1 //===-- ObjectFileMachO.cpp -----------------------------------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 9 #include "llvm/ADT/ScopeExit.h" 10 #include "llvm/ADT/StringRef.h" 11 12 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 13 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 14 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 15 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 16 #include "lldb/Core/Debugger.h" 17 #include "lldb/Core/FileSpecList.h" 18 #include "lldb/Core/Module.h" 19 #include "lldb/Core/ModuleSpec.h" 20 #include "lldb/Core/PluginManager.h" 21 #include "lldb/Core/Progress.h" 22 #include "lldb/Core/Section.h" 23 #include "lldb/Core/StreamFile.h" 24 #include "lldb/Host/Host.h" 25 #include "lldb/Symbol/DWARFCallFrameInfo.h" 26 #include "lldb/Symbol/LocateSymbolFile.h" 27 #include "lldb/Symbol/ObjectFile.h" 28 #include "lldb/Target/DynamicLoader.h" 29 #include "lldb/Target/MemoryRegionInfo.h" 30 #include "lldb/Target/Platform.h" 31 #include "lldb/Target/Process.h" 32 #include "lldb/Target/SectionLoadList.h" 33 #include "lldb/Target/Target.h" 34 #include "lldb/Target/Thread.h" 35 #include "lldb/Target/ThreadList.h" 36 #include "lldb/Utility/ArchSpec.h" 37 #include "lldb/Utility/DataBuffer.h" 38 #include "lldb/Utility/FileSpec.h" 39 #include "lldb/Utility/LLDBLog.h" 40 #include "lldb/Utility/Log.h" 41 #include "lldb/Utility/RangeMap.h" 42 #include "lldb/Utility/RegisterValue.h" 43 #include "lldb/Utility/Status.h" 44 #include "lldb/Utility/StreamString.h" 45 #include "lldb/Utility/Timer.h" 46 #include "lldb/Utility/UUID.h" 47 48 #include "lldb/Host/SafeMachO.h" 49 50 #include "llvm/ADT/DenseSet.h" 51 #include "llvm/Support/FormatVariadic.h" 52 #include "llvm/Support/MemoryBuffer.h" 53 54 #include "ObjectFileMachO.h" 55 56 #if defined(__APPLE__) 57 #include <TargetConditionals.h> 58 // GetLLDBSharedCacheUUID() needs to call dlsym() 59 #include <dlfcn.h> 60 #include <mach/mach_init.h> 61 #include <mach/vm_map.h> 62 #include <lldb/Host/SafeMachO.h> 63 #endif 64 65 #ifndef __APPLE__ 66 #include "Utility/UuidCompatibility.h" 67 #else 68 #include <uuid/uuid.h> 69 #endif 70 71 #include <bitset> 72 #include <memory> 73 74 // Unfortunately the signpost header pulls in the system MachO header, too. 75 #ifdef CPU_TYPE_ARM 76 #undef CPU_TYPE_ARM 77 #endif 78 #ifdef CPU_TYPE_ARM64 79 #undef CPU_TYPE_ARM64 80 #endif 81 #ifdef CPU_TYPE_ARM64_32 82 #undef CPU_TYPE_ARM64_32 83 #endif 84 #ifdef CPU_TYPE_I386 85 #undef CPU_TYPE_I386 86 #endif 87 #ifdef CPU_TYPE_X86_64 88 #undef CPU_TYPE_X86_64 89 #endif 90 #ifdef MH_DYLINKER 91 #undef MH_DYLINKER 92 #endif 93 #ifdef MH_OBJECT 94 #undef MH_OBJECT 95 #endif 96 #ifdef LC_VERSION_MIN_MACOSX 97 #undef LC_VERSION_MIN_MACOSX 98 #endif 99 #ifdef LC_VERSION_MIN_IPHONEOS 100 #undef LC_VERSION_MIN_IPHONEOS 101 #endif 102 #ifdef LC_VERSION_MIN_TVOS 103 #undef LC_VERSION_MIN_TVOS 104 #endif 105 #ifdef LC_VERSION_MIN_WATCHOS 106 #undef LC_VERSION_MIN_WATCHOS 107 #endif 108 #ifdef LC_BUILD_VERSION 109 #undef LC_BUILD_VERSION 110 #endif 111 #ifdef PLATFORM_MACOS 112 #undef PLATFORM_MACOS 113 #endif 114 #ifdef PLATFORM_MACCATALYST 115 #undef PLATFORM_MACCATALYST 116 #endif 117 #ifdef PLATFORM_IOS 118 #undef PLATFORM_IOS 119 #endif 120 #ifdef PLATFORM_IOSSIMULATOR 121 #undef PLATFORM_IOSSIMULATOR 122 #endif 123 #ifdef PLATFORM_TVOS 124 #undef PLATFORM_TVOS 125 #endif 126 #ifdef PLATFORM_TVOSSIMULATOR 127 #undef PLATFORM_TVOSSIMULATOR 128 #endif 129 #ifdef PLATFORM_WATCHOS 130 #undef PLATFORM_WATCHOS 131 #endif 132 #ifdef PLATFORM_WATCHOSSIMULATOR 133 #undef PLATFORM_WATCHOSSIMULATOR 134 #endif 135 136 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 137 using namespace lldb; 138 using namespace lldb_private; 139 using namespace llvm::MachO; 140 141 LLDB_PLUGIN_DEFINE(ObjectFileMachO) 142 143 // Some structure definitions needed for parsing the dyld shared cache files 144 // found on iOS devices. 145 146 struct lldb_copy_dyld_cache_header_v1 { 147 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 148 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 149 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 150 uint32_t imagesOffset; 151 uint32_t imagesCount; 152 uint64_t dyldBaseAddress; 153 uint64_t codeSignatureOffset; 154 uint64_t codeSignatureSize; 155 uint64_t slideInfoOffset; 156 uint64_t slideInfoSize; 157 uint64_t localSymbolsOffset; 158 uint64_t localSymbolsSize; 159 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 160 // and later 161 }; 162 163 static void PrintRegisterValue(RegisterContext *reg_ctx, const char *name, 164 const char *alt_name, size_t reg_byte_size, 165 Stream &data) { 166 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 167 if (reg_info == nullptr) 168 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 169 if (reg_info) { 170 lldb_private::RegisterValue reg_value; 171 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 172 if (reg_info->byte_size >= reg_byte_size) 173 data.Write(reg_value.GetBytes(), reg_byte_size); 174 else { 175 data.Write(reg_value.GetBytes(), reg_info->byte_size); 176 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; ++i) 177 data.PutChar(0); 178 } 179 return; 180 } 181 } 182 // Just write zeros if all else fails 183 for (size_t i = 0; i < reg_byte_size; ++i) 184 data.PutChar(0); 185 } 186 187 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 188 public: 189 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 190 const DataExtractor &data) 191 : RegisterContextDarwin_x86_64(thread, 0) { 192 SetRegisterDataFrom_LC_THREAD(data); 193 } 194 195 void InvalidateAllRegisters() override { 196 // Do nothing... registers are always valid... 197 } 198 199 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 200 lldb::offset_t offset = 0; 201 SetError(GPRRegSet, Read, -1); 202 SetError(FPURegSet, Read, -1); 203 SetError(EXCRegSet, Read, -1); 204 bool done = false; 205 206 while (!done) { 207 int flavor = data.GetU32(&offset); 208 if (flavor == 0) 209 done = true; 210 else { 211 uint32_t i; 212 uint32_t count = data.GetU32(&offset); 213 switch (flavor) { 214 case GPRRegSet: 215 for (i = 0; i < count; ++i) 216 (&gpr.rax)[i] = data.GetU64(&offset); 217 SetError(GPRRegSet, Read, 0); 218 done = true; 219 220 break; 221 case FPURegSet: 222 // TODO: fill in FPU regs.... 223 // SetError (FPURegSet, Read, -1); 224 done = true; 225 226 break; 227 case EXCRegSet: 228 exc.trapno = data.GetU32(&offset); 229 exc.err = data.GetU32(&offset); 230 exc.faultvaddr = data.GetU64(&offset); 231 SetError(EXCRegSet, Read, 0); 232 done = true; 233 break; 234 case 7: 235 case 8: 236 case 9: 237 // fancy flavors that encapsulate of the above flavors... 238 break; 239 240 default: 241 done = true; 242 break; 243 } 244 } 245 } 246 } 247 248 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 249 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 250 if (reg_ctx_sp) { 251 RegisterContext *reg_ctx = reg_ctx_sp.get(); 252 253 data.PutHex32(GPRRegSet); // Flavor 254 data.PutHex32(GPRWordCount); 255 PrintRegisterValue(reg_ctx, "rax", nullptr, 8, data); 256 PrintRegisterValue(reg_ctx, "rbx", nullptr, 8, data); 257 PrintRegisterValue(reg_ctx, "rcx", nullptr, 8, data); 258 PrintRegisterValue(reg_ctx, "rdx", nullptr, 8, data); 259 PrintRegisterValue(reg_ctx, "rdi", nullptr, 8, data); 260 PrintRegisterValue(reg_ctx, "rsi", nullptr, 8, data); 261 PrintRegisterValue(reg_ctx, "rbp", nullptr, 8, data); 262 PrintRegisterValue(reg_ctx, "rsp", nullptr, 8, data); 263 PrintRegisterValue(reg_ctx, "r8", nullptr, 8, data); 264 PrintRegisterValue(reg_ctx, "r9", nullptr, 8, data); 265 PrintRegisterValue(reg_ctx, "r10", nullptr, 8, data); 266 PrintRegisterValue(reg_ctx, "r11", nullptr, 8, data); 267 PrintRegisterValue(reg_ctx, "r12", nullptr, 8, data); 268 PrintRegisterValue(reg_ctx, "r13", nullptr, 8, data); 269 PrintRegisterValue(reg_ctx, "r14", nullptr, 8, data); 270 PrintRegisterValue(reg_ctx, "r15", nullptr, 8, data); 271 PrintRegisterValue(reg_ctx, "rip", nullptr, 8, data); 272 PrintRegisterValue(reg_ctx, "rflags", nullptr, 8, data); 273 PrintRegisterValue(reg_ctx, "cs", nullptr, 8, data); 274 PrintRegisterValue(reg_ctx, "fs", nullptr, 8, data); 275 PrintRegisterValue(reg_ctx, "gs", nullptr, 8, data); 276 277 // // Write out the FPU registers 278 // const size_t fpu_byte_size = sizeof(FPU); 279 // size_t bytes_written = 0; 280 // data.PutHex32 (FPURegSet); 281 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 282 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 283 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 284 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 285 // data); // uint16_t fcw; // "fctrl" 286 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 287 // data); // uint16_t fsw; // "fstat" 288 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 289 // data); // uint8_t ftw; // "ftag" 290 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 291 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 292 // data); // uint16_t fop; // "fop" 293 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 294 // data); // uint32_t ip; // "fioff" 295 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 296 // data); // uint16_t cs; // "fiseg" 297 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 298 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 299 // data); // uint32_t dp; // "fooff" 300 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 301 // data); // uint16_t ds; // "foseg" 302 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 303 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 304 // data); // uint32_t mxcsr; 305 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 306 // 4, data);// uint32_t mxcsrmask; 307 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 308 // sizeof(MMSReg), data); 309 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 310 // sizeof(MMSReg), data); 311 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 312 // sizeof(MMSReg), data); 313 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 314 // sizeof(MMSReg), data); 315 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 316 // sizeof(MMSReg), data); 317 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 318 // sizeof(MMSReg), data); 319 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 320 // sizeof(MMSReg), data); 321 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 322 // sizeof(MMSReg), data); 323 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 324 // sizeof(XMMReg), data); 325 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 326 // sizeof(XMMReg), data); 327 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 328 // sizeof(XMMReg), data); 329 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 330 // sizeof(XMMReg), data); 331 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 332 // sizeof(XMMReg), data); 333 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 334 // sizeof(XMMReg), data); 335 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 336 // sizeof(XMMReg), data); 337 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 338 // sizeof(XMMReg), data); 339 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 340 // sizeof(XMMReg), data); 341 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 342 // sizeof(XMMReg), data); 343 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 344 // sizeof(XMMReg), data); 345 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 346 // sizeof(XMMReg), data); 347 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 348 // sizeof(XMMReg), data); 349 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 350 // sizeof(XMMReg), data); 351 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 352 // sizeof(XMMReg), data); 353 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 354 // sizeof(XMMReg), data); 355 // 356 // // Fill rest with zeros 357 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 358 // i) 359 // data.PutChar(0); 360 361 // Write out the EXC registers 362 data.PutHex32(EXCRegSet); 363 data.PutHex32(EXCWordCount); 364 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 365 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 366 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 8, data); 367 return true; 368 } 369 return false; 370 } 371 372 protected: 373 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 374 375 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 376 377 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 378 379 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 380 return 0; 381 } 382 383 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 384 return 0; 385 } 386 387 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 388 return 0; 389 } 390 }; 391 392 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 393 public: 394 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 395 const DataExtractor &data) 396 : RegisterContextDarwin_i386(thread, 0) { 397 SetRegisterDataFrom_LC_THREAD(data); 398 } 399 400 void InvalidateAllRegisters() override { 401 // Do nothing... registers are always valid... 402 } 403 404 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 405 lldb::offset_t offset = 0; 406 SetError(GPRRegSet, Read, -1); 407 SetError(FPURegSet, Read, -1); 408 SetError(EXCRegSet, Read, -1); 409 bool done = false; 410 411 while (!done) { 412 int flavor = data.GetU32(&offset); 413 if (flavor == 0) 414 done = true; 415 else { 416 uint32_t i; 417 uint32_t count = data.GetU32(&offset); 418 switch (flavor) { 419 case GPRRegSet: 420 for (i = 0; i < count; ++i) 421 (&gpr.eax)[i] = data.GetU32(&offset); 422 SetError(GPRRegSet, Read, 0); 423 done = true; 424 425 break; 426 case FPURegSet: 427 // TODO: fill in FPU regs.... 428 // SetError (FPURegSet, Read, -1); 429 done = true; 430 431 break; 432 case EXCRegSet: 433 exc.trapno = data.GetU32(&offset); 434 exc.err = data.GetU32(&offset); 435 exc.faultvaddr = data.GetU32(&offset); 436 SetError(EXCRegSet, Read, 0); 437 done = true; 438 break; 439 case 7: 440 case 8: 441 case 9: 442 // fancy flavors that encapsulate of the above flavors... 443 break; 444 445 default: 446 done = true; 447 break; 448 } 449 } 450 } 451 } 452 453 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 454 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 455 if (reg_ctx_sp) { 456 RegisterContext *reg_ctx = reg_ctx_sp.get(); 457 458 data.PutHex32(GPRRegSet); // Flavor 459 data.PutHex32(GPRWordCount); 460 PrintRegisterValue(reg_ctx, "eax", nullptr, 4, data); 461 PrintRegisterValue(reg_ctx, "ebx", nullptr, 4, data); 462 PrintRegisterValue(reg_ctx, "ecx", nullptr, 4, data); 463 PrintRegisterValue(reg_ctx, "edx", nullptr, 4, data); 464 PrintRegisterValue(reg_ctx, "edi", nullptr, 4, data); 465 PrintRegisterValue(reg_ctx, "esi", nullptr, 4, data); 466 PrintRegisterValue(reg_ctx, "ebp", nullptr, 4, data); 467 PrintRegisterValue(reg_ctx, "esp", nullptr, 4, data); 468 PrintRegisterValue(reg_ctx, "ss", nullptr, 4, data); 469 PrintRegisterValue(reg_ctx, "eflags", nullptr, 4, data); 470 PrintRegisterValue(reg_ctx, "eip", nullptr, 4, data); 471 PrintRegisterValue(reg_ctx, "cs", nullptr, 4, data); 472 PrintRegisterValue(reg_ctx, "ds", nullptr, 4, data); 473 PrintRegisterValue(reg_ctx, "es", nullptr, 4, data); 474 PrintRegisterValue(reg_ctx, "fs", nullptr, 4, data); 475 PrintRegisterValue(reg_ctx, "gs", nullptr, 4, data); 476 477 // Write out the EXC registers 478 data.PutHex32(EXCRegSet); 479 data.PutHex32(EXCWordCount); 480 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 481 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 482 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 4, data); 483 return true; 484 } 485 return false; 486 } 487 488 protected: 489 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 490 491 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 492 493 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 494 495 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 496 return 0; 497 } 498 499 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 500 return 0; 501 } 502 503 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 504 return 0; 505 } 506 }; 507 508 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 509 public: 510 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 511 const DataExtractor &data) 512 : RegisterContextDarwin_arm(thread, 0) { 513 SetRegisterDataFrom_LC_THREAD(data); 514 } 515 516 void InvalidateAllRegisters() override { 517 // Do nothing... registers are always valid... 518 } 519 520 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 521 lldb::offset_t offset = 0; 522 SetError(GPRRegSet, Read, -1); 523 SetError(FPURegSet, Read, -1); 524 SetError(EXCRegSet, Read, -1); 525 bool done = false; 526 527 while (!done) { 528 int flavor = data.GetU32(&offset); 529 uint32_t count = data.GetU32(&offset); 530 lldb::offset_t next_thread_state = offset + (count * 4); 531 switch (flavor) { 532 case GPRAltRegSet: 533 case GPRRegSet: 534 // On ARM, the CPSR register is also included in the count but it is 535 // not included in gpr.r so loop until (count-1). 536 for (uint32_t i = 0; i < (count - 1); ++i) { 537 gpr.r[i] = data.GetU32(&offset); 538 } 539 // Save cpsr explicitly. 540 gpr.cpsr = data.GetU32(&offset); 541 542 SetError(GPRRegSet, Read, 0); 543 offset = next_thread_state; 544 break; 545 546 case FPURegSet: { 547 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 548 const int fpu_reg_buf_size = sizeof(fpu.floats); 549 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 550 fpu_reg_buf) == fpu_reg_buf_size) { 551 offset += fpu_reg_buf_size; 552 fpu.fpscr = data.GetU32(&offset); 553 SetError(FPURegSet, Read, 0); 554 } else { 555 done = true; 556 } 557 } 558 offset = next_thread_state; 559 break; 560 561 case EXCRegSet: 562 if (count == 3) { 563 exc.exception = data.GetU32(&offset); 564 exc.fsr = data.GetU32(&offset); 565 exc.far = data.GetU32(&offset); 566 SetError(EXCRegSet, Read, 0); 567 } 568 done = true; 569 offset = next_thread_state; 570 break; 571 572 // Unknown register set flavor, stop trying to parse. 573 default: 574 done = true; 575 } 576 } 577 } 578 579 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 580 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 581 if (reg_ctx_sp) { 582 RegisterContext *reg_ctx = reg_ctx_sp.get(); 583 584 data.PutHex32(GPRRegSet); // Flavor 585 data.PutHex32(GPRWordCount); 586 PrintRegisterValue(reg_ctx, "r0", nullptr, 4, data); 587 PrintRegisterValue(reg_ctx, "r1", nullptr, 4, data); 588 PrintRegisterValue(reg_ctx, "r2", nullptr, 4, data); 589 PrintRegisterValue(reg_ctx, "r3", nullptr, 4, data); 590 PrintRegisterValue(reg_ctx, "r4", nullptr, 4, data); 591 PrintRegisterValue(reg_ctx, "r5", nullptr, 4, data); 592 PrintRegisterValue(reg_ctx, "r6", nullptr, 4, data); 593 PrintRegisterValue(reg_ctx, "r7", nullptr, 4, data); 594 PrintRegisterValue(reg_ctx, "r8", nullptr, 4, data); 595 PrintRegisterValue(reg_ctx, "r9", nullptr, 4, data); 596 PrintRegisterValue(reg_ctx, "r10", nullptr, 4, data); 597 PrintRegisterValue(reg_ctx, "r11", nullptr, 4, data); 598 PrintRegisterValue(reg_ctx, "r12", nullptr, 4, data); 599 PrintRegisterValue(reg_ctx, "sp", nullptr, 4, data); 600 PrintRegisterValue(reg_ctx, "lr", nullptr, 4, data); 601 PrintRegisterValue(reg_ctx, "pc", nullptr, 4, data); 602 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 603 604 // Write out the EXC registers 605 // data.PutHex32 (EXCRegSet); 606 // data.PutHex32 (EXCWordCount); 607 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 608 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 609 // WriteRegister (reg_ctx, "far", NULL, 4, data); 610 return true; 611 } 612 return false; 613 } 614 615 protected: 616 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 617 618 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 619 620 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 621 622 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 623 624 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 625 return 0; 626 } 627 628 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 629 return 0; 630 } 631 632 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 633 return 0; 634 } 635 636 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 637 return -1; 638 } 639 }; 640 641 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 642 public: 643 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 644 const DataExtractor &data) 645 : RegisterContextDarwin_arm64(thread, 0) { 646 SetRegisterDataFrom_LC_THREAD(data); 647 } 648 649 void InvalidateAllRegisters() override { 650 // Do nothing... registers are always valid... 651 } 652 653 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 654 lldb::offset_t offset = 0; 655 SetError(GPRRegSet, Read, -1); 656 SetError(FPURegSet, Read, -1); 657 SetError(EXCRegSet, Read, -1); 658 bool done = false; 659 while (!done) { 660 int flavor = data.GetU32(&offset); 661 uint32_t count = data.GetU32(&offset); 662 lldb::offset_t next_thread_state = offset + (count * 4); 663 switch (flavor) { 664 case GPRRegSet: 665 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 666 // 32-bit register) 667 if (count >= (33 * 2) + 1) { 668 for (uint32_t i = 0; i < 29; ++i) 669 gpr.x[i] = data.GetU64(&offset); 670 gpr.fp = data.GetU64(&offset); 671 gpr.lr = data.GetU64(&offset); 672 gpr.sp = data.GetU64(&offset); 673 gpr.pc = data.GetU64(&offset); 674 gpr.cpsr = data.GetU32(&offset); 675 SetError(GPRRegSet, Read, 0); 676 } 677 offset = next_thread_state; 678 break; 679 case FPURegSet: { 680 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 681 const int fpu_reg_buf_size = sizeof(fpu); 682 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 683 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 684 fpu_reg_buf) == fpu_reg_buf_size) { 685 SetError(FPURegSet, Read, 0); 686 } else { 687 done = true; 688 } 689 } 690 offset = next_thread_state; 691 break; 692 case EXCRegSet: 693 if (count == 4) { 694 exc.far = data.GetU64(&offset); 695 exc.esr = data.GetU32(&offset); 696 exc.exception = data.GetU32(&offset); 697 SetError(EXCRegSet, Read, 0); 698 } 699 offset = next_thread_state; 700 break; 701 default: 702 done = true; 703 break; 704 } 705 } 706 } 707 708 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 709 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 710 if (reg_ctx_sp) { 711 RegisterContext *reg_ctx = reg_ctx_sp.get(); 712 713 data.PutHex32(GPRRegSet); // Flavor 714 data.PutHex32(GPRWordCount); 715 PrintRegisterValue(reg_ctx, "x0", nullptr, 8, data); 716 PrintRegisterValue(reg_ctx, "x1", nullptr, 8, data); 717 PrintRegisterValue(reg_ctx, "x2", nullptr, 8, data); 718 PrintRegisterValue(reg_ctx, "x3", nullptr, 8, data); 719 PrintRegisterValue(reg_ctx, "x4", nullptr, 8, data); 720 PrintRegisterValue(reg_ctx, "x5", nullptr, 8, data); 721 PrintRegisterValue(reg_ctx, "x6", nullptr, 8, data); 722 PrintRegisterValue(reg_ctx, "x7", nullptr, 8, data); 723 PrintRegisterValue(reg_ctx, "x8", nullptr, 8, data); 724 PrintRegisterValue(reg_ctx, "x9", nullptr, 8, data); 725 PrintRegisterValue(reg_ctx, "x10", nullptr, 8, data); 726 PrintRegisterValue(reg_ctx, "x11", nullptr, 8, data); 727 PrintRegisterValue(reg_ctx, "x12", nullptr, 8, data); 728 PrintRegisterValue(reg_ctx, "x13", nullptr, 8, data); 729 PrintRegisterValue(reg_ctx, "x14", nullptr, 8, data); 730 PrintRegisterValue(reg_ctx, "x15", nullptr, 8, data); 731 PrintRegisterValue(reg_ctx, "x16", nullptr, 8, data); 732 PrintRegisterValue(reg_ctx, "x17", nullptr, 8, data); 733 PrintRegisterValue(reg_ctx, "x18", nullptr, 8, data); 734 PrintRegisterValue(reg_ctx, "x19", nullptr, 8, data); 735 PrintRegisterValue(reg_ctx, "x20", nullptr, 8, data); 736 PrintRegisterValue(reg_ctx, "x21", nullptr, 8, data); 737 PrintRegisterValue(reg_ctx, "x22", nullptr, 8, data); 738 PrintRegisterValue(reg_ctx, "x23", nullptr, 8, data); 739 PrintRegisterValue(reg_ctx, "x24", nullptr, 8, data); 740 PrintRegisterValue(reg_ctx, "x25", nullptr, 8, data); 741 PrintRegisterValue(reg_ctx, "x26", nullptr, 8, data); 742 PrintRegisterValue(reg_ctx, "x27", nullptr, 8, data); 743 PrintRegisterValue(reg_ctx, "x28", nullptr, 8, data); 744 PrintRegisterValue(reg_ctx, "fp", nullptr, 8, data); 745 PrintRegisterValue(reg_ctx, "lr", nullptr, 8, data); 746 PrintRegisterValue(reg_ctx, "sp", nullptr, 8, data); 747 PrintRegisterValue(reg_ctx, "pc", nullptr, 8, data); 748 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 749 data.PutHex32(0); // uint32_t pad at the end 750 751 // Write out the EXC registers 752 data.PutHex32(EXCRegSet); 753 data.PutHex32(EXCWordCount); 754 PrintRegisterValue(reg_ctx, "far", nullptr, 8, data); 755 PrintRegisterValue(reg_ctx, "esr", nullptr, 4, data); 756 PrintRegisterValue(reg_ctx, "exception", nullptr, 4, data); 757 return true; 758 } 759 return false; 760 } 761 762 protected: 763 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 764 765 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 766 767 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 768 769 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 770 771 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 772 return 0; 773 } 774 775 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 776 return 0; 777 } 778 779 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 780 return 0; 781 } 782 783 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 784 return -1; 785 } 786 }; 787 788 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 789 switch (magic) { 790 case MH_MAGIC: 791 case MH_CIGAM: 792 return sizeof(struct llvm::MachO::mach_header); 793 794 case MH_MAGIC_64: 795 case MH_CIGAM_64: 796 return sizeof(struct llvm::MachO::mach_header_64); 797 break; 798 799 default: 800 break; 801 } 802 return 0; 803 } 804 805 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 806 807 char ObjectFileMachO::ID; 808 809 void ObjectFileMachO::Initialize() { 810 PluginManager::RegisterPlugin( 811 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 812 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 813 } 814 815 void ObjectFileMachO::Terminate() { 816 PluginManager::UnregisterPlugin(CreateInstance); 817 } 818 819 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 820 DataBufferSP data_sp, 821 lldb::offset_t data_offset, 822 const FileSpec *file, 823 lldb::offset_t file_offset, 824 lldb::offset_t length) { 825 if (!data_sp) { 826 data_sp = MapFileData(*file, length, file_offset); 827 if (!data_sp) 828 return nullptr; 829 data_offset = 0; 830 } 831 832 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 833 return nullptr; 834 835 // Update the data to contain the entire file if it doesn't already 836 if (data_sp->GetByteSize() < length) { 837 data_sp = MapFileData(*file, length, file_offset); 838 if (!data_sp) 839 return nullptr; 840 data_offset = 0; 841 } 842 auto objfile_up = std::make_unique<ObjectFileMachO>( 843 module_sp, data_sp, data_offset, file, file_offset, length); 844 if (!objfile_up || !objfile_up->ParseHeader()) 845 return nullptr; 846 847 return objfile_up.release(); 848 } 849 850 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 851 const lldb::ModuleSP &module_sp, DataBufferSP data_sp, 852 const ProcessSP &process_sp, lldb::addr_t header_addr) { 853 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 854 std::unique_ptr<ObjectFile> objfile_up( 855 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 856 if (objfile_up.get() && objfile_up->ParseHeader()) 857 return objfile_up.release(); 858 } 859 return nullptr; 860 } 861 862 size_t ObjectFileMachO::GetModuleSpecifications( 863 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 864 lldb::offset_t data_offset, lldb::offset_t file_offset, 865 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 866 const size_t initial_count = specs.GetSize(); 867 868 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 869 DataExtractor data; 870 data.SetData(data_sp); 871 llvm::MachO::mach_header header; 872 if (ParseHeader(data, &data_offset, header)) { 873 size_t header_and_load_cmds = 874 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 875 if (header_and_load_cmds >= data_sp->GetByteSize()) { 876 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 877 data.SetData(data_sp); 878 data_offset = MachHeaderSizeFromMagic(header.magic); 879 } 880 if (data_sp) { 881 ModuleSpec base_spec; 882 base_spec.GetFileSpec() = file; 883 base_spec.SetObjectOffset(file_offset); 884 base_spec.SetObjectSize(length); 885 GetAllArchSpecs(header, data, data_offset, base_spec, specs); 886 } 887 } 888 } 889 return specs.GetSize() - initial_count; 890 } 891 892 ConstString ObjectFileMachO::GetSegmentNameTEXT() { 893 static ConstString g_segment_name_TEXT("__TEXT"); 894 return g_segment_name_TEXT; 895 } 896 897 ConstString ObjectFileMachO::GetSegmentNameDATA() { 898 static ConstString g_segment_name_DATA("__DATA"); 899 return g_segment_name_DATA; 900 } 901 902 ConstString ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 903 static ConstString g_segment_name("__DATA_DIRTY"); 904 return g_segment_name; 905 } 906 907 ConstString ObjectFileMachO::GetSegmentNameDATA_CONST() { 908 static ConstString g_segment_name("__DATA_CONST"); 909 return g_segment_name; 910 } 911 912 ConstString ObjectFileMachO::GetSegmentNameOBJC() { 913 static ConstString g_segment_name_OBJC("__OBJC"); 914 return g_segment_name_OBJC; 915 } 916 917 ConstString ObjectFileMachO::GetSegmentNameLINKEDIT() { 918 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 919 return g_section_name_LINKEDIT; 920 } 921 922 ConstString ObjectFileMachO::GetSegmentNameDWARF() { 923 static ConstString g_section_name("__DWARF"); 924 return g_section_name; 925 } 926 927 ConstString ObjectFileMachO::GetSectionNameEHFrame() { 928 static ConstString g_section_name_eh_frame("__eh_frame"); 929 return g_section_name_eh_frame; 930 } 931 932 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 933 lldb::addr_t data_offset, 934 lldb::addr_t data_length) { 935 DataExtractor data; 936 data.SetData(data_sp, data_offset, data_length); 937 lldb::offset_t offset = 0; 938 uint32_t magic = data.GetU32(&offset); 939 return MachHeaderSizeFromMagic(magic) != 0; 940 } 941 942 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 943 DataBufferSP &data_sp, 944 lldb::offset_t data_offset, 945 const FileSpec *file, 946 lldb::offset_t file_offset, 947 lldb::offset_t length) 948 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 949 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 950 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 951 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 952 ::memset(&m_header, 0, sizeof(m_header)); 953 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 954 } 955 956 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 957 lldb::DataBufferSP &header_data_sp, 958 const lldb::ProcessSP &process_sp, 959 lldb::addr_t header_addr) 960 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 961 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 962 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 963 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 964 ::memset(&m_header, 0, sizeof(m_header)); 965 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 966 } 967 968 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 969 lldb::offset_t *data_offset_ptr, 970 llvm::MachO::mach_header &header) { 971 data.SetByteOrder(endian::InlHostByteOrder()); 972 // Leave magic in the original byte order 973 header.magic = data.GetU32(data_offset_ptr); 974 bool can_parse = false; 975 bool is_64_bit = false; 976 switch (header.magic) { 977 case MH_MAGIC: 978 data.SetByteOrder(endian::InlHostByteOrder()); 979 data.SetAddressByteSize(4); 980 can_parse = true; 981 break; 982 983 case MH_MAGIC_64: 984 data.SetByteOrder(endian::InlHostByteOrder()); 985 data.SetAddressByteSize(8); 986 can_parse = true; 987 is_64_bit = true; 988 break; 989 990 case MH_CIGAM: 991 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 992 ? eByteOrderLittle 993 : eByteOrderBig); 994 data.SetAddressByteSize(4); 995 can_parse = true; 996 break; 997 998 case MH_CIGAM_64: 999 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1000 ? eByteOrderLittle 1001 : eByteOrderBig); 1002 data.SetAddressByteSize(8); 1003 is_64_bit = true; 1004 can_parse = true; 1005 break; 1006 1007 default: 1008 break; 1009 } 1010 1011 if (can_parse) { 1012 data.GetU32(data_offset_ptr, &header.cputype, 6); 1013 if (is_64_bit) 1014 *data_offset_ptr += 4; 1015 return true; 1016 } else { 1017 memset(&header, 0, sizeof(header)); 1018 } 1019 return false; 1020 } 1021 1022 bool ObjectFileMachO::ParseHeader() { 1023 ModuleSP module_sp(GetModule()); 1024 if (!module_sp) 1025 return false; 1026 1027 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1028 bool can_parse = false; 1029 lldb::offset_t offset = 0; 1030 m_data.SetByteOrder(endian::InlHostByteOrder()); 1031 // Leave magic in the original byte order 1032 m_header.magic = m_data.GetU32(&offset); 1033 switch (m_header.magic) { 1034 case MH_MAGIC: 1035 m_data.SetByteOrder(endian::InlHostByteOrder()); 1036 m_data.SetAddressByteSize(4); 1037 can_parse = true; 1038 break; 1039 1040 case MH_MAGIC_64: 1041 m_data.SetByteOrder(endian::InlHostByteOrder()); 1042 m_data.SetAddressByteSize(8); 1043 can_parse = true; 1044 break; 1045 1046 case MH_CIGAM: 1047 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1048 ? eByteOrderLittle 1049 : eByteOrderBig); 1050 m_data.SetAddressByteSize(4); 1051 can_parse = true; 1052 break; 1053 1054 case MH_CIGAM_64: 1055 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1056 ? eByteOrderLittle 1057 : eByteOrderBig); 1058 m_data.SetAddressByteSize(8); 1059 can_parse = true; 1060 break; 1061 1062 default: 1063 break; 1064 } 1065 1066 if (can_parse) { 1067 m_data.GetU32(&offset, &m_header.cputype, 6); 1068 1069 ModuleSpecList all_specs; 1070 ModuleSpec base_spec; 1071 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 1072 base_spec, all_specs); 1073 1074 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 1075 ArchSpec mach_arch = 1076 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 1077 1078 // Check if the module has a required architecture 1079 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1080 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1081 continue; 1082 1083 if (SetModulesArchitecture(mach_arch)) { 1084 const size_t header_and_lc_size = 1085 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1086 if (m_data.GetByteSize() < header_and_lc_size) { 1087 DataBufferSP data_sp; 1088 ProcessSP process_sp(m_process_wp.lock()); 1089 if (process_sp) { 1090 data_sp = ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1091 } else { 1092 // Read in all only the load command data from the file on disk 1093 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset); 1094 if (data_sp->GetByteSize() != header_and_lc_size) 1095 continue; 1096 } 1097 if (data_sp) 1098 m_data.SetData(data_sp); 1099 } 1100 } 1101 return true; 1102 } 1103 // None found. 1104 return false; 1105 } else { 1106 memset(&m_header, 0, sizeof(struct llvm::MachO::mach_header)); 1107 } 1108 return false; 1109 } 1110 1111 ByteOrder ObjectFileMachO::GetByteOrder() const { 1112 return m_data.GetByteOrder(); 1113 } 1114 1115 bool ObjectFileMachO::IsExecutable() const { 1116 return m_header.filetype == MH_EXECUTE; 1117 } 1118 1119 bool ObjectFileMachO::IsDynamicLoader() const { 1120 return m_header.filetype == MH_DYLINKER; 1121 } 1122 1123 bool ObjectFileMachO::IsSharedCacheBinary() const { 1124 return m_header.flags & MH_DYLIB_IN_CACHE; 1125 } 1126 1127 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1128 return m_data.GetAddressByteSize(); 1129 } 1130 1131 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1132 Symtab *symtab = GetSymtab(); 1133 if (!symtab) 1134 return AddressClass::eUnknown; 1135 1136 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1137 if (symbol) { 1138 if (symbol->ValueIsAddress()) { 1139 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1140 if (section_sp) { 1141 const lldb::SectionType section_type = section_sp->GetType(); 1142 switch (section_type) { 1143 case eSectionTypeInvalid: 1144 return AddressClass::eUnknown; 1145 1146 case eSectionTypeCode: 1147 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1148 // For ARM we have a bit in the n_desc field of the symbol that 1149 // tells us ARM/Thumb which is bit 0x0008. 1150 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1151 return AddressClass::eCodeAlternateISA; 1152 } 1153 return AddressClass::eCode; 1154 1155 case eSectionTypeContainer: 1156 return AddressClass::eUnknown; 1157 1158 case eSectionTypeData: 1159 case eSectionTypeDataCString: 1160 case eSectionTypeDataCStringPointers: 1161 case eSectionTypeDataSymbolAddress: 1162 case eSectionTypeData4: 1163 case eSectionTypeData8: 1164 case eSectionTypeData16: 1165 case eSectionTypeDataPointers: 1166 case eSectionTypeZeroFill: 1167 case eSectionTypeDataObjCMessageRefs: 1168 case eSectionTypeDataObjCCFStrings: 1169 case eSectionTypeGoSymtab: 1170 return AddressClass::eData; 1171 1172 case eSectionTypeDebug: 1173 case eSectionTypeDWARFDebugAbbrev: 1174 case eSectionTypeDWARFDebugAbbrevDwo: 1175 case eSectionTypeDWARFDebugAddr: 1176 case eSectionTypeDWARFDebugAranges: 1177 case eSectionTypeDWARFDebugCuIndex: 1178 case eSectionTypeDWARFDebugFrame: 1179 case eSectionTypeDWARFDebugInfo: 1180 case eSectionTypeDWARFDebugInfoDwo: 1181 case eSectionTypeDWARFDebugLine: 1182 case eSectionTypeDWARFDebugLineStr: 1183 case eSectionTypeDWARFDebugLoc: 1184 case eSectionTypeDWARFDebugLocDwo: 1185 case eSectionTypeDWARFDebugLocLists: 1186 case eSectionTypeDWARFDebugLocListsDwo: 1187 case eSectionTypeDWARFDebugMacInfo: 1188 case eSectionTypeDWARFDebugMacro: 1189 case eSectionTypeDWARFDebugNames: 1190 case eSectionTypeDWARFDebugPubNames: 1191 case eSectionTypeDWARFDebugPubTypes: 1192 case eSectionTypeDWARFDebugRanges: 1193 case eSectionTypeDWARFDebugRngLists: 1194 case eSectionTypeDWARFDebugRngListsDwo: 1195 case eSectionTypeDWARFDebugStr: 1196 case eSectionTypeDWARFDebugStrDwo: 1197 case eSectionTypeDWARFDebugStrOffsets: 1198 case eSectionTypeDWARFDebugStrOffsetsDwo: 1199 case eSectionTypeDWARFDebugTuIndex: 1200 case eSectionTypeDWARFDebugTypes: 1201 case eSectionTypeDWARFDebugTypesDwo: 1202 case eSectionTypeDWARFAppleNames: 1203 case eSectionTypeDWARFAppleTypes: 1204 case eSectionTypeDWARFAppleNamespaces: 1205 case eSectionTypeDWARFAppleObjC: 1206 case eSectionTypeDWARFGNUDebugAltLink: 1207 return AddressClass::eDebug; 1208 1209 case eSectionTypeEHFrame: 1210 case eSectionTypeARMexidx: 1211 case eSectionTypeARMextab: 1212 case eSectionTypeCompactUnwind: 1213 return AddressClass::eRuntime; 1214 1215 case eSectionTypeAbsoluteAddress: 1216 case eSectionTypeELFSymbolTable: 1217 case eSectionTypeELFDynamicSymbols: 1218 case eSectionTypeELFRelocationEntries: 1219 case eSectionTypeELFDynamicLinkInfo: 1220 case eSectionTypeOther: 1221 return AddressClass::eUnknown; 1222 } 1223 } 1224 } 1225 1226 const SymbolType symbol_type = symbol->GetType(); 1227 switch (symbol_type) { 1228 case eSymbolTypeAny: 1229 return AddressClass::eUnknown; 1230 case eSymbolTypeAbsolute: 1231 return AddressClass::eUnknown; 1232 1233 case eSymbolTypeCode: 1234 case eSymbolTypeTrampoline: 1235 case eSymbolTypeResolver: 1236 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1237 // For ARM we have a bit in the n_desc field of the symbol that tells 1238 // us ARM/Thumb which is bit 0x0008. 1239 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1240 return AddressClass::eCodeAlternateISA; 1241 } 1242 return AddressClass::eCode; 1243 1244 case eSymbolTypeData: 1245 return AddressClass::eData; 1246 case eSymbolTypeRuntime: 1247 return AddressClass::eRuntime; 1248 case eSymbolTypeException: 1249 return AddressClass::eRuntime; 1250 case eSymbolTypeSourceFile: 1251 return AddressClass::eDebug; 1252 case eSymbolTypeHeaderFile: 1253 return AddressClass::eDebug; 1254 case eSymbolTypeObjectFile: 1255 return AddressClass::eDebug; 1256 case eSymbolTypeCommonBlock: 1257 return AddressClass::eDebug; 1258 case eSymbolTypeBlock: 1259 return AddressClass::eDebug; 1260 case eSymbolTypeLocal: 1261 return AddressClass::eData; 1262 case eSymbolTypeParam: 1263 return AddressClass::eData; 1264 case eSymbolTypeVariable: 1265 return AddressClass::eData; 1266 case eSymbolTypeVariableType: 1267 return AddressClass::eDebug; 1268 case eSymbolTypeLineEntry: 1269 return AddressClass::eDebug; 1270 case eSymbolTypeLineHeader: 1271 return AddressClass::eDebug; 1272 case eSymbolTypeScopeBegin: 1273 return AddressClass::eDebug; 1274 case eSymbolTypeScopeEnd: 1275 return AddressClass::eDebug; 1276 case eSymbolTypeAdditional: 1277 return AddressClass::eUnknown; 1278 case eSymbolTypeCompiler: 1279 return AddressClass::eDebug; 1280 case eSymbolTypeInstrumentation: 1281 return AddressClass::eDebug; 1282 case eSymbolTypeUndefined: 1283 return AddressClass::eUnknown; 1284 case eSymbolTypeObjCClass: 1285 return AddressClass::eRuntime; 1286 case eSymbolTypeObjCMetaClass: 1287 return AddressClass::eRuntime; 1288 case eSymbolTypeObjCIVar: 1289 return AddressClass::eRuntime; 1290 case eSymbolTypeReExported: 1291 return AddressClass::eRuntime; 1292 } 1293 } 1294 return AddressClass::eUnknown; 1295 } 1296 1297 bool ObjectFileMachO::IsStripped() { 1298 if (m_dysymtab.cmd == 0) { 1299 ModuleSP module_sp(GetModule()); 1300 if (module_sp) { 1301 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1302 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1303 const lldb::offset_t load_cmd_offset = offset; 1304 1305 llvm::MachO::load_command lc; 1306 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 1307 break; 1308 if (lc.cmd == LC_DYSYMTAB) { 1309 m_dysymtab.cmd = lc.cmd; 1310 m_dysymtab.cmdsize = lc.cmdsize; 1311 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1312 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1313 nullptr) { 1314 // Clear m_dysymtab if we were unable to read all items from the 1315 // load command 1316 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1317 } 1318 } 1319 offset = load_cmd_offset + lc.cmdsize; 1320 } 1321 } 1322 } 1323 if (m_dysymtab.cmd) 1324 return m_dysymtab.nlocalsym <= 1; 1325 return false; 1326 } 1327 1328 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1329 EncryptedFileRanges result; 1330 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1331 1332 llvm::MachO::encryption_info_command encryption_cmd; 1333 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1334 const lldb::offset_t load_cmd_offset = offset; 1335 if (m_data.GetU32(&offset, &encryption_cmd, 2) == nullptr) 1336 break; 1337 1338 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the 1339 // 3 fields we care about, so treat them the same. 1340 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1341 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1342 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1343 if (encryption_cmd.cryptid != 0) { 1344 EncryptedFileRanges::Entry entry; 1345 entry.SetRangeBase(encryption_cmd.cryptoff); 1346 entry.SetByteSize(encryption_cmd.cryptsize); 1347 result.Append(entry); 1348 } 1349 } 1350 } 1351 offset = load_cmd_offset + encryption_cmd.cmdsize; 1352 } 1353 1354 return result; 1355 } 1356 1357 void ObjectFileMachO::SanitizeSegmentCommand( 1358 llvm::MachO::segment_command_64 &seg_cmd, uint32_t cmd_idx) { 1359 if (m_length == 0 || seg_cmd.filesize == 0) 1360 return; 1361 1362 if (IsSharedCacheBinary() && !IsInMemory()) { 1363 // In shared cache images, the load commands are relative to the 1364 // shared cache file, and not the specific image we are 1365 // examining. Let's fix this up so that it looks like a normal 1366 // image. 1367 if (strncmp(seg_cmd.segname, "__TEXT", sizeof(seg_cmd.segname)) == 0) 1368 m_text_address = seg_cmd.vmaddr; 1369 if (strncmp(seg_cmd.segname, "__LINKEDIT", sizeof(seg_cmd.segname)) == 0) 1370 m_linkedit_original_offset = seg_cmd.fileoff; 1371 1372 seg_cmd.fileoff = seg_cmd.vmaddr - m_text_address; 1373 } 1374 1375 if (seg_cmd.fileoff > m_length) { 1376 // We have a load command that says it extends past the end of the file. 1377 // This is likely a corrupt file. We don't have any way to return an error 1378 // condition here (this method was likely invoked from something like 1379 // ObjectFile::GetSectionList()), so we just null out the section contents, 1380 // and dump a message to stdout. The most common case here is core file 1381 // debugging with a truncated file. 1382 const char *lc_segment_name = 1383 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1384 GetModule()->ReportWarning( 1385 "load command %u %s has a fileoff (0x%" PRIx64 1386 ") that extends beyond the end of the file (0x%" PRIx64 1387 "), ignoring this section", 1388 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1389 1390 seg_cmd.fileoff = 0; 1391 seg_cmd.filesize = 0; 1392 } 1393 1394 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1395 // We have a load command that says it extends past the end of the file. 1396 // This is likely a corrupt file. We don't have any way to return an error 1397 // condition here (this method was likely invoked from something like 1398 // ObjectFile::GetSectionList()), so we just null out the section contents, 1399 // and dump a message to stdout. The most common case here is core file 1400 // debugging with a truncated file. 1401 const char *lc_segment_name = 1402 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1403 GetModule()->ReportWarning( 1404 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1405 ") that extends beyond the end of the file (0x%" PRIx64 1406 "), the segment will be truncated to match", 1407 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1408 1409 // Truncate the length 1410 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1411 } 1412 } 1413 1414 static uint32_t 1415 GetSegmentPermissions(const llvm::MachO::segment_command_64 &seg_cmd) { 1416 uint32_t result = 0; 1417 if (seg_cmd.initprot & VM_PROT_READ) 1418 result |= ePermissionsReadable; 1419 if (seg_cmd.initprot & VM_PROT_WRITE) 1420 result |= ePermissionsWritable; 1421 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1422 result |= ePermissionsExecutable; 1423 return result; 1424 } 1425 1426 static lldb::SectionType GetSectionType(uint32_t flags, 1427 ConstString section_name) { 1428 1429 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1430 return eSectionTypeCode; 1431 1432 uint32_t mach_sect_type = flags & SECTION_TYPE; 1433 static ConstString g_sect_name_objc_data("__objc_data"); 1434 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1435 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1436 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1437 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1438 static ConstString g_sect_name_objc_const("__objc_const"); 1439 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1440 static ConstString g_sect_name_cfstring("__cfstring"); 1441 1442 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1443 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1444 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1445 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1446 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1447 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1448 static ConstString g_sect_name_dwarf_debug_loclists("__debug_loclists"); 1449 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1450 static ConstString g_sect_name_dwarf_debug_names("__debug_names"); 1451 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1452 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1453 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1454 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1455 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1456 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1457 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1458 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1459 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1460 static ConstString g_sect_name_eh_frame("__eh_frame"); 1461 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1462 static ConstString g_sect_name_text("__text"); 1463 static ConstString g_sect_name_data("__data"); 1464 static ConstString g_sect_name_go_symtab("__gosymtab"); 1465 1466 if (section_name == g_sect_name_dwarf_debug_abbrev) 1467 return eSectionTypeDWARFDebugAbbrev; 1468 if (section_name == g_sect_name_dwarf_debug_aranges) 1469 return eSectionTypeDWARFDebugAranges; 1470 if (section_name == g_sect_name_dwarf_debug_frame) 1471 return eSectionTypeDWARFDebugFrame; 1472 if (section_name == g_sect_name_dwarf_debug_info) 1473 return eSectionTypeDWARFDebugInfo; 1474 if (section_name == g_sect_name_dwarf_debug_line) 1475 return eSectionTypeDWARFDebugLine; 1476 if (section_name == g_sect_name_dwarf_debug_loc) 1477 return eSectionTypeDWARFDebugLoc; 1478 if (section_name == g_sect_name_dwarf_debug_loclists) 1479 return eSectionTypeDWARFDebugLocLists; 1480 if (section_name == g_sect_name_dwarf_debug_macinfo) 1481 return eSectionTypeDWARFDebugMacInfo; 1482 if (section_name == g_sect_name_dwarf_debug_names) 1483 return eSectionTypeDWARFDebugNames; 1484 if (section_name == g_sect_name_dwarf_debug_pubnames) 1485 return eSectionTypeDWARFDebugPubNames; 1486 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1487 return eSectionTypeDWARFDebugPubTypes; 1488 if (section_name == g_sect_name_dwarf_debug_ranges) 1489 return eSectionTypeDWARFDebugRanges; 1490 if (section_name == g_sect_name_dwarf_debug_str) 1491 return eSectionTypeDWARFDebugStr; 1492 if (section_name == g_sect_name_dwarf_debug_types) 1493 return eSectionTypeDWARFDebugTypes; 1494 if (section_name == g_sect_name_dwarf_apple_names) 1495 return eSectionTypeDWARFAppleNames; 1496 if (section_name == g_sect_name_dwarf_apple_types) 1497 return eSectionTypeDWARFAppleTypes; 1498 if (section_name == g_sect_name_dwarf_apple_namespaces) 1499 return eSectionTypeDWARFAppleNamespaces; 1500 if (section_name == g_sect_name_dwarf_apple_objc) 1501 return eSectionTypeDWARFAppleObjC; 1502 if (section_name == g_sect_name_objc_selrefs) 1503 return eSectionTypeDataCStringPointers; 1504 if (section_name == g_sect_name_objc_msgrefs) 1505 return eSectionTypeDataObjCMessageRefs; 1506 if (section_name == g_sect_name_eh_frame) 1507 return eSectionTypeEHFrame; 1508 if (section_name == g_sect_name_compact_unwind) 1509 return eSectionTypeCompactUnwind; 1510 if (section_name == g_sect_name_cfstring) 1511 return eSectionTypeDataObjCCFStrings; 1512 if (section_name == g_sect_name_go_symtab) 1513 return eSectionTypeGoSymtab; 1514 if (section_name == g_sect_name_objc_data || 1515 section_name == g_sect_name_objc_classrefs || 1516 section_name == g_sect_name_objc_superrefs || 1517 section_name == g_sect_name_objc_const || 1518 section_name == g_sect_name_objc_classlist) { 1519 return eSectionTypeDataPointers; 1520 } 1521 1522 switch (mach_sect_type) { 1523 // TODO: categorize sections by other flags for regular sections 1524 case S_REGULAR: 1525 if (section_name == g_sect_name_text) 1526 return eSectionTypeCode; 1527 if (section_name == g_sect_name_data) 1528 return eSectionTypeData; 1529 return eSectionTypeOther; 1530 case S_ZEROFILL: 1531 return eSectionTypeZeroFill; 1532 case S_CSTRING_LITERALS: // section with only literal C strings 1533 return eSectionTypeDataCString; 1534 case S_4BYTE_LITERALS: // section with only 4 byte literals 1535 return eSectionTypeData4; 1536 case S_8BYTE_LITERALS: // section with only 8 byte literals 1537 return eSectionTypeData8; 1538 case S_LITERAL_POINTERS: // section with only pointers to literals 1539 return eSectionTypeDataPointers; 1540 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1541 return eSectionTypeDataPointers; 1542 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1543 return eSectionTypeDataPointers; 1544 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1545 // the reserved2 field 1546 return eSectionTypeCode; 1547 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1548 // initialization 1549 return eSectionTypeDataPointers; 1550 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1551 // termination 1552 return eSectionTypeDataPointers; 1553 case S_COALESCED: 1554 return eSectionTypeOther; 1555 case S_GB_ZEROFILL: 1556 return eSectionTypeZeroFill; 1557 case S_INTERPOSING: // section with only pairs of function pointers for 1558 // interposing 1559 return eSectionTypeCode; 1560 case S_16BYTE_LITERALS: // section with only 16 byte literals 1561 return eSectionTypeData16; 1562 case S_DTRACE_DOF: 1563 return eSectionTypeDebug; 1564 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1565 return eSectionTypeDataPointers; 1566 default: 1567 return eSectionTypeOther; 1568 } 1569 } 1570 1571 struct ObjectFileMachO::SegmentParsingContext { 1572 const EncryptedFileRanges EncryptedRanges; 1573 lldb_private::SectionList &UnifiedList; 1574 uint32_t NextSegmentIdx = 0; 1575 uint32_t NextSectionIdx = 0; 1576 bool FileAddressesChanged = false; 1577 1578 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1579 lldb_private::SectionList &UnifiedList) 1580 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1581 }; 1582 1583 void ObjectFileMachO::ProcessSegmentCommand( 1584 const llvm::MachO::load_command &load_cmd_, lldb::offset_t offset, 1585 uint32_t cmd_idx, SegmentParsingContext &context) { 1586 llvm::MachO::segment_command_64 load_cmd; 1587 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1588 1589 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1590 return; 1591 1592 ModuleSP module_sp = GetModule(); 1593 const bool is_core = GetType() == eTypeCoreFile; 1594 const bool is_dsym = (m_header.filetype == MH_DSYM); 1595 bool add_section = true; 1596 bool add_to_unified = true; 1597 ConstString const_segname( 1598 load_cmd.segname, strnlen(load_cmd.segname, sizeof(load_cmd.segname))); 1599 1600 SectionSP unified_section_sp( 1601 context.UnifiedList.FindSectionByName(const_segname)); 1602 if (is_dsym && unified_section_sp) { 1603 if (const_segname == GetSegmentNameLINKEDIT()) { 1604 // We need to keep the __LINKEDIT segment private to this object file 1605 // only 1606 add_to_unified = false; 1607 } else { 1608 // This is the dSYM file and this section has already been created by the 1609 // object file, no need to create it. 1610 add_section = false; 1611 } 1612 } 1613 load_cmd.vmaddr = m_data.GetAddress(&offset); 1614 load_cmd.vmsize = m_data.GetAddress(&offset); 1615 load_cmd.fileoff = m_data.GetAddress(&offset); 1616 load_cmd.filesize = m_data.GetAddress(&offset); 1617 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1618 return; 1619 1620 SanitizeSegmentCommand(load_cmd, cmd_idx); 1621 1622 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1623 const bool segment_is_encrypted = 1624 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1625 1626 // Keep a list of mach segments around in case we need to get at data that 1627 // isn't stored in the abstracted Sections. 1628 m_mach_segments.push_back(load_cmd); 1629 1630 // Use a segment ID of the segment index shifted left by 8 so they never 1631 // conflict with any of the sections. 1632 SectionSP segment_sp; 1633 if (add_section && (const_segname || is_core)) { 1634 segment_sp = std::make_shared<Section>( 1635 module_sp, // Module to which this section belongs 1636 this, // Object file to which this sections belongs 1637 ++context.NextSegmentIdx 1638 << 8, // Section ID is the 1 based segment index 1639 // shifted right by 8 bits as not to collide with any of the 256 1640 // section IDs that are possible 1641 const_segname, // Name of this section 1642 eSectionTypeContainer, // This section is a container of other 1643 // sections. 1644 load_cmd.vmaddr, // File VM address == addresses as they are 1645 // found in the object file 1646 load_cmd.vmsize, // VM size in bytes of this section 1647 load_cmd.fileoff, // Offset to the data for this section in 1648 // the file 1649 load_cmd.filesize, // Size in bytes of this section as found 1650 // in the file 1651 0, // Segments have no alignment information 1652 load_cmd.flags); // Flags for this section 1653 1654 segment_sp->SetIsEncrypted(segment_is_encrypted); 1655 m_sections_up->AddSection(segment_sp); 1656 segment_sp->SetPermissions(segment_permissions); 1657 if (add_to_unified) 1658 context.UnifiedList.AddSection(segment_sp); 1659 } else if (unified_section_sp) { 1660 // If this is a dSYM and the file addresses in the dSYM differ from the 1661 // file addresses in the ObjectFile, we must use the file base address for 1662 // the Section from the dSYM for the DWARF to resolve correctly. 1663 // This only happens with binaries in the shared cache in practice; 1664 // normally a mismatch like this would give a binary & dSYM that do not 1665 // match UUIDs. When a binary is included in the shared cache, its 1666 // segments are rearranged to optimize the shared cache, so its file 1667 // addresses will differ from what the ObjectFile had originally, 1668 // and what the dSYM has. 1669 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1670 Log *log = GetLog(LLDBLog::Symbols); 1671 if (log) { 1672 log->Printf( 1673 "Installing dSYM's %s segment file address over ObjectFile's " 1674 "so symbol table/debug info resolves correctly for %s", 1675 const_segname.AsCString(), 1676 module_sp->GetFileSpec().GetFilename().AsCString()); 1677 } 1678 1679 // Make sure we've parsed the symbol table from the ObjectFile before 1680 // we go around changing its Sections. 1681 module_sp->GetObjectFile()->GetSymtab(); 1682 // eh_frame would present the same problems but we parse that on a per- 1683 // function basis as-needed so it's more difficult to remove its use of 1684 // the Sections. Realistically, the environments where this code path 1685 // will be taken will not have eh_frame sections. 1686 1687 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1688 1689 // Notify the module that the section addresses have been changed once 1690 // we're done so any file-address caches can be updated. 1691 context.FileAddressesChanged = true; 1692 } 1693 m_sections_up->AddSection(unified_section_sp); 1694 } 1695 1696 llvm::MachO::section_64 sect64; 1697 ::memset(§64, 0, sizeof(sect64)); 1698 // Push a section into our mach sections for the section at index zero 1699 // (NO_SECT) if we don't have any mach sections yet... 1700 if (m_mach_sections.empty()) 1701 m_mach_sections.push_back(sect64); 1702 uint32_t segment_sect_idx; 1703 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1704 1705 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1706 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1707 ++segment_sect_idx) { 1708 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1709 sizeof(sect64.sectname)) == nullptr) 1710 break; 1711 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1712 sizeof(sect64.segname)) == nullptr) 1713 break; 1714 sect64.addr = m_data.GetAddress(&offset); 1715 sect64.size = m_data.GetAddress(&offset); 1716 1717 if (m_data.GetU32(&offset, §64.offset, num_u32s) == nullptr) 1718 break; 1719 1720 if (IsSharedCacheBinary() && !IsInMemory()) { 1721 sect64.offset = sect64.addr - m_text_address; 1722 } 1723 1724 // Keep a list of mach sections around in case we need to get at data that 1725 // isn't stored in the abstracted Sections. 1726 m_mach_sections.push_back(sect64); 1727 1728 if (add_section) { 1729 ConstString section_name( 1730 sect64.sectname, strnlen(sect64.sectname, sizeof(sect64.sectname))); 1731 if (!const_segname) { 1732 // We have a segment with no name so we need to conjure up segments 1733 // that correspond to the section's segname if there isn't already such 1734 // a section. If there is such a section, we resize the section so that 1735 // it spans all sections. We also mark these sections as fake so 1736 // address matches don't hit if they land in the gaps between the child 1737 // sections. 1738 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1739 sizeof(sect64.segname)); 1740 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1741 if (segment_sp.get()) { 1742 Section *segment = segment_sp.get(); 1743 // Grow the section size as needed. 1744 const lldb::addr_t sect64_min_addr = sect64.addr; 1745 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1746 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1747 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1748 const lldb::addr_t curr_seg_max_addr = 1749 curr_seg_min_addr + curr_seg_byte_size; 1750 if (sect64_min_addr >= curr_seg_min_addr) { 1751 const lldb::addr_t new_seg_byte_size = 1752 sect64_max_addr - curr_seg_min_addr; 1753 // Only grow the section size if needed 1754 if (new_seg_byte_size > curr_seg_byte_size) 1755 segment->SetByteSize(new_seg_byte_size); 1756 } else { 1757 // We need to change the base address of the segment and adjust the 1758 // child section offsets for all existing children. 1759 const lldb::addr_t slide_amount = 1760 sect64_min_addr - curr_seg_min_addr; 1761 segment->Slide(slide_amount, false); 1762 segment->GetChildren().Slide(-slide_amount, false); 1763 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1764 } 1765 1766 // Grow the section size as needed. 1767 if (sect64.offset) { 1768 const lldb::addr_t segment_min_file_offset = 1769 segment->GetFileOffset(); 1770 const lldb::addr_t segment_max_file_offset = 1771 segment_min_file_offset + segment->GetFileSize(); 1772 1773 const lldb::addr_t section_min_file_offset = sect64.offset; 1774 const lldb::addr_t section_max_file_offset = 1775 section_min_file_offset + sect64.size; 1776 const lldb::addr_t new_file_offset = 1777 std::min(section_min_file_offset, segment_min_file_offset); 1778 const lldb::addr_t new_file_size = 1779 std::max(section_max_file_offset, segment_max_file_offset) - 1780 new_file_offset; 1781 segment->SetFileOffset(new_file_offset); 1782 segment->SetFileSize(new_file_size); 1783 } 1784 } else { 1785 // Create a fake section for the section's named segment 1786 segment_sp = std::make_shared<Section>( 1787 segment_sp, // Parent section 1788 module_sp, // Module to which this section belongs 1789 this, // Object file to which this section belongs 1790 ++context.NextSegmentIdx 1791 << 8, // Section ID is the 1 based segment index 1792 // shifted right by 8 bits as not to 1793 // collide with any of the 256 section IDs 1794 // that are possible 1795 const_segname, // Name of this section 1796 eSectionTypeContainer, // This section is a container of 1797 // other sections. 1798 sect64.addr, // File VM address == addresses as they are 1799 // found in the object file 1800 sect64.size, // VM size in bytes of this section 1801 sect64.offset, // Offset to the data for this section in 1802 // the file 1803 sect64.offset ? sect64.size : 0, // Size in bytes of 1804 // this section as 1805 // found in the file 1806 sect64.align, 1807 load_cmd.flags); // Flags for this section 1808 segment_sp->SetIsFake(true); 1809 segment_sp->SetPermissions(segment_permissions); 1810 m_sections_up->AddSection(segment_sp); 1811 if (add_to_unified) 1812 context.UnifiedList.AddSection(segment_sp); 1813 segment_sp->SetIsEncrypted(segment_is_encrypted); 1814 } 1815 } 1816 assert(segment_sp.get()); 1817 1818 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1819 1820 SectionSP section_sp(new Section( 1821 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1822 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1823 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1824 sect64.flags)); 1825 // Set the section to be encrypted to match the segment 1826 1827 bool section_is_encrypted = false; 1828 if (!segment_is_encrypted && load_cmd.filesize != 0) 1829 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1830 sect64.offset) != nullptr; 1831 1832 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1833 section_sp->SetPermissions(segment_permissions); 1834 segment_sp->GetChildren().AddSection(section_sp); 1835 1836 if (segment_sp->IsFake()) { 1837 segment_sp.reset(); 1838 const_segname.Clear(); 1839 } 1840 } 1841 } 1842 if (segment_sp && is_dsym) { 1843 if (first_segment_sectID <= context.NextSectionIdx) { 1844 lldb::user_id_t sect_uid; 1845 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1846 ++sect_uid) { 1847 SectionSP curr_section_sp( 1848 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1849 SectionSP next_section_sp; 1850 if (sect_uid + 1 <= context.NextSectionIdx) 1851 next_section_sp = 1852 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1853 1854 if (curr_section_sp.get()) { 1855 if (curr_section_sp->GetByteSize() == 0) { 1856 if (next_section_sp.get() != nullptr) 1857 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1858 curr_section_sp->GetFileAddress()); 1859 else 1860 curr_section_sp->SetByteSize(load_cmd.vmsize); 1861 } 1862 } 1863 } 1864 } 1865 } 1866 } 1867 1868 void ObjectFileMachO::ProcessDysymtabCommand( 1869 const llvm::MachO::load_command &load_cmd, lldb::offset_t offset) { 1870 m_dysymtab.cmd = load_cmd.cmd; 1871 m_dysymtab.cmdsize = load_cmd.cmdsize; 1872 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1873 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1874 } 1875 1876 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1877 if (m_sections_up) 1878 return; 1879 1880 m_sections_up = std::make_unique<SectionList>(); 1881 1882 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1883 // bool dump_sections = false; 1884 ModuleSP module_sp(GetModule()); 1885 1886 offset = MachHeaderSizeFromMagic(m_header.magic); 1887 1888 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1889 llvm::MachO::load_command load_cmd; 1890 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1891 const lldb::offset_t load_cmd_offset = offset; 1892 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 1893 break; 1894 1895 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1896 ProcessSegmentCommand(load_cmd, offset, i, context); 1897 else if (load_cmd.cmd == LC_DYSYMTAB) 1898 ProcessDysymtabCommand(load_cmd, offset); 1899 1900 offset = load_cmd_offset + load_cmd.cmdsize; 1901 } 1902 1903 if (context.FileAddressesChanged && module_sp) 1904 module_sp->SectionFileAddressesChanged(); 1905 } 1906 1907 class MachSymtabSectionInfo { 1908 public: 1909 MachSymtabSectionInfo(SectionList *section_list) 1910 : m_section_list(section_list), m_section_infos() { 1911 // Get the number of sections down to a depth of 1 to include all segments 1912 // and their sections, but no other sections that may be added for debug 1913 // map or 1914 m_section_infos.resize(section_list->GetNumSections(1)); 1915 } 1916 1917 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1918 if (n_sect == 0) 1919 return SectionSP(); 1920 if (n_sect < m_section_infos.size()) { 1921 if (!m_section_infos[n_sect].section_sp) { 1922 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1923 m_section_infos[n_sect].section_sp = section_sp; 1924 if (section_sp) { 1925 m_section_infos[n_sect].vm_range.SetBaseAddress( 1926 section_sp->GetFileAddress()); 1927 m_section_infos[n_sect].vm_range.SetByteSize( 1928 section_sp->GetByteSize()); 1929 } else { 1930 std::string filename = "<unknown>"; 1931 SectionSP first_section_sp(m_section_list->GetSectionAtIndex(0)); 1932 if (first_section_sp) 1933 filename = first_section_sp->GetObjectFile()->GetFileSpec().GetPath(); 1934 1935 Host::SystemLog(Host::eSystemLogError, 1936 "error: unable to find section %d for a symbol in " 1937 "%s, corrupt file?\n", 1938 n_sect, filename.c_str()); 1939 } 1940 } 1941 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1942 // Symbol is in section. 1943 return m_section_infos[n_sect].section_sp; 1944 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1945 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1946 file_addr) { 1947 // Symbol is in section with zero size, but has the same start address 1948 // as the section. This can happen with linker symbols (symbols that 1949 // start with the letter 'l' or 'L'. 1950 return m_section_infos[n_sect].section_sp; 1951 } 1952 } 1953 return m_section_list->FindSectionContainingFileAddress(file_addr); 1954 } 1955 1956 protected: 1957 struct SectionInfo { 1958 SectionInfo() : vm_range(), section_sp() {} 1959 1960 VMRange vm_range; 1961 SectionSP section_sp; 1962 }; 1963 SectionList *m_section_list; 1964 std::vector<SectionInfo> m_section_infos; 1965 }; 1966 1967 #define TRIE_SYMBOL_IS_THUMB (1ULL << 63) 1968 struct TrieEntry { 1969 void Dump() const { 1970 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1971 static_cast<unsigned long long>(address), 1972 static_cast<unsigned long long>(flags), 1973 static_cast<unsigned long long>(other), name.GetCString()); 1974 if (import_name) 1975 printf(" -> \"%s\"\n", import_name.GetCString()); 1976 else 1977 printf("\n"); 1978 } 1979 ConstString name; 1980 uint64_t address = LLDB_INVALID_ADDRESS; 1981 uint64_t flags = 1982 0; // EXPORT_SYMBOL_FLAGS_REEXPORT, EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER, 1983 // TRIE_SYMBOL_IS_THUMB 1984 uint64_t other = 0; 1985 ConstString import_name; 1986 }; 1987 1988 struct TrieEntryWithOffset { 1989 lldb::offset_t nodeOffset; 1990 TrieEntry entry; 1991 1992 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1993 1994 void Dump(uint32_t idx) const { 1995 printf("[%3u] 0x%16.16llx: ", idx, 1996 static_cast<unsigned long long>(nodeOffset)); 1997 entry.Dump(); 1998 } 1999 2000 bool operator<(const TrieEntryWithOffset &other) const { 2001 return (nodeOffset < other.nodeOffset); 2002 } 2003 }; 2004 2005 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2006 const bool is_arm, addr_t text_seg_base_addr, 2007 std::vector<llvm::StringRef> &nameSlices, 2008 std::set<lldb::addr_t> &resolver_addresses, 2009 std::vector<TrieEntryWithOffset> &reexports, 2010 std::vector<TrieEntryWithOffset> &ext_symbols) { 2011 if (!data.ValidOffset(offset)) 2012 return true; 2013 2014 // Terminal node -- end of a branch, possibly add this to 2015 // the symbol table or resolver table. 2016 const uint64_t terminalSize = data.GetULEB128(&offset); 2017 lldb::offset_t children_offset = offset + terminalSize; 2018 if (terminalSize != 0) { 2019 TrieEntryWithOffset e(offset); 2020 e.entry.flags = data.GetULEB128(&offset); 2021 const char *import_name = nullptr; 2022 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2023 e.entry.address = 0; 2024 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2025 import_name = data.GetCStr(&offset); 2026 } else { 2027 e.entry.address = data.GetULEB128(&offset); 2028 if (text_seg_base_addr != LLDB_INVALID_ADDRESS) 2029 e.entry.address += text_seg_base_addr; 2030 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2031 e.entry.other = data.GetULEB128(&offset); 2032 uint64_t resolver_addr = e.entry.other; 2033 if (text_seg_base_addr != LLDB_INVALID_ADDRESS) 2034 resolver_addr += text_seg_base_addr; 2035 if (is_arm) 2036 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2037 resolver_addresses.insert(resolver_addr); 2038 } else 2039 e.entry.other = 0; 2040 } 2041 bool add_this_entry = false; 2042 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT) && 2043 import_name && import_name[0]) { 2044 // add symbols that are reexport symbols with a valid import name. 2045 add_this_entry = true; 2046 } else if (e.entry.flags == 0 && 2047 (import_name == nullptr || import_name[0] == '\0')) { 2048 // add externally visible symbols, in case the nlist record has 2049 // been stripped/omitted. 2050 add_this_entry = true; 2051 } 2052 if (add_this_entry) { 2053 std::string name; 2054 if (!nameSlices.empty()) { 2055 for (auto name_slice : nameSlices) 2056 name.append(name_slice.data(), name_slice.size()); 2057 } 2058 if (name.size() > 1) { 2059 // Skip the leading '_' 2060 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2061 } 2062 if (import_name) { 2063 // Skip the leading '_' 2064 e.entry.import_name.SetCString(import_name + 1); 2065 } 2066 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT)) { 2067 reexports.push_back(e); 2068 } else { 2069 if (is_arm && (e.entry.address & 1)) { 2070 e.entry.flags |= TRIE_SYMBOL_IS_THUMB; 2071 e.entry.address &= THUMB_ADDRESS_BIT_MASK; 2072 } 2073 ext_symbols.push_back(e); 2074 } 2075 } 2076 } 2077 2078 const uint8_t childrenCount = data.GetU8(&children_offset); 2079 for (uint8_t i = 0; i < childrenCount; ++i) { 2080 const char *cstr = data.GetCStr(&children_offset); 2081 if (cstr) 2082 nameSlices.push_back(llvm::StringRef(cstr)); 2083 else 2084 return false; // Corrupt data 2085 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2086 if (childNodeOffset) { 2087 if (!ParseTrieEntries(data, childNodeOffset, is_arm, text_seg_base_addr, 2088 nameSlices, resolver_addresses, reexports, 2089 ext_symbols)) { 2090 return false; 2091 } 2092 } 2093 nameSlices.pop_back(); 2094 } 2095 return true; 2096 } 2097 2098 static SymbolType GetSymbolType(const char *&symbol_name, 2099 bool &demangled_is_synthesized, 2100 const SectionSP &text_section_sp, 2101 const SectionSP &data_section_sp, 2102 const SectionSP &data_dirty_section_sp, 2103 const SectionSP &data_const_section_sp, 2104 const SectionSP &symbol_section) { 2105 SymbolType type = eSymbolTypeInvalid; 2106 2107 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2108 if (symbol_section->IsDescendant(text_section_sp.get())) { 2109 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 2110 S_ATTR_SELF_MODIFYING_CODE | 2111 S_ATTR_SOME_INSTRUCTIONS)) 2112 type = eSymbolTypeData; 2113 else 2114 type = eSymbolTypeCode; 2115 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 2116 symbol_section->IsDescendant(data_dirty_section_sp.get()) || 2117 symbol_section->IsDescendant(data_const_section_sp.get())) { 2118 if (symbol_sect_name && 2119 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 2120 type = eSymbolTypeRuntime; 2121 2122 if (symbol_name) { 2123 llvm::StringRef symbol_name_ref(symbol_name); 2124 if (symbol_name_ref.startswith("OBJC_")) { 2125 static const llvm::StringRef g_objc_v2_prefix_class("OBJC_CLASS_$_"); 2126 static const llvm::StringRef g_objc_v2_prefix_metaclass( 2127 "OBJC_METACLASS_$_"); 2128 static const llvm::StringRef g_objc_v2_prefix_ivar("OBJC_IVAR_$_"); 2129 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 2130 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2131 type = eSymbolTypeObjCClass; 2132 demangled_is_synthesized = true; 2133 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 2134 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2135 type = eSymbolTypeObjCMetaClass; 2136 demangled_is_synthesized = true; 2137 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 2138 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2139 type = eSymbolTypeObjCIVar; 2140 demangled_is_synthesized = true; 2141 } 2142 } 2143 } 2144 } else if (symbol_sect_name && 2145 ::strstr(symbol_sect_name, "__gcc_except_tab") == 2146 symbol_sect_name) { 2147 type = eSymbolTypeException; 2148 } else { 2149 type = eSymbolTypeData; 2150 } 2151 } else if (symbol_sect_name && 2152 ::strstr(symbol_sect_name, "__IMPORT") == symbol_sect_name) { 2153 type = eSymbolTypeTrampoline; 2154 } 2155 return type; 2156 } 2157 2158 // Read the UUID out of a dyld_shared_cache file on-disk. 2159 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2160 const ByteOrder byte_order, 2161 const uint32_t addr_byte_size) { 2162 UUID dsc_uuid; 2163 DataBufferSP DscData = MapFileData( 2164 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2165 if (!DscData) 2166 return dsc_uuid; 2167 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2168 2169 char version_str[7]; 2170 lldb::offset_t offset = 0; 2171 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2172 version_str[6] = '\0'; 2173 if (strcmp(version_str, "dyld_v") == 0) { 2174 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2175 dsc_uuid = UUID::fromOptionalData( 2176 dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t)); 2177 } 2178 Log *log = GetLog(LLDBLog::Symbols); 2179 if (log && dsc_uuid.IsValid()) { 2180 LLDB_LOGF(log, "Shared cache %s has UUID %s", 2181 dyld_shared_cache.GetPath().c_str(), 2182 dsc_uuid.GetAsString().c_str()); 2183 } 2184 return dsc_uuid; 2185 } 2186 2187 static llvm::Optional<struct nlist_64> 2188 ParseNList(DataExtractor &nlist_data, lldb::offset_t &nlist_data_offset, 2189 size_t nlist_byte_size) { 2190 struct nlist_64 nlist; 2191 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2192 return {}; 2193 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 2194 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 2195 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 2196 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 2197 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 2198 return nlist; 2199 } 2200 2201 enum { DebugSymbols = true, NonDebugSymbols = false }; 2202 2203 void ObjectFileMachO::ParseSymtab(Symtab &symtab) { 2204 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 const bool pinned = dyld_shared_cache_pin_mapping(shared_cache); 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 if (pinned) 2773 dyld_shared_cache_unpin_mapping(shared_cache); 2774 }); 2775 if (nlist_buffer) { 2776 DataExtractor dsc_local_symbols_data(nlist_buffer, 2777 nlist_count * nlist_byte_size, 2778 byte_order, addr_byte_size); 2779 unmapped_local_symbols_found = nlist_count; 2780 2781 // The normal nlist code cannot correctly size the Symbols 2782 // array, we need to allocate it here. 2783 sym = symtab.Resize( 2784 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2785 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2786 num_syms = symtab.GetNumSymbols(); 2787 2788 lldb::offset_t nlist_data_offset = 0; 2789 2790 for (uint32_t nlist_index = 0; 2791 nlist_index < nlist_count; 2792 nlist_index++) { 2793 ///////////////////////////// 2794 { 2795 llvm::Optional<struct nlist_64> nlist_maybe = 2796 ParseNList(dsc_local_symbols_data, nlist_data_offset, 2797 nlist_byte_size); 2798 if (!nlist_maybe) 2799 break; 2800 struct nlist_64 nlist = *nlist_maybe; 2801 2802 SymbolType type = eSymbolTypeInvalid; 2803 const char *symbol_name = string_table + nlist.n_strx; 2804 2805 if (symbol_name == NULL) { 2806 // No symbol should be NULL, even the symbols with no 2807 // string values should have an offset zero which 2808 // points to an empty C-string 2809 Host::SystemLog( 2810 Host::eSystemLogError, 2811 "error: DSC unmapped local symbol[%u] has invalid " 2812 "string table offset 0x%x in %s, ignoring symbol\n", 2813 nlist_index, nlist.n_strx, 2814 module_sp->GetFileSpec().GetPath().c_str()); 2815 continue; 2816 } 2817 if (symbol_name[0] == '\0') 2818 symbol_name = NULL; 2819 2820 const char *symbol_name_non_abi_mangled = NULL; 2821 2822 SectionSP symbol_section; 2823 uint32_t symbol_byte_size = 0; 2824 bool add_nlist = true; 2825 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2826 bool demangled_is_synthesized = false; 2827 bool is_gsym = false; 2828 bool set_value = true; 2829 2830 assert(sym_idx < num_syms); 2831 2832 sym[sym_idx].SetDebug(is_debug); 2833 2834 if (is_debug) { 2835 switch (nlist.n_type) { 2836 case N_GSYM: 2837 // global symbol: name,,NO_SECT,type,0 2838 // Sometimes the N_GSYM value contains the address. 2839 2840 // FIXME: In the .o files, we have a GSYM and a debug 2841 // symbol for all the ObjC data. They 2842 // have the same address, but we want to ensure that 2843 // we always find only the real symbol, 'cause we 2844 // don't currently correctly attribute the 2845 // GSYM one to the ObjCClass/Ivar/MetaClass 2846 // symbol type. This is a temporary hack to make 2847 // sure the ObjectiveC symbols get treated correctly. 2848 // To do this right, we should coalesce all the GSYM 2849 // & global symbols that have the same address. 2850 2851 is_gsym = true; 2852 sym[sym_idx].SetExternal(true); 2853 2854 if (symbol_name && symbol_name[0] == '_' && 2855 symbol_name[1] == 'O') { 2856 llvm::StringRef symbol_name_ref(symbol_name); 2857 if (symbol_name_ref.startswith( 2858 g_objc_v2_prefix_class)) { 2859 symbol_name_non_abi_mangled = symbol_name + 1; 2860 symbol_name = 2861 symbol_name + g_objc_v2_prefix_class.size(); 2862 type = eSymbolTypeObjCClass; 2863 demangled_is_synthesized = true; 2864 2865 } else if (symbol_name_ref.startswith( 2866 g_objc_v2_prefix_metaclass)) { 2867 symbol_name_non_abi_mangled = symbol_name + 1; 2868 symbol_name = 2869 symbol_name + g_objc_v2_prefix_metaclass.size(); 2870 type = eSymbolTypeObjCMetaClass; 2871 demangled_is_synthesized = true; 2872 } else if (symbol_name_ref.startswith( 2873 g_objc_v2_prefix_ivar)) { 2874 symbol_name_non_abi_mangled = symbol_name + 1; 2875 symbol_name = 2876 symbol_name + g_objc_v2_prefix_ivar.size(); 2877 type = eSymbolTypeObjCIVar; 2878 demangled_is_synthesized = true; 2879 } 2880 } else { 2881 if (nlist.n_value != 0) 2882 symbol_section = section_info.GetSection( 2883 nlist.n_sect, nlist.n_value); 2884 type = eSymbolTypeData; 2885 } 2886 break; 2887 2888 case N_FNAME: 2889 // procedure name (f77 kludge): name,,NO_SECT,0,0 2890 type = eSymbolTypeCompiler; 2891 break; 2892 2893 case N_FUN: 2894 // procedure: name,,n_sect,linenumber,address 2895 if (symbol_name) { 2896 type = eSymbolTypeCode; 2897 symbol_section = section_info.GetSection( 2898 nlist.n_sect, nlist.n_value); 2899 2900 N_FUN_addr_to_sym_idx.insert( 2901 std::make_pair(nlist.n_value, sym_idx)); 2902 // We use the current number of symbols in the 2903 // symbol table in lieu of using nlist_idx in case 2904 // we ever start trimming entries out 2905 N_FUN_indexes.push_back(sym_idx); 2906 } else { 2907 type = eSymbolTypeCompiler; 2908 2909 if (!N_FUN_indexes.empty()) { 2910 // Copy the size of the function into the 2911 // original 2912 // STAB entry so we don't have 2913 // to hunt for it later 2914 symtab.SymbolAtIndex(N_FUN_indexes.back()) 2915 ->SetByteSize(nlist.n_value); 2916 N_FUN_indexes.pop_back(); 2917 // We don't really need the end function STAB as 2918 // it contains the size which we already placed 2919 // with the original symbol, so don't add it if 2920 // we want a minimal symbol table 2921 add_nlist = false; 2922 } 2923 } 2924 break; 2925 2926 case N_STSYM: 2927 // static symbol: name,,n_sect,type,address 2928 N_STSYM_addr_to_sym_idx.insert( 2929 std::make_pair(nlist.n_value, sym_idx)); 2930 symbol_section = section_info.GetSection(nlist.n_sect, 2931 nlist.n_value); 2932 if (symbol_name && symbol_name[0]) { 2933 type = ObjectFile::GetSymbolTypeFromName( 2934 symbol_name + 1, eSymbolTypeData); 2935 } 2936 break; 2937 2938 case N_LCSYM: 2939 // .lcomm symbol: name,,n_sect,type,address 2940 symbol_section = section_info.GetSection(nlist.n_sect, 2941 nlist.n_value); 2942 type = eSymbolTypeCommonBlock; 2943 break; 2944 2945 case N_BNSYM: 2946 // We use the current number of symbols in the symbol 2947 // table in lieu of using nlist_idx in case we ever 2948 // start trimming entries out Skip these if we want 2949 // minimal symbol tables 2950 add_nlist = false; 2951 break; 2952 2953 case N_ENSYM: 2954 // Set the size of the N_BNSYM to the terminating 2955 // index of this N_ENSYM so that we can always skip 2956 // the entire symbol if we need to navigate more 2957 // quickly at the source level when parsing STABS 2958 // Skip these if we want minimal symbol tables 2959 add_nlist = false; 2960 break; 2961 2962 case N_OPT: 2963 // emitted with gcc2_compiled and in gcc source 2964 type = eSymbolTypeCompiler; 2965 break; 2966 2967 case N_RSYM: 2968 // register sym: name,,NO_SECT,type,register 2969 type = eSymbolTypeVariable; 2970 break; 2971 2972 case N_SLINE: 2973 // src line: 0,,n_sect,linenumber,address 2974 symbol_section = section_info.GetSection(nlist.n_sect, 2975 nlist.n_value); 2976 type = eSymbolTypeLineEntry; 2977 break; 2978 2979 case N_SSYM: 2980 // structure elt: name,,NO_SECT,type,struct_offset 2981 type = eSymbolTypeVariableType; 2982 break; 2983 2984 case N_SO: 2985 // source file name 2986 type = eSymbolTypeSourceFile; 2987 if (symbol_name == NULL) { 2988 add_nlist = false; 2989 if (N_SO_index != UINT32_MAX) { 2990 // Set the size of the N_SO to the terminating 2991 // index of this N_SO so that we can always skip 2992 // the entire N_SO if we need to navigate more 2993 // quickly at the source level when parsing STABS 2994 symbol_ptr = symtab.SymbolAtIndex(N_SO_index); 2995 symbol_ptr->SetByteSize(sym_idx); 2996 symbol_ptr->SetSizeIsSibling(true); 2997 } 2998 N_NSYM_indexes.clear(); 2999 N_INCL_indexes.clear(); 3000 N_BRAC_indexes.clear(); 3001 N_COMM_indexes.clear(); 3002 N_FUN_indexes.clear(); 3003 N_SO_index = UINT32_MAX; 3004 } else { 3005 // We use the current number of symbols in the 3006 // symbol table in lieu of using nlist_idx in case 3007 // we ever start trimming entries out 3008 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3009 if (N_SO_has_full_path) { 3010 if ((N_SO_index == sym_idx - 1) && 3011 ((sym_idx - 1) < num_syms)) { 3012 // We have two consecutive N_SO entries where 3013 // the first contains a directory and the 3014 // second contains a full path. 3015 sym[sym_idx - 1].GetMangled().SetValue( 3016 ConstString(symbol_name), false); 3017 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3018 add_nlist = false; 3019 } else { 3020 // This is the first entry in a N_SO that 3021 // contains a directory or 3022 // a full path to the source file 3023 N_SO_index = sym_idx; 3024 } 3025 } else if ((N_SO_index == sym_idx - 1) && 3026 ((sym_idx - 1) < num_syms)) { 3027 // This is usually the second N_SO entry that 3028 // contains just the filename, so here we combine 3029 // it with the first one if we are minimizing the 3030 // symbol table 3031 const char *so_path = sym[sym_idx - 1] 3032 .GetMangled() 3033 .GetDemangledName() 3034 .AsCString(); 3035 if (so_path && so_path[0]) { 3036 std::string full_so_path(so_path); 3037 const size_t double_slash_pos = 3038 full_so_path.find("//"); 3039 if (double_slash_pos != std::string::npos) { 3040 // The linker has been generating bad N_SO 3041 // entries with doubled up paths 3042 // in the format "%s%s" where the first 3043 // string in the DW_AT_comp_dir, and the 3044 // second is the directory for the source 3045 // file so you end up with a path that looks 3046 // like "/tmp/src//tmp/src/" 3047 FileSpec so_dir(so_path); 3048 if (!FileSystem::Instance().Exists(so_dir)) { 3049 so_dir.SetFile( 3050 &full_so_path[double_slash_pos + 1], 3051 FileSpec::Style::native); 3052 if (FileSystem::Instance().Exists(so_dir)) { 3053 // Trim off the incorrect path 3054 full_so_path.erase(0, double_slash_pos + 1); 3055 } 3056 } 3057 } 3058 if (*full_so_path.rbegin() != '/') 3059 full_so_path += '/'; 3060 full_so_path += symbol_name; 3061 sym[sym_idx - 1].GetMangled().SetValue( 3062 ConstString(full_so_path.c_str()), false); 3063 add_nlist = false; 3064 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3065 } 3066 } else { 3067 // This could be a relative path to a N_SO 3068 N_SO_index = sym_idx; 3069 } 3070 } 3071 break; 3072 3073 case N_OSO: 3074 // object file name: name,,0,0,st_mtime 3075 type = eSymbolTypeObjectFile; 3076 break; 3077 3078 case N_LSYM: 3079 // local sym: name,,NO_SECT,type,offset 3080 type = eSymbolTypeLocal; 3081 break; 3082 3083 // INCL scopes 3084 case N_BINCL: 3085 // include file beginning: name,,NO_SECT,0,sum We use 3086 // the current number of symbols in the symbol table 3087 // in lieu of using nlist_idx in case we ever start 3088 // trimming entries out 3089 N_INCL_indexes.push_back(sym_idx); 3090 type = eSymbolTypeScopeBegin; 3091 break; 3092 3093 case N_EINCL: 3094 // include file end: name,,NO_SECT,0,0 3095 // Set the size of the N_BINCL to the terminating 3096 // index of this N_EINCL so that we can always skip 3097 // the entire symbol if we need to navigate more 3098 // quickly at the source level when parsing STABS 3099 if (!N_INCL_indexes.empty()) { 3100 symbol_ptr = 3101 symtab.SymbolAtIndex(N_INCL_indexes.back()); 3102 symbol_ptr->SetByteSize(sym_idx + 1); 3103 symbol_ptr->SetSizeIsSibling(true); 3104 N_INCL_indexes.pop_back(); 3105 } 3106 type = eSymbolTypeScopeEnd; 3107 break; 3108 3109 case N_SOL: 3110 // #included file name: name,,n_sect,0,address 3111 type = eSymbolTypeHeaderFile; 3112 3113 // We currently don't use the header files on darwin 3114 add_nlist = false; 3115 break; 3116 3117 case N_PARAMS: 3118 // compiler parameters: name,,NO_SECT,0,0 3119 type = eSymbolTypeCompiler; 3120 break; 3121 3122 case N_VERSION: 3123 // compiler version: name,,NO_SECT,0,0 3124 type = eSymbolTypeCompiler; 3125 break; 3126 3127 case N_OLEVEL: 3128 // compiler -O level: name,,NO_SECT,0,0 3129 type = eSymbolTypeCompiler; 3130 break; 3131 3132 case N_PSYM: 3133 // parameter: name,,NO_SECT,type,offset 3134 type = eSymbolTypeVariable; 3135 break; 3136 3137 case N_ENTRY: 3138 // alternate entry: name,,n_sect,linenumber,address 3139 symbol_section = section_info.GetSection(nlist.n_sect, 3140 nlist.n_value); 3141 type = eSymbolTypeLineEntry; 3142 break; 3143 3144 // Left and Right Braces 3145 case N_LBRAC: 3146 // left bracket: 0,,NO_SECT,nesting level,address We 3147 // use the current number of symbols in the symbol 3148 // table in lieu of using nlist_idx in case we ever 3149 // start trimming entries out 3150 symbol_section = section_info.GetSection(nlist.n_sect, 3151 nlist.n_value); 3152 N_BRAC_indexes.push_back(sym_idx); 3153 type = eSymbolTypeScopeBegin; 3154 break; 3155 3156 case N_RBRAC: 3157 // right bracket: 0,,NO_SECT,nesting level,address 3158 // Set the size of the N_LBRAC to the terminating 3159 // index of this N_RBRAC so that we can always skip 3160 // the entire symbol if we need to navigate more 3161 // quickly at the source level when parsing STABS 3162 symbol_section = section_info.GetSection(nlist.n_sect, 3163 nlist.n_value); 3164 if (!N_BRAC_indexes.empty()) { 3165 symbol_ptr = 3166 symtab.SymbolAtIndex(N_BRAC_indexes.back()); 3167 symbol_ptr->SetByteSize(sym_idx + 1); 3168 symbol_ptr->SetSizeIsSibling(true); 3169 N_BRAC_indexes.pop_back(); 3170 } 3171 type = eSymbolTypeScopeEnd; 3172 break; 3173 3174 case N_EXCL: 3175 // deleted include file: name,,NO_SECT,0,sum 3176 type = eSymbolTypeHeaderFile; 3177 break; 3178 3179 // COMM scopes 3180 case N_BCOMM: 3181 // begin common: name,,NO_SECT,0,0 3182 // We use the current number of symbols in the symbol 3183 // table in lieu of using nlist_idx in case we ever 3184 // start trimming entries out 3185 type = eSymbolTypeScopeBegin; 3186 N_COMM_indexes.push_back(sym_idx); 3187 break; 3188 3189 case N_ECOML: 3190 // end common (local name): 0,,n_sect,0,address 3191 symbol_section = section_info.GetSection(nlist.n_sect, 3192 nlist.n_value); 3193 // Fall through 3194 3195 case N_ECOMM: 3196 // end common: name,,n_sect,0,0 3197 // Set the size of the N_BCOMM to the terminating 3198 // index of this N_ECOMM/N_ECOML so that we can 3199 // always skip the entire symbol if we need to 3200 // navigate more quickly at the source level when 3201 // parsing STABS 3202 if (!N_COMM_indexes.empty()) { 3203 symbol_ptr = 3204 symtab.SymbolAtIndex(N_COMM_indexes.back()); 3205 symbol_ptr->SetByteSize(sym_idx + 1); 3206 symbol_ptr->SetSizeIsSibling(true); 3207 N_COMM_indexes.pop_back(); 3208 } 3209 type = eSymbolTypeScopeEnd; 3210 break; 3211 3212 case N_LENG: 3213 // second stab entry with length information 3214 type = eSymbolTypeAdditional; 3215 break; 3216 3217 default: 3218 break; 3219 } 3220 } else { 3221 // uint8_t n_pext = N_PEXT & nlist.n_type; 3222 uint8_t n_type = N_TYPE & nlist.n_type; 3223 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3224 3225 switch (n_type) { 3226 case N_INDR: { 3227 const char *reexport_name_cstr = 3228 strtab_data.PeekCStr(nlist.n_value); 3229 if (reexport_name_cstr && reexport_name_cstr[0]) { 3230 type = eSymbolTypeReExported; 3231 ConstString reexport_name( 3232 reexport_name_cstr + 3233 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3234 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3235 set_value = false; 3236 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3237 indirect_symbol_names.insert(ConstString( 3238 symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3239 } else 3240 type = eSymbolTypeUndefined; 3241 } break; 3242 3243 case N_UNDF: 3244 if (symbol_name && symbol_name[0]) { 3245 ConstString undefined_name( 3246 symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3247 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3248 } 3249 // Fall through 3250 case N_PBUD: 3251 type = eSymbolTypeUndefined; 3252 break; 3253 3254 case N_ABS: 3255 type = eSymbolTypeAbsolute; 3256 break; 3257 3258 case N_SECT: { 3259 symbol_section = section_info.GetSection(nlist.n_sect, 3260 nlist.n_value); 3261 3262 if (symbol_section == NULL) { 3263 // TODO: warn about this? 3264 add_nlist = false; 3265 break; 3266 } 3267 3268 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3269 type = eSymbolTypeException; 3270 } else { 3271 uint32_t section_type = 3272 symbol_section->Get() & SECTION_TYPE; 3273 3274 switch (section_type) { 3275 case S_CSTRING_LITERALS: 3276 type = eSymbolTypeData; 3277 break; // section with only literal C strings 3278 case S_4BYTE_LITERALS: 3279 type = eSymbolTypeData; 3280 break; // section with only 4 byte literals 3281 case S_8BYTE_LITERALS: 3282 type = eSymbolTypeData; 3283 break; // section with only 8 byte literals 3284 case S_LITERAL_POINTERS: 3285 type = eSymbolTypeTrampoline; 3286 break; // section with only pointers to literals 3287 case S_NON_LAZY_SYMBOL_POINTERS: 3288 type = eSymbolTypeTrampoline; 3289 break; // section with only non-lazy symbol 3290 // pointers 3291 case S_LAZY_SYMBOL_POINTERS: 3292 type = eSymbolTypeTrampoline; 3293 break; // section with only lazy symbol pointers 3294 case S_SYMBOL_STUBS: 3295 type = eSymbolTypeTrampoline; 3296 break; // section with only symbol stubs, byte 3297 // size of stub in the reserved2 field 3298 case S_MOD_INIT_FUNC_POINTERS: 3299 type = eSymbolTypeCode; 3300 break; // section with only function pointers for 3301 // initialization 3302 case S_MOD_TERM_FUNC_POINTERS: 3303 type = eSymbolTypeCode; 3304 break; // section with only function pointers for 3305 // termination 3306 case S_INTERPOSING: 3307 type = eSymbolTypeTrampoline; 3308 break; // section with only pairs of function 3309 // pointers for interposing 3310 case S_16BYTE_LITERALS: 3311 type = eSymbolTypeData; 3312 break; // section with only 16 byte literals 3313 case S_DTRACE_DOF: 3314 type = eSymbolTypeInstrumentation; 3315 break; 3316 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3317 type = eSymbolTypeTrampoline; 3318 break; 3319 default: 3320 switch (symbol_section->GetType()) { 3321 case lldb::eSectionTypeCode: 3322 type = eSymbolTypeCode; 3323 break; 3324 case eSectionTypeData: 3325 case eSectionTypeDataCString: // Inlined C string 3326 // data 3327 case eSectionTypeDataCStringPointers: // Pointers 3328 // to C 3329 // string 3330 // data 3331 case eSectionTypeDataSymbolAddress: // Address of 3332 // a symbol in 3333 // the symbol 3334 // table 3335 case eSectionTypeData4: 3336 case eSectionTypeData8: 3337 case eSectionTypeData16: 3338 type = eSymbolTypeData; 3339 break; 3340 default: 3341 break; 3342 } 3343 break; 3344 } 3345 3346 if (type == eSymbolTypeInvalid) { 3347 const char *symbol_sect_name = 3348 symbol_section->GetName().AsCString(); 3349 if (symbol_section->IsDescendant( 3350 text_section_sp.get())) { 3351 if (symbol_section->IsClear( 3352 S_ATTR_PURE_INSTRUCTIONS | 3353 S_ATTR_SELF_MODIFYING_CODE | 3354 S_ATTR_SOME_INSTRUCTIONS)) 3355 type = eSymbolTypeData; 3356 else 3357 type = eSymbolTypeCode; 3358 } else if (symbol_section->IsDescendant( 3359 data_section_sp.get()) || 3360 symbol_section->IsDescendant( 3361 data_dirty_section_sp.get()) || 3362 symbol_section->IsDescendant( 3363 data_const_section_sp.get())) { 3364 if (symbol_sect_name && 3365 ::strstr(symbol_sect_name, "__objc") == 3366 symbol_sect_name) { 3367 type = eSymbolTypeRuntime; 3368 3369 if (symbol_name) { 3370 llvm::StringRef symbol_name_ref(symbol_name); 3371 if (symbol_name_ref.startswith("_OBJC_")) { 3372 llvm::StringRef 3373 g_objc_v2_prefix_class( 3374 "_OBJC_CLASS_$_"); 3375 llvm::StringRef 3376 g_objc_v2_prefix_metaclass( 3377 "_OBJC_METACLASS_$_"); 3378 llvm::StringRef 3379 g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 3380 if (symbol_name_ref.startswith( 3381 g_objc_v2_prefix_class)) { 3382 symbol_name_non_abi_mangled = 3383 symbol_name + 1; 3384 symbol_name = 3385 symbol_name + 3386 g_objc_v2_prefix_class.size(); 3387 type = eSymbolTypeObjCClass; 3388 demangled_is_synthesized = true; 3389 } else if ( 3390 symbol_name_ref.startswith( 3391 g_objc_v2_prefix_metaclass)) { 3392 symbol_name_non_abi_mangled = 3393 symbol_name + 1; 3394 symbol_name = 3395 symbol_name + 3396 g_objc_v2_prefix_metaclass.size(); 3397 type = eSymbolTypeObjCMetaClass; 3398 demangled_is_synthesized = true; 3399 } else if (symbol_name_ref.startswith( 3400 g_objc_v2_prefix_ivar)) { 3401 symbol_name_non_abi_mangled = 3402 symbol_name + 1; 3403 symbol_name = 3404 symbol_name + 3405 g_objc_v2_prefix_ivar.size(); 3406 type = eSymbolTypeObjCIVar; 3407 demangled_is_synthesized = true; 3408 } 3409 } 3410 } 3411 } else if (symbol_sect_name && 3412 ::strstr(symbol_sect_name, 3413 "__gcc_except_tab") == 3414 symbol_sect_name) { 3415 type = eSymbolTypeException; 3416 } else { 3417 type = eSymbolTypeData; 3418 } 3419 } else if (symbol_sect_name && 3420 ::strstr(symbol_sect_name, "__IMPORT") == 3421 symbol_sect_name) { 3422 type = eSymbolTypeTrampoline; 3423 } else if (symbol_section->IsDescendant( 3424 objc_section_sp.get())) { 3425 type = eSymbolTypeRuntime; 3426 if (symbol_name && symbol_name[0] == '.') { 3427 llvm::StringRef symbol_name_ref(symbol_name); 3428 llvm::StringRef 3429 g_objc_v1_prefix_class(".objc_class_name_"); 3430 if (symbol_name_ref.startswith( 3431 g_objc_v1_prefix_class)) { 3432 symbol_name_non_abi_mangled = symbol_name; 3433 symbol_name = symbol_name + 3434 g_objc_v1_prefix_class.size(); 3435 type = eSymbolTypeObjCClass; 3436 demangled_is_synthesized = true; 3437 } 3438 } 3439 } 3440 } 3441 } 3442 } break; 3443 } 3444 } 3445 3446 if (add_nlist) { 3447 uint64_t symbol_value = nlist.n_value; 3448 if (symbol_name_non_abi_mangled) { 3449 sym[sym_idx].GetMangled().SetMangledName( 3450 ConstString(symbol_name_non_abi_mangled)); 3451 sym[sym_idx].GetMangled().SetDemangledName( 3452 ConstString(symbol_name)); 3453 } else { 3454 bool symbol_name_is_mangled = false; 3455 3456 if (symbol_name && symbol_name[0] == '_') { 3457 symbol_name_is_mangled = symbol_name[1] == '_'; 3458 symbol_name++; // Skip the leading underscore 3459 } 3460 3461 if (symbol_name) { 3462 ConstString const_symbol_name(symbol_name); 3463 sym[sym_idx].GetMangled().SetValue( 3464 const_symbol_name, symbol_name_is_mangled); 3465 if (is_gsym && is_debug) { 3466 const char *gsym_name = 3467 sym[sym_idx] 3468 .GetMangled() 3469 .GetName(Mangled::ePreferMangled) 3470 .GetCString(); 3471 if (gsym_name) 3472 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3473 } 3474 } 3475 } 3476 if (symbol_section) { 3477 const addr_t section_file_addr = 3478 symbol_section->GetFileAddress(); 3479 if (symbol_byte_size == 0 && 3480 function_starts_count > 0) { 3481 addr_t symbol_lookup_file_addr = nlist.n_value; 3482 // Do an exact address match for non-ARM addresses, 3483 // else get the closest since the symbol might be a 3484 // thumb symbol which has an address with bit zero 3485 // set 3486 FunctionStarts::Entry *func_start_entry = 3487 function_starts.FindEntry(symbol_lookup_file_addr, 3488 !is_arm); 3489 if (is_arm && func_start_entry) { 3490 // Verify that the function start address is the 3491 // symbol address (ARM) or the symbol address + 1 3492 // (thumb) 3493 if (func_start_entry->addr != 3494 symbol_lookup_file_addr && 3495 func_start_entry->addr != 3496 (symbol_lookup_file_addr + 1)) { 3497 // Not the right entry, NULL it out... 3498 func_start_entry = NULL; 3499 } 3500 } 3501 if (func_start_entry) { 3502 func_start_entry->data = true; 3503 3504 addr_t symbol_file_addr = func_start_entry->addr; 3505 uint32_t symbol_flags = 0; 3506 if (is_arm) { 3507 if (symbol_file_addr & 1) 3508 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3509 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3510 } 3511 3512 const FunctionStarts::Entry *next_func_start_entry = 3513 function_starts.FindNextEntry(func_start_entry); 3514 const addr_t section_end_file_addr = 3515 section_file_addr + 3516 symbol_section->GetByteSize(); 3517 if (next_func_start_entry) { 3518 addr_t next_symbol_file_addr = 3519 next_func_start_entry->addr; 3520 // Be sure the clear the Thumb address bit when 3521 // we calculate the size from the current and 3522 // next address 3523 if (is_arm) 3524 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3525 symbol_byte_size = std::min<lldb::addr_t>( 3526 next_symbol_file_addr - symbol_file_addr, 3527 section_end_file_addr - symbol_file_addr); 3528 } else { 3529 symbol_byte_size = 3530 section_end_file_addr - symbol_file_addr; 3531 } 3532 } 3533 } 3534 symbol_value -= section_file_addr; 3535 } 3536 3537 if (is_debug == false) { 3538 if (type == eSymbolTypeCode) { 3539 // See if we can find a N_FUN entry for any code 3540 // symbols. If we do find a match, and the name 3541 // matches, then we can merge the two into just the 3542 // function symbol to avoid duplicate entries in 3543 // the symbol table 3544 auto range = 3545 N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3546 if (range.first != range.second) { 3547 bool found_it = false; 3548 for (auto pos = range.first; pos != range.second; 3549 ++pos) { 3550 if (sym[sym_idx].GetMangled().GetName( 3551 Mangled::ePreferMangled) == 3552 sym[pos->second].GetMangled().GetName( 3553 Mangled::ePreferMangled)) { 3554 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3555 // We just need the flags from the linker 3556 // symbol, so put these flags 3557 // into the N_FUN flags to avoid duplicate 3558 // symbols in the symbol table 3559 sym[pos->second].SetExternal( 3560 sym[sym_idx].IsExternal()); 3561 sym[pos->second].SetFlags(nlist.n_type << 16 | 3562 nlist.n_desc); 3563 if (resolver_addresses.find(nlist.n_value) != 3564 resolver_addresses.end()) 3565 sym[pos->second].SetType(eSymbolTypeResolver); 3566 sym[sym_idx].Clear(); 3567 found_it = true; 3568 break; 3569 } 3570 } 3571 if (found_it) 3572 continue; 3573 } else { 3574 if (resolver_addresses.find(nlist.n_value) != 3575 resolver_addresses.end()) 3576 type = eSymbolTypeResolver; 3577 } 3578 } else if (type == eSymbolTypeData || 3579 type == eSymbolTypeObjCClass || 3580 type == eSymbolTypeObjCMetaClass || 3581 type == eSymbolTypeObjCIVar) { 3582 // See if we can find a N_STSYM entry for any data 3583 // symbols. If we do find a match, and the name 3584 // matches, then we can merge the two into just the 3585 // Static symbol to avoid duplicate entries in the 3586 // symbol table 3587 auto range = N_STSYM_addr_to_sym_idx.equal_range( 3588 nlist.n_value); 3589 if (range.first != range.second) { 3590 bool found_it = false; 3591 for (auto pos = range.first; pos != range.second; 3592 ++pos) { 3593 if (sym[sym_idx].GetMangled().GetName( 3594 Mangled::ePreferMangled) == 3595 sym[pos->second].GetMangled().GetName( 3596 Mangled::ePreferMangled)) { 3597 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3598 // We just need the flags from the linker 3599 // symbol, so put these flags 3600 // into the N_STSYM flags to avoid duplicate 3601 // symbols in the symbol table 3602 sym[pos->second].SetExternal( 3603 sym[sym_idx].IsExternal()); 3604 sym[pos->second].SetFlags(nlist.n_type << 16 | 3605 nlist.n_desc); 3606 sym[sym_idx].Clear(); 3607 found_it = true; 3608 break; 3609 } 3610 } 3611 if (found_it) 3612 continue; 3613 } else { 3614 const char *gsym_name = 3615 sym[sym_idx] 3616 .GetMangled() 3617 .GetName(Mangled::ePreferMangled) 3618 .GetCString(); 3619 if (gsym_name) { 3620 // Combine N_GSYM stab entries with the non 3621 // stab symbol 3622 ConstNameToSymbolIndexMap::const_iterator pos = 3623 N_GSYM_name_to_sym_idx.find(gsym_name); 3624 if (pos != N_GSYM_name_to_sym_idx.end()) { 3625 const uint32_t GSYM_sym_idx = pos->second; 3626 m_nlist_idx_to_sym_idx[nlist_idx] = 3627 GSYM_sym_idx; 3628 // Copy the address, because often the N_GSYM 3629 // address has an invalid address of zero 3630 // when the global is a common symbol 3631 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3632 symbol_section); 3633 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3634 symbol_value); 3635 add_symbol_addr(sym[GSYM_sym_idx] 3636 .GetAddress() 3637 .GetFileAddress()); 3638 // We just need the flags from the linker 3639 // symbol, so put these flags 3640 // into the N_GSYM flags to avoid duplicate 3641 // symbols in the symbol table 3642 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 3643 nlist.n_desc); 3644 sym[sym_idx].Clear(); 3645 continue; 3646 } 3647 } 3648 } 3649 } 3650 } 3651 3652 sym[sym_idx].SetID(nlist_idx); 3653 sym[sym_idx].SetType(type); 3654 if (set_value) { 3655 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 3656 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3657 add_symbol_addr( 3658 sym[sym_idx].GetAddress().GetFileAddress()); 3659 } 3660 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 3661 3662 if (symbol_byte_size > 0) 3663 sym[sym_idx].SetByteSize(symbol_byte_size); 3664 3665 if (demangled_is_synthesized) 3666 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3667 ++sym_idx; 3668 } else { 3669 sym[sym_idx].Clear(); 3670 } 3671 } 3672 ///////////////////////////// 3673 } 3674 } 3675 3676 for (const auto &pos : reexport_shlib_needs_fixup) { 3677 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3678 if (undef_pos != undefined_name_to_desc.end()) { 3679 const uint8_t dylib_ordinal = 3680 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3681 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3682 sym[pos.first].SetReExportedSymbolSharedLibrary( 3683 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3684 } 3685 } 3686 } 3687 3688 #endif 3689 lldb::offset_t nlist_data_offset = 0; 3690 3691 if (nlist_data.GetByteSize() > 0) { 3692 3693 // If the sym array was not created while parsing the DSC unmapped 3694 // symbols, create it now. 3695 if (sym == nullptr) { 3696 sym = 3697 symtab.Resize(symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3698 num_syms = symtab.GetNumSymbols(); 3699 } 3700 3701 if (unmapped_local_symbols_found) { 3702 assert(m_dysymtab.ilocalsym == 0); 3703 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3704 nlist_idx = m_dysymtab.nlocalsym; 3705 } else { 3706 nlist_idx = 0; 3707 } 3708 3709 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 3710 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 3711 UndefinedNameToDescMap undefined_name_to_desc; 3712 SymbolIndexToName reexport_shlib_needs_fixup; 3713 3714 // Symtab parsing is a huge mess. Everything is entangled and the code 3715 // requires access to a ridiculous amount of variables. LLDB depends 3716 // heavily on the proper merging of symbols and to get that right we need 3717 // to make sure we have parsed all the debug symbols first. Therefore we 3718 // invoke the lambda twice, once to parse only the debug symbols and then 3719 // once more to parse the remaining symbols. 3720 auto ParseSymbolLambda = [&](struct nlist_64 &nlist, uint32_t nlist_idx, 3721 bool debug_only) { 3722 const bool is_debug = ((nlist.n_type & N_STAB) != 0); 3723 if (is_debug != debug_only) 3724 return true; 3725 3726 const char *symbol_name_non_abi_mangled = nullptr; 3727 const char *symbol_name = nullptr; 3728 3729 if (have_strtab_data) { 3730 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3731 3732 if (symbol_name == nullptr) { 3733 // No symbol should be NULL, even the symbols with no string values 3734 // should have an offset zero which points to an empty C-string 3735 Host::SystemLog(Host::eSystemLogError, 3736 "error: symbol[%u] has invalid string table offset " 3737 "0x%x in %s, ignoring symbol\n", 3738 nlist_idx, nlist.n_strx, 3739 module_sp->GetFileSpec().GetPath().c_str()); 3740 return true; 3741 } 3742 if (symbol_name[0] == '\0') 3743 symbol_name = nullptr; 3744 } else { 3745 const addr_t str_addr = strtab_addr + nlist.n_strx; 3746 Status str_error; 3747 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3748 str_error)) 3749 symbol_name = memory_symbol_name.c_str(); 3750 } 3751 3752 SymbolType type = eSymbolTypeInvalid; 3753 SectionSP symbol_section; 3754 lldb::addr_t symbol_byte_size = 0; 3755 bool add_nlist = true; 3756 bool is_gsym = false; 3757 bool demangled_is_synthesized = false; 3758 bool set_value = true; 3759 3760 assert(sym_idx < num_syms); 3761 sym[sym_idx].SetDebug(is_debug); 3762 3763 if (is_debug) { 3764 switch (nlist.n_type) { 3765 case N_GSYM: 3766 // global symbol: name,,NO_SECT,type,0 3767 // Sometimes the N_GSYM value contains the address. 3768 3769 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3770 // the ObjC data. They 3771 // have the same address, but we want to ensure that we always find 3772 // only the real symbol, 'cause we don't currently correctly 3773 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3774 // type. This is a temporary hack to make sure the ObjectiveC 3775 // symbols get treated correctly. To do this right, we should 3776 // coalesce all the GSYM & global symbols that have the same 3777 // address. 3778 is_gsym = true; 3779 sym[sym_idx].SetExternal(true); 3780 3781 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3782 llvm::StringRef symbol_name_ref(symbol_name); 3783 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3784 symbol_name_non_abi_mangled = symbol_name + 1; 3785 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3786 type = eSymbolTypeObjCClass; 3787 demangled_is_synthesized = true; 3788 3789 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 3790 symbol_name_non_abi_mangled = symbol_name + 1; 3791 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3792 type = eSymbolTypeObjCMetaClass; 3793 demangled_is_synthesized = true; 3794 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3795 symbol_name_non_abi_mangled = symbol_name + 1; 3796 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3797 type = eSymbolTypeObjCIVar; 3798 demangled_is_synthesized = true; 3799 } 3800 } else { 3801 if (nlist.n_value != 0) 3802 symbol_section = 3803 section_info.GetSection(nlist.n_sect, nlist.n_value); 3804 type = eSymbolTypeData; 3805 } 3806 break; 3807 3808 case N_FNAME: 3809 // procedure name (f77 kludge): name,,NO_SECT,0,0 3810 type = eSymbolTypeCompiler; 3811 break; 3812 3813 case N_FUN: 3814 // procedure: name,,n_sect,linenumber,address 3815 if (symbol_name) { 3816 type = eSymbolTypeCode; 3817 symbol_section = 3818 section_info.GetSection(nlist.n_sect, nlist.n_value); 3819 3820 N_FUN_addr_to_sym_idx.insert( 3821 std::make_pair(nlist.n_value, sym_idx)); 3822 // We use the current number of symbols in the symbol table in 3823 // lieu of using nlist_idx in case we ever start trimming entries 3824 // out 3825 N_FUN_indexes.push_back(sym_idx); 3826 } else { 3827 type = eSymbolTypeCompiler; 3828 3829 if (!N_FUN_indexes.empty()) { 3830 // Copy the size of the function into the original STAB entry 3831 // so we don't have to hunt for it later 3832 symtab.SymbolAtIndex(N_FUN_indexes.back()) 3833 ->SetByteSize(nlist.n_value); 3834 N_FUN_indexes.pop_back(); 3835 // We don't really need the end function STAB as it contains 3836 // the size which we already placed with the original symbol, 3837 // so don't add it if we want a minimal symbol table 3838 add_nlist = false; 3839 } 3840 } 3841 break; 3842 3843 case N_STSYM: 3844 // static symbol: name,,n_sect,type,address 3845 N_STSYM_addr_to_sym_idx.insert( 3846 std::make_pair(nlist.n_value, sym_idx)); 3847 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3848 if (symbol_name && symbol_name[0]) { 3849 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3850 eSymbolTypeData); 3851 } 3852 break; 3853 3854 case N_LCSYM: 3855 // .lcomm symbol: name,,n_sect,type,address 3856 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3857 type = eSymbolTypeCommonBlock; 3858 break; 3859 3860 case N_BNSYM: 3861 // We use the current number of symbols in the symbol table in lieu 3862 // of using nlist_idx in case we ever start trimming entries out 3863 // Skip these if we want minimal symbol tables 3864 add_nlist = false; 3865 break; 3866 3867 case N_ENSYM: 3868 // Set the size of the N_BNSYM to the terminating index of this 3869 // N_ENSYM so that we can always skip the entire symbol if we need 3870 // to navigate more quickly at the source level when parsing STABS 3871 // Skip these if we want minimal symbol tables 3872 add_nlist = false; 3873 break; 3874 3875 case N_OPT: 3876 // emitted with gcc2_compiled and in gcc source 3877 type = eSymbolTypeCompiler; 3878 break; 3879 3880 case N_RSYM: 3881 // register sym: name,,NO_SECT,type,register 3882 type = eSymbolTypeVariable; 3883 break; 3884 3885 case N_SLINE: 3886 // src line: 0,,n_sect,linenumber,address 3887 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3888 type = eSymbolTypeLineEntry; 3889 break; 3890 3891 case N_SSYM: 3892 // structure elt: name,,NO_SECT,type,struct_offset 3893 type = eSymbolTypeVariableType; 3894 break; 3895 3896 case N_SO: 3897 // source file name 3898 type = eSymbolTypeSourceFile; 3899 if (symbol_name == nullptr) { 3900 add_nlist = false; 3901 if (N_SO_index != UINT32_MAX) { 3902 // Set the size of the N_SO to the terminating index of this 3903 // N_SO so that we can always skip the entire N_SO if we need 3904 // to navigate more quickly at the source level when parsing 3905 // STABS 3906 symbol_ptr = symtab.SymbolAtIndex(N_SO_index); 3907 symbol_ptr->SetByteSize(sym_idx); 3908 symbol_ptr->SetSizeIsSibling(true); 3909 } 3910 N_NSYM_indexes.clear(); 3911 N_INCL_indexes.clear(); 3912 N_BRAC_indexes.clear(); 3913 N_COMM_indexes.clear(); 3914 N_FUN_indexes.clear(); 3915 N_SO_index = UINT32_MAX; 3916 } else { 3917 // We use the current number of symbols in the symbol table in 3918 // lieu of using nlist_idx in case we ever start trimming entries 3919 // out 3920 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3921 if (N_SO_has_full_path) { 3922 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3923 // We have two consecutive N_SO entries where the first 3924 // contains a directory and the second contains a full path. 3925 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), 3926 false); 3927 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3928 add_nlist = false; 3929 } else { 3930 // This is the first entry in a N_SO that contains a 3931 // directory or a full path to the source file 3932 N_SO_index = sym_idx; 3933 } 3934 } else if ((N_SO_index == sym_idx - 1) && 3935 ((sym_idx - 1) < num_syms)) { 3936 // This is usually the second N_SO entry that contains just the 3937 // filename, so here we combine it with the first one if we are 3938 // minimizing the symbol table 3939 const char *so_path = 3940 sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3941 if (so_path && so_path[0]) { 3942 std::string full_so_path(so_path); 3943 const size_t double_slash_pos = full_so_path.find("//"); 3944 if (double_slash_pos != std::string::npos) { 3945 // The linker has been generating bad N_SO entries with 3946 // doubled up paths in the format "%s%s" where the first 3947 // string in the DW_AT_comp_dir, and the second is the 3948 // directory for the source file so you end up with a path 3949 // that looks like "/tmp/src//tmp/src/" 3950 FileSpec so_dir(so_path); 3951 if (!FileSystem::Instance().Exists(so_dir)) { 3952 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 3953 FileSpec::Style::native); 3954 if (FileSystem::Instance().Exists(so_dir)) { 3955 // Trim off the incorrect path 3956 full_so_path.erase(0, double_slash_pos + 1); 3957 } 3958 } 3959 } 3960 if (*full_so_path.rbegin() != '/') 3961 full_so_path += '/'; 3962 full_so_path += symbol_name; 3963 sym[sym_idx - 1].GetMangled().SetValue( 3964 ConstString(full_so_path.c_str()), false); 3965 add_nlist = false; 3966 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3967 } 3968 } else { 3969 // This could be a relative path to a N_SO 3970 N_SO_index = sym_idx; 3971 } 3972 } 3973 break; 3974 3975 case N_OSO: 3976 // object file name: name,,0,0,st_mtime 3977 type = eSymbolTypeObjectFile; 3978 break; 3979 3980 case N_LSYM: 3981 // local sym: name,,NO_SECT,type,offset 3982 type = eSymbolTypeLocal; 3983 break; 3984 3985 // INCL scopes 3986 case N_BINCL: 3987 // include file beginning: name,,NO_SECT,0,sum We use the current 3988 // number of symbols in the symbol table in lieu of using nlist_idx 3989 // in case we ever start trimming entries out 3990 N_INCL_indexes.push_back(sym_idx); 3991 type = eSymbolTypeScopeBegin; 3992 break; 3993 3994 case N_EINCL: 3995 // include file end: name,,NO_SECT,0,0 3996 // Set the size of the N_BINCL to the terminating index of this 3997 // N_EINCL so that we can always skip the entire symbol if we need 3998 // to navigate more quickly at the source level when parsing STABS 3999 if (!N_INCL_indexes.empty()) { 4000 symbol_ptr = symtab.SymbolAtIndex(N_INCL_indexes.back()); 4001 symbol_ptr->SetByteSize(sym_idx + 1); 4002 symbol_ptr->SetSizeIsSibling(true); 4003 N_INCL_indexes.pop_back(); 4004 } 4005 type = eSymbolTypeScopeEnd; 4006 break; 4007 4008 case N_SOL: 4009 // #included file name: name,,n_sect,0,address 4010 type = eSymbolTypeHeaderFile; 4011 4012 // We currently don't use the header files on darwin 4013 add_nlist = false; 4014 break; 4015 4016 case N_PARAMS: 4017 // compiler parameters: name,,NO_SECT,0,0 4018 type = eSymbolTypeCompiler; 4019 break; 4020 4021 case N_VERSION: 4022 // compiler version: name,,NO_SECT,0,0 4023 type = eSymbolTypeCompiler; 4024 break; 4025 4026 case N_OLEVEL: 4027 // compiler -O level: name,,NO_SECT,0,0 4028 type = eSymbolTypeCompiler; 4029 break; 4030 4031 case N_PSYM: 4032 // parameter: name,,NO_SECT,type,offset 4033 type = eSymbolTypeVariable; 4034 break; 4035 4036 case N_ENTRY: 4037 // alternate entry: name,,n_sect,linenumber,address 4038 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4039 type = eSymbolTypeLineEntry; 4040 break; 4041 4042 // Left and Right Braces 4043 case N_LBRAC: 4044 // left bracket: 0,,NO_SECT,nesting level,address We use the 4045 // current number of symbols in the symbol table in lieu of using 4046 // nlist_idx in case we ever start trimming entries out 4047 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4048 N_BRAC_indexes.push_back(sym_idx); 4049 type = eSymbolTypeScopeBegin; 4050 break; 4051 4052 case N_RBRAC: 4053 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4054 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4055 // can always skip the entire symbol if we need to navigate more 4056 // quickly at the source level when parsing STABS 4057 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4058 if (!N_BRAC_indexes.empty()) { 4059 symbol_ptr = symtab.SymbolAtIndex(N_BRAC_indexes.back()); 4060 symbol_ptr->SetByteSize(sym_idx + 1); 4061 symbol_ptr->SetSizeIsSibling(true); 4062 N_BRAC_indexes.pop_back(); 4063 } 4064 type = eSymbolTypeScopeEnd; 4065 break; 4066 4067 case N_EXCL: 4068 // deleted include file: name,,NO_SECT,0,sum 4069 type = eSymbolTypeHeaderFile; 4070 break; 4071 4072 // COMM scopes 4073 case N_BCOMM: 4074 // begin common: name,,NO_SECT,0,0 4075 // We use the current number of symbols in the symbol table in lieu 4076 // of using nlist_idx in case we ever start trimming entries out 4077 type = eSymbolTypeScopeBegin; 4078 N_COMM_indexes.push_back(sym_idx); 4079 break; 4080 4081 case N_ECOML: 4082 // end common (local name): 0,,n_sect,0,address 4083 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4084 LLVM_FALLTHROUGH; 4085 4086 case N_ECOMM: 4087 // end common: name,,n_sect,0,0 4088 // Set the size of the N_BCOMM to the terminating index of this 4089 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4090 // we need to navigate more quickly at the source level when 4091 // parsing STABS 4092 if (!N_COMM_indexes.empty()) { 4093 symbol_ptr = symtab.SymbolAtIndex(N_COMM_indexes.back()); 4094 symbol_ptr->SetByteSize(sym_idx + 1); 4095 symbol_ptr->SetSizeIsSibling(true); 4096 N_COMM_indexes.pop_back(); 4097 } 4098 type = eSymbolTypeScopeEnd; 4099 break; 4100 4101 case N_LENG: 4102 // second stab entry with length information 4103 type = eSymbolTypeAdditional; 4104 break; 4105 4106 default: 4107 break; 4108 } 4109 } else { 4110 uint8_t n_type = N_TYPE & nlist.n_type; 4111 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4112 4113 switch (n_type) { 4114 case N_INDR: { 4115 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4116 if (reexport_name_cstr && reexport_name_cstr[0]) { 4117 type = eSymbolTypeReExported; 4118 ConstString reexport_name(reexport_name_cstr + 4119 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4120 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4121 set_value = false; 4122 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4123 indirect_symbol_names.insert( 4124 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4125 } else 4126 type = eSymbolTypeUndefined; 4127 } break; 4128 4129 case N_UNDF: 4130 if (symbol_name && symbol_name[0]) { 4131 ConstString undefined_name(symbol_name + 4132 ((symbol_name[0] == '_') ? 1 : 0)); 4133 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4134 } 4135 LLVM_FALLTHROUGH; 4136 4137 case N_PBUD: 4138 type = eSymbolTypeUndefined; 4139 break; 4140 4141 case N_ABS: 4142 type = eSymbolTypeAbsolute; 4143 break; 4144 4145 case N_SECT: { 4146 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4147 4148 if (!symbol_section) { 4149 // TODO: warn about this? 4150 add_nlist = false; 4151 break; 4152 } 4153 4154 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4155 type = eSymbolTypeException; 4156 } else { 4157 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4158 4159 switch (section_type) { 4160 case S_CSTRING_LITERALS: 4161 type = eSymbolTypeData; 4162 break; // section with only literal C strings 4163 case S_4BYTE_LITERALS: 4164 type = eSymbolTypeData; 4165 break; // section with only 4 byte literals 4166 case S_8BYTE_LITERALS: 4167 type = eSymbolTypeData; 4168 break; // section with only 8 byte literals 4169 case S_LITERAL_POINTERS: 4170 type = eSymbolTypeTrampoline; 4171 break; // section with only pointers to literals 4172 case S_NON_LAZY_SYMBOL_POINTERS: 4173 type = eSymbolTypeTrampoline; 4174 break; // section with only non-lazy symbol pointers 4175 case S_LAZY_SYMBOL_POINTERS: 4176 type = eSymbolTypeTrampoline; 4177 break; // section with only lazy symbol pointers 4178 case S_SYMBOL_STUBS: 4179 type = eSymbolTypeTrampoline; 4180 break; // section with only symbol stubs, byte size of stub in 4181 // the reserved2 field 4182 case S_MOD_INIT_FUNC_POINTERS: 4183 type = eSymbolTypeCode; 4184 break; // section with only function pointers for initialization 4185 case S_MOD_TERM_FUNC_POINTERS: 4186 type = eSymbolTypeCode; 4187 break; // section with only function pointers for termination 4188 case S_INTERPOSING: 4189 type = eSymbolTypeTrampoline; 4190 break; // section with only pairs of function pointers for 4191 // interposing 4192 case S_16BYTE_LITERALS: 4193 type = eSymbolTypeData; 4194 break; // section with only 16 byte literals 4195 case S_DTRACE_DOF: 4196 type = eSymbolTypeInstrumentation; 4197 break; 4198 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4199 type = eSymbolTypeTrampoline; 4200 break; 4201 default: 4202 switch (symbol_section->GetType()) { 4203 case lldb::eSectionTypeCode: 4204 type = eSymbolTypeCode; 4205 break; 4206 case eSectionTypeData: 4207 case eSectionTypeDataCString: // Inlined C string data 4208 case eSectionTypeDataCStringPointers: // Pointers to C string 4209 // data 4210 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4211 // the symbol table 4212 case eSectionTypeData4: 4213 case eSectionTypeData8: 4214 case eSectionTypeData16: 4215 type = eSymbolTypeData; 4216 break; 4217 default: 4218 break; 4219 } 4220 break; 4221 } 4222 4223 if (type == eSymbolTypeInvalid) { 4224 const char *symbol_sect_name = 4225 symbol_section->GetName().AsCString(); 4226 if (symbol_section->IsDescendant(text_section_sp.get())) { 4227 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4228 S_ATTR_SELF_MODIFYING_CODE | 4229 S_ATTR_SOME_INSTRUCTIONS)) 4230 type = eSymbolTypeData; 4231 else 4232 type = eSymbolTypeCode; 4233 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 4234 symbol_section->IsDescendant( 4235 data_dirty_section_sp.get()) || 4236 symbol_section->IsDescendant( 4237 data_const_section_sp.get())) { 4238 if (symbol_sect_name && 4239 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 4240 type = eSymbolTypeRuntime; 4241 4242 if (symbol_name) { 4243 llvm::StringRef symbol_name_ref(symbol_name); 4244 if (symbol_name_ref.startswith("_OBJC_")) { 4245 llvm::StringRef g_objc_v2_prefix_class( 4246 "_OBJC_CLASS_$_"); 4247 llvm::StringRef g_objc_v2_prefix_metaclass( 4248 "_OBJC_METACLASS_$_"); 4249 llvm::StringRef g_objc_v2_prefix_ivar( 4250 "_OBJC_IVAR_$_"); 4251 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 4252 symbol_name_non_abi_mangled = symbol_name + 1; 4253 symbol_name = 4254 symbol_name + g_objc_v2_prefix_class.size(); 4255 type = eSymbolTypeObjCClass; 4256 demangled_is_synthesized = true; 4257 } else if (symbol_name_ref.startswith( 4258 g_objc_v2_prefix_metaclass)) { 4259 symbol_name_non_abi_mangled = symbol_name + 1; 4260 symbol_name = 4261 symbol_name + g_objc_v2_prefix_metaclass.size(); 4262 type = eSymbolTypeObjCMetaClass; 4263 demangled_is_synthesized = true; 4264 } else if (symbol_name_ref.startswith( 4265 g_objc_v2_prefix_ivar)) { 4266 symbol_name_non_abi_mangled = symbol_name + 1; 4267 symbol_name = 4268 symbol_name + g_objc_v2_prefix_ivar.size(); 4269 type = eSymbolTypeObjCIVar; 4270 demangled_is_synthesized = true; 4271 } 4272 } 4273 } 4274 } else if (symbol_sect_name && 4275 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4276 symbol_sect_name) { 4277 type = eSymbolTypeException; 4278 } else { 4279 type = eSymbolTypeData; 4280 } 4281 } else if (symbol_sect_name && 4282 ::strstr(symbol_sect_name, "__IMPORT") == 4283 symbol_sect_name) { 4284 type = eSymbolTypeTrampoline; 4285 } else if (symbol_section->IsDescendant(objc_section_sp.get())) { 4286 type = eSymbolTypeRuntime; 4287 if (symbol_name && symbol_name[0] == '.') { 4288 llvm::StringRef symbol_name_ref(symbol_name); 4289 llvm::StringRef g_objc_v1_prefix_class( 4290 ".objc_class_name_"); 4291 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4292 symbol_name_non_abi_mangled = symbol_name; 4293 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4294 type = eSymbolTypeObjCClass; 4295 demangled_is_synthesized = true; 4296 } 4297 } 4298 } 4299 } 4300 } 4301 } break; 4302 } 4303 } 4304 4305 if (!add_nlist) { 4306 sym[sym_idx].Clear(); 4307 return true; 4308 } 4309 4310 uint64_t symbol_value = nlist.n_value; 4311 4312 if (symbol_name_non_abi_mangled) { 4313 sym[sym_idx].GetMangled().SetMangledName( 4314 ConstString(symbol_name_non_abi_mangled)); 4315 sym[sym_idx].GetMangled().SetDemangledName(ConstString(symbol_name)); 4316 } else { 4317 bool symbol_name_is_mangled = false; 4318 4319 if (symbol_name && symbol_name[0] == '_') { 4320 symbol_name_is_mangled = symbol_name[1] == '_'; 4321 symbol_name++; // Skip the leading underscore 4322 } 4323 4324 if (symbol_name) { 4325 ConstString const_symbol_name(symbol_name); 4326 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4327 symbol_name_is_mangled); 4328 } 4329 } 4330 4331 if (is_gsym) { 4332 const char *gsym_name = sym[sym_idx] 4333 .GetMangled() 4334 .GetName(Mangled::ePreferMangled) 4335 .GetCString(); 4336 if (gsym_name) 4337 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4338 } 4339 4340 if (symbol_section) { 4341 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4342 if (symbol_byte_size == 0 && function_starts_count > 0) { 4343 addr_t symbol_lookup_file_addr = nlist.n_value; 4344 // Do an exact address match for non-ARM addresses, else get the 4345 // closest since the symbol might be a thumb symbol which has an 4346 // address with bit zero set. 4347 FunctionStarts::Entry *func_start_entry = 4348 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4349 if (is_arm && func_start_entry) { 4350 // Verify that the function start address is the symbol address 4351 // (ARM) or the symbol address + 1 (thumb). 4352 if (func_start_entry->addr != symbol_lookup_file_addr && 4353 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4354 // Not the right entry, NULL it out... 4355 func_start_entry = nullptr; 4356 } 4357 } 4358 if (func_start_entry) { 4359 func_start_entry->data = true; 4360 4361 addr_t symbol_file_addr = func_start_entry->addr; 4362 if (is_arm) 4363 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4364 4365 const FunctionStarts::Entry *next_func_start_entry = 4366 function_starts.FindNextEntry(func_start_entry); 4367 const addr_t section_end_file_addr = 4368 section_file_addr + symbol_section->GetByteSize(); 4369 if (next_func_start_entry) { 4370 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4371 // Be sure the clear the Thumb address bit when we calculate the 4372 // size from the current and next address 4373 if (is_arm) 4374 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4375 symbol_byte_size = std::min<lldb::addr_t>( 4376 next_symbol_file_addr - symbol_file_addr, 4377 section_end_file_addr - symbol_file_addr); 4378 } else { 4379 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4380 } 4381 } 4382 } 4383 symbol_value -= section_file_addr; 4384 } 4385 4386 if (!is_debug) { 4387 if (type == eSymbolTypeCode) { 4388 // See if we can find a N_FUN entry for any code symbols. If we do 4389 // find a match, and the name matches, then we can merge the two into 4390 // just the function symbol to avoid duplicate entries in the symbol 4391 // table. 4392 std::pair<ValueToSymbolIndexMap::const_iterator, 4393 ValueToSymbolIndexMap::const_iterator> 4394 range; 4395 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4396 if (range.first != range.second) { 4397 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4398 pos != range.second; ++pos) { 4399 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4400 sym[pos->second].GetMangled().GetName( 4401 Mangled::ePreferMangled)) { 4402 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4403 // We just need the flags from the linker symbol, so put these 4404 // flags into the N_FUN flags to avoid duplicate symbols in the 4405 // symbol table. 4406 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4407 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4408 if (resolver_addresses.find(nlist.n_value) != 4409 resolver_addresses.end()) 4410 sym[pos->second].SetType(eSymbolTypeResolver); 4411 sym[sym_idx].Clear(); 4412 return true; 4413 } 4414 } 4415 } else { 4416 if (resolver_addresses.find(nlist.n_value) != 4417 resolver_addresses.end()) 4418 type = eSymbolTypeResolver; 4419 } 4420 } else if (type == eSymbolTypeData || type == eSymbolTypeObjCClass || 4421 type == eSymbolTypeObjCMetaClass || 4422 type == eSymbolTypeObjCIVar) { 4423 // See if we can find a N_STSYM entry for any data symbols. If we do 4424 // find a match, and the name matches, then we can merge the two into 4425 // just the Static symbol to avoid duplicate entries in the symbol 4426 // table. 4427 std::pair<ValueToSymbolIndexMap::const_iterator, 4428 ValueToSymbolIndexMap::const_iterator> 4429 range; 4430 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4431 if (range.first != range.second) { 4432 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4433 pos != range.second; ++pos) { 4434 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4435 sym[pos->second].GetMangled().GetName( 4436 Mangled::ePreferMangled)) { 4437 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4438 // We just need the flags from the linker symbol, so put these 4439 // flags into the N_STSYM flags to avoid duplicate symbols in 4440 // the symbol table. 4441 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4442 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4443 sym[sym_idx].Clear(); 4444 return true; 4445 } 4446 } 4447 } else { 4448 // Combine N_GSYM stab entries with the non stab symbol. 4449 const char *gsym_name = sym[sym_idx] 4450 .GetMangled() 4451 .GetName(Mangled::ePreferMangled) 4452 .GetCString(); 4453 if (gsym_name) { 4454 ConstNameToSymbolIndexMap::const_iterator pos = 4455 N_GSYM_name_to_sym_idx.find(gsym_name); 4456 if (pos != N_GSYM_name_to_sym_idx.end()) { 4457 const uint32_t GSYM_sym_idx = pos->second; 4458 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4459 // Copy the address, because often the N_GSYM address has an 4460 // invalid address of zero when the global is a common symbol. 4461 sym[GSYM_sym_idx].GetAddressRef().SetSection(symbol_section); 4462 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4463 add_symbol_addr( 4464 sym[GSYM_sym_idx].GetAddress().GetFileAddress()); 4465 // We just need the flags from the linker symbol, so put these 4466 // flags into the N_GSYM flags to avoid duplicate symbols in 4467 // the symbol table. 4468 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4469 sym[sym_idx].Clear(); 4470 return true; 4471 } 4472 } 4473 } 4474 } 4475 } 4476 4477 sym[sym_idx].SetID(nlist_idx); 4478 sym[sym_idx].SetType(type); 4479 if (set_value) { 4480 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4481 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4482 if (symbol_section) 4483 add_symbol_addr(sym[sym_idx].GetAddress().GetFileAddress()); 4484 } 4485 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4486 if (nlist.n_desc & N_WEAK_REF) 4487 sym[sym_idx].SetIsWeak(true); 4488 4489 if (symbol_byte_size > 0) 4490 sym[sym_idx].SetByteSize(symbol_byte_size); 4491 4492 if (demangled_is_synthesized) 4493 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4494 4495 ++sym_idx; 4496 return true; 4497 }; 4498 4499 // First parse all the nlists but don't process them yet. See the next 4500 // comment for an explanation why. 4501 std::vector<struct nlist_64> nlists; 4502 nlists.reserve(symtab_load_command.nsyms); 4503 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 4504 if (auto nlist = 4505 ParseNList(nlist_data, nlist_data_offset, nlist_byte_size)) 4506 nlists.push_back(*nlist); 4507 else 4508 break; 4509 } 4510 4511 // Now parse all the debug symbols. This is needed to merge non-debug 4512 // symbols in the next step. Non-debug symbols are always coalesced into 4513 // the debug symbol. Doing this in one step would mean that some symbols 4514 // won't be merged. 4515 nlist_idx = 0; 4516 for (auto &nlist : nlists) { 4517 if (!ParseSymbolLambda(nlist, nlist_idx++, DebugSymbols)) 4518 break; 4519 } 4520 4521 // Finally parse all the non debug symbols. 4522 nlist_idx = 0; 4523 for (auto &nlist : nlists) { 4524 if (!ParseSymbolLambda(nlist, nlist_idx++, NonDebugSymbols)) 4525 break; 4526 } 4527 4528 for (const auto &pos : reexport_shlib_needs_fixup) { 4529 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4530 if (undef_pos != undefined_name_to_desc.end()) { 4531 const uint8_t dylib_ordinal = 4532 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4533 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4534 sym[pos.first].SetReExportedSymbolSharedLibrary( 4535 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4536 } 4537 } 4538 } 4539 4540 // Count how many trie symbols we'll add to the symbol table 4541 int trie_symbol_table_augment_count = 0; 4542 for (auto &e : external_sym_trie_entries) { 4543 if (symbols_added.find(e.entry.address) == symbols_added.end()) 4544 trie_symbol_table_augment_count++; 4545 } 4546 4547 if (num_syms < sym_idx + trie_symbol_table_augment_count) { 4548 num_syms = sym_idx + trie_symbol_table_augment_count; 4549 sym = symtab.Resize(num_syms); 4550 } 4551 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4552 4553 // Add symbols from the trie to the symbol table. 4554 for (auto &e : external_sym_trie_entries) { 4555 if (symbols_added.contains(e.entry.address)) 4556 continue; 4557 4558 // Find the section that this trie address is in, use that to annotate 4559 // symbol type as we add the trie address and name to the symbol table. 4560 Address symbol_addr; 4561 if (module_sp->ResolveFileAddress(e.entry.address, symbol_addr)) { 4562 SectionSP symbol_section(symbol_addr.GetSection()); 4563 const char *symbol_name = e.entry.name.GetCString(); 4564 bool demangled_is_synthesized = false; 4565 SymbolType type = 4566 GetSymbolType(symbol_name, demangled_is_synthesized, text_section_sp, 4567 data_section_sp, data_dirty_section_sp, 4568 data_const_section_sp, symbol_section); 4569 4570 sym[sym_idx].SetType(type); 4571 if (symbol_section) { 4572 sym[sym_idx].SetID(synthetic_sym_id++); 4573 sym[sym_idx].GetMangled().SetMangledName(ConstString(symbol_name)); 4574 if (demangled_is_synthesized) 4575 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4576 sym[sym_idx].SetIsSynthetic(true); 4577 sym[sym_idx].SetExternal(true); 4578 sym[sym_idx].GetAddressRef() = symbol_addr; 4579 add_symbol_addr(symbol_addr.GetFileAddress()); 4580 if (e.entry.flags & TRIE_SYMBOL_IS_THUMB) 4581 sym[sym_idx].SetFlags(MACHO_NLIST_ARM_SYMBOL_IS_THUMB); 4582 ++sym_idx; 4583 } 4584 } 4585 } 4586 4587 if (function_starts_count > 0) { 4588 uint32_t num_synthetic_function_symbols = 0; 4589 for (i = 0; i < function_starts_count; ++i) { 4590 if (symbols_added.find(function_starts.GetEntryRef(i).addr) == 4591 symbols_added.end()) 4592 ++num_synthetic_function_symbols; 4593 } 4594 4595 if (num_synthetic_function_symbols > 0) { 4596 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4597 num_syms = sym_idx + num_synthetic_function_symbols; 4598 sym = symtab.Resize(num_syms); 4599 } 4600 for (i = 0; i < function_starts_count; ++i) { 4601 const FunctionStarts::Entry *func_start_entry = 4602 function_starts.GetEntryAtIndex(i); 4603 if (symbols_added.find(func_start_entry->addr) == symbols_added.end()) { 4604 addr_t symbol_file_addr = func_start_entry->addr; 4605 uint32_t symbol_flags = 0; 4606 if (func_start_entry->data) 4607 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4608 Address symbol_addr; 4609 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4610 SectionSP symbol_section(symbol_addr.GetSection()); 4611 uint32_t symbol_byte_size = 0; 4612 if (symbol_section) { 4613 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4614 const FunctionStarts::Entry *next_func_start_entry = 4615 function_starts.FindNextEntry(func_start_entry); 4616 const addr_t section_end_file_addr = 4617 section_file_addr + symbol_section->GetByteSize(); 4618 if (next_func_start_entry) { 4619 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4620 if (is_arm) 4621 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4622 symbol_byte_size = std::min<lldb::addr_t>( 4623 next_symbol_file_addr - symbol_file_addr, 4624 section_end_file_addr - symbol_file_addr); 4625 } else { 4626 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4627 } 4628 sym[sym_idx].SetID(synthetic_sym_id++); 4629 // Don't set the name for any synthetic symbols, the Symbol 4630 // object will generate one if needed when the name is accessed 4631 // via accessors. 4632 sym[sym_idx].GetMangled().SetDemangledName(ConstString()); 4633 sym[sym_idx].SetType(eSymbolTypeCode); 4634 sym[sym_idx].SetIsSynthetic(true); 4635 sym[sym_idx].GetAddressRef() = symbol_addr; 4636 add_symbol_addr(symbol_addr.GetFileAddress()); 4637 if (symbol_flags) 4638 sym[sym_idx].SetFlags(symbol_flags); 4639 if (symbol_byte_size) 4640 sym[sym_idx].SetByteSize(symbol_byte_size); 4641 ++sym_idx; 4642 } 4643 } 4644 } 4645 } 4646 } 4647 } 4648 4649 // Trim our symbols down to just what we ended up with after removing any 4650 // symbols. 4651 if (sym_idx < num_syms) { 4652 num_syms = sym_idx; 4653 sym = symtab.Resize(num_syms); 4654 } 4655 4656 // Now synthesize indirect symbols 4657 if (m_dysymtab.nindirectsyms != 0) { 4658 if (indirect_symbol_index_data.GetByteSize()) { 4659 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4660 m_nlist_idx_to_sym_idx.end(); 4661 4662 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4663 ++sect_idx) { 4664 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4665 S_SYMBOL_STUBS) { 4666 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4667 if (symbol_stub_byte_size == 0) 4668 continue; 4669 4670 const uint32_t num_symbol_stubs = 4671 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4672 4673 if (num_symbol_stubs == 0) 4674 continue; 4675 4676 const uint32_t symbol_stub_index_offset = 4677 m_mach_sections[sect_idx].reserved1; 4678 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) { 4679 const uint32_t symbol_stub_index = 4680 symbol_stub_index_offset + stub_idx; 4681 const lldb::addr_t symbol_stub_addr = 4682 m_mach_sections[sect_idx].addr + 4683 (stub_idx * symbol_stub_byte_size); 4684 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4685 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4686 symbol_stub_offset, 4)) { 4687 const uint32_t stub_sym_id = 4688 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4689 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4690 continue; 4691 4692 NListIndexToSymbolIndexMap::const_iterator index_pos = 4693 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4694 Symbol *stub_symbol = nullptr; 4695 if (index_pos != end_index_pos) { 4696 // We have a remapping from the original nlist index to a 4697 // current symbol index, so just look this up by index 4698 stub_symbol = symtab.SymbolAtIndex(index_pos->second); 4699 } else { 4700 // We need to lookup a symbol using the original nlist symbol 4701 // index since this index is coming from the S_SYMBOL_STUBS 4702 stub_symbol = symtab.FindSymbolByID(stub_sym_id); 4703 } 4704 4705 if (stub_symbol) { 4706 Address so_addr(symbol_stub_addr, section_list); 4707 4708 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4709 // Change the external symbol into a trampoline that makes 4710 // sense These symbols were N_UNDF N_EXT, and are useless 4711 // to us, so we can re-use them so we don't have to make up 4712 // a synthetic symbol for no good reason. 4713 if (resolver_addresses.find(symbol_stub_addr) == 4714 resolver_addresses.end()) 4715 stub_symbol->SetType(eSymbolTypeTrampoline); 4716 else 4717 stub_symbol->SetType(eSymbolTypeResolver); 4718 stub_symbol->SetExternal(false); 4719 stub_symbol->GetAddressRef() = so_addr; 4720 stub_symbol->SetByteSize(symbol_stub_byte_size); 4721 } else { 4722 // Make a synthetic symbol to describe the trampoline stub 4723 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4724 if (sym_idx >= num_syms) { 4725 sym = symtab.Resize(++num_syms); 4726 stub_symbol = nullptr; // this pointer no longer valid 4727 } 4728 sym[sym_idx].SetID(synthetic_sym_id++); 4729 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4730 if (resolver_addresses.find(symbol_stub_addr) == 4731 resolver_addresses.end()) 4732 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4733 else 4734 sym[sym_idx].SetType(eSymbolTypeResolver); 4735 sym[sym_idx].SetIsSynthetic(true); 4736 sym[sym_idx].GetAddressRef() = so_addr; 4737 add_symbol_addr(so_addr.GetFileAddress()); 4738 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4739 ++sym_idx; 4740 } 4741 } else { 4742 if (log) 4743 log->Warning("symbol stub referencing symbol table symbol " 4744 "%u that isn't in our minimal symbol table, " 4745 "fix this!!!", 4746 stub_sym_id); 4747 } 4748 } 4749 } 4750 } 4751 } 4752 } 4753 } 4754 4755 if (!reexport_trie_entries.empty()) { 4756 for (const auto &e : reexport_trie_entries) { 4757 if (e.entry.import_name) { 4758 // Only add indirect symbols from the Trie entries if we didn't have 4759 // a N_INDR nlist entry for this already 4760 if (indirect_symbol_names.find(e.entry.name) == 4761 indirect_symbol_names.end()) { 4762 // Make a synthetic symbol to describe re-exported symbol. 4763 if (sym_idx >= num_syms) 4764 sym = symtab.Resize(++num_syms); 4765 sym[sym_idx].SetID(synthetic_sym_id++); 4766 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4767 sym[sym_idx].SetType(eSymbolTypeReExported); 4768 sym[sym_idx].SetIsSynthetic(true); 4769 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4770 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4771 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4772 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4773 } 4774 ++sym_idx; 4775 } 4776 } 4777 } 4778 } 4779 } 4780 4781 void ObjectFileMachO::Dump(Stream *s) { 4782 ModuleSP module_sp(GetModule()); 4783 if (module_sp) { 4784 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4785 s->Printf("%p: ", static_cast<void *>(this)); 4786 s->Indent(); 4787 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4788 s->PutCString("ObjectFileMachO64"); 4789 else 4790 s->PutCString("ObjectFileMachO32"); 4791 4792 *s << ", file = '" << m_file; 4793 ModuleSpecList all_specs; 4794 ModuleSpec base_spec; 4795 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 4796 base_spec, all_specs); 4797 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 4798 *s << "', triple"; 4799 if (e) 4800 s->Printf("[%d]", i); 4801 *s << " = "; 4802 *s << all_specs.GetModuleSpecRefAtIndex(i) 4803 .GetArchitecture() 4804 .GetTriple() 4805 .getTriple(); 4806 } 4807 *s << "\n"; 4808 SectionList *sections = GetSectionList(); 4809 if (sections) 4810 sections->Dump(s->AsRawOstream(), s->GetIndentLevel(), nullptr, true, 4811 UINT32_MAX); 4812 4813 if (m_symtab_up) 4814 m_symtab_up->Dump(s, nullptr, eSortOrderNone); 4815 } 4816 } 4817 4818 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4819 const lldb_private::DataExtractor &data, 4820 lldb::offset_t lc_offset) { 4821 uint32_t i; 4822 llvm::MachO::uuid_command load_cmd; 4823 4824 lldb::offset_t offset = lc_offset; 4825 for (i = 0; i < header.ncmds; ++i) { 4826 const lldb::offset_t cmd_offset = offset; 4827 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4828 break; 4829 4830 if (load_cmd.cmd == LC_UUID) { 4831 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4832 4833 if (uuid_bytes) { 4834 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4835 // We pretend these object files have no UUID to prevent crashing. 4836 4837 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4838 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4839 0xbb, 0x14, 0xf0, 0x0d}; 4840 4841 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4842 return UUID(); 4843 4844 return UUID::fromOptionalData(uuid_bytes, 16); 4845 } 4846 return UUID(); 4847 } 4848 offset = cmd_offset + load_cmd.cmdsize; 4849 } 4850 return UUID(); 4851 } 4852 4853 static llvm::StringRef GetOSName(uint32_t cmd) { 4854 switch (cmd) { 4855 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4856 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4857 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4858 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4859 case llvm::MachO::LC_VERSION_MIN_TVOS: 4860 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4861 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4862 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4863 default: 4864 llvm_unreachable("unexpected LC_VERSION load command"); 4865 } 4866 } 4867 4868 namespace { 4869 struct OSEnv { 4870 llvm::StringRef os_type; 4871 llvm::StringRef environment; 4872 OSEnv(uint32_t cmd) { 4873 switch (cmd) { 4874 case llvm::MachO::PLATFORM_MACOS: 4875 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4876 return; 4877 case llvm::MachO::PLATFORM_IOS: 4878 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4879 return; 4880 case llvm::MachO::PLATFORM_TVOS: 4881 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4882 return; 4883 case llvm::MachO::PLATFORM_WATCHOS: 4884 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4885 return; 4886 // TODO: add BridgeOS & DriverKit once in llvm/lib/Support/Triple.cpp 4887 // NEED_BRIDGEOS_TRIPLE 4888 // case llvm::MachO::PLATFORM_BRIDGEOS: 4889 // os_type = llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4890 // return; 4891 // case llvm::MachO::PLATFORM_DRIVERKIT: 4892 // os_type = llvm::Triple::getOSTypeName(llvm::Triple::DriverKit); 4893 // return; 4894 case llvm::MachO::PLATFORM_MACCATALYST: 4895 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4896 environment = llvm::Triple::getEnvironmentTypeName(llvm::Triple::MacABI); 4897 return; 4898 case llvm::MachO::PLATFORM_IOSSIMULATOR: 4899 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4900 environment = 4901 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4902 return; 4903 case llvm::MachO::PLATFORM_TVOSSIMULATOR: 4904 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4905 environment = 4906 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4907 return; 4908 case llvm::MachO::PLATFORM_WATCHOSSIMULATOR: 4909 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4910 environment = 4911 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4912 return; 4913 default: { 4914 Log *log(GetLog(LLDBLog::Symbols | LLDBLog::Process)); 4915 LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION"); 4916 } 4917 } 4918 } 4919 }; 4920 4921 struct MinOS { 4922 uint32_t major_version, minor_version, patch_version; 4923 MinOS(uint32_t version) 4924 : major_version(version >> 16), minor_version((version >> 8) & 0xffu), 4925 patch_version(version & 0xffu) {} 4926 }; 4927 } // namespace 4928 4929 void ObjectFileMachO::GetAllArchSpecs(const llvm::MachO::mach_header &header, 4930 const lldb_private::DataExtractor &data, 4931 lldb::offset_t lc_offset, 4932 ModuleSpec &base_spec, 4933 lldb_private::ModuleSpecList &all_specs) { 4934 auto &base_arch = base_spec.GetArchitecture(); 4935 base_arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4936 if (!base_arch.IsValid()) 4937 return; 4938 4939 bool found_any = false; 4940 auto add_triple = [&](const llvm::Triple &triple) { 4941 auto spec = base_spec; 4942 spec.GetArchitecture().GetTriple() = triple; 4943 if (spec.GetArchitecture().IsValid()) { 4944 spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset); 4945 all_specs.Append(spec); 4946 found_any = true; 4947 } 4948 }; 4949 4950 // Set OS to an unspecified unknown or a "*" so it can match any OS 4951 llvm::Triple base_triple = base_arch.GetTriple(); 4952 base_triple.setOS(llvm::Triple::UnknownOS); 4953 base_triple.setOSName(llvm::StringRef()); 4954 4955 if (header.filetype == MH_PRELOAD) { 4956 if (header.cputype == CPU_TYPE_ARM) { 4957 // If this is a 32-bit arm binary, and it's a standalone binary, force 4958 // the Vendor to Apple so we don't accidentally pick up the generic 4959 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4960 // frame pointer register; most other armv7 ABIs use a combination of 4961 // r7 and r11. 4962 base_triple.setVendor(llvm::Triple::Apple); 4963 } else { 4964 // Set vendor to an unspecified unknown or a "*" so it can match any 4965 // vendor This is required for correct behavior of EFI debugging on 4966 // x86_64 4967 base_triple.setVendor(llvm::Triple::UnknownVendor); 4968 base_triple.setVendorName(llvm::StringRef()); 4969 } 4970 return add_triple(base_triple); 4971 } 4972 4973 llvm::MachO::load_command load_cmd; 4974 4975 // See if there is an LC_VERSION_MIN_* load command that can give 4976 // us the OS type. 4977 lldb::offset_t offset = lc_offset; 4978 for (uint32_t i = 0; i < header.ncmds; ++i) { 4979 const lldb::offset_t cmd_offset = offset; 4980 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4981 break; 4982 4983 llvm::MachO::version_min_command version_min; 4984 switch (load_cmd.cmd) { 4985 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4986 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4987 case llvm::MachO::LC_VERSION_MIN_TVOS: 4988 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 4989 if (load_cmd.cmdsize != sizeof(version_min)) 4990 break; 4991 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 4992 data.GetByteOrder(), &version_min) == 0) 4993 break; 4994 MinOS min_os(version_min.version); 4995 llvm::SmallString<32> os_name; 4996 llvm::raw_svector_ostream os(os_name); 4997 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 4998 << min_os.minor_version << '.' << min_os.patch_version; 4999 5000 auto triple = base_triple; 5001 triple.setOSName(os.str()); 5002 5003 // Disambiguate legacy simulator platforms. 5004 if (load_cmd.cmd != llvm::MachO::LC_VERSION_MIN_MACOSX && 5005 (base_triple.getArch() == llvm::Triple::x86_64 || 5006 base_triple.getArch() == llvm::Triple::x86)) { 5007 // The combination of legacy LC_VERSION_MIN load command and 5008 // x86 architecture always indicates a simulator environment. 5009 // The combination of LC_VERSION_MIN and arm architecture only 5010 // appears for native binaries. Back-deploying simulator 5011 // binaries on Apple Silicon Macs use the modern unambigous 5012 // LC_BUILD_VERSION load commands; no special handling required. 5013 triple.setEnvironment(llvm::Triple::Simulator); 5014 } 5015 add_triple(triple); 5016 break; 5017 } 5018 default: 5019 break; 5020 } 5021 5022 offset = cmd_offset + load_cmd.cmdsize; 5023 } 5024 5025 // See if there are LC_BUILD_VERSION load commands that can give 5026 // us the OS type. 5027 offset = lc_offset; 5028 for (uint32_t i = 0; i < header.ncmds; ++i) { 5029 const lldb::offset_t cmd_offset = offset; 5030 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 5031 break; 5032 5033 do { 5034 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 5035 llvm::MachO::build_version_command build_version; 5036 if (load_cmd.cmdsize < sizeof(build_version)) { 5037 // Malformed load command. 5038 break; 5039 } 5040 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 5041 data.GetByteOrder(), &build_version) == 0) 5042 break; 5043 MinOS min_os(build_version.minos); 5044 OSEnv os_env(build_version.platform); 5045 llvm::SmallString<16> os_name; 5046 llvm::raw_svector_ostream os(os_name); 5047 os << os_env.os_type << min_os.major_version << '.' 5048 << min_os.minor_version << '.' << min_os.patch_version; 5049 auto triple = base_triple; 5050 triple.setOSName(os.str()); 5051 os_name.clear(); 5052 if (!os_env.environment.empty()) 5053 triple.setEnvironmentName(os_env.environment); 5054 add_triple(triple); 5055 } 5056 } while (false); 5057 offset = cmd_offset + load_cmd.cmdsize; 5058 } 5059 5060 if (!found_any) { 5061 add_triple(base_triple); 5062 } 5063 } 5064 5065 ArchSpec ObjectFileMachO::GetArchitecture( 5066 ModuleSP module_sp, const llvm::MachO::mach_header &header, 5067 const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) { 5068 ModuleSpecList all_specs; 5069 ModuleSpec base_spec; 5070 GetAllArchSpecs(header, data, MachHeaderSizeFromMagic(header.magic), 5071 base_spec, all_specs); 5072 5073 // If the object file offers multiple alternative load commands, 5074 // pick the one that matches the module. 5075 if (module_sp) { 5076 const ArchSpec &module_arch = module_sp->GetArchitecture(); 5077 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 5078 ArchSpec mach_arch = 5079 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 5080 if (module_arch.IsCompatibleMatch(mach_arch)) 5081 return mach_arch; 5082 } 5083 } 5084 5085 // Return the first arch we found. 5086 if (all_specs.GetSize() == 0) 5087 return {}; 5088 return all_specs.GetModuleSpecRefAtIndex(0).GetArchitecture(); 5089 } 5090 5091 UUID ObjectFileMachO::GetUUID() { 5092 ModuleSP module_sp(GetModule()); 5093 if (module_sp) { 5094 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5095 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5096 return GetUUID(m_header, m_data, offset); 5097 } 5098 return UUID(); 5099 } 5100 5101 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5102 uint32_t count = 0; 5103 ModuleSP module_sp(GetModule()); 5104 if (module_sp) { 5105 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5106 llvm::MachO::load_command load_cmd; 5107 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5108 std::vector<std::string> rpath_paths; 5109 std::vector<std::string> rpath_relative_paths; 5110 std::vector<std::string> at_exec_relative_paths; 5111 uint32_t i; 5112 for (i = 0; i < m_header.ncmds; ++i) { 5113 const uint32_t cmd_offset = offset; 5114 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5115 break; 5116 5117 switch (load_cmd.cmd) { 5118 case LC_RPATH: 5119 case LC_LOAD_DYLIB: 5120 case LC_LOAD_WEAK_DYLIB: 5121 case LC_REEXPORT_DYLIB: 5122 case LC_LOAD_DYLINKER: 5123 case LC_LOADFVMLIB: 5124 case LC_LOAD_UPWARD_DYLIB: { 5125 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5126 const char *path = m_data.PeekCStr(name_offset); 5127 if (path) { 5128 if (load_cmd.cmd == LC_RPATH) 5129 rpath_paths.push_back(path); 5130 else { 5131 if (path[0] == '@') { 5132 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5133 rpath_relative_paths.push_back(path + strlen("@rpath")); 5134 else if (strncmp(path, "@executable_path", 5135 strlen("@executable_path")) == 0) 5136 at_exec_relative_paths.push_back(path + 5137 strlen("@executable_path")); 5138 } else { 5139 FileSpec file_spec(path); 5140 if (files.AppendIfUnique(file_spec)) 5141 count++; 5142 } 5143 } 5144 } 5145 } break; 5146 5147 default: 5148 break; 5149 } 5150 offset = cmd_offset + load_cmd.cmdsize; 5151 } 5152 5153 FileSpec this_file_spec(m_file); 5154 FileSystem::Instance().Resolve(this_file_spec); 5155 5156 if (!rpath_paths.empty()) { 5157 // Fixup all LC_RPATH values to be absolute paths 5158 std::string loader_path("@loader_path"); 5159 std::string executable_path("@executable_path"); 5160 for (auto &rpath : rpath_paths) { 5161 if (llvm::StringRef(rpath).startswith(loader_path)) { 5162 rpath.erase(0, loader_path.size()); 5163 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5164 } else if (llvm::StringRef(rpath).startswith(executable_path)) { 5165 rpath.erase(0, executable_path.size()); 5166 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5167 } 5168 } 5169 5170 for (const auto &rpath_relative_path : rpath_relative_paths) { 5171 for (const auto &rpath : rpath_paths) { 5172 std::string path = rpath; 5173 path += rpath_relative_path; 5174 // It is OK to resolve this path because we must find a file on disk 5175 // for us to accept it anyway if it is rpath relative. 5176 FileSpec file_spec(path); 5177 FileSystem::Instance().Resolve(file_spec); 5178 if (FileSystem::Instance().Exists(file_spec) && 5179 files.AppendIfUnique(file_spec)) { 5180 count++; 5181 break; 5182 } 5183 } 5184 } 5185 } 5186 5187 // We may have @executable_paths but no RPATHS. Figure those out here. 5188 // Only do this if this object file is the executable. We have no way to 5189 // get back to the actual executable otherwise, so we won't get the right 5190 // path. 5191 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5192 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5193 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5194 FileSpec file_spec = 5195 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5196 if (FileSystem::Instance().Exists(file_spec) && 5197 files.AppendIfUnique(file_spec)) 5198 count++; 5199 } 5200 } 5201 } 5202 return count; 5203 } 5204 5205 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5206 // If the object file is not an executable it can't hold the entry point. 5207 // m_entry_point_address is initialized to an invalid address, so we can just 5208 // return that. If m_entry_point_address is valid it means we've found it 5209 // already, so return the cached value. 5210 5211 if ((!IsExecutable() && !IsDynamicLoader()) || 5212 m_entry_point_address.IsValid()) { 5213 return m_entry_point_address; 5214 } 5215 5216 // Otherwise, look for the UnixThread or Thread command. The data for the 5217 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5218 // 5219 // uint32_t flavor - this is the flavor argument you would pass to 5220 // thread_get_state 5221 // uint32_t count - this is the count of longs in the thread state data 5222 // struct XXX_thread_state state - this is the structure from 5223 // <machine/thread_status.h> corresponding to the flavor. 5224 // <repeat this trio> 5225 // 5226 // So we just keep reading the various register flavors till we find the GPR 5227 // one, then read the PC out of there. 5228 // FIXME: We will need to have a "RegisterContext data provider" class at some 5229 // point that can get all the registers 5230 // out of data in this form & attach them to a given thread. That should 5231 // underlie the MacOS X User process plugin, and we'll also need it for the 5232 // MacOS X Core File process plugin. When we have that we can also use it 5233 // here. 5234 // 5235 // For now we hard-code the offsets and flavors we need: 5236 // 5237 // 5238 5239 ModuleSP module_sp(GetModule()); 5240 if (module_sp) { 5241 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5242 llvm::MachO::load_command load_cmd; 5243 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5244 uint32_t i; 5245 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5246 bool done = false; 5247 5248 for (i = 0; i < m_header.ncmds; ++i) { 5249 const lldb::offset_t cmd_offset = offset; 5250 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5251 break; 5252 5253 switch (load_cmd.cmd) { 5254 case LC_UNIXTHREAD: 5255 case LC_THREAD: { 5256 while (offset < cmd_offset + load_cmd.cmdsize) { 5257 uint32_t flavor = m_data.GetU32(&offset); 5258 uint32_t count = m_data.GetU32(&offset); 5259 if (count == 0) { 5260 // We've gotten off somehow, log and exit; 5261 return m_entry_point_address; 5262 } 5263 5264 switch (m_header.cputype) { 5265 case llvm::MachO::CPU_TYPE_ARM: 5266 if (flavor == 1 || 5267 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 5268 // from mach/arm/thread_status.h 5269 { 5270 offset += 60; // This is the offset of pc in the GPR thread state 5271 // data structure. 5272 start_address = m_data.GetU32(&offset); 5273 done = true; 5274 } 5275 break; 5276 case llvm::MachO::CPU_TYPE_ARM64: 5277 case llvm::MachO::CPU_TYPE_ARM64_32: 5278 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5279 { 5280 offset += 256; // This is the offset of pc in the GPR thread state 5281 // data structure. 5282 start_address = m_data.GetU64(&offset); 5283 done = true; 5284 } 5285 break; 5286 case llvm::MachO::CPU_TYPE_I386: 5287 if (flavor == 5288 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5289 { 5290 offset += 40; // This is the offset of eip in the GPR thread state 5291 // data structure. 5292 start_address = m_data.GetU32(&offset); 5293 done = true; 5294 } 5295 break; 5296 case llvm::MachO::CPU_TYPE_X86_64: 5297 if (flavor == 5298 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5299 { 5300 offset += 16 * 8; // This is the offset of rip in the GPR thread 5301 // state data structure. 5302 start_address = m_data.GetU64(&offset); 5303 done = true; 5304 } 5305 break; 5306 default: 5307 return m_entry_point_address; 5308 } 5309 // Haven't found the GPR flavor yet, skip over the data for this 5310 // flavor: 5311 if (done) 5312 break; 5313 offset += count * 4; 5314 } 5315 } break; 5316 case LC_MAIN: { 5317 ConstString text_segment_name("__TEXT"); 5318 uint64_t entryoffset = m_data.GetU64(&offset); 5319 SectionSP text_segment_sp = 5320 GetSectionList()->FindSectionByName(text_segment_name); 5321 if (text_segment_sp) { 5322 done = true; 5323 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5324 } 5325 } break; 5326 5327 default: 5328 break; 5329 } 5330 if (done) 5331 break; 5332 5333 // Go to the next load command: 5334 offset = cmd_offset + load_cmd.cmdsize; 5335 } 5336 5337 if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) { 5338 if (GetSymtab()) { 5339 Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType( 5340 ConstString("_dyld_start"), SymbolType::eSymbolTypeCode, 5341 Symtab::eDebugAny, Symtab::eVisibilityAny); 5342 if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) { 5343 start_address = dyld_start_sym->GetAddress().GetFileAddress(); 5344 } 5345 } 5346 } 5347 5348 if (start_address != LLDB_INVALID_ADDRESS) { 5349 // We got the start address from the load commands, so now resolve that 5350 // address in the sections of this ObjectFile: 5351 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5352 start_address, GetSectionList())) { 5353 m_entry_point_address.Clear(); 5354 } 5355 } else { 5356 // We couldn't read the UnixThread load command - maybe it wasn't there. 5357 // As a fallback look for the "start" symbol in the main executable. 5358 5359 ModuleSP module_sp(GetModule()); 5360 5361 if (module_sp) { 5362 SymbolContextList contexts; 5363 SymbolContext context; 5364 module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5365 eSymbolTypeCode, contexts); 5366 if (contexts.GetSize()) { 5367 if (contexts.GetContextAtIndex(0, context)) 5368 m_entry_point_address = context.symbol->GetAddress(); 5369 } 5370 } 5371 } 5372 } 5373 5374 return m_entry_point_address; 5375 } 5376 5377 lldb_private::Address ObjectFileMachO::GetBaseAddress() { 5378 lldb_private::Address header_addr; 5379 SectionList *section_list = GetSectionList(); 5380 if (section_list) { 5381 SectionSP text_segment_sp( 5382 section_list->FindSectionByName(GetSegmentNameTEXT())); 5383 if (text_segment_sp) { 5384 header_addr.SetSection(text_segment_sp); 5385 header_addr.SetOffset(0); 5386 } 5387 } 5388 return header_addr; 5389 } 5390 5391 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5392 ModuleSP module_sp(GetModule()); 5393 if (module_sp) { 5394 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5395 if (!m_thread_context_offsets_valid) { 5396 m_thread_context_offsets_valid = true; 5397 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5398 FileRangeArray::Entry file_range; 5399 llvm::MachO::thread_command thread_cmd; 5400 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5401 const uint32_t cmd_offset = offset; 5402 if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr) 5403 break; 5404 5405 if (thread_cmd.cmd == LC_THREAD) { 5406 file_range.SetRangeBase(offset); 5407 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5408 m_thread_context_offsets.Append(file_range); 5409 } 5410 offset = cmd_offset + thread_cmd.cmdsize; 5411 } 5412 } 5413 } 5414 return m_thread_context_offsets.GetSize(); 5415 } 5416 5417 std::string ObjectFileMachO::GetIdentifierString() { 5418 std::string result; 5419 ModuleSP module_sp(GetModule()); 5420 if (module_sp) { 5421 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5422 5423 // First, look over the load commands for an LC_NOTE load command with 5424 // data_owner string "kern ver str" & use that if found. 5425 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5426 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5427 const uint32_t cmd_offset = offset; 5428 llvm::MachO::load_command lc; 5429 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5430 break; 5431 if (lc.cmd == LC_NOTE) { 5432 char data_owner[17]; 5433 m_data.CopyData(offset, 16, data_owner); 5434 data_owner[16] = '\0'; 5435 offset += 16; 5436 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5437 uint64_t size = m_data.GetU64_unchecked(&offset); 5438 5439 // "kern ver str" has a uint32_t version and then a nul terminated 5440 // c-string. 5441 if (strcmp("kern ver str", data_owner) == 0) { 5442 offset = fileoff; 5443 uint32_t version; 5444 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5445 if (version == 1) { 5446 uint32_t strsize = size - sizeof(uint32_t); 5447 char *buf = (char *)malloc(strsize); 5448 if (buf) { 5449 m_data.CopyData(offset, strsize, buf); 5450 buf[strsize - 1] = '\0'; 5451 result = buf; 5452 if (buf) 5453 free(buf); 5454 return result; 5455 } 5456 } 5457 } 5458 } 5459 } 5460 offset = cmd_offset + lc.cmdsize; 5461 } 5462 5463 // Second, make a pass over the load commands looking for an obsolete 5464 // LC_IDENT load command. 5465 offset = MachHeaderSizeFromMagic(m_header.magic); 5466 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5467 const uint32_t cmd_offset = offset; 5468 llvm::MachO::ident_command ident_command; 5469 if (m_data.GetU32(&offset, &ident_command, 2) == nullptr) 5470 break; 5471 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5472 char *buf = (char *)malloc(ident_command.cmdsize); 5473 if (buf != nullptr && m_data.CopyData(offset, ident_command.cmdsize, 5474 buf) == ident_command.cmdsize) { 5475 buf[ident_command.cmdsize - 1] = '\0'; 5476 result = buf; 5477 } 5478 if (buf) 5479 free(buf); 5480 } 5481 offset = cmd_offset + ident_command.cmdsize; 5482 } 5483 } 5484 return result; 5485 } 5486 5487 addr_t ObjectFileMachO::GetAddressMask() { 5488 addr_t mask = 0; 5489 ModuleSP module_sp(GetModule()); 5490 if (module_sp) { 5491 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5492 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5493 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5494 const uint32_t cmd_offset = offset; 5495 llvm::MachO::load_command lc; 5496 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5497 break; 5498 if (lc.cmd == LC_NOTE) { 5499 char data_owner[17]; 5500 m_data.CopyData(offset, 16, data_owner); 5501 data_owner[16] = '\0'; 5502 offset += 16; 5503 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5504 5505 // "addrable bits" has a uint32_t version and a uint32_t 5506 // number of bits used in addressing. 5507 if (strcmp("addrable bits", data_owner) == 0) { 5508 offset = fileoff; 5509 uint32_t version; 5510 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5511 if (version == 3) { 5512 uint32_t num_addr_bits = m_data.GetU32_unchecked(&offset); 5513 if (num_addr_bits != 0) { 5514 mask = ~((1ULL << num_addr_bits) - 1); 5515 } 5516 break; 5517 } 5518 } 5519 } 5520 } 5521 offset = cmd_offset + lc.cmdsize; 5522 } 5523 } 5524 return mask; 5525 } 5526 5527 bool ObjectFileMachO::GetCorefileMainBinaryInfo(addr_t &value, 5528 bool &value_is_offset, 5529 UUID &uuid, 5530 ObjectFile::BinaryType &type) { 5531 value = LLDB_INVALID_ADDRESS; 5532 value_is_offset = false; 5533 uuid.Clear(); 5534 uint32_t log2_pagesize = 0; // not currently passed up to caller 5535 uint32_t platform = 0; // not currently passed up to caller 5536 ModuleSP module_sp(GetModule()); 5537 if (module_sp) { 5538 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5539 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5540 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5541 const uint32_t cmd_offset = offset; 5542 llvm::MachO::load_command lc; 5543 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5544 break; 5545 if (lc.cmd == LC_NOTE) { 5546 char data_owner[17]; 5547 memset(data_owner, 0, sizeof(data_owner)); 5548 m_data.CopyData(offset, 16, data_owner); 5549 offset += 16; 5550 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5551 uint64_t size = m_data.GetU64_unchecked(&offset); 5552 5553 // struct main_bin_spec 5554 // { 5555 // uint32_t version; // currently 2 5556 // uint32_t type; // 0 == unspecified, 1 == kernel, 5557 // // 2 == user process, 5558 // // 3 == standalone binary 5559 // uint64_t address; // UINT64_MAX if address not specified 5560 // uint64_t slide; // slide, UINT64_MAX if unspecified 5561 // // 0 if no slide needs to be applied to 5562 // // file address 5563 // uuid_t uuid; // all zero's if uuid not specified 5564 // uint32_t log2_pagesize; // process page size in log base 2, 5565 // // e.g. 4k pages are 12. 5566 // // 0 for unspecified 5567 // uint32_t platform; // The Mach-O platform for this corefile. 5568 // // 0 for unspecified. 5569 // // The values are defined in 5570 // // <mach-o/loader.h>, PLATFORM_*. 5571 // } __attribute((packed)); 5572 5573 // "main bin spec" (main binary specification) data payload is 5574 // formatted: 5575 // uint32_t version [currently 1] 5576 // uint32_t type [0 == unspecified, 1 == kernel, 5577 // 2 == user process, 3 == firmware ] 5578 // uint64_t address [ UINT64_MAX if address not specified ] 5579 // uuid_t uuid [ all zero's if uuid not specified ] 5580 // uint32_t log2_pagesize [ process page size in log base 5581 // 2, e.g. 4k pages are 12. 5582 // 0 for unspecified ] 5583 // uint32_t unused [ for alignment ] 5584 5585 if (strcmp("main bin spec", data_owner) == 0 && size >= 32) { 5586 offset = fileoff; 5587 uint32_t version; 5588 if (m_data.GetU32(&offset, &version, 1) != nullptr && version <= 2) { 5589 uint32_t binspec_type = 0; 5590 uuid_t raw_uuid; 5591 memset(raw_uuid, 0, sizeof(uuid_t)); 5592 5593 if (!m_data.GetU32(&offset, &binspec_type, 1)) 5594 return false; 5595 if (!m_data.GetU64(&offset, &value, 1)) 5596 return false; 5597 uint64_t slide = LLDB_INVALID_ADDRESS; 5598 if (version > 1 && !m_data.GetU64(&offset, &slide, 1)) 5599 return false; 5600 if (value == LLDB_INVALID_ADDRESS && 5601 slide != LLDB_INVALID_ADDRESS) { 5602 value = slide; 5603 value_is_offset = true; 5604 } 5605 5606 if (m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5607 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5608 // convert the "main bin spec" type into our 5609 // ObjectFile::BinaryType enum 5610 switch (binspec_type) { 5611 case 0: 5612 type = eBinaryTypeUnknown; 5613 break; 5614 case 1: 5615 type = eBinaryTypeKernel; 5616 break; 5617 case 2: 5618 type = eBinaryTypeUser; 5619 break; 5620 case 3: 5621 type = eBinaryTypeStandalone; 5622 break; 5623 } 5624 if (!m_data.GetU32(&offset, &log2_pagesize, 1)) 5625 return false; 5626 if (version > 1 && !m_data.GetU32(&offset, &platform, 1)) 5627 return false; 5628 return true; 5629 } 5630 } 5631 } 5632 } 5633 offset = cmd_offset + lc.cmdsize; 5634 } 5635 } 5636 return false; 5637 } 5638 5639 lldb::RegisterContextSP 5640 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5641 lldb_private::Thread &thread) { 5642 lldb::RegisterContextSP reg_ctx_sp; 5643 5644 ModuleSP module_sp(GetModule()); 5645 if (module_sp) { 5646 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5647 if (!m_thread_context_offsets_valid) 5648 GetNumThreadContexts(); 5649 5650 const FileRangeArray::Entry *thread_context_file_range = 5651 m_thread_context_offsets.GetEntryAtIndex(idx); 5652 if (thread_context_file_range) { 5653 5654 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5655 thread_context_file_range->GetByteSize()); 5656 5657 switch (m_header.cputype) { 5658 case llvm::MachO::CPU_TYPE_ARM64: 5659 case llvm::MachO::CPU_TYPE_ARM64_32: 5660 reg_ctx_sp = 5661 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data); 5662 break; 5663 5664 case llvm::MachO::CPU_TYPE_ARM: 5665 reg_ctx_sp = 5666 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data); 5667 break; 5668 5669 case llvm::MachO::CPU_TYPE_I386: 5670 reg_ctx_sp = 5671 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data); 5672 break; 5673 5674 case llvm::MachO::CPU_TYPE_X86_64: 5675 reg_ctx_sp = 5676 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data); 5677 break; 5678 } 5679 } 5680 } 5681 return reg_ctx_sp; 5682 } 5683 5684 ObjectFile::Type ObjectFileMachO::CalculateType() { 5685 switch (m_header.filetype) { 5686 case MH_OBJECT: // 0x1u 5687 if (GetAddressByteSize() == 4) { 5688 // 32 bit kexts are just object files, but they do have a valid 5689 // UUID load command. 5690 if (GetUUID()) { 5691 // this checking for the UUID load command is not enough we could 5692 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5693 // this is required of kexts 5694 if (m_strata == eStrataInvalid) 5695 m_strata = eStrataKernel; 5696 return eTypeSharedLibrary; 5697 } 5698 } 5699 return eTypeObjectFile; 5700 5701 case MH_EXECUTE: 5702 return eTypeExecutable; // 0x2u 5703 case MH_FVMLIB: 5704 return eTypeSharedLibrary; // 0x3u 5705 case MH_CORE: 5706 return eTypeCoreFile; // 0x4u 5707 case MH_PRELOAD: 5708 return eTypeSharedLibrary; // 0x5u 5709 case MH_DYLIB: 5710 return eTypeSharedLibrary; // 0x6u 5711 case MH_DYLINKER: 5712 return eTypeDynamicLinker; // 0x7u 5713 case MH_BUNDLE: 5714 return eTypeSharedLibrary; // 0x8u 5715 case MH_DYLIB_STUB: 5716 return eTypeStubLibrary; // 0x9u 5717 case MH_DSYM: 5718 return eTypeDebugInfo; // 0xAu 5719 case MH_KEXT_BUNDLE: 5720 return eTypeSharedLibrary; // 0xBu 5721 default: 5722 break; 5723 } 5724 return eTypeUnknown; 5725 } 5726 5727 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5728 switch (m_header.filetype) { 5729 case MH_OBJECT: // 0x1u 5730 { 5731 // 32 bit kexts are just object files, but they do have a valid 5732 // UUID load command. 5733 if (GetUUID()) { 5734 // this checking for the UUID load command is not enough we could 5735 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5736 // this is required of kexts 5737 if (m_type == eTypeInvalid) 5738 m_type = eTypeSharedLibrary; 5739 5740 return eStrataKernel; 5741 } 5742 } 5743 return eStrataUnknown; 5744 5745 case MH_EXECUTE: // 0x2u 5746 // Check for the MH_DYLDLINK bit in the flags 5747 if (m_header.flags & MH_DYLDLINK) { 5748 return eStrataUser; 5749 } else { 5750 SectionList *section_list = GetSectionList(); 5751 if (section_list) { 5752 static ConstString g_kld_section_name("__KLD"); 5753 if (section_list->FindSectionByName(g_kld_section_name)) 5754 return eStrataKernel; 5755 } 5756 } 5757 return eStrataRawImage; 5758 5759 case MH_FVMLIB: 5760 return eStrataUser; // 0x3u 5761 case MH_CORE: 5762 return eStrataUnknown; // 0x4u 5763 case MH_PRELOAD: 5764 return eStrataRawImage; // 0x5u 5765 case MH_DYLIB: 5766 return eStrataUser; // 0x6u 5767 case MH_DYLINKER: 5768 return eStrataUser; // 0x7u 5769 case MH_BUNDLE: 5770 return eStrataUser; // 0x8u 5771 case MH_DYLIB_STUB: 5772 return eStrataUser; // 0x9u 5773 case MH_DSYM: 5774 return eStrataUnknown; // 0xAu 5775 case MH_KEXT_BUNDLE: 5776 return eStrataKernel; // 0xBu 5777 default: 5778 break; 5779 } 5780 return eStrataUnknown; 5781 } 5782 5783 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5784 ModuleSP module_sp(GetModule()); 5785 if (module_sp) { 5786 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5787 llvm::MachO::dylib_command load_cmd; 5788 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5789 uint32_t version_cmd = 0; 5790 uint64_t version = 0; 5791 uint32_t i; 5792 for (i = 0; i < m_header.ncmds; ++i) { 5793 const lldb::offset_t cmd_offset = offset; 5794 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5795 break; 5796 5797 if (load_cmd.cmd == LC_ID_DYLIB) { 5798 if (version_cmd == 0) { 5799 version_cmd = load_cmd.cmd; 5800 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr) 5801 break; 5802 version = load_cmd.dylib.current_version; 5803 } 5804 break; // Break for now unless there is another more complete version 5805 // number load command in the future. 5806 } 5807 offset = cmd_offset + load_cmd.cmdsize; 5808 } 5809 5810 if (version_cmd == LC_ID_DYLIB) { 5811 unsigned major = (version & 0xFFFF0000ull) >> 16; 5812 unsigned minor = (version & 0x0000FF00ull) >> 8; 5813 unsigned subminor = (version & 0x000000FFull); 5814 return llvm::VersionTuple(major, minor, subminor); 5815 } 5816 } 5817 return llvm::VersionTuple(); 5818 } 5819 5820 ArchSpec ObjectFileMachO::GetArchitecture() { 5821 ModuleSP module_sp(GetModule()); 5822 ArchSpec arch; 5823 if (module_sp) { 5824 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5825 5826 return GetArchitecture(module_sp, m_header, m_data, 5827 MachHeaderSizeFromMagic(m_header.magic)); 5828 } 5829 return arch; 5830 } 5831 5832 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, 5833 addr_t &base_addr, UUID &uuid) { 5834 uuid.Clear(); 5835 base_addr = LLDB_INVALID_ADDRESS; 5836 if (process && process->GetDynamicLoader()) { 5837 DynamicLoader *dl = process->GetDynamicLoader(); 5838 LazyBool using_shared_cache; 5839 LazyBool private_shared_cache; 5840 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5841 private_shared_cache); 5842 } 5843 Log *log(GetLog(LLDBLog::Symbols | LLDBLog::Process)); 5844 LLDB_LOGF( 5845 log, 5846 "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64, 5847 uuid.GetAsString().c_str(), base_addr); 5848 } 5849 5850 // From dyld SPI header dyld_process_info.h 5851 typedef void *dyld_process_info; 5852 struct lldb_copy__dyld_process_cache_info { 5853 uuid_t cacheUUID; // UUID of cache used by process 5854 uint64_t cacheBaseAddress; // load address of dyld shared cache 5855 bool noCache; // process is running without a dyld cache 5856 bool privateCache; // process is using a private copy of its dyld cache 5857 }; 5858 5859 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with 5860 // llvm enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile 5861 // errors. So we need to use the actual underlying types of task_t and 5862 // kern_return_t below. 5863 extern "C" unsigned int /*task_t*/ mach_task_self(); 5864 5865 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5866 uuid.Clear(); 5867 base_addr = LLDB_INVALID_ADDRESS; 5868 5869 #if defined(__APPLE__) 5870 uint8_t *(*dyld_get_all_image_infos)(void); 5871 dyld_get_all_image_infos = 5872 (uint8_t * (*)()) dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5873 if (dyld_get_all_image_infos) { 5874 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5875 if (dyld_all_image_infos_address) { 5876 uint32_t *version = (uint32_t *) 5877 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5878 if (*version >= 13) { 5879 uuid_t *sharedCacheUUID_address = 0; 5880 int wordsize = sizeof(uint8_t *); 5881 if (wordsize == 8) { 5882 sharedCacheUUID_address = 5883 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5884 160); // sharedCacheUUID <mach-o/dyld_images.h> 5885 if (*version >= 15) 5886 base_addr = 5887 *(uint64_t 5888 *)((uint8_t *)dyld_all_image_infos_address + 5889 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5890 } else { 5891 sharedCacheUUID_address = 5892 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5893 84); // sharedCacheUUID <mach-o/dyld_images.h> 5894 if (*version >= 15) { 5895 base_addr = 0; 5896 base_addr = 5897 *(uint32_t 5898 *)((uint8_t *)dyld_all_image_infos_address + 5899 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5900 } 5901 } 5902 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5903 } 5904 } 5905 } else { 5906 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5907 dyld_process_info (*dyld_process_info_create)( 5908 unsigned int /* task_t */ task, uint64_t timestamp, 5909 unsigned int /*kern_return_t*/ *kernelError); 5910 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5911 void (*dyld_process_info_release)(dyld_process_info info); 5912 5913 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, 5914 unsigned int /*kern_return_t*/ *)) 5915 dlsym(RTLD_DEFAULT, "_dyld_process_info_create"); 5916 dyld_process_info_get_cache = (void (*)(void *, void *))dlsym( 5917 RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5918 dyld_process_info_release = 5919 (void (*)(void *))dlsym(RTLD_DEFAULT, "_dyld_process_info_release"); 5920 5921 if (dyld_process_info_create && dyld_process_info_get_cache) { 5922 unsigned int /*kern_return_t */ kern_ret; 5923 dyld_process_info process_info = 5924 dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5925 if (process_info) { 5926 struct lldb_copy__dyld_process_cache_info sc_info; 5927 memset(&sc_info, 0, sizeof(struct lldb_copy__dyld_process_cache_info)); 5928 dyld_process_info_get_cache(process_info, &sc_info); 5929 if (sc_info.cacheBaseAddress != 0) { 5930 base_addr = sc_info.cacheBaseAddress; 5931 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 5932 } 5933 dyld_process_info_release(process_info); 5934 } 5935 } 5936 } 5937 Log *log(GetLog(LLDBLog::Symbols | LLDBLog::Process)); 5938 if (log && uuid.IsValid()) 5939 LLDB_LOGF(log, 5940 "lldb's in-memory shared cache has a UUID of %s base address of " 5941 "0x%" PRIx64, 5942 uuid.GetAsString().c_str(), base_addr); 5943 #endif 5944 } 5945 5946 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5947 if (!m_min_os_version) { 5948 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5949 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5950 const lldb::offset_t load_cmd_offset = offset; 5951 5952 llvm::MachO::version_min_command lc; 5953 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5954 break; 5955 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5956 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5957 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5958 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5959 if (m_data.GetU32(&offset, &lc.version, 5960 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5961 const uint32_t xxxx = lc.version >> 16; 5962 const uint32_t yy = (lc.version >> 8) & 0xffu; 5963 const uint32_t zz = lc.version & 0xffu; 5964 if (xxxx) { 5965 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5966 break; 5967 } 5968 } 5969 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5970 // struct build_version_command { 5971 // uint32_t cmd; /* LC_BUILD_VERSION */ 5972 // uint32_t cmdsize; /* sizeof(struct 5973 // build_version_command) plus */ 5974 // /* ntools * sizeof(struct 5975 // build_tool_version) */ 5976 // uint32_t platform; /* platform */ 5977 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5978 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded in 5979 // nibbles xxxx.yy.zz */ uint32_t ntools; /* number of 5980 // tool entries following this */ 5981 // }; 5982 5983 offset += 4; // skip platform 5984 uint32_t minos = m_data.GetU32(&offset); 5985 5986 const uint32_t xxxx = minos >> 16; 5987 const uint32_t yy = (minos >> 8) & 0xffu; 5988 const uint32_t zz = minos & 0xffu; 5989 if (xxxx) { 5990 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5991 break; 5992 } 5993 } 5994 5995 offset = load_cmd_offset + lc.cmdsize; 5996 } 5997 5998 if (!m_min_os_version) { 5999 // Set version to an empty value so we don't keep trying to 6000 m_min_os_version = llvm::VersionTuple(); 6001 } 6002 } 6003 6004 return *m_min_os_version; 6005 } 6006 6007 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { 6008 if (!m_sdk_versions.hasValue()) { 6009 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 6010 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6011 const lldb::offset_t load_cmd_offset = offset; 6012 6013 llvm::MachO::version_min_command lc; 6014 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6015 break; 6016 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 6017 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 6018 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 6019 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 6020 if (m_data.GetU32(&offset, &lc.version, 6021 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 6022 const uint32_t xxxx = lc.sdk >> 16; 6023 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 6024 const uint32_t zz = lc.sdk & 0xffu; 6025 if (xxxx) { 6026 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 6027 break; 6028 } else { 6029 GetModule()->ReportWarning("minimum OS version load command with " 6030 "invalid (0) version found."); 6031 } 6032 } 6033 } 6034 offset = load_cmd_offset + lc.cmdsize; 6035 } 6036 6037 if (!m_sdk_versions.hasValue()) { 6038 offset = MachHeaderSizeFromMagic(m_header.magic); 6039 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6040 const lldb::offset_t load_cmd_offset = offset; 6041 6042 llvm::MachO::version_min_command lc; 6043 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6044 break; 6045 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 6046 // struct build_version_command { 6047 // uint32_t cmd; /* LC_BUILD_VERSION */ 6048 // uint32_t cmdsize; /* sizeof(struct 6049 // build_version_command) plus */ 6050 // /* ntools * sizeof(struct 6051 // build_tool_version) */ 6052 // uint32_t platform; /* platform */ 6053 // uint32_t minos; /* X.Y.Z is encoded in nibbles 6054 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded 6055 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number 6056 // of tool entries following this */ 6057 // }; 6058 6059 offset += 4; // skip platform 6060 uint32_t minos = m_data.GetU32(&offset); 6061 6062 const uint32_t xxxx = minos >> 16; 6063 const uint32_t yy = (minos >> 8) & 0xffu; 6064 const uint32_t zz = minos & 0xffu; 6065 if (xxxx) { 6066 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 6067 break; 6068 } 6069 } 6070 offset = load_cmd_offset + lc.cmdsize; 6071 } 6072 } 6073 6074 if (!m_sdk_versions.hasValue()) 6075 m_sdk_versions = llvm::VersionTuple(); 6076 } 6077 6078 return m_sdk_versions.getValue(); 6079 } 6080 6081 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 6082 return m_header.filetype == llvm::MachO::MH_DYLINKER; 6083 } 6084 6085 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 6086 return m_allow_assembly_emulation_unwind_plans; 6087 } 6088 6089 Section *ObjectFileMachO::GetMachHeaderSection() { 6090 // Find the first address of the mach header which is the first non-zero file 6091 // sized section whose file offset is zero. This is the base file address of 6092 // the mach-o file which can be subtracted from the vmaddr of the other 6093 // segments found in memory and added to the load address 6094 ModuleSP module_sp = GetModule(); 6095 if (!module_sp) 6096 return nullptr; 6097 SectionList *section_list = GetSectionList(); 6098 if (!section_list) 6099 return nullptr; 6100 const size_t num_sections = section_list->GetSize(); 6101 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6102 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 6103 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 6104 return section; 6105 } 6106 6107 // We may have a binary in the shared cache that has a non-zero 6108 // file address for its first segment, traditionally the __TEXT segment. 6109 // Search for it by name and return it as our next best guess. 6110 SectionSP text_segment_sp = 6111 GetSectionList()->FindSectionByName(GetSegmentNameTEXT()); 6112 if (text_segment_sp.get() && SectionIsLoadable(text_segment_sp.get())) 6113 return text_segment_sp.get(); 6114 6115 return nullptr; 6116 } 6117 6118 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 6119 if (!section) 6120 return false; 6121 const bool is_dsym = (m_header.filetype == MH_DSYM); 6122 if (section->GetFileSize() == 0 && !is_dsym) 6123 return false; 6124 if (section->IsThreadSpecific()) 6125 return false; 6126 if (GetModule().get() != section->GetModule().get()) 6127 return false; 6128 // Be careful with __LINKEDIT and __DWARF segments 6129 if (section->GetName() == GetSegmentNameLINKEDIT() || 6130 section->GetName() == GetSegmentNameDWARF()) { 6131 // Only map __LINKEDIT and __DWARF if we have an in memory image and 6132 // this isn't a kernel binary like a kext or mach_kernel. 6133 const bool is_memory_image = (bool)m_process_wp.lock(); 6134 const Strata strata = GetStrata(); 6135 if (is_memory_image == false || strata == eStrataKernel) 6136 return false; 6137 } 6138 return true; 6139 } 6140 6141 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 6142 lldb::addr_t header_load_address, const Section *header_section, 6143 const Section *section) { 6144 ModuleSP module_sp = GetModule(); 6145 if (module_sp && header_section && section && 6146 header_load_address != LLDB_INVALID_ADDRESS) { 6147 lldb::addr_t file_addr = header_section->GetFileAddress(); 6148 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 6149 return section->GetFileAddress() - file_addr + header_load_address; 6150 } 6151 return LLDB_INVALID_ADDRESS; 6152 } 6153 6154 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 6155 bool value_is_offset) { 6156 ModuleSP module_sp = GetModule(); 6157 if (!module_sp) 6158 return false; 6159 6160 SectionList *section_list = GetSectionList(); 6161 if (!section_list) 6162 return false; 6163 6164 size_t num_loaded_sections = 0; 6165 const size_t num_sections = section_list->GetSize(); 6166 6167 if (value_is_offset) { 6168 // "value" is an offset to apply to each top level segment 6169 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6170 // Iterate through the object file sections to find all of the 6171 // sections that size on disk (to avoid __PAGEZERO) and load them 6172 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6173 if (SectionIsLoadable(section_sp.get())) 6174 if (target.GetSectionLoadList().SetSectionLoadAddress( 6175 section_sp, section_sp->GetFileAddress() + value)) 6176 ++num_loaded_sections; 6177 } 6178 } else { 6179 // "value" is the new base address of the mach_header, adjust each 6180 // section accordingly 6181 6182 Section *mach_header_section = GetMachHeaderSection(); 6183 if (mach_header_section) { 6184 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6185 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6186 6187 lldb::addr_t section_load_addr = 6188 CalculateSectionLoadAddressForMemoryImage( 6189 value, mach_header_section, section_sp.get()); 6190 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6191 if (target.GetSectionLoadList().SetSectionLoadAddress( 6192 section_sp, section_load_addr)) 6193 ++num_loaded_sections; 6194 } 6195 } 6196 } 6197 } 6198 return num_loaded_sections > 0; 6199 } 6200 6201 struct all_image_infos_header { 6202 uint32_t version; // currently 1 6203 uint32_t imgcount; // number of binary images 6204 uint64_t entries_fileoff; // file offset in the corefile of where the array of 6205 // struct entry's begin. 6206 uint32_t entries_size; // size of 'struct entry'. 6207 uint32_t unused; 6208 }; 6209 6210 struct image_entry { 6211 uint64_t filepath_offset; // offset in corefile to c-string of the file path, 6212 // UINT64_MAX if unavailable. 6213 uuid_t uuid; // uint8_t[16]. should be set to all zeroes if 6214 // uuid is unknown. 6215 uint64_t load_address; // UINT64_MAX if unknown. 6216 uint64_t seg_addrs_offset; // offset to the array of struct segment_vmaddr's. 6217 uint32_t segment_count; // The number of segments for this binary. 6218 uint32_t unused; 6219 6220 image_entry() { 6221 filepath_offset = UINT64_MAX; 6222 memset(&uuid, 0, sizeof(uuid_t)); 6223 segment_count = 0; 6224 load_address = UINT64_MAX; 6225 seg_addrs_offset = UINT64_MAX; 6226 unused = 0; 6227 } 6228 image_entry(const image_entry &rhs) { 6229 filepath_offset = rhs.filepath_offset; 6230 memcpy(&uuid, &rhs.uuid, sizeof(uuid_t)); 6231 segment_count = rhs.segment_count; 6232 seg_addrs_offset = rhs.seg_addrs_offset; 6233 load_address = rhs.load_address; 6234 unused = rhs.unused; 6235 } 6236 }; 6237 6238 struct segment_vmaddr { 6239 char segname[16]; 6240 uint64_t vmaddr; 6241 uint64_t unused; 6242 6243 segment_vmaddr() { 6244 memset(&segname, 0, 16); 6245 vmaddr = UINT64_MAX; 6246 unused = 0; 6247 } 6248 segment_vmaddr(const segment_vmaddr &rhs) { 6249 memcpy(&segname, &rhs.segname, 16); 6250 vmaddr = rhs.vmaddr; 6251 unused = rhs.unused; 6252 } 6253 }; 6254 6255 // Write the payload for the "all image infos" LC_NOTE into 6256 // the supplied all_image_infos_payload, assuming that this 6257 // will be written into the corefile starting at 6258 // initial_file_offset. 6259 // 6260 // The placement of this payload is a little tricky. We're 6261 // laying this out as 6262 // 6263 // 1. header (struct all_image_info_header) 6264 // 2. Array of fixed-size (struct image_entry)'s, one 6265 // per binary image present in the process. 6266 // 3. Arrays of (struct segment_vmaddr)'s, a varying number 6267 // for each binary image. 6268 // 4. Variable length c-strings of binary image filepaths, 6269 // one per binary. 6270 // 6271 // To compute where everything will be laid out in the 6272 // payload, we need to iterate over the images and calculate 6273 // how many segment_vmaddr structures each image will need, 6274 // and how long each image's filepath c-string is. There 6275 // are some multiple passes over the image list while calculating 6276 // everything. 6277 6278 static offset_t CreateAllImageInfosPayload( 6279 const lldb::ProcessSP &process_sp, offset_t initial_file_offset, 6280 StreamString &all_image_infos_payload, SaveCoreStyle core_style) { 6281 Target &target = process_sp->GetTarget(); 6282 ModuleList modules = target.GetImages(); 6283 6284 // stack-only corefiles have no reason to include binaries that 6285 // are not executing; we're trying to make the smallest corefile 6286 // we can, so leave the rest out. 6287 if (core_style == SaveCoreStyle::eSaveCoreStackOnly) 6288 modules.Clear(); 6289 6290 std::set<std::string> executing_uuids; 6291 ThreadList &thread_list(process_sp->GetThreadList()); 6292 for (uint32_t i = 0; i < thread_list.GetSize(); i++) { 6293 ThreadSP thread_sp = thread_list.GetThreadAtIndex(i); 6294 uint32_t stack_frame_count = thread_sp->GetStackFrameCount(); 6295 for (uint32_t j = 0; j < stack_frame_count; j++) { 6296 StackFrameSP stack_frame_sp = thread_sp->GetStackFrameAtIndex(j); 6297 Address pc = stack_frame_sp->GetFrameCodeAddress(); 6298 ModuleSP module_sp = pc.GetModule(); 6299 if (module_sp) { 6300 UUID uuid = module_sp->GetUUID(); 6301 if (uuid.IsValid()) { 6302 executing_uuids.insert(uuid.GetAsString()); 6303 modules.AppendIfNeeded(module_sp); 6304 } 6305 } 6306 } 6307 } 6308 size_t modules_count = modules.GetSize(); 6309 6310 struct all_image_infos_header infos; 6311 infos.version = 1; 6312 infos.imgcount = modules_count; 6313 infos.entries_size = sizeof(image_entry); 6314 infos.entries_fileoff = initial_file_offset + sizeof(all_image_infos_header); 6315 infos.unused = 0; 6316 6317 all_image_infos_payload.PutHex32(infos.version); 6318 all_image_infos_payload.PutHex32(infos.imgcount); 6319 all_image_infos_payload.PutHex64(infos.entries_fileoff); 6320 all_image_infos_payload.PutHex32(infos.entries_size); 6321 all_image_infos_payload.PutHex32(infos.unused); 6322 6323 // First create the structures for all of the segment name+vmaddr vectors 6324 // for each module, so we will know the size of them as we add the 6325 // module entries. 6326 std::vector<std::vector<segment_vmaddr>> modules_segment_vmaddrs; 6327 for (size_t i = 0; i < modules_count; i++) { 6328 ModuleSP module = modules.GetModuleAtIndex(i); 6329 6330 SectionList *sections = module->GetSectionList(); 6331 size_t sections_count = sections->GetSize(); 6332 std::vector<segment_vmaddr> segment_vmaddrs; 6333 for (size_t j = 0; j < sections_count; j++) { 6334 SectionSP section = sections->GetSectionAtIndex(j); 6335 if (!section->GetParent().get()) { 6336 addr_t vmaddr = section->GetLoadBaseAddress(&target); 6337 if (vmaddr == LLDB_INVALID_ADDRESS) 6338 continue; 6339 ConstString name = section->GetName(); 6340 segment_vmaddr seg_vmaddr; 6341 strncpy(seg_vmaddr.segname, name.AsCString(), 6342 sizeof(seg_vmaddr.segname)); 6343 seg_vmaddr.vmaddr = vmaddr; 6344 seg_vmaddr.unused = 0; 6345 segment_vmaddrs.push_back(seg_vmaddr); 6346 } 6347 } 6348 modules_segment_vmaddrs.push_back(segment_vmaddrs); 6349 } 6350 6351 offset_t size_of_vmaddr_structs = 0; 6352 for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) { 6353 size_of_vmaddr_structs += 6354 modules_segment_vmaddrs[i].size() * sizeof(segment_vmaddr); 6355 } 6356 6357 offset_t size_of_filepath_cstrings = 0; 6358 for (size_t i = 0; i < modules_count; i++) { 6359 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6360 size_of_filepath_cstrings += module_sp->GetFileSpec().GetPath().size() + 1; 6361 } 6362 6363 // Calculate the file offsets of our "all image infos" payload in the 6364 // corefile. initial_file_offset the original value passed in to this method. 6365 6366 offset_t start_of_entries = 6367 initial_file_offset + sizeof(all_image_infos_header); 6368 offset_t start_of_seg_vmaddrs = 6369 start_of_entries + sizeof(image_entry) * modules_count; 6370 offset_t start_of_filenames = start_of_seg_vmaddrs + size_of_vmaddr_structs; 6371 6372 offset_t final_file_offset = start_of_filenames + size_of_filepath_cstrings; 6373 6374 // Now write the one-per-module 'struct image_entry' into the 6375 // StringStream; keep track of where the struct segment_vmaddr 6376 // entries for each module will end up in the corefile. 6377 6378 offset_t current_string_offset = start_of_filenames; 6379 offset_t current_segaddrs_offset = start_of_seg_vmaddrs; 6380 std::vector<struct image_entry> image_entries; 6381 for (size_t i = 0; i < modules_count; i++) { 6382 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6383 6384 struct image_entry ent; 6385 memcpy(&ent.uuid, module_sp->GetUUID().GetBytes().data(), sizeof(ent.uuid)); 6386 if (modules_segment_vmaddrs[i].size() > 0) { 6387 ent.segment_count = modules_segment_vmaddrs[i].size(); 6388 ent.seg_addrs_offset = current_segaddrs_offset; 6389 } 6390 ent.filepath_offset = current_string_offset; 6391 ObjectFile *objfile = module_sp->GetObjectFile(); 6392 if (objfile) { 6393 Address base_addr(objfile->GetBaseAddress()); 6394 if (base_addr.IsValid()) { 6395 ent.load_address = base_addr.GetLoadAddress(&target); 6396 } 6397 } 6398 6399 all_image_infos_payload.PutHex64(ent.filepath_offset); 6400 all_image_infos_payload.PutRawBytes(ent.uuid, sizeof(ent.uuid)); 6401 all_image_infos_payload.PutHex64(ent.load_address); 6402 all_image_infos_payload.PutHex64(ent.seg_addrs_offset); 6403 all_image_infos_payload.PutHex32(ent.segment_count); 6404 6405 if (executing_uuids.find(module_sp->GetUUID().GetAsString()) != 6406 executing_uuids.end()) 6407 all_image_infos_payload.PutHex32(1); 6408 else 6409 all_image_infos_payload.PutHex32(0); 6410 6411 current_segaddrs_offset += ent.segment_count * sizeof(segment_vmaddr); 6412 current_string_offset += module_sp->GetFileSpec().GetPath().size() + 1; 6413 } 6414 6415 // Now write the struct segment_vmaddr entries into the StringStream. 6416 6417 for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) { 6418 if (modules_segment_vmaddrs[i].size() == 0) 6419 continue; 6420 for (struct segment_vmaddr segvm : modules_segment_vmaddrs[i]) { 6421 all_image_infos_payload.PutRawBytes(segvm.segname, sizeof(segvm.segname)); 6422 all_image_infos_payload.PutHex64(segvm.vmaddr); 6423 all_image_infos_payload.PutHex64(segvm.unused); 6424 } 6425 } 6426 6427 for (size_t i = 0; i < modules_count; i++) { 6428 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6429 std::string filepath = module_sp->GetFileSpec().GetPath(); 6430 all_image_infos_payload.PutRawBytes(filepath.data(), filepath.size() + 1); 6431 } 6432 6433 return final_file_offset; 6434 } 6435 6436 // Temp struct used to combine contiguous memory regions with 6437 // identical permissions. 6438 struct page_object { 6439 addr_t addr; 6440 addr_t size; 6441 uint32_t prot; 6442 }; 6443 6444 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6445 const FileSpec &outfile, 6446 lldb::SaveCoreStyle &core_style, Status &error) { 6447 if (!process_sp) 6448 return false; 6449 6450 // Default on macOS is to create a dirty-memory-only corefile. 6451 if (core_style == SaveCoreStyle::eSaveCoreUnspecified) { 6452 core_style = SaveCoreStyle::eSaveCoreDirtyOnly; 6453 } 6454 6455 Target &target = process_sp->GetTarget(); 6456 const ArchSpec target_arch = target.GetArchitecture(); 6457 const llvm::Triple &target_triple = target_arch.GetTriple(); 6458 if (target_triple.getVendor() == llvm::Triple::Apple && 6459 (target_triple.getOS() == llvm::Triple::MacOSX || 6460 target_triple.getOS() == llvm::Triple::IOS || 6461 target_triple.getOS() == llvm::Triple::WatchOS || 6462 target_triple.getOS() == llvm::Triple::TvOS)) { 6463 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) 6464 // { 6465 bool make_core = false; 6466 switch (target_arch.GetMachine()) { 6467 case llvm::Triple::aarch64: 6468 case llvm::Triple::aarch64_32: 6469 case llvm::Triple::arm: 6470 case llvm::Triple::thumb: 6471 case llvm::Triple::x86: 6472 case llvm::Triple::x86_64: 6473 make_core = true; 6474 break; 6475 default: 6476 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6477 target_triple.str().c_str()); 6478 break; 6479 } 6480 6481 if (make_core) { 6482 std::vector<llvm::MachO::segment_command_64> segment_load_commands; 6483 // uint32_t range_info_idx = 0; 6484 MemoryRegionInfo range_info; 6485 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6486 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6487 const ByteOrder byte_order = target_arch.GetByteOrder(); 6488 std::vector<page_object> pages_to_copy; 6489 6490 if (range_error.Success()) { 6491 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6492 // Calculate correct protections 6493 uint32_t prot = 0; 6494 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6495 prot |= VM_PROT_READ; 6496 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6497 prot |= VM_PROT_WRITE; 6498 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6499 prot |= VM_PROT_EXECUTE; 6500 6501 const addr_t addr = range_info.GetRange().GetRangeBase(); 6502 const addr_t size = range_info.GetRange().GetByteSize(); 6503 6504 if (size == 0) 6505 break; 6506 6507 bool include_this_region = true; 6508 bool dirty_pages_only = false; 6509 if (core_style == SaveCoreStyle::eSaveCoreStackOnly) { 6510 dirty_pages_only = true; 6511 if (range_info.IsStackMemory() != MemoryRegionInfo::eYes) { 6512 include_this_region = false; 6513 } 6514 } 6515 if (core_style == SaveCoreStyle::eSaveCoreDirtyOnly) { 6516 dirty_pages_only = true; 6517 } 6518 6519 if (prot != 0 && include_this_region) { 6520 addr_t pagesize = range_info.GetPageSize(); 6521 const llvm::Optional<std::vector<addr_t>> &dirty_page_list = 6522 range_info.GetDirtyPageList(); 6523 if (dirty_pages_only && dirty_page_list.hasValue()) { 6524 for (addr_t dirtypage : dirty_page_list.getValue()) { 6525 page_object obj; 6526 obj.addr = dirtypage; 6527 obj.size = pagesize; 6528 obj.prot = prot; 6529 pages_to_copy.push_back(obj); 6530 } 6531 } else { 6532 page_object obj; 6533 obj.addr = addr; 6534 obj.size = size; 6535 obj.prot = prot; 6536 pages_to_copy.push_back(obj); 6537 } 6538 } 6539 6540 range_error = process_sp->GetMemoryRegionInfo( 6541 range_info.GetRange().GetRangeEnd(), range_info); 6542 if (range_error.Fail()) 6543 break; 6544 } 6545 6546 // Combine contiguous entries that have the same 6547 // protections so we don't have an excess of 6548 // load commands. 6549 std::vector<page_object> combined_page_objects; 6550 page_object last_obj; 6551 last_obj.addr = LLDB_INVALID_ADDRESS; 6552 last_obj.size = 0; 6553 for (page_object obj : pages_to_copy) { 6554 if (last_obj.addr == LLDB_INVALID_ADDRESS) { 6555 last_obj = obj; 6556 continue; 6557 } 6558 if (last_obj.addr + last_obj.size == obj.addr && 6559 last_obj.prot == obj.prot) { 6560 last_obj.size += obj.size; 6561 continue; 6562 } 6563 combined_page_objects.push_back(last_obj); 6564 last_obj = obj; 6565 } 6566 // Add the last entry we were looking to combine 6567 // on to the array. 6568 if (last_obj.addr != LLDB_INVALID_ADDRESS && last_obj.size != 0) 6569 combined_page_objects.push_back(last_obj); 6570 6571 for (page_object obj : combined_page_objects) { 6572 uint32_t cmd_type = LC_SEGMENT_64; 6573 uint32_t segment_size = sizeof(llvm::MachO::segment_command_64); 6574 if (addr_byte_size == 4) { 6575 cmd_type = LC_SEGMENT; 6576 segment_size = sizeof(llvm::MachO::segment_command); 6577 } 6578 llvm::MachO::segment_command_64 segment = { 6579 cmd_type, // uint32_t cmd; 6580 segment_size, // uint32_t cmdsize; 6581 {0}, // char segname[16]; 6582 obj.addr, // uint64_t vmaddr; // uint32_t for 32-bit 6583 // Mach-O 6584 obj.size, // uint64_t vmsize; // uint32_t for 32-bit 6585 // Mach-O 6586 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6587 obj.size, // uint64_t filesize; // uint32_t for 32-bit 6588 // Mach-O 6589 obj.prot, // uint32_t maxprot; 6590 obj.prot, // uint32_t initprot; 6591 0, // uint32_t nsects; 6592 0}; // uint32_t flags; 6593 segment_load_commands.push_back(segment); 6594 } 6595 6596 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6597 6598 llvm::MachO::mach_header_64 mach_header; 6599 if (addr_byte_size == 8) { 6600 mach_header.magic = MH_MAGIC_64; 6601 } else { 6602 mach_header.magic = MH_MAGIC; 6603 } 6604 mach_header.cputype = target_arch.GetMachOCPUType(); 6605 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6606 mach_header.filetype = MH_CORE; 6607 mach_header.ncmds = segment_load_commands.size(); 6608 mach_header.flags = 0; 6609 mach_header.reserved = 0; 6610 ThreadList &thread_list = process_sp->GetThreadList(); 6611 const uint32_t num_threads = thread_list.GetSize(); 6612 6613 // Make an array of LC_THREAD data items. Each one contains the 6614 // contents of the LC_THREAD load command. The data doesn't contain 6615 // the load command + load command size, we will add the load command 6616 // and load command size as we emit the data. 6617 std::vector<StreamString> LC_THREAD_datas(num_threads); 6618 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6619 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6620 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6621 LC_THREAD_data.SetByteOrder(byte_order); 6622 } 6623 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) { 6624 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6625 if (thread_sp) { 6626 switch (mach_header.cputype) { 6627 case llvm::MachO::CPU_TYPE_ARM64: 6628 case llvm::MachO::CPU_TYPE_ARM64_32: 6629 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6630 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6631 break; 6632 6633 case llvm::MachO::CPU_TYPE_ARM: 6634 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6635 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6636 break; 6637 6638 case llvm::MachO::CPU_TYPE_I386: 6639 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6640 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6641 break; 6642 6643 case llvm::MachO::CPU_TYPE_X86_64: 6644 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6645 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6646 break; 6647 } 6648 } 6649 } 6650 6651 // The size of the load command is the size of the segments... 6652 if (addr_byte_size == 8) { 6653 mach_header.sizeofcmds = segment_load_commands.size() * 6654 sizeof(llvm::MachO::segment_command_64); 6655 } else { 6656 mach_header.sizeofcmds = segment_load_commands.size() * 6657 sizeof(llvm::MachO::segment_command); 6658 } 6659 6660 // and the size of all LC_THREAD load command 6661 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6662 ++mach_header.ncmds; 6663 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6664 } 6665 6666 // Bits will be set to indicate which bits are NOT used in 6667 // addressing in this process or 0 for unknown. 6668 uint64_t address_mask = process_sp->GetCodeAddressMask(); 6669 if (address_mask != 0) { 6670 // LC_NOTE "addrable bits" 6671 mach_header.ncmds++; 6672 mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); 6673 } 6674 6675 // LC_NOTE "all image infos" 6676 mach_header.ncmds++; 6677 mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); 6678 6679 // Write the mach header 6680 buffer.PutHex32(mach_header.magic); 6681 buffer.PutHex32(mach_header.cputype); 6682 buffer.PutHex32(mach_header.cpusubtype); 6683 buffer.PutHex32(mach_header.filetype); 6684 buffer.PutHex32(mach_header.ncmds); 6685 buffer.PutHex32(mach_header.sizeofcmds); 6686 buffer.PutHex32(mach_header.flags); 6687 if (addr_byte_size == 8) { 6688 buffer.PutHex32(mach_header.reserved); 6689 } 6690 6691 // Skip the mach header and all load commands and align to the next 6692 // 0x1000 byte boundary 6693 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6694 6695 file_offset = llvm::alignTo(file_offset, 16); 6696 std::vector<std::unique_ptr<LCNoteEntry>> lc_notes; 6697 6698 // Add "addrable bits" LC_NOTE when an address mask is available 6699 if (address_mask != 0) { 6700 std::unique_ptr<LCNoteEntry> addrable_bits_lcnote_up( 6701 new LCNoteEntry(addr_byte_size, byte_order)); 6702 addrable_bits_lcnote_up->name = "addrable bits"; 6703 addrable_bits_lcnote_up->payload_file_offset = file_offset; 6704 int bits = std::bitset<64>(~address_mask).count(); 6705 addrable_bits_lcnote_up->payload.PutHex32(3); // version 6706 addrable_bits_lcnote_up->payload.PutHex32( 6707 bits); // # of bits used for addressing 6708 addrable_bits_lcnote_up->payload.PutHex64(0); // unused 6709 6710 file_offset += addrable_bits_lcnote_up->payload.GetSize(); 6711 6712 lc_notes.push_back(std::move(addrable_bits_lcnote_up)); 6713 } 6714 6715 // Add "all image infos" LC_NOTE 6716 std::unique_ptr<LCNoteEntry> all_image_infos_lcnote_up( 6717 new LCNoteEntry(addr_byte_size, byte_order)); 6718 all_image_infos_lcnote_up->name = "all image infos"; 6719 all_image_infos_lcnote_up->payload_file_offset = file_offset; 6720 file_offset = CreateAllImageInfosPayload( 6721 process_sp, file_offset, all_image_infos_lcnote_up->payload, 6722 core_style); 6723 lc_notes.push_back(std::move(all_image_infos_lcnote_up)); 6724 6725 // Add LC_NOTE load commands 6726 for (auto &lcnote : lc_notes) { 6727 // Add the LC_NOTE load command to the file. 6728 buffer.PutHex32(LC_NOTE); 6729 buffer.PutHex32(sizeof(llvm::MachO::note_command)); 6730 char namebuf[16]; 6731 memset(namebuf, 0, sizeof(namebuf)); 6732 // this is the uncommon case where strncpy is exactly 6733 // the right one, doesn't need to be nul terminated. 6734 strncpy(namebuf, lcnote->name.c_str(), sizeof(namebuf)); 6735 buffer.PutRawBytes(namebuf, sizeof(namebuf)); 6736 buffer.PutHex64(lcnote->payload_file_offset); 6737 buffer.PutHex64(lcnote->payload.GetSize()); 6738 } 6739 6740 // Align to 4096-byte page boundary for the LC_SEGMENTs. 6741 file_offset = llvm::alignTo(file_offset, 4096); 6742 6743 for (auto &segment : segment_load_commands) { 6744 segment.fileoff = file_offset; 6745 file_offset += segment.filesize; 6746 } 6747 6748 // Write out all of the LC_THREAD load commands 6749 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6750 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6751 buffer.PutHex32(LC_THREAD); 6752 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6753 buffer.Write(LC_THREAD_data.GetString().data(), LC_THREAD_data_size); 6754 } 6755 6756 // Write out all of the segment load commands 6757 for (const auto &segment : segment_load_commands) { 6758 buffer.PutHex32(segment.cmd); 6759 buffer.PutHex32(segment.cmdsize); 6760 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6761 if (addr_byte_size == 8) { 6762 buffer.PutHex64(segment.vmaddr); 6763 buffer.PutHex64(segment.vmsize); 6764 buffer.PutHex64(segment.fileoff); 6765 buffer.PutHex64(segment.filesize); 6766 } else { 6767 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6768 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6769 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6770 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6771 } 6772 buffer.PutHex32(segment.maxprot); 6773 buffer.PutHex32(segment.initprot); 6774 buffer.PutHex32(segment.nsects); 6775 buffer.PutHex32(segment.flags); 6776 } 6777 6778 std::string core_file_path(outfile.GetPath()); 6779 auto core_file = FileSystem::Instance().Open( 6780 outfile, File::eOpenOptionWriteOnly | File::eOpenOptionTruncate | 6781 File::eOpenOptionCanCreate); 6782 if (!core_file) { 6783 error = core_file.takeError(); 6784 } else { 6785 // Read 1 page at a time 6786 uint8_t bytes[0x1000]; 6787 // Write the mach header and load commands out to the core file 6788 size_t bytes_written = buffer.GetString().size(); 6789 error = 6790 core_file.get()->Write(buffer.GetString().data(), bytes_written); 6791 if (error.Success()) { 6792 6793 for (auto &lcnote : lc_notes) { 6794 if (core_file.get()->SeekFromStart(lcnote->payload_file_offset) == 6795 -1) { 6796 error.SetErrorStringWithFormat("Unable to seek to corefile pos " 6797 "to write '%s' LC_NOTE payload", 6798 lcnote->name.c_str()); 6799 return false; 6800 } 6801 bytes_written = lcnote->payload.GetSize(); 6802 error = core_file.get()->Write(lcnote->payload.GetData(), 6803 bytes_written); 6804 if (!error.Success()) 6805 return false; 6806 } 6807 6808 // Now write the file data for all memory segments in the process 6809 for (const auto &segment : segment_load_commands) { 6810 if (core_file.get()->SeekFromStart(segment.fileoff) == -1) { 6811 error.SetErrorStringWithFormat( 6812 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6813 segment.fileoff, core_file_path.c_str()); 6814 break; 6815 } 6816 6817 target.GetDebugger().GetAsyncOutputStream()->Printf( 6818 "Saving %" PRId64 6819 " bytes of data for memory region at 0x%" PRIx64 "\n", 6820 segment.vmsize, segment.vmaddr); 6821 addr_t bytes_left = segment.vmsize; 6822 addr_t addr = segment.vmaddr; 6823 Status memory_read_error; 6824 while (bytes_left > 0 && error.Success()) { 6825 const size_t bytes_to_read = 6826 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6827 6828 // In a savecore setting, we don't really care about caching, 6829 // as the data is dumped and very likely never read again, 6830 // so we call ReadMemoryFromInferior to bypass it. 6831 const size_t bytes_read = process_sp->ReadMemoryFromInferior( 6832 addr, bytes, bytes_to_read, memory_read_error); 6833 6834 if (bytes_read == bytes_to_read) { 6835 size_t bytes_written = bytes_read; 6836 error = core_file.get()->Write(bytes, bytes_written); 6837 bytes_left -= bytes_read; 6838 addr += bytes_read; 6839 } else { 6840 // Some pages within regions are not readable, those should 6841 // be zero filled 6842 memset(bytes, 0, bytes_to_read); 6843 size_t bytes_written = bytes_to_read; 6844 error = core_file.get()->Write(bytes, bytes_written); 6845 bytes_left -= bytes_to_read; 6846 addr += bytes_to_read; 6847 } 6848 } 6849 } 6850 } 6851 } 6852 } else { 6853 error.SetErrorString( 6854 "process doesn't support getting memory region info"); 6855 } 6856 } 6857 return true; // This is the right plug to handle saving core files for 6858 // this process 6859 } 6860 return false; 6861 } 6862 6863 ObjectFileMachO::MachOCorefileAllImageInfos 6864 ObjectFileMachO::GetCorefileAllImageInfos() { 6865 MachOCorefileAllImageInfos image_infos; 6866 6867 // Look for an "all image infos" LC_NOTE. 6868 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 6869 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6870 const uint32_t cmd_offset = offset; 6871 llvm::MachO::load_command lc; 6872 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6873 break; 6874 if (lc.cmd == LC_NOTE) { 6875 char data_owner[17]; 6876 m_data.CopyData(offset, 16, data_owner); 6877 data_owner[16] = '\0'; 6878 offset += 16; 6879 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 6880 offset += 4; /* size unused */ 6881 6882 if (strcmp("all image infos", data_owner) == 0) { 6883 offset = fileoff; 6884 // Read the struct all_image_infos_header. 6885 uint32_t version = m_data.GetU32(&offset); 6886 if (version != 1) { 6887 return image_infos; 6888 } 6889 uint32_t imgcount = m_data.GetU32(&offset); 6890 uint64_t entries_fileoff = m_data.GetU64(&offset); 6891 offset += 4; // uint32_t entries_size; 6892 offset += 4; // uint32_t unused; 6893 6894 offset = entries_fileoff; 6895 for (uint32_t i = 0; i < imgcount; i++) { 6896 // Read the struct image_entry. 6897 offset_t filepath_offset = m_data.GetU64(&offset); 6898 uuid_t uuid; 6899 memcpy(&uuid, m_data.GetData(&offset, sizeof(uuid_t)), 6900 sizeof(uuid_t)); 6901 uint64_t load_address = m_data.GetU64(&offset); 6902 offset_t seg_addrs_offset = m_data.GetU64(&offset); 6903 uint32_t segment_count = m_data.GetU32(&offset); 6904 uint32_t currently_executing = m_data.GetU32(&offset); 6905 6906 MachOCorefileImageEntry image_entry; 6907 image_entry.filename = (const char *)m_data.GetCStr(&filepath_offset); 6908 image_entry.uuid = UUID::fromData(uuid, sizeof(uuid_t)); 6909 image_entry.load_address = load_address; 6910 image_entry.currently_executing = currently_executing; 6911 6912 offset_t seg_vmaddrs_offset = seg_addrs_offset; 6913 for (uint32_t j = 0; j < segment_count; j++) { 6914 char segname[17]; 6915 m_data.CopyData(seg_vmaddrs_offset, 16, segname); 6916 segname[16] = '\0'; 6917 seg_vmaddrs_offset += 16; 6918 uint64_t vmaddr = m_data.GetU64(&seg_vmaddrs_offset); 6919 seg_vmaddrs_offset += 8; /* unused */ 6920 6921 std::tuple<ConstString, addr_t> new_seg{ConstString(segname), 6922 vmaddr}; 6923 image_entry.segment_load_addresses.push_back(new_seg); 6924 } 6925 image_infos.all_image_infos.push_back(image_entry); 6926 } 6927 } else if (strcmp("load binary", data_owner) == 0) { 6928 uint32_t version = m_data.GetU32(&fileoff); 6929 if (version == 1) { 6930 uuid_t uuid; 6931 memcpy(&uuid, m_data.GetData(&fileoff, sizeof(uuid_t)), 6932 sizeof(uuid_t)); 6933 uint64_t load_address = m_data.GetU64(&fileoff); 6934 uint64_t slide = m_data.GetU64(&fileoff); 6935 std::string filename = m_data.GetCStr(&fileoff); 6936 6937 MachOCorefileImageEntry image_entry; 6938 image_entry.filename = filename; 6939 image_entry.uuid = UUID::fromData(uuid, sizeof(uuid_t)); 6940 image_entry.load_address = load_address; 6941 image_entry.slide = slide; 6942 image_infos.all_image_infos.push_back(image_entry); 6943 } 6944 } 6945 } 6946 offset = cmd_offset + lc.cmdsize; 6947 } 6948 6949 return image_infos; 6950 } 6951 6952 bool ObjectFileMachO::LoadCoreFileImages(lldb_private::Process &process) { 6953 MachOCorefileAllImageInfos image_infos = GetCorefileAllImageInfos(); 6954 Log *log = GetLog(LLDBLog::DynamicLoader); 6955 6956 ModuleList added_modules; 6957 for (const MachOCorefileImageEntry &image : image_infos.all_image_infos) { 6958 ModuleSpec module_spec; 6959 module_spec.GetUUID() = image.uuid; 6960 if (image.filename.empty()) { 6961 char namebuf[80]; 6962 if (image.load_address != LLDB_INVALID_ADDRESS) 6963 snprintf(namebuf, sizeof(namebuf), "mem-image-0x%" PRIx64, 6964 image.load_address); 6965 else 6966 snprintf(namebuf, sizeof(namebuf), "mem-image+0x%" PRIx64, image.slide); 6967 module_spec.GetFileSpec() = FileSpec(namebuf); 6968 } else { 6969 module_spec.GetFileSpec() = FileSpec(image.filename.c_str()); 6970 } 6971 if (image.currently_executing) { 6972 Symbols::DownloadObjectAndSymbolFile(module_spec, true); 6973 if (FileSystem::Instance().Exists(module_spec.GetFileSpec())) { 6974 process.GetTarget().GetOrCreateModule(module_spec, false); 6975 } 6976 } 6977 Status error; 6978 ModuleSP module_sp = 6979 process.GetTarget().GetOrCreateModule(module_spec, false, &error); 6980 if (!module_sp.get() || !module_sp->GetObjectFile()) { 6981 if (image.load_address != LLDB_INVALID_ADDRESS) { 6982 module_sp = process.ReadModuleFromMemory(module_spec.GetFileSpec(), 6983 image.load_address); 6984 } 6985 } 6986 if (module_sp.get()) { 6987 // Will call ModulesDidLoad with all modules once they've all 6988 // been added to the Target with load addresses. Don't notify 6989 // here, before the load address is set. 6990 const bool notify = false; 6991 process.GetTarget().GetImages().AppendIfNeeded(module_sp, notify); 6992 added_modules.Append(module_sp, notify); 6993 if (image.segment_load_addresses.size() > 0) { 6994 if (log) { 6995 std::string uuidstr = image.uuid.GetAsString(); 6996 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 6997 "UUID %s with section load addresses", 6998 image.filename.c_str(), uuidstr.c_str()); 6999 } 7000 for (auto name_vmaddr_tuple : image.segment_load_addresses) { 7001 SectionList *sectlist = module_sp->GetObjectFile()->GetSectionList(); 7002 if (sectlist) { 7003 SectionSP sect_sp = 7004 sectlist->FindSectionByName(std::get<0>(name_vmaddr_tuple)); 7005 if (sect_sp) { 7006 process.GetTarget().SetSectionLoadAddress( 7007 sect_sp, std::get<1>(name_vmaddr_tuple)); 7008 } 7009 } 7010 } 7011 } else if (image.load_address != LLDB_INVALID_ADDRESS) { 7012 if (log) { 7013 std::string uuidstr = image.uuid.GetAsString(); 7014 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 7015 "UUID %s with load address 0x%" PRIx64, 7016 image.filename.c_str(), uuidstr.c_str(), 7017 image.load_address); 7018 } 7019 const bool address_is_slide = false; 7020 bool changed = false; 7021 module_sp->SetLoadAddress(process.GetTarget(), image.load_address, 7022 address_is_slide, changed); 7023 } else if (image.slide != 0) { 7024 if (log) { 7025 std::string uuidstr = image.uuid.GetAsString(); 7026 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 7027 "UUID %s with slide amount 0x%" PRIx64, 7028 image.filename.c_str(), uuidstr.c_str(), image.slide); 7029 } 7030 const bool address_is_slide = true; 7031 bool changed = false; 7032 module_sp->SetLoadAddress(process.GetTarget(), image.slide, 7033 address_is_slide, changed); 7034 } else { 7035 if (log) { 7036 std::string uuidstr = image.uuid.GetAsString(); 7037 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 7038 "UUID %s at its file address, no slide applied", 7039 image.filename.c_str(), uuidstr.c_str()); 7040 } 7041 const bool address_is_slide = true; 7042 bool changed = false; 7043 module_sp->SetLoadAddress(process.GetTarget(), 0, address_is_slide, 7044 changed); 7045 } 7046 } 7047 } 7048 if (added_modules.GetSize() > 0) { 7049 process.GetTarget().ModulesDidLoad(added_modules); 7050 process.Flush(); 7051 return true; 7052 } 7053 return false; 7054 } 7055