_is_offset = false; 5530 uuid.Clear(); 5531 uint32_t log2_pagesize = 0; // not currently passed up to caller 5532 uint32_t platform = 0; // not currently passed up to caller 5533 ModuleSP module_sp(GetModule()); 5534 if (module_sp) { 5535 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5536 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5537 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5538 const uint32_t cmd_offset = offset; 5539 llvm::MachO::load_command lc; 5540 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5541 break; 5542 if (lc.cmd == LC_NOTE) { 5543 char data_owner[17]; 5544 memset(data_owner, 0, sizeof(data_owner)); 5545 m_data.CopyData(offset, 16, data_owner); 5546 offset += 16; 5547 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5548 uint64_t size = m_data.GetU64_unchecked(&offset); 5549 5550 // struct main_bin_spec 5551 // { 5552 // uint32_t version; // currently 2 5553 // uint32_t type; // 0 == unspecified, 1 == kernel, 5554 // // 2 == user process, 5555 // // 3 == standalone binary 5556 // uint64_t address; // UINT64_MAX if address not specified 5557 // uint64_t slide; // slide, UINT64_MAX if unspecified 5558 // // 0 if no slide needs to be applied to 5559 // // file address 5560 // uuid_t uuid; // all zero's if uuid not specified 5561 // uint32_t log2_pagesize; // process page size in log base 2, 5562 // // e.g. 4k pages are 12. 5563 // // 0 for unspecified 5564 // uint32_t platform; // The Mach-O platform for this corefile. 5565 // // 0 for unspecified. 5566 // // The values are defined in 5567 // // <mach-o/loader.h>, PLATFORM_*. 5568 // } __attribute((packed)); 5569 5570 // "main bin spec" (main binary specification) data payload is 5571 // formatted: 5572 // uint32_t version [currently 1] 5573 // uint32_t type [0 == unspecified, 1 == kernel, 5574 // 2 == user process, 3 == firmware ] 5575 // uint64_t address [ UINT64_MAX if address not specified ] 5576 // uuid_t uuid [ all zero's if uuid not specified ] 5577 // uint32_t log2_pagesize [ process page size in log base 5578 // 2, e.g. 4k pages are 12. 5579 // 0 for unspecified ] 5580 // uint32_t unused [ for alignment ] 5581 5582 if (strcmp("main bin spec", data_owner) == 0 && size >= 32) { 5583 offset = fileoff; 5584 uint32_t version; 5585 if (m_data.GetU32(&offset, &version, 1) != nullptr && version <= 2) { 5586 uint32_t binspec_type = 0; 5587 uuid_t raw_uuid; 5588 memset(raw_uuid, 0, sizeof(uuid_t)); 5589 5590 if (!m_data.GetU32(&offset, &binspec_type, 1)) 5591 return false; 5592 if (!m_data.GetU64(&offset, &value, 1)) 5593 return false; 5594 uint64_t slide = LLDB_INVALID_ADDRESS; 5595 if (version > 1 && !m_data.GetU64(&offset, &slide, 1)) 5596 return false; 5597 if (value == LLDB_INVALID_ADDRESS && 5598 slide != LLDB_INVALID_ADDRESS) { 5599 value = slide; 5600 value_is_offset = true; 5601 } 5602 5603 if (m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5604 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5605 // convert the "main bin spec" type into our 5606 // ObjectFile::BinaryType enum 5607 switch (binspec_type) { 5608 case 0: 5609 type = eBinaryTypeUnknown; 5610 break; 5611 case 1: 5612 type = eBinaryTypeKernel; 5613 break; 5614 case 2: 5615 type = eBinaryTypeUser; 5616 break; 5617 case 3: 5618 type = eBinaryTypeStandalone; 5619 break; 5620 } 5621 if (!m_data.GetU32(&offset, &log2_pagesize, 1)) 5622 return false; 5623 if (version > 1 && !m_data.GetU32(&offset, &platform, 1)) 5624 return false; 5625 return true; 5626 } 5627 } 5628 } 5629 } 5630 offset = cmd_offset + lc.cmdsize; 5631 } 5632 } 5633 return false; 5634 } 5635 5636 lldb::RegisterContextSP 5637 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5638 lldb_private::Thread &thread) { 5639 lldb::RegisterContextSP reg_ctx_sp; 5640 5641 ModuleSP module_sp(GetModule()); 5642 if (module_sp) { 5643 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5644 if (!m_thread_context_offsets_valid) 5645 GetNumThreadContexts(); 5646 5647 const FileRangeArray::Entry *thread_context_file_range = 5648 m_thread_context_offsets.GetEntryAtIndex(idx); 5649 if (thread_context_file_range) { 5650 5651 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5652 thread_context_file_range->GetByteSize()); 5653 5654 switch (m_header.cputype) { 5655 case llvm::MachO::CPU_TYPE_ARM64: 5656 case llvm::MachO::CPU_TYPE_ARM64_32: 5657 reg_ctx_sp = 5658 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data); 5659 break; 5660 5661 case llvm::MachO::CPU_TYPE_ARM: 5662 reg_ctx_sp = 5663 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data); 5664 break; 5665 5666 case llvm::MachO::CPU_TYPE_I386: 5667 reg_ctx_sp = 5668 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data); 5669 break; 5670 5671 case llvm::MachO::CPU_TYPE_X86_64: 5672 reg_ctx_sp = 5673 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data); 5674 break; 5675 } 5676 } 5677 } 5678 return reg_ctx_sp; 5679 } 5680 5681 ObjectFile::Type ObjectFileMachO::CalculateType() { 5682 switch (m_header.filetype) { 5683 case MH_OBJECT: // 0x1u 5684 if (GetAddressByteSize() == 4) { 5685 // 32 bit kexts are just object files, but they do have a valid 5686 // UUID load command. 5687 if (GetUUID()) { 5688 // this checking for the UUID load command is not enough we could 5689 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5690 // this is required of kexts 5691 if (m_strata == eStrataInvalid) 5692 m_strata = eStrataKernel; 5693 return eTypeSharedLibrary; 5694 } 5695 } 5696 return eTypeObjectFile; 5697 5698 case MH_EXECUTE: 5699 return eTypeExecutable; // 0x2u 5700 case MH_FVMLIB: 5701 return eTypeSharedLibrary; // 0x3u 5702 case MH_CORE: 5703 return eTypeCoreFile; // 0x4u 5704 case MH_PRELOAD: 5705 return eTypeSharedLibrary; // 0x5u 5706 case MH_DYLIB: 5707 return eTypeSharedLibrary; // 0x6u 5708 case MH_DYLINKER: 5709 return eTypeDynamicLinker; // 0x7u 5710 case MH_BUNDLE: 5711 return eTypeSharedLibrary; // 0x8u 5712 case MH_DYLIB_STUB: 5713 return eTypeStubLibrary; // 0x9u 5714 case MH_DSYM: 5715 return eTypeDebugInfo; // 0xAu 5716 case MH_KEXT_BUNDLE: 5717 return eTypeSharedLibrary; // 0xBu 5718 default: 5719 break; 5720 } 5721 return eTypeUnknown; 5722 } 5723 5724 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5725 switch (m_header.filetype) { 5726 case MH_OBJECT: // 0x1u 5727 { 5728 // 32 bit kexts are just object files, but they do have a valid 5729 // UUID load command. 5730 if (GetUUID()) { 5731 // this checking for the UUID load command is not enough we could 5732 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5733 // this is required of kexts 5734 if (m_type == eTypeInvalid) 5735 m_type = eTypeSharedLibrary; 5736 5737 return eStrataKernel; 5738 } 5739 } 5740 return eStrataUnknown; 5741 5742 case MH_EXECUTE: // 0x2u 5743 // Check for the MH_DYLDLINK bit in the flags 5744 if (m_header.flags & MH_DYLDLINK) { 5745 return eStrataUser; 5746 } else { 5747 SectionList *section_list = GetSectionList(); 5748 if (section_list) { 5749 static ConstString g_kld_section_name("__KLD"); 5750 if (section_list->FindSectionByName(g_kld_section_name)) 5751 return eStrataKernel; 5752 } 5753 } 5754 return eStrataRawImage; 5755 5756 case MH_FVMLIB: 5757 return eStrataUser; // 0x3u 5758 case MH_CORE: 5759 return eStrataUnknown; // 0x4u 5760 case MH_PRELOAD: 5761 return eStrataRawImage; // 0x5u 5762 case MH_DYLIB: 5763 return eStrataUser; // 0x6u 5764 case MH_DYLINKER: 5765 return eStrataUser; // 0x7u 5766 case MH_BUNDLE: 5767 return eStrataUser; // 0x8u 5768 case MH_DYLIB_STUB: 5769 return eStrataUser; // 0x9u 5770 case MH_DSYM: 5771 return eStrataUnknown; // 0xAu 5772 case MH_KEXT_BUNDLE: 5773 return eStrataKernel; // 0xBu 5774 default: 5775 break; 5776 } 5777 return eStrataUnknown; 5778 } 5779 5780 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5781 ModuleSP module_sp(GetModule()); 5782 if (module_sp) { 5783 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5784 llvm::MachO::dylib_command load_cmd; 5785 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5786 uint32_t version_cmd = 0; 5787 uint64_t version = 0; 5788 uint32_t i; 5789 for (i = 0; i < m_header.ncmds; ++i) { 5790 const lldb::offset_t cmd_offset = offset; 5791 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5792 break; 5793 5794 if (load_cmd.cmd == LC_ID_DYLIB) { 5795 if (version_cmd == 0) { 5796 version_cmd = load_cmd.cmd; 5797 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr) 5798 break; 5799 version = load_cmd.dylib.current_version; 5800 } 5801 break; // Break for now unless there is another more complete version 5802 // number load command in the future. 5803 } 5804 offset = cmd_offset + load_cmd.cmdsize; 5805 } 5806 5807 if (version_cmd == LC_ID_DYLIB) { 5808 unsigned major = (version & 0xFFFF0000ull) >> 16; 5809 unsigned minor = (version & 0x0000FF00ull) >> 8; 5810 unsigned subminor = (version & 0x000000FFull); 5811 return llvm::VersionTuple(major, minor, subminor); 5812 } 5813 } 5814 return llvm::VersionTuple(); 5815 } 5816 5817 ArchSpec ObjectFileMachO::GetArchitecture() { 5818 ModuleSP module_sp(GetModule()); 5819 ArchSpec arch; 5820 if (module_sp) { 5821 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5822 5823 return GetArchitecture(module_sp, m_header, m_data, 5824 MachHeaderSizeFromMagic(m_header.magic)); 5825 } 5826 return arch; 5827 } 5828 5829 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, 5830 addr_t &base_addr, UUID &uuid) { 5831 uuid.Clear(); 5832 base_addr = LLDB_INVALID_ADDRESS; 5833 if (process && process->GetDynamicLoader()) { 5834 DynamicLoader *dl = process->GetDynamicLoader(); 5835 LazyBool using_shared_cache; 5836 LazyBool private_shared_cache; 5837 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5838 private_shared_cache); 5839 } 5840 Log *log(GetLog(LLDBLog::Symbols | LLDBLog::Process)); 5841 LLDB_LOGF( 5842 log, 5843 "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64, 5844 uuid.GetAsString().c_str(), base_addr); 5845 } 5846 5847 // From dyld SPI header dyld_process_info.h 5848 typedef void *dyld_process_info; 5849 struct lldb_copy__dyld_process_cache_info { 5850 uuid_t cacheUUID; // UUID of cache used by process 5851 uint64_t cacheBaseAddress; // load address of dyld shared cache 5852 bool noCache; // process is running without a dyld cache 5853 bool privateCache; // process is using a private copy of its dyld cache 5854 }; 5855 5856 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with 5857 // llvm enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile 5858 // errors. So we need to use the actual underlying types of task_t and 5859 // kern_return_t below. 5860 extern "C" unsigned int /*task_t*/ mach_task_self(); 5861 5862 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5863 uuid.Clear(); 5864 base_addr = LLDB_INVALID_ADDRESS; 5865 5866 #if defined(__APPLE__) 5867 uint8_t *(*dyld_get_all_image_infos)(void); 5868 dyld_get_all_image_infos = 5869 (uint8_t * (*)()) dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5870 if (dyld_get_all_image_infos) { 5871 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5872 if (dyld_all_image_infos_address) { 5873 uint32_t *version = (uint32_t *) 5874 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5875 if (*version >= 13) { 5876 uuid_t *sharedCacheUUID_address = 0; 5877 int wordsize = sizeof(uint8_t *); 5878 if (wordsize == 8) { 5879 sharedCacheUUID_address = 5880 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5881 160); // sharedCacheUUID <mach-o/dyld_images.h> 5882 if (*version >= 15) 5883 base_addr = 5884 *(uint64_t 5885 *)((uint8_t *)dyld_all_image_infos_address + 5886 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5887 } else { 5888 sharedCacheUUID_address = 5889 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5890 84); // sharedCacheUUID <mach-o/dyld_images.h> 5891 if (*version >= 15) { 5892 base_addr = 0; 5893 base_addr = 5894 *(uint32_t 5895 *)((uint8_t *)dyld_all_image_infos_address + 5896 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5897 } 5898 } 5899 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5900 } 5901 } 5902 } else { 5903 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5904 dyld_process_info (*dyld_process_info_create)( 5905 unsigned int /* task_t */ task, uint64_t timestamp, 5906 unsigned int /*kern_return_t*/ *kernelError); 5907 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5908 void (*dyld_process_info_release)(dyld_process_info info); 5909 5910 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, 5911 unsigned int /*kern_return_t*/ *)) 5912 dlsym(RTLD_DEFAULT, "_dyld_process_info_create"); 5913 dyld_process_info_get_cache = (void (*)(void *, void *))dlsym( 5914 RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5915 dyld_process_info_release = 5916 (void (*)(void *))dlsym(RTLD_DEFAULT, "_dyld_process_info_release"); 5917 5918 if (dyld_process_info_create && dyld_process_info_get_cache) { 5919 unsigned int /*kern_return_t */ kern_ret; 5920 dyld_process_info process_info = 5921 dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5922 if (process_info) { 5923 struct lldb_copy__dyld_process_cache_info sc_info; 5924 memset(&sc_info, 0, sizeof(struct lldb_copy__dyld_process_cache_info)); 5925 dyld_process_info_get_cache(process_info, &sc_info); 5926 if (sc_info.cacheBaseAddress != 0) { 5927 base_addr = sc_info.cacheBaseAddress; 5928 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 5929 } 5930 dyld_process_info_release(process_info); 5931 } 5932 } 5933 } 5934 Log *log(GetLog(LLDBLog::Symbols | LLDBLog::Process)); 5935 if (log && uuid.IsValid()) 5936 LLDB_LOGF(log, 5937 "lldb's in-memory shared cache has a UUID of %s base address of " 5938 "0x%" PRIx64, 5939 uuid.GetAsString().c_str(), base_addr); 5940 #endif 5941 } 5942 5943 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5944 if (!m_min_os_version) { 5945 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5946 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5947 const lldb::offset_t load_cmd_offset = offset; 5948 5949 llvm::MachO::version_min_command lc; 5950 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5951 break; 5952 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5953 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5954 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5955 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5956 if (m_data.GetU32(&offset, &lc.version, 5957 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5958 const uint32_t xxxx = lc.version >> 16; 5959 const uint32_t yy = (lc.version >> 8) & 0xffu; 5960 const uint32_t zz = lc.version & 0xffu; 5961 if (xxxx) { 5962 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5963 break; 5964 } 5965 } 5966 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5967 // struct build_version_command { 5968 // uint32_t cmd; /* LC_BUILD_VERSION */ 5969 // uint32_t cmdsize; /* sizeof(struct 5970 // build_version_command) plus */ 5971 // /* ntools * sizeof(struct 5972 // build_tool_version) */ 5973 // uint32_t platform; /* platform */ 5974 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5975 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded in 5976 // nibbles xxxx.yy.zz */ uint32_t ntools; /* number of 5977 // tool entries following this */ 5978 // }; 5979 5980 offset += 4; // skip platform 5981 uint32_t minos = m_data.GetU32(&offset); 5982 5983 const uint32_t xxxx = minos >> 16; 5984 const uint32_t yy = (minos >> 8) & 0xffu; 5985 const uint32_t zz = minos & 0xffu; 5986 if (xxxx) { 5987 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5988 break; 5989 } 5990 } 5991 5992 offset = load_cmd_offset + lc.cmdsize; 5993 } 5994 5995 if (!m_min_os_version) { 5996 // Set version to an empty value so we don't keep trying to 5997 m_min_os_version = llvm::VersionTuple(); 5998 } 5999 } 6000 6001 return *m_min_os_version; 6002 } 6003 6004 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { 6005 if (!m_sdk_versions.hasValue()) { 6006 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 6007 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6008 const lldb::offset_t load_cmd_offset = offset; 6009 6010 llvm::MachO::version_min_command lc; 6011 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6012 break; 6013 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 6014 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 6015 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 6016 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 6017 if (m_data.GetU32(&offset, &lc.version, 6018 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 6019 const uint32_t xxxx = lc.sdk >> 16; 6020 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 6021 const uint32_t zz = lc.sdk & 0xffu; 6022 if (xxxx) { 6023 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 6024 break; 6025 } else { 6026 GetModule()->ReportWarning("minimum OS version load command with " 6027 "invalid (0) version found."); 6028 } 6029 } 6030 } 6031 offset = load_cmd_offset + lc.cmdsize; 6032 } 6033 6034 if (!m_sdk_versions.hasValue()) { 6035 offset = MachHeaderSizeFromMagic(m_header.magic); 6036 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6037 const lldb::offset_t load_cmd_offset = offset; 6038 6039 llvm::MachO::version_min_command lc; 6040 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6041 break; 6042 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 6043 // struct build_version_command { 6044 // uint32_t cmd; /* LC_BUILD_VERSION */ 6045 // uint32_t cmdsize; /* sizeof(struct 6046 // build_version_command) plus */ 6047 // /* ntools * sizeof(struct 6048 // build_tool_version) */ 6049 // uint32_t platform; /* platform */ 6050 // uint32_t minos; /* X.Y.Z is encoded in nibbles 6051 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded 6052 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number 6053 // of tool entries following this */ 6054 // }; 6055 6056 offset += 4; // skip platform 6057 uint32_t minos = m_data.GetU32(&offset); 6058 6059 const uint32_t xxxx = minos >> 16; 6060 const uint32_t yy = (minos >> 8) & 0xffu; 6061 const uint32_t zz = minos & 0xffu; 6062 if (xxxx) { 6063 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 6064 break; 6065 } 6066 } 6067 offset = load_cmd_offset + lc.cmdsize; 6068 } 6069 } 6070 6071 if (!m_sdk_versions.hasValue()) 6072 m_sdk_versions = llvm::VersionTuple(); 6073 } 6074 6075 return m_sdk_versions.getValue(); 6076 } 6077 6078 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 6079 return m_header.filetype == llvm::MachO::MH_DYLINKER; 6080 } 6081 6082 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 6083 return m_allow_assembly_emulation_unwind_plans; 6084 } 6085 6086 Section *ObjectFileMachO::GetMachHeaderSection() { 6087 // Find the first address of the mach header which is the first non-zero file 6088 // sized section whose file offset is zero. This is the base file address of 6089 // the mach-o file which can be subtracted from the vmaddr of the other 6090 // segments found in memory and added to the load address 6091 ModuleSP module_sp = GetModule(); 6092 if (!module_sp) 6093 return nullptr; 6094 SectionList *section_list = GetSectionList(); 6095 if (!section_list) 6096 return nullptr; 6097 const size_t num_sections = section_list->GetSize(); 6098 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6099 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 6100 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 6101 return section; 6102 } 6103 return nullptr; 6104 } 6105 6106 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 6107 if (!section) 6108 return false; 6109 const bool is_dsym = (m_header.filetype == MH_DSYM); 6110 if (section->GetFileSize() == 0 && !is_dsym) 6111 return false; 6112 if (section->IsThreadSpecific()) 6113 return false; 6114 if (GetModule().get() != section->GetModule().get()) 6115 return false; 6116 // Be careful with __LINKEDIT and __DWARF segments 6117 if (section->GetName() == GetSegmentNameLINKEDIT() || 6118 section->GetName() == GetSegmentNameDWARF()) { 6119 // Only map __LINKEDIT and __DWARF if we have an in memory image and 6120 // this isn't a kernel binary like a kext or mach_kernel. 6121 const bool is_memory_image = (bool)m_process_wp.lock(); 6122 const Strata strata = GetStrata(); 6123 if (is_memory_image == false || strata == eStrataKernel) 6124 return false; 6125 } 6126 return true; 6127 } 6128 6129 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 6130 lldb::addr_t header_load_address, const Section *header_section, 6131 const Section *section) { 6132 ModuleSP module_sp = GetModule(); 6133 if (module_sp && header_section && section && 6134 header_load_address != LLDB_INVALID_ADDRESS) { 6135 lldb::addr_t file_addr = header_section->GetFileAddress(); 6136 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 6137 return section->GetFileAddress() - file_addr + header_load_address; 6138 } 6139 return LLDB_INVALID_ADDRESS; 6140 } 6141 6142 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 6143 bool value_is_offset) { 6144 ModuleSP module_sp = GetModule(); 6145 if (!module_sp) 6146 return false; 6147 6148 SectionList *section_list = GetSectionList(); 6149 if (!section_list) 6150 return false; 6151 6152 size_t num_loaded_sections = 0; 6153 const size_t num_sections = section_list->GetSize(); 6154 6155 if (value_is_offset) { 6156 // "value" is an offset to apply to each top level segment 6157 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6158 // Iterate through the object file sections to find all of the 6159 // sections that size on disk (to avoid __PAGEZERO) and load them 6160 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6161 if (SectionIsLoadable(section_sp.get())) 6162 if (target.GetSectionLoadList().SetSectionLoadAddress( 6163 section_sp, section_sp->GetFileAddress() + value)) 6164 ++num_loaded_sections; 6165 } 6166 } else { 6167 // "value" is the new base address of the mach_header, adjust each 6168 // section accordingly 6169 6170 Section *mach_header_section = GetMachHeaderSection(); 6171 if (mach_header_section) { 6172 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6173 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6174 6175 lldb::addr_t section_load_addr = 6176 CalculateSectionLoadAddressForMemoryImage( 6177 value, mach_header_section, section_sp.get()); 6178 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6179 if (target.GetSectionLoadList().SetSectionLoadAddress( 6180 section_sp, section_load_addr)) 6181 ++num_loaded_sections; 6182 } 6183 } 6184 } 6185 } 6186 return num_loaded_sections > 0; 6187 } 6188 6189 struct all_image_infos_header { 6190 uint32_t version; // currently 1 6191 uint32_t imgcount; // number of binary images 6192 uint64_t entries_fileoff; // file offset in the corefile of where the array of 6193 // struct entry's begin. 6194 uint32_t entries_size; // size of 'struct entry'. 6195 uint32_t unused; 6196 }; 6197 6198 struct image_entry { 6199 uint64_t filepath_offset; // offset in corefile to c-string of the file path, 6200 // UINT64_MAX if unavailable. 6201 uuid_t uuid; // uint8_t[16]. should be set to all zeroes if 6202 // uuid is unknown. 6203 uint64_t load_address; // UINT64_MAX if unknown. 6204 uint64_t seg_addrs_offset; // offset to the array of struct segment_vmaddr's. 6205 uint32_t segment_count; // The number of segments for this binary. 6206 uint32_t unused; 6207 6208 image_entry() { 6209 filepath_offset = UINT64_MAX; 6210 memset(&uuid, 0, sizeof(uuid_t)); 6211 segment_count = 0; 6212 load_address = UINT64_MAX; 6213 seg_addrs_offset = UINT64_MAX; 6214 unused = 0; 6215 } 6216 image_entry(const image_entry &rhs) { 6217 filepath_offset = rhs.filepath_offset; 6218 memcpy(&uuid, &rhs.uuid, sizeof(uuid_t)); 6219 segment_count = rhs.segment_count; 6220 seg_addrs_offset = rhs.seg_addrs_offset; 6221 load_address = rhs.load_address; 6222 unused = rhs.unused; 6223 } 6224 }; 6225 6226 struct segment_vmaddr { 6227 char segname[16]; 6228 uint64_t vmaddr; 6229 uint64_t unused; 6230 6231 segment_vmaddr() { 6232 memset(&segname, 0, 16); 6233 vmaddr = UINT64_MAX; 6234 unused = 0; 6235 } 6236 segment_vmaddr(const segment_vmaddr &rhs) { 6237 memcpy(&segname, &rhs.segname, 16); 6238 vmaddr = rhs.vmaddr; 6239 unused = rhs.unused; 6240 } 6241 }; 6242 6243 // Write the payload for the "all image infos" LC_NOTE into 6244 // the supplied all_image_infos_payload, assuming that this 6245 // will be written into the corefile starting at 6246 // initial_file_offset. 6247 // 6248 // The placement of this payload is a little tricky. We're 6249 // laying this out as 6250 // 6251 // 1. header (struct all_image_info_header) 6252 // 2. Array of fixed-size (struct image_entry)'s, one 6253 // per binary image present in the process. 6254 // 3. Arrays of (struct segment_vmaddr)'s, a varying number 6255 // for each binary image. 6256 // 4. Variable length c-strings of binary image filepaths, 6257 // one per binary. 6258 // 6259 // To compute where everything will be laid out in the 6260 // payload, we need to iterate over the images and calculate 6261 // how many segment_vmaddr structures each image will need, 6262 // and how long each image's filepath c-string is. There 6263 // are some multiple passes over the image list while calculating 6264 // everything. 6265 6266 static offset_t CreateAllImageInfosPayload( 6267 const lldb::ProcessSP &process_sp, offset_t initial_file_offset, 6268 StreamString &all_image_infos_payload, SaveCoreStyle core_style) { 6269 Target &target = process_sp->GetTarget(); 6270 ModuleList modules = target.GetImages(); 6271 6272 // stack-only corefiles have no reason to include binaries that 6273 // are not executing; we're trying to make the smallest corefile 6274 // we can, so leave the rest out. 6275 if (core_style == SaveCoreStyle::eSaveCoreStackOnly) 6276 modules.Clear(); 6277 6278 std::set<std::string> executing_uuids; 6279 ThreadList &thread_list(process_sp->GetThreadList()); 6280 for (uint32_t i = 0; i < thread_list.GetSize(); i++) { 6281 ThreadSP thread_sp = thread_list.GetThreadAtIndex(i); 6282 uint32_t stack_frame_count = thread_sp->GetStackFrameCount(); 6283 for (uint32_t j = 0; j < stack_frame_count; j++) { 6284 StackFrameSP stack_frame_sp = thread_sp->GetStackFrameAtIndex(j); 6285 Address pc = stack_frame_sp->GetFrameCodeAddress(); 6286 ModuleSP module_sp = pc.GetModule(); 6287 if (module_sp) { 6288 UUID uuid = module_sp->GetUUID(); 6289 if (uuid.IsValid()) { 6290 executing_uuids.insert(uuid.GetAsString()); 6291 modules.AppendIfNeeded(module_sp); 6292 } 6293 } 6294 } 6295 } 6296 size_t modules_count = modules.GetSize(); 6297 6298 struct all_image_infos_header infos; 6299 infos.version = 1; 6300 infos.imgcount = modules_count; 6301 infos.entries_size = sizeof(image_entry); 6302 infos.entries_fileoff = initial_file_offset + sizeof(all_image_infos_header); 6303 infos.unused = 0; 6304 6305 all_image_infos_payload.PutHex32(infos.version); 6306 all_image_infos_payload.PutHex32(infos.imgcount); 6307 all_image_infos_payload.PutHex64(infos.entries_fileoff); 6308 all_image_infos_payload.PutHex32(infos.entries_size); 6309 all_image_infos_payload.PutHex32(infos.unused); 6310 6311 // First create the structures for all of the segment name+vmaddr vectors 6312 // for each module, so we will know the size of them as we add the 6313 // module entries. 6314 std::vector<std::vector<segment_vmaddr>> modules_segment_vmaddrs; 6315 for (size_t i = 0; i < modules_count; i++) { 6316 ModuleSP module = modules.GetModuleAtIndex(i); 6317 6318 SectionList *sections = module->GetSectionList(); 6319 size_t sections_count = sections->GetSize(); 6320 std::vector<segment_vmaddr> segment_vmaddrs; 6321 for (size_t j = 0; j < sections_count; j++) { 6322 SectionSP section = sections->GetSectionAtIndex(j); 6323 if (!section->GetParent().get()) { 6324 addr_t vmaddr = section->GetLoadBaseAddress(&target); 6325 if (vmaddr == LLDB_INVALID_ADDRESS) 6326 continue; 6327 ConstString name = section->GetName(); 6328 segment_vmaddr seg_vmaddr; 6329 strncpy(seg_vmaddr.segname, name.AsCString(), 6330 sizeof(seg_vmaddr.segname)); 6331 seg_vmaddr.vmaddr = vmaddr; 6332 seg_vmaddr.unused = 0; 6333 segment_vmaddrs.push_back(seg_vmaddr); 6334 } 6335 } 6336 modules_segment_vmaddrs.push_back(segment_vmaddrs); 6337 } 6338 6339 offset_t size_of_vmaddr_structs = 0; 6340 for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) { 6341 size_of_vmaddr_structs += 6342 modules_segment_vmaddrs[i].size() * sizeof(segment_vmaddr); 6343 } 6344 6345 offset_t size_of_filepath_cstrings = 0; 6346 for (size_t i = 0; i < modules_count; i++) { 6347 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6348 size_of_filepath_cstrings += module_sp->GetFileSpec().GetPath().size() + 1; 6349 } 6350 6351 // Calculate the file offsets of our "all image infos" payload in the 6352 // corefile. initial_file_offset the original value passed in to this method. 6353 6354 offset_t start_of_entries = 6355 initial_file_offset + sizeof(all_image_infos_header); 6356 offset_t start_of_seg_vmaddrs = 6357 start_of_entries + sizeof(image_entry) * modules_count; 6358 offset_t start_of_filenames = start_of_seg_vmaddrs + size_of_vmaddr_structs; 6359 6360 offset_t final_file_offset = start_of_filenames + size_of_filepath_cstrings; 6361 6362 // Now write the one-per-module 'struct image_entry' into the 6363 // StringStream; keep track of where the struct segment_vmaddr 6364 // entries for each module will end up in the corefile. 6365 6366 offset_t current_string_offset = start_of_filenames; 6367 offset_t current_segaddrs_offset = start_of_seg_vmaddrs; 6368 std::vector<struct image_entry> image_entries; 6369 for (size_t i = 0; i < modules_count; i++) { 6370 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6371 6372 struct image_entry ent; 6373 memcpy(&ent.uuid, module_sp->GetUUID().GetBytes().data(), sizeof(ent.uuid)); 6374 if (modules_segment_vmaddrs[i].size() > 0) { 6375 ent.segment_count = modules_segment_vmaddrs[i].size(); 6376 ent.seg_addrs_offset = current_segaddrs_offset; 6377 } 6378 ent.filepath_offset = current_string_offset; 6379 ObjectFile *objfile = module_sp->GetObjectFile(); 6380 if (objfile) { 6381 Address base_addr(objfile->GetBaseAddress()); 6382 if (base_addr.IsValid()) { 6383 ent.load_address = base_addr.GetLoadAddress(&target); 6384 } 6385 } 6386 6387 all_image_infos_payload.PutHex64(ent.filepath_offset); 6388 all_image_infos_payload.PutRawBytes(ent.uuid, sizeof(ent.uuid)); 6389 all_image_infos_payload.PutHex64(ent.load_address); 6390 all_image_infos_payload.PutHex64(ent.seg_addrs_offset); 6391 all_image_infos_payload.PutHex32(ent.segment_count); 6392 6393 if (executing_uuids.find(module_sp->GetUUID().GetAsString()) != 6394 executing_uuids.end()) 6395 all_image_infos_payload.PutHex32(1); 6396 else 6397 all_image_infos_payload.PutHex32(0); 6398 6399 current_segaddrs_offset += ent.segment_count * sizeof(segment_vmaddr); 6400 current_string_offset += module_sp->GetFileSpec().GetPath().size() + 1; 6401 } 6402 6403 // Now write the struct segment_vmaddr entries into the StringStream. 6404 6405 for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) { 6406 if (modules_segment_vmaddrs[i].size() == 0) 6407 continue; 6408 for (struct segment_vmaddr segvm : modules_segment_vmaddrs[i]) { 6409 all_image_infos_payload.PutRawBytes(segvm.segname, sizeof(segvm.segname)); 6410 all_image_infos_payload.PutHex64(segvm.vmaddr); 6411 all_image_infos_payload.PutHex64(segvm.unused); 6412 } 6413 } 6414 6415 for (size_t i = 0; i < modules_count; i++) { 6416 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6417 std::string filepath = module_sp->GetFileSpec().GetPath(); 6418 all_image_infos_payload.PutRawBytes(filepath.data(), filepath.size() + 1); 6419 } 6420 6421 return final_file_offset; 6422 } 6423 6424 // Temp struct used to combine contiguous memory regions with 6425 // identical permissions. 6426 struct page_object { 6427 addr_t addr; 6428 addr_t size; 6429 uint32_t prot; 6430 }; 6431 6432 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6433 const FileSpec &outfile, 6434 lldb::SaveCoreStyle &core_style, Status &error) { 6435 if (!process_sp) 6436 return false; 6437 6438 // Default on macOS is to create a dirty-memory-only corefile. 6439 if (core_style == SaveCoreStyle::eSaveCoreUnspecified) { 6440 core_style = SaveCoreStyle::eSaveCoreDirtyOnly; 6441 } 6442 6443 Target &target = process_sp->GetTarget(); 6444 const ArchSpec target_arch = target.GetArchitecture(); 6445 const llvm::Triple &target_triple = target_arch.GetTriple(); 6446 if (target_triple.getVendor() == llvm::Triple::Apple && 6447 (target_triple.getOS() == llvm::Triple::MacOSX || 6448 target_triple.getOS() == llvm::Triple::IOS || 6449 target_triple.getOS() == llvm::Triple::WatchOS || 6450 target_triple.getOS() == llvm::Triple::TvOS)) { 6451 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) 6452 // { 6453 bool make_core = false; 6454 switch (target_arch.GetMachine()) { 6455 case llvm::Triple::aarch64: 6456 case llvm::Triple::aarch64_32: 6457 case llvm::Triple::arm: 6458 case llvm::Triple::thumb: 6459 case llvm::Triple::x86: 6460 case llvm::Triple::x86_64: 6461 make_core = true; 6462 break; 6463 default: 6464 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6465 target_triple.str().c_str()); 6466 break; 6467 } 6468 6469 if (make_core) { 6470 std::vector<llvm::MachO::segment_command_64> segment_load_commands; 6471 // uint32_t range_info_idx = 0; 6472 MemoryRegionInfo range_info; 6473 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6474 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6475 const ByteOrder byte_order = target_arch.GetByteOrder(); 6476 std::vector<page_object> pages_to_copy; 6477 6478 if (range_error.Success()) { 6479 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6480 // Calculate correct protections 6481 uint32_t prot = 0; 6482 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6483 prot |= VM_PROT_READ; 6484 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6485 prot |= VM_PROT_WRITE; 6486 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6487 prot |= VM_PROT_EXECUTE; 6488 6489 const addr_t addr = range_info.GetRange().GetRangeBase(); 6490 const addr_t size = range_info.GetRange().GetByteSize(); 6491 6492 if (size == 0) 6493 break; 6494 6495 bool include_this_region = true; 6496 bool dirty_pages_only = false; 6497 if (core_style == SaveCoreStyle::eSaveCoreStackOnly) { 6498 dirty_pages_only = true; 6499 if (range_info.IsStackMemory() != MemoryRegionInfo::eYes) { 6500 include_this_region = false; 6501 } 6502 } 6503 if (core_style == SaveCoreStyle::eSaveCoreDirtyOnly) { 6504 dirty_pages_only = true; 6505 } 6506 6507 if (prot != 0 && include_this_region) { 6508 addr_t pagesize = range_info.GetPageSize(); 6509 const llvm::Optional<std::vector<addr_t>> &dirty_page_list = 6510 range_info.GetDirtyPageList(); 6511 if (dirty_pages_only && dirty_page_list.hasValue()) { 6512 for (addr_t dirtypage : dirty_page_list.getValue()) { 6513 page_object obj; 6514 obj.addr = dirtypage; 6515 obj.size = pagesize; 6516 obj.prot = prot; 6517 pages_to_copy.push_back(obj); 6518 } 6519 } else { 6520 page_object obj; 6521 obj.addr = addr; 6522 obj.size = size; 6523 obj.prot = prot; 6524 pages_to_copy.push_back(obj); 6525 } 6526 } 6527 6528 range_error = process_sp->GetMemoryRegionInfo( 6529 range_info.GetRange().GetRangeEnd(), range_info); 6530 if (range_error.Fail()) 6531 break; 6532 } 6533 6534 // Combine contiguous entries that have the same 6535 // protections so we don't have an excess of 6536 // load commands. 6537 std::vector<page_object> combined_page_objects; 6538 page_object last_obj; 6539 last_obj.addr = LLDB_INVALID_ADDRESS; 6540 last_obj.size = 0; 6541 for (page_object obj : pages_to_copy) { 6542 if (last_obj.addr == LLDB_INVALID_ADDRESS) { 6543 last_obj = obj; 6544 continue; 6545 } 6546 if (last_obj.addr + last_obj.size == obj.addr && 6547 last_obj.prot == obj.prot) { 6548 last_obj.size += obj.size; 6549 continue; 6550 } 6551 combined_page_objects.push_back(last_obj); 6552 last_obj = obj; 6553 } 6554 // Add the last entry we were looking to combine 6555 // on to the array. 6556 if (last_obj.addr != LLDB_INVALID_ADDRESS && last_obj.size != 0) 6557 combined_page_objects.push_back(last_obj); 6558 6559 for (page_object obj : combined_page_objects) { 6560 uint32_t cmd_type = LC_SEGMENT_64; 6561 uint32_t segment_size = sizeof(llvm::MachO::segment_command_64); 6562 if (addr_byte_size == 4) { 6563 cmd_type = LC_SEGMENT; 6564 segment_size = sizeof(llvm::MachO::segment_command); 6565 } 6566 llvm::MachO::segment_command_64 segment = { 6567 cmd_type, // uint32_t cmd; 6568 segment_size, // uint32_t cmdsize; 6569 {0}, // char segname[16]; 6570 obj.addr, // uint64_t vmaddr; // uint32_t for 32-bit 6571 // Mach-O 6572 obj.size, // uint64_t vmsize; // uint32_t for 32-bit 6573 // Mach-O 6574 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6575 obj.size, // uint64_t filesize; // uint32_t for 32-bit 6576 // Mach-O 6577 obj.prot, // uint32_t maxprot; 6578 obj.prot, // uint32_t initprot; 6579 0, // uint32_t nsects; 6580 0}; // uint32_t flags; 6581 segment_load_commands.push_back(segment); 6582 } 6583 6584 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6585 6586 llvm::MachO::mach_header_64 mach_header; 6587 if (addr_byte_size == 8) { 6588 mach_header.magic = MH_MAGIC_64; 6589 } else { 6590 mach_header.magic = MH_MAGIC; 6591 } 6592 mach_header.cputype = target_arch.GetMachOCPUType(); 6593 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6594 mach_header.filetype = MH_CORE; 6595 mach_header.ncmds = segment_load_commands.size(); 6596 mach_header.flags = 0; 6597 mach_header.reserved = 0; 6598 ThreadList &thread_list = process_sp->GetThreadList(); 6599 const uint32_t num_threads = thread_list.GetSize(); 6600 6601 // Make an array of LC_THREAD data items. Each one contains the 6602 // contents of the LC_THREAD load command. The data doesn't contain 6603 // the load command + load command size, we will add the load command 6604 // and load command size as we emit the data. 6605 std::vector<StreamString> LC_THREAD_datas(num_threads); 6606 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6607 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6608 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6609 LC_THREAD_data.SetByteOrder(byte_order); 6610 } 6611 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) { 6612 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6613 if (thread_sp) { 6614 switch (mach_header.cputype) { 6615 case llvm::MachO::CPU_TYPE_ARM64: 6616 case llvm::MachO::CPU_TYPE_ARM64_32: 6617 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6618 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6619 break; 6620 6621 case llvm::MachO::CPU_TYPE_ARM: 6622 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6623 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6624 break; 6625 6626 case llvm::MachO::CPU_TYPE_I386: 6627 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6628 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6629 break; 6630 6631 case llvm::MachO::CPU_TYPE_X86_64: 6632 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6633 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6634 break; 6635 } 6636 } 6637 } 6638 6639 // The size of the load command is the size of the segments... 6640 if (addr_byte_size == 8) { 6641 mach_header.sizeofcmds = segment_load_commands.size() * 6642 sizeof(llvm::MachO::segment_command_64); 6643 } else { 6644 mach_header.sizeofcmds = segment_load_commands.size() * 6645 sizeof(llvm::MachO::segment_command); 6646 } 6647 6648 // and the size of all LC_THREAD load command 6649 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6650 ++mach_header.ncmds; 6651 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6652 } 6653 6654 // Bits will be set to indicate which bits are NOT used in 6655 // addressing in this process or 0 for unknown. 6656 uint64_t address_mask = process_sp->GetCodeAddressMask(); 6657 if (address_mask != 0) { 6658 // LC_NOTE "addrable bits" 6659 mach_header.ncmds++; 6660 mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); 6661 } 6662 6663 // LC_NOTE "all image infos" 6664 mach_header.ncmds++; 6665 mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); 6666 6667 // Write the mach header 6668 buffer.PutHex32(mach_header.magic); 6669 buffer.PutHex32(mach_header.cputype); 6670 buffer.PutHex32(mach_header.cpusubtype); 6671 buffer.PutHex32(mach_header.filetype); 6672 buffer.PutHex32(mach_header.ncmds); 6673 buffer.PutHex32(mach_header.sizeofcmds); 6674 buffer.PutHex32(mach_header.flags); 6675 if (addr_byte_size == 8) { 6676 buffer.PutHex32(mach_header.reserved); 6677 } 6678 6679 // Skip the mach header and all load commands and align to the next 6680 // 0x1000 byte boundary 6681 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6682 6683 file_offset = llvm::alignTo(file_offset, 16); 6684 std::vector<std::unique_ptr<LCNoteEntry>> lc_notes; 6685 6686 // Add "addrable bits" LC_NOTE when an address mask is available 6687 if (address_mask != 0) { 6688 std::unique_ptr<LCNoteEntry> addrable_bits_lcnote_up( 6689 new LCNoteEntry(addr_byte_size, byte_order)); 6690 addrable_bits_lcnote_up->name = "addrable bits"; 6691 addrable_bits_lcnote_up->payload_file_offset = file_offset; 6692 int bits = std::bitset<64>(~address_mask).count(); 6693 addrable_bits_lcnote_up->payload.PutHex32(3); // version 6694 addrable_bits_lcnote_up->payload.PutHex32( 6695 bits); // # of bits used for addressing 6696 addrable_bits_lcnote_up->payload.PutHex64(0); // unused 6697 6698 file_offset += addrable_bits_lcnote_up->payload.GetSize(); 6699 6700 lc_notes.push_back(std::move(addrable_bits_lcnote_up)); 6701 } 6702 6703 // Add "all image infos" LC_NOTE 6704 std::unique_ptr<LCNoteEntry> all_image_infos_lcnote_up( 6705 new LCNoteEntry(addr_byte_size, byte_order)); 6706 all_image_infos_lcnote_up->name = "all image infos"; 6707 all_image_infos_lcnote_up->payload_file_offset = file_offset; 6708 file_offset = CreateAllImageInfosPayload( 6709 process_sp, file_offset, all_image_infos_lcnote_up->payload, 6710 core_style); 6711 lc_notes.push_back(std::move(all_image_infos_lcnote_up)); 6712 6713 // Add LC_NOTE load commands 6714 for (auto &lcnote : lc_notes) { 6715 // Add the LC_NOTE load command to the file. 6716 buffer.PutHex32(LC_NOTE); 6717 buffer.PutHex32(sizeof(llvm::MachO::note_command)); 6718 char namebuf[16]; 6719 memset(namebuf, 0, sizeof(namebuf)); 6720 // this is the uncommon case where strncpy is exactly 6721 // the right one, doesn't need to be nul terminated. 6722 strncpy(namebuf, lcnote->name.c_str(), sizeof(namebuf)); 6723 buffer.PutRawBytes(namebuf, sizeof(namebuf)); 6724 buffer.PutHex64(lcnote->payload_file_offset); 6725 buffer.PutHex64(lcnote->payload.GetSize()); 6726 } 6727 6728 // Align to 4096-byte page boundary for the LC_SEGMENTs. 6729 file_offset = llvm::alignTo(file_offset, 4096); 6730 6731 for (auto &segment : segment_load_commands) { 6732 segment.fileoff = file_offset; 6733 file_offset += segment.filesize; 6734 } 6735 6736 // Write out all of the LC_THREAD load commands 6737 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6738 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6739 buffer.PutHex32(LC_THREAD); 6740 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6741 buffer.Write(LC_THREAD_data.GetString().data(), LC_THREAD_data_size); 6742 } 6743 6744 // Write out all of the segment load commands 6745 for (const auto &segment : segment_load_commands) { 6746 buffer.PutHex32(segment.cmd); 6747 buffer.PutHex32(segment.cmdsize); 6748 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6749 if (addr_byte_size == 8) { 6750 buffer.PutHex64(segment.vmaddr); 6751 buffer.PutHex64(segment.vmsize); 6752 buffer.PutHex64(segment.fileoff); 6753 buffer.PutHex64(segment.filesize); 6754 } else { 6755 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6756 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6757 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6758 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6759 } 6760 buffer.PutHex32(segment.maxprot); 6761 buffer.PutHex32(segment.initprot); 6762 buffer.PutHex32(segment.nsects); 6763 buffer.PutHex32(segment.flags); 6764 } 6765 6766 std::string core_file_path(outfile.GetPath()); 6767 auto core_file = FileSystem::Instance().Open( 6768 outfile, File::eOpenOptionWriteOnly | File::eOpenOptionTruncate | 6769 File::eOpenOptionCanCreate); 6770 if (!core_file) { 6771 error = core_file.takeError(); 6772 } else { 6773 // Read 1 page at a time 6774 uint8_t bytes[0x1000]; 6775 // Write the mach header and load commands out to the core file 6776 size_t bytes_written = buffer.GetString().size(); 6777 error = 6778 core_file.get()->Write(buffer.GetString().data(), bytes_written); 6779 if (error.Success()) { 6780 6781 for (auto &lcnote : lc_notes) { 6782 if (core_file.get()->SeekFromStart(lcnote->payload_file_offset) == 6783 -1) { 6784 error.SetErrorStringWithFormat("Unable to seek to corefile pos " 6785 "to write '%s' LC_NOTE payload", 6786 lcnote->name.c_str()); 6787 return false; 6788 } 6789 bytes_written = lcnote->payload.GetSize(); 6790 error = core_file.get()->Write(lcnote->payload.GetData(), 6791 bytes_written); 6792 if (!error.Success()) 6793 return false; 6794 } 6795 6796 // Now write the file data for all memory segments in the process 6797 for (const auto &segment : segment_load_commands) { 6798 if (core_file.get()->SeekFromStart(segment.fileoff) == -1) { 6799 error.SetErrorStringWithFormat( 6800 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6801 segment.fileoff, core_file_path.c_str()); 6802 break; 6803 } 6804 6805 target.GetDebugger().GetAsyncOutputStream()->Printf( 6806 "Saving %" PRId64 6807 " bytes of data for memory region at 0x%" PRIx64 "\n", 6808 segment.vmsize, segment.vmaddr); 6809 addr_t bytes_left = segment.vmsize; 6810 addr_t addr = segment.vmaddr; 6811 Status memory_read_error; 6812 while (bytes_left > 0 && error.Success()) { 6813 const size_t bytes_to_read = 6814 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6815 6816 // In a savecore setting, we don't really care about caching, 6817 // as the data is dumped and very likely never read again, 6818 // so we call ReadMemoryFromInferior to bypass it. 6819 const size_t bytes_read = process_sp->ReadMemoryFromInferior( 6820 addr, bytes, bytes_to_read, memory_read_error); 6821 6822 if (bytes_read == bytes_to_read) { 6823 size_t bytes_written = bytes_read; 6824 error = core_file.get()->Write(bytes, bytes_written); 6825 bytes_left -= bytes_read; 6826 addr += bytes_read; 6827 } else { 6828 // Some pages within regions are not readable, those should 6829 // be zero filled 6830 memset(bytes, 0, bytes_to_read); 6831 size_t bytes_written = bytes_to_read; 6832 error = core_file.get()->Write(bytes, bytes_written); 6833 bytes_left -= bytes_to_read; 6834 addr += bytes_to_read; 6835 } 6836 } 6837 } 6838 } 6839 } 6840 } else { 6841 error.SetErrorString( 6842 "process doesn't support getting memory region info"); 6843 } 6844 } 6845 return true; // This is the right plug to handle saving core files for 6846 // this process 6847 } 6848 return false; 6849 } 6850 6851 ObjectFileMachO::MachOCorefileAllImageInfos 6852 ObjectFileMachO::GetCorefileAllImageInfos() { 6853 MachOCorefileAllImageInfos image_infos; 6854 6855 // Look for an "all image infos" LC_NOTE. 6856 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 6857 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6858 const uint32_t cmd_offset = offset; 6859 llvm::MachO::load_command lc; 6860 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6861 break; 6862 if (lc.cmd == LC_NOTE) { 6863 char data_owner[17]; 6864 m_data.CopyData(offset, 16, data_owner); 6865 data_owner[16] = '\0'; 6866 offset += 16; 6867 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 6868 offset += 4; /* size unused */ 6869 6870 if (strcmp("all image infos", data_owner) == 0) { 6871 offset = fileoff; 6872 // Read the struct all_image_infos_header. 6873 uint32_t version = m_data.GetU32(&offset); 6874 if (version != 1) { 6875 return image_infos; 6876 } 6877 uint32_t imgcount = m_data.GetU32(&offset); 6878 uint64_t entries_fileoff = m_data.GetU64(&offset); 6879 offset += 4; // uint32_t entries_size; 6880 offset += 4; // uint32_t unused; 6881 6882 offset = entries_fileoff; 6883 for (uint32_t i = 0; i < imgcount; i++) { 6884 // Read the struct image_entry. 6885 offset_t filepath_offset = m_data.GetU64(&offset); 6886 uuid_t uuid; 6887 memcpy(&uuid, m_data.GetData(&offset, sizeof(uuid_t)), 6888 sizeof(uuid_t)); 6889 uint64_t load_address = m_data.GetU64(&offset); 6890 offset_t seg_addrs_offset = m_data.GetU64(&offset); 6891 uint32_t segment_count = m_data.GetU32(&offset); 6892 uint32_t currently_executing = m_data.GetU32(&offset); 6893 6894 MachOCorefileImageEntry image_entry; 6895 image_entry.filename = (const char *)m_data.GetCStr(&filepath_offset); 6896 image_entry.uuid = UUID::fromData(uuid, sizeof(uuid_t)); 6897 image_entry.load_address = load_address; 6898 image_entry.currently_executing = currently_executing; 6899 6900 offset_t seg_vmaddrs_offset = seg_addrs_offset; 6901 for (uint32_t j = 0; j < segment_count; j++) { 6902 char segname[17]; 6903 m_data.CopyData(seg_vmaddrs_offset, 16, segname); 6904 segname[16] = '\0'; 6905 seg_vmaddrs_offset += 16; 6906 uint64_t vmaddr = m_data.GetU64(&seg_vmaddrs_offset); 6907 seg_vmaddrs_offset += 8; /* unused */ 6908 6909 std::tuple<ConstString, addr_t> new_seg{ConstString(segname), 6910 vmaddr}; 6911 image_entry.segment_load_addresses.push_back(new_seg); 6912 } 6913 image_infos.all_image_infos.push_back(image_entry); 6914 } 6915 } else if (strcmp("load binary", data_owner) == 0) { 6916 uint32_t version = m_data.GetU32(&fileoff); 6917 if (version == 1) { 6918 uuid_t uuid; 6919 memcpy(&uuid, m_data.GetData(&fileoff, sizeof(uuid_t)), 6920 sizeof(uuid_t)); 6921 uint64_t load_address = m_data.GetU64(&fileoff); 6922 uint64_t slide = m_data.GetU64(&fileoff); 6923 std::string filename = m_data.GetCStr(&fileoff); 6924 6925 MachOCorefileImageEntry image_entry; 6926 image_entry.filename = filename; 6927 image_entry.uuid = UUID::fromData(uuid, sizeof(uuid_t)); 6928 image_entry.load_address = load_address; 6929 image_entry.slide = slide; 6930 image_infos.all_image_infos.push_back(image_entry); 6931 } 6932 } 6933 } 6934 offset = cmd_offset + lc.cmdsize; 6935 } 6936 6937 return image_infos; 6938 } 6939 6940 bool ObjectFileMachO::LoadCoreFileImages(lldb_private::Process &process) { 6941 MachOCorefileAllImageInfos image_infos = GetCorefileAllImageInfos(); 6942 Log *log = GetLog(LLDBLog::DynamicLoader); 6943 6944 ModuleList added_modules; 6945 for (const MachOCorefileImageEntry &image : image_infos.all_image_infos) { 6946 ModuleSpec module_spec; 6947 module_spec.GetUUID() = image.uuid; 6948 if (image.filename.empty()) { 6949 char namebuf[80]; 6950 if (image.load_address != LLDB_INVALID_ADDRESS) 6951 snprintf(namebuf, sizeof(namebuf), "mem-image-0x%" PRIx64, 6952 image.load_address); 6953 else 6954 snprintf(namebuf, sizeof(namebuf), "mem-image+0x%" PRIx64, image.slide); 6955 module_spec.GetFileSpec() = FileSpec(namebuf); 6956 } else { 6957 module_spec.GetFileSpec() = FileSpec(image.filename.c_str()); 6958 } 6959 if (image.currently_executing) { 6960 Symbols::DownloadObjectAndSymbolFile(module_spec, true); 6961 if (FileSystem::Instance().Exists(module_spec.GetFileSpec())) { 6962 process.GetTarget().GetOrCreateModule(module_spec, false); 6963 } 6964 } 6965 Status error; 6966 ModuleSP module_sp = 6967 process.GetTarget().GetOrCreateModule(module_spec, false, &error); 6968 if (!module_sp.get() || !module_sp->GetObjectFile()) { 6969 if (image.load_address != LLDB_INVALID_ADDRESS) { 6970 module_sp = process.ReadModuleFromMemory(module_spec.GetFileSpec(), 6971 image.load_address); 6972 } 6973 } 6974 if (module_sp.get()) { 6975 // Will call ModulesDidLoad with all modules once they've all 6976 // been added to the Target with load addresses. Don't notify 6977 // here, before the load address is set. 6978 const bool notify = false; 6979 process.GetTarget().GetImages().AppendIfNeeded(module_sp, notify); 6980 added_modules.Append(module_sp, notify); 6981 if (image.segment_load_addresses.size() > 0) { 6982 if (log) { 6983 std::string uuidstr = image.uuid.GetAsString(); 6984 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 6985 "UUID %s with section load addresses", 6986 image.filename.c_str(), uuidstr.c_str()); 6987 } 6988 for (auto name_vmaddr_tuple : image.segment_load_addresses) { 6989 SectionList *sectlist = module_sp->GetObjectFile()->GetSectionList(); 6990 if (sectlist) { 6991 SectionSP sect_sp = 6992 sectlist->FindSectionByName(std::get<0>(name_vmaddr_tuple)); 6993 if (sect_sp) { 6994 process.GetTarget().SetSectionLoadAddress( 6995 sect_sp, std::get<1>(name_vmaddr_tuple)); 6996 } 6997 } 6998 } 6999 } else if (image.load_address != LLDB_INVALID_ADDRESS) { 7000 if (log) { 7001 std::string uuidstr = image.uuid.GetAsString(); 7002 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 7003 "UUID %s with load address 0x%" PRIx64, 7004 image.filename.c_str(), uuidstr.c_str(), 7005 image.load_address); 7006 } 7007 const bool address_is_slide = false; 7008 bool changed = false; 7009 module_sp->SetLoadAddress(process.GetTarget(), image.load_address, 7010 address_is_slide, changed); 7011 } else if (image.slide != 0) { 7012 if (log) { 7013 std::string uuidstr = image.uuid.GetAsString(); 7014 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 7015 "UUID %s with slide amount 0x%" PRIx64, 7016 image.filename.c_str(), uuidstr.c_str(), image.slide); 7017 } 7018 const bool address_is_slide = true; 7019 bool changed = false; 7020 module_sp->SetLoadAddress(process.GetTarget(), image.slide, 7021 address_is_slide, changed); 7022 } else { 7023 if (log) { 7024 std::string uuidstr = image.uuid.GetAsString(); 7025 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 7026 "UUID %s at its file address, no slide applied", 7027 image.filename.c_str(), uuidstr.c_str()); 7028 } 7029 const bool address_is_slide = true; 7030 bool changed = false; 7031 module_sp->SetLoadAddress(process.GetTarget(), 0, address_is_slide, 7032 changed); 7033 } 7034 } 7035 } 7036 if (added_modules.GetSize() > 0) { 7037 process.GetTarget().ModulesDidLoad(added_modules); 7038 process.Flush(); 7039 return true; 7040 } 7041 return false; 7042 } 7043