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 <mach/mach_init.h> 13 #include <mach/vm_map.h> 14 #include <lldb/Host/SafeMachO.h> 15 16 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 17 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 18 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 19 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 20 #include "lldb/Core/Debugger.h" 21 #include "lldb/Core/FileSpecList.h" 22 #include "lldb/Core/Module.h" 23 #include "lldb/Core/ModuleSpec.h" 24 #include "lldb/Core/PluginManager.h" 25 #include "lldb/Core/Progress.h" 26 #include "lldb/Core/Section.h" 27 #include "lldb/Core/StreamFile.h" 28 #include "lldb/Host/Host.h" 29 #include "lldb/Symbol/DWARFCallFrameInfo.h" 30 #include "lldb/Symbol/LocateSymbolFile.h" 31 #include "lldb/Symbol/ObjectFile.h" 32 #include "lldb/Target/DynamicLoader.h" 33 #include "lldb/Target/MemoryRegionInfo.h" 34 #include "lldb/Target/Platform.h" 35 #include "lldb/Target/Process.h" 36 #include "lldb/Target/SectionLoadList.h" 37 #include "lldb/Target/Target.h" 38 #include "lldb/Target/Thread.h" 39 #include "lldb/Target/ThreadList.h" 40 #include "lldb/Utility/ArchSpec.h" 41 #include "lldb/Utility/DataBuffer.h" 42 #include "lldb/Utility/FileSpec.h" 43 #include "lldb/Utility/Log.h" 44 #include "lldb/Utility/RangeMap.h" 45 #include "lldb/Utility/RegisterValue.h" 46 #include "lldb/Utility/Status.h" 47 #include "lldb/Utility/StreamString.h" 48 #include "lldb/Utility/Timer.h" 49 #include "lldb/Utility/UUID.h" 50 51 #include "lldb/Host/SafeMachO.h" 52 53 #include "llvm/ADT/DenseSet.h" 54 #include "llvm/Support/FormatVariadic.h" 55 #include "llvm/Support/MemoryBuffer.h" 56 57 #include "ObjectFileMachO.h" 58 59 #if defined(__APPLE__) 60 #include <TargetConditionals.h> 61 // GetLLDBSharedCacheUUID() needs to call dlsym() 62 #include <dlfcn.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(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_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(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2179 if (log && dsc_uuid.IsValid()) { 2180 LLDB_LOGF(log, "Shared cache %s has UUID %s", 2181 dyld_shared_cache.GetPath().c_str(), 2182 dsc_uuid.GetAsString().c_str()); 2183 } 2184 return dsc_uuid; 2185 } 2186 2187 static llvm::Optional<struct nlist_64> 2188 ParseNList(DataExtractor &nlist_data, lldb::offset_t &nlist_data_offset, 2189 size_t nlist_byte_size) { 2190 struct nlist_64 nlist; 2191 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2192 return {}; 2193 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 2194 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 2195 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 2196 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 2197 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 2198 return nlist; 2199 } 2200 2201 enum { DebugSymbols = true, NonDebugSymbols = false }; 2202 2203 void ObjectFileMachO::ParseSymtab(Symtab &symtab) { 2204 LLDB_SCOPED_TIMERF("ObjectFileMachO::ParseSymtab () module = %s", 2205 m_file.GetFilename().AsCString("")); 2206 ModuleSP module_sp(GetModule()); 2207 if (!module_sp) 2208 return; 2209 2210 Progress progress(llvm::formatv("Parsing symbol table for {0}", 2211 m_file.GetFilename().AsCString("<Unknown>"))); 2212 2213 llvm::MachO::symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2214 llvm::MachO::linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2215 llvm::MachO::linkedit_data_command exports_trie_load_command = {0, 0, 0, 0}; 2216 llvm::MachO::dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2217 // The data element of type bool indicates that this entry is thumb 2218 // code. 2219 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2220 2221 // Record the address of every function/data that we add to the symtab. 2222 // We add symbols to the table in the order of most information (nlist 2223 // records) to least (function starts), and avoid duplicating symbols 2224 // via this set. 2225 llvm::DenseSet<addr_t> symbols_added; 2226 2227 // We are using a llvm::DenseSet for "symbols_added" so we must be sure we 2228 // do not add the tombstone or empty keys to the set. 2229 auto add_symbol_addr = [&symbols_added](lldb::addr_t file_addr) { 2230 // Don't add the tombstone or empty keys. 2231 if (file_addr == UINT64_MAX || file_addr == UINT64_MAX - 1) 2232 return; 2233 symbols_added.insert(file_addr); 2234 }; 2235 FunctionStarts function_starts; 2236 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2237 uint32_t i; 2238 FileSpecList dylib_files; 2239 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2240 llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2241 llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2242 llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2243 UUID image_uuid; 2244 2245 for (i = 0; i < m_header.ncmds; ++i) { 2246 const lldb::offset_t cmd_offset = offset; 2247 // Read in the load command and load command size 2248 llvm::MachO::load_command lc; 2249 if (m_data.GetU32(&offset, &lc, 2) == nullptr) 2250 break; 2251 // Watch for the symbol table load command 2252 switch (lc.cmd) { 2253 case LC_SYMTAB: 2254 symtab_load_command.cmd = lc.cmd; 2255 symtab_load_command.cmdsize = lc.cmdsize; 2256 // Read in the rest of the symtab load command 2257 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2258 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2259 return; 2260 break; 2261 2262 case LC_DYLD_INFO: 2263 case LC_DYLD_INFO_ONLY: 2264 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2265 dyld_info.cmd = lc.cmd; 2266 dyld_info.cmdsize = lc.cmdsize; 2267 } else { 2268 memset(&dyld_info, 0, sizeof(dyld_info)); 2269 } 2270 break; 2271 2272 case LC_LOAD_DYLIB: 2273 case LC_LOAD_WEAK_DYLIB: 2274 case LC_REEXPORT_DYLIB: 2275 case LC_LOADFVMLIB: 2276 case LC_LOAD_UPWARD_DYLIB: { 2277 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2278 const char *path = m_data.PeekCStr(name_offset); 2279 if (path) { 2280 FileSpec file_spec(path); 2281 // Strip the path if there is @rpath, @executable, etc so we just use 2282 // the basename 2283 if (path[0] == '@') 2284 file_spec.GetDirectory().Clear(); 2285 2286 if (lc.cmd == LC_REEXPORT_DYLIB) { 2287 m_reexported_dylibs.AppendIfUnique(file_spec); 2288 } 2289 2290 dylib_files.Append(file_spec); 2291 } 2292 } break; 2293 2294 case LC_DYLD_EXPORTS_TRIE: 2295 exports_trie_load_command.cmd = lc.cmd; 2296 exports_trie_load_command.cmdsize = lc.cmdsize; 2297 if (m_data.GetU32(&offset, &exports_trie_load_command.dataoff, 2) == 2298 nullptr) // fill in offset and size fields 2299 memset(&exports_trie_load_command, 0, 2300 sizeof(exports_trie_load_command)); 2301 break; 2302 case LC_FUNCTION_STARTS: 2303 function_starts_load_command.cmd = lc.cmd; 2304 function_starts_load_command.cmdsize = lc.cmdsize; 2305 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2306 nullptr) // fill in data offset and size fields 2307 memset(&function_starts_load_command, 0, 2308 sizeof(function_starts_load_command)); 2309 break; 2310 2311 case LC_UUID: { 2312 const uint8_t *uuid_bytes = m_data.PeekData(offset, 16); 2313 2314 if (uuid_bytes) 2315 image_uuid = UUID::fromOptionalData(uuid_bytes, 16); 2316 break; 2317 } 2318 2319 default: 2320 break; 2321 } 2322 offset = cmd_offset + lc.cmdsize; 2323 } 2324 2325 if (!symtab_load_command.cmd) 2326 return; 2327 2328 SectionList *section_list = GetSectionList(); 2329 if (section_list == nullptr) 2330 return; 2331 2332 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2333 const ByteOrder byte_order = m_data.GetByteOrder(); 2334 bool bit_width_32 = addr_byte_size == 4; 2335 const size_t nlist_byte_size = 2336 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2337 2338 DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size); 2339 DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size); 2340 DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size); 2341 DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order, 2342 addr_byte_size); 2343 DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size); 2344 2345 const addr_t nlist_data_byte_size = 2346 symtab_load_command.nsyms * nlist_byte_size; 2347 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2348 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2349 2350 ProcessSP process_sp(m_process_wp.lock()); 2351 Process *process = process_sp.get(); 2352 2353 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2354 bool is_shared_cache_image = IsSharedCacheBinary(); 2355 bool is_local_shared_cache_image = is_shared_cache_image && !IsInMemory(); 2356 SectionSP linkedit_section_sp( 2357 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2358 2359 if (process && m_header.filetype != llvm::MachO::MH_OBJECT && 2360 !is_local_shared_cache_image) { 2361 Target &target = process->GetTarget(); 2362 2363 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2364 2365 // Reading mach file from memory in a process or core file... 2366 2367 if (linkedit_section_sp) { 2368 addr_t linkedit_load_addr = 2369 linkedit_section_sp->GetLoadBaseAddress(&target); 2370 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2371 // We might be trying to access the symbol table before the 2372 // __LINKEDIT's load address has been set in the target. We can't 2373 // fail to read the symbol table, so calculate the right address 2374 // manually 2375 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2376 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2377 } 2378 2379 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2380 const addr_t symoff_addr = linkedit_load_addr + 2381 symtab_load_command.symoff - 2382 linkedit_file_offset; 2383 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2384 linkedit_file_offset; 2385 2386 // Always load dyld - the dynamic linker - from memory if we didn't 2387 // find a binary anywhere else. lldb will not register 2388 // dylib/framework/bundle loads/unloads if we don't have the dyld 2389 // symbols, we force dyld to load from memory despite the user's 2390 // target.memory-module-load-level setting. 2391 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2392 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2393 DataBufferSP nlist_data_sp( 2394 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2395 if (nlist_data_sp) 2396 nlist_data.SetData(nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2397 if (m_dysymtab.nindirectsyms != 0) { 2398 const addr_t indirect_syms_addr = linkedit_load_addr + 2399 m_dysymtab.indirectsymoff - 2400 linkedit_file_offset; 2401 DataBufferSP indirect_syms_data_sp(ReadMemory( 2402 process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 2403 if (indirect_syms_data_sp) 2404 indirect_symbol_index_data.SetData( 2405 indirect_syms_data_sp, 0, 2406 indirect_syms_data_sp->GetByteSize()); 2407 // If this binary is outside the shared cache, 2408 // cache the string table. 2409 // Binaries in the shared cache all share a giant string table, 2410 // and we can't share the string tables across multiple 2411 // ObjectFileMachO's, so we'd end up re-reading this mega-strtab 2412 // for every binary in the shared cache - it would be a big perf 2413 // problem. For binaries outside the shared cache, it's faster to 2414 // read the entire strtab at once instead of piece-by-piece as we 2415 // process the nlist records. 2416 if (!is_shared_cache_image) { 2417 DataBufferSP strtab_data_sp( 2418 ReadMemory(process_sp, strtab_addr, strtab_data_byte_size)); 2419 if (strtab_data_sp) { 2420 strtab_data.SetData(strtab_data_sp, 0, 2421 strtab_data_sp->GetByteSize()); 2422 } 2423 } 2424 } 2425 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) { 2426 if (function_starts_load_command.cmd) { 2427 const addr_t func_start_addr = 2428 linkedit_load_addr + function_starts_load_command.dataoff - 2429 linkedit_file_offset; 2430 DataBufferSP func_start_data_sp( 2431 ReadMemory(process_sp, func_start_addr, 2432 function_starts_load_command.datasize)); 2433 if (func_start_data_sp) 2434 function_starts_data.SetData(func_start_data_sp, 0, 2435 func_start_data_sp->GetByteSize()); 2436 } 2437 } 2438 } 2439 } 2440 } else { 2441 if (is_local_shared_cache_image) { 2442 // The load commands in shared cache images are relative to the 2443 // beginning of the shared cache, not the library image. The 2444 // data we get handed when creating the ObjectFileMachO starts 2445 // at the beginning of a specific library and spans to the end 2446 // of the cache to be able to reach the shared LINKEDIT 2447 // segments. We need to convert the load command offsets to be 2448 // relative to the beginning of our specific image. 2449 lldb::addr_t linkedit_offset = linkedit_section_sp->GetFileOffset(); 2450 lldb::offset_t linkedit_slide = 2451 linkedit_offset - m_linkedit_original_offset; 2452 symtab_load_command.symoff += linkedit_slide; 2453 symtab_load_command.stroff += linkedit_slide; 2454 dyld_info.export_off += linkedit_slide; 2455 m_dysymtab.indirectsymoff += linkedit_slide; 2456 function_starts_load_command.dataoff += linkedit_slide; 2457 exports_trie_load_command.dataoff += linkedit_slide; 2458 } 2459 2460 nlist_data.SetData(m_data, symtab_load_command.symoff, 2461 nlist_data_byte_size); 2462 strtab_data.SetData(m_data, symtab_load_command.stroff, 2463 strtab_data_byte_size); 2464 2465 // We shouldn't have exports data from both the LC_DYLD_INFO command 2466 // AND the LC_DYLD_EXPORTS_TRIE command in the same binary: 2467 lldbassert(!((dyld_info.export_size > 0) 2468 && (exports_trie_load_command.datasize > 0))); 2469 if (dyld_info.export_size > 0) { 2470 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2471 dyld_info.export_size); 2472 } else if (exports_trie_load_command.datasize > 0) { 2473 dyld_trie_data.SetData(m_data, exports_trie_load_command.dataoff, 2474 exports_trie_load_command.datasize); 2475 } 2476 2477 if (m_dysymtab.nindirectsyms != 0) { 2478 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2479 m_dysymtab.nindirectsyms * 4); 2480 } 2481 if (function_starts_load_command.cmd) { 2482 function_starts_data.SetData(m_data, function_starts_load_command.dataoff, 2483 function_starts_load_command.datasize); 2484 } 2485 } 2486 2487 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2488 2489 ConstString g_segment_name_TEXT = GetSegmentNameTEXT(); 2490 ConstString g_segment_name_DATA = GetSegmentNameDATA(); 2491 ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2492 ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2493 ConstString g_segment_name_OBJC = GetSegmentNameOBJC(); 2494 ConstString g_section_name_eh_frame = GetSectionNameEHFrame(); 2495 SectionSP text_section_sp( 2496 section_list->FindSectionByName(g_segment_name_TEXT)); 2497 SectionSP data_section_sp( 2498 section_list->FindSectionByName(g_segment_name_DATA)); 2499 SectionSP data_dirty_section_sp( 2500 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2501 SectionSP data_const_section_sp( 2502 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2503 SectionSP objc_section_sp( 2504 section_list->FindSectionByName(g_segment_name_OBJC)); 2505 SectionSP eh_frame_section_sp; 2506 if (text_section_sp.get()) 2507 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2508 g_section_name_eh_frame); 2509 else 2510 eh_frame_section_sp = 2511 section_list->FindSectionByName(g_section_name_eh_frame); 2512 2513 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2514 const bool always_thumb = GetArchitecture().IsAlwaysThumbInstructions(); 2515 2516 // lldb works best if it knows the start address of all functions in a 2517 // module. Linker symbols or debug info are normally the best source of 2518 // information for start addr / size but they may be stripped in a released 2519 // binary. Two additional sources of information exist in Mach-O binaries: 2520 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2521 // function's start address in the 2522 // binary, relative to the text section. 2523 // eh_frame - the eh_frame FDEs have the start addr & size of 2524 // each function 2525 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2526 // all modern binaries. 2527 // Binaries built to run on older releases may need to use eh_frame 2528 // information. 2529 2530 if (text_section_sp && function_starts_data.GetByteSize()) { 2531 FunctionStarts::Entry function_start_entry; 2532 function_start_entry.data = false; 2533 lldb::offset_t function_start_offset = 0; 2534 function_start_entry.addr = text_section_sp->GetFileAddress(); 2535 uint64_t delta; 2536 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2537 0) { 2538 // Now append the current entry 2539 function_start_entry.addr += delta; 2540 if (is_arm) { 2541 if (function_start_entry.addr & 1) { 2542 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2543 function_start_entry.data = true; 2544 } else if (always_thumb) { 2545 function_start_entry.data = true; 2546 } 2547 } 2548 function_starts.Append(function_start_entry); 2549 } 2550 } else { 2551 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2552 // load command claiming an eh_frame but it doesn't actually have the 2553 // eh_frame content. And if we have a dSYM, we don't need to do any of 2554 // this fill-in-the-missing-symbols works anyway - the debug info should 2555 // give us all the functions in the module. 2556 if (text_section_sp.get() && eh_frame_section_sp.get() && 2557 m_type != eTypeDebugInfo) { 2558 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2559 DWARFCallFrameInfo::EH); 2560 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2561 eh_frame.GetFunctionAddressAndSizeVector(functions); 2562 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2563 size_t count = functions.GetSize(); 2564 for (size_t i = 0; i < count; ++i) { 2565 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2566 functions.GetEntryAtIndex(i); 2567 if (func) { 2568 FunctionStarts::Entry function_start_entry; 2569 function_start_entry.addr = func->base - text_base_addr; 2570 if (is_arm) { 2571 if (function_start_entry.addr & 1) { 2572 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2573 function_start_entry.data = true; 2574 } else if (always_thumb) { 2575 function_start_entry.data = true; 2576 } 2577 } 2578 function_starts.Append(function_start_entry); 2579 } 2580 } 2581 } 2582 } 2583 2584 const size_t function_starts_count = function_starts.GetSize(); 2585 2586 // For user process binaries (executables, dylibs, frameworks, bundles), if 2587 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2588 // going to assume the binary has been stripped. Don't allow assembly 2589 // language instruction emulation because we don't know proper function 2590 // start boundaries. 2591 // 2592 // For all other types of binaries (kernels, stand-alone bare board 2593 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2594 // sections - we should not make any assumptions about them based on that. 2595 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2596 m_allow_assembly_emulation_unwind_plans = false; 2597 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2598 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2599 2600 if (unwind_or_symbol_log) 2601 module_sp->LogMessage( 2602 unwind_or_symbol_log, 2603 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2604 } 2605 2606 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() 2607 ? eh_frame_section_sp->GetID() 2608 : static_cast<user_id_t>(NO_SECT); 2609 2610 uint32_t N_SO_index = UINT32_MAX; 2611 2612 MachSymtabSectionInfo section_info(section_list); 2613 std::vector<uint32_t> N_FUN_indexes; 2614 std::vector<uint32_t> N_NSYM_indexes; 2615 std::vector<uint32_t> N_INCL_indexes; 2616 std::vector<uint32_t> N_BRAC_indexes; 2617 std::vector<uint32_t> N_COMM_indexes; 2618 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2619 typedef llvm::DenseMap<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2620 typedef llvm::DenseMap<const char *, uint32_t> ConstNameToSymbolIndexMap; 2621 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2622 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2623 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2624 // Any symbols that get merged into another will get an entry in this map 2625 // so we know 2626 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2627 uint32_t nlist_idx = 0; 2628 Symbol *symbol_ptr = nullptr; 2629 2630 uint32_t sym_idx = 0; 2631 Symbol *sym = nullptr; 2632 size_t num_syms = 0; 2633 std::string memory_symbol_name; 2634 uint32_t unmapped_local_symbols_found = 0; 2635 2636 std::vector<TrieEntryWithOffset> reexport_trie_entries; 2637 std::vector<TrieEntryWithOffset> external_sym_trie_entries; 2638 std::set<lldb::addr_t> resolver_addresses; 2639 2640 if (dyld_trie_data.GetByteSize() > 0) { 2641 ConstString text_segment_name("__TEXT"); 2642 SectionSP text_segment_sp = 2643 GetSectionList()->FindSectionByName(text_segment_name); 2644 lldb::addr_t text_segment_file_addr = LLDB_INVALID_ADDRESS; 2645 if (text_segment_sp) 2646 text_segment_file_addr = text_segment_sp->GetFileAddress(); 2647 std::vector<llvm::StringRef> nameSlices; 2648 ParseTrieEntries(dyld_trie_data, 0, is_arm, text_segment_file_addr, 2649 nameSlices, resolver_addresses, reexport_trie_entries, 2650 external_sym_trie_entries); 2651 } 2652 2653 typedef std::set<ConstString> IndirectSymbols; 2654 IndirectSymbols indirect_symbol_names; 2655 2656 #if TARGET_OS_IPHONE 2657 2658 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2659 // optimized by moving LOCAL symbols out of the memory mapped portion of 2660 // the DSC. The symbol information has all been retained, but it isn't 2661 // available in the normal nlist data. However, there *are* duplicate 2662 // entries of *some* 2663 // LOCAL symbols in the normal nlist data. To handle this situation 2664 // correctly, we must first attempt 2665 // to parse any DSC unmapped symbol information. If we find any, we set a 2666 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2667 2668 if (IsSharedCacheBinary()) { 2669 // Before we can start mapping the DSC, we need to make certain the 2670 // target process is actually using the cache we can find. 2671 2672 // Next we need to determine the correct path for the dyld shared cache. 2673 2674 ArchSpec header_arch = GetArchitecture(); 2675 2676 UUID dsc_uuid; 2677 UUID process_shared_cache_uuid; 2678 addr_t process_shared_cache_base_addr; 2679 2680 if (process) { 2681 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, 2682 process_shared_cache_uuid); 2683 } 2684 2685 __block bool found_image = false; 2686 __block void *nlist_buffer = nullptr; 2687 __block unsigned nlist_count = 0; 2688 __block char *string_table = nullptr; 2689 __block vm_offset_t vm_nlist_memory = 0; 2690 __block mach_msg_type_number_t vm_nlist_bytes_read = 0; 2691 __block vm_offset_t vm_string_memory = 0; 2692 __block mach_msg_type_number_t vm_string_bytes_read = 0; 2693 2694 auto _ = llvm::make_scope_exit(^{ 2695 if (vm_nlist_memory) 2696 vm_deallocate(mach_task_self(), vm_nlist_memory, vm_nlist_bytes_read); 2697 if (vm_string_memory) 2698 vm_deallocate(mach_task_self(), vm_string_memory, vm_string_bytes_read); 2699 }); 2700 2701 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 2702 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 2703 UndefinedNameToDescMap undefined_name_to_desc; 2704 SymbolIndexToName reexport_shlib_needs_fixup; 2705 2706 dyld_for_each_installed_shared_cache(^(dyld_shared_cache_t shared_cache) { 2707 uuid_t cache_uuid; 2708 dyld_shared_cache_copy_uuid(shared_cache, &cache_uuid); 2709 if (found_image) 2710 return; 2711 2712 if (process_shared_cache_uuid.IsValid() && 2713 process_shared_cache_uuid != UUID::fromOptionalData(&cache_uuid, 16)) 2714 return; 2715 2716 dyld_shared_cache_for_each_image(shared_cache, ^(dyld_image_t image) { 2717 uuid_t dsc_image_uuid; 2718 if (found_image) 2719 return; 2720 2721 dyld_image_copy_uuid(image, &dsc_image_uuid); 2722 if (image_uuid != UUID::fromOptionalData(dsc_image_uuid, 16)) 2723 return; 2724 2725 found_image = true; 2726 2727 // Compute the size of the string table. We need to ask dyld for a 2728 // new SPI to avoid this step. 2729 dyld_image_local_nlist_content_4Symbolication( 2730 image, ^(const void *nlistStart, uint64_t nlistCount, 2731 const char *stringTable) { 2732 if (!nlistStart || !nlistCount) 2733 return; 2734 2735 // The buffers passed here are valid only inside the block. 2736 // Use vm_read to make a cheap copy of them available for our 2737 // processing later. 2738 kern_return_t ret = 2739 vm_read(mach_task_self(), (vm_address_t)nlistStart, 2740 nlist_byte_size * nlistCount, &vm_nlist_memory, 2741 &vm_nlist_bytes_read); 2742 if (ret != KERN_SUCCESS) 2743 return; 2744 assert(vm_nlist_bytes_read == nlist_byte_size * nlistCount); 2745 2746 // We don't know the size of the string table. It's cheaper 2747 // to map the whol VM region than to determine the size by 2748 // parsing all teh nlist entries. 2749 vm_address_t string_address = (vm_address_t)stringTable; 2750 vm_size_t region_size; 2751 mach_msg_type_number_t info_count = VM_REGION_BASIC_INFO_COUNT_64; 2752 vm_region_basic_info_data_t info; 2753 memory_object_name_t object; 2754 ret = vm_region_64(mach_task_self(), &string_address, 2755 ®ion_size, VM_REGION_BASIC_INFO_64, 2756 (vm_region_info_t)&info, &info_count, &object); 2757 if (ret != KERN_SUCCESS) 2758 return; 2759 2760 ret = vm_read(mach_task_self(), (vm_address_t)stringTable, 2761 region_size - 2762 ((vm_address_t)stringTable - string_address), 2763 &vm_string_memory, &vm_string_bytes_read); 2764 if (ret != KERN_SUCCESS) 2765 return; 2766 2767 nlist_buffer = (void *)vm_nlist_memory; 2768 string_table = (char *)vm_string_memory; 2769 nlist_count = nlistCount; 2770 }); 2771 }); 2772 }); 2773 if (nlist_buffer) { 2774 DataExtractor dsc_local_symbols_data(nlist_buffer, 2775 nlist_count * nlist_byte_size, 2776 byte_order, addr_byte_size); 2777 unmapped_local_symbols_found = nlist_count; 2778 2779 // The normal nlist code cannot correctly size the Symbols 2780 // array, we need to allocate it here. 2781 sym = symtab.Resize( 2782 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2783 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2784 num_syms = symtab.GetNumSymbols(); 2785 2786 lldb::offset_t nlist_data_offset = 0; 2787 2788 for (uint32_t nlist_index = 0; 2789 nlist_index < nlist_count; 2790 nlist_index++) { 2791 ///////////////////////////// 2792 { 2793 llvm::Optional<struct nlist_64> nlist_maybe = 2794 ParseNList(dsc_local_symbols_data, nlist_data_offset, 2795 nlist_byte_size); 2796 if (!nlist_maybe) 2797 break; 2798 struct nlist_64 nlist = *nlist_maybe; 2799 2800 SymbolType type = eSymbolTypeInvalid; 2801 const char *symbol_name = string_table + nlist.n_strx; 2802 2803 if (symbol_name == NULL) { 2804 // No symbol should be NULL, even the symbols with no 2805 // string values should have an offset zero which 2806 // points to an empty C-string 2807 Host::SystemLog( 2808 Host::eSystemLogError, 2809 "error: DSC unmapped local symbol[%u] has invalid " 2810 "string table offset 0x%x in %s, ignoring symbol\n", 2811 nlist_index, nlist.n_strx, 2812 module_sp->GetFileSpec().GetPath().c_str()); 2813 continue; 2814 } 2815 if (symbol_name[0] == '\0') 2816 symbol_name = NULL; 2817 2818 const char *symbol_name_non_abi_mangled = NULL; 2819 2820 SectionSP symbol_section; 2821 uint32_t symbol_byte_size = 0; 2822 bool add_nlist = true; 2823 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2824 bool demangled_is_synthesized = false; 2825 bool is_gsym = false; 2826 bool set_value = true; 2827 2828 assert(sym_idx < num_syms); 2829 2830 sym[sym_idx].SetDebug(is_debug); 2831 2832 if (is_debug) { 2833 switch (nlist.n_type) { 2834 case N_GSYM: 2835 // global symbol: name,,NO_SECT,type,0 2836 // Sometimes the N_GSYM value contains the address. 2837 2838 // FIXME: In the .o files, we have a GSYM and a debug 2839 // symbol for all the ObjC data. They 2840 // have the same address, but we want to ensure that 2841 // we always find only the real symbol, 'cause we 2842 // don't currently correctly attribute the 2843 // GSYM one to the ObjCClass/Ivar/MetaClass 2844 // symbol type. This is a temporary hack to make 2845 // sure the ObjectiveC symbols get treated correctly. 2846 // To do this right, we should coalesce all the GSYM 2847 // & global symbols that have the same address. 2848 2849 is_gsym = true; 2850 sym[sym_idx].SetExternal(true); 2851 2852 if (symbol_name && symbol_name[0] == '_' && 2853 symbol_name[1] == 'O') { 2854 llvm::StringRef symbol_name_ref(symbol_name); 2855 if (symbol_name_ref.startswith( 2856 g_objc_v2_prefix_class)) { 2857 symbol_name_non_abi_mangled = symbol_name + 1; 2858 symbol_name = 2859 symbol_name + g_objc_v2_prefix_class.size(); 2860 type = eSymbolTypeObjCClass; 2861 demangled_is_synthesized = true; 2862 2863 } else if (symbol_name_ref.startswith( 2864 g_objc_v2_prefix_metaclass)) { 2865 symbol_name_non_abi_mangled = symbol_name + 1; 2866 symbol_name = 2867 symbol_name + g_objc_v2_prefix_metaclass.size(); 2868 type = eSymbolTypeObjCMetaClass; 2869 demangled_is_synthesized = true; 2870 } else if (symbol_name_ref.startswith( 2871 g_objc_v2_prefix_ivar)) { 2872 symbol_name_non_abi_mangled = symbol_name + 1; 2873 symbol_name = 2874 symbol_name + g_objc_v2_prefix_ivar.size(); 2875 type = eSymbolTypeObjCIVar; 2876 demangled_is_synthesized = true; 2877 } 2878 } else { 2879 if (nlist.n_value != 0) 2880 symbol_section = section_info.GetSection( 2881 nlist.n_sect, nlist.n_value); 2882 type = eSymbolTypeData; 2883 } 2884 break; 2885 2886 case N_FNAME: 2887 // procedure name (f77 kludge): name,,NO_SECT,0,0 2888 type = eSymbolTypeCompiler; 2889 break; 2890 2891 case N_FUN: 2892 // procedure: name,,n_sect,linenumber,address 2893 if (symbol_name) { 2894 type = eSymbolTypeCode; 2895 symbol_section = section_info.GetSection( 2896 nlist.n_sect, nlist.n_value); 2897 2898 N_FUN_addr_to_sym_idx.insert( 2899 std::make_pair(nlist.n_value, sym_idx)); 2900 // We use the current number of symbols in the 2901 // symbol table in lieu of using nlist_idx in case 2902 // we ever start trimming entries out 2903 N_FUN_indexes.push_back(sym_idx); 2904 } else { 2905 type = eSymbolTypeCompiler; 2906 2907 if (!N_FUN_indexes.empty()) { 2908 // Copy the size of the function into the 2909 // original 2910 // STAB entry so we don't have 2911 // to hunt for it later 2912 symtab.SymbolAtIndex(N_FUN_indexes.back()) 2913 ->SetByteSize(nlist.n_value); 2914 N_FUN_indexes.pop_back(); 2915 // We don't really need the end function STAB as 2916 // it contains the size which we already placed 2917 // with the original symbol, so don't add it if 2918 // we want a minimal symbol table 2919 add_nlist = false; 2920 } 2921 } 2922 break; 2923 2924 case N_STSYM: 2925 // static symbol: name,,n_sect,type,address 2926 N_STSYM_addr_to_sym_idx.insert( 2927 std::make_pair(nlist.n_value, sym_idx)); 2928 symbol_section = section_info.GetSection(nlist.n_sect, 2929 nlist.n_value); 2930 if (symbol_name && symbol_name[0]) { 2931 type = ObjectFile::GetSymbolTypeFromName( 2932 symbol_name + 1, eSymbolTypeData); 2933 } 2934 break; 2935 2936 case N_LCSYM: 2937 // .lcomm symbol: name,,n_sect,type,address 2938 symbol_section = section_info.GetSection(nlist.n_sect, 2939 nlist.n_value); 2940 type = eSymbolTypeCommonBlock; 2941 break; 2942 2943 case N_BNSYM: 2944 // We use the current number of symbols in the symbol 2945 // table in lieu of using nlist_idx in case we ever 2946 // start trimming entries out Skip these if we want 2947 // minimal symbol tables 2948 add_nlist = false; 2949 break; 2950 2951 case N_ENSYM: 2952 // Set the size of the N_BNSYM to the terminating 2953 // index of this N_ENSYM so that we can always skip 2954 // the entire symbol if we need to navigate more 2955 // quickly at the source level when parsing STABS 2956 // Skip these if we want minimal symbol tables 2957 add_nlist = false; 2958 break; 2959 2960 case N_OPT: 2961 // emitted with gcc2_compiled and in gcc source 2962 type = eSymbolTypeCompiler; 2963 break; 2964 2965 case N_RSYM: 2966 // register sym: name,,NO_SECT,type,register 2967 type = eSymbolTypeVariable; 2968 break; 2969 2970 case N_SLINE: 2971 // src line: 0,,n_sect,linenumber,address 2972 symbol_section = section_info.GetSection(nlist.n_sect, 2973 nlist.n_value); 2974 type = eSymbolTypeLineEntry; 2975 break; 2976 2977 case N_SSYM: 2978 // structure elt: name,,NO_SECT,type,struct_offset 2979 type = eSymbolTypeVariableType; 2980 break; 2981 2982 case N_SO: 2983 // source file name 2984 type = eSymbolTypeSourceFile; 2985 if (symbol_name == NULL) { 2986 add_nlist = false; 2987 if (N_SO_index != UINT32_MAX) { 2988 // Set the size of the N_SO to the terminating 2989 // index of this N_SO so that we can always skip 2990 // the entire N_SO if we need to navigate more 2991 // quickly at the source level when parsing STABS 2992 symbol_ptr = symtab.SymbolAtIndex(N_SO_index); 2993 symbol_ptr->SetByteSize(sym_idx); 2994 symbol_ptr->SetSizeIsSibling(true); 2995 } 2996 N_NSYM_indexes.clear(); 2997 N_INCL_indexes.clear(); 2998 N_BRAC_indexes.clear(); 2999 N_COMM_indexes.clear(); 3000 N_FUN_indexes.clear(); 3001 N_SO_index = UINT32_MAX; 3002 } else { 3003 // We use the current number of symbols in the 3004 // symbol table in lieu of using nlist_idx in case 3005 // we ever start trimming entries out 3006 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3007 if (N_SO_has_full_path) { 3008 if ((N_SO_index == sym_idx - 1) && 3009 ((sym_idx - 1) < num_syms)) { 3010 // We have two consecutive N_SO entries where 3011 // the first contains a directory and the 3012 // second contains a full path. 3013 sym[sym_idx - 1].GetMangled().SetValue( 3014 ConstString(symbol_name), false); 3015 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3016 add_nlist = false; 3017 } else { 3018 // This is the first entry in a N_SO that 3019 // contains a directory or 3020 // a full path to the source file 3021 N_SO_index = sym_idx; 3022 } 3023 } else if ((N_SO_index == sym_idx - 1) && 3024 ((sym_idx - 1) < num_syms)) { 3025 // This is usually the second N_SO entry that 3026 // contains just the filename, so here we combine 3027 // it with the first one if we are minimizing the 3028 // symbol table 3029 const char *so_path = sym[sym_idx - 1] 3030 .GetMangled() 3031 .GetDemangledName() 3032 .AsCString(); 3033 if (so_path && so_path[0]) { 3034 std::string full_so_path(so_path); 3035 const size_t double_slash_pos = 3036 full_so_path.find("//"); 3037 if (double_slash_pos != std::string::npos) { 3038 // The linker has been generating bad N_SO 3039 // entries with doubled up paths 3040 // in the format "%s%s" where the first 3041 // string in the DW_AT_comp_dir, and the 3042 // second is the directory for the source 3043 // file so you end up with a path that looks 3044 // like "/tmp/src//tmp/src/" 3045 FileSpec so_dir(so_path); 3046 if (!FileSystem::Instance().Exists(so_dir)) { 3047 so_dir.SetFile( 3048 &full_so_path[double_slash_pos + 1], 3049 FileSpec::Style::native); 3050 if (FileSystem::Instance().Exists(so_dir)) { 3051 // Trim off the incorrect path 3052 full_so_path.erase(0, double_slash_pos + 1); 3053 } 3054 } 3055 } 3056 if (*full_so_path.rbegin() != '/') 3057 full_so_path += '/'; 3058 full_so_path += symbol_name; 3059 sym[sym_idx - 1].GetMangled().SetValue( 3060 ConstString(full_so_path.c_str()), false); 3061 add_nlist = false; 3062 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3063 } 3064 } else { 3065 // This could be a relative path to a N_SO 3066 N_SO_index = sym_idx; 3067 } 3068 } 3069 break; 3070 3071 case N_OSO: 3072 // object file name: name,,0,0,st_mtime 3073 type = eSymbolTypeObjectFile; 3074 break; 3075 3076 case N_LSYM: 3077 // local sym: name,,NO_SECT,type,offset 3078 type = eSymbolTypeLocal; 3079 break; 3080 3081 // INCL scopes 3082 case N_BINCL: 3083 // include file beginning: name,,NO_SECT,0,sum We use 3084 // the current number of symbols in the symbol table 3085 // in lieu of using nlist_idx in case we ever start 3086 // trimming entries out 3087 N_INCL_indexes.push_back(sym_idx); 3088 type = eSymbolTypeScopeBegin; 3089 break; 3090 3091 case N_EINCL: 3092 // include file end: name,,NO_SECT,0,0 3093 // Set the size of the N_BINCL to the terminating 3094 // index of this N_EINCL so that we can always skip 3095 // the entire symbol if we need to navigate more 3096 // quickly at the source level when parsing STABS 3097 if (!N_INCL_indexes.empty()) { 3098 symbol_ptr = 3099 symtab.SymbolAtIndex(N_INCL_indexes.back()); 3100 symbol_ptr->SetByteSize(sym_idx + 1); 3101 symbol_ptr->SetSizeIsSibling(true); 3102 N_INCL_indexes.pop_back(); 3103 } 3104 type = eSymbolTypeScopeEnd; 3105 break; 3106 3107 case N_SOL: 3108 // #included file name: name,,n_sect,0,address 3109 type = eSymbolTypeHeaderFile; 3110 3111 // We currently don't use the header files on darwin 3112 add_nlist = false; 3113 break; 3114 3115 case N_PARAMS: 3116 // compiler parameters: name,,NO_SECT,0,0 3117 type = eSymbolTypeCompiler; 3118 break; 3119 3120 case N_VERSION: 3121 // compiler version: name,,NO_SECT,0,0 3122 type = eSymbolTypeCompiler; 3123 break; 3124 3125 case N_OLEVEL: 3126 // compiler -O level: name,,NO_SECT,0,0 3127 type = eSymbolTypeCompiler; 3128 break; 3129 3130 case N_PSYM: 3131 // parameter: name,,NO_SECT,type,offset 3132 type = eSymbolTypeVariable; 3133 break; 3134 3135 case N_ENTRY: 3136 // alternate entry: name,,n_sect,linenumber,address 3137 symbol_section = section_info.GetSection(nlist.n_sect, 3138 nlist.n_value); 3139 type = eSymbolTypeLineEntry; 3140 break; 3141 3142 // Left and Right Braces 3143 case N_LBRAC: 3144 // left bracket: 0,,NO_SECT,nesting level,address We 3145 // use the current number of symbols in the symbol 3146 // table in lieu of using nlist_idx in case we ever 3147 // start trimming entries out 3148 symbol_section = section_info.GetSection(nlist.n_sect, 3149 nlist.n_value); 3150 N_BRAC_indexes.push_back(sym_idx); 3151 type = eSymbolTypeScopeBegin; 3152 break; 3153 3154 case N_RBRAC: 3155 // right bracket: 0,,NO_SECT,nesting level,address 3156 // Set the size of the N_LBRAC to the terminating 3157 // index of this N_RBRAC so that we can always skip 3158 // the entire symbol if we need to navigate more 3159 // quickly at the source level when parsing STABS 3160 symbol_section = section_info.GetSection(nlist.n_sect, 3161 nlist.n_value); 3162 if (!N_BRAC_indexes.empty()) { 3163 symbol_ptr = 3164 symtab.SymbolAtIndex(N_BRAC_indexes.back()); 3165 symbol_ptr->SetByteSize(sym_idx + 1); 3166 symbol_ptr->SetSizeIsSibling(true); 3167 N_BRAC_indexes.pop_back(); 3168 } 3169 type = eSymbolTypeScopeEnd; 3170 break; 3171 3172 case N_EXCL: 3173 // deleted include file: name,,NO_SECT,0,sum 3174 type = eSymbolTypeHeaderFile; 3175 break; 3176 3177 // COMM scopes 3178 case N_BCOMM: 3179 // begin common: name,,NO_SECT,0,0 3180 // We use the current number of symbols in the symbol 3181 // table in lieu of using nlist_idx in case we ever 3182 // start trimming entries out 3183 type = eSymbolTypeScopeBegin; 3184 N_COMM_indexes.push_back(sym_idx); 3185 break; 3186 3187 case N_ECOML: 3188 // end common (local name): 0,,n_sect,0,address 3189 symbol_section = section_info.GetSection(nlist.n_sect, 3190 nlist.n_value); 3191 // Fall through 3192 3193 case N_ECOMM: 3194 // end common: name,,n_sect,0,0 3195 // Set the size of the N_BCOMM to the terminating 3196 // index of this N_ECOMM/N_ECOML so that we can 3197 // always skip the entire symbol if we need to 3198 // navigate more quickly at the source level when 3199 // parsing STABS 3200 if (!N_COMM_indexes.empty()) { 3201 symbol_ptr = 3202 symtab.SymbolAtIndex(N_COMM_indexes.back()); 3203 symbol_ptr->SetByteSize(sym_idx + 1); 3204 symbol_ptr->SetSizeIsSibling(true); 3205 N_COMM_indexes.pop_back(); 3206 } 3207 type = eSymbolTypeScopeEnd; 3208 break; 3209 3210 case N_LENG: 3211 // second stab entry with length information 3212 type = eSymbolTypeAdditional; 3213 break; 3214 3215 default: 3216 break; 3217 } 3218 } else { 3219 // uint8_t n_pext = N_PEXT & nlist.n_type; 3220 uint8_t n_type = N_TYPE & nlist.n_type; 3221 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3222 3223 switch (n_type) { 3224 case N_INDR: { 3225 const char *reexport_name_cstr = 3226 strtab_data.PeekCStr(nlist.n_value); 3227 if (reexport_name_cstr && reexport_name_cstr[0]) { 3228 type = eSymbolTypeReExported; 3229 ConstString reexport_name( 3230 reexport_name_cstr + 3231 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3232 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3233 set_value = false; 3234 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3235 indirect_symbol_names.insert(ConstString( 3236 symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3237 } else 3238 type = eSymbolTypeUndefined; 3239 } break; 3240 3241 case N_UNDF: 3242 if (symbol_name && symbol_name[0]) { 3243 ConstString undefined_name( 3244 symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3245 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3246 } 3247 // Fall through 3248 case N_PBUD: 3249 type = eSymbolTypeUndefined; 3250 break; 3251 3252 case N_ABS: 3253 type = eSymbolTypeAbsolute; 3254 break; 3255 3256 case N_SECT: { 3257 symbol_section = section_info.GetSection(nlist.n_sect, 3258 nlist.n_value); 3259 3260 if (symbol_section == NULL) { 3261 // TODO: warn about this? 3262 add_nlist = false; 3263 break; 3264 } 3265 3266 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3267 type = eSymbolTypeException; 3268 } else { 3269 uint32_t section_type = 3270 symbol_section->Get() & SECTION_TYPE; 3271 3272 switch (section_type) { 3273 case S_CSTRING_LITERALS: 3274 type = eSymbolTypeData; 3275 break; // section with only literal C strings 3276 case S_4BYTE_LITERALS: 3277 type = eSymbolTypeData; 3278 break; // section with only 4 byte literals 3279 case S_8BYTE_LITERALS: 3280 type = eSymbolTypeData; 3281 break; // section with only 8 byte literals 3282 case S_LITERAL_POINTERS: 3283 type = eSymbolTypeTrampoline; 3284 break; // section with only pointers to literals 3285 case S_NON_LAZY_SYMBOL_POINTERS: 3286 type = eSymbolTypeTrampoline; 3287 break; // section with only non-lazy symbol 3288 // pointers 3289 case S_LAZY_SYMBOL_POINTERS: 3290 type = eSymbolTypeTrampoline; 3291 break; // section with only lazy symbol pointers 3292 case S_SYMBOL_STUBS: 3293 type = eSymbolTypeTrampoline; 3294 break; // section with only symbol stubs, byte 3295 // size of stub in the reserved2 field 3296 case S_MOD_INIT_FUNC_POINTERS: 3297 type = eSymbolTypeCode; 3298 break; // section with only function pointers for 3299 // initialization 3300 case S_MOD_TERM_FUNC_POINTERS: 3301 type = eSymbolTypeCode; 3302 break; // section with only function pointers for 3303 // termination 3304 case S_INTERPOSING: 3305 type = eSymbolTypeTrampoline; 3306 break; // section with only pairs of function 3307 // pointers for interposing 3308 case S_16BYTE_LITERALS: 3309 type = eSymbolTypeData; 3310 break; // section with only 16 byte literals 3311 case S_DTRACE_DOF: 3312 type = eSymbolTypeInstrumentation; 3313 break; 3314 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3315 type = eSymbolTypeTrampoline; 3316 break; 3317 default: 3318 switch (symbol_section->GetType()) { 3319 case lldb::eSectionTypeCode: 3320 type = eSymbolTypeCode; 3321 break; 3322 case eSectionTypeData: 3323 case eSectionTypeDataCString: // Inlined C string 3324 // data 3325 case eSectionTypeDataCStringPointers: // Pointers 3326 // to C 3327 // string 3328 // data 3329 case eSectionTypeDataSymbolAddress: // Address of 3330 // a symbol in 3331 // the symbol 3332 // table 3333 case eSectionTypeData4: 3334 case eSectionTypeData8: 3335 case eSectionTypeData16: 3336 type = eSymbolTypeData; 3337 break; 3338 default: 3339 break; 3340 } 3341 break; 3342 } 3343 3344 if (type == eSymbolTypeInvalid) { 3345 const char *symbol_sect_name = 3346 symbol_section->GetName().AsCString(); 3347 if (symbol_section->IsDescendant( 3348 text_section_sp.get())) { 3349 if (symbol_section->IsClear( 3350 S_ATTR_PURE_INSTRUCTIONS | 3351 S_ATTR_SELF_MODIFYING_CODE | 3352 S_ATTR_SOME_INSTRUCTIONS)) 3353 type = eSymbolTypeData; 3354 else 3355 type = eSymbolTypeCode; 3356 } else if (symbol_section->IsDescendant( 3357 data_section_sp.get()) || 3358 symbol_section->IsDescendant( 3359 data_dirty_section_sp.get()) || 3360 symbol_section->IsDescendant( 3361 data_const_section_sp.get())) { 3362 if (symbol_sect_name && 3363 ::strstr(symbol_sect_name, "__objc") == 3364 symbol_sect_name) { 3365 type = eSymbolTypeRuntime; 3366 3367 if (symbol_name) { 3368 llvm::StringRef symbol_name_ref(symbol_name); 3369 if (symbol_name_ref.startswith("_OBJC_")) { 3370 llvm::StringRef 3371 g_objc_v2_prefix_class( 3372 "_OBJC_CLASS_$_"); 3373 llvm::StringRef 3374 g_objc_v2_prefix_metaclass( 3375 "_OBJC_METACLASS_$_"); 3376 llvm::StringRef 3377 g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 3378 if (symbol_name_ref.startswith( 3379 g_objc_v2_prefix_class)) { 3380 symbol_name_non_abi_mangled = 3381 symbol_name + 1; 3382 symbol_name = 3383 symbol_name + 3384 g_objc_v2_prefix_class.size(); 3385 type = eSymbolTypeObjCClass; 3386 demangled_is_synthesized = true; 3387 } else if ( 3388 symbol_name_ref.startswith( 3389 g_objc_v2_prefix_metaclass)) { 3390 symbol_name_non_abi_mangled = 3391 symbol_name + 1; 3392 symbol_name = 3393 symbol_name + 3394 g_objc_v2_prefix_metaclass.size(); 3395 type = eSymbolTypeObjCMetaClass; 3396 demangled_is_synthesized = true; 3397 } else if (symbol_name_ref.startswith( 3398 g_objc_v2_prefix_ivar)) { 3399 symbol_name_non_abi_mangled = 3400 symbol_name + 1; 3401 symbol_name = 3402 symbol_name + 3403 g_objc_v2_prefix_ivar.size(); 3404 type = eSymbolTypeObjCIVar; 3405 demangled_is_synthesized = true; 3406 } 3407 } 3408 } 3409 } else if (symbol_sect_name && 3410 ::strstr(symbol_sect_name, 3411 "__gcc_except_tab") == 3412 symbol_sect_name) { 3413 type = eSymbolTypeException; 3414 } else { 3415 type = eSymbolTypeData; 3416 } 3417 } else if (symbol_sect_name && 3418 ::strstr(symbol_sect_name, "__IMPORT") == 3419 symbol_sect_name) { 3420 type = eSymbolTypeTrampoline; 3421 } else if (symbol_section->IsDescendant( 3422 objc_section_sp.get())) { 3423 type = eSymbolTypeRuntime; 3424 if (symbol_name && symbol_name[0] == '.') { 3425 llvm::StringRef symbol_name_ref(symbol_name); 3426 llvm::StringRef 3427 g_objc_v1_prefix_class(".objc_class_name_"); 3428 if (symbol_name_ref.startswith( 3429 g_objc_v1_prefix_class)) { 3430 symbol_name_non_abi_mangled = symbol_name; 3431 symbol_name = symbol_name + 3432 g_objc_v1_prefix_class.size(); 3433 type = eSymbolTypeObjCClass; 3434 demangled_is_synthesized = true; 3435 } 3436 } 3437 } 3438 } 3439 } 3440 } break; 3441 } 3442 } 3443 3444 if (add_nlist) { 3445 uint64_t symbol_value = nlist.n_value; 3446 if (symbol_name_non_abi_mangled) { 3447 sym[sym_idx].GetMangled().SetMangledName( 3448 ConstString(symbol_name_non_abi_mangled)); 3449 sym[sym_idx].GetMangled().SetDemangledName( 3450 ConstString(symbol_name)); 3451 } else { 3452 bool symbol_name_is_mangled = false; 3453 3454 if (symbol_name && symbol_name[0] == '_') { 3455 symbol_name_is_mangled = symbol_name[1] == '_'; 3456 symbol_name++; // Skip the leading underscore 3457 } 3458 3459 if (symbol_name) { 3460 ConstString const_symbol_name(symbol_name); 3461 sym[sym_idx].GetMangled().SetValue( 3462 const_symbol_name, symbol_name_is_mangled); 3463 if (is_gsym && is_debug) { 3464 const char *gsym_name = 3465 sym[sym_idx] 3466 .GetMangled() 3467 .GetName(Mangled::ePreferMangled) 3468 .GetCString(); 3469 if (gsym_name) 3470 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3471 } 3472 } 3473 } 3474 if (symbol_section) { 3475 const addr_t section_file_addr = 3476 symbol_section->GetFileAddress(); 3477 if (symbol_byte_size == 0 && 3478 function_starts_count > 0) { 3479 addr_t symbol_lookup_file_addr = nlist.n_value; 3480 // Do an exact address match for non-ARM addresses, 3481 // else get the closest since the symbol might be a 3482 // thumb symbol which has an address with bit zero 3483 // set 3484 FunctionStarts::Entry *func_start_entry = 3485 function_starts.FindEntry(symbol_lookup_file_addr, 3486 !is_arm); 3487 if (is_arm && func_start_entry) { 3488 // Verify that the function start address is the 3489 // symbol address (ARM) or the symbol address + 1 3490 // (thumb) 3491 if (func_start_entry->addr != 3492 symbol_lookup_file_addr && 3493 func_start_entry->addr != 3494 (symbol_lookup_file_addr + 1)) { 3495 // Not the right entry, NULL it out... 3496 func_start_entry = NULL; 3497 } 3498 } 3499 if (func_start_entry) { 3500 func_start_entry->data = true; 3501 3502 addr_t symbol_file_addr = func_start_entry->addr; 3503 uint32_t symbol_flags = 0; 3504 if (is_arm) { 3505 if (symbol_file_addr & 1) 3506 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3507 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3508 } 3509 3510 const FunctionStarts::Entry *next_func_start_entry = 3511 function_starts.FindNextEntry(func_start_entry); 3512 const addr_t section_end_file_addr = 3513 section_file_addr + 3514 symbol_section->GetByteSize(); 3515 if (next_func_start_entry) { 3516 addr_t next_symbol_file_addr = 3517 next_func_start_entry->addr; 3518 // Be sure the clear the Thumb address bit when 3519 // we calculate the size from the current and 3520 // next address 3521 if (is_arm) 3522 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3523 symbol_byte_size = std::min<lldb::addr_t>( 3524 next_symbol_file_addr - symbol_file_addr, 3525 section_end_file_addr - symbol_file_addr); 3526 } else { 3527 symbol_byte_size = 3528 section_end_file_addr - symbol_file_addr; 3529 } 3530 } 3531 } 3532 symbol_value -= section_file_addr; 3533 } 3534 3535 if (is_debug == false) { 3536 if (type == eSymbolTypeCode) { 3537 // See if we can find a N_FUN entry for any code 3538 // symbols. If we do find a match, and the name 3539 // matches, then we can merge the two into just the 3540 // function symbol to avoid duplicate entries in 3541 // the symbol table 3542 auto range = 3543 N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3544 if (range.first != range.second) { 3545 bool found_it = false; 3546 for (auto pos = range.first; pos != range.second; 3547 ++pos) { 3548 if (sym[sym_idx].GetMangled().GetName( 3549 Mangled::ePreferMangled) == 3550 sym[pos->second].GetMangled().GetName( 3551 Mangled::ePreferMangled)) { 3552 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3553 // We just need the flags from the linker 3554 // symbol, so put these flags 3555 // into the N_FUN flags to avoid duplicate 3556 // symbols in the symbol table 3557 sym[pos->second].SetExternal( 3558 sym[sym_idx].IsExternal()); 3559 sym[pos->second].SetFlags(nlist.n_type << 16 | 3560 nlist.n_desc); 3561 if (resolver_addresses.find(nlist.n_value) != 3562 resolver_addresses.end()) 3563 sym[pos->second].SetType(eSymbolTypeResolver); 3564 sym[sym_idx].Clear(); 3565 found_it = true; 3566 break; 3567 } 3568 } 3569 if (found_it) 3570 continue; 3571 } else { 3572 if (resolver_addresses.find(nlist.n_value) != 3573 resolver_addresses.end()) 3574 type = eSymbolTypeResolver; 3575 } 3576 } else if (type == eSymbolTypeData || 3577 type == eSymbolTypeObjCClass || 3578 type == eSymbolTypeObjCMetaClass || 3579 type == eSymbolTypeObjCIVar) { 3580 // See if we can find a N_STSYM entry for any data 3581 // symbols. If we do find a match, and the name 3582 // matches, then we can merge the two into just the 3583 // Static symbol to avoid duplicate entries in the 3584 // symbol table 3585 auto range = N_STSYM_addr_to_sym_idx.equal_range( 3586 nlist.n_value); 3587 if (range.first != range.second) { 3588 bool found_it = false; 3589 for (auto pos = range.first; pos != range.second; 3590 ++pos) { 3591 if (sym[sym_idx].GetMangled().GetName( 3592 Mangled::ePreferMangled) == 3593 sym[pos->second].GetMangled().GetName( 3594 Mangled::ePreferMangled)) { 3595 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3596 // We just need the flags from the linker 3597 // symbol, so put these flags 3598 // into the N_STSYM flags to avoid duplicate 3599 // symbols in the symbol table 3600 sym[pos->second].SetExternal( 3601 sym[sym_idx].IsExternal()); 3602 sym[pos->second].SetFlags(nlist.n_type << 16 | 3603 nlist.n_desc); 3604 sym[sym_idx].Clear(); 3605 found_it = true; 3606 break; 3607 } 3608 } 3609 if (found_it) 3610 continue; 3611 } else { 3612 const char *gsym_name = 3613 sym[sym_idx] 3614 .GetMangled() 3615 .GetName(Mangled::ePreferMangled) 3616 .GetCString(); 3617 if (gsym_name) { 3618 // Combine N_GSYM stab entries with the non 3619 // stab symbol 3620 ConstNameToSymbolIndexMap::const_iterator pos = 3621 N_GSYM_name_to_sym_idx.find(gsym_name); 3622 if (pos != N_GSYM_name_to_sym_idx.end()) { 3623 const uint32_t GSYM_sym_idx = pos->second; 3624 m_nlist_idx_to_sym_idx[nlist_idx] = 3625 GSYM_sym_idx; 3626 // Copy the address, because often the N_GSYM 3627 // address has an invalid address of zero 3628 // when the global is a common symbol 3629 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3630 symbol_section); 3631 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3632 symbol_value); 3633 add_symbol_addr(sym[GSYM_sym_idx] 3634 .GetAddress() 3635 .GetFileAddress()); 3636 // We just need the flags from the linker 3637 // symbol, so put these flags 3638 // into the N_GSYM flags to avoid duplicate 3639 // symbols in the symbol table 3640 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 3641 nlist.n_desc); 3642 sym[sym_idx].Clear(); 3643 continue; 3644 } 3645 } 3646 } 3647 } 3648 } 3649 3650 sym[sym_idx].SetID(nlist_idx); 3651 sym[sym_idx].SetType(type); 3652 if (set_value) { 3653 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 3654 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3655 add_symbol_addr( 3656 sym[sym_idx].GetAddress().GetFileAddress()); 3657 } 3658 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 3659 3660 if (symbol_byte_size > 0) 3661 sym[sym_idx].SetByteSize(symbol_byte_size); 3662 3663 if (demangled_is_synthesized) 3664 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3665 ++sym_idx; 3666 } else { 3667 sym[sym_idx].Clear(); 3668 } 3669 } 3670 ///////////////////////////// 3671 } 3672 } 3673 3674 for (const auto &pos : reexport_shlib_needs_fixup) { 3675 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3676 if (undef_pos != undefined_name_to_desc.end()) { 3677 const uint8_t dylib_ordinal = 3678 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3679 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3680 sym[pos.first].SetReExportedSymbolSharedLibrary( 3681 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3682 } 3683 } 3684 } 3685 3686 #endif 3687 lldb::offset_t nlist_data_offset = 0; 3688 3689 if (nlist_data.GetByteSize() > 0) { 3690 3691 // If the sym array was not created while parsing the DSC unmapped 3692 // symbols, create it now. 3693 if (sym == nullptr) { 3694 sym = 3695 symtab.Resize(symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3696 num_syms = symtab.GetNumSymbols(); 3697 } 3698 3699 if (unmapped_local_symbols_found) { 3700 assert(m_dysymtab.ilocalsym == 0); 3701 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3702 nlist_idx = m_dysymtab.nlocalsym; 3703 } else { 3704 nlist_idx = 0; 3705 } 3706 3707 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 3708 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 3709 UndefinedNameToDescMap undefined_name_to_desc; 3710 SymbolIndexToName reexport_shlib_needs_fixup; 3711 3712 // Symtab parsing is a huge mess. Everything is entangled and the code 3713 // requires access to a ridiculous amount of variables. LLDB depends 3714 // heavily on the proper merging of symbols and to get that right we need 3715 // to make sure we have parsed all the debug symbols first. Therefore we 3716 // invoke the lambda twice, once to parse only the debug symbols and then 3717 // once more to parse the remaining symbols. 3718 auto ParseSymbolLambda = [&](struct nlist_64 &nlist, uint32_t nlist_idx, 3719 bool debug_only) { 3720 const bool is_debug = ((nlist.n_type & N_STAB) != 0); 3721 if (is_debug != debug_only) 3722 return true; 3723 3724 const char *symbol_name_non_abi_mangled = nullptr; 3725 const char *symbol_name = nullptr; 3726 3727 if (have_strtab_data) { 3728 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3729 3730 if (symbol_name == nullptr) { 3731 // No symbol should be NULL, even the symbols with no string values 3732 // should have an offset zero which points to an empty C-string 3733 Host::SystemLog(Host::eSystemLogError, 3734 "error: symbol[%u] has invalid string table offset " 3735 "0x%x in %s, ignoring symbol\n", 3736 nlist_idx, nlist.n_strx, 3737 module_sp->GetFileSpec().GetPath().c_str()); 3738 return true; 3739 } 3740 if (symbol_name[0] == '\0') 3741 symbol_name = nullptr; 3742 } else { 3743 const addr_t str_addr = strtab_addr + nlist.n_strx; 3744 Status str_error; 3745 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3746 str_error)) 3747 symbol_name = memory_symbol_name.c_str(); 3748 } 3749 3750 SymbolType type = eSymbolTypeInvalid; 3751 SectionSP symbol_section; 3752 lldb::addr_t symbol_byte_size = 0; 3753 bool add_nlist = true; 3754 bool is_gsym = false; 3755 bool demangled_is_synthesized = false; 3756 bool set_value = true; 3757 3758 assert(sym_idx < num_syms); 3759 sym[sym_idx].SetDebug(is_debug); 3760 3761 if (is_debug) { 3762 switch (nlist.n_type) { 3763 case N_GSYM: 3764 // global symbol: name,,NO_SECT,type,0 3765 // Sometimes the N_GSYM value contains the address. 3766 3767 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3768 // the ObjC data. They 3769 // have the same address, but we want to ensure that we always find 3770 // only the real symbol, 'cause we don't currently correctly 3771 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3772 // type. This is a temporary hack to make sure the ObjectiveC 3773 // symbols get treated correctly. To do this right, we should 3774 // coalesce all the GSYM & global symbols that have the same 3775 // address. 3776 is_gsym = true; 3777 sym[sym_idx].SetExternal(true); 3778 3779 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3780 llvm::StringRef symbol_name_ref(symbol_name); 3781 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3782 symbol_name_non_abi_mangled = symbol_name + 1; 3783 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3784 type = eSymbolTypeObjCClass; 3785 demangled_is_synthesized = true; 3786 3787 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 3788 symbol_name_non_abi_mangled = symbol_name + 1; 3789 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3790 type = eSymbolTypeObjCMetaClass; 3791 demangled_is_synthesized = true; 3792 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3793 symbol_name_non_abi_mangled = symbol_name + 1; 3794 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3795 type = eSymbolTypeObjCIVar; 3796 demangled_is_synthesized = true; 3797 } 3798 } else { 3799 if (nlist.n_value != 0) 3800 symbol_section = 3801 section_info.GetSection(nlist.n_sect, nlist.n_value); 3802 type = eSymbolTypeData; 3803 } 3804 break; 3805 3806 case N_FNAME: 3807 // procedure name (f77 kludge): name,,NO_SECT,0,0 3808 type = eSymbolTypeCompiler; 3809 break; 3810 3811 case N_FUN: 3812 // procedure: name,,n_sect,linenumber,address 3813 if (symbol_name) { 3814 type = eSymbolTypeCode; 3815 symbol_section = 3816 section_info.GetSection(nlist.n_sect, nlist.n_value); 3817 3818 N_FUN_addr_to_sym_idx.insert( 3819 std::make_pair(nlist.n_value, sym_idx)); 3820 // We use the current number of symbols in the symbol table in 3821 // lieu of using nlist_idx in case we ever start trimming entries 3822 // out 3823 N_FUN_indexes.push_back(sym_idx); 3824 } else { 3825 type = eSymbolTypeCompiler; 3826 3827 if (!N_FUN_indexes.empty()) { 3828 // Copy the size of the function into the original STAB entry 3829 // so we don't have to hunt for it later 3830 symtab.SymbolAtIndex(N_FUN_indexes.back()) 3831 ->SetByteSize(nlist.n_value); 3832 N_FUN_indexes.pop_back(); 3833 // We don't really need the end function STAB as it contains 3834 // the size which we already placed with the original symbol, 3835 // so don't add it if we want a minimal symbol table 3836 add_nlist = false; 3837 } 3838 } 3839 break; 3840 3841 case N_STSYM: 3842 // static symbol: name,,n_sect,type,address 3843 N_STSYM_addr_to_sym_idx.insert( 3844 std::make_pair(nlist.n_value, sym_idx)); 3845 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3846 if (symbol_name && symbol_name[0]) { 3847 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3848 eSymbolTypeData); 3849 } 3850 break; 3851 3852 case N_LCSYM: 3853 // .lcomm symbol: name,,n_sect,type,address 3854 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3855 type = eSymbolTypeCommonBlock; 3856 break; 3857 3858 case N_BNSYM: 3859 // We use the current number of symbols in the symbol table in lieu 3860 // of using nlist_idx in case we ever start trimming entries out 3861 // Skip these if we want minimal symbol tables 3862 add_nlist = false; 3863 break; 3864 3865 case N_ENSYM: 3866 // Set the size of the N_BNSYM to the terminating index of this 3867 // N_ENSYM so that we can always skip the entire symbol if we need 3868 // to navigate more quickly at the source level when parsing STABS 3869 // Skip these if we want minimal symbol tables 3870 add_nlist = false; 3871 break; 3872 3873 case N_OPT: 3874 // emitted with gcc2_compiled and in gcc source 3875 type = eSymbolTypeCompiler; 3876 break; 3877 3878 case N_RSYM: 3879 // register sym: name,,NO_SECT,type,register 3880 type = eSymbolTypeVariable; 3881 break; 3882 3883 case N_SLINE: 3884 // src line: 0,,n_sect,linenumber,address 3885 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3886 type = eSymbolTypeLineEntry; 3887 break; 3888 3889 case N_SSYM: 3890 // structure elt: name,,NO_SECT,type,struct_offset 3891 type = eSymbolTypeVariableType; 3892 break; 3893 3894 case N_SO: 3895 // source file name 3896 type = eSymbolTypeSourceFile; 3897 if (symbol_name == nullptr) { 3898 add_nlist = false; 3899 if (N_SO_index != UINT32_MAX) { 3900 // Set the size of the N_SO to the terminating index of this 3901 // N_SO so that we can always skip the entire N_SO if we need 3902 // to navigate more quickly at the source level when parsing 3903 // STABS 3904 symbol_ptr = symtab.SymbolAtIndex(N_SO_index); 3905 symbol_ptr->SetByteSize(sym_idx); 3906 symbol_ptr->SetSizeIsSibling(true); 3907 } 3908 N_NSYM_indexes.clear(); 3909 N_INCL_indexes.clear(); 3910 N_BRAC_indexes.clear(); 3911 N_COMM_indexes.clear(); 3912 N_FUN_indexes.clear(); 3913 N_SO_index = UINT32_MAX; 3914 } else { 3915 // We use the current number of symbols in the symbol table in 3916 // lieu of using nlist_idx in case we ever start trimming entries 3917 // out 3918 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3919 if (N_SO_has_full_path) { 3920 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3921 // We have two consecutive N_SO entries where the first 3922 // contains a directory and the second contains a full path. 3923 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), 3924 false); 3925 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3926 add_nlist = false; 3927 } else { 3928 // This is the first entry in a N_SO that contains a 3929 // directory or a full path to the source file 3930 N_SO_index = sym_idx; 3931 } 3932 } else if ((N_SO_index == sym_idx - 1) && 3933 ((sym_idx - 1) < num_syms)) { 3934 // This is usually the second N_SO entry that contains just the 3935 // filename, so here we combine it with the first one if we are 3936 // minimizing the symbol table 3937 const char *so_path = 3938 sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3939 if (so_path && so_path[0]) { 3940 std::string full_so_path(so_path); 3941 const size_t double_slash_pos = full_so_path.find("//"); 3942 if (double_slash_pos != std::string::npos) { 3943 // The linker has been generating bad N_SO entries with 3944 // doubled up paths in the format "%s%s" where the first 3945 // string in the DW_AT_comp_dir, and the second is the 3946 // directory for the source file so you end up with a path 3947 // that looks like "/tmp/src//tmp/src/" 3948 FileSpec so_dir(so_path); 3949 if (!FileSystem::Instance().Exists(so_dir)) { 3950 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 3951 FileSpec::Style::native); 3952 if (FileSystem::Instance().Exists(so_dir)) { 3953 // Trim off the incorrect path 3954 full_so_path.erase(0, double_slash_pos + 1); 3955 } 3956 } 3957 } 3958 if (*full_so_path.rbegin() != '/') 3959 full_so_path += '/'; 3960 full_so_path += symbol_name; 3961 sym[sym_idx - 1].GetMangled().SetValue( 3962 ConstString(full_so_path.c_str()), false); 3963 add_nlist = false; 3964 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3965 } 3966 } else { 3967 // This could be a relative path to a N_SO 3968 N_SO_index = sym_idx; 3969 } 3970 } 3971 break; 3972 3973 case N_OSO: 3974 // object file name: name,,0,0,st_mtime 3975 type = eSymbolTypeObjectFile; 3976 break; 3977 3978 case N_LSYM: 3979 // local sym: name,,NO_SECT,type,offset 3980 type = eSymbolTypeLocal; 3981 break; 3982 3983 // INCL scopes 3984 case N_BINCL: 3985 // include file beginning: name,,NO_SECT,0,sum We use the current 3986 // number of symbols in the symbol table in lieu of using nlist_idx 3987 // in case we ever start trimming entries out 3988 N_INCL_indexes.push_back(sym_idx); 3989 type = eSymbolTypeScopeBegin; 3990 break; 3991 3992 case N_EINCL: 3993 // include file end: name,,NO_SECT,0,0 3994 // Set the size of the N_BINCL to the terminating index of this 3995 // N_EINCL so that we can always skip the entire symbol if we need 3996 // to navigate more quickly at the source level when parsing STABS 3997 if (!N_INCL_indexes.empty()) { 3998 symbol_ptr = symtab.SymbolAtIndex(N_INCL_indexes.back()); 3999 symbol_ptr->SetByteSize(sym_idx + 1); 4000 symbol_ptr->SetSizeIsSibling(true); 4001 N_INCL_indexes.pop_back(); 4002 } 4003 type = eSymbolTypeScopeEnd; 4004 break; 4005 4006 case N_SOL: 4007 // #included file name: name,,n_sect,0,address 4008 type = eSymbolTypeHeaderFile; 4009 4010 // We currently don't use the header files on darwin 4011 add_nlist = false; 4012 break; 4013 4014 case N_PARAMS: 4015 // compiler parameters: name,,NO_SECT,0,0 4016 type = eSymbolTypeCompiler; 4017 break; 4018 4019 case N_VERSION: 4020 // compiler version: name,,NO_SECT,0,0 4021 type = eSymbolTypeCompiler; 4022 break; 4023 4024 case N_OLEVEL: 4025 // compiler -O level: name,,NO_SECT,0,0 4026 type = eSymbolTypeCompiler; 4027 break; 4028 4029 case N_PSYM: 4030 // parameter: name,,NO_SECT,type,offset 4031 type = eSymbolTypeVariable; 4032 break; 4033 4034 case N_ENTRY: 4035 // alternate entry: name,,n_sect,linenumber,address 4036 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4037 type = eSymbolTypeLineEntry; 4038 break; 4039 4040 // Left and Right Braces 4041 case N_LBRAC: 4042 // left bracket: 0,,NO_SECT,nesting level,address We use the 4043 // current number of symbols in the symbol table in lieu of using 4044 // nlist_idx in case we ever start trimming entries out 4045 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4046 N_BRAC_indexes.push_back(sym_idx); 4047 type = eSymbolTypeScopeBegin; 4048 break; 4049 4050 case N_RBRAC: 4051 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4052 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4053 // can always skip the entire symbol if we need to navigate more 4054 // quickly at the source level when parsing STABS 4055 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4056 if (!N_BRAC_indexes.empty()) { 4057 symbol_ptr = symtab.SymbolAtIndex(N_BRAC_indexes.back()); 4058 symbol_ptr->SetByteSize(sym_idx + 1); 4059 symbol_ptr->SetSizeIsSibling(true); 4060 N_BRAC_indexes.pop_back(); 4061 } 4062 type = eSymbolTypeScopeEnd; 4063 break; 4064 4065 case N_EXCL: 4066 // deleted include file: name,,NO_SECT,0,sum 4067 type = eSymbolTypeHeaderFile; 4068 break; 4069 4070 // COMM scopes 4071 case N_BCOMM: 4072 // begin common: name,,NO_SECT,0,0 4073 // We use the current number of symbols in the symbol table in lieu 4074 // of using nlist_idx in case we ever start trimming entries out 4075 type = eSymbolTypeScopeBegin; 4076 N_COMM_indexes.push_back(sym_idx); 4077 break; 4078 4079 case N_ECOML: 4080 // end common (local name): 0,,n_sect,0,address 4081 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4082 LLVM_FALLTHROUGH; 4083 4084 case N_ECOMM: 4085 // end common: name,,n_sect,0,0 4086 // Set the size of the N_BCOMM to the terminating index of this 4087 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4088 // we need to navigate more quickly at the source level when 4089 // parsing STABS 4090 if (!N_COMM_indexes.empty()) { 4091 symbol_ptr = symtab.SymbolAtIndex(N_COMM_indexes.back()); 4092 symbol_ptr->SetByteSize(sym_idx + 1); 4093 symbol_ptr->SetSizeIsSibling(true); 4094 N_COMM_indexes.pop_back(); 4095 } 4096 type = eSymbolTypeScopeEnd; 4097 break; 4098 4099 case N_LENG: 4100 // second stab entry with length information 4101 type = eSymbolTypeAdditional; 4102 break; 4103 4104 default: 4105 break; 4106 } 4107 } else { 4108 uint8_t n_type = N_TYPE & nlist.n_type; 4109 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4110 4111 switch (n_type) { 4112 case N_INDR: { 4113 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4114 if (reexport_name_cstr && reexport_name_cstr[0]) { 4115 type = eSymbolTypeReExported; 4116 ConstString reexport_name(reexport_name_cstr + 4117 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4118 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4119 set_value = false; 4120 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4121 indirect_symbol_names.insert( 4122 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4123 } else 4124 type = eSymbolTypeUndefined; 4125 } break; 4126 4127 case N_UNDF: 4128 if (symbol_name && symbol_name[0]) { 4129 ConstString undefined_name(symbol_name + 4130 ((symbol_name[0] == '_') ? 1 : 0)); 4131 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4132 } 4133 LLVM_FALLTHROUGH; 4134 4135 case N_PBUD: 4136 type = eSymbolTypeUndefined; 4137 break; 4138 4139 case N_ABS: 4140 type = eSymbolTypeAbsolute; 4141 break; 4142 4143 case N_SECT: { 4144 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4145 4146 if (!symbol_section) { 4147 // TODO: warn about this? 4148 add_nlist = false; 4149 break; 4150 } 4151 4152 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4153 type = eSymbolTypeException; 4154 } else { 4155 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4156 4157 switch (section_type) { 4158 case S_CSTRING_LITERALS: 4159 type = eSymbolTypeData; 4160 break; // section with only literal C strings 4161 case S_4BYTE_LITERALS: 4162 type = eSymbolTypeData; 4163 break; // section with only 4 byte literals 4164 case S_8BYTE_LITERALS: 4165 type = eSymbolTypeData; 4166 break; // section with only 8 byte literals 4167 case S_LITERAL_POINTERS: 4168 type = eSymbolTypeTrampoline; 4169 break; // section with only pointers to literals 4170 case S_NON_LAZY_SYMBOL_POINTERS: 4171 type = eSymbolTypeTrampoline; 4172 break; // section with only non-lazy symbol pointers 4173 case S_LAZY_SYMBOL_POINTERS: 4174 type = eSymbolTypeTrampoline; 4175 break; // section with only lazy symbol pointers 4176 case S_SYMBOL_STUBS: 4177 type = eSymbolTypeTrampoline; 4178 break; // section with only symbol stubs, byte size of stub in 4179 // the reserved2 field 4180 case S_MOD_INIT_FUNC_POINTERS: 4181 type = eSymbolTypeCode; 4182 break; // section with only function pointers for initialization 4183 case S_MOD_TERM_FUNC_POINTERS: 4184 type = eSymbolTypeCode; 4185 break; // section with only function pointers for termination 4186 case S_INTERPOSING: 4187 type = eSymbolTypeTrampoline; 4188 break; // section with only pairs of function pointers for 4189 // interposing 4190 case S_16BYTE_LITERALS: 4191 type = eSymbolTypeData; 4192 break; // section with only 16 byte literals 4193 case S_DTRACE_DOF: 4194 type = eSymbolTypeInstrumentation; 4195 break; 4196 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4197 type = eSymbolTypeTrampoline; 4198 break; 4199 default: 4200 switch (symbol_section->GetType()) { 4201 case lldb::eSectionTypeCode: 4202 type = eSymbolTypeCode; 4203 break; 4204 case eSectionTypeData: 4205 case eSectionTypeDataCString: // Inlined C string data 4206 case eSectionTypeDataCStringPointers: // Pointers to C string 4207 // data 4208 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4209 // the symbol table 4210 case eSectionTypeData4: 4211 case eSectionTypeData8: 4212 case eSectionTypeData16: 4213 type = eSymbolTypeData; 4214 break; 4215 default: 4216 break; 4217 } 4218 break; 4219 } 4220 4221 if (type == eSymbolTypeInvalid) { 4222 const char *symbol_sect_name = 4223 symbol_section->GetName().AsCString(); 4224 if (symbol_section->IsDescendant(text_section_sp.get())) { 4225 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4226 S_ATTR_SELF_MODIFYING_CODE | 4227 S_ATTR_SOME_INSTRUCTIONS)) 4228 type = eSymbolTypeData; 4229 else 4230 type = eSymbolTypeCode; 4231 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 4232 symbol_section->IsDescendant( 4233 data_dirty_section_sp.get()) || 4234 symbol_section->IsDescendant( 4235 data_const_section_sp.get())) { 4236 if (symbol_sect_name && 4237 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 4238 type = eSymbolTypeRuntime; 4239 4240 if (symbol_name) { 4241 llvm::StringRef symbol_name_ref(symbol_name); 4242 if (symbol_name_ref.startswith("_OBJC_")) { 4243 llvm::StringRef g_objc_v2_prefix_class( 4244 "_OBJC_CLASS_$_"); 4245 llvm::StringRef g_objc_v2_prefix_metaclass( 4246 "_OBJC_METACLASS_$_"); 4247 llvm::StringRef g_objc_v2_prefix_ivar( 4248 "_OBJC_IVAR_$_"); 4249 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 4250 symbol_name_non_abi_mangled = symbol_name + 1; 4251 symbol_name = 4252 symbol_name + g_objc_v2_prefix_class.size(); 4253 type = eSymbolTypeObjCClass; 4254 demangled_is_synthesized = true; 4255 } else if (symbol_name_ref.startswith( 4256 g_objc_v2_prefix_metaclass)) { 4257 symbol_name_non_abi_mangled = symbol_name + 1; 4258 symbol_name = 4259 symbol_name + g_objc_v2_prefix_metaclass.size(); 4260 type = eSymbolTypeObjCMetaClass; 4261 demangled_is_synthesized = true; 4262 } else if (symbol_name_ref.startswith( 4263 g_objc_v2_prefix_ivar)) { 4264 symbol_name_non_abi_mangled = symbol_name + 1; 4265 symbol_name = 4266 symbol_name + g_objc_v2_prefix_ivar.size(); 4267 type = eSymbolTypeObjCIVar; 4268 demangled_is_synthesized = true; 4269 } 4270 } 4271 } 4272 } else if (symbol_sect_name && 4273 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4274 symbol_sect_name) { 4275 type = eSymbolTypeException; 4276 } else { 4277 type = eSymbolTypeData; 4278 } 4279 } else if (symbol_sect_name && 4280 ::strstr(symbol_sect_name, "__IMPORT") == 4281 symbol_sect_name) { 4282 type = eSymbolTypeTrampoline; 4283 } else if (symbol_section->IsDescendant(objc_section_sp.get())) { 4284 type = eSymbolTypeRuntime; 4285 if (symbol_name && symbol_name[0] == '.') { 4286 llvm::StringRef symbol_name_ref(symbol_name); 4287 llvm::StringRef g_objc_v1_prefix_class( 4288 ".objc_class_name_"); 4289 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4290 symbol_name_non_abi_mangled = symbol_name; 4291 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4292 type = eSymbolTypeObjCClass; 4293 demangled_is_synthesized = true; 4294 } 4295 } 4296 } 4297 } 4298 } 4299 } break; 4300 } 4301 } 4302 4303 if (!add_nlist) { 4304 sym[sym_idx].Clear(); 4305 return true; 4306 } 4307 4308 uint64_t symbol_value = nlist.n_value; 4309 4310 if (symbol_name_non_abi_mangled) { 4311 sym[sym_idx].GetMangled().SetMangledName( 4312 ConstString(symbol_name_non_abi_mangled)); 4313 sym[sym_idx].GetMangled().SetDemangledName(ConstString(symbol_name)); 4314 } else { 4315 bool symbol_name_is_mangled = false; 4316 4317 if (symbol_name && symbol_name[0] == '_') { 4318 symbol_name_is_mangled = symbol_name[1] == '_'; 4319 symbol_name++; // Skip the leading underscore 4320 } 4321 4322 if (symbol_name) { 4323 ConstString const_symbol_name(symbol_name); 4324 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4325 symbol_name_is_mangled); 4326 } 4327 } 4328 4329 if (is_gsym) { 4330 const char *gsym_name = sym[sym_idx] 4331 .GetMangled() 4332 .GetName(Mangled::ePreferMangled) 4333 .GetCString(); 4334 if (gsym_name) 4335 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4336 } 4337 4338 if (symbol_section) { 4339 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4340 if (symbol_byte_size == 0 && function_starts_count > 0) { 4341 addr_t symbol_lookup_file_addr = nlist.n_value; 4342 // Do an exact address match for non-ARM addresses, else get the 4343 // closest since the symbol might be a thumb symbol which has an 4344 // address with bit zero set. 4345 FunctionStarts::Entry *func_start_entry = 4346 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4347 if (is_arm && func_start_entry) { 4348 // Verify that the function start address is the symbol address 4349 // (ARM) or the symbol address + 1 (thumb). 4350 if (func_start_entry->addr != symbol_lookup_file_addr && 4351 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4352 // Not the right entry, NULL it out... 4353 func_start_entry = nullptr; 4354 } 4355 } 4356 if (func_start_entry) { 4357 func_start_entry->data = true; 4358 4359 addr_t symbol_file_addr = func_start_entry->addr; 4360 if (is_arm) 4361 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4362 4363 const FunctionStarts::Entry *next_func_start_entry = 4364 function_starts.FindNextEntry(func_start_entry); 4365 const addr_t section_end_file_addr = 4366 section_file_addr + symbol_section->GetByteSize(); 4367 if (next_func_start_entry) { 4368 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4369 // Be sure the clear the Thumb address bit when we calculate the 4370 // size from the current and next address 4371 if (is_arm) 4372 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4373 symbol_byte_size = std::min<lldb::addr_t>( 4374 next_symbol_file_addr - symbol_file_addr, 4375 section_end_file_addr - symbol_file_addr); 4376 } else { 4377 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4378 } 4379 } 4380 } 4381 symbol_value -= section_file_addr; 4382 } 4383 4384 if (!is_debug) { 4385 if (type == eSymbolTypeCode) { 4386 // See if we can find a N_FUN entry for any code symbols. If we do 4387 // find a match, and the name matches, then we can merge the two into 4388 // just the function symbol to avoid duplicate entries in the symbol 4389 // table. 4390 std::pair<ValueToSymbolIndexMap::const_iterator, 4391 ValueToSymbolIndexMap::const_iterator> 4392 range; 4393 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4394 if (range.first != range.second) { 4395 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4396 pos != range.second; ++pos) { 4397 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4398 sym[pos->second].GetMangled().GetName( 4399 Mangled::ePreferMangled)) { 4400 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4401 // We just need the flags from the linker symbol, so put these 4402 // flags into the N_FUN flags to avoid duplicate symbols in the 4403 // symbol table. 4404 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4405 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4406 if (resolver_addresses.find(nlist.n_value) != 4407 resolver_addresses.end()) 4408 sym[pos->second].SetType(eSymbolTypeResolver); 4409 sym[sym_idx].Clear(); 4410 return true; 4411 } 4412 } 4413 } else { 4414 if (resolver_addresses.find(nlist.n_value) != 4415 resolver_addresses.end()) 4416 type = eSymbolTypeResolver; 4417 } 4418 } else if (type == eSymbolTypeData || type == eSymbolTypeObjCClass || 4419 type == eSymbolTypeObjCMetaClass || 4420 type == eSymbolTypeObjCIVar) { 4421 // See if we can find a N_STSYM entry for any data symbols. If we do 4422 // find a match, and the name matches, then we can merge the two into 4423 // just the Static symbol to avoid duplicate entries in the symbol 4424 // table. 4425 std::pair<ValueToSymbolIndexMap::const_iterator, 4426 ValueToSymbolIndexMap::const_iterator> 4427 range; 4428 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4429 if (range.first != range.second) { 4430 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4431 pos != range.second; ++pos) { 4432 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4433 sym[pos->second].GetMangled().GetName( 4434 Mangled::ePreferMangled)) { 4435 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4436 // We just need the flags from the linker symbol, so put these 4437 // flags into the N_STSYM flags to avoid duplicate symbols in 4438 // the symbol table. 4439 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4440 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4441 sym[sym_idx].Clear(); 4442 return true; 4443 } 4444 } 4445 } else { 4446 // Combine N_GSYM stab entries with the non stab symbol. 4447 const char *gsym_name = sym[sym_idx] 4448 .GetMangled() 4449 .GetName(Mangled::ePreferMangled) 4450 .GetCString(); 4451 if (gsym_name) { 4452 ConstNameToSymbolIndexMap::const_iterator pos = 4453 N_GSYM_name_to_sym_idx.find(gsym_name); 4454 if (pos != N_GSYM_name_to_sym_idx.end()) { 4455 const uint32_t GSYM_sym_idx = pos->second; 4456 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4457 // Copy the address, because often the N_GSYM address has an 4458 // invalid address of zero when the global is a common symbol. 4459 sym[GSYM_sym_idx].GetAddressRef().SetSection(symbol_section); 4460 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4461 add_symbol_addr( 4462 sym[GSYM_sym_idx].GetAddress().GetFileAddress()); 4463 // We just need the flags from the linker symbol, so put these 4464 // flags into the N_GSYM flags to avoid duplicate symbols in 4465 // the symbol table. 4466 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4467 sym[sym_idx].Clear(); 4468 return true; 4469 } 4470 } 4471 } 4472 } 4473 } 4474 4475 sym[sym_idx].SetID(nlist_idx); 4476 sym[sym_idx].SetType(type); 4477 if (set_value) { 4478 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4479 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4480 if (symbol_section) 4481 add_symbol_addr(sym[sym_idx].GetAddress().GetFileAddress()); 4482 } 4483 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4484 if (nlist.n_desc & N_WEAK_REF) 4485 sym[sym_idx].SetIsWeak(true); 4486 4487 if (symbol_byte_size > 0) 4488 sym[sym_idx].SetByteSize(symbol_byte_size); 4489 4490 if (demangled_is_synthesized) 4491 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4492 4493 ++sym_idx; 4494 return true; 4495 }; 4496 4497 // First parse all the nlists but don't process them yet. See the next 4498 // comment for an explanation why. 4499 std::vector<struct nlist_64> nlists; 4500 nlists.reserve(symtab_load_command.nsyms); 4501 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 4502 if (auto nlist = 4503 ParseNList(nlist_data, nlist_data_offset, nlist_byte_size)) 4504 nlists.push_back(*nlist); 4505 else 4506 break; 4507 } 4508 4509 // Now parse all the debug symbols. This is needed to merge non-debug 4510 // symbols in the next step. Non-debug symbols are always coalesced into 4511 // the debug symbol. Doing this in one step would mean that some symbols 4512 // won't be merged. 4513 nlist_idx = 0; 4514 for (auto &nlist : nlists) { 4515 if (!ParseSymbolLambda(nlist, nlist_idx++, DebugSymbols)) 4516 break; 4517 } 4518 4519 // Finally parse all the non debug symbols. 4520 nlist_idx = 0; 4521 for (auto &nlist : nlists) { 4522 if (!ParseSymbolLambda(nlist, nlist_idx++, NonDebugSymbols)) 4523 break; 4524 } 4525 4526 for (const auto &pos : reexport_shlib_needs_fixup) { 4527 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4528 if (undef_pos != undefined_name_to_desc.end()) { 4529 const uint8_t dylib_ordinal = 4530 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4531 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4532 sym[pos.first].SetReExportedSymbolSharedLibrary( 4533 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4534 } 4535 } 4536 } 4537 4538 // Count how many trie symbols we'll add to the symbol table 4539 int trie_symbol_table_augment_count = 0; 4540 for (auto &e : external_sym_trie_entries) { 4541 if (symbols_added.find(e.entry.address) == symbols_added.end()) 4542 trie_symbol_table_augment_count++; 4543 } 4544 4545 if (num_syms < sym_idx + trie_symbol_table_augment_count) { 4546 num_syms = sym_idx + trie_symbol_table_augment_count; 4547 sym = symtab.Resize(num_syms); 4548 } 4549 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4550 4551 // Add symbols from the trie to the symbol table. 4552 for (auto &e : external_sym_trie_entries) { 4553 if (symbols_added.contains(e.entry.address)) 4554 continue; 4555 4556 // Find the section that this trie address is in, use that to annotate 4557 // symbol type as we add the trie address and name to the symbol table. 4558 Address symbol_addr; 4559 if (module_sp->ResolveFileAddress(e.entry.address, symbol_addr)) { 4560 SectionSP symbol_section(symbol_addr.GetSection()); 4561 const char *symbol_name = e.entry.name.GetCString(); 4562 bool demangled_is_synthesized = false; 4563 SymbolType type = 4564 GetSymbolType(symbol_name, demangled_is_synthesized, text_section_sp, 4565 data_section_sp, data_dirty_section_sp, 4566 data_const_section_sp, symbol_section); 4567 4568 sym[sym_idx].SetType(type); 4569 if (symbol_section) { 4570 sym[sym_idx].SetID(synthetic_sym_id++); 4571 sym[sym_idx].GetMangled().SetMangledName(ConstString(symbol_name)); 4572 if (demangled_is_synthesized) 4573 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4574 sym[sym_idx].SetIsSynthetic(true); 4575 sym[sym_idx].SetExternal(true); 4576 sym[sym_idx].GetAddressRef() = symbol_addr; 4577 add_symbol_addr(symbol_addr.GetFileAddress()); 4578 if (e.entry.flags & TRIE_SYMBOL_IS_THUMB) 4579 sym[sym_idx].SetFlags(MACHO_NLIST_ARM_SYMBOL_IS_THUMB); 4580 ++sym_idx; 4581 } 4582 } 4583 } 4584 4585 if (function_starts_count > 0) { 4586 uint32_t num_synthetic_function_symbols = 0; 4587 for (i = 0; i < function_starts_count; ++i) { 4588 if (symbols_added.find(function_starts.GetEntryRef(i).addr) == 4589 symbols_added.end()) 4590 ++num_synthetic_function_symbols; 4591 } 4592 4593 if (num_synthetic_function_symbols > 0) { 4594 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4595 num_syms = sym_idx + num_synthetic_function_symbols; 4596 sym = symtab.Resize(num_syms); 4597 } 4598 for (i = 0; i < function_starts_count; ++i) { 4599 const FunctionStarts::Entry *func_start_entry = 4600 function_starts.GetEntryAtIndex(i); 4601 if (symbols_added.find(func_start_entry->addr) == symbols_added.end()) { 4602 addr_t symbol_file_addr = func_start_entry->addr; 4603 uint32_t symbol_flags = 0; 4604 if (func_start_entry->data) 4605 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4606 Address symbol_addr; 4607 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4608 SectionSP symbol_section(symbol_addr.GetSection()); 4609 uint32_t symbol_byte_size = 0; 4610 if (symbol_section) { 4611 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4612 const FunctionStarts::Entry *next_func_start_entry = 4613 function_starts.FindNextEntry(func_start_entry); 4614 const addr_t section_end_file_addr = 4615 section_file_addr + symbol_section->GetByteSize(); 4616 if (next_func_start_entry) { 4617 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4618 if (is_arm) 4619 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4620 symbol_byte_size = std::min<lldb::addr_t>( 4621 next_symbol_file_addr - symbol_file_addr, 4622 section_end_file_addr - symbol_file_addr); 4623 } else { 4624 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4625 } 4626 sym[sym_idx].SetID(synthetic_sym_id++); 4627 // Don't set the name for any synthetic symbols, the Symbol 4628 // object will generate one if needed when the name is accessed 4629 // via accessors. 4630 sym[sym_idx].GetMangled().SetDemangledName(ConstString()); 4631 sym[sym_idx].SetType(eSymbolTypeCode); 4632 sym[sym_idx].SetIsSynthetic(true); 4633 sym[sym_idx].GetAddressRef() = symbol_addr; 4634 add_symbol_addr(symbol_addr.GetFileAddress()); 4635 if (symbol_flags) 4636 sym[sym_idx].SetFlags(symbol_flags); 4637 if (symbol_byte_size) 4638 sym[sym_idx].SetByteSize(symbol_byte_size); 4639 ++sym_idx; 4640 } 4641 } 4642 } 4643 } 4644 } 4645 } 4646 4647 // Trim our symbols down to just what we ended up with after removing any 4648 // symbols. 4649 if (sym_idx < num_syms) { 4650 num_syms = sym_idx; 4651 sym = symtab.Resize(num_syms); 4652 } 4653 4654 // Now synthesize indirect symbols 4655 if (m_dysymtab.nindirectsyms != 0) { 4656 if (indirect_symbol_index_data.GetByteSize()) { 4657 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4658 m_nlist_idx_to_sym_idx.end(); 4659 4660 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4661 ++sect_idx) { 4662 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4663 S_SYMBOL_STUBS) { 4664 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4665 if (symbol_stub_byte_size == 0) 4666 continue; 4667 4668 const uint32_t num_symbol_stubs = 4669 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4670 4671 if (num_symbol_stubs == 0) 4672 continue; 4673 4674 const uint32_t symbol_stub_index_offset = 4675 m_mach_sections[sect_idx].reserved1; 4676 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) { 4677 const uint32_t symbol_stub_index = 4678 symbol_stub_index_offset + stub_idx; 4679 const lldb::addr_t symbol_stub_addr = 4680 m_mach_sections[sect_idx].addr + 4681 (stub_idx * symbol_stub_byte_size); 4682 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4683 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4684 symbol_stub_offset, 4)) { 4685 const uint32_t stub_sym_id = 4686 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4687 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4688 continue; 4689 4690 NListIndexToSymbolIndexMap::const_iterator index_pos = 4691 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4692 Symbol *stub_symbol = nullptr; 4693 if (index_pos != end_index_pos) { 4694 // We have a remapping from the original nlist index to a 4695 // current symbol index, so just look this up by index 4696 stub_symbol = symtab.SymbolAtIndex(index_pos->second); 4697 } else { 4698 // We need to lookup a symbol using the original nlist symbol 4699 // index since this index is coming from the S_SYMBOL_STUBS 4700 stub_symbol = symtab.FindSymbolByID(stub_sym_id); 4701 } 4702 4703 if (stub_symbol) { 4704 Address so_addr(symbol_stub_addr, section_list); 4705 4706 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4707 // Change the external symbol into a trampoline that makes 4708 // sense These symbols were N_UNDF N_EXT, and are useless 4709 // to us, so we can re-use them so we don't have to make up 4710 // a synthetic symbol for no good reason. 4711 if (resolver_addresses.find(symbol_stub_addr) == 4712 resolver_addresses.end()) 4713 stub_symbol->SetType(eSymbolTypeTrampoline); 4714 else 4715 stub_symbol->SetType(eSymbolTypeResolver); 4716 stub_symbol->SetExternal(false); 4717 stub_symbol->GetAddressRef() = so_addr; 4718 stub_symbol->SetByteSize(symbol_stub_byte_size); 4719 } else { 4720 // Make a synthetic symbol to describe the trampoline stub 4721 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4722 if (sym_idx >= num_syms) { 4723 sym = symtab.Resize(++num_syms); 4724 stub_symbol = nullptr; // this pointer no longer valid 4725 } 4726 sym[sym_idx].SetID(synthetic_sym_id++); 4727 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4728 if (resolver_addresses.find(symbol_stub_addr) == 4729 resolver_addresses.end()) 4730 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4731 else 4732 sym[sym_idx].SetType(eSymbolTypeResolver); 4733 sym[sym_idx].SetIsSynthetic(true); 4734 sym[sym_idx].GetAddressRef() = so_addr; 4735 add_symbol_addr(so_addr.GetFileAddress()); 4736 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4737 ++sym_idx; 4738 } 4739 } else { 4740 if (log) 4741 log->Warning("symbol stub referencing symbol table symbol " 4742 "%u that isn't in our minimal symbol table, " 4743 "fix this!!!", 4744 stub_sym_id); 4745 } 4746 } 4747 } 4748 } 4749 } 4750 } 4751 } 4752 4753 if (!reexport_trie_entries.empty()) { 4754 for (const auto &e : reexport_trie_entries) { 4755 if (e.entry.import_name) { 4756 // Only add indirect symbols from the Trie entries if we didn't have 4757 // a N_INDR nlist entry for this already 4758 if (indirect_symbol_names.find(e.entry.name) == 4759 indirect_symbol_names.end()) { 4760 // Make a synthetic symbol to describe re-exported symbol. 4761 if (sym_idx >= num_syms) 4762 sym = symtab.Resize(++num_syms); 4763 sym[sym_idx].SetID(synthetic_sym_id++); 4764 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4765 sym[sym_idx].SetType(eSymbolTypeReExported); 4766 sym[sym_idx].SetIsSynthetic(true); 4767 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4768 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4769 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4770 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4771 } 4772 ++sym_idx; 4773 } 4774 } 4775 } 4776 } 4777 } 4778 4779 void ObjectFileMachO::Dump(Stream *s) { 4780 ModuleSP module_sp(GetModule()); 4781 if (module_sp) { 4782 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4783 s->Printf("%p: ", static_cast<void *>(this)); 4784 s->Indent(); 4785 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4786 s->PutCString("ObjectFileMachO64"); 4787 else 4788 s->PutCString("ObjectFileMachO32"); 4789 4790 *s << ", file = '" << m_file; 4791 ModuleSpecList all_specs; 4792 ModuleSpec base_spec; 4793 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 4794 base_spec, all_specs); 4795 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 4796 *s << "', triple"; 4797 if (e) 4798 s->Printf("[%d]", i); 4799 *s << " = "; 4800 *s << all_specs.GetModuleSpecRefAtIndex(i) 4801 .GetArchitecture() 4802 .GetTriple() 4803 .getTriple(); 4804 } 4805 *s << "\n"; 4806 SectionList *sections = GetSectionList(); 4807 if (sections) 4808 sections->Dump(s->AsRawOstream(), s->GetIndentLevel(), nullptr, true, 4809 UINT32_MAX); 4810 4811 if (m_symtab_up) 4812 m_symtab_up->Dump(s, nullptr, eSortOrderNone); 4813 } 4814 } 4815 4816 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4817 const lldb_private::DataExtractor &data, 4818 lldb::offset_t lc_offset) { 4819 uint32_t i; 4820 llvm::MachO::uuid_command load_cmd; 4821 4822 lldb::offset_t offset = lc_offset; 4823 for (i = 0; i < header.ncmds; ++i) { 4824 const lldb::offset_t cmd_offset = offset; 4825 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4826 break; 4827 4828 if (load_cmd.cmd == LC_UUID) { 4829 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4830 4831 if (uuid_bytes) { 4832 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4833 // We pretend these object files have no UUID to prevent crashing. 4834 4835 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4836 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4837 0xbb, 0x14, 0xf0, 0x0d}; 4838 4839 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4840 return UUID(); 4841 4842 return UUID::fromOptionalData(uuid_bytes, 16); 4843 } 4844 return UUID(); 4845 } 4846 offset = cmd_offset + load_cmd.cmdsize; 4847 } 4848 return UUID(); 4849 } 4850 4851 static llvm::StringRef GetOSName(uint32_t cmd) { 4852 switch (cmd) { 4853 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4854 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4855 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4856 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4857 case llvm::MachO::LC_VERSION_MIN_TVOS: 4858 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4859 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4860 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4861 default: 4862 llvm_unreachable("unexpected LC_VERSION load command"); 4863 } 4864 } 4865 4866 namespace { 4867 struct OSEnv { 4868 llvm::StringRef os_type; 4869 llvm::StringRef environment; 4870 OSEnv(uint32_t cmd) { 4871 switch (cmd) { 4872 case llvm::MachO::PLATFORM_MACOS: 4873 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4874 return; 4875 case llvm::MachO::PLATFORM_IOS: 4876 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4877 return; 4878 case llvm::MachO::PLATFORM_TVOS: 4879 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4880 return; 4881 case llvm::MachO::PLATFORM_WATCHOS: 4882 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4883 return; 4884 // TODO: add BridgeOS & DriverKit once in llvm/lib/Support/Triple.cpp 4885 // NEED_BRIDGEOS_TRIPLE 4886 // case llvm::MachO::PLATFORM_BRIDGEOS: 4887 // os_type = llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4888 // return; 4889 // case llvm::MachO::PLATFORM_DRIVERKIT: 4890 // os_type = llvm::Triple::getOSTypeName(llvm::Triple::DriverKit); 4891 // return; 4892 case llvm::MachO::PLATFORM_MACCATALYST: 4893 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4894 environment = llvm::Triple::getEnvironmentTypeName(llvm::Triple::MacABI); 4895 return; 4896 case llvm::MachO::PLATFORM_IOSSIMULATOR: 4897 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4898 environment = 4899 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4900 return; 4901 case llvm::MachO::PLATFORM_TVOSSIMULATOR: 4902 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4903 environment = 4904 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4905 return; 4906 case llvm::MachO::PLATFORM_WATCHOSSIMULATOR: 4907 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4908 environment = 4909 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4910 return; 4911 default: { 4912 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 4913 LIBLLDB_LOG_PROCESS)); 4914 LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION"); 4915 } 4916 } 4917 } 4918 }; 4919 4920 struct MinOS { 4921 uint32_t major_version, minor_version, patch_version; 4922 MinOS(uint32_t version) 4923 : major_version(version >> 16), minor_version((version >> 8) & 0xffu), 4924 patch_version(version & 0xffu) {} 4925 }; 4926 } // namespace 4927 4928 void ObjectFileMachO::GetAllArchSpecs(const llvm::MachO::mach_header &header, 4929 const lldb_private::DataExtractor &data, 4930 lldb::offset_t lc_offset, 4931 ModuleSpec &base_spec, 4932 lldb_private::ModuleSpecList &all_specs) { 4933 auto &base_arch = base_spec.GetArchitecture(); 4934 base_arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4935 if (!base_arch.IsValid()) 4936 return; 4937 4938 bool found_any = false; 4939 auto add_triple = [&](const llvm::Triple &triple) { 4940 auto spec = base_spec; 4941 spec.GetArchitecture().GetTriple() = triple; 4942 if (spec.GetArchitecture().IsValid()) { 4943 spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset); 4944 all_specs.Append(spec); 4945 found_any = true; 4946 } 4947 }; 4948 4949 // Set OS to an unspecified unknown or a "*" so it can match any OS 4950 llvm::Triple base_triple = base_arch.GetTriple(); 4951 base_triple.setOS(llvm::Triple::UnknownOS); 4952 base_triple.setOSName(llvm::StringRef()); 4953 4954 if (header.filetype == MH_PRELOAD) { 4955 if (header.cputype == CPU_TYPE_ARM) { 4956 // If this is a 32-bit arm binary, and it's a standalone binary, force 4957 // the Vendor to Apple so we don't accidentally pick up the generic 4958 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4959 // frame pointer register; most other armv7 ABIs use a combination of 4960 // r7 and r11. 4961 base_triple.setVendor(llvm::Triple::Apple); 4962 } else { 4963 // Set vendor to an unspecified unknown or a "*" so it can match any 4964 // vendor This is required for correct behavior of EFI debugging on 4965 // x86_64 4966 base_triple.setVendor(llvm::Triple::UnknownVendor); 4967 base_triple.setVendorName(llvm::StringRef()); 4968 } 4969 return add_triple(base_triple); 4970 } 4971 4972 llvm::MachO::load_command load_cmd; 4973 4974 // See if there is an LC_VERSION_MIN_* load command that can give 4975 // us the OS type. 4976 lldb::offset_t offset = lc_offset; 4977 for (uint32_t i = 0; i < header.ncmds; ++i) { 4978 const lldb::offset_t cmd_offset = offset; 4979 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4980 break; 4981 4982 llvm::MachO::version_min_command version_min; 4983 switch (load_cmd.cmd) { 4984 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4985 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4986 case llvm::MachO::LC_VERSION_MIN_TVOS: 4987 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 4988 if (load_cmd.cmdsize != sizeof(version_min)) 4989 break; 4990 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 4991 data.GetByteOrder(), &version_min) == 0) 4992 break; 4993 MinOS min_os(version_min.version); 4994 llvm::SmallString<32> os_name; 4995 llvm::raw_svector_ostream os(os_name); 4996 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 4997 << min_os.minor_version << '.' << min_os.patch_version; 4998 4999 auto triple = base_triple; 5000 triple.setOSName(os.str()); 5001 5002 // Disambiguate legacy simulator platforms. 5003 if (load_cmd.cmd != llvm::MachO::LC_VERSION_MIN_MACOSX && 5004 (base_triple.getArch() == llvm::Triple::x86_64 || 5005 base_triple.getArch() == llvm::Triple::x86)) { 5006 // The combination of legacy LC_VERSION_MIN load command and 5007 // x86 architecture always indicates a simulator environment. 5008 // The combination of LC_VERSION_MIN and arm architecture only 5009 // appears for native binaries. Back-deploying simulator 5010 // binaries on Apple Silicon Macs use the modern unambigous 5011 // LC_BUILD_VERSION load commands; no special handling required. 5012 triple.setEnvironment(llvm::Triple::Simulator); 5013 } 5014 add_triple(triple); 5015 break; 5016 } 5017 default: 5018 break; 5019 } 5020 5021 offset = cmd_offset + load_cmd.cmdsize; 5022 } 5023 5024 // See if there are LC_BUILD_VERSION load commands that can give 5025 // us the OS type. 5026 offset = lc_offset; 5027 for (uint32_t i = 0; i < header.ncmds; ++i) { 5028 const lldb::offset_t cmd_offset = offset; 5029 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 5030 break; 5031 5032 do { 5033 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 5034 llvm::MachO::build_version_command build_version; 5035 if (load_cmd.cmdsize < sizeof(build_version)) { 5036 // Malformed load command. 5037 break; 5038 } 5039 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 5040 data.GetByteOrder(), &build_version) == 0) 5041 break; 5042 MinOS min_os(build_version.minos); 5043 OSEnv os_env(build_version.platform); 5044 llvm::SmallString<16> os_name; 5045 llvm::raw_svector_ostream os(os_name); 5046 os << os_env.os_type << min_os.major_version << '.' 5047 << min_os.minor_version << '.' << min_os.patch_version; 5048 auto triple = base_triple; 5049 triple.setOSName(os.str()); 5050 os_name.clear(); 5051 if (!os_env.environment.empty()) 5052 triple.setEnvironmentName(os_env.environment); 5053 add_triple(triple); 5054 } 5055 } while (false); 5056 offset = cmd_offset + load_cmd.cmdsize; 5057 } 5058 5059 if (!found_any) { 5060 add_triple(base_triple); 5061 } 5062 } 5063 5064 ArchSpec ObjectFileMachO::GetArchitecture( 5065 ModuleSP module_sp, const llvm::MachO::mach_header &header, 5066 const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) { 5067 ModuleSpecList all_specs; 5068 ModuleSpec base_spec; 5069 GetAllArchSpecs(header, data, MachHeaderSizeFromMagic(header.magic), 5070 base_spec, all_specs); 5071 5072 // If the object file offers multiple alternative load commands, 5073 // pick the one that matches the module. 5074 if (module_sp) { 5075 const ArchSpec &module_arch = module_sp->GetArchitecture(); 5076 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 5077 ArchSpec mach_arch = 5078 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 5079 if (module_arch.IsCompatibleMatch(mach_arch)) 5080 return mach_arch; 5081 } 5082 } 5083 5084 // Return the first arch we found. 5085 if (all_specs.GetSize() == 0) 5086 return {}; 5087 return all_specs.GetModuleSpecRefAtIndex(0).GetArchitecture(); 5088 } 5089 5090 UUID ObjectFileMachO::GetUUID() { 5091 ModuleSP module_sp(GetModule()); 5092 if (module_sp) { 5093 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5094 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5095 return GetUUID(m_header, m_data, offset); 5096 } 5097 return UUID(); 5098 } 5099 5100 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5101 uint32_t count = 0; 5102 ModuleSP module_sp(GetModule()); 5103 if (module_sp) { 5104 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5105 llvm::MachO::load_command load_cmd; 5106 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5107 std::vector<std::string> rpath_paths; 5108 std::vector<std::string> rpath_relative_paths; 5109 std::vector<std::string> at_exec_relative_paths; 5110 uint32_t i; 5111 for (i = 0; i < m_header.ncmds; ++i) { 5112 const uint32_t cmd_offset = offset; 5113 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5114 break; 5115 5116 switch (load_cmd.cmd) { 5117 case LC_RPATH: 5118 case LC_LOAD_DYLIB: 5119 case LC_LOAD_WEAK_DYLIB: 5120 case LC_REEXPORT_DYLIB: 5121 case LC_LOAD_DYLINKER: 5122 case LC_LOADFVMLIB: 5123 case LC_LOAD_UPWARD_DYLIB: { 5124 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5125 const char *path = m_data.PeekCStr(name_offset); 5126 if (path) { 5127 if (load_cmd.cmd == LC_RPATH) 5128 rpath_paths.push_back(path); 5129 else { 5130 if (path[0] == '@') { 5131 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5132 rpath_relative_paths.push_back(path + strlen("@rpath")); 5133 else if (strncmp(path, "@executable_path", 5134 strlen("@executable_path")) == 0) 5135 at_exec_relative_paths.push_back(path + 5136 strlen("@executable_path")); 5137 } else { 5138 FileSpec file_spec(path); 5139 if (files.AppendIfUnique(file_spec)) 5140 count++; 5141 } 5142 } 5143 } 5144 } break; 5145 5146 default: 5147 break; 5148 } 5149 offset = cmd_offset + load_cmd.cmdsize; 5150 } 5151 5152 FileSpec this_file_spec(m_file); 5153 FileSystem::Instance().Resolve(this_file_spec); 5154 5155 if (!rpath_paths.empty()) { 5156 // Fixup all LC_RPATH values to be absolute paths 5157 std::string loader_path("@loader_path"); 5158 std::string executable_path("@executable_path"); 5159 for (auto &rpath : rpath_paths) { 5160 if (llvm::StringRef(rpath).startswith(loader_path)) { 5161 rpath.erase(0, loader_path.size()); 5162 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5163 } else if (llvm::StringRef(rpath).startswith(executable_path)) { 5164 rpath.erase(0, executable_path.size()); 5165 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5166 } 5167 } 5168 5169 for (const auto &rpath_relative_path : rpath_relative_paths) { 5170 for (const auto &rpath : rpath_paths) { 5171 std::string path = rpath; 5172 path += rpath_relative_path; 5173 // It is OK to resolve this path because we must find a file on disk 5174 // for us to accept it anyway if it is rpath relative. 5175 FileSpec file_spec(path); 5176 FileSystem::Instance().Resolve(file_spec); 5177 if (FileSystem::Instance().Exists(file_spec) && 5178 files.AppendIfUnique(file_spec)) { 5179 count++; 5180 break; 5181 } 5182 } 5183 } 5184 } 5185 5186 // We may have @executable_paths but no RPATHS. Figure those out here. 5187 // Only do this if this object file is the executable. We have no way to 5188 // get back to the actual executable otherwise, so we won't get the right 5189 // path. 5190 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5191 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5192 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5193 FileSpec file_spec = 5194 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5195 if (FileSystem::Instance().Exists(file_spec) && 5196 files.AppendIfUnique(file_spec)) 5197 count++; 5198 } 5199 } 5200 } 5201 return count; 5202 } 5203 5204 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5205 // If the object file is not an executable it can't hold the entry point. 5206 // m_entry_point_address is initialized to an invalid address, so we can just 5207 // return that. If m_entry_point_address is valid it means we've found it 5208 // already, so return the cached value. 5209 5210 if ((!IsExecutable() && !IsDynamicLoader()) || 5211 m_entry_point_address.IsValid()) { 5212 return m_entry_point_address; 5213 } 5214 5215 // Otherwise, look for the UnixThread or Thread command. The data for the 5216 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5217 // 5218 // uint32_t flavor - this is the flavor argument you would pass to 5219 // thread_get_state 5220 // uint32_t count - this is the count of longs in the thread state data 5221 // struct XXX_thread_state state - this is the structure from 5222 // <machine/thread_status.h> corresponding to the flavor. 5223 // <repeat this trio> 5224 // 5225 // So we just keep reading the various register flavors till we find the GPR 5226 // one, then read the PC out of there. 5227 // FIXME: We will need to have a "RegisterContext data provider" class at some 5228 // point that can get all the registers 5229 // out of data in this form & attach them to a given thread. That should 5230 // underlie the MacOS X User process plugin, and we'll also need it for the 5231 // MacOS X Core File process plugin. When we have that we can also use it 5232 // here. 5233 // 5234 // For now we hard-code the offsets and flavors we need: 5235 // 5236 // 5237 5238 ModuleSP module_sp(GetModule()); 5239 if (module_sp) { 5240 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5241 llvm::MachO::load_command load_cmd; 5242 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5243 uint32_t i; 5244 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5245 bool done = false; 5246 5247 for (i = 0; i < m_header.ncmds; ++i) { 5248 const lldb::offset_t cmd_offset = offset; 5249 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5250 break; 5251 5252 switch (load_cmd.cmd) { 5253 case LC_UNIXTHREAD: 5254 case LC_THREAD: { 5255 while (offset < cmd_offset + load_cmd.cmdsize) { 5256 uint32_t flavor = m_data.GetU32(&offset); 5257 uint32_t count = m_data.GetU32(&offset); 5258 if (count == 0) { 5259 // We've gotten off somehow, log and exit; 5260 return m_entry_point_address; 5261 } 5262 5263 switch (m_header.cputype) { 5264 case llvm::MachO::CPU_TYPE_ARM: 5265 if (flavor == 1 || 5266 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 5267 // from mach/arm/thread_status.h 5268 { 5269 offset += 60; // This is the offset of pc in the GPR thread state 5270 // data structure. 5271 start_address = m_data.GetU32(&offset); 5272 done = true; 5273 } 5274 break; 5275 case llvm::MachO::CPU_TYPE_ARM64: 5276 case llvm::MachO::CPU_TYPE_ARM64_32: 5277 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5278 { 5279 offset += 256; // This is the offset of pc in the GPR thread state 5280 // data structure. 5281 start_address = m_data.GetU64(&offset); 5282 done = true; 5283 } 5284 break; 5285 case llvm::MachO::CPU_TYPE_I386: 5286 if (flavor == 5287 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5288 { 5289 offset += 40; // This is the offset of eip in the GPR thread state 5290 // data structure. 5291 start_address = m_data.GetU32(&offset); 5292 done = true; 5293 } 5294 break; 5295 case llvm::MachO::CPU_TYPE_X86_64: 5296 if (flavor == 5297 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5298 { 5299 offset += 16 * 8; // This is the offset of rip in the GPR thread 5300 // state data structure. 5301 start_address = m_data.GetU64(&offset); 5302 done = true; 5303 } 5304 break; 5305 default: 5306 return m_entry_point_address; 5307 } 5308 // Haven't found the GPR flavor yet, skip over the data for this 5309 // flavor: 5310 if (done) 5311 break; 5312 offset += count * 4; 5313 } 5314 } break; 5315 case LC_MAIN: { 5316 ConstString text_segment_name("__TEXT"); 5317 uint64_t entryoffset = m_data.GetU64(&offset); 5318 SectionSP text_segment_sp = 5319 GetSectionList()->FindSectionByName(text_segment_name); 5320 if (text_segment_sp) { 5321 done = true; 5322 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5323 } 5324 } break; 5325 5326 default: 5327 break; 5328 } 5329 if (done) 5330 break; 5331 5332 // Go to the next load command: 5333 offset = cmd_offset + load_cmd.cmdsize; 5334 } 5335 5336 if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) { 5337 if (GetSymtab()) { 5338 Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType( 5339 ConstString("_dyld_start"), SymbolType::eSymbolTypeCode, 5340 Symtab::eDebugAny, Symtab::eVisibilityAny); 5341 if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) { 5342 start_address = dyld_start_sym->GetAddress().GetFileAddress(); 5343 } 5344 } 5345 } 5346 5347 if (start_address != LLDB_INVALID_ADDRESS) { 5348 // We got the start address from the load commands, so now resolve that 5349 // address in the sections of this ObjectFile: 5350 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5351 start_address, GetSectionList())) { 5352 m_entry_point_address.Clear(); 5353 } 5354 } else { 5355 // We couldn't read the UnixThread load command - maybe it wasn't there. 5356 // As a fallback look for the "start" symbol in the main executable. 5357 5358 ModuleSP module_sp(GetModule()); 5359 5360 if (module_sp) { 5361 SymbolContextList contexts; 5362 SymbolContext context; 5363 module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5364 eSymbolTypeCode, contexts); 5365 if (contexts.GetSize()) { 5366 if (contexts.GetContextAtIndex(0, context)) 5367 m_entry_point_address = context.symbol->GetAddress(); 5368 } 5369 } 5370 } 5371 } 5372 5373 return m_entry_point_address; 5374 } 5375 5376 lldb_private::Address ObjectFileMachO::GetBaseAddress() { 5377 lldb_private::Address header_addr; 5378 SectionList *section_list = GetSectionList(); 5379 if (section_list) { 5380 SectionSP text_segment_sp( 5381 section_list->FindSectionByName(GetSegmentNameTEXT())); 5382 if (text_segment_sp) { 5383 header_addr.SetSection(text_segment_sp); 5384 header_addr.SetOffset(0); 5385 } 5386 } 5387 return header_addr; 5388 } 5389 5390 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5391 ModuleSP module_sp(GetModule()); 5392 if (module_sp) { 5393 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5394 if (!m_thread_context_offsets_valid) { 5395 m_thread_context_offsets_valid = true; 5396 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5397 FileRangeArray::Entry file_range; 5398 llvm::MachO::thread_command thread_cmd; 5399 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5400 const uint32_t cmd_offset = offset; 5401 if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr) 5402 break; 5403 5404 if (thread_cmd.cmd == LC_THREAD) { 5405 file_range.SetRangeBase(offset); 5406 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5407 m_thread_context_offsets.Append(file_range); 5408 } 5409 offset = cmd_offset + thread_cmd.cmdsize; 5410 } 5411 } 5412 } 5413 return m_thread_context_offsets.GetSize(); 5414 } 5415 5416 std::string ObjectFileMachO::GetIdentifierString() { 5417 std::string result; 5418 ModuleSP module_sp(GetModule()); 5419 if (module_sp) { 5420 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5421 5422 // First, look over the load commands for an LC_NOTE load command with 5423 // data_owner string "kern ver str" & use that if found. 5424 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5425 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5426 const uint32_t cmd_offset = offset; 5427 llvm::MachO::load_command lc; 5428 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5429 break; 5430 if (lc.cmd == LC_NOTE) { 5431 char data_owner[17]; 5432 m_data.CopyData(offset, 16, data_owner); 5433 data_owner[16] = '\0'; 5434 offset += 16; 5435 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5436 uint64_t size = m_data.GetU64_unchecked(&offset); 5437 5438 // "kern ver str" has a uint32_t version and then a nul terminated 5439 // c-string. 5440 if (strcmp("kern ver str", data_owner) == 0) { 5441 offset = fileoff; 5442 uint32_t version; 5443 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5444 if (version == 1) { 5445 uint32_t strsize = size - sizeof(uint32_t); 5446 char *buf = (char *)malloc(strsize); 5447 if (buf) { 5448 m_data.CopyData(offset, strsize, buf); 5449 buf[strsize - 1] = '\0'; 5450 result = buf; 5451 if (buf) 5452 free(buf); 5453 return result; 5454 } 5455 } 5456 } 5457 } 5458 } 5459 offset = cmd_offset + lc.cmdsize; 5460 } 5461 5462 // Second, make a pass over the load commands looking for an obsolete 5463 // LC_IDENT load command. 5464 offset = MachHeaderSizeFromMagic(m_header.magic); 5465 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5466 const uint32_t cmd_offset = offset; 5467 llvm::MachO::ident_command ident_command; 5468 if (m_data.GetU32(&offset, &ident_command, 2) == nullptr) 5469 break; 5470 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5471 char *buf = (char *)malloc(ident_command.cmdsize); 5472 if (buf != nullptr && m_data.CopyData(offset, ident_command.cmdsize, 5473 buf) == ident_command.cmdsize) { 5474 buf[ident_command.cmdsize - 1] = '\0'; 5475 result = buf; 5476 } 5477 if (buf) 5478 free(buf); 5479 } 5480 offset = cmd_offset + ident_command.cmdsize; 5481 } 5482 } 5483 return result; 5484 } 5485 5486 addr_t ObjectFileMachO::GetAddressMask() { 5487 addr_t mask = 0; 5488 ModuleSP module_sp(GetModule()); 5489 if (module_sp) { 5490 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5491 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5492 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5493 const uint32_t cmd_offset = offset; 5494 llvm::MachO::load_command lc; 5495 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5496 break; 5497 if (lc.cmd == LC_NOTE) { 5498 char data_owner[17]; 5499 m_data.CopyData(offset, 16, data_owner); 5500 data_owner[16] = '\0'; 5501 offset += 16; 5502 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5503 5504 // "addrable bits" has a uint32_t version and a uint32_t 5505 // number of bits used in addressing. 5506 if (strcmp("addrable bits", data_owner) == 0) { 5507 offset = fileoff; 5508 uint32_t version; 5509 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5510 if (version == 3) { 5511 uint32_t num_addr_bits = m_data.GetU32_unchecked(&offset); 5512 if (num_addr_bits != 0) { 5513 mask = ~((1ULL << num_addr_bits) - 1); 5514 } 5515 break; 5516 } 5517 } 5518 } 5519 } 5520 offset = cmd_offset + lc.cmdsize; 5521 } 5522 } 5523 return mask; 5524 } 5525 5526 bool ObjectFileMachO::GetCorefileMainBinaryInfo(addr_t &value, 5527 bool &value_is_offset, 5528 UUID &uuid, 5529 ObjectFile::BinaryType &type) { 5530 value = LLDB_INVALID_ADDRESS; 5531 value_is_offset = false; 5532 uuid.Clear(); 5533 uint32_t log2_pagesize = 0; // not currently passed up to caller 5534 uint32_t platform = 0; // not currently passed up to caller 5535 ModuleSP module_sp(GetModule()); 5536 if (module_sp) { 5537 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5538 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5539 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5540 const uint32_t cmd_offset = offset; 5541 llvm::MachO::load_command lc; 5542 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5543 break; 5544 if (lc.cmd == LC_NOTE) { 5545 char data_owner[17]; 5546 memset(data_owner, 0, sizeof(data_owner)); 5547 m_data.CopyData(offset, 16, data_owner); 5548 offset += 16; 5549 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5550 uint64_t size = m_data.GetU64_unchecked(&offset); 5551 5552 // struct main_bin_spec 5553 // { 5554 // uint32_t version; // currently 2 5555 // uint32_t type; // 0 == unspecified, 1 == kernel, 5556 // // 2 == user process, 5557 // // 3 == standalone binary 5558 // uint64_t address; // UINT64_MAX if address not specified 5559 // uint64_t slide; // slide, UINT64_MAX if unspecified 5560 // // 0 if no slide needs to be applied to 5561 // // file address 5562 // uuid_t uuid; // all zero's if uuid not specified 5563 // uint32_t log2_pagesize; // process page size in log base 2, 5564 // // e.g. 4k pages are 12. 5565 // // 0 for unspecified 5566 // uint32_t platform; // The Mach-O platform for this corefile. 5567 // // 0 for unspecified. 5568 // // The values are defined in 5569 // // <mach-o/loader.h>, PLATFORM_*. 5570 // } __attribute((packed)); 5571 5572 // "main bin spec" (main binary specification) data payload is 5573 // formatted: 5574 // uint32_t version [currently 1] 5575 // uint32_t type [0 == unspecified, 1 == kernel, 5576 // 2 == user process, 3 == firmware ] 5577 // uint64_t address [ UINT64_MAX if address not specified ] 5578 // uuid_t uuid [ all zero's if uuid not specified ] 5579 // uint32_t log2_pagesize [ process page size in log base 5580 // 2, e.g. 4k pages are 12. 5581 // 0 for unspecified ] 5582 // uint32_t unused [ for alignment ] 5583 5584 if (strcmp("main bin spec", data_owner) == 0 && size >= 32) { 5585 offset = fileoff; 5586 uint32_t version; 5587 if (m_data.GetU32(&offset, &version, 1) != nullptr && version <= 2) { 5588 uint32_t binspec_type = 0; 5589 uuid_t raw_uuid; 5590 memset(raw_uuid, 0, sizeof(uuid_t)); 5591 5592 if (!m_data.GetU32(&offset, &binspec_type, 1)) 5593 return false; 5594 if (!m_data.GetU64(&offset, &value, 1)) 5595 return false; 5596 uint64_t slide = LLDB_INVALID_ADDRESS; 5597 if (version > 1 && !m_data.GetU64(&offset, &slide, 1)) 5598 return false; 5599 if (value == LLDB_INVALID_ADDRESS && 5600 slide != LLDB_INVALID_ADDRESS) { 5601 value = slide; 5602 value_is_offset = true; 5603 } 5604 5605 if (m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5606 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5607 // convert the "main bin spec" type into our 5608 // ObjectFile::BinaryType enum 5609 switch (binspec_type) { 5610 case 0: 5611 type = eBinaryTypeUnknown; 5612 break; 5613 case 1: 5614 type = eBinaryTypeKernel; 5615 break; 5616 case 2: 5617 type = eBinaryTypeUser; 5618 break; 5619 case 3: 5620 type = eBinaryTypeStandalone; 5621 break; 5622 } 5623 if (!m_data.GetU32(&offset, &log2_pagesize, 1)) 5624 return false; 5625 if (version > 1 && !m_data.GetU32(&offset, &platform, 1)) 5626 return false; 5627 return true; 5628 } 5629 } 5630 } 5631 } 5632 offset = cmd_offset + lc.cmdsize; 5633 } 5634 } 5635 return false; 5636 } 5637 5638 lldb::RegisterContextSP 5639 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5640 lldb_private::Thread &thread) { 5641 lldb::RegisterContextSP reg_ctx_sp; 5642 5643 ModuleSP module_sp(GetModule()); 5644 if (module_sp) { 5645 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5646 if (!m_thread_context_offsets_valid) 5647 GetNumThreadContexts(); 5648 5649 const FileRangeArray::Entry *thread_context_file_range = 5650 m_thread_context_offsets.GetEntryAtIndex(idx); 5651 if (thread_context_file_range) { 5652 5653 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5654 thread_context_file_range->GetByteSize()); 5655 5656 switch (m_header.cputype) { 5657 case llvm::MachO::CPU_TYPE_ARM64: 5658 case llvm::MachO::CPU_TYPE_ARM64_32: 5659 reg_ctx_sp = 5660 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data); 5661 break; 5662 5663 case llvm::MachO::CPU_TYPE_ARM: 5664 reg_ctx_sp = 5665 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data); 5666 break; 5667 5668 case llvm::MachO::CPU_TYPE_I386: 5669 reg_ctx_sp = 5670 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data); 5671 break; 5672 5673 case llvm::MachO::CPU_TYPE_X86_64: 5674 reg_ctx_sp = 5675 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data); 5676 break; 5677 } 5678 } 5679 } 5680 return reg_ctx_sp; 5681 } 5682 5683 ObjectFile::Type ObjectFileMachO::CalculateType() { 5684 switch (m_header.filetype) { 5685 case MH_OBJECT: // 0x1u 5686 if (GetAddressByteSize() == 4) { 5687 // 32 bit kexts are just object files, but they do have a valid 5688 // UUID load command. 5689 if (GetUUID()) { 5690 // this checking for the UUID load command is not enough we could 5691 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5692 // this is required of kexts 5693 if (m_strata == eStrataInvalid) 5694 m_strata = eStrataKernel; 5695 return eTypeSharedLibrary; 5696 } 5697 } 5698 return eTypeObjectFile; 5699 5700 case MH_EXECUTE: 5701 return eTypeExecutable; // 0x2u 5702 case MH_FVMLIB: 5703 return eTypeSharedLibrary; // 0x3u 5704 case MH_CORE: 5705 return eTypeCoreFile; // 0x4u 5706 case MH_PRELOAD: 5707 return eTypeSharedLibrary; // 0x5u 5708 case MH_DYLIB: 5709 return eTypeSharedLibrary; // 0x6u 5710 case MH_DYLINKER: 5711 return eTypeDynamicLinker; // 0x7u 5712 case MH_BUNDLE: 5713 return eTypeSharedLibrary; // 0x8u 5714 case MH_DYLIB_STUB: 5715 return eTypeStubLibrary; // 0x9u 5716 case MH_DSYM: 5717 return eTypeDebugInfo; // 0xAu 5718 case MH_KEXT_BUNDLE: 5719 return eTypeSharedLibrary; // 0xBu 5720 default: 5721 break; 5722 } 5723 return eTypeUnknown; 5724 } 5725 5726 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5727 switch (m_header.filetype) { 5728 case MH_OBJECT: // 0x1u 5729 { 5730 // 32 bit kexts are just object files, but they do have a valid 5731 // UUID load command. 5732 if (GetUUID()) { 5733 // this checking for the UUID load command is not enough we could 5734 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5735 // this is required of kexts 5736 if (m_type == eTypeInvalid) 5737 m_type = eTypeSharedLibrary; 5738 5739 return eStrataKernel; 5740 } 5741 } 5742 return eStrataUnknown; 5743 5744 case MH_EXECUTE: // 0x2u 5745 // Check for the MH_DYLDLINK bit in the flags 5746 if (m_header.flags & MH_DYLDLINK) { 5747 return eStrataUser; 5748 } else { 5749 SectionList *section_list = GetSectionList(); 5750 if (section_list) { 5751 static ConstString g_kld_section_name("__KLD"); 5752 if (section_list->FindSectionByName(g_kld_section_name)) 5753 return eStrataKernel; 5754 } 5755 } 5756 return eStrataRawImage; 5757 5758 case MH_FVMLIB: 5759 return eStrataUser; // 0x3u 5760 case MH_CORE: 5761 return eStrataUnknown; // 0x4u 5762 case MH_PRELOAD: 5763 return eStrataRawImage; // 0x5u 5764 case MH_DYLIB: 5765 return eStrataUser; // 0x6u 5766 case MH_DYLINKER: 5767 return eStrataUser; // 0x7u 5768 case MH_BUNDLE: 5769 return eStrataUser; // 0x8u 5770 case MH_DYLIB_STUB: 5771 return eStrataUser; // 0x9u 5772 case MH_DSYM: 5773 return eStrataUnknown; // 0xAu 5774 case MH_KEXT_BUNDLE: 5775 return eStrataKernel; // 0xBu 5776 default: 5777 break; 5778 } 5779 return eStrataUnknown; 5780 } 5781 5782 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5783 ModuleSP module_sp(GetModule()); 5784 if (module_sp) { 5785 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5786 llvm::MachO::dylib_command load_cmd; 5787 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5788 uint32_t version_cmd = 0; 5789 uint64_t version = 0; 5790 uint32_t i; 5791 for (i = 0; i < m_header.ncmds; ++i) { 5792 const lldb::offset_t cmd_offset = offset; 5793 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5794 break; 5795 5796 if (load_cmd.cmd == LC_ID_DYLIB) { 5797 if (version_cmd == 0) { 5798 version_cmd = load_cmd.cmd; 5799 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr) 5800 break; 5801 version = load_cmd.dylib.current_version; 5802 } 5803 break; // Break for now unless there is another more complete version 5804 // number load command in the future. 5805 } 5806 offset = cmd_offset + load_cmd.cmdsize; 5807 } 5808 5809 if (version_cmd == LC_ID_DYLIB) { 5810 unsigned major = (version & 0xFFFF0000ull) >> 16; 5811 unsigned minor = (version & 0x0000FF00ull) >> 8; 5812 unsigned subminor = (version & 0x000000FFull); 5813 return llvm::VersionTuple(major, minor, subminor); 5814 } 5815 } 5816 return llvm::VersionTuple(); 5817 } 5818 5819 ArchSpec ObjectFileMachO::GetArchitecture() { 5820 ModuleSP module_sp(GetModule()); 5821 ArchSpec arch; 5822 if (module_sp) { 5823 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5824 5825 return GetArchitecture(module_sp, m_header, m_data, 5826 MachHeaderSizeFromMagic(m_header.magic)); 5827 } 5828 return arch; 5829 } 5830 5831 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, 5832 addr_t &base_addr, UUID &uuid) { 5833 uuid.Clear(); 5834 base_addr = LLDB_INVALID_ADDRESS; 5835 if (process && process->GetDynamicLoader()) { 5836 DynamicLoader *dl = process->GetDynamicLoader(); 5837 LazyBool using_shared_cache; 5838 LazyBool private_shared_cache; 5839 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5840 private_shared_cache); 5841 } 5842 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5843 LIBLLDB_LOG_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(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5938 LIBLLDB_LOG_PROCESS)); 5939 if (log && uuid.IsValid()) 5940 LLDB_LOGF(log, 5941 "lldb's in-memory shared cache has a UUID of %s base address of " 5942 "0x%" PRIx64, 5943 uuid.GetAsString().c_str(), base_addr); 5944 #endif 5945 } 5946 5947 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5948 if (!m_min_os_version) { 5949 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5950 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5951 const lldb::offset_t load_cmd_offset = offset; 5952 5953 llvm::MachO::version_min_command lc; 5954 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5955 break; 5956 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5957 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5958 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5959 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5960 if (m_data.GetU32(&offset, &lc.version, 5961 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5962 const uint32_t xxxx = lc.version >> 16; 5963 const uint32_t yy = (lc.version >> 8) & 0xffu; 5964 const uint32_t zz = lc.version & 0xffu; 5965 if (xxxx) { 5966 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5967 break; 5968 } 5969 } 5970 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5971 // struct build_version_command { 5972 // uint32_t cmd; /* LC_BUILD_VERSION */ 5973 // uint32_t cmdsize; /* sizeof(struct 5974 // build_version_command) plus */ 5975 // /* ntools * sizeof(struct 5976 // build_tool_version) */ 5977 // uint32_t platform; /* platform */ 5978 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5979 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded in 5980 // nibbles xxxx.yy.zz */ uint32_t ntools; /* number of 5981 // tool entries following this */ 5982 // }; 5983 5984 offset += 4; // skip platform 5985 uint32_t minos = m_data.GetU32(&offset); 5986 5987 const uint32_t xxxx = minos >> 16; 5988 const uint32_t yy = (minos >> 8) & 0xffu; 5989 const uint32_t zz = minos & 0xffu; 5990 if (xxxx) { 5991 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5992 break; 5993 } 5994 } 5995 5996 offset = load_cmd_offset + lc.cmdsize; 5997 } 5998 5999 if (!m_min_os_version) { 6000 // Set version to an empty value so we don't keep trying to 6001 m_min_os_version = llvm::VersionTuple(); 6002 } 6003 } 6004 6005 return *m_min_os_version; 6006 } 6007 6008 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { 6009 if (!m_sdk_versions.hasValue()) { 6010 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 6011 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6012 const lldb::offset_t load_cmd_offset = offset; 6013 6014 llvm::MachO::version_min_command lc; 6015 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6016 break; 6017 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 6018 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 6019 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 6020 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 6021 if (m_data.GetU32(&offset, &lc.version, 6022 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 6023 const uint32_t xxxx = lc.sdk >> 16; 6024 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 6025 const uint32_t zz = lc.sdk & 0xffu; 6026 if (xxxx) { 6027 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 6028 break; 6029 } else { 6030 GetModule()->ReportWarning("minimum OS version load command with " 6031 "invalid (0) version found."); 6032 } 6033 } 6034 } 6035 offset = load_cmd_offset + lc.cmdsize; 6036 } 6037 6038 if (!m_sdk_versions.hasValue()) { 6039 offset = MachHeaderSizeFromMagic(m_header.magic); 6040 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6041 const lldb::offset_t load_cmd_offset = offset; 6042 6043 llvm::MachO::version_min_command lc; 6044 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6045 break; 6046 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 6047 // struct build_version_command { 6048 // uint32_t cmd; /* LC_BUILD_VERSION */ 6049 // uint32_t cmdsize; /* sizeof(struct 6050 // build_version_command) plus */ 6051 // /* ntools * sizeof(struct 6052 // build_tool_version) */ 6053 // uint32_t platform; /* platform */ 6054 // uint32_t minos; /* X.Y.Z is encoded in nibbles 6055 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded 6056 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number 6057 // of tool entries following this */ 6058 // }; 6059 6060 offset += 4; // skip platform 6061 uint32_t minos = m_data.GetU32(&offset); 6062 6063 const uint32_t xxxx = minos >> 16; 6064 const uint32_t yy = (minos >> 8) & 0xffu; 6065 const uint32_t zz = minos & 0xffu; 6066 if (xxxx) { 6067 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 6068 break; 6069 } 6070 } 6071 offset = load_cmd_offset + lc.cmdsize; 6072 } 6073 } 6074 6075 if (!m_sdk_versions.hasValue()) 6076 m_sdk_versions = llvm::VersionTuple(); 6077 } 6078 6079 return m_sdk_versions.getValue(); 6080 } 6081 6082 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 6083 return m_header.filetype == llvm::MachO::MH_DYLINKER; 6084 } 6085 6086 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 6087 return m_allow_assembly_emulation_unwind_plans; 6088 } 6089 6090 Section *ObjectFileMachO::GetMachHeaderSection() { 6091 // Find the first address of the mach header which is the first non-zero file 6092 // sized section whose file offset is zero. This is the base file address of 6093 // the mach-o file which can be subtracted from the vmaddr of the other 6094 // segments found in memory and added to the load address 6095 ModuleSP module_sp = GetModule(); 6096 if (!module_sp) 6097 return nullptr; 6098 SectionList *section_list = GetSectionList(); 6099 if (!section_list) 6100 return nullptr; 6101 const size_t num_sections = section_list->GetSize(); 6102 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6103 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 6104 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 6105 return section; 6106 } 6107 return nullptr; 6108 } 6109 6110 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 6111 if (!section) 6112 return false; 6113 const bool is_dsym = (m_header.filetype == MH_DSYM); 6114 if (section->GetFileSize() == 0 && !is_dsym) 6115 return false; 6116 if (section->IsThreadSpecific()) 6117 return false; 6118 if (GetModule().get() != section->GetModule().get()) 6119 return false; 6120 // Be careful with __LINKEDIT and __DWARF segments 6121 if (section->GetName() == GetSegmentNameLINKEDIT() || 6122 section->GetName() == GetSegmentNameDWARF()) { 6123 // Only map __LINKEDIT and __DWARF if we have an in memory image and 6124 // this isn't a kernel binary like a kext or mach_kernel. 6125 const bool is_memory_image = (bool)m_process_wp.lock(); 6126 const Strata strata = GetStrata(); 6127 if (is_memory_image == false || strata == eStrataKernel) 6128 return false; 6129 } 6130 return true; 6131 } 6132 6133 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 6134 lldb::addr_t header_load_address, const Section *header_section, 6135 const Section *section) { 6136 ModuleSP module_sp = GetModule(); 6137 if (module_sp && header_section && section && 6138 header_load_address != LLDB_INVALID_ADDRESS) { 6139 lldb::addr_t file_addr = header_section->GetFileAddress(); 6140 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 6141 return section->GetFileAddress() - file_addr + header_load_address; 6142 } 6143 return LLDB_INVALID_ADDRESS; 6144 } 6145 6146 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 6147 bool value_is_offset) { 6148 ModuleSP module_sp = GetModule(); 6149 if (!module_sp) 6150 return false; 6151 6152 SectionList *section_list = GetSectionList(); 6153 if (!section_list) 6154 return false; 6155 6156 size_t num_loaded_sections = 0; 6157 const size_t num_sections = section_list->GetSize(); 6158 6159 if (value_is_offset) { 6160 // "value" is an offset to apply to each top level segment 6161 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6162 // Iterate through the object file sections to find all of the 6163 // sections that size on disk (to avoid __PAGEZERO) and load them 6164 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6165 if (SectionIsLoadable(section_sp.get())) 6166 if (target.GetSectionLoadList().SetSectionLoadAddress( 6167 section_sp, section_sp->GetFileAddress() + value)) 6168 ++num_loaded_sections; 6169 } 6170 } else { 6171 // "value" is the new base address of the mach_header, adjust each 6172 // section accordingly 6173 6174 Section *mach_header_section = GetMachHeaderSection(); 6175 if (mach_header_section) { 6176 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6177 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6178 6179 lldb::addr_t section_load_addr = 6180 CalculateSectionLoadAddressForMemoryImage( 6181 value, mach_header_section, section_sp.get()); 6182 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6183 if (target.GetSectionLoadList().SetSectionLoadAddress( 6184 section_sp, section_load_addr)) 6185 ++num_loaded_sections; 6186 } 6187 } 6188 } 6189 } 6190 return num_loaded_sections > 0; 6191 } 6192 6193 struct all_image_infos_header { 6194 uint32_t version; // currently 1 6195 uint32_t imgcount; // number of binary images 6196 uint64_t entries_fileoff; // file offset in the corefile of where the array of 6197 // struct entry's begin. 6198 uint32_t entries_size; // size of 'struct entry'. 6199 uint32_t unused; 6200 }; 6201 6202 struct image_entry { 6203 uint64_t filepath_offset; // offset in corefile to c-string of the file path, 6204 // UINT64_MAX if unavailable. 6205 uuid_t uuid; // uint8_t[16]. should be set to all zeroes if 6206 // uuid is unknown. 6207 uint64_t load_address; // UINT64_MAX if unknown. 6208 uint64_t seg_addrs_offset; // offset to the array of struct segment_vmaddr's. 6209 uint32_t segment_count; // The number of segments for this binary. 6210 uint32_t unused; 6211 6212 image_entry() { 6213 filepath_offset = UINT64_MAX; 6214 memset(&uuid, 0, sizeof(uuid_t)); 6215 segment_count = 0; 6216 load_address = UINT64_MAX; 6217 seg_addrs_offset = UINT64_MAX; 6218 unused = 0; 6219 } 6220 image_entry(const image_entry &rhs) { 6221 filepath_offset = rhs.filepath_offset; 6222 memcpy(&uuid, &rhs.uuid, sizeof(uuid_t)); 6223 segment_count = rhs.segment_count; 6224 seg_addrs_offset = rhs.seg_addrs_offset; 6225 load_address = rhs.load_address; 6226 unused = rhs.unused; 6227 } 6228 }; 6229 6230 struct segment_vmaddr { 6231 char segname[16]; 6232 uint64_t vmaddr; 6233 uint64_t unused; 6234 6235 segment_vmaddr() { 6236 memset(&segname, 0, 16); 6237 vmaddr = UINT64_MAX; 6238 unused = 0; 6239 } 6240 segment_vmaddr(const segment_vmaddr &rhs) { 6241 memcpy(&segname, &rhs.segname, 16); 6242 vmaddr = rhs.vmaddr; 6243 unused = rhs.unused; 6244 } 6245 }; 6246 6247 // Write the payload for the "all image infos" LC_NOTE into 6248 // the supplied all_image_infos_payload, assuming that this 6249 // will be written into the corefile starting at 6250 // initial_file_offset. 6251 // 6252 // The placement of this payload is a little tricky. We're 6253 // laying this out as 6254 // 6255 // 1. header (struct all_image_info_header) 6256 // 2. Array of fixed-size (struct image_entry)'s, one 6257 // per binary image present in the process. 6258 // 3. Arrays of (struct segment_vmaddr)'s, a varying number 6259 // for each binary image. 6260 // 4. Variable length c-strings of binary image filepaths, 6261 // one per binary. 6262 // 6263 // To compute where everything will be laid out in the 6264 // payload, we need to iterate over the images and calculate 6265 // how many segment_vmaddr structures each image will need, 6266 // and how long each image's filepath c-string is. There 6267 // are some multiple passes over the image list while calculating 6268 // everything. 6269 6270 static offset_t CreateAllImageInfosPayload( 6271 const lldb::ProcessSP &process_sp, offset_t initial_file_offset, 6272 StreamString &all_image_infos_payload, SaveCoreStyle core_style) { 6273 Target &target = process_sp->GetTarget(); 6274 ModuleList modules = target.GetImages(); 6275 6276 // stack-only corefiles have no reason to include binaries that 6277 // are not executing; we're trying to make the smallest corefile 6278 // we can, so leave the rest out. 6279 if (core_style == SaveCoreStyle::eSaveCoreStackOnly) 6280 modules.Clear(); 6281 6282 std::set<std::string> executing_uuids; 6283 ThreadList &thread_list(process_sp->GetThreadList()); 6284 for (uint32_t i = 0; i < thread_list.GetSize(); i++) { 6285 ThreadSP thread_sp = thread_list.GetThreadAtIndex(i); 6286 uint32_t stack_frame_count = thread_sp->GetStackFrameCount(); 6287 for (uint32_t j = 0; j < stack_frame_count; j++) { 6288 StackFrameSP stack_frame_sp = thread_sp->GetStackFrameAtIndex(j); 6289 Address pc = stack_frame_sp->GetFrameCodeAddress(); 6290 ModuleSP module_sp = pc.GetModule(); 6291 if (module_sp) { 6292 UUID uuid = module_sp->GetUUID(); 6293 if (uuid.IsValid()) { 6294 executing_uuids.insert(uuid.GetAsString()); 6295 modules.AppendIfNeeded(module_sp); 6296 } 6297 } 6298 } 6299 } 6300 size_t modules_count = modules.GetSize(); 6301 6302 struct all_image_infos_header infos; 6303 infos.version = 1; 6304 infos.imgcount = modules_count; 6305 infos.entries_size = sizeof(image_entry); 6306 infos.entries_fileoff = initial_file_offset + sizeof(all_image_infos_header); 6307 infos.unused = 0; 6308 6309 all_image_infos_payload.PutHex32(infos.version); 6310 all_image_infos_payload.PutHex32(infos.imgcount); 6311 all_image_infos_payload.PutHex64(infos.entries_fileoff); 6312 all_image_infos_payload.PutHex32(infos.entries_size); 6313 all_image_infos_payload.PutHex32(infos.unused); 6314 6315 // First create the structures for all of the segment name+vmaddr vectors 6316 // for each module, so we will know the size of them as we add the 6317 // module entries. 6318 std::vector<std::vector<segment_vmaddr>> modules_segment_vmaddrs; 6319 for (size_t i = 0; i < modules_count; i++) { 6320 ModuleSP module = modules.GetModuleAtIndex(i); 6321 6322 SectionList *sections = module->GetSectionList(); 6323 size_t sections_count = sections->GetSize(); 6324 std::vector<segment_vmaddr> segment_vmaddrs; 6325 for (size_t j = 0; j < sections_count; j++) { 6326 SectionSP section = sections->GetSectionAtIndex(j); 6327 if (!section->GetParent().get()) { 6328 addr_t vmaddr = section->GetLoadBaseAddress(&target); 6329 if (vmaddr == LLDB_INVALID_ADDRESS) 6330 continue; 6331 ConstString name = section->GetName(); 6332 segment_vmaddr seg_vmaddr; 6333 strncpy(seg_vmaddr.segname, name.AsCString(), 6334 sizeof(seg_vmaddr.segname)); 6335 seg_vmaddr.vmaddr = vmaddr; 6336 seg_vmaddr.unused = 0; 6337 segment_vmaddrs.push_back(seg_vmaddr); 6338 } 6339 } 6340 modules_segment_vmaddrs.push_back(segment_vmaddrs); 6341 } 6342 6343 offset_t size_of_vmaddr_structs = 0; 6344 for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) { 6345 size_of_vmaddr_structs += 6346 modules_segment_vmaddrs[i].size() * sizeof(segment_vmaddr); 6347 } 6348 6349 offset_t size_of_filepath_cstrings = 0; 6350 for (size_t i = 0; i < modules_count; i++) { 6351 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6352 size_of_filepath_cstrings += module_sp->GetFileSpec().GetPath().size() + 1; 6353 } 6354 6355 // Calculate the file offsets of our "all image infos" payload in the 6356 // corefile. initial_file_offset the original value passed in to this method. 6357 6358 offset_t start_of_entries = 6359 initial_file_offset + sizeof(all_image_infos_header); 6360 offset_t start_of_seg_vmaddrs = 6361 start_of_entries + sizeof(image_entry) * modules_count; 6362 offset_t start_of_filenames = start_of_seg_vmaddrs + size_of_vmaddr_structs; 6363 6364 offset_t final_file_offset = start_of_filenames + size_of_filepath_cstrings; 6365 6366 // Now write the one-per-module 'struct image_entry' into the 6367 // StringStream; keep track of where the struct segment_vmaddr 6368 // entries for each module will end up in the corefile. 6369 6370 offset_t current_string_offset = start_of_filenames; 6371 offset_t current_segaddrs_offset = start_of_seg_vmaddrs; 6372 std::vector<struct image_entry> image_entries; 6373 for (size_t i = 0; i < modules_count; i++) { 6374 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6375 6376 struct image_entry ent; 6377 memcpy(&ent.uuid, module_sp->GetUUID().GetBytes().data(), sizeof(ent.uuid)); 6378 if (modules_segment_vmaddrs[i].size() > 0) { 6379 ent.segment_count = modules_segment_vmaddrs[i].size(); 6380 ent.seg_addrs_offset = current_segaddrs_offset; 6381 } 6382 ent.filepath_offset = current_string_offset; 6383 ObjectFile *objfile = module_sp->GetObjectFile(); 6384 if (objfile) { 6385 Address base_addr(objfile->GetBaseAddress()); 6386 if (base_addr.IsValid()) { 6387 ent.load_address = base_addr.GetLoadAddress(&target); 6388 } 6389 } 6390 6391 all_image_infos_payload.PutHex64(ent.filepath_offset); 6392 all_image_infos_payload.PutRawBytes(ent.uuid, sizeof(ent.uuid)); 6393 all_image_infos_payload.PutHex64(ent.load_address); 6394 all_image_infos_payload.PutHex64(ent.seg_addrs_offset); 6395 all_image_infos_payload.PutHex32(ent.segment_count); 6396 6397 if (executing_uuids.find(module_sp->GetUUID().GetAsString()) != 6398 executing_uuids.end()) 6399 all_image_infos_payload.PutHex32(1); 6400 else 6401 all_image_infos_payload.PutHex32(0); 6402 6403 current_segaddrs_offset += ent.segment_count * sizeof(segment_vmaddr); 6404 current_string_offset += module_sp->GetFileSpec().GetPath().size() + 1; 6405 } 6406 6407 // Now write the struct segment_vmaddr entries into the StringStream. 6408 6409 for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) { 6410 if (modules_segment_vmaddrs[i].size() == 0) 6411 continue; 6412 for (struct segment_vmaddr segvm : modules_segment_vmaddrs[i]) { 6413 all_image_infos_payload.PutRawBytes(segvm.segname, sizeof(segvm.segname)); 6414 all_image_infos_payload.PutHex64(segvm.vmaddr); 6415 all_image_infos_payload.PutHex64(segvm.unused); 6416 } 6417 } 6418 6419 for (size_t i = 0; i < modules_count; i++) { 6420 ModuleSP module_sp = modules.GetModuleAtIndex(i); 6421 std::string filepath = module_sp->GetFileSpec().GetPath(); 6422 all_image_infos_payload.PutRawBytes(filepath.data(), filepath.size() + 1); 6423 } 6424 6425 return final_file_offset; 6426 } 6427 6428 // Temp struct used to combine contiguous memory regions with 6429 // identical permissions. 6430 struct page_object { 6431 addr_t addr; 6432 addr_t size; 6433 uint32_t prot; 6434 }; 6435 6436 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6437 const FileSpec &outfile, 6438 lldb::SaveCoreStyle &core_style, Status &error) { 6439 if (!process_sp) 6440 return false; 6441 6442 // Default on macOS is to create a dirty-memory-only corefile. 6443 if (core_style == SaveCoreStyle::eSaveCoreUnspecified) { 6444 core_style = SaveCoreStyle::eSaveCoreDirtyOnly; 6445 } 6446 6447 Target &target = process_sp->GetTarget(); 6448 const ArchSpec target_arch = target.GetArchitecture(); 6449 const llvm::Triple &target_triple = target_arch.GetTriple(); 6450 if (target_triple.getVendor() == llvm::Triple::Apple && 6451 (target_triple.getOS() == llvm::Triple::MacOSX || 6452 target_triple.getOS() == llvm::Triple::IOS || 6453 target_triple.getOS() == llvm::Triple::WatchOS || 6454 target_triple.getOS() == llvm::Triple::TvOS)) { 6455 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) 6456 // { 6457 bool make_core = false; 6458 switch (target_arch.GetMachine()) { 6459 case llvm::Triple::aarch64: 6460 case llvm::Triple::aarch64_32: 6461 case llvm::Triple::arm: 6462 case llvm::Triple::thumb: 6463 case llvm::Triple::x86: 6464 case llvm::Triple::x86_64: 6465 make_core = true; 6466 break; 6467 default: 6468 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6469 target_triple.str().c_str()); 6470 break; 6471 } 6472 6473 if (make_core) { 6474 std::vector<llvm::MachO::segment_command_64> segment_load_commands; 6475 // uint32_t range_info_idx = 0; 6476 MemoryRegionInfo range_info; 6477 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6478 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6479 const ByteOrder byte_order = target_arch.GetByteOrder(); 6480 std::vector<page_object> pages_to_copy; 6481 6482 if (range_error.Success()) { 6483 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6484 // Calculate correct protections 6485 uint32_t prot = 0; 6486 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6487 prot |= VM_PROT_READ; 6488 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6489 prot |= VM_PROT_WRITE; 6490 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6491 prot |= VM_PROT_EXECUTE; 6492 6493 const addr_t addr = range_info.GetRange().GetRangeBase(); 6494 const addr_t size = range_info.GetRange().GetByteSize(); 6495 6496 if (size == 0) 6497 break; 6498 6499 bool include_this_region = true; 6500 bool dirty_pages_only = false; 6501 if (core_style == SaveCoreStyle::eSaveCoreStackOnly) { 6502 dirty_pages_only = true; 6503 if (range_info.IsStackMemory() != MemoryRegionInfo::eYes) { 6504 include_this_region = false; 6505 } 6506 } 6507 if (core_style == SaveCoreStyle::eSaveCoreDirtyOnly) { 6508 dirty_pages_only = true; 6509 } 6510 6511 if (prot != 0 && include_this_region) { 6512 addr_t pagesize = range_info.GetPageSize(); 6513 const llvm::Optional<std::vector<addr_t>> &dirty_page_list = 6514 range_info.GetDirtyPageList(); 6515 if (dirty_pages_only && dirty_page_list.hasValue()) { 6516 for (addr_t dirtypage : dirty_page_list.getValue()) { 6517 page_object obj; 6518 obj.addr = dirtypage; 6519 obj.size = pagesize; 6520 obj.prot = prot; 6521 pages_to_copy.push_back(obj); 6522 } 6523 } else { 6524 page_object obj; 6525 obj.addr = addr; 6526 obj.size = size; 6527 obj.prot = prot; 6528 pages_to_copy.push_back(obj); 6529 } 6530 } 6531 6532 range_error = process_sp->GetMemoryRegionInfo( 6533 range_info.GetRange().GetRangeEnd(), range_info); 6534 if (range_error.Fail()) 6535 break; 6536 } 6537 6538 // Combine contiguous entries that have the same 6539 // protections so we don't have an excess of 6540 // load commands. 6541 std::vector<page_object> combined_page_objects; 6542 page_object last_obj; 6543 last_obj.addr = LLDB_INVALID_ADDRESS; 6544 last_obj.size = 0; 6545 for (page_object obj : pages_to_copy) { 6546 if (last_obj.addr == LLDB_INVALID_ADDRESS) { 6547 last_obj = obj; 6548 continue; 6549 } 6550 if (last_obj.addr + last_obj.size == obj.addr && 6551 last_obj.prot == obj.prot) { 6552 last_obj.size += obj.size; 6553 continue; 6554 } 6555 combined_page_objects.push_back(last_obj); 6556 last_obj = obj; 6557 } 6558 // Add the last entry we were looking to combine 6559 // on to the array. 6560 if (last_obj.addr != LLDB_INVALID_ADDRESS && last_obj.size != 0) 6561 combined_page_objects.push_back(last_obj); 6562 6563 for (page_object obj : combined_page_objects) { 6564 uint32_t cmd_type = LC_SEGMENT_64; 6565 uint32_t segment_size = sizeof(llvm::MachO::segment_command_64); 6566 if (addr_byte_size == 4) { 6567 cmd_type = LC_SEGMENT; 6568 segment_size = sizeof(llvm::MachO::segment_command); 6569 } 6570 llvm::MachO::segment_command_64 segment = { 6571 cmd_type, // uint32_t cmd; 6572 segment_size, // uint32_t cmdsize; 6573 {0}, // char segname[16]; 6574 obj.addr, // uint64_t vmaddr; // uint32_t for 32-bit 6575 // Mach-O 6576 obj.size, // uint64_t vmsize; // uint32_t for 32-bit 6577 // Mach-O 6578 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6579 obj.size, // uint64_t filesize; // uint32_t for 32-bit 6580 // Mach-O 6581 obj.prot, // uint32_t maxprot; 6582 obj.prot, // uint32_t initprot; 6583 0, // uint32_t nsects; 6584 0}; // uint32_t flags; 6585 segment_load_commands.push_back(segment); 6586 } 6587 6588 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6589 6590 llvm::MachO::mach_header_64 mach_header; 6591 if (addr_byte_size == 8) { 6592 mach_header.magic = MH_MAGIC_64; 6593 } else { 6594 mach_header.magic = MH_MAGIC; 6595 } 6596 mach_header.cputype = target_arch.GetMachOCPUType(); 6597 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6598 mach_header.filetype = MH_CORE; 6599 mach_header.ncmds = segment_load_commands.size(); 6600 mach_header.flags = 0; 6601 mach_header.reserved = 0; 6602 ThreadList &thread_list = process_sp->GetThreadList(); 6603 const uint32_t num_threads = thread_list.GetSize(); 6604 6605 // Make an array of LC_THREAD data items. Each one contains the 6606 // contents of the LC_THREAD load command. The data doesn't contain 6607 // the load command + load command size, we will add the load command 6608 // and load command size as we emit the data. 6609 std::vector<StreamString> LC_THREAD_datas(num_threads); 6610 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6611 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6612 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6613 LC_THREAD_data.SetByteOrder(byte_order); 6614 } 6615 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) { 6616 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6617 if (thread_sp) { 6618 switch (mach_header.cputype) { 6619 case llvm::MachO::CPU_TYPE_ARM64: 6620 case llvm::MachO::CPU_TYPE_ARM64_32: 6621 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6622 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6623 break; 6624 6625 case llvm::MachO::CPU_TYPE_ARM: 6626 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6627 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6628 break; 6629 6630 case llvm::MachO::CPU_TYPE_I386: 6631 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6632 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6633 break; 6634 6635 case llvm::MachO::CPU_TYPE_X86_64: 6636 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6637 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6638 break; 6639 } 6640 } 6641 } 6642 6643 // The size of the load command is the size of the segments... 6644 if (addr_byte_size == 8) { 6645 mach_header.sizeofcmds = segment_load_commands.size() * 6646 sizeof(llvm::MachO::segment_command_64); 6647 } else { 6648 mach_header.sizeofcmds = segment_load_commands.size() * 6649 sizeof(llvm::MachO::segment_command); 6650 } 6651 6652 // and the size of all LC_THREAD load command 6653 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6654 ++mach_header.ncmds; 6655 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6656 } 6657 6658 // Bits will be set to indicate which bits are NOT used in 6659 // addressing in this process or 0 for unknown. 6660 uint64_t address_mask = process_sp->GetCodeAddressMask(); 6661 if (address_mask != 0) { 6662 // LC_NOTE "addrable bits" 6663 mach_header.ncmds++; 6664 mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); 6665 } 6666 6667 // LC_NOTE "all image infos" 6668 mach_header.ncmds++; 6669 mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); 6670 6671 // Write the mach header 6672 buffer.PutHex32(mach_header.magic); 6673 buffer.PutHex32(mach_header.cputype); 6674 buffer.PutHex32(mach_header.cpusubtype); 6675 buffer.PutHex32(mach_header.filetype); 6676 buffer.PutHex32(mach_header.ncmds); 6677 buffer.PutHex32(mach_header.sizeofcmds); 6678 buffer.PutHex32(mach_header.flags); 6679 if (addr_byte_size == 8) { 6680 buffer.PutHex32(mach_header.reserved); 6681 } 6682 6683 // Skip the mach header and all load commands and align to the next 6684 // 0x1000 byte boundary 6685 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6686 6687 file_offset = llvm::alignTo(file_offset, 16); 6688 std::vector<std::unique_ptr<LCNoteEntry>> lc_notes; 6689 6690 // Add "addrable bits" LC_NOTE when an address mask is available 6691 if (address_mask != 0) { 6692 std::unique_ptr<LCNoteEntry> addrable_bits_lcnote_up( 6693 new LCNoteEntry(addr_byte_size, byte_order)); 6694 addrable_bits_lcnote_up->name = "addrable bits"; 6695 addrable_bits_lcnote_up->payload_file_offset = file_offset; 6696 int bits = std::bitset<64>(~address_mask).count(); 6697 addrable_bits_lcnote_up->payload.PutHex32(3); // version 6698 addrable_bits_lcnote_up->payload.PutHex32( 6699 bits); // # of bits used for addressing 6700 addrable_bits_lcnote_up->payload.PutHex64(0); // unused 6701 6702 file_offset += addrable_bits_lcnote_up->payload.GetSize(); 6703 6704 lc_notes.push_back(std::move(addrable_bits_lcnote_up)); 6705 } 6706 6707 // Add "all image infos" LC_NOTE 6708 std::unique_ptr<LCNoteEntry> all_image_infos_lcnote_up( 6709 new LCNoteEntry(addr_byte_size, byte_order)); 6710 all_image_infos_lcnote_up->name = "all image infos"; 6711 all_image_infos_lcnote_up->payload_file_offset = file_offset; 6712 file_offset = CreateAllImageInfosPayload( 6713 process_sp, file_offset, all_image_infos_lcnote_up->payload, 6714 core_style); 6715 lc_notes.push_back(std::move(all_image_infos_lcnote_up)); 6716 6717 // Add LC_NOTE load commands 6718 for (auto &lcnote : lc_notes) { 6719 // Add the LC_NOTE load command to the file. 6720 buffer.PutHex32(LC_NOTE); 6721 buffer.PutHex32(sizeof(llvm::MachO::note_command)); 6722 char namebuf[16]; 6723 memset(namebuf, 0, sizeof(namebuf)); 6724 // this is the uncommon case where strncpy is exactly 6725 // the right one, doesn't need to be nul terminated. 6726 strncpy(namebuf, lcnote->name.c_str(), sizeof(namebuf)); 6727 buffer.PutRawBytes(namebuf, sizeof(namebuf)); 6728 buffer.PutHex64(lcnote->payload_file_offset); 6729 buffer.PutHex64(lcnote->payload.GetSize()); 6730 } 6731 6732 // Align to 4096-byte page boundary for the LC_SEGMENTs. 6733 file_offset = llvm::alignTo(file_offset, 4096); 6734 6735 for (auto &segment : segment_load_commands) { 6736 segment.fileoff = file_offset; 6737 file_offset += segment.filesize; 6738 } 6739 6740 // Write out all of the LC_THREAD load commands 6741 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6742 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6743 buffer.PutHex32(LC_THREAD); 6744 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6745 buffer.Write(LC_THREAD_data.GetString().data(), LC_THREAD_data_size); 6746 } 6747 6748 // Write out all of the segment load commands 6749 for (const auto &segment : segment_load_commands) { 6750 buffer.PutHex32(segment.cmd); 6751 buffer.PutHex32(segment.cmdsize); 6752 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6753 if (addr_byte_size == 8) { 6754 buffer.PutHex64(segment.vmaddr); 6755 buffer.PutHex64(segment.vmsize); 6756 buffer.PutHex64(segment.fileoff); 6757 buffer.PutHex64(segment.filesize); 6758 } else { 6759 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6760 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6761 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6762 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6763 } 6764 buffer.PutHex32(segment.maxprot); 6765 buffer.PutHex32(segment.initprot); 6766 buffer.PutHex32(segment.nsects); 6767 buffer.PutHex32(segment.flags); 6768 } 6769 6770 std::string core_file_path(outfile.GetPath()); 6771 auto core_file = FileSystem::Instance().Open( 6772 outfile, File::eOpenOptionWriteOnly | File::eOpenOptionTruncate | 6773 File::eOpenOptionCanCreate); 6774 if (!core_file) { 6775 error = core_file.takeError(); 6776 } else { 6777 // Read 1 page at a time 6778 uint8_t bytes[0x1000]; 6779 // Write the mach header and load commands out to the core file 6780 size_t bytes_written = buffer.GetString().size(); 6781 error = 6782 core_file.get()->Write(buffer.GetString().data(), bytes_written); 6783 if (error.Success()) { 6784 6785 for (auto &lcnote : lc_notes) { 6786 if (core_file.get()->SeekFromStart(lcnote->payload_file_offset) == 6787 -1) { 6788 error.SetErrorStringWithFormat("Unable to seek to corefile pos " 6789 "to write '%s' LC_NOTE payload", 6790 lcnote->name.c_str()); 6791 return false; 6792 } 6793 bytes_written = lcnote->payload.GetSize(); 6794 error = core_file.get()->Write(lcnote->payload.GetData(), 6795 bytes_written); 6796 if (!error.Success()) 6797 return false; 6798 } 6799 6800 // Now write the file data for all memory segments in the process 6801 for (const auto &segment : segment_load_commands) { 6802 if (core_file.get()->SeekFromStart(segment.fileoff) == -1) { 6803 error.SetErrorStringWithFormat( 6804 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6805 segment.fileoff, core_file_path.c_str()); 6806 break; 6807 } 6808 6809 target.GetDebugger().GetAsyncOutputStream()->Printf( 6810 "Saving %" PRId64 6811 " bytes of data for memory region at 0x%" PRIx64 "\n", 6812 segment.vmsize, segment.vmaddr); 6813 addr_t bytes_left = segment.vmsize; 6814 addr_t addr = segment.vmaddr; 6815 Status memory_read_error; 6816 while (bytes_left > 0 && error.Success()) { 6817 const size_t bytes_to_read = 6818 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6819 6820 // In a savecore setting, we don't really care about caching, 6821 // as the data is dumped and very likely never read again, 6822 // so we call ReadMemoryFromInferior to bypass it. 6823 const size_t bytes_read = process_sp->ReadMemoryFromInferior( 6824 addr, bytes, bytes_to_read, memory_read_error); 6825 6826 if (bytes_read == bytes_to_read) { 6827 size_t bytes_written = bytes_read; 6828 error = core_file.get()->Write(bytes, bytes_written); 6829 bytes_left -= bytes_read; 6830 addr += bytes_read; 6831 } else { 6832 // Some pages within regions are not readable, those should 6833 // be zero filled 6834 memset(bytes, 0, bytes_to_read); 6835 size_t bytes_written = bytes_to_read; 6836 error = core_file.get()->Write(bytes, bytes_written); 6837 bytes_left -= bytes_to_read; 6838 addr += bytes_to_read; 6839 } 6840 } 6841 } 6842 } 6843 } 6844 } else { 6845 error.SetErrorString( 6846 "process doesn't support getting memory region info"); 6847 } 6848 } 6849 return true; // This is the right plug to handle saving core files for 6850 // this process 6851 } 6852 return false; 6853 } 6854 6855 ObjectFileMachO::MachOCorefileAllImageInfos 6856 ObjectFileMachO::GetCorefileAllImageInfos() { 6857 MachOCorefileAllImageInfos image_infos; 6858 6859 // Look for an "all image infos" LC_NOTE. 6860 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 6861 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 6862 const uint32_t cmd_offset = offset; 6863 llvm::MachO::load_command lc; 6864 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 6865 break; 6866 if (lc.cmd == LC_NOTE) { 6867 char data_owner[17]; 6868 m_data.CopyData(offset, 16, data_owner); 6869 data_owner[16] = '\0'; 6870 offset += 16; 6871 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 6872 offset += 4; /* size unused */ 6873 6874 if (strcmp("all image infos", data_owner) == 0) { 6875 offset = fileoff; 6876 // Read the struct all_image_infos_header. 6877 uint32_t version = m_data.GetU32(&offset); 6878 if (version != 1) { 6879 return image_infos; 6880 } 6881 uint32_t imgcount = m_data.GetU32(&offset); 6882 uint64_t entries_fileoff = m_data.GetU64(&offset); 6883 offset += 4; // uint32_t entries_size; 6884 offset += 4; // uint32_t unused; 6885 6886 offset = entries_fileoff; 6887 for (uint32_t i = 0; i < imgcount; i++) { 6888 // Read the struct image_entry. 6889 offset_t filepath_offset = m_data.GetU64(&offset); 6890 uuid_t uuid; 6891 memcpy(&uuid, m_data.GetData(&offset, sizeof(uuid_t)), 6892 sizeof(uuid_t)); 6893 uint64_t load_address = m_data.GetU64(&offset); 6894 offset_t seg_addrs_offset = m_data.GetU64(&offset); 6895 uint32_t segment_count = m_data.GetU32(&offset); 6896 uint32_t currently_executing = m_data.GetU32(&offset); 6897 6898 MachOCorefileImageEntry image_entry; 6899 image_entry.filename = (const char *)m_data.GetCStr(&filepath_offset); 6900 image_entry.uuid = UUID::fromData(uuid, sizeof(uuid_t)); 6901 image_entry.load_address = load_address; 6902 image_entry.currently_executing = currently_executing; 6903 6904 offset_t seg_vmaddrs_offset = seg_addrs_offset; 6905 for (uint32_t j = 0; j < segment_count; j++) { 6906 char segname[17]; 6907 m_data.CopyData(seg_vmaddrs_offset, 16, segname); 6908 segname[16] = '\0'; 6909 seg_vmaddrs_offset += 16; 6910 uint64_t vmaddr = m_data.GetU64(&seg_vmaddrs_offset); 6911 seg_vmaddrs_offset += 8; /* unused */ 6912 6913 std::tuple<ConstString, addr_t> new_seg{ConstString(segname), 6914 vmaddr}; 6915 image_entry.segment_load_addresses.push_back(new_seg); 6916 } 6917 image_infos.all_image_infos.push_back(image_entry); 6918 } 6919 } else if (strcmp("load binary", data_owner) == 0) { 6920 uint32_t version = m_data.GetU32(&fileoff); 6921 if (version == 1) { 6922 uuid_t uuid; 6923 memcpy(&uuid, m_data.GetData(&fileoff, sizeof(uuid_t)), 6924 sizeof(uuid_t)); 6925 uint64_t load_address = m_data.GetU64(&fileoff); 6926 uint64_t slide = m_data.GetU64(&fileoff); 6927 std::string filename = m_data.GetCStr(&fileoff); 6928 6929 MachOCorefileImageEntry image_entry; 6930 image_entry.filename = filename; 6931 image_entry.uuid = UUID::fromData(uuid, sizeof(uuid_t)); 6932 image_entry.load_address = load_address; 6933 image_entry.slide = slide; 6934 image_infos.all_image_infos.push_back(image_entry); 6935 } 6936 } 6937 } 6938 offset = cmd_offset + lc.cmdsize; 6939 } 6940 6941 return image_infos; 6942 } 6943 6944 bool ObjectFileMachO::LoadCoreFileImages(lldb_private::Process &process) { 6945 MachOCorefileAllImageInfos image_infos = GetCorefileAllImageInfos(); 6946 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_DYNAMIC_LOADER)); 6947 6948 ModuleList added_modules; 6949 for (const MachOCorefileImageEntry &image : image_infos.all_image_infos) { 6950 ModuleSpec module_spec; 6951 module_spec.GetUUID() = image.uuid; 6952 if (image.filename.empty()) { 6953 char namebuf[80]; 6954 if (image.load_address != LLDB_INVALID_ADDRESS) 6955 snprintf(namebuf, sizeof(namebuf), "mem-image-0x%" PRIx64, 6956 image.load_address); 6957 else 6958 snprintf(namebuf, sizeof(namebuf), "mem-image+0x%" PRIx64, image.slide); 6959 module_spec.GetFileSpec() = FileSpec(namebuf); 6960 } else { 6961 module_spec.GetFileSpec() = FileSpec(image.filename.c_str()); 6962 } 6963 if (image.currently_executing) { 6964 Symbols::DownloadObjectAndSymbolFile(module_spec, true); 6965 if (FileSystem::Instance().Exists(module_spec.GetFileSpec())) { 6966 process.GetTarget().GetOrCreateModule(module_spec, false); 6967 } 6968 } 6969 Status error; 6970 ModuleSP module_sp = 6971 process.GetTarget().GetOrCreateModule(module_spec, false, &error); 6972 if (!module_sp.get() || !module_sp->GetObjectFile()) { 6973 if (image.load_address != LLDB_INVALID_ADDRESS) { 6974 module_sp = process.ReadModuleFromMemory(module_spec.GetFileSpec(), 6975 image.load_address); 6976 } 6977 } 6978 if (module_sp.get()) { 6979 // Will call ModulesDidLoad with all modules once they've all 6980 // been added to the Target with load addresses. Don't notify 6981 // here, before the load address is set. 6982 const bool notify = false; 6983 process.GetTarget().GetImages().AppendIfNeeded(module_sp, notify); 6984 added_modules.Append(module_sp, notify); 6985 if (image.segment_load_addresses.size() > 0) { 6986 if (log) { 6987 std::string uuidstr = image.uuid.GetAsString(); 6988 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 6989 "UUID %s with section load addresses", 6990 image.filename.c_str(), uuidstr.c_str()); 6991 } 6992 for (auto name_vmaddr_tuple : image.segment_load_addresses) { 6993 SectionList *sectlist = module_sp->GetObjectFile()->GetSectionList(); 6994 if (sectlist) { 6995 SectionSP sect_sp = 6996 sectlist->FindSectionByName(std::get<0>(name_vmaddr_tuple)); 6997 if (sect_sp) { 6998 process.GetTarget().SetSectionLoadAddress( 6999 sect_sp, std::get<1>(name_vmaddr_tuple)); 7000 } 7001 } 7002 } 7003 } else if (image.load_address != LLDB_INVALID_ADDRESS) { 7004 if (log) { 7005 std::string uuidstr = image.uuid.GetAsString(); 7006 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 7007 "UUID %s with load address 0x%" PRIx64, 7008 image.filename.c_str(), uuidstr.c_str(), 7009 image.load_address); 7010 } 7011 const bool address_is_slide = false; 7012 bool changed = false; 7013 module_sp->SetLoadAddress(process.GetTarget(), image.load_address, 7014 address_is_slide, changed); 7015 } else if (image.slide != 0) { 7016 if (log) { 7017 std::string uuidstr = image.uuid.GetAsString(); 7018 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 7019 "UUID %s with slide amount 0x%" PRIx64, 7020 image.filename.c_str(), uuidstr.c_str(), image.slide); 7021 } 7022 const bool address_is_slide = true; 7023 bool changed = false; 7024 module_sp->SetLoadAddress(process.GetTarget(), image.slide, 7025 address_is_slide, changed); 7026 } else { 7027 if (log) { 7028 std::string uuidstr = image.uuid.GetAsString(); 7029 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' " 7030 "UUID %s at its file address, no slide applied", 7031 image.filename.c_str(), uuidstr.c_str()); 7032 } 7033 const bool address_is_slide = true; 7034 bool changed = false; 7035 module_sp->SetLoadAddress(process.GetTarget(), 0, address_is_slide, 7036 changed); 7037 } 7038 } 7039 } 7040 if (added_modules.GetSize() > 0) { 7041 process.GetTarget().ModulesDidLoad(added_modules); 7042 process.Flush(); 7043 return true; 7044 } 7045 return false; 7046 } 7047