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, WritableDataBufferSP 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::WritableDataBufferSP header_data_sp, 958 const lldb::ProcessSP &process_sp, 959 lldb::addr_t header_addr) 960 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 961 m_mach_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 ModuleSP module_sp(GetModule()); 2205 if (!module_sp) 2206 return; 2207 2208 const FileSpec &file = m_file ? m_file : module_sp->GetFileSpec(); 2209 const char *file_name = file.GetFilename().AsCString("<Unknown>"); 2210 LLDB_SCOPED_TIMERF("ObjectFileMachO::ParseSymtab () module = %s", file_name); 2211 Progress progress(llvm::formatv("Parsing symbol table for {0}", file_name)); 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 llvm::MachO::dysymtab_command dysymtab = m_dysymtab; 2218 // The data element of type bool indicates that this entry is thumb 2219 // code. 2220 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2221 2222 // Record the address of every function/data that we add to the symtab. 2223 // We add symbols to the table in the order of most information (nlist 2224 // records) to least (function starts), and avoid duplicating symbols 2225 // via this set. 2226 llvm::DenseSet<addr_t> symbols_added; 2227 2228 // We are using a llvm::DenseSet for "symbols_added" so we must be sure we 2229 // do not add the tombstone or empty keys to the set. 2230 auto add_symbol_addr = [&symbols_added](lldb::addr_t file_addr) { 2231 // Don't add the tombstone or empty keys. 2232 if (file_addr == UINT64_MAX || file_addr == UINT64_MAX - 1) 2233 return; 2234 symbols_added.insert(file_addr); 2235 }; 2236 FunctionStarts function_starts; 2237 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2238 uint32_t i; 2239 FileSpecList dylib_files; 2240 Log *log = GetLog(LLDBLog::Symbols); 2241 llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2242 llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2243 llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2244 UUID image_uuid; 2245 2246 for (i = 0; i < m_header.ncmds; ++i) { 2247 const lldb::offset_t cmd_offset = offset; 2248 // Read in the load command and load command size 2249 llvm::MachO::load_command lc; 2250 if (m_data.GetU32(&offset, &lc, 2) == nullptr) 2251 break; 2252 // Watch for the symbol table load command 2253 switch (lc.cmd) { 2254 case LC_SYMTAB: 2255 symtab_load_command.cmd = lc.cmd; 2256 symtab_load_command.cmdsize = lc.cmdsize; 2257 // Read in the rest of the symtab load command 2258 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2259 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2260 return; 2261 break; 2262 2263 case LC_DYLD_INFO: 2264 case LC_DYLD_INFO_ONLY: 2265 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2266 dyld_info.cmd = lc.cmd; 2267 dyld_info.cmdsize = lc.cmdsize; 2268 } else { 2269 memset(&dyld_info, 0, sizeof(dyld_info)); 2270 } 2271 break; 2272 2273 case LC_LOAD_DYLIB: 2274 case LC_LOAD_WEAK_DYLIB: 2275 case LC_REEXPORT_DYLIB: 2276 case LC_LOADFVMLIB: 2277 case LC_LOAD_UPWARD_DYLIB: { 2278 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2279 const char *path = m_data.PeekCStr(name_offset); 2280 if (path) { 2281 FileSpec file_spec(path); 2282 // Strip the path if there is @rpath, @executable, etc so we just use 2283 // the basename 2284 if (path[0] == '@') 2285 file_spec.GetDirectory().Clear(); 2286 2287 if (lc.cmd == LC_REEXPORT_DYLIB) { 2288 m_reexported_dylibs.AppendIfUnique(file_spec); 2289 } 2290 2291 dylib_files.Append(file_spec); 2292 } 2293 } break; 2294 2295 case LC_DYLD_EXPORTS_TRIE: 2296 exports_trie_load_command.cmd = lc.cmd; 2297 exports_trie_load_command.cmdsize = lc.cmdsize; 2298 if (m_data.GetU32(&offset, &exports_trie_load_command.dataoff, 2) == 2299 nullptr) // fill in offset and size fields 2300 memset(&exports_trie_load_command, 0, 2301 sizeof(exports_trie_load_command)); 2302 break; 2303 case LC_FUNCTION_STARTS: 2304 function_starts_load_command.cmd = lc.cmd; 2305 function_starts_load_command.cmdsize = lc.cmdsize; 2306 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2307 nullptr) // fill in data offset and size fields 2308 memset(&function_starts_load_command, 0, 2309 sizeof(function_starts_load_command)); 2310 break; 2311 2312 case LC_UUID: { 2313 const uint8_t *uuid_bytes = m_data.PeekData(offset, 16); 2314 2315 if (uuid_bytes) 2316 image_uuid = UUID::fromOptionalData(uuid_bytes, 16); 2317 break; 2318 } 2319 2320 default: 2321 break; 2322 } 2323 offset = cmd_offset + lc.cmdsize; 2324 } 2325 2326 if (!symtab_load_command.cmd) 2327 return; 2328 2329 SectionList *section_list = GetSectionList(); 2330 if (section_list == nullptr) 2331 return; 2332 2333 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2334 const ByteOrder byte_order = m_data.GetByteOrder(); 2335 bool bit_width_32 = addr_byte_size == 4; 2336 const size_t nlist_byte_size = 2337 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2338 2339 DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size); 2340 DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size); 2341 DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size); 2342 DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order, 2343 addr_byte_size); 2344 DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size); 2345 2346 const addr_t nlist_data_byte_size = 2347 symtab_load_command.nsyms * nlist_byte_size; 2348 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2349 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2350 2351 ProcessSP process_sp(m_process_wp.lock()); 2352 Process *process = process_sp.get(); 2353 2354 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2355 bool is_shared_cache_image = IsSharedCacheBinary(); 2356 bool is_local_shared_cache_image = is_shared_cache_image && !IsInMemory(); 2357 SectionSP linkedit_section_sp( 2358 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2359 2360 if (process && m_header.filetype != llvm::MachO::MH_OBJECT && 2361 !is_local_shared_cache_image) { 2362 Target &target = process->GetTarget(); 2363 2364 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2365 2366 // Reading mach file from memory in a process or core file... 2367 2368 if (linkedit_section_sp) { 2369 addr_t linkedit_load_addr = 2370 linkedit_section_sp->GetLoadBaseAddress(&target); 2371 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2372 // We might be trying to access the symbol table before the 2373 // __LINKEDIT's load address has been set in the target. We can't 2374 // fail to read the symbol table, so calculate the right address 2375 // manually 2376 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2377 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2378 } 2379 2380 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2381 const addr_t symoff_addr = linkedit_load_addr + 2382 symtab_load_command.symoff - 2383 linkedit_file_offset; 2384 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2385 linkedit_file_offset; 2386 2387 // Always load dyld - the dynamic linker - from memory if we didn't 2388 // find a binary anywhere else. lldb will not register 2389 // dylib/framework/bundle loads/unloads if we don't have the dyld 2390 // symbols, we force dyld to load from memory despite the user's 2391 // target.memory-module-load-level setting. 2392 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2393 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2394 DataBufferSP nlist_data_sp( 2395 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2396 if (nlist_data_sp) 2397 nlist_data.SetData(nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2398 if (dysymtab.nindirectsyms != 0) { 2399 const addr_t indirect_syms_addr = linkedit_load_addr + 2400 dysymtab.indirectsymoff - 2401 linkedit_file_offset; 2402 DataBufferSP indirect_syms_data_sp(ReadMemory( 2403 process_sp, indirect_syms_addr, dysymtab.nindirectsyms * 4)); 2404 if (indirect_syms_data_sp) 2405 indirect_symbol_index_data.SetData( 2406 indirect_syms_data_sp, 0, 2407 indirect_syms_data_sp->GetByteSize()); 2408 // If this binary is outside the shared cache, 2409 // cache the string table. 2410 // Binaries in the shared cache all share a giant string table, 2411 // and we can't share the string tables across multiple 2412 // ObjectFileMachO's, so we'd end up re-reading this mega-strtab 2413 // for every binary in the shared cache - it would be a big perf 2414 // problem. For binaries outside the shared cache, it's faster to 2415 // read the entire strtab at once instead of piece-by-piece as we 2416 // process the nlist records. 2417 if (!is_shared_cache_image) { 2418 DataBufferSP strtab_data_sp( 2419 ReadMemory(process_sp, strtab_addr, strtab_data_byte_size)); 2420 if (strtab_data_sp) { 2421 strtab_data.SetData(strtab_data_sp, 0, 2422 strtab_data_sp->GetByteSize()); 2423 } 2424 } 2425 } 2426 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) { 2427 if (function_starts_load_command.cmd) { 2428 const addr_t func_start_addr = 2429 linkedit_load_addr + function_starts_load_command.dataoff - 2430 linkedit_file_offset; 2431 DataBufferSP func_start_data_sp( 2432 ReadMemory(process_sp, func_start_addr, 2433 function_starts_load_command.datasize)); 2434 if (func_start_data_sp) 2435 function_starts_data.SetData(func_start_data_sp, 0, 2436 func_start_data_sp->GetByteSize()); 2437 } 2438 } 2439 } 2440 } 2441 } else { 2442 if (is_local_shared_cache_image) { 2443 // The load commands in shared cache images are relative to the 2444 // beginning of the shared cache, not the library image. The 2445 // data we get handed when creating the ObjectFileMachO starts 2446 // at the beginning of a specific library and spans to the end 2447 // of the cache to be able to reach the shared LINKEDIT 2448 // segments. We need to convert the load command offsets to be 2449 // relative to the beginning of our specific image. 2450 lldb::addr_t linkedit_offset = linkedit_section_sp->GetFileOffset(); 2451 lldb::offset_t linkedit_slide = 2452 linkedit_offset - m_linkedit_original_offset; 2453 symtab_load_command.symoff += linkedit_slide; 2454 symtab_load_command.stroff += linkedit_slide; 2455 dyld_info.export_off += linkedit_slide; 2456 dysymtab.indirectsymoff += linkedit_slide; 2457 function_starts_load_command.dataoff += linkedit_slide; 2458 exports_trie_load_command.dataoff += linkedit_slide; 2459 } 2460 2461 nlist_data.SetData(m_data, symtab_load_command.symoff, 2462 nlist_data_byte_size); 2463 strtab_data.SetData(m_data, symtab_load_command.stroff, 2464 strtab_data_byte_size); 2465 2466 // We shouldn't have exports data from both the LC_DYLD_INFO command 2467 // AND the LC_DYLD_EXPORTS_TRIE command in the same binary: 2468 lldbassert(!((dyld_info.export_size > 0) 2469 && (exports_trie_load_command.datasize > 0))); 2470 if (dyld_info.export_size > 0) { 2471 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2472 dyld_info.export_size); 2473 } else if (exports_trie_load_command.datasize > 0) { 2474 dyld_trie_data.SetData(m_data, exports_trie_load_command.dataoff, 2475 exports_trie_load_command.datasize); 2476 } 2477 2478 if (dysymtab.nindirectsyms != 0) { 2479 indirect_symbol_index_data.SetData(m_data, dysymtab.indirectsymoff, 2480 dysymtab.nindirectsyms * 4); 2481 } 2482 if (function_starts_load_command.cmd) { 2483 function_starts_data.SetData(m_data, function_starts_load_command.dataoff, 2484 function_starts_load_command.datasize); 2485 } 2486 } 2487 2488 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2489 2490 ConstString g_segment_name_TEXT = GetSegmentNameTEXT(); 2491 ConstString g_segment_name_DATA = GetSegmentNameDATA(); 2492 ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2493 ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2494 ConstString g_segment_name_OBJC = GetSegmentNameOBJC(); 2495 ConstString g_section_name_eh_frame = GetSectionNameEHFrame(); 2496 SectionSP text_section_sp( 2497 section_list->FindSectionByName(g_segment_name_TEXT)); 2498 SectionSP data_section_sp( 2499 section_list->FindSectionByName(g_segment_name_DATA)); 2500 SectionSP data_dirty_section_sp( 2501 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2502 SectionSP data_const_section_sp( 2503 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2504 SectionSP objc_section_sp( 2505 section_list->FindSectionByName(g_segment_name_OBJC)); 2506 SectionSP eh_frame_section_sp; 2507 if (text_section_sp.get()) 2508 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2509 g_section_name_eh_frame); 2510 else 2511 eh_frame_section_sp = 2512 section_list->FindSectionByName(g_section_name_eh_frame); 2513 2514 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2515 const bool always_thumb = GetArchitecture().IsAlwaysThumbInstructions(); 2516 2517 // lldb works best if it knows the start address of all functions in a 2518 // module. Linker symbols or debug info are normally the best source of 2519 // information for start addr / size but they may be stripped in a released 2520 // binary. Two additional sources of information exist in Mach-O binaries: 2521 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2522 // function's start address in the 2523 // binary, relative to the text section. 2524 // eh_frame - the eh_frame FDEs have the start addr & size of 2525 // each function 2526 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2527 // all modern binaries. 2528 // Binaries built to run on older releases may need to use eh_frame 2529 // information. 2530 2531 if (text_section_sp && function_starts_data.GetByteSize()) { 2532 FunctionStarts::Entry function_start_entry; 2533 function_start_entry.data = false; 2534 lldb::offset_t function_start_offset = 0; 2535 function_start_entry.addr = text_section_sp->GetFileAddress(); 2536 uint64_t delta; 2537 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2538 0) { 2539 // Now append the current entry 2540 function_start_entry.addr += delta; 2541 if (is_arm) { 2542 if (function_start_entry.addr & 1) { 2543 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2544 function_start_entry.data = true; 2545 } else if (always_thumb) { 2546 function_start_entry.data = true; 2547 } 2548 } 2549 function_starts.Append(function_start_entry); 2550 } 2551 } else { 2552 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2553 // load command claiming an eh_frame but it doesn't actually have the 2554 // eh_frame content. And if we have a dSYM, we don't need to do any of 2555 // this fill-in-the-missing-symbols works anyway - the debug info should 2556 // give us all the functions in the module. 2557 if (text_section_sp.get() && eh_frame_section_sp.get() && 2558 m_type != eTypeDebugInfo) { 2559 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2560 DWARFCallFrameInfo::EH); 2561 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2562 eh_frame.GetFunctionAddressAndSizeVector(functions); 2563 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2564 size_t count = functions.GetSize(); 2565 for (size_t i = 0; i < count; ++i) { 2566 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2567 functions.GetEntryAtIndex(i); 2568 if (func) { 2569 FunctionStarts::Entry function_start_entry; 2570 function_start_entry.addr = func->base - text_base_addr; 2571 if (is_arm) { 2572 if (function_start_entry.addr & 1) { 2573 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2574 function_start_entry.data = true; 2575 } else if (always_thumb) { 2576 function_start_entry.data = true; 2577 } 2578 } 2579 function_starts.Append(function_start_entry); 2580 } 2581 } 2582 } 2583 } 2584 2585 const size_t function_starts_count = function_starts.GetSize(); 2586 2587 // For user process binaries (executables, dylibs, frameworks, bundles), if 2588 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2589 // going to assume the binary has been stripped. Don't allow assembly 2590 // language instruction emulation because we don't know proper function 2591 // start boundaries. 2592 // 2593 // For all other types of binaries (kernels, stand-alone bare board 2594 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2595 // sections - we should not make any assumptions about them based on that. 2596 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2597 m_allow_assembly_emulation_unwind_plans = false; 2598 Log *unwind_or_symbol_log(GetLog(LLDBLog::Symbols | LLDBLog::Unwind)); 2599 2600 if (unwind_or_symbol_log) 2601 module_sp->LogMessage( 2602 unwind_or_symbol_log, 2603 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2604 } 2605 2606 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() 2607 ? eh_frame_section_sp->GetID() 2608 : static_cast<user_id_t>(NO_SECT); 2609 2610 uint32_t N_SO_index = UINT32_MAX; 2611 2612 MachSymtabSectionInfo section_info(section_list); 2613 std::vector<uint32_t> N_FUN_indexes; 2614 std::vector<uint32_t> N_NSYM_indexes; 2615 std::vector<uint32_t> N_INCL_indexes; 2616 std::vector<uint32_t> N_BRAC_indexes; 2617 std::vector<uint32_t> N_COMM_indexes; 2618 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2619 typedef llvm::DenseMap<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2620 typedef llvm::DenseMap<const char *, uint32_t> ConstNameToSymbolIndexMap; 2621 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2622 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2623 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2624 // Any symbols that get merged into another will get an entry in this map 2625 // so we know 2626 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2627 uint32_t nlist_idx = 0; 2628 Symbol *symbol_ptr = nullptr; 2629 2630 uint32_t sym_idx = 0; 2631 Symbol *sym = nullptr; 2632 size_t num_syms = 0; 2633 std::string memory_symbol_name; 2634 uint32_t unmapped_local_symbols_found = 0; 2635 2636 std::vector<TrieEntryWithOffset> reexport_trie_entries; 2637 std::vector<TrieEntryWithOffset> external_sym_trie_entries; 2638 std::set<lldb::addr_t> resolver_addresses; 2639 2640 if (dyld_trie_data.GetByteSize() > 0) { 2641 ConstString text_segment_name("__TEXT"); 2642 SectionSP text_segment_sp = 2643 GetSectionList()->FindSectionByName(text_segment_name); 2644 lldb::addr_t text_segment_file_addr = LLDB_INVALID_ADDRESS; 2645 if (text_segment_sp) 2646 text_segment_file_addr = text_segment_sp->GetFileAddress(); 2647 std::vector<llvm::StringRef> nameSlices; 2648 ParseTrieEntries(dyld_trie_data, 0, is_arm, text_segment_file_addr, 2649 nameSlices, resolver_addresses, reexport_trie_entries, 2650 external_sym_trie_entries); 2651 } 2652 2653 typedef std::set<ConstString> IndirectSymbols; 2654 IndirectSymbols indirect_symbol_names; 2655 2656 #if TARGET_OS_IPHONE 2657 2658 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2659 // optimized by moving LOCAL symbols out of the memory mapped portion of 2660 // the DSC. The symbol information has all been retained, but it isn't 2661 // available in the normal nlist data. However, there *are* duplicate 2662 // entries of *some* 2663 // LOCAL symbols in the normal nlist data. To handle this situation 2664 // correctly, we must first attempt 2665 // to parse any DSC unmapped symbol information. If we find any, we set a 2666 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2667 2668 if (IsSharedCacheBinary()) { 2669 // Before we can start mapping the DSC, we need to make certain the 2670 // target process is actually using the cache we can find. 2671 2672 // Next we need to determine the correct path for the dyld shared cache. 2673 2674 ArchSpec header_arch = GetArchitecture(); 2675 2676 UUID dsc_uuid; 2677 UUID process_shared_cache_uuid; 2678 addr_t process_shared_cache_base_addr; 2679 2680 if (process) { 2681 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, 2682 process_shared_cache_uuid); 2683 } 2684 2685 __block bool found_image = false; 2686 __block void *nlist_buffer = nullptr; 2687 __block unsigned nlist_count = 0; 2688 __block char *string_table = nullptr; 2689 __block vm_offset_t vm_nlist_memory = 0; 2690 __block mach_msg_type_number_t vm_nlist_bytes_read = 0; 2691 __block vm_offset_t vm_string_memory = 0; 2692 __block mach_msg_type_number_t vm_string_bytes_read = 0; 2693 2694 auto _ = llvm::make_scope_exit(^{ 2695 if (vm_nlist_memory) 2696 vm_deallocate(mach_task_self(), vm_nlist_memory, vm_nlist_bytes_read); 2697 if (vm_string_memory) 2698 vm_deallocate(mach_task_self(), vm_string_memory, vm_string_bytes_read); 2699 }); 2700 2701 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 2702 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 2703 UndefinedNameToDescMap undefined_name_to_desc; 2704 SymbolIndexToName reexport_shlib_needs_fixup; 2705 2706 dyld_for_each_installed_shared_cache(^(dyld_shared_cache_t shared_cache) { 2707 uuid_t cache_uuid; 2708 dyld_shared_cache_copy_uuid(shared_cache, &cache_uuid); 2709 if (found_image) 2710 return; 2711 2712 if (process_shared_cache_uuid.IsValid() && 2713 process_shared_cache_uuid != UUID::fromOptionalData(&cache_uuid, 16)) 2714 return; 2715 2716 dyld_shared_cache_for_each_image(shared_cache, ^(dyld_image_t image) { 2717 uuid_t dsc_image_uuid; 2718 if (found_image) 2719 return; 2720 2721 dyld_image_copy_uuid(image, &dsc_image_uuid); 2722 if (image_uuid != UUID::fromOptionalData(dsc_image_uuid, 16)) 2723 return; 2724 2725 found_image = true; 2726 2727 // Compute the size of the string table. We need to ask dyld for a 2728 // new SPI to avoid this step. 2729 dyld_image_local_nlist_content_4Symbolication( 2730 image, ^(const void *nlistStart, uint64_t nlistCount, 2731 const char *stringTable) { 2732 if (!nlistStart || !nlistCount) 2733 return; 2734 2735 // The buffers passed here are valid only inside the block. 2736 // Use vm_read to make a cheap copy of them available for our 2737 // processing later. 2738 kern_return_t ret = 2739 vm_read(mach_task_self(), (vm_address_t)nlistStart, 2740 nlist_byte_size * nlistCount, &vm_nlist_memory, 2741 &vm_nlist_bytes_read); 2742 if (ret != KERN_SUCCESS) 2743 return; 2744 assert(vm_nlist_bytes_read == nlist_byte_size * nlistCount); 2745 2746 // We don't know the size of the string table. It's cheaper 2747 // to map the whol VM region than to determine the size by 2748 // parsing all teh nlist entries. 2749 vm_address_t string_address = (vm_address_t)stringTable; 2750 vm_size_t region_size; 2751 mach_msg_type_number_t info_count = VM_REGION_BASIC_INFO_COUNT_64; 2752 vm_region_basic_info_data_t info; 2753 memory_object_name_t object; 2754 ret = vm_region_64(mach_task_self(), &string_address, 2755 ®ion_size, VM_REGION_BASIC_INFO_64, 2756 (vm_region_info_t)&info, &info_count, &object); 2757 if (ret != KERN_SUCCESS) 2758 return; 2759 2760 ret = vm_read(mach_task_self(), (vm_address_t)stringTable, 2761 region_size - 2762 ((vm_address_t)stringTable - string_address), 2763 &vm_string_memory, &vm_string_bytes_read); 2764 if (ret != KERN_SUCCESS) 2765 return; 2766 2767 nlist_buffer = (void *)vm_nlist_memory; 2768 string_table = (char *)vm_string_memory; 2769 nlist_count = nlistCount; 2770 }); 2771 }); 2772 }); 2773 if (nlist_buffer) { 2774 DataExtractor dsc_local_symbols_data(nlist_buffer, 2775 nlist_count * nlist_byte_size, 2776 byte_order, addr_byte_size); 2777 unmapped_local_symbols_found = nlist_count; 2778 2779 // The normal nlist code cannot correctly size the Symbols 2780 // array, we need to allocate it here. 2781 sym = symtab.Resize( 2782 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2783 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2784 num_syms = symtab.GetNumSymbols(); 2785 2786 lldb::offset_t nlist_data_offset = 0; 2787 2788 for (uint32_t nlist_index = 0; 2789 nlist_index < nlist_count; 2790 nlist_index++) { 2791 ///////////////////////////// 2792 { 2793 llvm::Optional<struct nlist_64> nlist_maybe = 2794 ParseNList(dsc_local_symbols_data, nlist_data_offset, 2795 nlist_byte_size); 2796 if (!nlist_maybe) 2797 break; 2798 struct nlist_64 nlist = *nlist_maybe; 2799 2800 SymbolType type = eSymbolTypeInvalid; 2801 const char *symbol_name = string_table + nlist.n_strx; 2802 2803 if (symbol_name == NULL) { 2804 // No symbol should be NULL, even the symbols with no 2805 // string values should have an offset zero which 2806 // points to an empty C-string 2807 Host::SystemLog( 2808 Host::eSystemLogError, 2809 "error: DSC unmapped local symbol[%u] has invalid " 2810 "string table offset 0x%x in %s, ignoring symbol\n", 2811 nlist_index, nlist.n_strx, 2812 module_sp->GetFileSpec().GetPath().c_str()); 2813 continue; 2814 } 2815 if (symbol_name[0] == '\0') 2816 symbol_name = NULL; 2817 2818 const char *symbol_name_non_abi_mangled = NULL; 2819 2820 SectionSP symbol_section; 2821 uint32_t symbol_byte_size = 0; 2822 bool add_nlist = true; 2823 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2824 bool demangled_is_synthesized = false; 2825 bool is_gsym = false; 2826 bool set_value = true; 2827 2828 assert(sym_idx < num_syms); 2829 2830 sym[sym_idx].SetDebug(is_debug); 2831 2832 if (is_debug) { 2833 switch (nlist.n_type) { 2834 case N_GSYM: 2835 // global symbol: name,,NO_SECT,type,0 2836 // Sometimes the N_GSYM value contains the address. 2837 2838 // FIXME: In the .o files, we have a GSYM and a debug 2839 // symbol for all the ObjC data. They 2840 // have the same address, but we want to ensure that 2841 // we always find only the real symbol, 'cause we 2842 // don't currently correctly attribute the 2843 // GSYM one to the ObjCClass/Ivar/MetaClass 2844 // symbol type. This is a temporary hack to make 2845 // sure the ObjectiveC symbols get treated correctly. 2846 // To do this right, we should coalesce all the GSYM 2847 // & global symbols that have the same address. 2848 2849 is_gsym = true; 2850 sym[sym_idx].SetExternal(true); 2851 2852 if (symbol_name && symbol_name[0] == '_' && 2853 symbol_name[1] == 'O') { 2854 llvm::StringRef symbol_name_ref(symbol_name); 2855 if (symbol_name_ref.startswith( 2856 g_objc_v2_prefix_class)) { 2857 symbol_name_non_abi_mangled = symbol_name + 1; 2858 symbol_name = 2859 symbol_name + g_objc_v2_prefix_class.size(); 2860 type = eSymbolTypeObjCClass; 2861 demangled_is_synthesized = true; 2862 2863 } else if (symbol_name_ref.startswith( 2864 g_objc_v2_prefix_metaclass)) { 2865 symbol_name_non_abi_mangled = symbol_name + 1; 2866 symbol_name = 2867 symbol_name + g_objc_v2_prefix_metaclass.size(); 2868 type = eSymbolTypeObjCMetaClass; 2869 demangled_is_synthesized = true; 2870 } else if (symbol_name_ref.startswith( 2871 g_objc_v2_prefix_ivar)) { 2872 symbol_name_non_abi_mangled = symbol_name + 1; 2873 symbol_name = 2874 symbol_name + g_objc_v2_prefix_ivar.size(); 2875 type = eSymbolTypeObjCIVar; 2876 demangled_is_synthesized = true; 2877 } 2878 } else { 2879 if (nlist.n_value != 0) 2880 symbol_section = section_info.GetSection( 2881 nlist.n_sect, nlist.n_value); 2882 type = eSymbolTypeData; 2883 } 2884 break; 2885 2886 case N_FNAME: 2887 // procedure name (f77 kludge): name,,NO_SECT,0,0 2888 type = eSymbolTypeCompiler; 2889 break; 2890 2891 case N_FUN: 2892 // procedure: name,,n_sect,linenumber,address 2893 if (symbol_name) { 2894 type = eSymbolTypeCode; 2895 symbol_section = section_info.GetSection( 2896 nlist.n_sect, nlist.n_value); 2897 2898 N_FUN_addr_to_sym_idx.insert( 2899 std::make_pair(nlist.n_value, sym_idx)); 2900 // We use the current number of symbols in the 2901 // symbol table in lieu of using nlist_idx in case 2902 // we ever start trimming entries out 2903 N_FUN_indexes.push_back(sym_idx); 2904 } else { 2905 type = eSymbolTypeCompiler; 2906 2907 if (!N_FUN_indexes.empty()) { 2908 // Copy the size of the function into the 2909 // original 2910 // STAB entry so we don't have 2911 // to hunt for it later 2912 symtab.SymbolAtIndex(N_FUN_indexes.back()) 2913 ->SetByteSize(nlist.n_value); 2914 N_FUN_indexes.pop_back(); 2915 // We don't really need the end function STAB as 2916 // it contains the size which we already placed 2917 // with the original symbol, so don't add it if 2918 // we want a minimal symbol table 2919 add_nlist = false; 2920 } 2921 } 2922 break; 2923 2924 case N_STSYM: 2925 // static symbol: name,,n_sect,type,address 2926 N_STSYM_addr_to_sym_idx.insert( 2927 std::make_pair(nlist.n_value, sym_idx)); 2928 symbol_section = section_info.GetSection(nlist.n_sect, 2929 nlist.n_value); 2930 if (symbol_name && symbol_name[0]) { 2931 type = ObjectFile::GetSymbolTypeFromName( 2932 symbol_name + 1, eSymbolTypeData); 2933 } 2934 break; 2935 2936 case N_LCSYM: 2937 // .lcomm symbol: name,,n_sect,type,address 2938 symbol_section = section_info.GetSection(nlist.n_sect, 2939 nlist.n_value); 2940 type = eSymbolTypeCommonBlock; 2941 break; 2942 2943 case N_BNSYM: 2944 // We use the current number of symbols in the symbol 2945 // table in lieu of using nlist_idx in case we ever 2946 // start trimming entries out Skip these if we want 2947 // minimal symbol tables 2948 add_nlist = false; 2949 break; 2950 2951 case N_ENSYM: 2952 // Set the size of the N_BNSYM to the terminating 2953 // index of this N_ENSYM so that we can always skip 2954 // the entire symbol if we need to navigate more 2955 // quickly at the source level when parsing STABS 2956 // Skip these if we want minimal symbol tables 2957 add_nlist = false; 2958 break; 2959 2960 case N_OPT: 2961 // emitted with gcc2_compiled and in gcc source 2962 type = eSymbolTypeCompiler; 2963 break; 2964 2965 case N_RSYM: 2966 // register sym: name,,NO_SECT,type,register 2967 type = eSymbolTypeVariable; 2968 break; 2969 2970 case N_SLINE: 2971 // src line: 0,,n_sect,linenumber,address 2972 symbol_section = section_info.GetSection(nlist.n_sect, 2973 nlist.n_value); 2974 type = eSymbolTypeLineEntry; 2975 break; 2976 2977 case N_SSYM: 2978 // structure elt: name,,NO_SECT,type,struct_offset 2979 type = eSymbolTypeVariableType; 2980 break; 2981 2982 case N_SO: 2983 // source file name 2984 type = eSymbolTypeSourceFile; 2985 if (symbol_name == NULL) { 2986 add_nlist = false; 2987 if (N_SO_index != UINT32_MAX) { 2988 // Set the size of the N_SO to the terminating 2989 // index of this N_SO so that we can always skip 2990 // the entire N_SO if we need to navigate more 2991 // quickly at the source level when parsing STABS 2992 symbol_ptr = symtab.SymbolAtIndex(N_SO_index); 2993 symbol_ptr->SetByteSize(sym_idx); 2994 symbol_ptr->SetSizeIsSibling(true); 2995 } 2996 N_NSYM_indexes.clear(); 2997 N_INCL_indexes.clear(); 2998 N_BRAC_indexes.clear(); 2999 N_COMM_indexes.clear(); 3000 N_FUN_indexes.clear(); 3001 N_SO_index = UINT32_MAX; 3002 } else { 3003 // We use the current number of symbols in the 3004 // symbol table in lieu of using nlist_idx in case 3005 // we ever start trimming entries out 3006 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3007 if (N_SO_has_full_path) { 3008 if ((N_SO_index == sym_idx - 1) && 3009 ((sym_idx - 1) < num_syms)) { 3010 // We have two consecutive N_SO entries where 3011 // the first contains a directory and the 3012 // second contains a full path. 3013 sym[sym_idx - 1].GetMangled().SetValue( 3014 ConstString(symbol_name), false); 3015 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3016 add_nlist = false; 3017 } else { 3018 // This is the first entry in a N_SO that 3019 // contains a directory or 3020 // a full path to the source file 3021 N_SO_index = sym_idx; 3022 } 3023 } else if ((N_SO_index == sym_idx - 1) && 3024 ((sym_idx - 1) < num_syms)) { 3025 // This is usually the second N_SO entry that 3026 // contains just the filename, so here we combine 3027 // it with the first one if we are minimizing the 3028 // symbol table 3029 const char *so_path = sym[sym_idx - 1] 3030 .GetMangled() 3031 .GetDemangledName() 3032 .AsCString(); 3033 if (so_path && so_path[0]) { 3034 std::string full_so_path(so_path); 3035 const size_t double_slash_pos = 3036 full_so_path.find("//"); 3037 if (double_slash_pos != std::string::npos) { 3038 // The linker has been generating bad N_SO 3039 // entries with doubled up paths 3040 // in the format "%s%s" where the first 3041 // string in the DW_AT_comp_dir, and the 3042 // second is the directory for the source 3043 // file so you end up with a path that looks 3044 // like "/tmp/src//tmp/src/" 3045 FileSpec so_dir(so_path); 3046 if (!FileSystem::Instance().Exists(so_dir)) { 3047 so_dir.SetFile( 3048 &full_so_path[double_slash_pos + 1], 3049 FileSpec::Style::native); 3050 if (FileSystem::Instance().Exists(so_dir)) { 3051 // Trim off the incorrect path 3052 full_so_path.erase(0, double_slash_pos + 1); 3053 } 3054 } 3055 } 3056 if (*full_so_path.rbegin() != '/') 3057 full_so_path += '/'; 3058 full_so_path += symbol_name; 3059 sym[sym_idx - 1].GetMangled().SetValue( 3060 ConstString(full_so_path.c_str()), false); 3061 add_nlist = false; 3062 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3063 } 3064 } else { 3065 // This could be a relative path to a N_SO 3066 N_SO_index = sym_idx; 3067 } 3068 } 3069 break; 3070 3071 case N_OSO: 3072 // object file name: name,,0,0,st_mtime 3073 type = eSymbolTypeObjectFile; 3074 break; 3075 3076 case N_LSYM: 3077 // local sym: name,,NO_SECT,type,offset 3078 type = eSymbolTypeLocal; 3079 break; 3080 3081 // INCL scopes 3082 case N_BINCL: 3083 // include file beginning: name,,NO_SECT,0,sum We use 3084 // the current number of symbols in the symbol table 3085 // in lieu of using nlist_idx in case we ever start 3086 // trimming entries out 3087 N_INCL_indexes.push_back(sym_idx); 3088 type = eSymbolTypeScopeBegin; 3089 break; 3090 3091 case N_EINCL: 3092 // include file end: name,,NO_SECT,0,0 3093 // Set the size of the N_BINCL to the terminating 3094 // index of this N_EINCL so that we can always skip 3095 // the entire symbol if we need to navigate more 3096 // quickly at the source level when parsing STABS 3097 if (!N_INCL_indexes.empty()) { 3098 symbol_ptr = 3099 symtab.SymbolAtIndex(N_INCL_indexes.back()); 3100 symbol_ptr->SetByteSize(sym_idx + 1); 3101 symbol_ptr->SetSizeIsSibling(true); 3102 N_INCL_indexes.pop_back(); 3103 } 3104 type = eSymbolTypeScopeEnd; 3105 break; 3106 3107 case N_SOL: 3108 // #included file name: name,,n_sect,0,address 3109 type = eSymbolTypeHeaderFile; 3110 3111 // We currently don't use the header files on darwin 3112 add_nlist = false; 3113 break; 3114 3115 case N_PARAMS: 3116 // compiler parameters: name,,NO_SECT,0,0 3117 type = eSymbolTypeCompiler; 3118 break; 3119 3120 case N_VERSION: 3121 // compiler version: name,,NO_SECT,0,0 3122 type = eSymbolTypeCompiler; 3123 break; 3124 3125 case N_OLEVEL: 3126 // compiler -O level: name,,NO_SECT,0,0 3127 type = eSymbolTypeCompiler; 3128 break; 3129 3130 case N_PSYM: 3131 // parameter: name,,NO_SECT,type,offset 3132 type = eSymbolTypeVariable; 3133 break; 3134 3135 case N_ENTRY: 3136 // alternate entry: name,,n_sect,linenumber,address 3137 symbol_section = section_info.GetSection(nlist.n_sect, 3138 nlist.n_value); 3139 type = eSymbolTypeLineEntry; 3140 break; 3141 3142 // Left and Right Braces 3143 case N_LBRAC: 3144 // left bracket: 0,,NO_SECT,nesting level,address We 3145 // use the current number of symbols in the symbol 3146 // table in lieu of using nlist_idx in case we ever 3147 // start trimming entries out 3148 symbol_section = section_info.GetSection(nlist.n_sect, 3149 nlist.n_value); 3150 N_BRAC_indexes.push_back(sym_idx); 3151 type = eSymbolTypeScopeBegin; 3152 break; 3153 3154 case N_RBRAC: 3155 // right bracket: 0,,NO_SECT,nesting level,address 3156 // Set the size of the N_LBRAC to the terminating 3157 // index of this N_RBRAC so that we can always skip 3158 // the entire symbol if we need to navigate more 3159 // quickly at the source level when parsing STABS 3160 symbol_section = section_info.GetSection(nlist.n_sect, 3161 nlist.n_value); 3162 if (!N_BRAC_indexes.empty()) { 3163 symbol_ptr = 3164 symtab.SymbolAtIndex(N_BRAC_indexes.back()); 3165 symbol_ptr->SetByteSize(sym_idx + 1); 3166 symbol_ptr->SetSizeIsSibling(true); 3167 N_BRAC_indexes.pop_back(); 3168 } 3169 type = eSymbolTypeScopeEnd; 3170 break; 3171 3172 case N_EXCL: 3173 // deleted include file: name,,NO_SECT,0,sum 3174 type = eSymbolTypeHeaderFile; 3175 break; 3176 3177 // COMM scopes 3178 case N_BCOMM: 3179 // begin common: name,,NO_SECT,0,0 3180 // We use the current number of symbols in the symbol 3181 // table in lieu of using nlist_idx in case we ever 3182 // start trimming entries out 3183 type = eSymbolTypeScopeBegin; 3184 N_COMM_indexes.push_back(sym_idx); 3185 break; 3186 3187 case N_ECOML: 3188 // end common (local name): 0,,n_sect,0,address 3189 symbol_section = section_info.GetSection(nlist.n_sect, 3190 nlist.n_value); 3191 // Fall through 3192 3193 case N_ECOMM: 3194 // end common: name,,n_sect,0,0 3195 // Set the size of the N_BCOMM to the terminating 3196 // index of this N_ECOMM/N_ECOML so that we can 3197 // always skip the entire symbol if we need to 3198 // navigate more quickly at the source level when 3199 // parsing STABS 3200 if (!N_COMM_indexes.empty()) { 3201 symbol_ptr = 3202 symtab.SymbolAtIndex(N_COMM_indexes.back()); 3203 symbol_ptr->SetByteSize(sym_idx + 1); 3204 symbol_ptr->SetSizeIsSibling(true); 3205 N_COMM_indexes.pop_back(); 3206 } 3207 type = eSymbolTypeScopeEnd; 3208 break; 3209 3210 case N_LENG: 3211 // second stab entry with length information 3212 type = eSymbolTypeAdditional; 3213 break; 3214 3215 default: 3216 break; 3217 } 3218 } else { 3219 // uint8_t n_pext = N_PEXT & nlist.n_type; 3220 uint8_t n_type = N_TYPE & nlist.n_type; 3221 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3222 3223 switch (n_type) { 3224 case N_INDR: { 3225 const char *reexport_name_cstr = 3226 strtab_data.PeekCStr(nlist.n_value); 3227 if (reexport_name_cstr && reexport_name_cstr[0]) { 3228 type = eSymbolTypeReExported; 3229 ConstString reexport_name( 3230 reexport_name_cstr + 3231 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3232 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3233 set_value = false; 3234 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3235 indirect_symbol_names.insert(ConstString( 3236 symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3237 } else 3238 type = eSymbolTypeUndefined; 3239 } break; 3240 3241 case N_UNDF: 3242 if (symbol_name && symbol_name[0]) { 3243 ConstString undefined_name( 3244 symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3245 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3246 } 3247 // Fall through 3248 case N_PBUD: 3249 type = eSymbolTypeUndefined; 3250 break; 3251 3252 case N_ABS: 3253 type = eSymbolTypeAbsolute; 3254 break; 3255 3256 case N_SECT: { 3257 symbol_section = section_info.GetSection(nlist.n_sect, 3258 nlist.n_value); 3259 3260 if (symbol_section == NULL) { 3261 // TODO: warn about this? 3262 add_nlist = false; 3263 break; 3264 } 3265 3266 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3267 type = eSymbolTypeException; 3268 } else { 3269 uint32_t section_type = 3270 symbol_section->Get() & SECTION_TYPE; 3271 3272 switch (section_type) { 3273 case S_CSTRING_LITERALS: 3274 type = eSymbolTypeData; 3275 break; // section with only literal C strings 3276 case S_4BYTE_LITERALS: 3277 type = eSymbolTypeData; 3278 break; // section with only 4 byte literals 3279 case S_8BYTE_LITERALS: 3280 type = eSymbolTypeData; 3281 break; // section with only 8 byte literals 3282 case S_LITERAL_POINTERS: 3283 type = eSymbolTypeTrampoline; 3284 break; // section with only pointers to literals 3285 case S_NON_LAZY_SYMBOL_POINTERS: 3286 type = eSymbolTypeTrampoline; 3287 break; // section with only non-lazy symbol 3288 // pointers 3289 case S_LAZY_SYMBOL_POINTERS: 3290 type = eSymbolTypeTrampoline; 3291 break; // section with only lazy symbol pointers 3292 case S_SYMBOL_STUBS: 3293 type = eSymbolTypeTrampoline; 3294 break; // section with only symbol stubs, byte 3295 // size of stub in the reserved2 field 3296 case S_MOD_INIT_FUNC_POINTERS: 3297 type = eSymbolTypeCode; 3298 break; // section with only function pointers for 3299 // initialization 3300 case S_MOD_TERM_FUNC_POINTERS: 3301 type = eSymbolTypeCode; 3302 break; // section with only function pointers for 3303 // termination 3304 case S_INTERPOSING: 3305 type = eSymbolTypeTrampoline; 3306 break; // section with only pairs of function 3307 // pointers for interposing 3308 case S_16BYTE_LITERALS: 3309 type = eSymbolTypeData; 3310 break; // section with only 16 byte literals 3311 case S_DTRACE_DOF: 3312 type = eSymbolTypeInstrumentation; 3313 break; 3314 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3315 type = eSymbolTypeTrampoline; 3316 break; 3317 default: 3318 switch (symbol_section->GetType()) { 3319 case lldb::eSectionTypeCode: 3320 type = eSymbolTypeCode; 3321 break; 3322 case eSectionTypeData: 3323 case eSectionTypeDataCString: // Inlined C string 3324 // data 3325 case eSectionTypeDataCStringPointers: // Pointers 3326 // to C 3327 // string 3328 // data 3329 case eSectionTypeDataSymbolAddress: // Address of 3330 // a symbol in 3331 // the symbol 3332 // table 3333 case eSectionTypeData4: 3334 case eSectionTypeData8: 3335 case eSectionTypeData16: 3336 type = eSymbolTypeData; 3337 break; 3338 default: 3339 break; 3340 } 3341 break; 3342 } 3343 3344 if (type == eSymbolTypeInvalid) { 3345 const char *symbol_sect_name = 3346 symbol_section->GetName().AsCString(); 3347 if (symbol_section->IsDescendant( 3348 text_section_sp.get())) { 3349 if (symbol_section->IsClear( 3350 S_ATTR_PURE_INSTRUCTIONS | 3351 S_ATTR_SELF_MODIFYING_CODE | 3352 S_ATTR_SOME_INSTRUCTIONS)) 3353 type = eSymbolTypeData; 3354 else 3355 type = eSymbolTypeCode; 3356 } else if (symbol_section->IsDescendant( 3357 data_section_sp.get()) || 3358 symbol_section->IsDescendant( 3359 data_dirty_section_sp.get()) || 3360 symbol_section->IsDescendant( 3361 data_const_section_sp.get())) { 3362 if (symbol_sect_name && 3363 ::strstr(symbol_sect_name, "__objc") == 3364 symbol_sect_name) { 3365 type = eSymbolTypeRuntime; 3366 3367 if (symbol_name) { 3368 llvm::StringRef symbol_name_ref(symbol_name); 3369 if (symbol_name_ref.startswith("_OBJC_")) { 3370 llvm::StringRef 3371 g_objc_v2_prefix_class( 3372 "_OBJC_CLASS_$_"); 3373 llvm::StringRef 3374 g_objc_v2_prefix_metaclass( 3375 "_OBJC_METACLASS_$_"); 3376 llvm::StringRef 3377 g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 3378 if (symbol_name_ref.startswith( 3379 g_objc_v2_prefix_class)) { 3380 symbol_name_non_abi_mangled = 3381 symbol_name + 1; 3382 symbol_name = 3383 symbol_name + 3384 g_objc_v2_prefix_class.size(); 3385 type = eSymbolTypeObjCClass; 3386 demangled_is_synthesized = true; 3387 } else if ( 3388 symbol_name_ref.startswith( 3389 g_objc_v2_prefix_metaclass)) { 3390 symbol_name_non_abi_mangled = 3391 symbol_name + 1; 3392 symbol_name = 3393 symbol_name + 3394 g_objc_v2_prefix_metaclass.size(); 3395 type = eSymbolTypeObjCMetaClass; 3396 demangled_is_synthesized = true; 3397 } else if (symbol_name_ref.startswith( 3398 g_objc_v2_prefix_ivar)) { 3399 symbol_name_non_abi_mangled = 3400 symbol_name + 1; 3401 symbol_name = 3402 symbol_name + 3403 g_objc_v2_prefix_ivar.size(); 3404 type = eSymbolTypeObjCIVar; 3405 demangled_is_synthesized = true; 3406 } 3407 } 3408 } 3409 } else if (symbol_sect_name && 3410 ::strstr(symbol_sect_name, 3411 "__gcc_except_tab") == 3412 symbol_sect_name) { 3413 type = eSymbolTypeException; 3414 } else { 3415 type = eSymbolTypeData; 3416 } 3417 } else if (symbol_sect_name && 3418 ::strstr(symbol_sect_name, "__IMPORT") == 3419 symbol_sect_name) { 3420 type = eSymbolTypeTrampoline; 3421 } else if (symbol_section->IsDescendant( 3422 objc_section_sp.get())) { 3423 type = eSymbolTypeRuntime; 3424 if (symbol_name && symbol_name[0] == '.') { 3425 llvm::StringRef symbol_name_ref(symbol_name); 3426 llvm::StringRef 3427 g_objc_v1_prefix_class(".objc_class_name_"); 3428 if (symbol_name_ref.startswith( 3429 g_objc_v1_prefix_class)) { 3430 symbol_name_non_abi_mangled = symbol_name; 3431 symbol_name = symbol_name + 3432 g_objc_v1_prefix_class.size(); 3433 type = eSymbolTypeObjCClass; 3434 demangled_is_synthesized = true; 3435 } 3436 } 3437 } 3438 } 3439 } 3440 } break; 3441 } 3442 } 3443 3444 if (add_nlist) { 3445 uint64_t symbol_value = nlist.n_value; 3446 if (symbol_name_non_abi_mangled) { 3447 sym[sym_idx].GetMangled().SetMangledName( 3448 ConstString(symbol_name_non_abi_mangled)); 3449 sym[sym_idx].GetMangled().SetDemangledName( 3450 ConstString(symbol_name)); 3451 } else { 3452 bool symbol_name_is_mangled = false; 3453 3454 if (symbol_name && symbol_name[0] == '_') { 3455 symbol_name_is_mangled = symbol_name[1] == '_'; 3456 symbol_name++; // Skip the leading underscore 3457 } 3458 3459 if (symbol_name) { 3460 ConstString const_symbol_name(symbol_name); 3461 sym[sym_idx].GetMangled().SetValue( 3462 const_symbol_name, symbol_name_is_mangled); 3463 if (is_gsym && is_debug) { 3464 const char *gsym_name = 3465 sym[sym_idx] 3466 .GetMangled() 3467 .GetName(Mangled::ePreferMangled) 3468 .GetCString(); 3469 if (gsym_name) 3470 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3471 } 3472 } 3473 } 3474 if (symbol_section) { 3475 const addr_t section_file_addr = 3476 symbol_section->GetFileAddress(); 3477 if (symbol_byte_size == 0 && 3478 function_starts_count > 0) { 3479 addr_t symbol_lookup_file_addr = nlist.n_value; 3480 // Do an exact address match for non-ARM addresses, 3481 // else get the closest since the symbol might be a 3482 // thumb symbol which has an address with bit zero 3483 // set 3484 FunctionStarts::Entry *func_start_entry = 3485 function_starts.FindEntry(symbol_lookup_file_addr, 3486 !is_arm); 3487 if (is_arm && func_start_entry) { 3488 // Verify that the function start address is the 3489 // symbol address (ARM) or the symbol address + 1 3490 // (thumb) 3491 if (func_start_entry->addr != 3492 symbol_lookup_file_addr && 3493 func_start_entry->addr != 3494 (symbol_lookup_file_addr + 1)) { 3495 // Not the right entry, NULL it out... 3496 func_start_entry = NULL; 3497 } 3498 } 3499 if (func_start_entry) { 3500 func_start_entry->data = true; 3501 3502 addr_t symbol_file_addr = func_start_entry->addr; 3503 uint32_t symbol_flags = 0; 3504 if (is_arm) { 3505 if (symbol_file_addr & 1) 3506 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3507 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3508 } 3509 3510 const FunctionStarts::Entry *next_func_start_entry = 3511 function_starts.FindNextEntry(func_start_entry); 3512 const addr_t section_end_file_addr = 3513 section_file_addr + 3514 symbol_section->GetByteSize(); 3515 if (next_func_start_entry) { 3516 addr_t next_symbol_file_addr = 3517 next_func_start_entry->addr; 3518 // Be sure the clear the Thumb address bit when 3519 // we calculate the size from the current and 3520 // next address 3521 if (is_arm) 3522 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3523 symbol_byte_size = std::min<lldb::addr_t>( 3524 next_symbol_file_addr - symbol_file_addr, 3525 section_end_file_addr - symbol_file_addr); 3526 } else { 3527 symbol_byte_size = 3528 section_end_file_addr - symbol_file_addr; 3529 } 3530 } 3531 } 3532 symbol_value -= section_file_addr; 3533 } 3534 3535 if (is_debug == false) { 3536 if (type == eSymbolTypeCode) { 3537 // See if we can find a N_FUN entry for any code 3538 // symbols. If we do find a match, and the name 3539 // matches, then we can merge the two into just the 3540 // function symbol to avoid duplicate entries in 3541 // the symbol table 3542 auto range = 3543 N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3544 if (range.first != range.second) { 3545 bool found_it = false; 3546 for (auto pos = range.first; pos != range.second; 3547 ++pos) { 3548 if (sym[sym_idx].GetMangled().GetName( 3549 Mangled::ePreferMangled) == 3550 sym[pos->second].GetMangled().GetName( 3551 Mangled::ePreferMangled)) { 3552 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3553 // We just need the flags from the linker 3554 // symbol, so put these flags 3555 // into the N_FUN flags to avoid duplicate 3556 // symbols in the symbol table 3557 sym[pos->second].SetExternal( 3558 sym[sym_idx].IsExternal()); 3559 sym[pos->second].SetFlags(nlist.n_type << 16 | 3560 nlist.n_desc); 3561 if (resolver_addresses.find(nlist.n_value) != 3562 resolver_addresses.end()) 3563 sym[pos->second].SetType(eSymbolTypeResolver); 3564 sym[sym_idx].Clear(); 3565 found_it = true; 3566 break; 3567 } 3568 } 3569 if (found_it) 3570 continue; 3571 } else { 3572 if (resolver_addresses.find(nlist.n_value) != 3573 resolver_addresses.end()) 3574 type = eSymbolTypeResolver; 3575 } 3576 } else if (type == eSymbolTypeData || 3577 type == eSymbolTypeObjCClass || 3578 type == eSymbolTypeObjCMetaClass || 3579 type == eSymbolTypeObjCIVar) { 3580 // See if we can find a N_STSYM entry for any data 3581 // symbols. If we do find a match, and the name 3582 // matches, then we can merge the two into just the 3583 // Static symbol to avoid duplicate entries in the 3584 // symbol table 3585 auto range = N_STSYM_addr_to_sym_idx.equal_range( 3586 nlist.n_value); 3587 if (range.first != range.second) { 3588 bool found_it = false; 3589 for (auto pos = range.first; pos != range.second; 3590 ++pos) { 3591 if (sym[sym_idx].GetMangled().GetName( 3592 Mangled::ePreferMangled) == 3593 sym[pos->second].GetMangled().GetName( 3594 Mangled::ePreferMangled)) { 3595 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3596 // We just need the flags from the linker 3597 // symbol, so put these flags 3598 // into the N_STSYM flags to avoid duplicate 3599 // symbols in the symbol table 3600 sym[pos->second].SetExternal( 3601 sym[sym_idx].IsExternal()); 3602 sym[pos->second].SetFlags(nlist.n_type << 16 | 3603 nlist.n_desc); 3604 sym[sym_idx].Clear(); 3605 found_it = true; 3606 break; 3607 } 3608 } 3609 if (found_it) 3610 continue; 3611 } else { 3612 const char *gsym_name = 3613 sym[sym_idx] 3614 .GetMangled() 3615 .GetName(Mangled::ePreferMangled) 3616 .GetCString(); 3617 if (gsym_name) { 3618 // Combine N_GSYM stab entries with the non 3619 // stab symbol 3620 ConstNameToSymbolIndexMap::const_iterator pos = 3621 N_GSYM_name_to_sym_idx.find(gsym_name); 3622 if (pos != N_GSYM_name_to_sym_idx.end()) { 3623 const uint32_t GSYM_sym_idx = pos->second; 3624 m_nlist_idx_to_sym_idx[nlist_idx] = 3625 GSYM_sym_idx; 3626 // Copy the address, because often the N_GSYM 3627 // address has an invalid address of zero 3628 // when the global is a common symbol 3629 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3630 symbol_section); 3631 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3632 symbol_value); 3633 add_symbol_addr(sym[GSYM_sym_idx] 3634 .GetAddress() 3635 .GetFileAddress()); 3636 // We just need the flags from the linker 3637 // symbol, so put these flags 3638 // into the N_GSYM flags to avoid duplicate 3639 // symbols in the symbol table 3640 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 3641 nlist.n_desc); 3642 sym[sym_idx].Clear(); 3643 continue; 3644 } 3645 } 3646 } 3647 } 3648 } 3649 3650 sym[sym_idx].SetID(nlist_idx); 3651 sym[sym_idx].SetType(type); 3652 if (set_value) { 3653 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 3654 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3655 add_symbol_addr( 3656 sym[sym_idx].GetAddress().GetFileAddress()); 3657 } 3658 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 3659 3660 if (symbol_byte_size > 0) 3661 sym[sym_idx].SetByteSize(symbol_byte_size); 3662 3663 if (demangled_is_synthesized) 3664 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3665 ++sym_idx; 3666 } else { 3667 sym[sym_idx].Clear(); 3668 } 3669 } 3670 ///////////////////////////// 3671 } 3672 } 3673 3674 for (const auto &pos : reexport_shlib_needs_fixup) { 3675 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3676 if (undef_pos != undefined_name_to_desc.end()) { 3677 const uint8_t dylib_ordinal = 3678 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3679 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3680 sym[pos.first].SetReExportedSymbolSharedLibrary( 3681 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3682 } 3683 } 3684 } 3685 3686 #endif 3687 lldb::offset_t nlist_data_offset = 0; 3688 3689 if (nlist_data.GetByteSize() > 0) { 3690 3691 // If the sym array was not created while parsing the DSC unmapped 3692 // symbols, create it now. 3693 if (sym == nullptr) { 3694 sym = 3695 symtab.Resize(symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3696 num_syms = symtab.GetNumSymbols(); 3697 } 3698 3699 if (unmapped_local_symbols_found) { 3700 assert(m_dysymtab.ilocalsym == 0); 3701 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3702 nlist_idx = m_dysymtab.nlocalsym; 3703 } else { 3704 nlist_idx = 0; 3705 } 3706 3707 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 3708 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 3709 UndefinedNameToDescMap undefined_name_to_desc; 3710 SymbolIndexToName reexport_shlib_needs_fixup; 3711 3712 // Symtab parsing is a huge mess. Everything is entangled and the code 3713 // requires access to a ridiculous amount of variables. LLDB depends 3714 // heavily on the proper merging of symbols and to get that right we need 3715 // to make sure we have parsed all the debug symbols first. Therefore we 3716 // invoke the lambda twice, once to parse only the debug symbols and then 3717 // once more to parse the remaining symbols. 3718 auto ParseSymbolLambda = [&](struct nlist_64 &nlist, uint32_t nlist_idx, 3719 bool debug_only) { 3720 const bool is_debug = ((nlist.n_type & N_STAB) != 0); 3721 if (is_debug != debug_only) 3722 return true; 3723 3724 const char *symbol_name_non_abi_mangled = nullptr; 3725 const char *symbol_name = nullptr; 3726 3727 if (have_strtab_data) { 3728 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3729 3730 if (symbol_name == nullptr) { 3731 // No symbol should be NULL, even the symbols with no string values 3732 // should have an offset zero which points to an empty C-string 3733 Host::SystemLog(Host::eSystemLogError, 3734 "error: symbol[%u] has invalid string table offset " 3735 "0x%x in %s, ignoring symbol\n", 3736 nlist_idx, nlist.n_strx, 3737 module_sp->GetFileSpec().GetPath().c_str()); 3738 return true; 3739 } 3740 if (symbol_name[0] == '\0') 3741 symbol_name = nullptr; 3742 } else { 3743 const addr_t str_addr = strtab_addr + nlist.n_strx; 3744 Status str_error; 3745 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3746 str_error)) 3747 symbol_name = memory_symbol_name.c_str(); 3748 } 3749 3750 SymbolType type = eSymbolTypeInvalid; 3751 SectionSP symbol_section; 3752 lldb::addr_t symbol_byte_size = 0; 3753 bool add_nlist = true; 3754 bool is_gsym = false; 3755 bool demangled_is_synthesized = false; 3756 bool set_value = true; 3757 3758 assert(sym_idx < num_syms); 3759 sym[sym_idx].SetDebug(is_debug); 3760 3761 if (is_debug) { 3762 switch (nlist.n_type) { 3763 case N_GSYM: 3764 // global symbol: name,,NO_SECT,type,0 3765 // Sometimes the N_GSYM value contains the address. 3766 3767 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3768 // the ObjC data. They 3769 // have the same address, but we want to ensure that we always find 3770 // only the real symbol, 'cause we don't currently correctly 3771 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3772 // type. This is a temporary hack to make sure the ObjectiveC 3773 // symbols get treated correctly. To do this right, we should 3774 // coalesce all the GSYM & global symbols that have the same 3775 // address. 3776 is_gsym = true; 3777 sym[sym_idx].SetExternal(true); 3778 3779 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3780 llvm::StringRef symbol_name_ref(symbol_name); 3781 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3782 symbol_name_non_abi_mangled = symbol_name + 1; 3783 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3784 type = eSymbolTypeObjCClass; 3785 demangled_is_synthesized = true; 3786 3787 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 3788 symbol_name_non_abi_mangled = symbol_name + 1; 3789 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3790 type = eSymbolTypeObjCMetaClass; 3791 demangled_is_synthesized = true; 3792 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3793 symbol_name_non_abi_mangled = symbol_name + 1; 3794 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3795 type = eSymbolTypeObjCIVar; 3796 demangled_is_synthesized = true; 3797 } 3798 } else { 3799 if (nlist.n_value != 0) 3800 symbol_section = 3801 section_info.GetSection(nlist.n_sect, nlist.n_value); 3802 type = eSymbolTypeData; 3803 } 3804 break; 3805 3806 case N_FNAME: 3807 // procedure name (f77 kludge): name,,NO_SECT,0,0 3808 type = eSymbolTypeCompiler; 3809 break; 3810 3811 case N_FUN: 3812 // procedure: name,,n_sect,linenumber,address 3813 if (symbol_name) { 3814 type = eSymbolTypeCode; 3815 symbol_section = 3816 section_info.GetSection(nlist.n_sect, nlist.n_value); 3817 3818 N_FUN_addr_to_sym_idx.insert( 3819 std::make_pair(nlist.n_value, sym_idx)); 3820 // We use the current number of symbols in the symbol table in 3821 // lieu of using nlist_idx in case we ever start trimming entries 3822 // out 3823 N_FUN_indexes.push_back(sym_idx); 3824 } else { 3825 type = eSymbolTypeCompiler; 3826 3827 if (!N_FUN_indexes.empty()) { 3828 // Copy the size of the function into the original STAB entry 3829 // so we don't have to hunt for it later 3830 symtab.SymbolAtIndex(N_FUN_indexes.back()) 3831 ->SetByteSize(nlist.n_value); 3832 N_FUN_indexes.pop_back(); 3833 // We don't really need the end function STAB as it contains 3834 // the size which we already placed with the original symbol, 3835 // so don't add it if we want a minimal symbol table 3836 add_nlist = false; 3837 } 3838 } 3839 break; 3840 3841 case N_STSYM: 3842 // static symbol: name,,n_sect,type,address 3843 N_STSYM_addr_to_sym_idx.insert( 3844 std::make_pair(nlist.n_value, sym_idx)); 3845 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3846 if (symbol_name && symbol_name[0]) { 3847 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3848 eSymbolTypeData); 3849 } 3850 break; 3851 3852 case N_LCSYM: 3853 // .lcomm symbol: name,,n_sect,type,address 3854 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3855 type = eSymbolTypeCommonBlock; 3856 break; 3857 3858 case N_BNSYM: 3859 // We use the current number of symbols in the symbol table in lieu 3860 // of using nlist_idx in case we ever start trimming entries out 3861 // Skip these if we want minimal symbol tables 3862 add_nlist = false; 3863 break; 3864 3865 case N_ENSYM: 3866 // Set the size of the N_BNSYM to the terminating index of this 3867 // N_ENSYM so that we can always skip the entire symbol if we need 3868 // to navigate more quickly at the source level when parsing STABS 3869 // Skip these if we want minimal symbol tables 3870 add_nlist = false; 3871 break; 3872 3873 case N_OPT: 3874 // emitted with gcc2_compiled and in gcc source 3875 type = eSymbolTypeCompiler; 3876 break; 3877 3878 case N_RSYM: 3879 // register sym: name,,NO_SECT,type,register 3880 type = eSymbolTypeVariable; 3881 break; 3882 3883 case N_SLINE: 3884 // src line: 0,,n_sect,linenumber,address 3885 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3886 type = eSymbolTypeLineEntry; 3887 break; 3888 3889 case N_SSYM: 3890 // structure elt: name,,NO_SECT,type,struct_offset 3891 type = eSymbolTypeVariableType; 3892 break; 3893 3894 case N_SO: 3895 // source file name 3896 type = eSymbolTypeSourceFile; 3897 if (symbol_name == nullptr) { 3898 add_nlist = false; 3899 if (N_SO_index != UINT32_MAX) { 3900 // Set the size of the N_SO to the terminating index of this 3901 // N_SO so that we can always skip the entire N_SO if we need 3902 // to navigate more quickly at the source level when parsing 3903 // STABS 3904 symbol_ptr = symtab.SymbolAtIndex(N_SO_index); 3905 symbol_ptr->SetByteSize(sym_idx); 3906 symbol_ptr->SetSizeIsSibling(true); 3907 } 3908 N_NSYM_indexes.clear(); 3909 N_INCL_indexes.clear(); 3910 N_BRAC_indexes.clear(); 3911 N_COMM_indexes.clear(); 3912 N_FUN_indexes.clear(); 3913 N_SO_index = UINT32_MAX; 3914 } else { 3915 // We use the current number of symbols in the symbol table in 3916 // lieu of using nlist_idx in case we ever start trimming entries 3917 // out 3918 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3919 if (N_SO_has_full_path) { 3920 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3921 // We have two consecutive N_SO entries where the first 3922 // contains a directory and the second contains a full path. 3923 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), 3924 false); 3925 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3926 add_nlist = false; 3927 } else { 3928 // This is the first entry in a N_SO that contains a 3929 // directory or a full path to the source file 3930 N_SO_index = sym_idx; 3931 } 3932 } else if ((N_SO_index == sym_idx - 1) && 3933 ((sym_idx - 1) < num_syms)) { 3934 // This is usually the second N_SO entry that contains just the 3935 // filename, so here we combine it with the first one if we are 3936 // minimizing the symbol table 3937 const char *so_path = 3938 sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3939 if (so_path && so_path[0]) { 3940 std::string full_so_path(so_path); 3941 const size_t double_slash_pos = full_so_path.find("//"); 3942 if (double_slash_pos != std::string::npos) { 3943 // The linker has been generating bad N_SO entries with 3944 // doubled up paths in the format "%s%s" where the first 3945 // string in the DW_AT_comp_dir, and the second is the 3946 // directory for the source file so you end up with a path 3947 // that looks like "/tmp/src//tmp/src/" 3948 FileSpec so_dir(so_path); 3949 if (!FileSystem::Instance().Exists(so_dir)) { 3950 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 3951 FileSpec::Style::native); 3952 if (FileSystem::Instance().Exists(so_dir)) { 3953 // Trim off the incorrect path 3954 full_so_path.erase(0, double_slash_pos + 1); 3955 } 3956 } 3957 } 3958 if (*full_so_path.rbegin() != '/') 3959 full_so_path += '/'; 3960 full_so_path += symbol_name; 3961 sym[sym_idx - 1].GetMangled().SetValue( 3962 ConstString(full_so_path.c_str()), false); 3963 add_nlist = false; 3964 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3965 } 3966 } else { 3967 // This could be a relative path to a N_SO 3968 N_SO_index = sym_idx; 3969 } 3970 } 3971 break; 3972 3973 case N_OSO: 3974 // object file name: name,,0,0,st_mtime 3975 type = eSymbolTypeObjectFile; 3976 break; 3977 3978 case N_LSYM: 3979 // local sym: name,,NO_SECT,type,offset 3980 type = eSymbolTypeLocal; 3981 break; 3982 3983 // INCL scopes 3984 case N_BINCL: 3985 // include file beginning: name,,NO_SECT,0,sum We use the current 3986 // number of symbols in the symbol table in lieu of using nlist_idx 3987 // in case we ever start trimming entries out 3988 N_INCL_indexes.push_back(sym_idx); 3989 type = eSymbolTypeScopeBegin; 3990 break; 3991 3992 case N_EINCL: 3993 // include file end: name,,NO_SECT,0,0 3994 // Set the size of the N_BINCL to the terminating index of this 3995 // N_EINCL so that we can always skip the entire symbol if we need 3996 // to navigate more quickly at the source level when parsing STABS 3997 if (!N_INCL_indexes.empty()) { 3998 symbol_ptr = symtab.SymbolAtIndex(N_INCL_indexes.back()); 3999 symbol_ptr->SetByteSize(sym_idx + 1); 4000 symbol_ptr->SetSizeIsSibling(true); 4001 N_INCL_indexes.pop_back(); 4002 } 4003 type = eSymbolTypeScopeEnd; 4004 break; 4005 4006 case N_SOL: 4007 // #included file name: name,,n_sect,0,address 4008 type = eSymbolTypeHeaderFile; 4009 4010 // We currently don't use the header files on darwin 4011 add_nlist = false; 4012 break; 4013 4014 case N_PARAMS: 4015 // compiler parameters: name,,NO_SECT,0,0 4016 type = eSymbolTypeCompiler; 4017 break; 4018 4019 case N_VERSION: 4020 // compiler version: name,,NO_SECT,0,0 4021 type = eSymbolTypeCompiler; 4022 break; 4023 4024 case N_OLEVEL: 4025 // compiler -O level: name,,NO_SECT,0,0 4026 type = eSymbolTypeCompiler; 4027 break; 4028 4029 case N_PSYM: 4030 // parameter: name,,NO_SECT,type,offset 4031 type = eSymbolTypeVariable; 4032 break; 4033 4034 case N_ENTRY: 4035 // alternate entry: name,,n_sect,linenumber,address 4036 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4037 type = eSymbolTypeLineEntry; 4038 break; 4039 4040 // Left and Right Braces 4041 case N_LBRAC: 4042 // left bracket: 0,,NO_SECT,nesting level,address We use the 4043 // current number of symbols in the symbol table in lieu of using 4044 // nlist_idx in case we ever start trimming entries out 4045 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4046 N_BRAC_indexes.push_back(sym_idx); 4047 type = eSymbolTypeScopeBegin; 4048 break; 4049 4050 case N_RBRAC: 4051 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4052 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4053 // can always skip the entire symbol if we need to navigate more 4054 // quickly at the source level when parsing STABS 4055 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4056 if (!N_BRAC_indexes.empty()) { 4057 symbol_ptr = symtab.SymbolAtIndex(N_BRAC_indexes.back()); 4058 symbol_ptr->SetByteSize(sym_idx + 1); 4059 symbol_ptr->SetSizeIsSibling(true); 4060 N_BRAC_indexes.pop_back(); 4061 } 4062 type = eSymbolTypeScopeEnd; 4063 break; 4064 4065 case N_EXCL: 4066 // deleted include file: name,,NO_SECT,0,sum 4067 type = eSymbolTypeHeaderFile; 4068 break; 4069 4070 // COMM scopes 4071 case N_BCOMM: 4072 // begin common: name,,NO_SECT,0,0 4073 // We use the current number of symbols in the symbol table in lieu 4074 // of using nlist_idx in case we ever start trimming entries out 4075 type = eSymbolTypeScopeBegin; 4076 N_COMM_indexes.push_back(sym_idx); 4077 break; 4078 4079 case N_ECOML: 4080 // end common (local name): 0,,n_sect,0,address 4081 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4082 LLVM_FALLTHROUGH; 4083 4084 case N_ECOMM: 4085 // end common: name,,n_sect,0,0 4086 // Set the size of the N_BCOMM to the terminating index of this 4087 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4088 // we need to navigate more quickly at the source level when 4089 // parsing STABS 4090 if (!N_COMM_indexes.empty()) { 4091 symbol_ptr = symtab.SymbolAtIndex(N_COMM_indexes.back()); 4092 symbol_ptr->SetByteSize(sym_idx + 1); 4093 symbol_ptr->SetSizeIsSibling(true); 4094 N_COMM_indexes.pop_back(); 4095 } 4096 type = eSymbolTypeScopeEnd; 4097 break; 4098 4099 case N_LENG: 4100 // second stab entry with length information 4101 type = eSymbolTypeAdditional; 4102 break; 4103 4104 default: 4105 break; 4106 } 4107 } else { 4108 uint8_t n_type = N_TYPE & nlist.n_type; 4109 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4110 4111 switch (n_type) { 4112 case N_INDR: { 4113 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4114 if (reexport_name_cstr && reexport_name_cstr[0]) { 4115 type = eSymbolTypeReExported; 4116 ConstString reexport_name(reexport_name_cstr + 4117 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4118 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4119 set_value = false; 4120 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4121 indirect_symbol_names.insert( 4122 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4123 } else 4124 type = eSymbolTypeUndefined; 4125 } break; 4126 4127 case N_UNDF: 4128 if (symbol_name && symbol_name[0]) { 4129 ConstString undefined_name(symbol_name + 4130 ((symbol_name[0] == '_') ? 1 : 0)); 4131 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4132 } 4133 LLVM_FALLTHROUGH; 4134 4135 case N_PBUD: 4136 type = eSymbolTypeUndefined; 4137 break; 4138 4139 case N_ABS: 4140 type = eSymbolTypeAbsolute; 4141 break; 4142 4143 case N_SECT: { 4144 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4145 4146 if (!symbol_section) { 4147 // TODO: warn about this? 4148 add_nlist = false; 4149 break; 4150 } 4151 4152 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4153 type = eSymbolTypeException; 4154 } else { 4155 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4156 4157 switch (section_type) { 4158 case S_CSTRING_LITERALS: 4159 type = eSymbolTypeData; 4160 break; // section with only literal C strings 4161 case S_4BYTE_LITERALS: 4162 type = eSymbolTypeData; 4163 break; // section with only 4 byte literals 4164 case S_8BYTE_LITERALS: 4165 type = eSymbolTypeData; 4166 break; // section with only 8 byte literals 4167 case S_LITERAL_POINTERS: 4168 type = eSymbolTypeTrampoline; 4169 break; // section with only pointers to literals 4170 case S_NON_LAZY_SYMBOL_POINTERS: 4171 type = eSymbolTypeTrampoline; 4172 break; // section with only non-lazy symbol pointers 4173 case S_LAZY_SYMBOL_POINTERS: 4174 type = eSymbolTypeTrampoline; 4175 break; // section with only lazy symbol pointers 4176 case S_SYMBOL_STUBS: 4177 type = eSymbolTypeTrampoline; 4178 break; // section with only symbol stubs, byte size of stub in 4179 // the reserved2 field 4180 case S_MOD_INIT_FUNC_POINTERS: 4181 type = eSymbolTypeCode; 4182 break; // section with only function pointers for initialization 4183 case S_MOD_TERM_FUNC_POINTERS: 4184 type = eSymbolTypeCode; 4185 break; // section with only function pointers for termination 4186 case S_INTERPOSING: 4187 type = eSymbolTypeTrampoline; 4188 break; // section with only pairs of function pointers for 4189 // interposing 4190 case S_16BYTE_LITERALS: 4191 type = eSymbolTypeData; 4192 break; // section with only 16 byte literals 4193 case S_DTRACE_DOF: 4194 type = eSymbolTypeInstrumentation; 4195 break; 4196 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4197 type = eSymbolTypeTrampoline; 4198 break; 4199 default: 4200 switch (symbol_section->GetType()) { 4201 case lldb::eSectionTypeCode: 4202 type = eSymbolTypeCode; 4203 break; 4204 case eSectionTypeData: 4205 case eSectionTypeDataCString: // Inlined C string data 4206 case eSectionTypeDataCStringPointers: // Pointers to C string 4207 // data 4208 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4209 // the symbol table 4210 case eSectionTypeData4: 4211 case eSectionTypeData8: 4212 case eSectionTypeData16: 4213 type = eSymbolTypeData; 4214 break; 4215 default: 4216 break; 4217 } 4218 break; 4219 } 4220 4221 if (type == eSymbolTypeInvalid) { 4222 const char *symbol_sect_name = 4223 symbol_section->GetName().AsCString(); 4224 if (symbol_section->IsDescendant(text_section_sp.get())) { 4225 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4226 S_ATTR_SELF_MODIFYING_CODE | 4227 S_ATTR_SOME_INSTRUCTIONS)) 4228 type = eSymbolTypeData; 4229 else 4230 type = eSymbolTypeCode; 4231 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 4232 symbol_section->IsDescendant( 4233 data_dirty_section_sp.get()) || 4234 symbol_section->IsDescendant( 4235 data_const_section_sp.get())) { 4236 if (symbol_sect_name && 4237 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 4238 type = eSymbolTypeRuntime; 4239 4240 if (symbol_name) { 4241 llvm::StringRef symbol_name_ref(symbol_name); 4242 if (symbol_name_ref.startswith("_OBJC_")) { 4243 llvm::StringRef g_objc_v2_prefix_class( 4244 "_OBJC_CLASS_$_"); 4245 llvm::StringRef g_objc_v2_prefix_metaclass( 4246 "_OBJC_METACLASS_$_"); 4247 llvm::StringRef g_objc_v2_prefix_ivar( 4248 "_OBJC_IVAR_$_"); 4249 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 4250 symbol_name_non_abi_mangled = symbol_name + 1; 4251 symbol_name = 4252 symbol_name + g_objc_v2_prefix_class.size(); 4253 type = eSymbolTypeObjCClass; 4254 demangled_is_synthesized = true; 4255 } else if (symbol_name_ref.startswith( 4256 g_objc_v2_prefix_metaclass)) { 4257 symbol_name_non_abi_mangled = symbol_name + 1; 4258 symbol_name = 4259 symbol_name + g_objc_v2_prefix_metaclass.size(); 4260 type = eSymbolTypeObjCMetaClass; 4261 demangled_is_synthesized = true; 4262 } else if (symbol_name_ref.startswith( 4263 g_objc_v2_prefix_ivar)) { 4264 symbol_name_non_abi_mangled = symbol_name + 1; 4265 symbol_name = 4266 symbol_name + g_objc_v2_prefix_ivar.size(); 4267 type = eSymbolTypeObjCIVar; 4268 demangled_is_synthesized = true; 4269 } 4270 } 4271 } 4272 } else if (symbol_sect_name && 4273 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4274 symbol_sect_name) { 4275 type = eSymbolTypeException; 4276 } else { 4277 type = eSymbolTypeData; 4278 } 4279 } else if (symbol_sect_name && 4280 ::strstr(symbol_sect_name, "__IMPORT") == 4281 symbol_sect_name) { 4282 type = eSymbolTypeTrampoline; 4283 } else if (symbol_section->IsDescendant(objc_section_sp.get())) { 4284 type = eSymbolTypeRuntime; 4285 if (symbol_name && symbol_name[0] == '.') { 4286 llvm::StringRef symbol_name_ref(symbol_name); 4287 llvm::StringRef g_objc_v1_prefix_class( 4288 ".objc_class_name_"); 4289 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4290 symbol_name_non_abi_mangled = symbol_name; 4291 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4292 type = eSymbolTypeObjCClass; 4293 demangled_is_synthesized = true; 4294 } 4295 } 4296 } 4297 } 4298 } 4299 } break; 4300 } 4301 } 4302 4303 if (!add_nlist) { 4304 sym[sym_idx].Clear(); 4305 return true; 4306 } 4307 4308 uint64_t symbol_value = nlist.n_value; 4309 4310 if (symbol_name_non_abi_mangled) { 4311 sym[sym_idx].GetMangled().SetMangledName( 4312 ConstString(symbol_name_non_abi_mangled)); 4313 sym[sym_idx].GetMangled().SetDemangledName(ConstString(symbol_name)); 4314 } else { 4315 bool symbol_name_is_mangled = false; 4316 4317 if (symbol_name && symbol_name[0] == '_') { 4318 symbol_name_is_mangled = symbol_name[1] == '_'; 4319 symbol_name++; // Skip the leading underscore 4320 } 4321 4322 if (symbol_name) { 4323 ConstString const_symbol_name(symbol_name); 4324 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4325 symbol_name_is_mangled); 4326 } 4327 } 4328 4329 if (is_gsym) { 4330 const char *gsym_name = sym[sym_idx] 4331 .GetMangled() 4332 .GetName(Mangled::ePreferMangled) 4333 .GetCString(); 4334 if (gsym_name) 4335 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4336 } 4337 4338 if (symbol_section) { 4339 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4340 if (symbol_byte_size == 0 && function_starts_count > 0) { 4341 addr_t symbol_lookup_file_addr = nlist.n_value; 4342 // Do an exact address match for non-ARM addresses, else get the 4343 // closest since the symbol might be a thumb symbol which has an 4344 // address with bit zero set. 4345 FunctionStarts::Entry *func_start_entry = 4346 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4347 if (is_arm && func_start_entry) { 4348 // Verify that the function start address is the symbol address 4349 // (ARM) or the symbol address + 1 (thumb). 4350 if (func_start_entry->addr != symbol_lookup_file_addr && 4351 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4352 // Not the right entry, NULL it out... 4353 func_start_entry = nullptr; 4354 } 4355 } 4356 if (func_start_entry) { 4357 func_start_entry->data = true; 4358 4359 addr_t symbol_file_addr = func_start_entry->addr; 4360 if (is_arm) 4361 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4362 4363 const FunctionStarts::Entry *next_func_start_entry = 4364 function_starts.FindNextEntry(func_start_entry); 4365 const addr_t section_end_file_addr = 4366 section_file_addr + symbol_section->GetByteSize(); 4367 if (next_func_start_entry) { 4368 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4369 // Be sure the clear the Thumb address bit when we calculate the 4370 // size from the current and next address 4371 if (is_arm) 4372 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4373 symbol_byte_size = std::min<lldb::addr_t>( 4374 next_symbol_file_addr - symbol_file_addr, 4375 section_end_file_addr - symbol_file_addr); 4376 } else { 4377 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4378 } 4379 } 4380 } 4381 symbol_value -= section_file_addr; 4382 } 4383 4384 if (!is_debug) { 4385 if (type == eSymbolTypeCode) { 4386 // See if we can find a N_FUN entry for any code symbols. If we do 4387 // find a match, and the name matches, then we can merge the two into 4388 // just the function symbol to avoid duplicate entries in the symbol 4389 // table. 4390 std::pair<ValueToSymbolIndexMap::const_iterator, 4391 ValueToSymbolIndexMap::const_iterator> 4392 range; 4393 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4394 if (range.first != range.second) { 4395 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4396 pos != range.second; ++pos) { 4397 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4398 sym[pos->second].GetMangled().GetName( 4399 Mangled::ePreferMangled)) { 4400 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4401 // We just need the flags from the linker symbol, so put these 4402 // flags into the N_FUN flags to avoid duplicate symbols in the 4403 // symbol table. 4404 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4405 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4406 if (resolver_addresses.find(nlist.n_value) != 4407 resolver_addresses.end()) 4408 sym[pos->second].SetType(eSymbolTypeResolver); 4409 sym[sym_idx].Clear(); 4410 return true; 4411 } 4412 } 4413 } else { 4414 if (resolver_addresses.find(nlist.n_value) != 4415 resolver_addresses.end()) 4416 type = eSymbolTypeResolver; 4417 } 4418 } else if (type == eSymbolTypeData || type == eSymbolTypeObjCClass || 4419 type == eSymbolTypeObjCMetaClass || 4420 type == eSymbolTypeObjCIVar) { 4421 // See if we can find a N_STSYM entry for any data symbols. If we do 4422 // find a match, and the name matches, then we can merge the two into 4423 // just the Static symbol to avoid duplicate entries in the symbol 4424 // table. 4425 std::pair<ValueToSymbolIndexMap::const_iterator, 4426 ValueToSymbolIndexMap::const_iterator> 4427 range; 4428 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4429 if (range.first != range.second) { 4430 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4431 pos != range.second; ++pos) { 4432 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4433 sym[pos->second].GetMangled().GetName( 4434 Mangled::ePreferMangled)) { 4435 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4436 // We just need the flags from the linker symbol, so put these 4437 // flags into the N_STSYM flags to avoid duplicate symbols in 4438 // the symbol table. 4439 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4440 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4441 sym[sym_idx].Clear(); 4442 return true; 4443 } 4444 } 4445 } else { 4446 // Combine N_GSYM stab entries with the non stab symbol. 4447 const char *gsym_name = sym[sym_idx] 4448 .GetMangled() 4449 .GetName(Mangled::ePreferMangled) 4450 .GetCString(); 4451 if (gsym_name) { 4452 ConstNameToSymbolIndexMap::const_iterator pos = 4453 N_GSYM_name_to_sym_idx.find(gsym_name); 4454 if (pos != N_GSYM_name_to_sym_idx.end()) { 4455 const uint32_t GSYM_sym_idx = pos->second; 4456 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4457 // Copy the address, because often the N_GSYM address has an 4458 // invalid address of zero when the global is a common symbol. 4459 sym[GSYM_sym_idx].GetAddressRef().SetSection(symbol_section); 4460 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4461 add_symbol_addr( 4462 sym[GSYM_sym_idx].GetAddress().GetFileAddress()); 4463 // We just need the flags from the linker symbol, so put these 4464 // flags into the N_GSYM flags to avoid duplicate symbols in 4465 // the symbol table. 4466 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4467 sym[sym_idx].Clear(); 4468 return true; 4469 } 4470 } 4471 } 4472 } 4473 } 4474 4475 sym[sym_idx].SetID(nlist_idx); 4476 sym[sym_idx].SetType(type); 4477 if (set_value) { 4478 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4479 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4480 if (symbol_section) 4481 add_symbol_addr(sym[sym_idx].GetAddress().GetFileAddress()); 4482 } 4483 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4484 if (nlist.n_desc & N_WEAK_REF) 4485 sym[sym_idx].SetIsWeak(true); 4486 4487 if (symbol_byte_size > 0) 4488 sym[sym_idx].SetByteSize(symbol_byte_size); 4489 4490 if (demangled_is_synthesized) 4491 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4492 4493 ++sym_idx; 4494 return true; 4495 }; 4496 4497 // First parse all the nlists but don't process them yet. See the next 4498 // comment for an explanation why. 4499 std::vector<struct nlist_64> nlists; 4500 nlists.reserve(symtab_load_command.nsyms); 4501 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 4502 if (auto nlist = 4503 ParseNList(nlist_data, nlist_data_offset, nlist_byte_size)) 4504 nlists.push_back(*nlist); 4505 else 4506 break; 4507 } 4508 4509 // Now parse all the debug symbols. This is needed to merge non-debug 4510 // symbols in the next step. Non-debug symbols are always coalesced into 4511 // the debug symbol. Doing this in one step would mean that some symbols 4512 // won't be merged. 4513 nlist_idx = 0; 4514 for (auto &nlist : nlists) { 4515 if (!ParseSymbolLambda(nlist, nlist_idx++, DebugSymbols)) 4516 break; 4517 } 4518 4519 // Finally parse all the non debug symbols. 4520 nlist_idx = 0; 4521 for (auto &nlist : nlists) { 4522 if (!ParseSymbolLambda(nlist, nlist_idx++, NonDebugSymbols)) 4523 break; 4524 } 4525 4526 for (const auto &pos : reexport_shlib_needs_fixup) { 4527 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4528 if (undef_pos != undefined_name_to_desc.end()) { 4529 const uint8_t dylib_ordinal = 4530 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4531 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4532 sym[pos.first].SetReExportedSymbolSharedLibrary( 4533 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4534 } 4535 } 4536 } 4537 4538 // Count how many trie symbols we'll add to the symbol table 4539 int trie_symbol_table_augment_count = 0; 4540 for (auto &e : external_sym_trie_entries) { 4541 if (symbols_added.find(e.entry.address) == symbols_added.end()) 4542 trie_symbol_table_augment_count++; 4543 } 4544 4545 if (num_syms < sym_idx + trie_symbol_table_augment_count) { 4546 num_syms = sym_idx + trie_symbol_table_augment_count; 4547 sym = symtab.Resize(num_syms); 4548 } 4549 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4550 4551 // Add symbols from the trie to the symbol table. 4552 for (auto &e : external_sym_trie_entries) { 4553 if (symbols_added.contains(e.entry.address)) 4554 continue; 4555 4556 // Find the section that this trie address is in, use that to annotate 4557 // symbol type as we add the trie address and name to the symbol table. 4558 Address symbol_addr; 4559 if (module_sp->ResolveFileAddress(e.entry.address, symbol_addr)) { 4560 SectionSP symbol_section(symbol_addr.GetSection()); 4561 const char *symbol_name = e.entry.name.GetCString(); 4562 bool demangled_is_synthesized = false; 4563 SymbolType type = 4564 GetSymbolType(symbol_name, demangled_is_synthesized, text_section_sp, 4565 data_section_sp, data_dirty_section_sp, 4566 data_const_section_sp, symbol_section); 4567 4568 sym[sym_idx].SetType(type); 4569 if (symbol_section) { 4570 sym[sym_idx].SetID(synthetic_sym_id++); 4571 sym[sym_idx].GetMangled().SetMangledName(ConstString(symbol_name)); 4572 if (demangled_is_synthesized) 4573 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4574 sym[sym_idx].SetIsSynthetic(true); 4575 sym[sym_idx].SetExternal(true); 4576 sym[sym_idx].GetAddressRef() = symbol_addr; 4577 add_symbol_addr(symbol_addr.GetFileAddress()); 4578 if (e.entry.flags & TRIE_SYMBOL_IS_THUMB) 4579 sym[sym_idx].SetFlags(MACHO_NLIST_ARM_SYMBOL_IS_THUMB); 4580 ++sym_idx; 4581 } 4582 } 4583 } 4584 4585 if (function_starts_count > 0) { 4586 uint32_t num_synthetic_function_symbols = 0; 4587 for (i = 0; i < function_starts_count; ++i) { 4588 if (symbols_added.find(function_starts.GetEntryRef(i).addr) == 4589 symbols_added.end()) 4590 ++num_synthetic_function_symbols; 4591 } 4592 4593 if (num_synthetic_function_symbols > 0) { 4594 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4595 num_syms = sym_idx + num_synthetic_function_symbols; 4596 sym = symtab.Resize(num_syms); 4597 } 4598 for (i = 0; i < function_starts_count; ++i) { 4599 const FunctionStarts::Entry *func_start_entry = 4600 function_starts.GetEntryAtIndex(i); 4601 if (symbols_added.find(func_start_entry->addr) == symbols_added.end()) { 4602 addr_t symbol_file_addr = func_start_entry->addr; 4603 uint32_t symbol_flags = 0; 4604 if (func_start_entry->data) 4605 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4606 Address symbol_addr; 4607 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4608 SectionSP symbol_section(symbol_addr.GetSection()); 4609 uint32_t symbol_byte_size = 0; 4610 if (symbol_section) { 4611 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4612 const FunctionStarts::Entry *next_func_start_entry = 4613 function_starts.FindNextEntry(func_start_entry); 4614 const addr_t section_end_file_addr = 4615 section_file_addr + symbol_section->GetByteSize(); 4616 if (next_func_start_entry) { 4617 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4618 if (is_arm) 4619 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4620 symbol_byte_size = std::min<lldb::addr_t>( 4621 next_symbol_file_addr - symbol_file_addr, 4622 section_end_file_addr - symbol_file_addr); 4623 } else { 4624 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4625 } 4626 sym[sym_idx].SetID(synthetic_sym_id++); 4627 // Don't set the name for any synthetic symbols, the Symbol 4628 // object will generate one if needed when the name is accessed 4629 // via accessors. 4630 sym[sym_idx].GetMangled().SetDemangledName(ConstString()); 4631 sym[sym_idx].SetType(eSymbolTypeCode); 4632 sym[sym_idx].SetIsSynthetic(true); 4633 sym[sym_idx].GetAddressRef() = symbol_addr; 4634 add_symbol_addr(symbol_addr.GetFileAddress()); 4635 if (symbol_flags) 4636 sym[sym_idx].SetFlags(symbol_flags); 4637 if (symbol_byte_size) 4638 sym[sym_idx].SetByteSize(symbol_byte_size); 4639 ++sym_idx; 4640 } 4641 } 4642 } 4643 } 4644 } 4645 } 4646 4647 // Trim our symbols down to just what we ended up with after removing any 4648 // symbols. 4649 if (sym_idx < num_syms) { 4650 num_syms = sym_idx; 4651 sym = symtab.Resize(num_syms); 4652 } 4653 4654 // Now synthesize indirect symbols 4655 if (m_dysymtab.nindirectsyms != 0) { 4656 if (indirect_symbol_index_data.GetByteSize()) { 4657 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4658 m_nlist_idx_to_sym_idx.end(); 4659 4660 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4661 ++sect_idx) { 4662 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4663 S_SYMBOL_STUBS) { 4664 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4665 if (symbol_stub_byte_size == 0) 4666 continue; 4667 4668 const uint32_t num_symbol_stubs = 4669 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4670 4671 if (num_symbol_stubs == 0) 4672 continue; 4673 4674 const uint32_t symbol_stub_index_offset = 4675 m_mach_sections[sect_idx].reserved1; 4676 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) { 4677 const uint32_t symbol_stub_index = 4678 symbol_stub_index_offset + stub_idx; 4679 const lldb::addr_t symbol_stub_addr = 4680 m_mach_sections[sect_idx].addr + 4681 (stub_idx * symbol_stub_byte_size); 4682 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4683 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4684 symbol_stub_offset, 4)) { 4685 const uint32_t stub_sym_id = 4686 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4687 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4688 continue; 4689 4690 NListIndexToSymbolIndexMap::const_iterator index_pos = 4691 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4692 Symbol *stub_symbol = nullptr; 4693 if (index_pos != end_index_pos) { 4694 // We have a remapping from the original nlist index to a 4695 // current symbol index, so just look this up by index 4696 stub_symbol = symtab.SymbolAtIndex(index_pos->second); 4697 } else { 4698 // We need to lookup a symbol using the original nlist symbol 4699 // index since this index is coming from the S_SYMBOL_STUBS 4700 stub_symbol = symtab.FindSymbolByID(stub_sym_id); 4701 } 4702 4703 if (stub_symbol) { 4704 Address so_addr(symbol_stub_addr, section_list); 4705 4706 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4707 // Change the external symbol into a trampoline that makes 4708 // sense These symbols were N_UNDF N_EXT, and are useless 4709 // to us, so we can re-use them so we don't have to make up 4710 // a synthetic symbol for no good reason. 4711 if (resolver_addresses.find(symbol_stub_addr) == 4712 resolver_addresses.end()) 4713 stub_symbol->SetType(eSymbolTypeTrampoline); 4714 else 4715 stub_symbol->SetType(eSymbolTypeResolver); 4716 stub_symbol->SetExternal(false); 4717 stub_symbol->GetAddressRef() = so_addr; 4718 stub_symbol->SetByteSize(symbol_stub_byte_size); 4719 } else { 4720 // Make a synthetic symbol to describe the trampoline stub 4721 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4722 if (sym_idx >= num_syms) { 4723 sym = symtab.Resize(++num_syms); 4724 stub_symbol = nullptr; // this pointer no longer valid 4725 } 4726 sym[sym_idx].SetID(synthetic_sym_id++); 4727 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4728 if (resolver_addresses.find(symbol_stub_addr) == 4729 resolver_addresses.end()) 4730 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4731 else 4732 sym[sym_idx].SetType(eSymbolTypeResolver); 4733 sym[sym_idx].SetIsSynthetic(true); 4734 sym[sym_idx].GetAddressRef() = so_addr; 4735 add_symbol_addr(so_addr.GetFileAddress()); 4736 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4737 ++sym_idx; 4738 } 4739 } else { 4740 if (log) 4741 log->Warning("symbol stub referencing symbol table symbol " 4742 "%u that isn't in our minimal symbol table, " 4743 "fix this!!!", 4744 stub_sym_id); 4745 } 4746 } 4747 } 4748 } 4749 } 4750 } 4751 } 4752 4753 if (!reexport_trie_entries.empty()) { 4754 for (const auto &e : reexport_trie_entries) { 4755 if (e.entry.import_name) { 4756 // Only add indirect symbols from the Trie entries if we didn't have 4757 // a N_INDR nlist entry for this already 4758 if (indirect_symbol_names.find(e.entry.name) == 4759 indirect_symbol_names.end()) { 4760 // Make a synthetic symbol to describe re-exported symbol. 4761 if (sym_idx >= num_syms) 4762 sym = symtab.Resize(++num_syms); 4763 sym[sym_idx].SetID(synthetic_sym_id++); 4764 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4765 sym[sym_idx].SetType(eSymbolTypeReExported); 4766 sym[sym_idx].SetIsSynthetic(true); 4767 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4768 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4769 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4770 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4771 } 4772 ++sym_idx; 4773 } 4774 } 4775 } 4776 } 4777 } 4778 4779 void ObjectFileMachO::Dump(Stream *s) { 4780 ModuleSP module_sp(GetModule()); 4781 if (module_sp) { 4782 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4783 s->Printf("%p: ", static_cast<void *>(this)); 4784 s->Indent(); 4785 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4786 s->PutCString("ObjectFileMachO64"); 4787 else 4788 s->PutCString("ObjectFileMachO32"); 4789 4790 *s << ", file = '" << m_file; 4791 ModuleSpecList all_specs; 4792 ModuleSpec base_spec; 4793 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 4794 base_spec, all_specs); 4795 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 4796 *s << "', triple"; 4797 if (e) 4798 s->Printf("[%d]", i); 4799 *s << " = "; 4800 *s << all_specs.GetModuleSpecRefAtIndex(i) 4801 .GetArchitecture() 4802 .GetTriple() 4803 .getTriple(); 4804 } 4805 *s << "\n"; 4806 SectionList *sections = GetSectionList(); 4807 if (sections) 4808 sections->Dump(s->AsRawOstream(), s->GetIndentLevel(), nullptr, true, 4809 UINT32_MAX); 4810 4811 if (m_symtab_up) 4812 m_symtab_up->Dump(s, nullptr, eSortOrderNone); 4813 } 4814 } 4815 4816 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4817 const lldb_private::DataExtractor &data, 4818 lldb::offset_t lc_offset) { 4819 uint32_t i; 4820 llvm::MachO::uuid_command load_cmd; 4821 4822 lldb::offset_t offset = lc_offset; 4823 for (i = 0; i < header.ncmds; ++i) { 4824 const lldb::offset_t cmd_offset = offset; 4825 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4826 break; 4827 4828 if (load_cmd.cmd == LC_UUID) { 4829 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4830 4831 if (uuid_bytes) { 4832 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4833 // We pretend these object files have no UUID to prevent crashing. 4834 4835 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4836 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4837 0xbb, 0x14, 0xf0, 0x0d}; 4838 4839 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4840 return UUID(); 4841 4842 return UUID::fromOptionalData(uuid_bytes, 16); 4843 } 4844 return UUID(); 4845 } 4846 offset = cmd_offset + load_cmd.cmdsize; 4847 } 4848 return UUID(); 4849 } 4850 4851 static llvm::StringRef GetOSName(uint32_t cmd) { 4852 switch (cmd) { 4853 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4854 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4855 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4856 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4857 case llvm::MachO::LC_VERSION_MIN_TVOS: 4858 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4859 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4860 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4861 default: 4862 llvm_unreachable("unexpected LC_VERSION load command"); 4863 } 4864 } 4865 4866 namespace { 4867 struct OSEnv { 4868 llvm::StringRef os_type; 4869 llvm::StringRef environment; 4870 OSEnv(uint32_t cmd) { 4871 switch (cmd) { 4872 case llvm::MachO::PLATFORM_MACOS: 4873 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4874 return; 4875 case llvm::MachO::PLATFORM_IOS: 4876 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4877 return; 4878 case llvm::MachO::PLATFORM_TVOS: 4879 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4880 return; 4881 case llvm::MachO::PLATFORM_WATCHOS: 4882 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4883 return; 4884 // TODO: add BridgeOS & DriverKit once in llvm/lib/Support/Triple.cpp 4885 // NEED_BRIDGEOS_TRIPLE 4886 // case llvm::MachO::PLATFORM_BRIDGEOS: 4887 // os_type = llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4888 // return; 4889 // case llvm::MachO::PLATFORM_DRIVERKIT: 4890 // os_type = llvm::Triple::getOSTypeName(llvm::Triple::DriverKit); 4891 // return; 4892 case llvm::MachO::PLATFORM_MACCATALYST: 4893 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4894 environment = llvm::Triple::getEnvironmentTypeName(llvm::Triple::MacABI); 4895 return; 4896 case llvm::MachO::PLATFORM_IOSSIMULATOR: 4897 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4898 environment = 4899 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4900 return; 4901 case llvm::MachO::PLATFORM_TVOSSIMULATOR: 4902 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4903 environment = 4904 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4905 return; 4906 case llvm::MachO::PLATFORM_WATCHOSSIMULATOR: 4907 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4908 environment = 4909 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4910 return; 4911 default: { 4912 Log *log(GetLog(LLDBLog::Symbols | LLDBLog::Process)); 4913 LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION"); 4914 } 4915 } 4916 } 4917 }; 4918 4919 struct MinOS { 4920 uint32_t major_version, minor_version, patch_version; 4921 MinOS(uint32_t version) 4922 : major_version(version >> 16), minor_version((version >> 8) & 0xffu), 4923 patch_version(version & 0xffu) {} 4924 }; 4925 } // namespace 4926 4927 void ObjectFileMachO::GetAllArchSpecs(const llvm::MachO::mach_header &header, 4928 const lldb_private::DataExtractor &data, 4929 lldb::offset_t lc_offset, 4930 ModuleSpec &base_spec, 4931 lldb_private::ModuleSpecList &all_specs) { 4932 auto &base_arch = base_spec.GetArchitecture(); 4933 base_arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4934 if (!base_arch.IsValid()) 4935 return; 4936 4937 bool found_any = false; 4938 auto add_triple = [&](const llvm::Triple &triple) { 4939 auto spec = base_spec; 4940 spec.GetArchitecture().GetTriple() = triple; 4941 if (spec.GetArchitecture().IsValid()) { 4942 spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset); 4943 all_specs.Append(spec); 4944 found_any = true; 4945 } 4946 }; 4947 4948 // Set OS to an unspecified unknown or a "*" so it can match any OS 4949 llvm::Triple base_triple = base_arch.GetTriple(); 4950 base_triple.setOS(llvm::Triple::UnknownOS); 4951 base_triple.setOSName(llvm::StringRef()); 4952 4953 if (header.filetype == MH_PRELOAD) { 4954 if (header.cputype == CPU_TYPE_ARM) { 4955 // If this is a 32-bit arm binary, and it's a standalone binary, force 4956 // the Vendor to Apple so we don't accidentally pick up the generic 4957 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4958 // frame pointer register; most other armv7 ABIs use a combination of 4959 // r7 and r11. 4960 base_triple.setVendor(llvm::Triple::Apple); 4961 } else { 4962 // Set vendor to an unspecified unknown or a "*" so it can match any 4963 // vendor This is required for correct behavior of EFI debugging on 4964 // x86_64 4965 base_triple.setVendor(llvm::Triple::UnknownVendor); 4966 base_triple.setVendorName(llvm::StringRef()); 4967 } 4968 return add_triple(base_triple); 4969 } 4970 4971 llvm::MachO::load_command load_cmd; 4972 4973 // See if there is an LC_VERSION_MIN_* load command that can give 4974 // us the OS type. 4975 lldb::offset_t offset = lc_offset; 4976 for (uint32_t i = 0; i < header.ncmds; ++i) { 4977 const lldb::offset_t cmd_offset = offset; 4978 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4979 break; 4980 4981 llvm::MachO::version_min_command version_min; 4982 switch (load_cmd.cmd) { 4983 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4984 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4985 case llvm::MachO::LC_VERSION_MIN_TVOS: 4986 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 4987 if (load_cmd.cmdsize != sizeof(version_min)) 4988 break; 4989 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 4990 data.GetByteOrder(), &version_min) == 0) 4991 break; 4992 MinOS min_os(version_min.version); 4993 llvm::SmallString<32> os_name; 4994 llvm::raw_svector_ostream os(os_name); 4995 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 4996 << min_os.minor_version << '.' << min_os.patch_version; 4997 4998 auto triple = base_triple; 4999 triple.setOSName(os.str()); 5000 5001 // Disambiguate legacy simulator platforms. 5002 if (load_cmd.cmd != llvm::MachO::LC_VERSION_MIN_MACOSX && 5003 (base_triple.getArch() == llvm::Triple::x86_64 || 5004 base_triple.getArch() == llvm::Triple::x86)) { 5005 // The combination of legacy LC_VERSION_MIN load command and 5006 // x86 architecture always indicates a simulator environment. 5007 // The combination of LC_VERSION_MIN and arm architecture only 5008 // appears for native binaries. Back-deploying simulator 5009 // binaries on Apple Silicon Macs use the modern unambigous 5010 // LC_BUILD_VERSION load commands; no special handling required. 5011 triple.setEnvironment(llvm::Triple::Simulator); 5012 } 5013 add_triple(triple); 5014 break; 5015 } 5016 default: 5017 break; 5018 } 5019 5020 offset = cmd_offset + load_cmd.cmdsize; 5021 } 5022 5023 // See if there are LC_BUILD_VERSION load commands that can give 5024 // us the OS type. 5025 offset = lc_offset; 5026 for (uint32_t i = 0; i < header.ncmds; ++i) { 5027 const lldb::offset_t cmd_offset = offset; 5028 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 5029 break; 5030 5031 do { 5032 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 5033 llvm::MachO::build_version_command build_version; 5034 if (load_cmd.cmdsize < sizeof(build_version)) { 5035 // Malformed load command. 5036 break; 5037 } 5038 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 5039 data.GetByteOrder(), &build_version) == 0) 5040 break; 5041 MinOS min_os(build_version.minos); 5042 OSEnv os_env(build_version.platform); 5043 llvm::SmallString<16> os_name; 5044 llvm::raw_svector_ostream os(os_name); 5045 os << os_env.os_type << min_os.major_version << '.' 5046 << min_os.minor_version << '.' << min_os.patch_version; 5047 auto triple = base_triple; 5048 triple.setOSName(os.str()); 5049 os_name.clear(); 5050 if (!os_env.environment.empty()) 5051 triple.setEnvironmentName(os_env.environment); 5052 add_triple(triple); 5053 } 5054 } while (false); 5055 offset = cmd_offset + load_cmd.cmdsize; 5056 } 5057 5058 if (!found_any) { 5059 add_triple(base_triple); 5060 } 5061 } 5062 5063 ArchSpec ObjectFileMachO::GetArchitecture( 5064 ModuleSP module_sp, const llvm::MachO::mach_header &header, 5065 const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) { 5066 ModuleSpecList all_specs; 5067 ModuleSpec base_spec; 5068 GetAllArchSpecs(header, data, MachHeaderSizeFromMagic(header.magic), 5069 base_spec, all_specs); 5070 5071 // If the object file offers multiple alternative load commands, 5072 // pick the one that matches the module. 5073 if (module_sp) { 5074 const ArchSpec &module_arch = module_sp->GetArchitecture(); 5075 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 5076 ArchSpec mach_arch = 5077 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 5078 if (module_arch.IsCompatibleMatch(mach_arch)) 5079 return mach_arch; 5080 } 5081 } 5082 5083 // Return the first arch we found. 5084 if (all_specs.GetSize() == 0) 5085 return {}; 5086 return all_specs.GetModuleSpecRefAtIndex(0).GetArchitecture(); 5087 } 5088 5089 UUID ObjectFileMachO::GetUUID() { 5090 ModuleSP module_sp(GetModule()); 5091 if (module_sp) { 5092 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5093 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5094 return GetUUID(m_header, m_data, offset); 5095 } 5096 return UUID(); 5097 } 5098 5099 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5100 uint32_t count = 0; 5101 ModuleSP module_sp(GetModule()); 5102 if (module_sp) { 5103 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5104 llvm::MachO::load_command load_cmd; 5105 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5106 std::vector<std::string> rpath_paths; 5107 std::vector<std::string> rpath_relative_paths; 5108 std::vector<std::string> at_exec_relative_paths; 5109 uint32_t i; 5110 for (i = 0; i < m_header.ncmds; ++i) { 5111 const uint32_t cmd_offset = offset; 5112 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5113 break; 5114 5115 switch (load_cmd.cmd) { 5116 case LC_RPATH: 5117 case LC_LOAD_DYLIB: 5118 case LC_LOAD_WEAK_DYLIB: 5119 case LC_REEXPORT_DYLIB: 5120 case LC_LOAD_DYLINKER: 5121 case LC_LOADFVMLIB: 5122 case LC_LOAD_UPWARD_DYLIB: { 5123 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5124 const char *path = m_data.PeekCStr(name_offset); 5125 if (path) { 5126 if (load_cmd.cmd == LC_RPATH) 5127 rpath_paths.push_back(path); 5128 else { 5129 if (path[0] == '@') { 5130 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5131 rpath_relative_paths.push_back(path + strlen("@rpath")); 5132 else if (strncmp(path, "@executable_path", 5133 strlen("@executable_path")) == 0) 5134 at_exec_relative_paths.push_back(path + 5135 strlen("@executable_path")); 5136 } else { 5137 FileSpec file_spec(path); 5138 if (files.AppendIfUnique(file_spec)) 5139 count++; 5140 } 5141 } 5142 } 5143 } break; 5144 5145 default: 5146 break; 5147 } 5148 offset = cmd_offset + load_cmd.cmdsize; 5149 } 5150 5151 FileSpec this_file_spec(m_file); 5152 FileSystem::Instance().Resolve(this_file_spec); 5153 5154 if (!rpath_paths.empty()) { 5155 // Fixup all LC_RPATH values to be absolute paths 5156 std::string loader_path("@loader_path"); 5157 std::string executable_path("@executable_path"); 5158 for (auto &rpath : rpath_paths) { 5159 if (llvm::StringRef(rpath).startswith(loader_path)) { 5160 rpath.erase(0, loader_path.size()); 5161 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5162 } else if (llvm::StringRef(rpath).startswith(executable_path)) { 5163 rpath.erase(0, executable_path.size()); 5164 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5165 } 5166 } 5167 5168 for (const auto &rpath_relative_path : rpath_relative_paths) { 5169 for (const auto &rpath : rpath_paths) { 5170 std::string path = rpath; 5171 path += rpath_relative_path; 5172 // It is OK to resolve this path because we must find a file on disk 5173 // for us to accept it anyway if it is rpath relative. 5174 FileSpec file_spec(path); 5175 FileSystem::Instance().Resolve(file_spec); 5176 if (FileSystem::Instance().Exists(file_spec) && 5177 files.AppendIfUnique(file_spec)) { 5178 count++; 5179 break; 5180 } 5181 } 5182 } 5183 } 5184 5185 // We may have @executable_paths but no RPATHS. Figure those out here. 5186 // Only do this if this object file is the executable. We have no way to 5187 // get back to the actual executable otherwise, so we won't get the right 5188 // path. 5189 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5190 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5191 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5192 FileSpec file_spec = 5193 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5194 if (FileSystem::Instance().Exists(file_spec) && 5195 files.AppendIfUnique(file_spec)) 5196 count++; 5197 } 5198 } 5199 } 5200 return count; 5201 } 5202 5203 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5204 // If the object file is not an executable it can't hold the entry point. 5205 // m_entry_point_address is initialized to an invalid address, so we can just 5206 // return that. If m_entry_point_address is valid it means we've found it 5207 // already, so return the cached value. 5208 5209 if ((!IsExecutable() && !IsDynamicLoader()) || 5210 m_entry_point_address.IsValid()) { 5211 return m_entry_point_address; 5212 } 5213 5214 // Otherwise, look for the UnixThread or Thread command. The data for the 5215 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5216 // 5217 // uint32_t flavor - this is the flavor argument you would pass to 5218 // thread_get_state 5219 // uint32_t count - this is the count of longs in the thread state data 5220 // struct XXX_thread_state state - this is the structure from 5221 // <machine/thread_status.h> corresponding to the flavor. 5222 // <repeat this trio> 5223 // 5224 // So we just keep reading the various register flavors till we find the GPR 5225 // one, then read the PC out of there. 5226 // FIXME: We will need to have a "RegisterContext data provider" class at some 5227 // point that can get all the registers 5228 // out of data in this form & attach them to a given thread. That should 5229 // underlie the MacOS X User process plugin, and we'll also need it for the 5230 // MacOS X Core File process plugin. When we have that we can also use it 5231 // here. 5232 // 5233 // For now we hard-code the offsets and flavors we need: 5234 // 5235 // 5236 5237 ModuleSP module_sp(GetModule()); 5238 if (module_sp) { 5239 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5240 llvm::MachO::load_command load_cmd; 5241 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5242 uint32_t i; 5243 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5244 bool done = false; 5245 5246 for (i = 0; i < m_header.ncmds; ++i) { 5247 const lldb::offset_t cmd_offset = offset; 5248 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5249 break; 5250 5251 switch (load_cmd.cmd) { 5252 case LC_UNIXTHREAD: 5253 case LC_THREAD: { 5254 while (offset < cmd_offset + load_cmd.cmdsize) { 5255 uint32_t flavor = m_data.GetU32(&offset); 5256 uint32_t count = m_data.GetU32(&offset); 5257 if (count == 0) { 5258 // We've gotten off somehow, log and exit; 5259 return m_entry_point_address; 5260 } 5261 5262 switch (m_header.cputype) { 5263 case llvm::MachO::CPU_TYPE_ARM: 5264 if (flavor == 1 || 5265 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 5266 // from mach/arm/thread_status.h 5267 { 5268 offset += 60; // This is the offset of pc in the GPR thread state 5269 // data structure. 5270 start_address = m_data.GetU32(&offset); 5271 done = true; 5272 } 5273 break; 5274 case llvm::MachO::CPU_TYPE_ARM64: 5275 case llvm::MachO::CPU_TYPE_ARM64_32: 5276 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5277 { 5278 offset += 256; // This is the offset of pc in the GPR thread state 5279 // data structure. 5280 start_address = m_data.GetU64(&offset); 5281 done = true; 5282 } 5283 break; 5284 case llvm::MachO::CPU_TYPE_I386: 5285 if (flavor == 5286 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5287 { 5288 offset += 40; // This is the offset of eip in the GPR thread state 5289 // data structure. 5290 start_address = m_data.GetU32(&offset); 5291 done = true; 5292 } 5293 break; 5294 case llvm::MachO::CPU_TYPE_X86_64: 5295 if (flavor == 5296 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5297 { 5298 offset += 16 * 8; // This is the offset of rip in the GPR thread 5299 // state data structure. 5300 start_address = m_data.GetU64(&offset); 5301 done = true; 5302 } 5303 break; 5304 default: 5305 return m_entry_point_address; 5306 } 5307 // Haven't found the GPR flavor yet, skip over the data for this 5308 // flavor: 5309 if (done) 5310 break; 5311 offset += count * 4; 5312 } 5313 } break; 5314 case LC_MAIN: { 5315 ConstString text_segment_name("__TEXT"); 5316 uint64_t entryoffset = m_data.GetU64(&offset); 5317 SectionSP text_segment_sp = 5318 GetSectionList()->FindSectionByName(text_segment_name); 5319 if (text_segment_sp) { 5320 done = true; 5321 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5322 } 5323 } break; 5324 5325 default: 5326 break; 5327 } 5328 if (done) 5329 break; 5330 5331 // Go to the next load command: 5332 offset = cmd_offset + load_cmd.cmdsize; 5333 } 5334 5335 if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) { 5336 if (GetSymtab()) { 5337 Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType( 5338 ConstString("_dyld_start"), SymbolType::eSymbolTypeCode, 5339 Symtab::eDebugAny, Symtab::eVisibilityAny); 5340 if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) { 5341 start_address = dyld_start_sym->GetAddress().GetFileAddress(); 5342 } 5343 } 5344 } 5345 5346 if (start_address != LLDB_INVALID_ADDRESS) { 5347 // We got the start address from the load commands, so now resolve that 5348 // address in the sections of this ObjectFile: 5349 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5350 start_address, GetSectionList())) { 5351 m_entry_point_address.Clear(); 5352 } 5353 } else { 5354 // We couldn't read the UnixThread load command - maybe it wasn't there. 5355 // As a fallback look for the "start" symbol in the main executable. 5356 5357 ModuleSP module_sp(GetModule()); 5358 5359 if (module_sp) { 5360 SymbolContextList contexts; 5361 SymbolContext context; 5362 module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5363 eSymbolTypeCode, contexts); 5364 if (contexts.GetSize()) { 5365 if (contexts.GetContextAtIndex(0, context)) 5366 m_entry_point_address = context.symbol->GetAddress(); 5367 } 5368 } 5369 } 5370 } 5371 5372 return m_entry_point_address; 5373 } 5374 5375 lldb_private::Address ObjectFileMachO::GetBaseAddress() { 5376 lldb_private::Address header_addr; 5377 SectionList *section_list = GetSectionList(); 5378 if (section_list) { 5379 SectionSP text_segment_sp( 5380 section_list->FindSectionByName(GetSegmentNameTEXT())); 5381 if (text_segment_sp) { 5382 header_addr.SetSection(text_segment_sp); 5383 header_addr.SetOffset(0); 5384 } 5385 } 5386 return header_addr; 5387 } 5388 5389 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5390 ModuleSP module_sp(GetModule()); 5391 if (module_sp) { 5392 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5393 if (!m_thread_context_offsets_valid) { 5394 m_thread_context_offsets_valid = true; 5395 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5396 FileRangeArray::Entry file_range; 5397 llvm::MachO::thread_command thread_cmd; 5398 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5399 const uint32_t cmd_offset = offset; 5400 if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr) 5401 break; 5402 5403 if (thread_cmd.cmd == LC_THREAD) { 5404 file_range.SetRangeBase(offset); 5405 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5406 m_thread_context_offsets.Append(file_range); 5407 } 5408 offset = cmd_offset + thread_cmd.cmdsize; 5409 } 5410 } 5411 } 5412 return m_thread_context_offsets.GetSize(); 5413 } 5414 5415 std::string ObjectFileMachO::GetIdentifierString() { 5416 std::string result; 5417 ModuleSP module_sp(GetModule()); 5418 if (module_sp) { 5419 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5420 5421 // First, look over the load commands for an LC_NOTE load command with 5422 // data_owner string "kern ver str" & use that if found. 5423 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5424 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5425 const uint32_t cmd_offset = offset; 5426 llvm::MachO::load_command lc; 5427 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5428 break; 5429 if (lc.cmd == LC_NOTE) { 5430 char data_owner[17]; 5431 m_data.CopyData(offset, 16, data_owner); 5432 data_owner[16] = '\0'; 5433 offset += 16; 5434 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5435 uint64_t size = m_data.GetU64_unchecked(&offset); 5436 5437 // "kern ver str" has a uint32_t version and then a nul terminated 5438 // c-string. 5439 if (strcmp("kern ver str", data_owner) == 0) { 5440 offset = fileoff; 5441 uint32_t version; 5442 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5443 if (version == 1) { 5444 uint32_t strsize = size - sizeof(uint32_t); 5445 char *buf = (char *)malloc(strsize); 5446 if (buf) { 5447 m_data.CopyData(offset, strsize, buf); 5448 buf[strsize - 1] = '\0'; 5449 result = buf; 5450 if (buf) 5451 free(buf); 5452 return result; 5453 } 5454 } 5455 } 5456 } 5457 } 5458 offset = cmd_offset + lc.cmdsize; 5459 } 5460 5461 // Second, make a pass over the load commands looking for an obsolete 5462 // LC_IDENT load command. 5463 offset = MachHeaderSizeFromMagic(m_header.magic); 5464 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5465 const uint32_t cmd_offset = offset; 5466 llvm::MachO::ident_command ident_command; 5467 if (m_data.GetU32(&offset, &ident_command, 2) == nullptr) 5468 break; 5469 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5470 char *buf = (char *)malloc(ident_command.cmdsize); 5471 if (buf != nullptr && m_data.CopyData(offset, ident_command.cmdsize, 5472 buf) == ident_command.cmdsize) { 5473 buf[ident_command.cmdsize - 1] = '\0'; 5474 result = buf; 5475 } 5476 if (buf) 5477 free(buf); 5478 } 5479 offset = cmd_offset + ident_command.cmdsize; 5480 } 5481 } 5482 return result; 5483 } 5484 5485 addr_t ObjectFileMachO::GetAddressMask() { 5486 addr_t mask = 0; 5487 ModuleSP module_sp(GetModule()); 5488 if (module_sp) { 5489 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5490 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5491 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5492 const uint32_t cmd_offset = offset; 5493 llvm::MachO::load_command lc; 5494 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5495 break; 5496 if (lc.cmd == LC_NOTE) { 5497 char data_owner[17]; 5498 m_data.CopyData(offset, 16, data_owner); 5499 data_owner[16] = '\0'; 5500 offset += 16; 5501 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5502 5503 // "addrable bits" has a uint32_t version and a uint32_t 5504 // number of bits used in addressing. 5505 if (strcmp("addrable bits", data_owner) == 0) { 5506 offset = fileoff; 5507 uint32_t version; 5508 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5509 if (version == 3) { 5510 uint32_t num_addr_bits = m_data.GetU32_unchecked(&offset); 5511 if (num_addr_bits != 0) { 5512 mask = ~((1ULL << num_addr_bits) - 1); 5513 } 5514 break; 5515 } 5516 } 5517 } 5518 } 5519 offset = cmd_offset + lc.cmdsize; 5520 } 5521 } 5522 return mask; 5523 } 5524 5525 bool ObjectFileMachO::GetCorefileMainBinaryInfo(addr_t &value, 5526 bool &value_is_offset, 5527 UUID &uuid, 5528 ObjectFile::BinaryType &type) { 5529 value = LLDB_INVALID_ADDRESS; 5530 value_is_offset = false; 5531 uuid.Clear(); 5532 uint32_t log2_pagesize = 0; // not currently passed up to caller 5533 uint32_t platform = 0; // not currently passed up to caller 5534 ModuleSP module_sp(GetModule()); 5535 if (module_sp) { 5536 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5537 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5538 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5539 const uint32_t cmd_offset = offset; 5540 llvm::MachO::load_command lc; 5541 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5542 break; 5543 if (lc.cmd == LC_NOTE) { 5544 char data_owner[17]; 5545 memset(data_owner, 0, sizeof(data_owner)); 5546 m_data.CopyData(offset, 16, data_owner); 5547 offset += 16; 5548 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5549 uint64_t size = m_data.GetU64_unchecked(&offset); 5550 5551 // struct main_bin_spec 5552 // { 5553 // uint32_t version; // currently 2 5554 // uint32_t type; // 0 == unspecified, 1 == kernel, 5555 // // 2 == user process, 5556 // // 3 == standalone binary 5557 // uint64_t address; // UINT64_MAX if address not specified 5558 // uint64_t slide; // slide, UINT64_MAX if unspecified 5559 // // 0 if no slide needs to be applied to 5560 // // file address 5561 // uuid_t uuid; // all zero's if uuid not specified 5562 // uint32_t log2_pagesize; // process page size in log base 2, 5563 // // e.g. 4k pages are 12. 5564 // // 0 for unspecified 5565 // uint32_t platform; // The Mach-O platform for this corefile. 5566 // // 0 for unspecified. 5567 // // The values are defined in 5568 // // <mach-o/loader.h>, PLATFORM_*. 5569 // } __attribute((packed)); 5570 5571 // "main bin spec" (main binary specification) data payload is 5572 // formatted: 5573 // uint32_t version [currently 1] 5574 // uint32_t type [0 == unspecified, 1 == kernel, 5575 // 2 == user process, 3 == firmware ] 5576 // uint64_t address [ UINT64_MAX if address not specified ] 5577 // uuid_t uuid [ all zero's if uuid not specified ] 5578 // uint32_t log2_pagesize [ process page size in log base 5579 // 2, e.g. 4k pages are 12. 5580 // 0 for unspecified ] 5581 // uint32_t unused [ for alignment ] 5582 5583 if (strcmp("main bin spec", data_owner) == 0 && size >= 32) { 5584 offset = fileoff; 5585 uint32_t version; 5586 if (m_data.GetU32(&offset, &version, 1) != nullptr && version <= 2) { 5587 uint32_t binspec_type = 0; 5588 uuid_t raw_uuid; 5589 memset(raw_uuid, 0, sizeof(uuid_t)); 5590 5591 if (!m_data.GetU32(&offset, &binspec_type, 1)) 5592 return false; 5593 if (!m_data.GetU64(&offset, &value, 1)) 5594 return false; 5595 uint64_t slide = LLDB_INVALID_ADDRESS; 5596 if (version > 1 && !m_data.GetU64(&offset, &slide, 1)) 5597 return false; 5598 if (value == LLDB_INVALID_ADDRESS && 5599 slide != LLDB_INVALID_ADDRESS) { 5600 value = slide; 5601 value_is_offset = true; 5602 } 5603 5604 if (m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5605 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5606 // convert the "main bin spec" type into our 5607 // ObjectFile::BinaryType enum 5608 switch (binspec_type) { 5609 case 0: 5610 type = eBinaryTypeUnknown; 5611 break; 5612 case 1: 5613 type = eBinaryTypeKernel; 5614 break; 5615 case 2: 5616 type = eBinaryTypeUser; 5617 break; 5618 case 3: 5619 type = eBinaryTypeStandalone; 5620 break; 5621 } 5622 if (!m_data.GetU32(&offset, &log2_pagesize, 1)) 5623 return false; 5624 if (version > 1 && !m_data.GetU32(&offset, &platform, 1)) 5625 return false; 5626 return true; 5627 } 5628 } 5629 } 5630 } 5631 offset = cmd_offset + lc.cmdsize; 5632 } 5633 } 5634 return false; 5635 } 5636 5637 lldb::RegisterContextSP 5638 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5639 lldb_private::Thread &thread) { 5640 lldb::RegisterContextSP reg_ctx_sp; 5641 5642 ModuleSP module_sp(GetModule()); 5643 if (module_sp) { 5644 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5645 if (!m_thread_context_offsets_valid) 5646 GetNumThreadContexts(); 5647 5648 const FileRangeArray::Entry *thread_context_file_range = 5649 m_thread_context_offsets.GetEntryAtIndex(idx); 5650 if (thread_context_file_range) { 5651 5652 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5653 thread_context_file_range->GetByteSize()); 5654 5655 switch (m_header.cputype) { 5656 case llvm::MachO::CPU_TYPE_ARM64: 5657 case llvm::MachO::CPU_TYPE_ARM64_32: 5658 reg_ctx_sp = 5659 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data); 5660 break; 5661 5662 case llvm::MachO::CPU_TYPE_ARM: 5663 reg_ctx_sp = 5664 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data); 5665 break; 5666 5667 case llvm::MachO::CPU_TYPE_I386: 5668 reg_ctx_sp = 5669 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data); 5670 break; 5671 5672 case llvm::MachO::CPU_TYPE_X86_64: 5673 reg_ctx_sp = 5674 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data); 5675 break; 5676 } 5677 } 5678 } 5679 return reg_ctx_sp; 5680 } 5681 5682 ObjectFile::Type ObjectFileMachO::CalculateType() { 5683 switch (m_header.filetype) { 5684 case MH_OBJECT: // 0x1u 5685 if (GetAddressByteSize() == 4) { 5686 // 32 bit kexts are just object files, but they do have a valid 5687 // UUID load command. 5688 if (GetUUID()) { 5689 // this checking for the UUID load command is not enough we could 5690 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5691 // this is required of kexts 5692 if (m_strata == eStrataInvalid) 5693 m_strata = eStrataKernel; 5694 return eTypeSharedLibrary; 5695 } 5696 } 5697 return eTypeObjectFile; 5698 5699 case MH_EXECUTE: 5700 return eTypeExecutable; // 0x2u 5701 case MH_FVMLIB: 5702 return eTypeSharedLibrary; // 0x3u 5703 case MH_CORE: 5704 return eTypeCoreFile; // 0x4u 5705 case MH_PRELOAD: 5706 return eTypeSharedLibrary; // 0x5u 5707 case MH_DYLIB: 5708 return eTypeSharedLibrary; // 0x6u 5709 case MH_DYLINKER: 5710 return eTypeDynamicLinker; // 0x7u 5711 case MH_BUNDLE: 5712 return eTypeSharedLibrary; // 0x8u 5713 case MH_DYLIB_STUB: 5714 return eTypeStubLibrary; // 0x9u 5715 case MH_DSYM: 5716 return eTypeDebugInfo; // 0xAu 5717 case MH_KEXT_BUNDLE: 5718 return eTypeSharedLibrary; // 0xBu 5719 default: 5720 break; 5721 } 5722 return eTypeUnknown; 5723 } 5724 5725 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5726 switch (m_header.filetype) { 5727 case MH_OBJECT: // 0x1u 5728 { 5729 // 32 bit kexts are just object files, but they do have a valid 5730 // UUID load command. 5731 if (GetUUID()) { 5732 // this checking for the UUID load command is not enough we could 5733 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5734 // this is required of kexts 5735 if (m_type == eTypeInvalid) 5736 m_type = eTypeSharedLibrary; 5737 5738 return eStrataKernel; 5739 } 5740 } 5741 return eStrataUnknown; 5742 5743 case MH_EXECUTE: // 0x2u 5744 // Check for the MH_DYLDLINK bit in the flags 5745 if (m_header.flags & MH_DYLDLINK) { 5746 return eStrataUser; 5747 } else { 5748 SectionList *section_list = GetSectionList(); 5749 if (section_list) { 5750 static ConstString g_kld_section_name("__KLD"); 5751 if (section_list->FindSectionByName(g_kld_section_name)) 5752 return eStrataKernel; 5753 } 5754 } 5755 return eStrataRawImage; 5756 5757 case MH_FVMLIB: 5758 return eStrataUser; // 0x3u 5759 case MH_CORE: 5760 return eStrataUnknown; // 0x4u 5761 case MH_PRELOAD: 5762 return eStrataRawImage; // 0x5u 5763 case MH_DYLIB: 5764 return eStrataUser; // 0x6u 5765 case MH_DYLINKER: 5766 return eStrataUser; // 0x7u 5767 case MH_BUNDLE: 5768 return eStrataUser; // 0x8u 5769 case MH_DYLIB_STUB: 5770 return eStrataUser; // 0x9u 5771 case MH_DSYM: 5772 return eStrataUnknown; // 0xAu 5773 case MH_KEXT_BUNDLE: 5774 return eStrataKernel; // 0xBu 5775 default: 5776 break; 5777 } 5778 return eStrataUnknown; 5779 } 5780 5781 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5782 ModuleSP module_sp(GetModule()); 5783 if (module_sp) { 5784 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5785 llvm::MachO::dylib_command load_cmd; 5786 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5787 uint32_t version_cmd = 0; 5788 uint64_t version = 0; 5789 uint32_t i; 5790 for (i = 0; i < m_header.ncmds; ++i) { 5791 const lldb::offset_t cmd_offset = offset; 5792 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5793 break; 5794 5795 if (load_cmd.cmd == LC_ID_DYLIB) { 5796 if (version_cmd == 0) { 5797 version_cmd = load_cmd.cmd; 5798 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr) 5799 break; 5800 version = load_cmd.dylib.current_version; 5801 } 5802 break; // Break for now unless there is another more complete version 5803 // number load command in the future. 5804 } 5805 offset = cmd_offset + load_cmd.cmdsize; 5806 } 5807 5808 if (version_cmd == LC_ID_DYLIB) { 5809 unsigned major = (version & 0xFFFF0000ull) >> 16; 5810 unsigned minor = (version & 0x0000FF00ull) >> 8; 5811 unsigned subminor = (version & 0x000000FFull); 5812 return llvm::VersionTuple(major, minor, subminor); 5813 } 5814 } 5815 return llvm::VersionTuple(); 5816 } 5817 5818 ArchSpec ObjectFileMachO::GetArchitecture() { 5819 ModuleSP module_sp(GetModule()); 5820 ArchSpec arch; 5821 if (module_sp) { 5822 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5823 5824 return GetArchitecture(module_sp, m_header, m_data, 5825 MachHeaderSizeFromMagic(m_header.magic)); 5826 } 5827 return arch; 5828 } 5829 5830 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, 5831 addr_t &base_addr, UUID &uuid) { 5832 uuid.Clear(); 5833 base_addr = LLDB_INVALID_ADDRESS; 5834 if (process && process->GetDynamicLoader()) { 5835 DynamicLoader *dl = process->GetDynamicLoader(); 5836 LazyBool using_shared_cache; 5837 LazyBool private_shared_cache; 5838 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5839 private_shared_cache); 5840 } 5841 Log *log(GetLog(LLDBLog::Symbols | LLDBLog::Process)); 5842 LLDB_LOGF( 5843 log, 5844 "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64, 5845 uuid.GetAsString().c_str(), base_addr); 5846 } 5847 5848 // From dyld SPI header dyld_process_info.h 5849 typedef void *dyld_process_info; 5850 struct lldb_copy__dyld_process_cache_info { 5851 uuid_t cacheUUID; // UUID of cache used by process 5852 uint64_t cacheBaseAddress; // load address of dyld shared cache 5853 bool noCache; // process is running without a dyld cache 5854 bool privateCache; // process is using a private copy of its dyld cache 5855 }; 5856 5857 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with 5858 // llvm enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile 5859 // errors. So we need to use the actual underlying types of task_t and 5860 // kern_return_t below. 5861 extern "C" unsigned int /*task_t*/ mach_task_self(); 5862 5863 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5864 uuid.Clear(); 5865 base_addr = LLDB_INVALID_ADDRESS; 5866 5867 #if defined(__APPLE__) 5868 uint8_t *(*dyld_get_all_image_infos)(void); 5869 dyld_get_all_image_infos = 5870 (uint8_t * (*)()) dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5871 if (dyld_get_all_image_infos) { 5872 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5873 if (dyld_all_image_infos_address) { 5874 uint32_t *version = (uint32_t *) 5875 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5876 if (*version >= 13) { 5877 uuid_t *sharedCacheUUID_address = 0; 5878 int wordsize = sizeof(uint8_t *); 5879 if (wordsize == 8) { 5880 sharedCacheUUID_address = 5881 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5882 160); // sharedCacheUUID <mach-o/dyld_images.h> 5883 if (*version >= 15) 5884 base_addr = 5885 *(uint64_t 5886 *)((uint8_t *)dyld_all_image_infos_address + 5887 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5888 } else { 5889 sharedCacheUUID_address = 5890 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5891 84); // sharedCacheUUID <mach-o/dyld_images.h> 5892 if (*version >= 15) { 5893 base_addr = 0; 5894 base_addr = 5895 *(uint32_t 5896 *)((uint8_t *)dyld_all_image_infos_address + 5897 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5898 } 5899 } 5900 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5901 } 5902 } 5903 } else { 5904 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5905 dyld_process_info (*dyld_process_info_create)( 5906 unsigned int /* task_t */ task, uint64_t timestamp, 5907 unsigned int /*kern_return_t*/ *kernelError); 5908 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5909 void (*dyld_process_info_release)(dyld_process_info info); 5910 5911 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, 5912 unsigned int /*kern_return_t*/ *)) 5913 dlsym(RTLD_DEFAULT, "_dyld_process_info_create"); 5914 dyld_process_info_get_cache = (void (*)(void *, void *))dlsym( 5915 RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5916 dyld_process_info_release = 5917 (void (*)(void *))dlsym(RTLD_DEFAULT, "_dyld_process_info_release"); 5918 5919 if (dyld_process_info_create && dyld_process_info_get_cache) { 5920 unsigned int /*kern_return_t */ kern_ret; 5921 dyld_process_info process_info = 5922 dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5923 if (process_info) { 5924 struct lldb_copy__dyld_process_cache_info sc_info; 5925 memset(&sc_info, 0, sizeof(struct lldb_copy__dyld_process_cache_info)); 5926 dyld_process_info_get_cache(process_info, &sc_info); 5927 if (sc_info.cacheBaseAddress != 0) { 5928 base_addr = sc_info.cacheBaseAddress; 5929 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 5930 } 5931 dyld_process_info_release(process_info); 5932 } 5933 } 5934 } 5935 Log *log(GetLog(LLDBLog::Symbols | LLDBLog::Process)); 5936 if (log && uuid.IsValid()) 5937 LLDB_LOGF(log, 5938 "lldb's in-memory shared cache has a UUID of %s base address of " 5939 "0x%" PRIx64, 5940 uuid.GetAsString().c_str(), base_addr); 5941 #endif 5942 } 5943 5944 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5945 if (!m_min_os_version) { 5946 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5947 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5948 const lldb::offset_t load_cmd_offset = offset; 5949 5950 llvm::MachO::version_min_command lc; 5951 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5952 break; 5953 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5954 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5955 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5956 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5957 if (m_data.GetU32(&offset, &lc.version, 5958 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5959 const uint32_t xxxx = lc.version >> 16; 5960 const uint32_t yy = (lc.version >> 8) & 0xffu; 5961 const uint32_t zz = lc.version & 0xffu; 5962 if (xxxx) { 5963 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5964 break; 5965 } 5966 } 5967 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5968 // struct build_version_command { 5969 // uint32_t cmd; /* LC_BUILD_VERSION */ 5970 // uint32_t cmdsize; /* sizeof(struct 5971 // build_version_command) plus */ 5972 // /* ntools * sizeof(struct 5973 // build_tool_version) */ 5974 // uint32_t platform; /* platform */ 5975 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5976 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded in 5977 // nibbles xxxx.yy.zz */ uint32_t ntools; /* number of 5978 // tool entries following this */ 5979 // }; 5980 5981 offset += 4; // skip platform 5982 uint32_t minos = m_data.GetU32(&offset); 5983 5984 const uint32_t xxxx = minos >> 16; 5985 const uint32_t yy = (minos >> 8) & 0xffu; 5986 const uint32_t zz = minos & 0xffu; 5987 if (xxxx) { 5988 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5989 break; 5990 } 5991 } 5992 5993 offset = load_cmd_offset + lc.cmdsize; 5994 } 5995 5996 if (!m_min_os_version) { 5997 // Set version to an empty value so we don't keep trying to 5998 m_min_os_version = llvm::VersionTuple(); 5999 } 6000 } 6001 6002 return *m_min_os_version; 6003 } 6004 6005 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { 6006 if (!m_sdk_versions) { 6007 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 6008 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6009 const lldb::offset_t load_cmd_offset = offset; 6010 6011 llvm::MachO::version_min_command lc; 6012 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6013 break; 6014 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 6015 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 6016 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 6017 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 6018 if (m_data.GetU32(&offset, &lc.version, 6019 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 6020 const uint32_t xxxx = lc.sdk >> 16; 6021 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 6022 const uint32_t zz = lc.sdk & 0xffu; 6023 if (xxxx) { 6024 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 6025 break; 6026 } else { 6027 GetModule()->ReportWarning("minimum OS version load command with " 6028 "invalid (0) version found."); 6029 } 6030 } 6031 } 6032 offset = load_cmd_offset + lc.cmdsize; 6033 } 6034 6035 if (!m_sdk_versions) { 6036 offset = MachHeaderSizeFromMagic(m_header.magic); 6037 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6038 const lldb::offset_t load_cmd_offset = offset; 6039 6040 llvm::MachO::version_min_command lc; 6041 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6042 break; 6043 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 6044 // struct build_version_command { 6045 // uint32_t cmd; /* LC_BUILD_VERSION */ 6046 // uint32_t cmdsize; /* sizeof(struct 6047 // build_version_command) plus */ 6048 // /* ntools * sizeof(struct 6049 // build_tool_version) */ 6050 // uint32_t platform; /* platform */ 6051 // uint32_t minos; /* X.Y.Z is encoded in nibbles 6052 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded 6053 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number 6054 // of tool entries following this */ 6055 // }; 6056 6057 offset += 4; // skip platform 6058 uint32_t minos = m_data.GetU32(&offset); 6059 6060 const uint32_t xxxx = minos >> 16; 6061 const uint32_t yy = (minos >> 8) & 0xffu; 6062 const uint32_t zz = minos & 0xffu; 6063 if (xxxx) { 6064 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 6065 break; 6066 } 6067 } 6068 offset = load_cmd_offset + lc.cmdsize; 6069 } 6070 } 6071 6072 if (!m_sdk_versions) 6073 m_sdk_versions = llvm::VersionTuple(); 6074 } 6075 6076 return *m_sdk_versions; 6077 } 6078 6079 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 6080 return m_header.filetype == llvm::MachO::MH_DYLINKER; 6081 } 6082 6083 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 6084 return m_allow_assembly_emulation_unwind_plans; 6085 } 6086 6087 Section *ObjectFileMachO::GetMachHeaderSection() { 6088 // Find the first address of the mach header which is the first non-zero file 6089 // sized section whose file offset is zero. This is the base file address of 6090 // the mach-o file which can be subtracted from the vmaddr of the other 6091 // segments found in memory and added to the load address 6092 ModuleSP module_sp = GetModule(); 6093 if (!module_sp) 6094 return nullptr; 6095 SectionList *section_list = GetSectionList(); 6096 if (!section_list) 6097 return nullptr; 6098 const size_t num_sections = section_list->GetSize(); 6099 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6100 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 6101 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 6102 return section; 6103 } 6104 6105 // We may have a binary in the shared cache that has a non-zero 6106 // file address for its first segment, traditionally the __TEXT segment. 6107 // Search for it by name and return it as our next best guess. 6108 SectionSP text_segment_sp = 6109 GetSectionList()->FindSectionByName(GetSegmentNameTEXT()); 6110 if (text_segment_sp.get() && SectionIsLoadable(text_segment_sp.get())) 6111 return text_segment_sp.get(); 6112 6113 return nullptr; 6114 } 6115 6116 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 6117 if (!section) 6118 return false; 6119 const bool is_dsym = (m_header.filetype == MH_DSYM); 6120 if (section->GetFileSize() == 0 && !is_dsym) 6121 return false; 6122 if (section->IsThreadSpecific()) 6123 return false; 6124 if (GetModule().get() != section->GetModule().get()) 6125 return false; 6126 // Be careful with __LINKEDIT and __DWARF segments 6127 if (section->GetName() == GetSegmentNameLINKEDIT() || 6128 section->GetName() == GetSegmentNameDWARF()) { 6129 // Only map __LINKEDIT and __DWARF if we have an in memory image and 6130 // this isn't a kernel binary like a kext or mach_kernel. 6131 const bool is_memory_image = (bool)m_process_wp.lock(); 6132 const Strata strata = GetStrata(); 6133 if (is_memory_image == false || strata == eStrataKernel) 6134 return false; 6135 } 6136 return true; 6137 } 6138 6139 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 6140 lldb::addr_t header_load_address, const Section *header_section, 6141 const Section *section) { 6142 ModuleSP module_sp = GetModule(); 6143 if (module_sp && header_section && section && 6144 header_load_address != LLDB_INVALID_ADDRESS) { 6145 lldb::addr_t file_addr = header_section->GetFileAddress(); 6146 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 6147 return section->GetFileAddress() - file_addr + header_load_address; 6148 } 6149 return LLDB_INVALID_ADDRESS; 6150 } 6151 6152 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 6153 bool value_is_offset) { 6154 ModuleSP module_sp = GetModule(); 6155 if (!module_sp) 6156 return false; 6157 6158 SectionList *section_list = GetSectionList(); 6159 if (!section_list) 6160 return false; 6161 6162 size_t num_loaded_sections = 0; 6163 const size_t num_sections = section_list->GetSize(); 6164 6165 if (value_is_offset) { 6166 // "value" is an offset to apply to each top level segment 6167 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6168 // Iterate through the object file sections to find all of the 6169 // sections that size on disk (to avoid __PAGEZERO) and load them 6170 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6171 if (SectionIsLoadable(section_sp.get())) 6172 if (target.GetSectionLoadList().SetSectionLoadAddress( 6173 section_sp, section_sp->GetFileAddress() + value)) 6174 ++num_loaded_sections; 6175 } 6176 } else { 6177 // "value" is the new base address of the mach_header, adjust each 6178 // section accordingly 6179 6180 Section *mach_header_section = GetMachHeaderSection(); 6181 if (mach_header_section) { 6182 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6183 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6184 6185 lldb::addr_t section_load_addr = 6186 CalculateSectionLoadAddressForMemoryImage( 6187 value, mach_header_section, section_sp.get()); 6188 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6189 if (target.GetSectionLoadList().SetSectionLoadAddress( 6190 section_sp, section_load_addr)) 6191 ++num_loaded_sections; 6192 } 6193 } 6194 } 6195 } 6196 return num_loaded_sections > 0; 6197 } 6198 6199 struct all_image_infos_header { 6200 uint32_t version; // currently 1 6201 uint32_t imgcount; // number of binary images 6202 uint64_t entries_fileoff; // file offset in the corefile of where the array of 6203 // struct entry's begin. 6204 uint32_t entries_size; // size of 'struct entry'. 6205 uint32_t unused; 6206 }; 6207 6208 struct image_entry { 6209 uint64_t filepath_offset; // offset in corefile to c-string of the file path, 6210 // UINT64_MAX if unavailable. 6211 uuid_t uuid; // uint8_t[16]. should be set to all zeroes if 6212 // uuid is unknown. 6213 uint64_t load_address; // UINT64_MAX if unknown. 6214 uint64_t seg_addrs_offset; // offset to the array of struct segment_vmaddr's. 6215 uint32_t segment_count; // The number of segments for this binary. 6216 uint32_t unused; 6217 6218 image_entry() { 6219 filepath_offset = UINT64_MAX; 6220 memset(&uuid, 0, sizeof(uuid_t)); 6221 segment_count = 0; 6222 load_address = UINT64_MAX; 6223 seg_addrs_offset = UINT64_MAX; 6224 unused = 0; 6225 } 6226 image_entry(const image_entry &rhs) { 6227 filepath_offset = rhs.filepath_offset; 6228 memcpy(&uuid, &rhs.uuid, sizeof(uuid_t)); 6229 segment_count = rhs.segment_count; 6230 seg_addrs_offset = rhs.seg_addrs_offset; 6231 load_address = rhs.load_address; 6232 unused = rhs.unused; 6233 } 6234 }; 6235 6236 struct segment_vmaddr { 6237 char segname[16]; 6238 uint64_t vmaddr; 6239 uint64_t unused; 6240 6241 segment_vmaddr() { 6242 memset(&segname, 0, 16); 6243 vmaddr = UINT64_MAX; 6244 unused = 0; 6245 } 6246 segment_vmaddr(const segment_vmaddr &rhs) { 6247 memcpy(&segname, &rhs.segname, 16); 6248 vmaddr = rhs.vmaddr; 6249 unused = rhs.unused; 6250 } 6251 }; 6252 6253 // Write the payload for the "all image infos" LC_NOTE into 6254 // the supplied all_image_infos_payload, assuming that this 6255 // will be written into the corefile starting at 6256 // initial_file_offset. 6257 // 6258 // The placement of this payload is a little tricky. We're 6259 // laying this out as 6260 // 6261 // 1. header (struct all_image_info_header) 6262 // 2. Array of fixed-size (struct image_entry)'s, one 6263 // per binary image present in the process. 6264 // 3. Arrays of (struct segment_vmaddr)'s, a varying number 6265 // for each binary image. 6266 // 4. Variable length c-strings of binary image filepaths, 6267 // one per binary. 6268 // 6269 // To compute where everything will be laid out in the 6270 // payload, we need to iterate over the images and calculate 6271 // how many segment_vmaddr structures each image will need, 6272 // and how long each image's filepath c-string is. There 6273 // are some multiple passes over the image list while calculating 6274 // everything. 6275 6276 static offset_t CreateAllImageInfosPayload( 6277 const lldb::ProcessSP &process_sp, offset_t initial_file_offset, 6278 StreamString &all_image_infos_payload, SaveCoreStyle core_style) { 6279 Target &target = process_sp->GetTarget(); 6280 ModuleList modules = target.GetImages(); 6281 6282 // stack-only corefiles have no reason to include binaries that 6283 // are not executing; we're trying to make the smallest corefile 6284 // we can, so leave the rest out. 6285 if (core_style == SaveCoreStyle::eSaveCoreStackOnly) 6286 modules.Clear(); 6287 6288 std::set<std::string> executing_uuids; 6289 ThreadList &thread_list(process_sp->GetThreadList()); 6290 for (uint32_t i = 0; i < thread_list.GetSize(); i++) { 6291 ThreadSP thread_sp = thread_list.GetThreadAtIndex(i); 6292 uint32_t stack_frame_count = thread_sp->GetStackFrameCount(); 6293 for (uint32_t j = 0; j < stack_frame_count; j++) { 6294 StackFrameSP stack_frame_sp = thread_sp->GetStackFrameAtIndex(j); 6295 Address pc = stack_frame_sp->GetFrameCodeAddress(); 6296 ModuleSP module_sp = pc.GetModule(); 6297 if (module_sp) { 6298 UUID uuid = module_sp->GetUUID(); 6299 if (uuid.IsValid()) { 6300 executing_uuids.insert(uuid.GetAsString()); 6301 modules.AppendIfNeeded(module_sp); 6302 } 6303 } 6304 } 6305 } 6306 size_t modules_count = modules.GetSize(); 6307 6308 struct all_image_infos_header infos; 6309 infos.version = 1; 6310 infos.imgcount = modules_count; 6311 infos.entries_size = sizeof(image_entry); 6312 infos.entries_fileoff = initial_file_offset + sizeof(all_image_infos_header); 6313 infos.unused = 0; 6314 6315 all_image_infos_payload.PutHex32(infos.version); 6316 all_image_infos_payload.PutHex32(infos.imgcount); 6317 all_image_infos_payload.PutHex64(infos.entries_fileoff); 6318 all_image_infos_payload.PutHex32(infos.entries_size); 6319 all_image_infos_payload.PutHex32(infos.unused); 6320 6321 // First create the structures for all of the segment name+vmaddr vectors 6322 // for each module, so we will know the size of them as we add the 6323 // module entries. 6324 std::vector<std::vector<segment_vmaddr>> modules_segment_vmaddrs; 6325 for (size_t i = 0; i < modules_count; i++) { 6326 ModuleSP module = modules.GetModuleAtIndex(i); 6327 6328 SectionList *sections = module->GetSectionList(); 6329 size_t sections_count = sections->GetSize(); 6330 std::vector<segment_vmaddr> segment_vmaddrs; 6331 for (size_t j = 0; j < sections_count; j++) { 6332 SectionSP section = sections->GetSectionAtIndex(j); 6333 if (!section->GetParent().get()) { 6334 addr_t vmaddr = section->GetLoadBaseAddress(&target); 6335 if (vmaddr == LLDB_INVALID_ADDRESS) 6336 continue; 6337 ConstString name = section->GetName(); 6338 segment_vmaddr seg_vmaddr; 6339 strncpy(seg_vmaddr.segname, name.AsCString(), 6340 sizeof(seg_vmaddr.segname)); 6341 seg_vmaddr.vmaddr = vmaddr; 6342 seg_vmaddr.unused = 0; 6343 segment_vmaddrs.push_back(seg_vmaddr); 6344 } 6345 } 6346 modules_segment_vmaddrs.push_back(segment_vmaddrs); 6347 } 6348 6349 offset_t size_of_vmaddr_structs = 0; 6350 for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) { 6351 size_of_vmaddr_structs += 6352 modules_segment_vmaddrs[i].size() * sizeof(segment_vmaddr); 6353 } 6354 6355 offset_t size_of_filepath_cstrings = 0; 6356 for (size_t i = 0; i < modules_count; i++) { 6357 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6358 size_of_filepath_cstrings += module_sp->GetFileSpec().GetPath().size() + 1; 6359 } 6360 6361 // Calculate the file offsets of our "all image infos" payload in the 6362 // corefile. initial_file_offset the original value passed in to this method. 6363 6364 offset_t start_of_entries = 6365 initial_file_offset + sizeof(all_image_infos_header); 6366 offset_t start_of_seg_vmaddrs = 6367 start_of_entries + sizeof(image_entry) * modules_count; 6368 offset_t start_of_filenames = start_of_seg_vmaddrs + size_of_vmaddr_structs; 6369 6370 offset_t final_file_offset = start_of_filenames + size_of_filepath_cstrings; 6371 6372 // Now write the one-per-module 'struct image_entry' into the 6373 // StringStream; keep track of where the struct segment_vmaddr 6374 // entries for each module will end up in the corefile. 6375 6376 offset_t current_string_offset = start_of_filenames; 6377 offset_t current_segaddrs_offset = start_of_seg_vmaddrs; 6378 std::vector<struct image_entry> image_entries; 6379 for (size_t i = 0; i < modules_count; i++) { 6380 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6381 6382 struct image_entry ent; 6383 memcpy(&ent.uuid, module_sp->GetUUID().GetBytes().data(), sizeof(ent.uuid)); 6384 if (modules_segment_vmaddrs[i].size() > 0) { 6385 ent.segment_count = modules_segment_vmaddrs[i].size(); 6386 ent.seg_addrs_offset = current_segaddrs_offset; 6387 } 6388 ent.filepath_offset = current_string_offset; 6389 ObjectFile *objfile = module_sp->GetObjectFile(); 6390 if (objfile) { 6391 Address base_addr(objfile->GetBaseAddress()); 6392 if (base_addr.IsValid()) { 6393 ent.load_address = base_addr.GetLoadAddress(&target); 6394 } 6395 } 6396 6397 all_image_infos_payload.PutHex64(ent.filepath_offset); 6398 all_image_infos_payload.PutRawBytes(ent.uuid, sizeof(ent.uuid)); 6399 all_image_infos_payload.PutHex64(ent.load_address); 6400 all_image_infos_payload.PutHex64(ent.seg_addrs_offset); 6401 all_image_infos_payload.PutHex32(ent.segment_count); 6402 6403 if (executing_uuids.find(module_sp->GetUUID().GetAsString()) != 6404 executing_uuids.end()) 6405 all_image_infos_payload.PutHex32(1); 6406 else 6407 all_image_infos_payload.PutHex32(0); 6408 6409 current_segaddrs_offset += ent.segment_count * sizeof(segment_vmaddr); 6410 current_string_offset += module_sp->GetFileSpec().GetPath().size() + 1; 6411 } 6412 6413 // Now write the struct segment_vmaddr entries into the StringStream. 6414 6415 for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) { 6416 if (modules_segment_vmaddrs[i].size() == 0) 6417 continue; 6418 for (struct segment_vmaddr segvm : modules_segment_vmaddrs[i]) { 6419 all_image_infos_payload.PutRawBytes(segvm.segname, sizeof(segvm.segname)); 6420 all_image_infos_payload.PutHex64(segvm.vmaddr); 6421 all_image_infos_payload.PutHex64(segvm.unused); 6422 } 6423 } 6424 6425 for (size_t i = 0; i < modules_count; i++) { 6426 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6427 std::string filepath = module_sp->GetFileSpec().GetPath(); 6428 all_image_infos_payload.PutRawBytes(filepath.data(), filepath.size() + 1); 6429 } 6430 6431 return final_file_offset; 6432 } 6433 6434 // Temp struct used to combine contiguous memory regions with 6435 // identical permissions. 6436 struct page_object { 6437 addr_t addr; 6438 addr_t size; 6439 uint32_t prot; 6440 }; 6441 6442 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6443 const FileSpec &outfile, 6444 lldb::SaveCoreStyle &core_style, Status &error) { 6445 if (!process_sp) 6446 return false; 6447 6448 // Default on macOS is to create a dirty-memory-only corefile. 6449 if (core_style == SaveCoreStyle::eSaveCoreUnspecified) { 6450 core_style = SaveCoreStyle::eSaveCoreDirtyOnly; 6451 } 6452 6453 Target &target = process_sp->GetTarget(); 6454 const ArchSpec target_arch = target.GetArchitecture(); 6455 const llvm::Triple &target_triple = target_arch.GetTriple(); 6456 if (target_triple.getVendor() == llvm::Triple::Apple && 6457 (target_triple.getOS() == llvm::Triple::MacOSX || 6458 target_triple.getOS() == llvm::Triple::IOS || 6459 target_triple.getOS() == llvm::Triple::WatchOS || 6460 target_triple.getOS() == llvm::Triple::TvOS)) { 6461 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) 6462 // { 6463 bool make_core = false; 6464 switch (target_arch.GetMachine()) { 6465 case llvm::Triple::aarch64: 6466 case llvm::Triple::aarch64_32: 6467 case llvm::Triple::arm: 6468 case llvm::Triple::thumb: 6469 case llvm::Triple::x86: 6470 case llvm::Triple::x86_64: 6471 make_core = true; 6472 break; 6473 default: 6474 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6475 target_triple.str().c_str()); 6476 break; 6477 } 6478 6479 if (make_core) { 6480 std::vector<llvm::MachO::segment_command_64> segment_load_commands; 6481 // uint32_t range_info_idx = 0; 6482 MemoryRegionInfo range_info; 6483 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6484 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6485 const ByteOrder byte_order = target_arch.GetByteOrder(); 6486 std::vector<page_object> pages_to_copy; 6487 6488 if (range_error.Success()) { 6489 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6490 // Calculate correct protections 6491 uint32_t prot = 0; 6492 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6493 prot |= VM_PROT_READ; 6494 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6495 prot |= VM_PROT_WRITE; 6496 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6497 prot |= VM_PROT_EXECUTE; 6498 6499 const addr_t addr = range_info.GetRange().GetRangeBase(); 6500 const addr_t size = range_info.GetRange().GetByteSize(); 6501 6502 if (size == 0) 6503 break; 6504 6505 bool include_this_region = true; 6506 bool dirty_pages_only = false; 6507 if (core_style == SaveCoreStyle::eSaveCoreStackOnly) { 6508 dirty_pages_only = true; 6509 if (range_info.IsStackMemory() != MemoryRegionInfo::eYes) { 6510 include_this_region = false; 6511 } 6512 } 6513 if (core_style == SaveCoreStyle::eSaveCoreDirtyOnly) { 6514 dirty_pages_only = true; 6515 } 6516 6517 if (prot != 0 && include_this_region) { 6518 addr_t pagesize = range_info.GetPageSize(); 6519 const llvm::Optional<std::vector<addr_t>> &dirty_page_list = 6520 range_info.GetDirtyPageList(); 6521 if (dirty_pages_only && dirty_page_list.hasValue()) { 6522 for (addr_t dirtypage : dirty_page_list.getValue()) { 6523 page_object obj; 6524 obj.addr = dirtypage; 6525 obj.size = pagesize; 6526 obj.prot = prot; 6527 pages_to_copy.push_back(obj); 6528 } 6529 } else { 6530 page_object obj; 6531 obj.addr = addr; 6532 obj.size = size; 6533 obj.prot = prot; 6534 pages_to_copy.push_back(obj); 6535 } 6536 } 6537 6538 range_error = process_sp->GetMemoryRegionInfo( 6539 range_info.GetRange().GetRangeEnd(), range_info); 6540 if (range_error.Fail()) 6541 break; 6542 } 6543 6544 // Combine contiguous entries that have the same 6545 // protections so we don't have an excess of 6546 // load commands. 6547 std::vector<page_object> combined_page_objects; 6548 page_object last_obj; 6549 last_obj.addr = LLDB_INVALID_ADDRESS; 6550 last_obj.size = 0; 6551 for (page_object obj : pages_to_copy) { 6552 if (last_obj.addr == LLDB_INVALID_ADDRESS) { 6553 last_obj = obj; 6554 continue; 6555 } 6556 if (last_obj.addr + last_obj.size == obj.addr && 6557 last_obj.prot == obj.prot) { 6558 last_obj.size += obj.size; 6559 continue; 6560 } 6561 combined_page_objects.push_back(last_obj); 6562 last_obj = obj; 6563 } 6564 // Add the last entry we were looking to combine 6565 // on to the array. 6566 if (last_obj.addr != LLDB_INVALID_ADDRESS && last_obj.size != 0) 6567 combined_page_objects.push_back(last_obj); 6568 6569 for (page_object obj : combined_page_objects) { 6570 uint32_t cmd_type = LC_SEGMENT_64; 6571 uint32_t segment_size = sizeof(llvm::MachO::segment_command_64); 6572 if (addr_byte_size == 4) { 6573 cmd_type = LC_SEGMENT; 6574 segment_size = sizeof(llvm::MachO::segment_command); 6575 } 6576 llvm::MachO::segment_command_64 segment = { 6577 cmd_type, // uint32_t cmd; 6578 segment_size, // uint32_t cmdsize; 6579 {0}, // char segname[16]; 6580 obj.addr, // uint64_t vmaddr; // uint32_t for 32-bit 6581 // Mach-O 6582 obj.size, // uint64_t vmsize; // uint32_t for 32-bit 6583 // Mach-O 6584 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6585 obj.size, // uint64_t filesize; // uint32_t for 32-bit 6586 // Mach-O 6587 obj.prot, // uint32_t maxprot; 6588 obj.prot, // uint32_t initprot; 6589 0, // uint32_t nsects; 6590 0}; // uint32_t flags; 6591 segment_load_commands.push_back(segment); 6592 } 6593 6594 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6595 6596 llvm::MachO::mach_header_64 mach_header; 6597 if (addr_byte_size == 8) { 6598 mach_header.magic = MH_MAGIC_64; 6599 } else { 6600 mach_header.magic = MH_MAGIC; 6601 } 6602 mach_header.cputype = target_arch.GetMachOCPUType(); 6603 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6604 mach_header.filetype = MH_CORE; 6605 mach_header.ncmds = segment_load_commands.size(); 6606 mach_header.flags = 0; 6607 mach_header.reserved = 0; 6608 ThreadList &thread_list = process_sp->GetThreadList(); 6609 const uint32_t num_threads = thread_list.GetSize(); 6610 6611 // Make an array of LC_THREAD data items. Each one contains the 6612 // contents of the LC_THREAD load command. The data doesn't contain 6613 // the load command + load command size, we will add the load command 6614 // and load command size as we emit the data. 6615 std::vector<StreamString> LC_THREAD_datas(num_threads); 6616 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6617 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6618 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6619 LC_THREAD_data.SetByteOrder(byte_order); 6620 } 6621 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) { 6622 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6623 if (thread_sp) { 6624 switch (mach_header.cputype) { 6625 case llvm::MachO::CPU_TYPE_ARM64: 6626 case llvm::MachO::CPU_TYPE_ARM64_32: 6627 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6628 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6629 break; 6630 6631 case llvm::MachO::CPU_TYPE_ARM: 6632 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6633 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6634 break; 6635 6636 case llvm::MachO::CPU_TYPE_I386: 6637 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6638 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6639 break; 6640 6641 case llvm::MachO::CPU_TYPE_X86_64: 6642 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6643 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6644 break; 6645 } 6646 } 6647 } 6648 6649 // The size of the load command is the size of the segments... 6650 if (addr_byte_size == 8) { 6651 mach_header.sizeofcmds = segment_load_commands.size() * 6652 sizeof(llvm::MachO::segment_command_64); 6653 } else { 6654 mach_header.sizeofcmds = segment_load_commands.size() * 6655 sizeof(llvm::MachO::segment_command); 6656 } 6657 6658 // and the size of all LC_THREAD load command 6659 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6660 ++mach_header.ncmds; 6661 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6662 } 6663 6664 // Bits will be set to indicate which bits are NOT used in 6665 // addressing in this process or 0 for unknown. 6666 uint64_t address_mask = process_sp->GetCodeAddressMask(); 6667 if (address_mask != 0) { 6668 // LC_NOTE "addrable bits" 6669 mach_header.ncmds++; 6670 mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); 6671 } 6672 6673 // LC_NOTE "all image infos" 6674 mach_header.ncmds++; 6675 mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); 6676 6677 // Write the mach header 6678 buffer.PutHex32(mach_header.magic); 6679 buffer.PutHex32(mach_header.cputype); 6680 buffer.PutHex32(mach_header.cpusubtype); 6681 buffer.PutHex32(mach_header.filetype); 6682 buffer.PutHex32(mach_header.ncmds); 6683 buffer.PutHex32(mach_header.sizeofcmds); 6684 buffer.PutHex32(mach_header.flags); 6685 if (addr_byte_size == 8) { 6686 buffer.PutHex32(mach_header.reserved); 6687 } 6688 6689 // Skip the mach header and all load commands and align to the next 6690 // 0x1000 byte boundary 6691 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6692 6693 file_offset = llvm::alignTo(file_offset, 16); 6694 std::vector<std::unique_ptr<LCNoteEntry>> lc_notes; 6695 6696 // Add "addrable bits" LC_NOTE when an address mask is available 6697 if (address_mask != 0) { 6698 std::unique_ptr<LCNoteEntry> addrable_bits_lcnote_up( 6699 new LCNoteEntry(addr_byte_size, byte_order)); 6700 addrable_bits_lcnote_up->name = "addrable bits"; 6701 addrable_bits_lcnote_up->payload_file_offset = file_offset; 6702 int bits = std::bitset<64>(~address_mask).count(); 6703 addrable_bits_lcnote_up->payload.PutHex32(3); // version 6704 addrable_bits_lcnote_up->payload.PutHex32( 6705 bits); // # of bits used for addressing 6706 addrable_bits_lcnote_up->payload.PutHex64(0); // unused 6707 6708 file_offset += addrable_bits_lcnote_up->payload.GetSize(); 6709 6710 lc_notes.push_back(std::move(addrable_bits_lcnote_up)); 6711 } 6712 6713 // Add "all image infos" LC_NOTE 6714 std::unique_ptr<LCNoteEntry> all_image_infos_lcnote_up( 6715 new LCNoteEntry(addr_byte_size, byte_order)); 6716 all_image_infos_lcnote_up->name = "all image infos"; 6717 all_image_infos_lcnote_up->payload_file_offset = file_offset; 6718 file_offset = CreateAllImageInfosPayload( 6719 process_sp, file_offset, all_image_infos_lcnote_up->payload, 6720 core_style); 6721 lc_notes.push_back(std::move(all_image_infos_lcnote_up)); 6722 6723 // Add LC_NOTE load commands 6724 for (auto &lcnote : lc_notes) { 6725 // Add the LC_NOTE load command to the file. 6726 buffer.PutHex32(LC_NOTE); 6727 buffer.PutHex32(sizeof(llvm::MachO::note_command)); 6728 char namebuf[16]; 6729 memset(namebuf, 0, sizeof(namebuf)); 6730 // this is the uncommon case where strncpy is exactly 6731 // the right one, doesn't need to be nul terminated. 6732 strncpy(namebuf, lcnote->name.c_str(), sizeof(namebuf)); 6733 buffer.PutRawBytes(namebuf, sizeof(namebuf)); 6734 buffer.PutHex64(lcnote->payload_file_offset); 6735 buffer.PutHex64(lcnote->payload.GetSize()); 6736 } 6737 6738 // Align to 4096-byte page boundary for the LC_SEGMENTs. 6739 file_offset = llvm::alignTo(file_offset, 4096); 6740 6741 for (auto &segment : segment_load_commands) { 6742 segment.fileoff = file_offset; 6743 file_offset += segment.filesize; 6744 } 6745 6746 // Write out all of the LC_THREAD load commands 6747 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6748 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6749 buffer.PutHex32(LC_THREAD); 6750 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6751 buffer.Write(LC_THREAD_data.GetString().data(), LC_THREAD_data_size); 6752 } 6753 6754 // Write out all of the segment load commands 6755 for (const auto &segment : segment_load_commands) { 6756 buffer.PutHex32(segment.cmd); 6757 buffer.PutHex32(segment.cmdsize); 6758 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6759 if (addr_byte_size == 8) { 6760 buffer.PutHex64(segment.vmaddr); 6761 buffer.PutHex64(segment.vmsize); 6762 buffer.PutHex64(segment.fileoff); 6763 buffer.PutHex64(segment.filesize); 6764 } else { 6765 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6766 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6767 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6768 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6769 } 6770 buffer.PutHex32(segment.maxprot); 6771 buffer.PutHex32(segment.initprot); 6772 buffer.PutHex32(segment.nsects); 6773 buffer.PutHex32(segment.flags); 6774 } 6775 6776 std::string core_file_path(outfile.GetPath()); 6777 auto core_file = FileSystem::Instance().Open( 6778 outfile, File::eOpenOptionWriteOnly | File::eOpenOptionTruncate | 6779 File::eOpenOptionCanCreate); 6780 if (!core_file) { 6781 error = core_file.takeError(); 6782 } else { 6783 // Read 1 page at a time 6784 uint8_t bytes[0x1000]; 6785 // Write the mach header and load commands out to the core file 6786 size_t bytes_written = buffer.GetString().size(); 6787 error = 6788 core_file.get()->Write(buffer.GetString().data(), bytes_written); 6789 if (error.Success()) { 6790 6791 for (auto &lcnote : lc_notes) { 6792 if (core_file.get()->SeekFromStart(lcnote->payload_file_offset) == 6793 -1) { 6794 error.SetErrorStringWithFormat("Unable to seek to corefile pos " 6795 "to write '%s' LC_NOTE payload", 6796 lcnote->name.c_str()); 6797 return false; 6798 } 6799 bytes_written = lcnote->payload.GetSize(); 6800 error = core_file.get()->Write(lcnote->payload.GetData(), 6801 bytes_written); 6802 if (!error.Success()) 6803 return false; 6804 } 6805 6806 // Now write the file data for all memory segments in the process 6807 for (const auto &segment : segment_load_commands) { 6808 if (core_file.get()->SeekFromStart(segment.fileoff) == -1) { 6809 error.SetErrorStringWithFormat( 6810 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6811 segment.fileoff, core_file_path.c_str()); 6812 break; 6813 } 6814 6815 target.GetDebugger().GetAsyncOutputStream()->Printf( 6816 "Saving %" PRId64 6817 " bytes of data for memory region at 0x%" PRIx64 "\n", 6818 segment.vmsize, segment.vmaddr); 6819 addr_t bytes_left = segment.vmsize; 6820 addr_t addr = segment.vmaddr; 6821 Status memory_read_error; 6822 while (bytes_left > 0 && error.Success()) { 6823 const size_t bytes_to_read = 6824 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6825 6826 // In a savecore setting, we don't really care about caching, 6827 // as the data is dumped and very likely never read again, 6828 // so we call ReadMemoryFromInferior to bypass it. 6829 const size_t bytes_read = process_sp->ReadMemoryFromInferior( 6830 addr, bytes, bytes_to_read, memory_read_error); 6831 6832 if (bytes_read == bytes_to_read) { 6833 size_t bytes_written = bytes_read; 6834 error = core_file.get()->Write(bytes, bytes_written); 6835 bytes_left -= bytes_read; 6836 addr += bytes_read; 6837 } else { 6838 // Some pages within regions are not readable, those should 6839 // be zero filled 6840 memset(bytes, 0, bytes_to_read); 6841 size_t bytes_written = bytes_to_read; 6842 error = core_file.get()->Write(bytes, bytes_written); 6843 bytes_left -= bytes_to_read; 6844 addr += bytes_to_read; 6845 } 6846 } 6847 } 6848 } 6849 } 6850 } else { 6851 error.SetErrorString( 6852 "process doesn't support getting memory region info"); 6853 } 6854 } 6855 return true; // This is the right plug to handle saving core files for 6856 // this process 6857 } 6858 return false; 6859 } 6860 6861 ObjectFileMachO::MachOCorefileAllImageInfos 6862 ObjectFileMachO::GetCorefileAllImageInfos() { 6863 MachOCorefileAllImageInfos image_infos; 6864 6865 // Look for an "all image infos" LC_NOTE. 6866 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 6867 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6868 const uint32_t cmd_offset = offset; 6869 llvm::MachO::load_command lc; 6870 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6871 break; 6872 if (lc.cmd == LC_NOTE) { 6873 char data_owner[17]; 6874 m_data.CopyData(offset, 16, data_owner); 6875 data_owner[16] = '\0'; 6876 offset += 16; 6877 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 6878 offset += 4; /* size unused */ 6879 6880 if (strcmp("all image infos", data_owner) == 0) { 6881 offset = fileoff; 6882 // Read the struct all_image_infos_header. 6883 uint32_t version = m_data.GetU32(&offset); 6884 if (version != 1) { 6885 return image_infos; 6886 } 6887 uint32_t imgcount = m_data.GetU32(&offset); 6888 uint64_t entries_fileoff = m_data.GetU64(&offset); 6889 offset += 4; // uint32_t entries_size; 6890 offset += 4; // uint32_t unused; 6891 6892 offset = entries_fileoff; 6893 for (uint32_t i = 0; i < imgcount; i++) { 6894 // Read the struct image_entry. 6895 offset_t filepath_offset = m_data.GetU64(&offset); 6896 uuid_t uuid; 6897 memcpy(&uuid, m_data.GetData(&offset, sizeof(uuid_t)), 6898 sizeof(uuid_t)); 6899 uint64_t load_address = m_data.GetU64(&offset); 6900 offset_t seg_addrs_offset = m_data.GetU64(&offset); 6901 uint32_t segment_count = m_data.GetU32(&offset); 6902 uint32_t currently_executing = m_data.GetU32(&offset); 6903 6904 MachOCorefileImageEntry image_entry; 6905 image_entry.filename = (const char *)m_data.GetCStr(&filepath_offset); 6906 image_entry.uuid = UUID::fromData(uuid, sizeof(uuid_t)); 6907 image_entry.load_address = load_address; 6908 image_entry.currently_executing = currently_executing; 6909 6910 offset_t seg_vmaddrs_offset = seg_addrs_offset; 6911 for (uint32_t j = 0; j < segment_count; j++) { 6912 char segname[17]; 6913 m_data.CopyData(seg_vmaddrs_offset, 16, segname); 6914 segname[16] = '\0'; 6915 seg_vmaddrs_offset += 16; 6916 uint64_t vmaddr = m_data.GetU64(&seg_vmaddrs_offset); 6917 seg_vmaddrs_offset += 8; /* unused */ 6918 6919 std::tuple<ConstString, addr_t> new_seg{ConstString(segname), 6920 vmaddr}; 6921 image_entry.segment_load_addresses.push_back(new_seg); 6922 } 6923 image_infos.all_image_infos.push_back(image_entry); 6924 } 6925 } else if (strcmp("load binary", data_owner) == 0) { 6926 uint32_t version = m_data.GetU32(&fileoff); 6927 if (version == 1) { 6928 uuid_t uuid; 6929 memcpy(&uuid, m_data.GetData(&fileoff, sizeof(uuid_t)), 6930 sizeof(uuid_t)); 6931 uint64_t load_address = m_data.GetU64(&fileoff); 6932 uint64_t slide = m_data.GetU64(&fileoff); 6933 std::string filename = m_data.GetCStr(&fileoff); 6934 6935 MachOCorefileImageEntry image_entry; 6936 image_entry.filename = filename; 6937 image_entry.uuid = UUID::fromData(uuid, sizeof(uuid_t)); 6938 image_entry.load_address = load_address; 6939 image_entry.slide = slide; 6940 image_infos.all_image_infos.push_back(image_entry); 6941 } 6942 } 6943 } 6944 offset = cmd_offset + lc.cmdsize; 6945 } 6946 6947 return image_infos; 6948 } 6949 6950 bool ObjectFileMachO::LoadCoreFileImages(lldb_private::Process &process) { 6951 MachOCorefileAllImageInfos image_infos = GetCorefileAllImageInfos(); 6952 Log *log = GetLog(LLDBLog::DynamicLoader); 6953 6954 ModuleList added_modules; 6955 for (const MachOCorefileImageEntry &image : image_infos.all_image_infos) { 6956 ModuleSpec module_spec; 6957 module_spec.GetUUID() = image.uuid; 6958 if (image.filename.empty()) { 6959 char namebuf[80]; 6960 if (image.load_address != LLDB_INVALID_ADDRESS) 6961 snprintf(namebuf, sizeof(namebuf), "mem-image-0x%" PRIx64, 6962 image.load_address); 6963 else 6964 snprintf(namebuf, sizeof(namebuf), "mem-image+0x%" PRIx64, image.slide); 6965 module_spec.GetFileSpec() = FileSpec(namebuf); 6966 } else { 6967 module_spec.GetFileSpec() = FileSpec(image.filename.c_str()); 6968 } 6969 if (image.currently_executing) { 6970 Status error; 6971 Symbols::DownloadObjectAndSymbolFile(module_spec, error, true); 6972 if (FileSystem::Instance().Exists(module_spec.GetFileSpec())) { 6973 process.GetTarget().GetOrCreateModule(module_spec, false); 6974 } 6975 } 6976 Status error; 6977 ModuleSP module_sp = 6978 process.GetTarget().GetOrCreateModule(module_spec, false, &error); 6979 if (!module_sp.get() || !module_sp->GetObjectFile()) { 6980 if (image.load_address != LLDB_INVALID_ADDRESS) { 6981 module_sp = process.ReadModuleFromMemory(module_spec.GetFileSpec(), 6982 image.load_address); 6983 } 6984 } 6985 if (module_sp.get()) { 6986 // Will call ModulesDidLoad with all modules once they've all 6987 // been added to the Target with load addresses. Don't notify 6988 // here, before the load address is set. 6989 const bool notify = false; 6990 process.GetTarget().GetImages().AppendIfNeeded(module_sp, notify); 6991 added_modules.Append(module_sp, notify); 6992 if (image.segment_load_addresses.size() > 0) { 6993 if (log) { 6994 std::string uuidstr = image.uuid.GetAsString(); 6995 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 6996 "UUID %s with section load addresses", 6997 image.filename.c_str(), uuidstr.c_str()); 6998 } 6999 for (auto name_vmaddr_tuple : image.segment_load_addresses) { 7000 SectionList *sectlist = module_sp->GetObjectFile()->GetSectionList(); 7001 if (sectlist) { 7002 SectionSP sect_sp = 7003 sectlist->FindSectionByName(std::get<0>(name_vmaddr_tuple)); 7004 if (sect_sp) { 7005 process.GetTarget().SetSectionLoadAddress( 7006 sect_sp, std::get<1>(name_vmaddr_tuple)); 7007 } 7008 } 7009 } 7010 } else if (image.load_address != LLDB_INVALID_ADDRESS) { 7011 if (log) { 7012 std::string uuidstr = image.uuid.GetAsString(); 7013 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 7014 "UUID %s with load address 0x%" PRIx64, 7015 image.filename.c_str(), uuidstr.c_str(), 7016 image.load_address); 7017 } 7018 const bool address_is_slide = false; 7019 bool changed = false; 7020 module_sp->SetLoadAddress(process.GetTarget(), image.load_address, 7021 address_is_slide, changed); 7022 } else if (image.slide != 0) { 7023 if (log) { 7024 std::string uuidstr = image.uuid.GetAsString(); 7025 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 7026 "UUID %s with slide amount 0x%" PRIx64, 7027 image.filename.c_str(), uuidstr.c_str(), image.slide); 7028 } 7029 const bool address_is_slide = true; 7030 bool changed = false; 7031 module_sp->SetLoadAddress(process.GetTarget(), image.slide, 7032 address_is_slide, changed); 7033 } else { 7034 if (log) { 7035 std::string uuidstr = image.uuid.GetAsString(); 7036 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 7037 "UUID %s at its file address, no slide applied", 7038 image.filename.c_str(), uuidstr.c_str()); 7039 } 7040 const bool address_is_slide = true; 7041 bool changed = false; 7042 module_sp->SetLoadAddress(process.GetTarget(), 0, address_is_slide, 7043 changed); 7044 } 7045 } 7046 } 7047 if (added_modules.GetSize() > 0) { 7048 process.GetTarget().ModulesDidLoad(added_modules); 7049 process.Flush(); 7050 return true; 7051 } 7052 return false; 7053 } 7054