1 //===-- ObjectFileMachO.cpp -------------------------------------*- C++ -*-===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 10 // C Includes 11 // C++ Includes 12 // Other libraries and framework includes 13 #include "llvm/ADT/StringRef.h" 14 15 // Project includes 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/RangeMap.h" 26 #include "lldb/Core/RegisterValue.h" 27 #include "lldb/Core/Section.h" 28 #include "lldb/Core/StreamFile.h" 29 #include "lldb/Host/Host.h" 30 #include "lldb/Symbol/DWARFCallFrameInfo.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/Status.h" 45 #include "lldb/Utility/StreamString.h" 46 #include "lldb/Utility/Timer.h" 47 #include "lldb/Utility/UUID.h" 48 49 #include "lldb/Utility/SafeMachO.h" 50 51 #include "llvm/Support/MemoryBuffer.h" 52 53 #include "ObjectFileMachO.h" 54 55 #if defined(__APPLE__) && \ 56 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 57 // GetLLDBSharedCacheUUID() needs to call dlsym() 58 #include <dlfcn.h> 59 #endif 60 61 #ifndef __APPLE__ 62 #include "Utility/UuidCompatibility.h" 63 #else 64 #include <uuid/uuid.h> 65 #endif 66 67 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 68 using namespace lldb; 69 using namespace lldb_private; 70 using namespace llvm::MachO; 71 72 // Some structure definitions needed for parsing the dyld shared cache files 73 // found on iOS devices. 74 75 struct lldb_copy_dyld_cache_header_v1 { 76 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 77 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 78 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 79 uint32_t imagesOffset; 80 uint32_t imagesCount; 81 uint64_t dyldBaseAddress; 82 uint64_t codeSignatureOffset; 83 uint64_t codeSignatureSize; 84 uint64_t slideInfoOffset; 85 uint64_t slideInfoSize; 86 uint64_t localSymbolsOffset; 87 uint64_t localSymbolsSize; 88 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 89 // and later 90 }; 91 92 struct lldb_copy_dyld_cache_mapping_info { 93 uint64_t address; 94 uint64_t size; 95 uint64_t fileOffset; 96 uint32_t maxProt; 97 uint32_t initProt; 98 }; 99 100 struct lldb_copy_dyld_cache_local_symbols_info { 101 uint32_t nlistOffset; 102 uint32_t nlistCount; 103 uint32_t stringsOffset; 104 uint32_t stringsSize; 105 uint32_t entriesOffset; 106 uint32_t entriesCount; 107 }; 108 struct lldb_copy_dyld_cache_local_symbols_entry { 109 uint32_t dylibOffset; 110 uint32_t nlistStartIndex; 111 uint32_t nlistCount; 112 }; 113 114 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 115 public: 116 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 117 const DataExtractor &data) 118 : RegisterContextDarwin_x86_64(thread, 0) { 119 SetRegisterDataFrom_LC_THREAD(data); 120 } 121 122 void InvalidateAllRegisters() override { 123 // Do nothing... registers are always valid... 124 } 125 126 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 127 lldb::offset_t offset = 0; 128 SetError(GPRRegSet, Read, -1); 129 SetError(FPURegSet, Read, -1); 130 SetError(EXCRegSet, Read, -1); 131 bool done = false; 132 133 while (!done) { 134 int flavor = data.GetU32(&offset); 135 if (flavor == 0) 136 done = true; 137 else { 138 uint32_t i; 139 uint32_t count = data.GetU32(&offset); 140 switch (flavor) { 141 case GPRRegSet: 142 for (i = 0; i < count; ++i) 143 (&gpr.rax)[i] = data.GetU64(&offset); 144 SetError(GPRRegSet, Read, 0); 145 done = true; 146 147 break; 148 case FPURegSet: 149 // TODO: fill in FPU regs.... 150 // SetError (FPURegSet, Read, -1); 151 done = true; 152 153 break; 154 case EXCRegSet: 155 exc.trapno = data.GetU32(&offset); 156 exc.err = data.GetU32(&offset); 157 exc.faultvaddr = data.GetU64(&offset); 158 SetError(EXCRegSet, Read, 0); 159 done = true; 160 break; 161 case 7: 162 case 8: 163 case 9: 164 // fancy flavors that encapsulate of the above 165 // flavors... 166 break; 167 168 default: 169 done = true; 170 break; 171 } 172 } 173 } 174 } 175 176 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 177 const char *alt_name, size_t reg_byte_size, 178 Stream &data) { 179 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 180 if (reg_info == NULL) 181 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 182 if (reg_info) { 183 lldb_private::RegisterValue reg_value; 184 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 185 if (reg_info->byte_size >= reg_byte_size) 186 data.Write(reg_value.GetBytes(), reg_byte_size); 187 else { 188 data.Write(reg_value.GetBytes(), reg_info->byte_size); 189 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 190 ++i) 191 data.PutChar(0); 192 } 193 return reg_byte_size; 194 } 195 } 196 // Just write zeros if all else fails 197 for (size_t i = 0; i < reg_byte_size; ++i) 198 data.PutChar(0); 199 return reg_byte_size; 200 } 201 202 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 203 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 204 if (reg_ctx_sp) { 205 RegisterContext *reg_ctx = reg_ctx_sp.get(); 206 207 data.PutHex32(GPRRegSet); // Flavor 208 data.PutHex32(GPRWordCount); 209 WriteRegister(reg_ctx, "rax", NULL, 8, data); 210 WriteRegister(reg_ctx, "rbx", NULL, 8, data); 211 WriteRegister(reg_ctx, "rcx", NULL, 8, data); 212 WriteRegister(reg_ctx, "rdx", NULL, 8, data); 213 WriteRegister(reg_ctx, "rdi", NULL, 8, data); 214 WriteRegister(reg_ctx, "rsi", NULL, 8, data); 215 WriteRegister(reg_ctx, "rbp", NULL, 8, data); 216 WriteRegister(reg_ctx, "rsp", NULL, 8, data); 217 WriteRegister(reg_ctx, "r8", NULL, 8, data); 218 WriteRegister(reg_ctx, "r9", NULL, 8, data); 219 WriteRegister(reg_ctx, "r10", NULL, 8, data); 220 WriteRegister(reg_ctx, "r11", NULL, 8, data); 221 WriteRegister(reg_ctx, "r12", NULL, 8, data); 222 WriteRegister(reg_ctx, "r13", NULL, 8, data); 223 WriteRegister(reg_ctx, "r14", NULL, 8, data); 224 WriteRegister(reg_ctx, "r15", NULL, 8, data); 225 WriteRegister(reg_ctx, "rip", NULL, 8, data); 226 WriteRegister(reg_ctx, "rflags", NULL, 8, data); 227 WriteRegister(reg_ctx, "cs", NULL, 8, data); 228 WriteRegister(reg_ctx, "fs", NULL, 8, data); 229 WriteRegister(reg_ctx, "gs", NULL, 8, data); 230 231 // // Write out the FPU registers 232 // const size_t fpu_byte_size = sizeof(FPU); 233 // size_t bytes_written = 0; 234 // data.PutHex32 (FPURegSet); 235 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 236 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 237 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 238 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 239 // data); // uint16_t fcw; // "fctrl" 240 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 241 // data); // uint16_t fsw; // "fstat" 242 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 243 // data); // uint8_t ftw; // "ftag" 244 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 245 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 246 // data); // uint16_t fop; // "fop" 247 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 248 // data); // uint32_t ip; // "fioff" 249 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 250 // data); // uint16_t cs; // "fiseg" 251 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 252 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 253 // data); // uint32_t dp; // "fooff" 254 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 255 // data); // uint16_t ds; // "foseg" 256 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 257 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 258 // data); // uint32_t mxcsr; 259 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 260 // 4, data);// uint32_t mxcsrmask; 261 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 262 // sizeof(MMSReg), data); 263 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 264 // sizeof(MMSReg), data); 265 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 266 // sizeof(MMSReg), data); 267 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 268 // sizeof(MMSReg), data); 269 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 270 // sizeof(MMSReg), data); 271 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 272 // sizeof(MMSReg), data); 273 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 274 // sizeof(MMSReg), data); 275 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 276 // sizeof(MMSReg), data); 277 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 278 // sizeof(XMMReg), data); 279 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 280 // sizeof(XMMReg), data); 281 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 282 // sizeof(XMMReg), data); 283 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 284 // sizeof(XMMReg), data); 285 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 286 // sizeof(XMMReg), data); 287 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 288 // sizeof(XMMReg), data); 289 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 290 // sizeof(XMMReg), data); 291 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 292 // sizeof(XMMReg), data); 293 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 294 // sizeof(XMMReg), data); 295 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 296 // sizeof(XMMReg), data); 297 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 298 // sizeof(XMMReg), data); 299 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 300 // sizeof(XMMReg), data); 301 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 302 // sizeof(XMMReg), data); 303 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 304 // sizeof(XMMReg), data); 305 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 306 // sizeof(XMMReg), data); 307 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 308 // sizeof(XMMReg), data); 309 // 310 // // Fill rest with zeros 311 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 312 // i) 313 // data.PutChar(0); 314 315 // Write out the EXC registers 316 data.PutHex32(EXCRegSet); 317 data.PutHex32(EXCWordCount); 318 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 319 WriteRegister(reg_ctx, "err", NULL, 4, data); 320 WriteRegister(reg_ctx, "faultvaddr", NULL, 8, data); 321 return true; 322 } 323 return false; 324 } 325 326 protected: 327 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 328 329 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 330 331 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 332 333 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 334 return 0; 335 } 336 337 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 338 return 0; 339 } 340 341 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 342 return 0; 343 } 344 }; 345 346 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 347 public: 348 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 349 const DataExtractor &data) 350 : RegisterContextDarwin_i386(thread, 0) { 351 SetRegisterDataFrom_LC_THREAD(data); 352 } 353 354 void InvalidateAllRegisters() override { 355 // Do nothing... registers are always valid... 356 } 357 358 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 359 lldb::offset_t offset = 0; 360 SetError(GPRRegSet, Read, -1); 361 SetError(FPURegSet, Read, -1); 362 SetError(EXCRegSet, Read, -1); 363 bool done = false; 364 365 while (!done) { 366 int flavor = data.GetU32(&offset); 367 if (flavor == 0) 368 done = true; 369 else { 370 uint32_t i; 371 uint32_t count = data.GetU32(&offset); 372 switch (flavor) { 373 case GPRRegSet: 374 for (i = 0; i < count; ++i) 375 (&gpr.eax)[i] = data.GetU32(&offset); 376 SetError(GPRRegSet, Read, 0); 377 done = true; 378 379 break; 380 case FPURegSet: 381 // TODO: fill in FPU regs.... 382 // SetError (FPURegSet, Read, -1); 383 done = true; 384 385 break; 386 case EXCRegSet: 387 exc.trapno = data.GetU32(&offset); 388 exc.err = data.GetU32(&offset); 389 exc.faultvaddr = data.GetU32(&offset); 390 SetError(EXCRegSet, Read, 0); 391 done = true; 392 break; 393 case 7: 394 case 8: 395 case 9: 396 // fancy flavors that encapsulate of the above 397 // flavors... 398 break; 399 400 default: 401 done = true; 402 break; 403 } 404 } 405 } 406 } 407 408 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 409 const char *alt_name, size_t reg_byte_size, 410 Stream &data) { 411 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 412 if (reg_info == NULL) 413 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 414 if (reg_info) { 415 lldb_private::RegisterValue reg_value; 416 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 417 if (reg_info->byte_size >= reg_byte_size) 418 data.Write(reg_value.GetBytes(), reg_byte_size); 419 else { 420 data.Write(reg_value.GetBytes(), reg_info->byte_size); 421 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 422 ++i) 423 data.PutChar(0); 424 } 425 return reg_byte_size; 426 } 427 } 428 // Just write zeros if all else fails 429 for (size_t i = 0; i < reg_byte_size; ++i) 430 data.PutChar(0); 431 return reg_byte_size; 432 } 433 434 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 435 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 436 if (reg_ctx_sp) { 437 RegisterContext *reg_ctx = reg_ctx_sp.get(); 438 439 data.PutHex32(GPRRegSet); // Flavor 440 data.PutHex32(GPRWordCount); 441 WriteRegister(reg_ctx, "eax", NULL, 4, data); 442 WriteRegister(reg_ctx, "ebx", NULL, 4, data); 443 WriteRegister(reg_ctx, "ecx", NULL, 4, data); 444 WriteRegister(reg_ctx, "edx", NULL, 4, data); 445 WriteRegister(reg_ctx, "edi", NULL, 4, data); 446 WriteRegister(reg_ctx, "esi", NULL, 4, data); 447 WriteRegister(reg_ctx, "ebp", NULL, 4, data); 448 WriteRegister(reg_ctx, "esp", NULL, 4, data); 449 WriteRegister(reg_ctx, "ss", NULL, 4, data); 450 WriteRegister(reg_ctx, "eflags", NULL, 4, data); 451 WriteRegister(reg_ctx, "eip", NULL, 4, data); 452 WriteRegister(reg_ctx, "cs", NULL, 4, data); 453 WriteRegister(reg_ctx, "ds", NULL, 4, data); 454 WriteRegister(reg_ctx, "es", NULL, 4, data); 455 WriteRegister(reg_ctx, "fs", NULL, 4, data); 456 WriteRegister(reg_ctx, "gs", NULL, 4, data); 457 458 // Write out the EXC registers 459 data.PutHex32(EXCRegSet); 460 data.PutHex32(EXCWordCount); 461 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 462 WriteRegister(reg_ctx, "err", NULL, 4, data); 463 WriteRegister(reg_ctx, "faultvaddr", NULL, 4, data); 464 return true; 465 } 466 return false; 467 } 468 469 protected: 470 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 471 472 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 473 474 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 475 476 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 477 return 0; 478 } 479 480 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 481 return 0; 482 } 483 484 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 485 return 0; 486 } 487 }; 488 489 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 490 public: 491 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 492 const DataExtractor &data) 493 : RegisterContextDarwin_arm(thread, 0) { 494 SetRegisterDataFrom_LC_THREAD(data); 495 } 496 497 void InvalidateAllRegisters() override { 498 // Do nothing... registers are always valid... 499 } 500 501 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 502 lldb::offset_t offset = 0; 503 SetError(GPRRegSet, Read, -1); 504 SetError(FPURegSet, Read, -1); 505 SetError(EXCRegSet, Read, -1); 506 bool done = false; 507 508 while (!done) { 509 int flavor = data.GetU32(&offset); 510 uint32_t count = data.GetU32(&offset); 511 lldb::offset_t next_thread_state = offset + (count * 4); 512 switch (flavor) { 513 case GPRAltRegSet: 514 case GPRRegSet: 515 for (uint32_t i = 0; i < count; ++i) { 516 gpr.r[i] = data.GetU32(&offset); 517 } 518 519 // Note that gpr.cpsr is also copied by the above loop; this loop 520 // technically extends 521 // one element past the end of the gpr.r[] array. 522 523 SetError(GPRRegSet, Read, 0); 524 offset = next_thread_state; 525 break; 526 527 case FPURegSet: { 528 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 529 const int fpu_reg_buf_size = sizeof(fpu.floats); 530 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 531 fpu_reg_buf) == fpu_reg_buf_size) { 532 offset += fpu_reg_buf_size; 533 fpu.fpscr = data.GetU32(&offset); 534 SetError(FPURegSet, Read, 0); 535 } else { 536 done = true; 537 } 538 } 539 offset = next_thread_state; 540 break; 541 542 case EXCRegSet: 543 if (count == 3) { 544 exc.exception = data.GetU32(&offset); 545 exc.fsr = data.GetU32(&offset); 546 exc.far = data.GetU32(&offset); 547 SetError(EXCRegSet, Read, 0); 548 } 549 done = true; 550 offset = next_thread_state; 551 break; 552 553 // Unknown register set flavor, stop trying to parse. 554 default: 555 done = true; 556 } 557 } 558 } 559 560 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 561 const char *alt_name, size_t reg_byte_size, 562 Stream &data) { 563 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 564 if (reg_info == NULL) 565 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 566 if (reg_info) { 567 lldb_private::RegisterValue reg_value; 568 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 569 if (reg_info->byte_size >= reg_byte_size) 570 data.Write(reg_value.GetBytes(), reg_byte_size); 571 else { 572 data.Write(reg_value.GetBytes(), reg_info->byte_size); 573 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 574 ++i) 575 data.PutChar(0); 576 } 577 return reg_byte_size; 578 } 579 } 580 // Just write zeros if all else fails 581 for (size_t i = 0; i < reg_byte_size; ++i) 582 data.PutChar(0); 583 return reg_byte_size; 584 } 585 586 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 587 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 588 if (reg_ctx_sp) { 589 RegisterContext *reg_ctx = reg_ctx_sp.get(); 590 591 data.PutHex32(GPRRegSet); // Flavor 592 data.PutHex32(GPRWordCount); 593 WriteRegister(reg_ctx, "r0", NULL, 4, data); 594 WriteRegister(reg_ctx, "r1", NULL, 4, data); 595 WriteRegister(reg_ctx, "r2", NULL, 4, data); 596 WriteRegister(reg_ctx, "r3", NULL, 4, data); 597 WriteRegister(reg_ctx, "r4", NULL, 4, data); 598 WriteRegister(reg_ctx, "r5", NULL, 4, data); 599 WriteRegister(reg_ctx, "r6", NULL, 4, data); 600 WriteRegister(reg_ctx, "r7", NULL, 4, data); 601 WriteRegister(reg_ctx, "r8", NULL, 4, data); 602 WriteRegister(reg_ctx, "r9", NULL, 4, data); 603 WriteRegister(reg_ctx, "r10", NULL, 4, data); 604 WriteRegister(reg_ctx, "r11", NULL, 4, data); 605 WriteRegister(reg_ctx, "r12", NULL, 4, data); 606 WriteRegister(reg_ctx, "sp", NULL, 4, data); 607 WriteRegister(reg_ctx, "lr", NULL, 4, data); 608 WriteRegister(reg_ctx, "pc", NULL, 4, data); 609 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 610 611 // Write out the EXC registers 612 // data.PutHex32 (EXCRegSet); 613 // data.PutHex32 (EXCWordCount); 614 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 615 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 616 // WriteRegister (reg_ctx, "far", NULL, 4, data); 617 return true; 618 } 619 return false; 620 } 621 622 protected: 623 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 624 625 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 626 627 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 628 629 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 630 631 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 632 return 0; 633 } 634 635 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 636 return 0; 637 } 638 639 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 640 return 0; 641 } 642 643 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 644 return -1; 645 } 646 }; 647 648 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 649 public: 650 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 651 const DataExtractor &data) 652 : RegisterContextDarwin_arm64(thread, 0) { 653 SetRegisterDataFrom_LC_THREAD(data); 654 } 655 656 void InvalidateAllRegisters() override { 657 // Do nothing... registers are always valid... 658 } 659 660 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 661 lldb::offset_t offset = 0; 662 SetError(GPRRegSet, Read, -1); 663 SetError(FPURegSet, Read, -1); 664 SetError(EXCRegSet, Read, -1); 665 bool done = false; 666 while (!done) { 667 int flavor = data.GetU32(&offset); 668 uint32_t count = data.GetU32(&offset); 669 lldb::offset_t next_thread_state = offset + (count * 4); 670 switch (flavor) { 671 case GPRRegSet: 672 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 673 // 32-bit register) 674 if (count >= (33 * 2) + 1) { 675 for (uint32_t i = 0; i < 29; ++i) 676 gpr.x[i] = data.GetU64(&offset); 677 gpr.fp = data.GetU64(&offset); 678 gpr.lr = data.GetU64(&offset); 679 gpr.sp = data.GetU64(&offset); 680 gpr.pc = data.GetU64(&offset); 681 gpr.cpsr = data.GetU32(&offset); 682 SetError(GPRRegSet, Read, 0); 683 } 684 offset = next_thread_state; 685 break; 686 case FPURegSet: { 687 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 688 const int fpu_reg_buf_size = sizeof(fpu); 689 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 690 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 691 fpu_reg_buf) == fpu_reg_buf_size) { 692 SetError(FPURegSet, Read, 0); 693 } else { 694 done = true; 695 } 696 } 697 offset = next_thread_state; 698 break; 699 case EXCRegSet: 700 if (count == 4) { 701 exc.far = data.GetU64(&offset); 702 exc.esr = data.GetU32(&offset); 703 exc.exception = data.GetU32(&offset); 704 SetError(EXCRegSet, Read, 0); 705 } 706 offset = next_thread_state; 707 break; 708 default: 709 done = true; 710 break; 711 } 712 } 713 } 714 715 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 716 const char *alt_name, size_t reg_byte_size, 717 Stream &data) { 718 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 719 if (reg_info == NULL) 720 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 721 if (reg_info) { 722 lldb_private::RegisterValue reg_value; 723 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 724 if (reg_info->byte_size >= reg_byte_size) 725 data.Write(reg_value.GetBytes(), reg_byte_size); 726 else { 727 data.Write(reg_value.GetBytes(), reg_info->byte_size); 728 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 729 ++i) 730 data.PutChar(0); 731 } 732 return reg_byte_size; 733 } 734 } 735 // Just write zeros if all else fails 736 for (size_t i = 0; i < reg_byte_size; ++i) 737 data.PutChar(0); 738 return reg_byte_size; 739 } 740 741 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 742 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 743 if (reg_ctx_sp) { 744 RegisterContext *reg_ctx = reg_ctx_sp.get(); 745 746 data.PutHex32(GPRRegSet); // Flavor 747 data.PutHex32(GPRWordCount); 748 WriteRegister(reg_ctx, "x0", NULL, 8, data); 749 WriteRegister(reg_ctx, "x1", NULL, 8, data); 750 WriteRegister(reg_ctx, "x2", NULL, 8, data); 751 WriteRegister(reg_ctx, "x3", NULL, 8, data); 752 WriteRegister(reg_ctx, "x4", NULL, 8, data); 753 WriteRegister(reg_ctx, "x5", NULL, 8, data); 754 WriteRegister(reg_ctx, "x6", NULL, 8, data); 755 WriteRegister(reg_ctx, "x7", NULL, 8, data); 756 WriteRegister(reg_ctx, "x8", NULL, 8, data); 757 WriteRegister(reg_ctx, "x9", NULL, 8, data); 758 WriteRegister(reg_ctx, "x10", NULL, 8, data); 759 WriteRegister(reg_ctx, "x11", NULL, 8, data); 760 WriteRegister(reg_ctx, "x12", NULL, 8, data); 761 WriteRegister(reg_ctx, "x13", NULL, 8, data); 762 WriteRegister(reg_ctx, "x14", NULL, 8, data); 763 WriteRegister(reg_ctx, "x15", NULL, 8, data); 764 WriteRegister(reg_ctx, "x16", NULL, 8, data); 765 WriteRegister(reg_ctx, "x17", NULL, 8, data); 766 WriteRegister(reg_ctx, "x18", NULL, 8, data); 767 WriteRegister(reg_ctx, "x19", NULL, 8, data); 768 WriteRegister(reg_ctx, "x20", NULL, 8, data); 769 WriteRegister(reg_ctx, "x21", NULL, 8, data); 770 WriteRegister(reg_ctx, "x22", NULL, 8, data); 771 WriteRegister(reg_ctx, "x23", NULL, 8, data); 772 WriteRegister(reg_ctx, "x24", NULL, 8, data); 773 WriteRegister(reg_ctx, "x25", NULL, 8, data); 774 WriteRegister(reg_ctx, "x26", NULL, 8, data); 775 WriteRegister(reg_ctx, "x27", NULL, 8, data); 776 WriteRegister(reg_ctx, "x28", NULL, 8, data); 777 WriteRegister(reg_ctx, "fp", NULL, 8, data); 778 WriteRegister(reg_ctx, "lr", NULL, 8, data); 779 WriteRegister(reg_ctx, "sp", NULL, 8, data); 780 WriteRegister(reg_ctx, "pc", NULL, 8, data); 781 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 782 783 // Write out the EXC registers 784 // data.PutHex32 (EXCRegSet); 785 // data.PutHex32 (EXCWordCount); 786 // WriteRegister (reg_ctx, "far", NULL, 8, data); 787 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 788 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 789 return true; 790 } 791 return false; 792 } 793 794 protected: 795 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 796 797 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 798 799 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 800 801 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 802 803 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 804 return 0; 805 } 806 807 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 808 return 0; 809 } 810 811 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 812 return 0; 813 } 814 815 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 816 return -1; 817 } 818 }; 819 820 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 821 switch (magic) { 822 case MH_MAGIC: 823 case MH_CIGAM: 824 return sizeof(struct mach_header); 825 826 case MH_MAGIC_64: 827 case MH_CIGAM_64: 828 return sizeof(struct mach_header_64); 829 break; 830 831 default: 832 break; 833 } 834 return 0; 835 } 836 837 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 838 839 void ObjectFileMachO::Initialize() { 840 PluginManager::RegisterPlugin( 841 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 842 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 843 } 844 845 void ObjectFileMachO::Terminate() { 846 PluginManager::UnregisterPlugin(CreateInstance); 847 } 848 849 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 850 static ConstString g_name("mach-o"); 851 return g_name; 852 } 853 854 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 855 return "Mach-o object file reader (32 and 64 bit)"; 856 } 857 858 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 859 DataBufferSP &data_sp, 860 lldb::offset_t data_offset, 861 const FileSpec *file, 862 lldb::offset_t file_offset, 863 lldb::offset_t length) { 864 if (!data_sp) { 865 data_sp = MapFileData(*file, length, file_offset); 866 if (!data_sp) 867 return nullptr; 868 data_offset = 0; 869 } 870 871 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 872 return nullptr; 873 874 // Update the data to contain the entire file if it doesn't already 875 if (data_sp->GetByteSize() < length) { 876 data_sp = MapFileData(*file, length, file_offset); 877 if (!data_sp) 878 return nullptr; 879 data_offset = 0; 880 } 881 auto objfile_ap = llvm::make_unique<ObjectFileMachO>( 882 module_sp, data_sp, data_offset, file, file_offset, length); 883 if (!objfile_ap || !objfile_ap->ParseHeader()) 884 return nullptr; 885 886 return objfile_ap.release(); 887 } 888 889 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 890 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 891 const ProcessSP &process_sp, lldb::addr_t header_addr) { 892 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 893 std::unique_ptr<ObjectFile> objfile_ap( 894 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 895 if (objfile_ap.get() && objfile_ap->ParseHeader()) 896 return objfile_ap.release(); 897 } 898 return NULL; 899 } 900 901 size_t ObjectFileMachO::GetModuleSpecifications( 902 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 903 lldb::offset_t data_offset, lldb::offset_t file_offset, 904 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 905 const size_t initial_count = specs.GetSize(); 906 907 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 908 DataExtractor data; 909 data.SetData(data_sp); 910 llvm::MachO::mach_header header; 911 if (ParseHeader(data, &data_offset, header)) { 912 size_t header_and_load_cmds = 913 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 914 if (header_and_load_cmds >= data_sp->GetByteSize()) { 915 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 916 data.SetData(data_sp); 917 data_offset = MachHeaderSizeFromMagic(header.magic); 918 } 919 if (data_sp) { 920 ModuleSpec spec; 921 spec.GetFileSpec() = file; 922 spec.SetObjectOffset(file_offset); 923 spec.SetObjectSize(length); 924 925 if (GetArchitecture(header, data, data_offset, 926 spec.GetArchitecture())) { 927 if (spec.GetArchitecture().IsValid()) { 928 GetUUID(header, data, data_offset, spec.GetUUID()); 929 specs.Append(spec); 930 } 931 } 932 } 933 } 934 } 935 return specs.GetSize() - initial_count; 936 } 937 938 const ConstString &ObjectFileMachO::GetSegmentNameTEXT() { 939 static ConstString g_segment_name_TEXT("__TEXT"); 940 return g_segment_name_TEXT; 941 } 942 943 const ConstString &ObjectFileMachO::GetSegmentNameDATA() { 944 static ConstString g_segment_name_DATA("__DATA"); 945 return g_segment_name_DATA; 946 } 947 948 const ConstString &ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 949 static ConstString g_segment_name("__DATA_DIRTY"); 950 return g_segment_name; 951 } 952 953 const ConstString &ObjectFileMachO::GetSegmentNameDATA_CONST() { 954 static ConstString g_segment_name("__DATA_CONST"); 955 return g_segment_name; 956 } 957 958 const ConstString &ObjectFileMachO::GetSegmentNameOBJC() { 959 static ConstString g_segment_name_OBJC("__OBJC"); 960 return g_segment_name_OBJC; 961 } 962 963 const ConstString &ObjectFileMachO::GetSegmentNameLINKEDIT() { 964 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 965 return g_section_name_LINKEDIT; 966 } 967 968 const ConstString &ObjectFileMachO::GetSectionNameEHFrame() { 969 static ConstString g_section_name_eh_frame("__eh_frame"); 970 return g_section_name_eh_frame; 971 } 972 973 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 974 lldb::addr_t data_offset, 975 lldb::addr_t data_length) { 976 DataExtractor data; 977 data.SetData(data_sp, data_offset, data_length); 978 lldb::offset_t offset = 0; 979 uint32_t magic = data.GetU32(&offset); 980 return MachHeaderSizeFromMagic(magic) != 0; 981 } 982 983 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 984 DataBufferSP &data_sp, 985 lldb::offset_t data_offset, 986 const FileSpec *file, 987 lldb::offset_t file_offset, 988 lldb::offset_t length) 989 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 990 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 991 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 992 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 993 ::memset(&m_header, 0, sizeof(m_header)); 994 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 995 } 996 997 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 998 lldb::DataBufferSP &header_data_sp, 999 const lldb::ProcessSP &process_sp, 1000 lldb::addr_t header_addr) 1001 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 1002 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 1003 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 1004 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 1005 ::memset(&m_header, 0, sizeof(m_header)); 1006 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1007 } 1008 1009 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 1010 lldb::offset_t *data_offset_ptr, 1011 llvm::MachO::mach_header &header) { 1012 data.SetByteOrder(endian::InlHostByteOrder()); 1013 // Leave magic in the original byte order 1014 header.magic = data.GetU32(data_offset_ptr); 1015 bool can_parse = false; 1016 bool is_64_bit = false; 1017 switch (header.magic) { 1018 case MH_MAGIC: 1019 data.SetByteOrder(endian::InlHostByteOrder()); 1020 data.SetAddressByteSize(4); 1021 can_parse = true; 1022 break; 1023 1024 case MH_MAGIC_64: 1025 data.SetByteOrder(endian::InlHostByteOrder()); 1026 data.SetAddressByteSize(8); 1027 can_parse = true; 1028 is_64_bit = true; 1029 break; 1030 1031 case MH_CIGAM: 1032 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1033 ? eByteOrderLittle 1034 : eByteOrderBig); 1035 data.SetAddressByteSize(4); 1036 can_parse = true; 1037 break; 1038 1039 case MH_CIGAM_64: 1040 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1041 ? eByteOrderLittle 1042 : eByteOrderBig); 1043 data.SetAddressByteSize(8); 1044 is_64_bit = true; 1045 can_parse = true; 1046 break; 1047 1048 default: 1049 break; 1050 } 1051 1052 if (can_parse) { 1053 data.GetU32(data_offset_ptr, &header.cputype, 6); 1054 if (is_64_bit) 1055 *data_offset_ptr += 4; 1056 return true; 1057 } else { 1058 memset(&header, 0, sizeof(header)); 1059 } 1060 return false; 1061 } 1062 1063 bool ObjectFileMachO::ParseHeader() { 1064 ModuleSP module_sp(GetModule()); 1065 if (module_sp) { 1066 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1067 bool can_parse = false; 1068 lldb::offset_t offset = 0; 1069 m_data.SetByteOrder(endian::InlHostByteOrder()); 1070 // Leave magic in the original byte order 1071 m_header.magic = m_data.GetU32(&offset); 1072 switch (m_header.magic) { 1073 case MH_MAGIC: 1074 m_data.SetByteOrder(endian::InlHostByteOrder()); 1075 m_data.SetAddressByteSize(4); 1076 can_parse = true; 1077 break; 1078 1079 case MH_MAGIC_64: 1080 m_data.SetByteOrder(endian::InlHostByteOrder()); 1081 m_data.SetAddressByteSize(8); 1082 can_parse = true; 1083 break; 1084 1085 case MH_CIGAM: 1086 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1087 ? eByteOrderLittle 1088 : eByteOrderBig); 1089 m_data.SetAddressByteSize(4); 1090 can_parse = true; 1091 break; 1092 1093 case MH_CIGAM_64: 1094 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1095 ? eByteOrderLittle 1096 : eByteOrderBig); 1097 m_data.SetAddressByteSize(8); 1098 can_parse = true; 1099 break; 1100 1101 default: 1102 break; 1103 } 1104 1105 if (can_parse) { 1106 m_data.GetU32(&offset, &m_header.cputype, 6); 1107 1108 ArchSpec mach_arch; 1109 1110 if (GetArchitecture(mach_arch)) { 1111 // Check if the module has a required architecture 1112 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1113 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1114 return false; 1115 1116 if (SetModulesArchitecture(mach_arch)) { 1117 const size_t header_and_lc_size = 1118 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1119 if (m_data.GetByteSize() < header_and_lc_size) { 1120 DataBufferSP data_sp; 1121 ProcessSP process_sp(m_process_wp.lock()); 1122 if (process_sp) { 1123 data_sp = 1124 ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1125 } else { 1126 // Read in all only the load command data from the file on disk 1127 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset); 1128 if (data_sp->GetByteSize() != header_and_lc_size) 1129 return false; 1130 } 1131 if (data_sp) 1132 m_data.SetData(data_sp); 1133 } 1134 } 1135 return true; 1136 } 1137 } else { 1138 memset(&m_header, 0, sizeof(struct mach_header)); 1139 } 1140 } 1141 return false; 1142 } 1143 1144 ByteOrder ObjectFileMachO::GetByteOrder() const { 1145 return m_data.GetByteOrder(); 1146 } 1147 1148 bool ObjectFileMachO::IsExecutable() const { 1149 return m_header.filetype == MH_EXECUTE; 1150 } 1151 1152 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1153 return m_data.GetAddressByteSize(); 1154 } 1155 1156 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1157 Symtab *symtab = GetSymtab(); 1158 if (symtab) { 1159 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1160 if (symbol) { 1161 if (symbol->ValueIsAddress()) { 1162 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1163 if (section_sp) { 1164 const lldb::SectionType section_type = section_sp->GetType(); 1165 switch (section_type) { 1166 case eSectionTypeInvalid: 1167 return eAddressClassUnknown; 1168 1169 case eSectionTypeCode: 1170 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1171 // For ARM we have a bit in the n_desc field of the symbol 1172 // that tells us ARM/Thumb which is bit 0x0008. 1173 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1174 return eAddressClassCodeAlternateISA; 1175 } 1176 return eAddressClassCode; 1177 1178 case eSectionTypeContainer: 1179 return eAddressClassUnknown; 1180 1181 case eSectionTypeData: 1182 case eSectionTypeDataCString: 1183 case eSectionTypeDataCStringPointers: 1184 case eSectionTypeDataSymbolAddress: 1185 case eSectionTypeData4: 1186 case eSectionTypeData8: 1187 case eSectionTypeData16: 1188 case eSectionTypeDataPointers: 1189 case eSectionTypeZeroFill: 1190 case eSectionTypeDataObjCMessageRefs: 1191 case eSectionTypeDataObjCCFStrings: 1192 case eSectionTypeGoSymtab: 1193 return eAddressClassData; 1194 1195 case eSectionTypeDebug: 1196 case eSectionTypeDWARFDebugAbbrev: 1197 case eSectionTypeDWARFDebugAddr: 1198 case eSectionTypeDWARFDebugAranges: 1199 case eSectionTypeDWARFDebugCuIndex: 1200 case eSectionTypeDWARFDebugFrame: 1201 case eSectionTypeDWARFDebugInfo: 1202 case eSectionTypeDWARFDebugLine: 1203 case eSectionTypeDWARFDebugLoc: 1204 case eSectionTypeDWARFDebugMacInfo: 1205 case eSectionTypeDWARFDebugMacro: 1206 case eSectionTypeDWARFDebugPubNames: 1207 case eSectionTypeDWARFDebugPubTypes: 1208 case eSectionTypeDWARFDebugRanges: 1209 case eSectionTypeDWARFDebugStr: 1210 case eSectionTypeDWARFDebugStrOffsets: 1211 case eSectionTypeDWARFAppleNames: 1212 case eSectionTypeDWARFAppleTypes: 1213 case eSectionTypeDWARFAppleNamespaces: 1214 case eSectionTypeDWARFAppleObjC: 1215 return eAddressClassDebug; 1216 1217 case eSectionTypeEHFrame: 1218 case eSectionTypeARMexidx: 1219 case eSectionTypeARMextab: 1220 case eSectionTypeCompactUnwind: 1221 return eAddressClassRuntime; 1222 1223 case eSectionTypeAbsoluteAddress: 1224 case eSectionTypeELFSymbolTable: 1225 case eSectionTypeELFDynamicSymbols: 1226 case eSectionTypeELFRelocationEntries: 1227 case eSectionTypeELFDynamicLinkInfo: 1228 case eSectionTypeOther: 1229 return eAddressClassUnknown; 1230 } 1231 } 1232 } 1233 1234 const SymbolType symbol_type = symbol->GetType(); 1235 switch (symbol_type) { 1236 case eSymbolTypeAny: 1237 return eAddressClassUnknown; 1238 case eSymbolTypeAbsolute: 1239 return eAddressClassUnknown; 1240 1241 case eSymbolTypeCode: 1242 case eSymbolTypeTrampoline: 1243 case eSymbolTypeResolver: 1244 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1245 // For ARM we have a bit in the n_desc field of the symbol 1246 // that tells us ARM/Thumb which is bit 0x0008. 1247 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1248 return eAddressClassCodeAlternateISA; 1249 } 1250 return eAddressClassCode; 1251 1252 case eSymbolTypeData: 1253 return eAddressClassData; 1254 case eSymbolTypeRuntime: 1255 return eAddressClassRuntime; 1256 case eSymbolTypeException: 1257 return eAddressClassRuntime; 1258 case eSymbolTypeSourceFile: 1259 return eAddressClassDebug; 1260 case eSymbolTypeHeaderFile: 1261 return eAddressClassDebug; 1262 case eSymbolTypeObjectFile: 1263 return eAddressClassDebug; 1264 case eSymbolTypeCommonBlock: 1265 return eAddressClassDebug; 1266 case eSymbolTypeBlock: 1267 return eAddressClassDebug; 1268 case eSymbolTypeLocal: 1269 return eAddressClassData; 1270 case eSymbolTypeParam: 1271 return eAddressClassData; 1272 case eSymbolTypeVariable: 1273 return eAddressClassData; 1274 case eSymbolTypeVariableType: 1275 return eAddressClassDebug; 1276 case eSymbolTypeLineEntry: 1277 return eAddressClassDebug; 1278 case eSymbolTypeLineHeader: 1279 return eAddressClassDebug; 1280 case eSymbolTypeScopeBegin: 1281 return eAddressClassDebug; 1282 case eSymbolTypeScopeEnd: 1283 return eAddressClassDebug; 1284 case eSymbolTypeAdditional: 1285 return eAddressClassUnknown; 1286 case eSymbolTypeCompiler: 1287 return eAddressClassDebug; 1288 case eSymbolTypeInstrumentation: 1289 return eAddressClassDebug; 1290 case eSymbolTypeUndefined: 1291 return eAddressClassUnknown; 1292 case eSymbolTypeObjCClass: 1293 return eAddressClassRuntime; 1294 case eSymbolTypeObjCMetaClass: 1295 return eAddressClassRuntime; 1296 case eSymbolTypeObjCIVar: 1297 return eAddressClassRuntime; 1298 case eSymbolTypeReExported: 1299 return eAddressClassRuntime; 1300 } 1301 } 1302 } 1303 return eAddressClassUnknown; 1304 } 1305 1306 Symtab *ObjectFileMachO::GetSymtab() { 1307 ModuleSP module_sp(GetModule()); 1308 if (module_sp) { 1309 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1310 if (m_symtab_ap.get() == NULL) { 1311 m_symtab_ap.reset(new Symtab(this)); 1312 std::lock_guard<std::recursive_mutex> symtab_guard( 1313 m_symtab_ap->GetMutex()); 1314 ParseSymtab(); 1315 m_symtab_ap->Finalize(); 1316 } 1317 } 1318 return m_symtab_ap.get(); 1319 } 1320 1321 bool ObjectFileMachO::IsStripped() { 1322 if (m_dysymtab.cmd == 0) { 1323 ModuleSP module_sp(GetModule()); 1324 if (module_sp) { 1325 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1326 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1327 const lldb::offset_t load_cmd_offset = offset; 1328 1329 load_command lc; 1330 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1331 break; 1332 if (lc.cmd == LC_DYSYMTAB) { 1333 m_dysymtab.cmd = lc.cmd; 1334 m_dysymtab.cmdsize = lc.cmdsize; 1335 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1336 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1337 NULL) { 1338 // Clear m_dysymtab if we were unable to read all items from the 1339 // load command 1340 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1341 } 1342 } 1343 offset = load_cmd_offset + lc.cmdsize; 1344 } 1345 } 1346 } 1347 if (m_dysymtab.cmd) 1348 return m_dysymtab.nlocalsym <= 1; 1349 return false; 1350 } 1351 1352 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1353 EncryptedFileRanges result; 1354 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1355 1356 encryption_info_command encryption_cmd; 1357 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1358 const lldb::offset_t load_cmd_offset = offset; 1359 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1360 break; 1361 1362 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for 1363 // the 3 fields we care about, so treat them the same. 1364 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1365 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1366 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1367 if (encryption_cmd.cryptid != 0) { 1368 EncryptedFileRanges::Entry entry; 1369 entry.SetRangeBase(encryption_cmd.cryptoff); 1370 entry.SetByteSize(encryption_cmd.cryptsize); 1371 result.Append(entry); 1372 } 1373 } 1374 } 1375 offset = load_cmd_offset + encryption_cmd.cmdsize; 1376 } 1377 1378 return result; 1379 } 1380 1381 void ObjectFileMachO::SanitizeSegmentCommand(segment_command_64 &seg_cmd, 1382 uint32_t cmd_idx) { 1383 if (m_length == 0 || seg_cmd.filesize == 0) 1384 return; 1385 1386 if (seg_cmd.fileoff > m_length) { 1387 // We have a load command that says it extends past the end of the file. 1388 // This is likely a corrupt file. We don't have any way to return an error 1389 // condition here (this method was likely invoked from something like 1390 // ObjectFile::GetSectionList()), so we just null out the section contents, 1391 // and dump a message to stdout. The most common case here is core file 1392 // debugging with a truncated file. 1393 const char *lc_segment_name = 1394 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1395 GetModule()->ReportWarning( 1396 "load command %u %s has a fileoff (0x%" PRIx64 1397 ") that extends beyond the end of the file (0x%" PRIx64 1398 "), ignoring this section", 1399 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1400 1401 seg_cmd.fileoff = 0; 1402 seg_cmd.filesize = 0; 1403 } 1404 1405 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1406 // We have a load command that says it extends past the end of the file. 1407 // This is likely a corrupt file. We don't have any way to return an error 1408 // condition here (this method was likely invoked from something like 1409 // ObjectFile::GetSectionList()), so we just null out the section contents, 1410 // and dump a message to stdout. The most common case here is core file 1411 // debugging with a truncated file. 1412 const char *lc_segment_name = 1413 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1414 GetModule()->ReportWarning( 1415 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1416 ") that extends beyond the end of the file (0x%" PRIx64 1417 "), the segment will be truncated to match", 1418 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1419 1420 // Truncate the length 1421 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1422 } 1423 } 1424 1425 static uint32_t GetSegmentPermissions(const segment_command_64 &seg_cmd) { 1426 uint32_t result = 0; 1427 if (seg_cmd.initprot & VM_PROT_READ) 1428 result |= ePermissionsReadable; 1429 if (seg_cmd.initprot & VM_PROT_WRITE) 1430 result |= ePermissionsWritable; 1431 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1432 result |= ePermissionsExecutable; 1433 return result; 1434 } 1435 1436 static lldb::SectionType GetSectionType(uint32_t flags, 1437 ConstString section_name) { 1438 1439 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1440 return eSectionTypeCode; 1441 1442 uint32_t mach_sect_type = flags & SECTION_TYPE; 1443 static ConstString g_sect_name_objc_data("__objc_data"); 1444 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1445 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1446 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1447 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1448 static ConstString g_sect_name_objc_const("__objc_const"); 1449 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1450 static ConstString g_sect_name_cfstring("__cfstring"); 1451 1452 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1453 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1454 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1455 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1456 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1457 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1458 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1459 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1460 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1461 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1462 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1463 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1464 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1465 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1466 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1467 static ConstString g_sect_name_eh_frame("__eh_frame"); 1468 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1469 static ConstString g_sect_name_text("__text"); 1470 static ConstString g_sect_name_data("__data"); 1471 static ConstString g_sect_name_go_symtab("__gosymtab"); 1472 1473 if (section_name == g_sect_name_dwarf_debug_abbrev) 1474 return eSectionTypeDWARFDebugAbbrev; 1475 if (section_name == g_sect_name_dwarf_debug_aranges) 1476 return eSectionTypeDWARFDebugAranges; 1477 if (section_name == g_sect_name_dwarf_debug_frame) 1478 return eSectionTypeDWARFDebugFrame; 1479 if (section_name == g_sect_name_dwarf_debug_info) 1480 return eSectionTypeDWARFDebugInfo; 1481 if (section_name == g_sect_name_dwarf_debug_line) 1482 return eSectionTypeDWARFDebugLine; 1483 if (section_name == g_sect_name_dwarf_debug_loc) 1484 return eSectionTypeDWARFDebugLoc; 1485 if (section_name == g_sect_name_dwarf_debug_macinfo) 1486 return eSectionTypeDWARFDebugMacInfo; 1487 if (section_name == g_sect_name_dwarf_debug_pubnames) 1488 return eSectionTypeDWARFDebugPubNames; 1489 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1490 return eSectionTypeDWARFDebugPubTypes; 1491 if (section_name == g_sect_name_dwarf_debug_ranges) 1492 return eSectionTypeDWARFDebugRanges; 1493 if (section_name == g_sect_name_dwarf_debug_str) 1494 return eSectionTypeDWARFDebugStr; 1495 if (section_name == g_sect_name_dwarf_apple_names) 1496 return eSectionTypeDWARFAppleNames; 1497 if (section_name == g_sect_name_dwarf_apple_types) 1498 return eSectionTypeDWARFAppleTypes; 1499 if (section_name == g_sect_name_dwarf_apple_namespaces) 1500 return eSectionTypeDWARFAppleNamespaces; 1501 if (section_name == g_sect_name_dwarf_apple_objc) 1502 return eSectionTypeDWARFAppleObjC; 1503 if (section_name == g_sect_name_objc_selrefs) 1504 return eSectionTypeDataCStringPointers; 1505 if (section_name == g_sect_name_objc_msgrefs) 1506 return eSectionTypeDataObjCMessageRefs; 1507 if (section_name == g_sect_name_eh_frame) 1508 return eSectionTypeEHFrame; 1509 if (section_name == g_sect_name_compact_unwind) 1510 return eSectionTypeCompactUnwind; 1511 if (section_name == g_sect_name_cfstring) 1512 return eSectionTypeDataObjCCFStrings; 1513 if (section_name == g_sect_name_go_symtab) 1514 return eSectionTypeGoSymtab; 1515 if (section_name == g_sect_name_objc_data || 1516 section_name == g_sect_name_objc_classrefs || 1517 section_name == g_sect_name_objc_superrefs || 1518 section_name == g_sect_name_objc_const || 1519 section_name == g_sect_name_objc_classlist) { 1520 return eSectionTypeDataPointers; 1521 } 1522 1523 switch (mach_sect_type) { 1524 // TODO: categorize sections by other flags for regular sections 1525 case S_REGULAR: 1526 if (section_name == g_sect_name_text) 1527 return eSectionTypeCode; 1528 if (section_name == g_sect_name_data) 1529 return eSectionTypeData; 1530 return eSectionTypeOther; 1531 case S_ZEROFILL: 1532 return eSectionTypeZeroFill; 1533 case S_CSTRING_LITERALS: // section with only literal C strings 1534 return eSectionTypeDataCString; 1535 case S_4BYTE_LITERALS: // section with only 4 byte literals 1536 return eSectionTypeData4; 1537 case S_8BYTE_LITERALS: // section with only 8 byte literals 1538 return eSectionTypeData8; 1539 case S_LITERAL_POINTERS: // section with only pointers to literals 1540 return eSectionTypeDataPointers; 1541 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1542 return eSectionTypeDataPointers; 1543 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1544 return eSectionTypeDataPointers; 1545 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1546 // the reserved2 field 1547 return eSectionTypeCode; 1548 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1549 // initialization 1550 return eSectionTypeDataPointers; 1551 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1552 // termination 1553 return eSectionTypeDataPointers; 1554 case S_COALESCED: 1555 return eSectionTypeOther; 1556 case S_GB_ZEROFILL: 1557 return eSectionTypeZeroFill; 1558 case S_INTERPOSING: // section with only pairs of function pointers for 1559 // interposing 1560 return eSectionTypeCode; 1561 case S_16BYTE_LITERALS: // section with only 16 byte literals 1562 return eSectionTypeData16; 1563 case S_DTRACE_DOF: 1564 return eSectionTypeDebug; 1565 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1566 return eSectionTypeDataPointers; 1567 default: 1568 return eSectionTypeOther; 1569 } 1570 } 1571 1572 struct ObjectFileMachO::SegmentParsingContext { 1573 const EncryptedFileRanges EncryptedRanges; 1574 lldb_private::SectionList &UnifiedList; 1575 uint32_t NextSegmentIdx = 0; 1576 uint32_t NextSectionIdx = 0; 1577 bool FileAddressesChanged = false; 1578 1579 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1580 lldb_private::SectionList &UnifiedList) 1581 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1582 }; 1583 1584 void ObjectFileMachO::ProcessSegmentCommand(const load_command &load_cmd_, 1585 lldb::offset_t offset, 1586 uint32_t cmd_idx, 1587 SegmentParsingContext &context) { 1588 segment_command_64 load_cmd; 1589 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1590 1591 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1592 return; 1593 1594 ModuleSP module_sp = GetModule(); 1595 const bool is_core = GetType() == eTypeCoreFile; 1596 const bool is_dsym = (m_header.filetype == MH_DSYM); 1597 bool add_section = true; 1598 bool add_to_unified = true; 1599 ConstString const_segname( 1600 load_cmd.segname, 1601 std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 1602 1603 SectionSP unified_section_sp( 1604 context.UnifiedList.FindSectionByName(const_segname)); 1605 if (is_dsym && unified_section_sp) { 1606 if (const_segname == GetSegmentNameLINKEDIT()) { 1607 // We need to keep the __LINKEDIT segment private to this object 1608 // file only 1609 add_to_unified = false; 1610 } else { 1611 // This is the dSYM file and this section has already been created 1612 // by the object file, no need to create it. 1613 add_section = false; 1614 } 1615 } 1616 load_cmd.vmaddr = m_data.GetAddress(&offset); 1617 load_cmd.vmsize = m_data.GetAddress(&offset); 1618 load_cmd.fileoff = m_data.GetAddress(&offset); 1619 load_cmd.filesize = m_data.GetAddress(&offset); 1620 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1621 return; 1622 1623 SanitizeSegmentCommand(load_cmd, cmd_idx); 1624 1625 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1626 const bool segment_is_encrypted = 1627 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1628 1629 // Keep a list of mach segments around in case we need to 1630 // get at data that isn't stored in the abstracted Sections. 1631 m_mach_segments.push_back(load_cmd); 1632 1633 // Use a segment ID of the segment index shifted left by 8 so they 1634 // never conflict with any of the sections. 1635 SectionSP segment_sp; 1636 if (add_section && (const_segname || is_core)) { 1637 segment_sp.reset(new Section( 1638 module_sp, // Module to which this section belongs 1639 this, // Object file to which this sections belongs 1640 ++context.NextSegmentIdx 1641 << 8, // Section ID is the 1 based segment index 1642 // shifted right by 8 bits as not to collide 1643 // with any of the 256 section IDs that are 1644 // possible 1645 const_segname, // Name of this section 1646 eSectionTypeContainer, // This section is a container of other 1647 // sections. 1648 load_cmd.vmaddr, // File VM address == addresses as they are 1649 // found in the object file 1650 load_cmd.vmsize, // VM size in bytes of this section 1651 load_cmd.fileoff, // Offset to the data for this section in 1652 // the file 1653 load_cmd.filesize, // Size in bytes of this section as found 1654 // in the file 1655 0, // Segments have no alignment information 1656 load_cmd.flags)); // Flags for this section 1657 1658 segment_sp->SetIsEncrypted(segment_is_encrypted); 1659 m_sections_ap->AddSection(segment_sp); 1660 segment_sp->SetPermissions(segment_permissions); 1661 if (add_to_unified) 1662 context.UnifiedList.AddSection(segment_sp); 1663 } else if (unified_section_sp) { 1664 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1665 // Check to see if the module was read from memory? 1666 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) { 1667 // We have a module that is in memory and needs to have its 1668 // file address adjusted. We need to do this because when we 1669 // load a file from memory, its addresses will be slid already, 1670 // yet the addresses in the new symbol file will still be 1671 // unslid. Since everything is stored as section offset, this 1672 // shouldn't cause any problems. 1673 1674 // Make sure we've parsed the symbol table from the 1675 // ObjectFile before we go around changing its Sections. 1676 module_sp->GetObjectFile()->GetSymtab(); 1677 // eh_frame would present the same problems but we parse that 1678 // on a per-function basis as-needed so it's more difficult to 1679 // remove its use of the Sections. Realistically, the 1680 // environments where this code path will be taken will not 1681 // have eh_frame sections. 1682 1683 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1684 1685 // Notify the module that the section addresses have been 1686 // changed once we're done so any file-address caches can be 1687 // updated. 1688 context.FileAddressesChanged = true; 1689 } 1690 } 1691 m_sections_ap->AddSection(unified_section_sp); 1692 } 1693 1694 struct section_64 sect64; 1695 ::memset(§64, 0, sizeof(sect64)); 1696 // Push a section into our mach sections for the section at 1697 // index zero (NO_SECT) if we don't have any mach sections yet... 1698 if (m_mach_sections.empty()) 1699 m_mach_sections.push_back(sect64); 1700 uint32_t segment_sect_idx; 1701 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1702 1703 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1704 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1705 ++segment_sect_idx) { 1706 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1707 sizeof(sect64.sectname)) == NULL) 1708 break; 1709 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1710 sizeof(sect64.segname)) == NULL) 1711 break; 1712 sect64.addr = m_data.GetAddress(&offset); 1713 sect64.size = m_data.GetAddress(&offset); 1714 1715 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1716 break; 1717 1718 // Keep a list of mach sections around in case we need to 1719 // get at data that isn't stored in the abstracted Sections. 1720 m_mach_sections.push_back(sect64); 1721 1722 if (add_section) { 1723 ConstString section_name( 1724 sect64.sectname, 1725 std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1726 if (!const_segname) { 1727 // We have a segment with no name so we need to conjure up 1728 // segments that correspond to the section's segname if there 1729 // isn't already such a section. If there is such a section, we 1730 // resize the section so that it spans all sections. We also 1731 // mark these sections as fake so address matches don't hit if 1732 // they land in the gaps between the child sections. 1733 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1734 sizeof(sect64.segname)); 1735 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1736 if (segment_sp.get()) { 1737 Section *segment = segment_sp.get(); 1738 // Grow the section size as needed. 1739 const lldb::addr_t sect64_min_addr = sect64.addr; 1740 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1741 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1742 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1743 const lldb::addr_t curr_seg_max_addr = 1744 curr_seg_min_addr + curr_seg_byte_size; 1745 if (sect64_min_addr >= curr_seg_min_addr) { 1746 const lldb::addr_t new_seg_byte_size = 1747 sect64_max_addr - curr_seg_min_addr; 1748 // Only grow the section size if needed 1749 if (new_seg_byte_size > curr_seg_byte_size) 1750 segment->SetByteSize(new_seg_byte_size); 1751 } else { 1752 // We need to change the base address of the segment and 1753 // adjust the child section offsets for all existing 1754 // children. 1755 const lldb::addr_t slide_amount = 1756 sect64_min_addr - curr_seg_min_addr; 1757 segment->Slide(slide_amount, false); 1758 segment->GetChildren().Slide(-slide_amount, false); 1759 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1760 } 1761 1762 // Grow the section size as needed. 1763 if (sect64.offset) { 1764 const lldb::addr_t segment_min_file_offset = 1765 segment->GetFileOffset(); 1766 const lldb::addr_t segment_max_file_offset = 1767 segment_min_file_offset + segment->GetFileSize(); 1768 1769 const lldb::addr_t section_min_file_offset = sect64.offset; 1770 const lldb::addr_t section_max_file_offset = 1771 section_min_file_offset + sect64.size; 1772 const lldb::addr_t new_file_offset = 1773 std::min(section_min_file_offset, segment_min_file_offset); 1774 const lldb::addr_t new_file_size = 1775 std::max(section_max_file_offset, segment_max_file_offset) - 1776 new_file_offset; 1777 segment->SetFileOffset(new_file_offset); 1778 segment->SetFileSize(new_file_size); 1779 } 1780 } else { 1781 // Create a fake section for the section's named segment 1782 segment_sp.reset(new Section( 1783 segment_sp, // Parent section 1784 module_sp, // Module to which this section belongs 1785 this, // Object file to which this section belongs 1786 ++context.NextSegmentIdx 1787 << 8, // Section ID is the 1 based segment index 1788 // shifted right by 8 bits as not to 1789 // collide with any of the 256 section IDs 1790 // that are possible 1791 const_segname, // Name of this section 1792 eSectionTypeContainer, // This section is a container of 1793 // other sections. 1794 sect64.addr, // File VM address == addresses as they are 1795 // found in the object file 1796 sect64.size, // VM size in bytes of this section 1797 sect64.offset, // Offset to the data for this section in 1798 // the file 1799 sect64.offset ? sect64.size : 0, // Size in bytes of 1800 // this section as 1801 // found in the file 1802 sect64.align, 1803 load_cmd.flags)); // Flags for this section 1804 segment_sp->SetIsFake(true); 1805 segment_sp->SetPermissions(segment_permissions); 1806 m_sections_ap->AddSection(segment_sp); 1807 if (add_to_unified) 1808 context.UnifiedList.AddSection(segment_sp); 1809 segment_sp->SetIsEncrypted(segment_is_encrypted); 1810 } 1811 } 1812 assert(segment_sp.get()); 1813 1814 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1815 1816 SectionSP section_sp(new Section( 1817 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1818 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1819 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1820 sect64.flags)); 1821 // Set the section to be encrypted to match the segment 1822 1823 bool section_is_encrypted = false; 1824 if (!segment_is_encrypted && load_cmd.filesize != 0) 1825 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1826 sect64.offset) != NULL; 1827 1828 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1829 section_sp->SetPermissions(segment_permissions); 1830 segment_sp->GetChildren().AddSection(section_sp); 1831 1832 if (segment_sp->IsFake()) { 1833 segment_sp.reset(); 1834 const_segname.Clear(); 1835 } 1836 } 1837 } 1838 if (segment_sp && is_dsym) { 1839 if (first_segment_sectID <= context.NextSectionIdx) { 1840 lldb::user_id_t sect_uid; 1841 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1842 ++sect_uid) { 1843 SectionSP curr_section_sp( 1844 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1845 SectionSP next_section_sp; 1846 if (sect_uid + 1 <= context.NextSectionIdx) 1847 next_section_sp = 1848 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1849 1850 if (curr_section_sp.get()) { 1851 if (curr_section_sp->GetByteSize() == 0) { 1852 if (next_section_sp.get() != NULL) 1853 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1854 curr_section_sp->GetFileAddress()); 1855 else 1856 curr_section_sp->SetByteSize(load_cmd.vmsize); 1857 } 1858 } 1859 } 1860 } 1861 } 1862 } 1863 1864 void ObjectFileMachO::ProcessDysymtabCommand(const load_command &load_cmd, 1865 lldb::offset_t offset) { 1866 m_dysymtab.cmd = load_cmd.cmd; 1867 m_dysymtab.cmdsize = load_cmd.cmdsize; 1868 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1869 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1870 } 1871 1872 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1873 if (m_sections_ap) 1874 return; 1875 1876 m_sections_ap.reset(new SectionList()); 1877 1878 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1879 // bool dump_sections = false; 1880 ModuleSP module_sp(GetModule()); 1881 1882 offset = MachHeaderSizeFromMagic(m_header.magic); 1883 1884 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1885 struct load_command load_cmd; 1886 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1887 const lldb::offset_t load_cmd_offset = offset; 1888 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1889 break; 1890 1891 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1892 ProcessSegmentCommand(load_cmd, offset, i, context); 1893 else if (load_cmd.cmd == LC_DYSYMTAB) 1894 ProcessDysymtabCommand(load_cmd, offset); 1895 1896 offset = load_cmd_offset + load_cmd.cmdsize; 1897 } 1898 1899 if (context.FileAddressesChanged && module_sp) 1900 module_sp->SectionFileAddressesChanged(); 1901 } 1902 1903 class MachSymtabSectionInfo { 1904 public: 1905 MachSymtabSectionInfo(SectionList *section_list) 1906 : m_section_list(section_list), m_section_infos() { 1907 // Get the number of sections down to a depth of 1 to include 1908 // all segments and their sections, but no other sections that 1909 // may be added for debug map or 1910 m_section_infos.resize(section_list->GetNumSections(1)); 1911 } 1912 1913 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1914 if (n_sect == 0) 1915 return SectionSP(); 1916 if (n_sect < m_section_infos.size()) { 1917 if (!m_section_infos[n_sect].section_sp) { 1918 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1919 m_section_infos[n_sect].section_sp = section_sp; 1920 if (section_sp) { 1921 m_section_infos[n_sect].vm_range.SetBaseAddress( 1922 section_sp->GetFileAddress()); 1923 m_section_infos[n_sect].vm_range.SetByteSize( 1924 section_sp->GetByteSize()); 1925 } else { 1926 Host::SystemLog(Host::eSystemLogError, 1927 "error: unable to find section for section %u\n", 1928 n_sect); 1929 } 1930 } 1931 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1932 // Symbol is in section. 1933 return m_section_infos[n_sect].section_sp; 1934 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1935 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1936 file_addr) { 1937 // Symbol is in section with zero size, but has the same start 1938 // address as the section. This can happen with linker symbols 1939 // (symbols that start with the letter 'l' or 'L'. 1940 return m_section_infos[n_sect].section_sp; 1941 } 1942 } 1943 return m_section_list->FindSectionContainingFileAddress(file_addr); 1944 } 1945 1946 protected: 1947 struct SectionInfo { 1948 SectionInfo() : vm_range(), section_sp() {} 1949 1950 VMRange vm_range; 1951 SectionSP section_sp; 1952 }; 1953 SectionList *m_section_list; 1954 std::vector<SectionInfo> m_section_infos; 1955 }; 1956 1957 struct TrieEntry { 1958 TrieEntry() 1959 : name(), address(LLDB_INVALID_ADDRESS), flags(0), other(0), 1960 import_name() {} 1961 1962 void Clear() { 1963 name.Clear(); 1964 address = LLDB_INVALID_ADDRESS; 1965 flags = 0; 1966 other = 0; 1967 import_name.Clear(); 1968 } 1969 1970 void Dump() const { 1971 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1972 static_cast<unsigned long long>(address), 1973 static_cast<unsigned long long>(flags), 1974 static_cast<unsigned long long>(other), name.GetCString()); 1975 if (import_name) 1976 printf(" -> \"%s\"\n", import_name.GetCString()); 1977 else 1978 printf("\n"); 1979 } 1980 ConstString name; 1981 uint64_t address; 1982 uint64_t flags; 1983 uint64_t other; 1984 ConstString import_name; 1985 }; 1986 1987 struct TrieEntryWithOffset { 1988 lldb::offset_t nodeOffset; 1989 TrieEntry entry; 1990 1991 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1992 1993 void Dump(uint32_t idx) const { 1994 printf("[%3u] 0x%16.16llx: ", idx, 1995 static_cast<unsigned long long>(nodeOffset)); 1996 entry.Dump(); 1997 } 1998 1999 bool operator<(const TrieEntryWithOffset &other) const { 2000 return (nodeOffset < other.nodeOffset); 2001 } 2002 }; 2003 2004 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2005 const bool is_arm, 2006 std::vector<llvm::StringRef> &nameSlices, 2007 std::set<lldb::addr_t> &resolver_addresses, 2008 std::vector<TrieEntryWithOffset> &output) { 2009 if (!data.ValidOffset(offset)) 2010 return true; 2011 2012 const uint64_t terminalSize = data.GetULEB128(&offset); 2013 lldb::offset_t children_offset = offset + terminalSize; 2014 if (terminalSize != 0) { 2015 TrieEntryWithOffset e(offset); 2016 e.entry.flags = data.GetULEB128(&offset); 2017 const char *import_name = NULL; 2018 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2019 e.entry.address = 0; 2020 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2021 import_name = data.GetCStr(&offset); 2022 } else { 2023 e.entry.address = data.GetULEB128(&offset); 2024 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2025 e.entry.other = data.GetULEB128(&offset); 2026 uint64_t resolver_addr = e.entry.other; 2027 if (is_arm) 2028 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2029 resolver_addresses.insert(resolver_addr); 2030 } else 2031 e.entry.other = 0; 2032 } 2033 // Only add symbols that are reexport symbols with a valid import name 2034 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && 2035 import_name[0]) { 2036 std::string name; 2037 if (!nameSlices.empty()) { 2038 for (auto name_slice : nameSlices) 2039 name.append(name_slice.data(), name_slice.size()); 2040 } 2041 if (name.size() > 1) { 2042 // Skip the leading '_' 2043 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2044 } 2045 if (import_name) { 2046 // Skip the leading '_' 2047 e.entry.import_name.SetCString(import_name + 1); 2048 } 2049 output.push_back(e); 2050 } 2051 } 2052 2053 const uint8_t childrenCount = data.GetU8(&children_offset); 2054 for (uint8_t i = 0; i < childrenCount; ++i) { 2055 const char *cstr = data.GetCStr(&children_offset); 2056 if (cstr) 2057 nameSlices.push_back(llvm::StringRef(cstr)); 2058 else 2059 return false; // Corrupt data 2060 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2061 if (childNodeOffset) { 2062 if (!ParseTrieEntries(data, childNodeOffset, is_arm, nameSlices, 2063 resolver_addresses, output)) { 2064 return false; 2065 } 2066 } 2067 nameSlices.pop_back(); 2068 } 2069 return true; 2070 } 2071 2072 // Read the UUID out of a dyld_shared_cache file on-disk. 2073 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2074 const ByteOrder byte_order, 2075 const uint32_t addr_byte_size) { 2076 UUID dsc_uuid; 2077 DataBufferSP DscData = MapFileData( 2078 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2079 if (!DscData) 2080 return dsc_uuid; 2081 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2082 2083 char version_str[7]; 2084 lldb::offset_t offset = 0; 2085 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2086 version_str[6] = '\0'; 2087 if (strcmp(version_str, "dyld_v") == 0) { 2088 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2089 uint8_t uuid_bytes[sizeof(uuid_t)]; 2090 memcpy(uuid_bytes, dsc_header_data.GetData(&offset, sizeof(uuid_t)), 2091 sizeof(uuid_t)); 2092 dsc_uuid.SetBytes(uuid_bytes); 2093 } 2094 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2095 if (log && dsc_uuid.IsValid()) { 2096 log->Printf("Shared cache %s has UUID %s", dyld_shared_cache.GetPath().c_str(), 2097 dsc_uuid.GetAsString().c_str()); 2098 } 2099 return dsc_uuid; 2100 } 2101 2102 size_t ObjectFileMachO::ParseSymtab() { 2103 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2104 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2105 m_file.GetFilename().AsCString("")); 2106 ModuleSP module_sp(GetModule()); 2107 if (!module_sp) 2108 return 0; 2109 2110 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2111 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2112 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2113 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2114 FunctionStarts function_starts; 2115 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2116 uint32_t i; 2117 FileSpecList dylib_files; 2118 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2119 static const llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2120 static const llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2121 static const llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2122 2123 for (i = 0; i < m_header.ncmds; ++i) { 2124 const lldb::offset_t cmd_offset = offset; 2125 // Read in the load command and load command size 2126 struct load_command lc; 2127 if (m_data.GetU32(&offset, &lc, 2) == NULL) 2128 break; 2129 // Watch for the symbol table load command 2130 switch (lc.cmd) { 2131 case LC_SYMTAB: 2132 symtab_load_command.cmd = lc.cmd; 2133 symtab_load_command.cmdsize = lc.cmdsize; 2134 // Read in the rest of the symtab load command 2135 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2136 0) // fill in symoff, nsyms, stroff, strsize fields 2137 return 0; 2138 if (symtab_load_command.symoff == 0) { 2139 if (log) 2140 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2141 return 0; 2142 } 2143 2144 if (symtab_load_command.stroff == 0) { 2145 if (log) 2146 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2147 return 0; 2148 } 2149 2150 if (symtab_load_command.nsyms == 0) { 2151 if (log) 2152 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2153 return 0; 2154 } 2155 2156 if (symtab_load_command.strsize == 0) { 2157 if (log) 2158 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2159 return 0; 2160 } 2161 break; 2162 2163 case LC_DYLD_INFO: 2164 case LC_DYLD_INFO_ONLY: 2165 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2166 dyld_info.cmd = lc.cmd; 2167 dyld_info.cmdsize = lc.cmdsize; 2168 } else { 2169 memset(&dyld_info, 0, sizeof(dyld_info)); 2170 } 2171 break; 2172 2173 case LC_LOAD_DYLIB: 2174 case LC_LOAD_WEAK_DYLIB: 2175 case LC_REEXPORT_DYLIB: 2176 case LC_LOADFVMLIB: 2177 case LC_LOAD_UPWARD_DYLIB: { 2178 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2179 const char *path = m_data.PeekCStr(name_offset); 2180 if (path) { 2181 FileSpec file_spec(path, false); 2182 // Strip the path if there is @rpath, @executable, etc so we just use 2183 // the basename 2184 if (path[0] == '@') 2185 file_spec.GetDirectory().Clear(); 2186 2187 if (lc.cmd == LC_REEXPORT_DYLIB) { 2188 m_reexported_dylibs.AppendIfUnique(file_spec); 2189 } 2190 2191 dylib_files.Append(file_spec); 2192 } 2193 } break; 2194 2195 case LC_FUNCTION_STARTS: 2196 function_starts_load_command.cmd = lc.cmd; 2197 function_starts_load_command.cmdsize = lc.cmdsize; 2198 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2199 NULL) // fill in symoff, nsyms, stroff, strsize fields 2200 memset(&function_starts_load_command, 0, 2201 sizeof(function_starts_load_command)); 2202 break; 2203 2204 default: 2205 break; 2206 } 2207 offset = cmd_offset + lc.cmdsize; 2208 } 2209 2210 if (symtab_load_command.cmd) { 2211 Symtab *symtab = m_symtab_ap.get(); 2212 SectionList *section_list = GetSectionList(); 2213 if (section_list == NULL) 2214 return 0; 2215 2216 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2217 const ByteOrder byte_order = m_data.GetByteOrder(); 2218 bool bit_width_32 = addr_byte_size == 4; 2219 const size_t nlist_byte_size = 2220 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2221 2222 DataExtractor nlist_data(NULL, 0, byte_order, addr_byte_size); 2223 DataExtractor strtab_data(NULL, 0, byte_order, addr_byte_size); 2224 DataExtractor function_starts_data(NULL, 0, byte_order, addr_byte_size); 2225 DataExtractor indirect_symbol_index_data(NULL, 0, byte_order, 2226 addr_byte_size); 2227 DataExtractor dyld_trie_data(NULL, 0, byte_order, addr_byte_size); 2228 2229 const addr_t nlist_data_byte_size = 2230 symtab_load_command.nsyms * nlist_byte_size; 2231 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2232 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2233 2234 ProcessSP process_sp(m_process_wp.lock()); 2235 Process *process = process_sp.get(); 2236 2237 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2238 2239 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2240 Target &target = process->GetTarget(); 2241 2242 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2243 2244 SectionSP linkedit_section_sp( 2245 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2246 // Reading mach file from memory in a process or core file... 2247 2248 if (linkedit_section_sp) { 2249 addr_t linkedit_load_addr = 2250 linkedit_section_sp->GetLoadBaseAddress(&target); 2251 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2252 // We might be trying to access the symbol table before the 2253 // __LINKEDIT's load 2254 // address has been set in the target. We can't fail to read the 2255 // symbol table, 2256 // so calculate the right address manually 2257 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2258 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2259 } 2260 2261 const addr_t linkedit_file_offset = 2262 linkedit_section_sp->GetFileOffset(); 2263 const addr_t symoff_addr = linkedit_load_addr + 2264 symtab_load_command.symoff - 2265 linkedit_file_offset; 2266 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2267 linkedit_file_offset; 2268 2269 bool data_was_read = false; 2270 2271 #if defined(__APPLE__) && \ 2272 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2273 if (m_header.flags & 0x80000000u && 2274 process->GetAddressByteSize() == sizeof(void *)) { 2275 // This mach-o memory file is in the dyld shared cache. If this 2276 // program is not remote and this is iOS, then this process will 2277 // share the same shared cache as the process we are debugging and 2278 // we can read the entire __LINKEDIT from the address space in this 2279 // process. This is a needed optimization that is used for local iOS 2280 // debugging only since all shared libraries in the shared cache do 2281 // not have corresponding files that exist in the file system of the 2282 // device. They have been combined into a single file. This means we 2283 // always have to load these files from memory. All of the symbol and 2284 // string tables from all of the __LINKEDIT sections from the shared 2285 // libraries in the shared cache have been merged into a single large 2286 // symbol and string table. Reading all of this symbol and string 2287 // table 2288 // data across can slow down debug launch times, so we optimize this 2289 // by 2290 // reading the memory for the __LINKEDIT section from this process. 2291 2292 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 2293 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 2294 bool use_lldb_cache = true; 2295 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2296 lldb_shared_cache != process_shared_cache) { 2297 use_lldb_cache = false; 2298 ModuleSP module_sp(GetModule()); 2299 if (module_sp) 2300 module_sp->ReportWarning("shared cache in process does not match " 2301 "lldb's own shared cache, startup will " 2302 "be slow."); 2303 } 2304 2305 PlatformSP platform_sp(target.GetPlatform()); 2306 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2307 data_was_read = true; 2308 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2309 eByteOrderLittle); 2310 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2311 eByteOrderLittle); 2312 if (function_starts_load_command.cmd) { 2313 const addr_t func_start_addr = 2314 linkedit_load_addr + function_starts_load_command.dataoff - 2315 linkedit_file_offset; 2316 function_starts_data.SetData( 2317 (void *)func_start_addr, 2318 function_starts_load_command.datasize, eByteOrderLittle); 2319 } 2320 } 2321 } 2322 #endif 2323 2324 if (!data_was_read) { 2325 // Always load dyld - the dynamic linker - from memory if we didn't 2326 // find a binary anywhere else. 2327 // lldb will not register dylib/framework/bundle loads/unloads if we 2328 // don't have the dyld symbols, 2329 // we force dyld to load from memory despite the user's 2330 // target.memory-module-load-level setting. 2331 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2332 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2333 DataBufferSP nlist_data_sp( 2334 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2335 if (nlist_data_sp) 2336 nlist_data.SetData(nlist_data_sp, 0, 2337 nlist_data_sp->GetByteSize()); 2338 // Load strings individually from memory when loading from memory 2339 // since shared cache 2340 // string tables contain strings for all symbols from all shared 2341 // cached libraries 2342 // DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, 2343 // strtab_data_byte_size)); 2344 // if (strtab_data_sp) 2345 // strtab_data.SetData (strtab_data_sp, 0, 2346 // strtab_data_sp->GetByteSize()); 2347 if (m_dysymtab.nindirectsyms != 0) { 2348 const addr_t indirect_syms_addr = linkedit_load_addr + 2349 m_dysymtab.indirectsymoff - 2350 linkedit_file_offset; 2351 DataBufferSP indirect_syms_data_sp( 2352 ReadMemory(process_sp, indirect_syms_addr, 2353 m_dysymtab.nindirectsyms * 4)); 2354 if (indirect_syms_data_sp) 2355 indirect_symbol_index_data.SetData( 2356 indirect_syms_data_sp, 0, 2357 indirect_syms_data_sp->GetByteSize()); 2358 } 2359 } else if (memory_module_load_level >= 2360 eMemoryModuleLoadLevelPartial) { 2361 if (function_starts_load_command.cmd) { 2362 const addr_t func_start_addr = 2363 linkedit_load_addr + function_starts_load_command.dataoff - 2364 linkedit_file_offset; 2365 DataBufferSP func_start_data_sp( 2366 ReadMemory(process_sp, func_start_addr, 2367 function_starts_load_command.datasize)); 2368 if (func_start_data_sp) 2369 function_starts_data.SetData(func_start_data_sp, 0, 2370 func_start_data_sp->GetByteSize()); 2371 } 2372 } 2373 } 2374 } 2375 } else { 2376 nlist_data.SetData(m_data, symtab_load_command.symoff, 2377 nlist_data_byte_size); 2378 strtab_data.SetData(m_data, symtab_load_command.stroff, 2379 strtab_data_byte_size); 2380 2381 if (dyld_info.export_size > 0) { 2382 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2383 dyld_info.export_size); 2384 } 2385 2386 if (m_dysymtab.nindirectsyms != 0) { 2387 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2388 m_dysymtab.nindirectsyms * 4); 2389 } 2390 if (function_starts_load_command.cmd) { 2391 function_starts_data.SetData(m_data, 2392 function_starts_load_command.dataoff, 2393 function_starts_load_command.datasize); 2394 } 2395 } 2396 2397 if (nlist_data.GetByteSize() == 0 && 2398 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2399 if (log) 2400 module_sp->LogMessage(log, "failed to read nlist data"); 2401 return 0; 2402 } 2403 2404 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2405 if (!have_strtab_data) { 2406 if (process) { 2407 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2408 if (log) 2409 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2410 return 0; 2411 } 2412 } else { 2413 if (log) 2414 module_sp->LogMessage(log, "failed to read strtab data"); 2415 return 0; 2416 } 2417 } 2418 2419 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2420 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2421 const ConstString &g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2422 const ConstString &g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2423 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2424 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2425 SectionSP text_section_sp( 2426 section_list->FindSectionByName(g_segment_name_TEXT)); 2427 SectionSP data_section_sp( 2428 section_list->FindSectionByName(g_segment_name_DATA)); 2429 SectionSP data_dirty_section_sp( 2430 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2431 SectionSP data_const_section_sp( 2432 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2433 SectionSP objc_section_sp( 2434 section_list->FindSectionByName(g_segment_name_OBJC)); 2435 SectionSP eh_frame_section_sp; 2436 if (text_section_sp.get()) 2437 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2438 g_section_name_eh_frame); 2439 else 2440 eh_frame_section_sp = 2441 section_list->FindSectionByName(g_section_name_eh_frame); 2442 2443 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2444 2445 // lldb works best if it knows the start address of all functions in a 2446 // module. 2447 // Linker symbols or debug info are normally the best source of information 2448 // for start addr / size but 2449 // they may be stripped in a released binary. 2450 // Two additional sources of information exist in Mach-O binaries: 2451 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2452 // function's start address in the 2453 // binary, relative to the text section. 2454 // eh_frame - the eh_frame FDEs have the start addr & size of 2455 // each function 2456 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2457 // all modern binaries. 2458 // Binaries built to run on older releases may need to use eh_frame 2459 // information. 2460 2461 if (text_section_sp && function_starts_data.GetByteSize()) { 2462 FunctionStarts::Entry function_start_entry; 2463 function_start_entry.data = false; 2464 lldb::offset_t function_start_offset = 0; 2465 function_start_entry.addr = text_section_sp->GetFileAddress(); 2466 uint64_t delta; 2467 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2468 0) { 2469 // Now append the current entry 2470 function_start_entry.addr += delta; 2471 function_starts.Append(function_start_entry); 2472 } 2473 } else { 2474 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2475 // load command claiming an eh_frame 2476 // but it doesn't actually have the eh_frame content. And if we have a 2477 // dSYM, we don't need to do any 2478 // of this fill-in-the-missing-symbols works anyway - the debug info 2479 // should give us all the functions in 2480 // the module. 2481 if (text_section_sp.get() && eh_frame_section_sp.get() && 2482 m_type != eTypeDebugInfo) { 2483 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2484 DWARFCallFrameInfo::EH); 2485 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2486 eh_frame.GetFunctionAddressAndSizeVector(functions); 2487 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2488 size_t count = functions.GetSize(); 2489 for (size_t i = 0; i < count; ++i) { 2490 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2491 functions.GetEntryAtIndex(i); 2492 if (func) { 2493 FunctionStarts::Entry function_start_entry; 2494 function_start_entry.addr = func->base - text_base_addr; 2495 function_starts.Append(function_start_entry); 2496 } 2497 } 2498 } 2499 } 2500 2501 const size_t function_starts_count = function_starts.GetSize(); 2502 2503 // For user process binaries (executables, dylibs, frameworks, bundles), if 2504 // we don't have 2505 // LC_FUNCTION_STARTS/eh_frame section in this binary, we're going to assume 2506 // the binary 2507 // has been stripped. Don't allow assembly language instruction emulation 2508 // because we don't 2509 // know proper function start boundaries. 2510 // 2511 // For all other types of binaries (kernels, stand-alone bare board 2512 // binaries, kexts), they 2513 // may not have LC_FUNCTION_STARTS / eh_frame sections - we should not make 2514 // any assumptions 2515 // about them based on that. 2516 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2517 m_allow_assembly_emulation_unwind_plans = false; 2518 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2519 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2520 2521 if (unwind_or_symbol_log) 2522 module_sp->LogMessage( 2523 unwind_or_symbol_log, 2524 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2525 } 2526 2527 const user_id_t TEXT_eh_frame_sectID = 2528 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2529 : static_cast<user_id_t>(NO_SECT); 2530 2531 lldb::offset_t nlist_data_offset = 0; 2532 2533 uint32_t N_SO_index = UINT32_MAX; 2534 2535 MachSymtabSectionInfo section_info(section_list); 2536 std::vector<uint32_t> N_FUN_indexes; 2537 std::vector<uint32_t> N_NSYM_indexes; 2538 std::vector<uint32_t> N_INCL_indexes; 2539 std::vector<uint32_t> N_BRAC_indexes; 2540 std::vector<uint32_t> N_COMM_indexes; 2541 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2542 typedef std::map<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2543 typedef std::map<const char *, uint32_t> ConstNameToSymbolIndexMap; 2544 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2545 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2546 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2547 // Any symbols that get merged into another will get an entry 2548 // in this map so we know 2549 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2550 uint32_t nlist_idx = 0; 2551 Symbol *symbol_ptr = NULL; 2552 2553 uint32_t sym_idx = 0; 2554 Symbol *sym = NULL; 2555 size_t num_syms = 0; 2556 std::string memory_symbol_name; 2557 uint32_t unmapped_local_symbols_found = 0; 2558 2559 std::vector<TrieEntryWithOffset> trie_entries; 2560 std::set<lldb::addr_t> resolver_addresses; 2561 2562 if (dyld_trie_data.GetByteSize() > 0) { 2563 std::vector<llvm::StringRef> nameSlices; 2564 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, 2565 resolver_addresses, trie_entries); 2566 2567 ConstString text_segment_name("__TEXT"); 2568 SectionSP text_segment_sp = 2569 GetSectionList()->FindSectionByName(text_segment_name); 2570 if (text_segment_sp) { 2571 const lldb::addr_t text_segment_file_addr = 2572 text_segment_sp->GetFileAddress(); 2573 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2574 for (auto &e : trie_entries) 2575 e.entry.address += text_segment_file_addr; 2576 } 2577 } 2578 } 2579 2580 typedef std::set<ConstString> IndirectSymbols; 2581 IndirectSymbols indirect_symbol_names; 2582 2583 #if defined(__APPLE__) && \ 2584 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2585 2586 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2587 // optimized by moving LOCAL 2588 // symbols out of the memory mapped portion of the DSC. The symbol 2589 // information has all been retained, 2590 // but it isn't available in the normal nlist data. However, there *are* 2591 // duplicate entries of *some* 2592 // LOCAL symbols in the normal nlist data. To handle this situation 2593 // correctly, we must first attempt 2594 // to parse any DSC unmapped symbol information. If we find any, we set a 2595 // flag that tells the normal 2596 // nlist parser to ignore all LOCAL symbols. 2597 2598 if (m_header.flags & 0x80000000u) { 2599 // Before we can start mapping the DSC, we need to make certain the target 2600 // process is actually 2601 // using the cache we can find. 2602 2603 // Next we need to determine the correct path for the dyld shared cache. 2604 2605 ArchSpec header_arch; 2606 GetArchitecture(header_arch); 2607 char dsc_path[PATH_MAX]; 2608 char dsc_path_development[PATH_MAX]; 2609 2610 snprintf( 2611 dsc_path, sizeof(dsc_path), "%s%s%s", 2612 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2613 */ 2614 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2615 header_arch.GetArchitectureName()); 2616 2617 snprintf( 2618 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2619 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2620 */ 2621 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2622 header_arch.GetArchitectureName(), ".development"); 2623 2624 FileSpec dsc_nondevelopment_filespec(dsc_path, false); 2625 FileSpec dsc_development_filespec(dsc_path_development, false); 2626 FileSpec dsc_filespec; 2627 2628 UUID dsc_uuid; 2629 UUID process_shared_cache_uuid; 2630 2631 if (process) { 2632 process_shared_cache_uuid = GetProcessSharedCacheUUID(process); 2633 } 2634 2635 // First see if we can find an exact match for the inferior process shared 2636 // cache UUID in 2637 // the development or non-development shared caches on disk. 2638 if (process_shared_cache_uuid.IsValid()) { 2639 if (dsc_development_filespec.Exists()) { 2640 UUID dsc_development_uuid = GetSharedCacheUUID( 2641 dsc_development_filespec, byte_order, addr_byte_size); 2642 if (dsc_development_uuid.IsValid() && 2643 dsc_development_uuid == process_shared_cache_uuid) { 2644 dsc_filespec = dsc_development_filespec; 2645 dsc_uuid = dsc_development_uuid; 2646 } 2647 } 2648 if (!dsc_uuid.IsValid() && dsc_nondevelopment_filespec.Exists()) { 2649 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2650 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2651 if (dsc_nondevelopment_uuid.IsValid() && 2652 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2653 dsc_filespec = dsc_nondevelopment_filespec; 2654 dsc_uuid = dsc_nondevelopment_uuid; 2655 } 2656 } 2657 } 2658 2659 // Failing a UUID match, prefer the development dyld_shared cache if both 2660 // are present. 2661 if (!dsc_filespec.Exists()) { 2662 if (dsc_development_filespec.Exists()) { 2663 dsc_filespec = dsc_development_filespec; 2664 } else { 2665 dsc_filespec = dsc_nondevelopment_filespec; 2666 } 2667 } 2668 2669 /* The dyld_cache_header has a pointer to the 2670 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2671 The dyld_cache_local_symbols_info structure gives us three things: 2672 1. The start and count of the nlist records in the dyld_shared_cache 2673 file 2674 2. The start and size of the strings for these nlist records 2675 3. The start and count of dyld_cache_local_symbols_entry entries 2676 2677 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2678 dyld shared cache. 2679 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2680 the dyld shared cache. 2681 The dyld_cache_local_symbols_entry also lists the start of this 2682 dylib/framework's nlist records 2683 and the count of how many nlist records there are for this 2684 dylib/framework. 2685 */ 2686 2687 // Process the dyld shared cache header to find the unmapped symbols 2688 2689 DataBufferSP dsc_data_sp = MapFileData( 2690 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2691 if (!dsc_uuid.IsValid()) { 2692 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2693 } 2694 if (dsc_data_sp) { 2695 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2696 2697 bool uuid_match = true; 2698 if (dsc_uuid.IsValid() && process) { 2699 if (process_shared_cache_uuid.IsValid() && 2700 dsc_uuid != process_shared_cache_uuid) { 2701 // The on-disk dyld_shared_cache file is not the same as the one in 2702 // this 2703 // process' memory, don't use it. 2704 uuid_match = false; 2705 ModuleSP module_sp(GetModule()); 2706 if (module_sp) 2707 module_sp->ReportWarning("process shared cache does not match " 2708 "on-disk dyld_shared_cache file, some " 2709 "symbol names will be missing."); 2710 } 2711 } 2712 2713 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2714 2715 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2716 2717 // If the mappingOffset points to a location inside the header, we've 2718 // opened an old dyld shared cache, and should not proceed further. 2719 if (uuid_match && 2720 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2721 2722 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2723 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2724 mappingOffset); 2725 2726 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2727 byte_order, addr_byte_size); 2728 offset = 0; 2729 2730 // The File addresses (from the in-memory Mach-O load commands) for 2731 // the shared libraries 2732 // in the shared library cache need to be adjusted by an offset to 2733 // match up with the 2734 // dylibOffset identifying field in the 2735 // dyld_cache_local_symbol_entry's. This offset is 2736 // recorded in mapping_offset_value. 2737 const uint64_t mapping_offset_value = 2738 dsc_mapping_info_data.GetU64(&offset); 2739 2740 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, 2741 localSymbolsOffset); 2742 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2743 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2744 2745 if (localSymbolsOffset && localSymbolsSize) { 2746 // Map the local symbols 2747 DataBufferSP dsc_local_symbols_data_sp = 2748 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2749 2750 if (dsc_local_symbols_data_sp) { 2751 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2752 byte_order, addr_byte_size); 2753 2754 offset = 0; 2755 2756 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2757 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2758 UndefinedNameToDescMap undefined_name_to_desc; 2759 SymbolIndexToName reexport_shlib_needs_fixup; 2760 2761 // Read the local_symbols_infos struct in one shot 2762 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2763 dsc_local_symbols_data.GetU32(&offset, 2764 &local_symbols_info.nlistOffset, 6); 2765 2766 SectionSP text_section_sp( 2767 section_list->FindSectionByName(GetSegmentNameTEXT())); 2768 2769 uint32_t header_file_offset = 2770 (text_section_sp->GetFileAddress() - mapping_offset_value); 2771 2772 offset = local_symbols_info.entriesOffset; 2773 for (uint32_t entry_index = 0; 2774 entry_index < local_symbols_info.entriesCount; 2775 entry_index++) { 2776 struct lldb_copy_dyld_cache_local_symbols_entry 2777 local_symbols_entry; 2778 local_symbols_entry.dylibOffset = 2779 dsc_local_symbols_data.GetU32(&offset); 2780 local_symbols_entry.nlistStartIndex = 2781 dsc_local_symbols_data.GetU32(&offset); 2782 local_symbols_entry.nlistCount = 2783 dsc_local_symbols_data.GetU32(&offset); 2784 2785 if (header_file_offset == local_symbols_entry.dylibOffset) { 2786 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2787 2788 // The normal nlist code cannot correctly size the Symbols 2789 // array, we need to allocate it here. 2790 sym = symtab->Resize( 2791 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2792 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2793 num_syms = symtab->GetNumSymbols(); 2794 2795 nlist_data_offset = 2796 local_symbols_info.nlistOffset + 2797 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2798 uint32_t string_table_offset = 2799 local_symbols_info.stringsOffset; 2800 2801 for (uint32_t nlist_index = 0; 2802 nlist_index < local_symbols_entry.nlistCount; 2803 nlist_index++) { 2804 ///////////////////////////// 2805 { 2806 struct nlist_64 nlist; 2807 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize( 2808 nlist_data_offset, nlist_byte_size)) 2809 break; 2810 2811 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked( 2812 &nlist_data_offset); 2813 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked( 2814 &nlist_data_offset); 2815 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked( 2816 &nlist_data_offset); 2817 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked( 2818 &nlist_data_offset); 2819 nlist.n_value = 2820 dsc_local_symbols_data.GetAddress_unchecked( 2821 &nlist_data_offset); 2822 2823 SymbolType type = eSymbolTypeInvalid; 2824 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2825 string_table_offset + nlist.n_strx); 2826 2827 if (symbol_name == NULL) { 2828 // No symbol should be NULL, even the symbols with no 2829 // string values should have an offset zero which points 2830 // to an empty C-string 2831 Host::SystemLog( 2832 Host::eSystemLogError, 2833 "error: DSC unmapped local symbol[%u] has invalid " 2834 "string table offset 0x%x in %s, ignoring symbol\n", 2835 entry_index, nlist.n_strx, 2836 module_sp->GetFileSpec().GetPath().c_str()); 2837 continue; 2838 } 2839 if (symbol_name[0] == '\0') 2840 symbol_name = NULL; 2841 2842 const char *symbol_name_non_abi_mangled = NULL; 2843 2844 SectionSP symbol_section; 2845 uint32_t symbol_byte_size = 0; 2846 bool add_nlist = true; 2847 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2848 bool demangled_is_synthesized = false; 2849 bool is_gsym = false; 2850 bool set_value = true; 2851 2852 assert(sym_idx < num_syms); 2853 2854 sym[sym_idx].SetDebug(is_debug); 2855 2856 if (is_debug) { 2857 switch (nlist.n_type) { 2858 case N_GSYM: 2859 // global symbol: name,,NO_SECT,type,0 2860 // Sometimes the N_GSYM value contains the address. 2861 2862 // FIXME: In the .o files, we have a GSYM and a debug 2863 // symbol for all the ObjC data. They 2864 // have the same address, but we want to ensure that 2865 // we always find only the real symbol, 2866 // 'cause we don't currently correctly attribute the 2867 // GSYM one to the ObjCClass/Ivar/MetaClass 2868 // symbol type. This is a temporary hack to make sure 2869 // the ObjectiveC symbols get treated 2870 // correctly. To do this right, we should coalesce 2871 // all the GSYM & global symbols that have the 2872 // same address. 2873 2874 is_gsym = true; 2875 sym[sym_idx].SetExternal(true); 2876 2877 if (symbol_name && symbol_name[0] == '_' && 2878 symbol_name[1] == 'O') { 2879 llvm::StringRef symbol_name_ref(symbol_name); 2880 if (symbol_name_ref.startswith( 2881 g_objc_v2_prefix_class)) { 2882 symbol_name_non_abi_mangled = symbol_name + 1; 2883 symbol_name = 2884 symbol_name + g_objc_v2_prefix_class.size(); 2885 type = eSymbolTypeObjCClass; 2886 demangled_is_synthesized = true; 2887 2888 } else if (symbol_name_ref.startswith( 2889 g_objc_v2_prefix_metaclass)) { 2890 symbol_name_non_abi_mangled = symbol_name + 1; 2891 symbol_name = symbol_name + 2892 g_objc_v2_prefix_metaclass.size(); 2893 type = eSymbolTypeObjCMetaClass; 2894 demangled_is_synthesized = true; 2895 } else if (symbol_name_ref.startswith( 2896 g_objc_v2_prefix_ivar)) { 2897 symbol_name_non_abi_mangled = symbol_name + 1; 2898 symbol_name = 2899 symbol_name + g_objc_v2_prefix_ivar.size(); 2900 type = eSymbolTypeObjCIVar; 2901 demangled_is_synthesized = true; 2902 } 2903 } else { 2904 if (nlist.n_value != 0) 2905 symbol_section = section_info.GetSection( 2906 nlist.n_sect, nlist.n_value); 2907 type = eSymbolTypeData; 2908 } 2909 break; 2910 2911 case N_FNAME: 2912 // procedure name (f77 kludge): name,,NO_SECT,0,0 2913 type = eSymbolTypeCompiler; 2914 break; 2915 2916 case N_FUN: 2917 // procedure: name,,n_sect,linenumber,address 2918 if (symbol_name) { 2919 type = eSymbolTypeCode; 2920 symbol_section = section_info.GetSection( 2921 nlist.n_sect, nlist.n_value); 2922 2923 N_FUN_addr_to_sym_idx.insert( 2924 std::make_pair(nlist.n_value, sym_idx)); 2925 // We use the current number of symbols in the 2926 // symbol table in lieu of 2927 // using nlist_idx in case we ever start trimming 2928 // entries out 2929 N_FUN_indexes.push_back(sym_idx); 2930 } else { 2931 type = eSymbolTypeCompiler; 2932 2933 if (!N_FUN_indexes.empty()) { 2934 // Copy the size of the function into the original 2935 // STAB entry so we don't have 2936 // to hunt for it later 2937 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2938 ->SetByteSize(nlist.n_value); 2939 N_FUN_indexes.pop_back(); 2940 // We don't really need the end function STAB as 2941 // it contains the size which 2942 // we already placed with the original symbol, so 2943 // don't add it if we want a 2944 // minimal symbol table 2945 add_nlist = false; 2946 } 2947 } 2948 break; 2949 2950 case N_STSYM: 2951 // static symbol: name,,n_sect,type,address 2952 N_STSYM_addr_to_sym_idx.insert( 2953 std::make_pair(nlist.n_value, sym_idx)); 2954 symbol_section = section_info.GetSection( 2955 nlist.n_sect, nlist.n_value); 2956 if (symbol_name && symbol_name[0]) { 2957 type = ObjectFile::GetSymbolTypeFromName( 2958 symbol_name + 1, eSymbolTypeData); 2959 } 2960 break; 2961 2962 case N_LCSYM: 2963 // .lcomm symbol: name,,n_sect,type,address 2964 symbol_section = section_info.GetSection( 2965 nlist.n_sect, nlist.n_value); 2966 type = eSymbolTypeCommonBlock; 2967 break; 2968 2969 case N_BNSYM: 2970 // We use the current number of symbols in the symbol 2971 // table in lieu of 2972 // using nlist_idx in case we ever start trimming 2973 // entries out 2974 // Skip these if we want minimal symbol tables 2975 add_nlist = false; 2976 break; 2977 2978 case N_ENSYM: 2979 // Set the size of the N_BNSYM to the terminating 2980 // index of this N_ENSYM 2981 // so that we can always skip the entire symbol if we 2982 // need to navigate 2983 // more quickly at the source level when parsing STABS 2984 // Skip these if we want minimal symbol tables 2985 add_nlist = false; 2986 break; 2987 2988 case N_OPT: 2989 // emitted with gcc2_compiled and in gcc source 2990 type = eSymbolTypeCompiler; 2991 break; 2992 2993 case N_RSYM: 2994 // register sym: name,,NO_SECT,type,register 2995 type = eSymbolTypeVariable; 2996 break; 2997 2998 case N_SLINE: 2999 // src line: 0,,n_sect,linenumber,address 3000 symbol_section = section_info.GetSection( 3001 nlist.n_sect, nlist.n_value); 3002 type = eSymbolTypeLineEntry; 3003 break; 3004 3005 case N_SSYM: 3006 // structure elt: name,,NO_SECT,type,struct_offset 3007 type = eSymbolTypeVariableType; 3008 break; 3009 3010 case N_SO: 3011 // source file name 3012 type = eSymbolTypeSourceFile; 3013 if (symbol_name == NULL) { 3014 add_nlist = false; 3015 if (N_SO_index != UINT32_MAX) { 3016 // Set the size of the N_SO to the terminating 3017 // index of this N_SO 3018 // so that we can always skip the entire N_SO if 3019 // we need to navigate 3020 // more quickly at the source level when parsing 3021 // STABS 3022 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3023 symbol_ptr->SetByteSize(sym_idx); 3024 symbol_ptr->SetSizeIsSibling(true); 3025 } 3026 N_NSYM_indexes.clear(); 3027 N_INCL_indexes.clear(); 3028 N_BRAC_indexes.clear(); 3029 N_COMM_indexes.clear(); 3030 N_FUN_indexes.clear(); 3031 N_SO_index = UINT32_MAX; 3032 } else { 3033 // We use the current number of symbols in the 3034 // symbol table in lieu of 3035 // using nlist_idx in case we ever start trimming 3036 // entries out 3037 const bool N_SO_has_full_path = 3038 symbol_name[0] == '/'; 3039 if (N_SO_has_full_path) { 3040 if ((N_SO_index == sym_idx - 1) && 3041 ((sym_idx - 1) < num_syms)) { 3042 // We have two consecutive N_SO entries where 3043 // the first contains a directory 3044 // and the second contains a full path. 3045 sym[sym_idx - 1].GetMangled().SetValue( 3046 ConstString(symbol_name), false); 3047 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3048 add_nlist = false; 3049 } else { 3050 // This is the first entry in a N_SO that 3051 // contains a directory or 3052 // a full path to the source file 3053 N_SO_index = sym_idx; 3054 } 3055 } else if ((N_SO_index == sym_idx - 1) && 3056 ((sym_idx - 1) < num_syms)) { 3057 // This is usually the second N_SO entry that 3058 // contains just the filename, 3059 // so here we combine it with the first one if we 3060 // are minimizing the symbol table 3061 const char *so_path = 3062 sym[sym_idx - 1] 3063 .GetMangled() 3064 .GetDemangledName( 3065 lldb::eLanguageTypeUnknown) 3066 .AsCString(); 3067 if (so_path && so_path[0]) { 3068 std::string full_so_path(so_path); 3069 const size_t double_slash_pos = 3070 full_so_path.find("//"); 3071 if (double_slash_pos != std::string::npos) { 3072 // The linker has been generating bad N_SO 3073 // entries with doubled up paths 3074 // in the format "%s%s" where the first string 3075 // in the DW_AT_comp_dir, 3076 // and the second is the directory for the 3077 // source file so you end up with 3078 // a path that looks like "/tmp/src//tmp/src/" 3079 FileSpec so_dir(so_path, false); 3080 if (!so_dir.Exists()) { 3081 so_dir.SetFile( 3082 &full_so_path[double_slash_pos + 1], 3083 false); 3084 if (so_dir.Exists()) { 3085 // Trim off the incorrect path 3086 full_so_path.erase(0, 3087 double_slash_pos + 1); 3088 } 3089 } 3090 } 3091 if (*full_so_path.rbegin() != '/') 3092 full_so_path += '/'; 3093 full_so_path += symbol_name; 3094 sym[sym_idx - 1].GetMangled().SetValue( 3095 ConstString(full_so_path.c_str()), false); 3096 add_nlist = false; 3097 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3098 } 3099 } else { 3100 // This could be a relative path to a N_SO 3101 N_SO_index = sym_idx; 3102 } 3103 } 3104 break; 3105 3106 case N_OSO: 3107 // object file name: name,,0,0,st_mtime 3108 type = eSymbolTypeObjectFile; 3109 break; 3110 3111 case N_LSYM: 3112 // local sym: name,,NO_SECT,type,offset 3113 type = eSymbolTypeLocal; 3114 break; 3115 3116 //---------------------------------------------------------------------- 3117 // INCL scopes 3118 //---------------------------------------------------------------------- 3119 case N_BINCL: 3120 // include file beginning: name,,NO_SECT,0,sum 3121 // We use the current number of symbols in the symbol 3122 // table in lieu of 3123 // using nlist_idx in case we ever start trimming 3124 // entries out 3125 N_INCL_indexes.push_back(sym_idx); 3126 type = eSymbolTypeScopeBegin; 3127 break; 3128 3129 case N_EINCL: 3130 // include file end: name,,NO_SECT,0,0 3131 // Set the size of the N_BINCL to the terminating 3132 // index of this N_EINCL 3133 // so that we can always skip the entire symbol if we 3134 // need to navigate 3135 // more quickly at the source level when parsing STABS 3136 if (!N_INCL_indexes.empty()) { 3137 symbol_ptr = 3138 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3139 symbol_ptr->SetByteSize(sym_idx + 1); 3140 symbol_ptr->SetSizeIsSibling(true); 3141 N_INCL_indexes.pop_back(); 3142 } 3143 type = eSymbolTypeScopeEnd; 3144 break; 3145 3146 case N_SOL: 3147 // #included file name: name,,n_sect,0,address 3148 type = eSymbolTypeHeaderFile; 3149 3150 // We currently don't use the header files on darwin 3151 add_nlist = false; 3152 break; 3153 3154 case N_PARAMS: 3155 // compiler parameters: name,,NO_SECT,0,0 3156 type = eSymbolTypeCompiler; 3157 break; 3158 3159 case N_VERSION: 3160 // compiler version: name,,NO_SECT,0,0 3161 type = eSymbolTypeCompiler; 3162 break; 3163 3164 case N_OLEVEL: 3165 // compiler -O level: name,,NO_SECT,0,0 3166 type = eSymbolTypeCompiler; 3167 break; 3168 3169 case N_PSYM: 3170 // parameter: name,,NO_SECT,type,offset 3171 type = eSymbolTypeVariable; 3172 break; 3173 3174 case N_ENTRY: 3175 // alternate entry: name,,n_sect,linenumber,address 3176 symbol_section = section_info.GetSection( 3177 nlist.n_sect, nlist.n_value); 3178 type = eSymbolTypeLineEntry; 3179 break; 3180 3181 //---------------------------------------------------------------------- 3182 // Left and Right Braces 3183 //---------------------------------------------------------------------- 3184 case N_LBRAC: 3185 // left bracket: 0,,NO_SECT,nesting level,address 3186 // We use the current number of symbols in the symbol 3187 // table in lieu of 3188 // using nlist_idx in case we ever start trimming 3189 // entries out 3190 symbol_section = section_info.GetSection( 3191 nlist.n_sect, nlist.n_value); 3192 N_BRAC_indexes.push_back(sym_idx); 3193 type = eSymbolTypeScopeBegin; 3194 break; 3195 3196 case N_RBRAC: 3197 // right bracket: 0,,NO_SECT,nesting level,address 3198 // Set the size of the N_LBRAC to the terminating 3199 // index of this N_RBRAC 3200 // so that we can always skip the entire symbol if we 3201 // need to navigate 3202 // more quickly at the source level when parsing STABS 3203 symbol_section = section_info.GetSection( 3204 nlist.n_sect, nlist.n_value); 3205 if (!N_BRAC_indexes.empty()) { 3206 symbol_ptr = 3207 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3208 symbol_ptr->SetByteSize(sym_idx + 1); 3209 symbol_ptr->SetSizeIsSibling(true); 3210 N_BRAC_indexes.pop_back(); 3211 } 3212 type = eSymbolTypeScopeEnd; 3213 break; 3214 3215 case N_EXCL: 3216 // deleted include file: name,,NO_SECT,0,sum 3217 type = eSymbolTypeHeaderFile; 3218 break; 3219 3220 //---------------------------------------------------------------------- 3221 // COMM scopes 3222 //---------------------------------------------------------------------- 3223 case N_BCOMM: 3224 // begin common: name,,NO_SECT,0,0 3225 // We use the current number of symbols in the symbol 3226 // table in lieu of 3227 // using nlist_idx in case we ever start trimming 3228 // entries out 3229 type = eSymbolTypeScopeBegin; 3230 N_COMM_indexes.push_back(sym_idx); 3231 break; 3232 3233 case N_ECOML: 3234 // end common (local name): 0,,n_sect,0,address 3235 symbol_section = section_info.GetSection( 3236 nlist.n_sect, nlist.n_value); 3237 // Fall through 3238 3239 case N_ECOMM: 3240 // end common: name,,n_sect,0,0 3241 // Set the size of the N_BCOMM to the terminating 3242 // index of this N_ECOMM/N_ECOML 3243 // so that we can always skip the entire symbol if we 3244 // need to navigate 3245 // more quickly at the source level when parsing STABS 3246 if (!N_COMM_indexes.empty()) { 3247 symbol_ptr = 3248 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3249 symbol_ptr->SetByteSize(sym_idx + 1); 3250 symbol_ptr->SetSizeIsSibling(true); 3251 N_COMM_indexes.pop_back(); 3252 } 3253 type = eSymbolTypeScopeEnd; 3254 break; 3255 3256 case N_LENG: 3257 // second stab entry with length information 3258 type = eSymbolTypeAdditional; 3259 break; 3260 3261 default: 3262 break; 3263 } 3264 } else { 3265 // uint8_t n_pext = N_PEXT & nlist.n_type; 3266 uint8_t n_type = N_TYPE & nlist.n_type; 3267 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3268 3269 switch (n_type) { 3270 case N_INDR: { 3271 const char *reexport_name_cstr = 3272 strtab_data.PeekCStr(nlist.n_value); 3273 if (reexport_name_cstr && reexport_name_cstr[0]) { 3274 type = eSymbolTypeReExported; 3275 ConstString reexport_name( 3276 reexport_name_cstr + 3277 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3278 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3279 set_value = false; 3280 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3281 indirect_symbol_names.insert( 3282 ConstString(symbol_name + 3283 ((symbol_name[0] == '_') ? 1 : 0))); 3284 } else 3285 type = eSymbolTypeUndefined; 3286 } break; 3287 3288 case N_UNDF: 3289 if (symbol_name && symbol_name[0]) { 3290 ConstString undefined_name( 3291 symbol_name + 3292 ((symbol_name[0] == '_') ? 1 : 0)); 3293 undefined_name_to_desc[undefined_name] = 3294 nlist.n_desc; 3295 } 3296 // Fall through 3297 case N_PBUD: 3298 type = eSymbolTypeUndefined; 3299 break; 3300 3301 case N_ABS: 3302 type = eSymbolTypeAbsolute; 3303 break; 3304 3305 case N_SECT: { 3306 symbol_section = section_info.GetSection( 3307 nlist.n_sect, nlist.n_value); 3308 3309 if (symbol_section == NULL) { 3310 // TODO: warn about this? 3311 add_nlist = false; 3312 break; 3313 } 3314 3315 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3316 type = eSymbolTypeException; 3317 } else { 3318 uint32_t section_type = 3319 symbol_section->Get() & SECTION_TYPE; 3320 3321 switch (section_type) { 3322 case S_CSTRING_LITERALS: 3323 type = eSymbolTypeData; 3324 break; // section with only literal C strings 3325 case S_4BYTE_LITERALS: 3326 type = eSymbolTypeData; 3327 break; // section with only 4 byte literals 3328 case S_8BYTE_LITERALS: 3329 type = eSymbolTypeData; 3330 break; // section with only 8 byte literals 3331 case S_LITERAL_POINTERS: 3332 type = eSymbolTypeTrampoline; 3333 break; // section with only pointers to literals 3334 case S_NON_LAZY_SYMBOL_POINTERS: 3335 type = eSymbolTypeTrampoline; 3336 break; // section with only non-lazy symbol 3337 // pointers 3338 case S_LAZY_SYMBOL_POINTERS: 3339 type = eSymbolTypeTrampoline; 3340 break; // section with only lazy symbol pointers 3341 case S_SYMBOL_STUBS: 3342 type = eSymbolTypeTrampoline; 3343 break; // section with only symbol stubs, byte 3344 // size of stub in the reserved2 field 3345 case S_MOD_INIT_FUNC_POINTERS: 3346 type = eSymbolTypeCode; 3347 break; // section with only function pointers for 3348 // initialization 3349 case S_MOD_TERM_FUNC_POINTERS: 3350 type = eSymbolTypeCode; 3351 break; // section with only function pointers for 3352 // termination 3353 case S_INTERPOSING: 3354 type = eSymbolTypeTrampoline; 3355 break; // section with only pairs of function 3356 // pointers for interposing 3357 case S_16BYTE_LITERALS: 3358 type = eSymbolTypeData; 3359 break; // section with only 16 byte literals 3360 case S_DTRACE_DOF: 3361 type = eSymbolTypeInstrumentation; 3362 break; 3363 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3364 type = eSymbolTypeTrampoline; 3365 break; 3366 default: 3367 switch (symbol_section->GetType()) { 3368 case lldb::eSectionTypeCode: 3369 type = eSymbolTypeCode; 3370 break; 3371 case eSectionTypeData: 3372 case eSectionTypeDataCString: // Inlined C string 3373 // data 3374 case eSectionTypeDataCStringPointers: // Pointers 3375 // to C 3376 // string 3377 // data 3378 case eSectionTypeDataSymbolAddress: // Address of 3379 // a symbol in 3380 // the symbol 3381 // table 3382 case eSectionTypeData4: 3383 case eSectionTypeData8: 3384 case eSectionTypeData16: 3385 type = eSymbolTypeData; 3386 break; 3387 default: 3388 break; 3389 } 3390 break; 3391 } 3392 3393 if (type == eSymbolTypeInvalid) { 3394 const char *symbol_sect_name = 3395 symbol_section->GetName().AsCString(); 3396 if (symbol_section->IsDescendant( 3397 text_section_sp.get())) { 3398 if (symbol_section->IsClear( 3399 S_ATTR_PURE_INSTRUCTIONS | 3400 S_ATTR_SELF_MODIFYING_CODE | 3401 S_ATTR_SOME_INSTRUCTIONS)) 3402 type = eSymbolTypeData; 3403 else 3404 type = eSymbolTypeCode; 3405 } else if (symbol_section->IsDescendant( 3406 data_section_sp.get()) || 3407 symbol_section->IsDescendant( 3408 data_dirty_section_sp.get()) || 3409 symbol_section->IsDescendant( 3410 data_const_section_sp.get())) { 3411 if (symbol_sect_name && 3412 ::strstr(symbol_sect_name, "__objc") == 3413 symbol_sect_name) { 3414 type = eSymbolTypeRuntime; 3415 3416 if (symbol_name) { 3417 llvm::StringRef symbol_name_ref( 3418 symbol_name); 3419 if (symbol_name_ref.startswith("_OBJC_")) { 3420 static const llvm::StringRef 3421 g_objc_v2_prefix_class( 3422 "_OBJC_CLASS_$_"); 3423 static const llvm::StringRef 3424 g_objc_v2_prefix_metaclass( 3425 "_OBJC_METACLASS_$_"); 3426 static const llvm::StringRef 3427 g_objc_v2_prefix_ivar( 3428 "_OBJC_IVAR_$_"); 3429 if (symbol_name_ref.startswith( 3430 g_objc_v2_prefix_class)) { 3431 symbol_name_non_abi_mangled = 3432 symbol_name + 1; 3433 symbol_name = 3434 symbol_name + 3435 g_objc_v2_prefix_class.size(); 3436 type = eSymbolTypeObjCClass; 3437 demangled_is_synthesized = true; 3438 } else if ( 3439 symbol_name_ref.startswith( 3440 g_objc_v2_prefix_metaclass)) { 3441 symbol_name_non_abi_mangled = 3442 symbol_name + 1; 3443 symbol_name = 3444 symbol_name + 3445 g_objc_v2_prefix_metaclass.size(); 3446 type = eSymbolTypeObjCMetaClass; 3447 demangled_is_synthesized = true; 3448 } else if (symbol_name_ref.startswith( 3449 g_objc_v2_prefix_ivar)) { 3450 symbol_name_non_abi_mangled = 3451 symbol_name + 1; 3452 symbol_name = 3453 symbol_name + 3454 g_objc_v2_prefix_ivar.size(); 3455 type = eSymbolTypeObjCIVar; 3456 demangled_is_synthesized = true; 3457 } 3458 } 3459 } 3460 } else if (symbol_sect_name && 3461 ::strstr(symbol_sect_name, 3462 "__gcc_except_tab") == 3463 symbol_sect_name) { 3464 type = eSymbolTypeException; 3465 } else { 3466 type = eSymbolTypeData; 3467 } 3468 } else if (symbol_sect_name && 3469 ::strstr(symbol_sect_name, 3470 "__IMPORT") == 3471 symbol_sect_name) { 3472 type = eSymbolTypeTrampoline; 3473 } else if (symbol_section->IsDescendant( 3474 objc_section_sp.get())) { 3475 type = eSymbolTypeRuntime; 3476 if (symbol_name && symbol_name[0] == '.') { 3477 llvm::StringRef symbol_name_ref(symbol_name); 3478 static const llvm::StringRef 3479 g_objc_v1_prefix_class( 3480 ".objc_class_name_"); 3481 if (symbol_name_ref.startswith( 3482 g_objc_v1_prefix_class)) { 3483 symbol_name_non_abi_mangled = symbol_name; 3484 symbol_name = symbol_name + 3485 g_objc_v1_prefix_class.size(); 3486 type = eSymbolTypeObjCClass; 3487 demangled_is_synthesized = true; 3488 } 3489 } 3490 } 3491 } 3492 } 3493 } break; 3494 } 3495 } 3496 3497 if (add_nlist) { 3498 uint64_t symbol_value = nlist.n_value; 3499 if (symbol_name_non_abi_mangled) { 3500 sym[sym_idx].GetMangled().SetMangledName( 3501 ConstString(symbol_name_non_abi_mangled)); 3502 sym[sym_idx].GetMangled().SetDemangledName( 3503 ConstString(symbol_name)); 3504 } else { 3505 bool symbol_name_is_mangled = false; 3506 3507 if (symbol_name && symbol_name[0] == '_') { 3508 symbol_name_is_mangled = symbol_name[1] == '_'; 3509 symbol_name++; // Skip the leading underscore 3510 } 3511 3512 if (symbol_name) { 3513 ConstString const_symbol_name(symbol_name); 3514 sym[sym_idx].GetMangled().SetValue( 3515 const_symbol_name, symbol_name_is_mangled); 3516 if (is_gsym && is_debug) { 3517 const char *gsym_name = 3518 sym[sym_idx] 3519 .GetMangled() 3520 .GetName(lldb::eLanguageTypeUnknown, 3521 Mangled::ePreferMangled) 3522 .GetCString(); 3523 if (gsym_name) 3524 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3525 } 3526 } 3527 } 3528 if (symbol_section) { 3529 const addr_t section_file_addr = 3530 symbol_section->GetFileAddress(); 3531 if (symbol_byte_size == 0 && 3532 function_starts_count > 0) { 3533 addr_t symbol_lookup_file_addr = nlist.n_value; 3534 // Do an exact address match for non-ARM addresses, 3535 // else get the closest since 3536 // the symbol might be a thumb symbol which has an 3537 // address with bit zero set 3538 FunctionStarts::Entry *func_start_entry = 3539 function_starts.FindEntry( 3540 symbol_lookup_file_addr, !is_arm); 3541 if (is_arm && func_start_entry) { 3542 // Verify that the function start address is the 3543 // symbol address (ARM) 3544 // or the symbol address + 1 (thumb) 3545 if (func_start_entry->addr != 3546 symbol_lookup_file_addr && 3547 func_start_entry->addr != 3548 (symbol_lookup_file_addr + 1)) { 3549 // Not the right entry, NULL it out... 3550 func_start_entry = NULL; 3551 } 3552 } 3553 if (func_start_entry) { 3554 func_start_entry->data = true; 3555 3556 addr_t symbol_file_addr = func_start_entry->addr; 3557 uint32_t symbol_flags = 0; 3558 if (is_arm) { 3559 if (symbol_file_addr & 1) 3560 symbol_flags = 3561 MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3562 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3563 } 3564 3565 const FunctionStarts::Entry 3566 *next_func_start_entry = 3567 function_starts.FindNextEntry( 3568 func_start_entry); 3569 const addr_t section_end_file_addr = 3570 section_file_addr + 3571 symbol_section->GetByteSize(); 3572 if (next_func_start_entry) { 3573 addr_t next_symbol_file_addr = 3574 next_func_start_entry->addr; 3575 // Be sure the clear the Thumb address bit when 3576 // we calculate the size 3577 // from the current and next address 3578 if (is_arm) 3579 next_symbol_file_addr &= 3580 THUMB_ADDRESS_BIT_MASK; 3581 symbol_byte_size = std::min<lldb::addr_t>( 3582 next_symbol_file_addr - symbol_file_addr, 3583 section_end_file_addr - symbol_file_addr); 3584 } else { 3585 symbol_byte_size = 3586 section_end_file_addr - symbol_file_addr; 3587 } 3588 } 3589 } 3590 symbol_value -= section_file_addr; 3591 } 3592 3593 if (is_debug == false) { 3594 if (type == eSymbolTypeCode) { 3595 // See if we can find a N_FUN entry for any code 3596 // symbols. 3597 // If we do find a match, and the name matches, then 3598 // we 3599 // can merge the two into just the function symbol 3600 // to avoid 3601 // duplicate entries in the symbol table 3602 std::pair<ValueToSymbolIndexMap::const_iterator, 3603 ValueToSymbolIndexMap::const_iterator> 3604 range; 3605 range = N_FUN_addr_to_sym_idx.equal_range( 3606 nlist.n_value); 3607 if (range.first != range.second) { 3608 bool found_it = false; 3609 for (ValueToSymbolIndexMap::const_iterator pos = 3610 range.first; 3611 pos != range.second; ++pos) { 3612 if (sym[sym_idx].GetMangled().GetName( 3613 lldb::eLanguageTypeUnknown, 3614 Mangled::ePreferMangled) == 3615 sym[pos->second].GetMangled().GetName( 3616 lldb::eLanguageTypeUnknown, 3617 Mangled::ePreferMangled)) { 3618 m_nlist_idx_to_sym_idx[nlist_idx] = 3619 pos->second; 3620 // We just need the flags from the linker 3621 // symbol, so put these flags 3622 // into the N_FUN flags to avoid duplicate 3623 // symbols in the symbol table 3624 sym[pos->second].SetExternal( 3625 sym[sym_idx].IsExternal()); 3626 sym[pos->second].SetFlags(nlist.n_type << 16 | 3627 nlist.n_desc); 3628 if (resolver_addresses.find(nlist.n_value) != 3629 resolver_addresses.end()) 3630 sym[pos->second].SetType( 3631 eSymbolTypeResolver); 3632 sym[sym_idx].Clear(); 3633 found_it = true; 3634 break; 3635 } 3636 } 3637 if (found_it) 3638 continue; 3639 } else { 3640 if (resolver_addresses.find(nlist.n_value) != 3641 resolver_addresses.end()) 3642 type = eSymbolTypeResolver; 3643 } 3644 } else if (type == eSymbolTypeData || 3645 type == eSymbolTypeObjCClass || 3646 type == eSymbolTypeObjCMetaClass || 3647 type == eSymbolTypeObjCIVar) { 3648 // See if we can find a N_STSYM entry for any data 3649 // symbols. 3650 // If we do find a match, and the name matches, then 3651 // we 3652 // can merge the two into just the Static symbol to 3653 // avoid 3654 // duplicate entries in the symbol table 3655 std::pair<ValueToSymbolIndexMap::const_iterator, 3656 ValueToSymbolIndexMap::const_iterator> 3657 range; 3658 range = N_STSYM_addr_to_sym_idx.equal_range( 3659 nlist.n_value); 3660 if (range.first != range.second) { 3661 bool found_it = false; 3662 for (ValueToSymbolIndexMap::const_iterator pos = 3663 range.first; 3664 pos != range.second; ++pos) { 3665 if (sym[sym_idx].GetMangled().GetName( 3666 lldb::eLanguageTypeUnknown, 3667 Mangled::ePreferMangled) == 3668 sym[pos->second].GetMangled().GetName( 3669 lldb::eLanguageTypeUnknown, 3670 Mangled::ePreferMangled)) { 3671 m_nlist_idx_to_sym_idx[nlist_idx] = 3672 pos->second; 3673 // We just need the flags from the linker 3674 // symbol, so put these flags 3675 // into the N_STSYM flags to avoid duplicate 3676 // symbols in the symbol table 3677 sym[pos->second].SetExternal( 3678 sym[sym_idx].IsExternal()); 3679 sym[pos->second].SetFlags(nlist.n_type << 16 | 3680 nlist.n_desc); 3681 sym[sym_idx].Clear(); 3682 found_it = true; 3683 break; 3684 } 3685 } 3686 if (found_it) 3687 continue; 3688 } else { 3689 const char *gsym_name = 3690 sym[sym_idx] 3691 .GetMangled() 3692 .GetName(lldb::eLanguageTypeUnknown, 3693 Mangled::ePreferMangled) 3694 .GetCString(); 3695 if (gsym_name) { 3696 // Combine N_GSYM stab entries with the non stab 3697 // symbol 3698 ConstNameToSymbolIndexMap::const_iterator pos = 3699 N_GSYM_name_to_sym_idx.find(gsym_name); 3700 if (pos != N_GSYM_name_to_sym_idx.end()) { 3701 const uint32_t GSYM_sym_idx = pos->second; 3702 m_nlist_idx_to_sym_idx[nlist_idx] = 3703 GSYM_sym_idx; 3704 // Copy the address, because often the N_GSYM 3705 // address has an invalid address of zero 3706 // when the global is a common symbol 3707 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3708 symbol_section); 3709 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3710 symbol_value); 3711 // We just need the flags from the linker 3712 // symbol, so put these flags 3713 // into the N_GSYM flags to avoid duplicate 3714 // symbols in the symbol table 3715 sym[GSYM_sym_idx].SetFlags( 3716 nlist.n_type << 16 | nlist.n_desc); 3717 sym[sym_idx].Clear(); 3718 continue; 3719 } 3720 } 3721 } 3722 } 3723 } 3724 3725 sym[sym_idx].SetID(nlist_idx); 3726 sym[sym_idx].SetType(type); 3727 if (set_value) { 3728 sym[sym_idx].GetAddressRef().SetSection( 3729 symbol_section); 3730 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3731 } 3732 sym[sym_idx].SetFlags(nlist.n_type << 16 | 3733 nlist.n_desc); 3734 3735 if (symbol_byte_size > 0) 3736 sym[sym_idx].SetByteSize(symbol_byte_size); 3737 3738 if (demangled_is_synthesized) 3739 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3740 ++sym_idx; 3741 } else { 3742 sym[sym_idx].Clear(); 3743 } 3744 } 3745 ///////////////////////////// 3746 } 3747 break; // No more entries to consider 3748 } 3749 } 3750 3751 for (const auto &pos : reexport_shlib_needs_fixup) { 3752 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3753 if (undef_pos != undefined_name_to_desc.end()) { 3754 const uint8_t dylib_ordinal = 3755 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3756 if (dylib_ordinal > 0 && 3757 dylib_ordinal < dylib_files.GetSize()) 3758 sym[pos.first].SetReExportedSymbolSharedLibrary( 3759 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3760 } 3761 } 3762 } 3763 } 3764 } 3765 } 3766 } 3767 3768 // Must reset this in case it was mutated above! 3769 nlist_data_offset = 0; 3770 #endif 3771 3772 if (nlist_data.GetByteSize() > 0) { 3773 3774 // If the sym array was not created while parsing the DSC unmapped 3775 // symbols, create it now. 3776 if (sym == NULL) { 3777 sym = symtab->Resize(symtab_load_command.nsyms + 3778 m_dysymtab.nindirectsyms); 3779 num_syms = symtab->GetNumSymbols(); 3780 } 3781 3782 if (unmapped_local_symbols_found) { 3783 assert(m_dysymtab.ilocalsym == 0); 3784 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3785 nlist_idx = m_dysymtab.nlocalsym; 3786 } else { 3787 nlist_idx = 0; 3788 } 3789 3790 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3791 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3792 UndefinedNameToDescMap undefined_name_to_desc; 3793 SymbolIndexToName reexport_shlib_needs_fixup; 3794 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 3795 struct nlist_64 nlist; 3796 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, 3797 nlist_byte_size)) 3798 break; 3799 3800 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3801 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 3802 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 3803 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 3804 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 3805 3806 SymbolType type = eSymbolTypeInvalid; 3807 const char *symbol_name = NULL; 3808 3809 if (have_strtab_data) { 3810 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3811 3812 if (symbol_name == NULL) { 3813 // No symbol should be NULL, even the symbols with no 3814 // string values should have an offset zero which points 3815 // to an empty C-string 3816 Host::SystemLog(Host::eSystemLogError, 3817 "error: symbol[%u] has invalid string table offset " 3818 "0x%x in %s, ignoring symbol\n", 3819 nlist_idx, nlist.n_strx, 3820 module_sp->GetFileSpec().GetPath().c_str()); 3821 continue; 3822 } 3823 if (symbol_name[0] == '\0') 3824 symbol_name = NULL; 3825 } else { 3826 const addr_t str_addr = strtab_addr + nlist.n_strx; 3827 Status str_error; 3828 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3829 str_error)) 3830 symbol_name = memory_symbol_name.c_str(); 3831 } 3832 const char *symbol_name_non_abi_mangled = NULL; 3833 3834 SectionSP symbol_section; 3835 lldb::addr_t symbol_byte_size = 0; 3836 bool add_nlist = true; 3837 bool is_gsym = false; 3838 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3839 bool demangled_is_synthesized = false; 3840 bool set_value = true; 3841 assert(sym_idx < num_syms); 3842 3843 sym[sym_idx].SetDebug(is_debug); 3844 3845 if (is_debug) { 3846 switch (nlist.n_type) { 3847 case N_GSYM: 3848 // global symbol: name,,NO_SECT,type,0 3849 // Sometimes the N_GSYM value contains the address. 3850 3851 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3852 // the ObjC data. They 3853 // have the same address, but we want to ensure that we always find 3854 // only the real symbol, 3855 // 'cause we don't currently correctly attribute the GSYM one to the 3856 // ObjCClass/Ivar/MetaClass 3857 // symbol type. This is a temporary hack to make sure the 3858 // ObjectiveC symbols get treated 3859 // correctly. To do this right, we should coalesce all the GSYM & 3860 // global symbols that have the 3861 // same address. 3862 is_gsym = true; 3863 sym[sym_idx].SetExternal(true); 3864 3865 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3866 llvm::StringRef symbol_name_ref(symbol_name); 3867 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3868 symbol_name_non_abi_mangled = symbol_name + 1; 3869 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3870 type = eSymbolTypeObjCClass; 3871 demangled_is_synthesized = true; 3872 3873 } else if (symbol_name_ref.startswith( 3874 g_objc_v2_prefix_metaclass)) { 3875 symbol_name_non_abi_mangled = symbol_name + 1; 3876 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3877 type = eSymbolTypeObjCMetaClass; 3878 demangled_is_synthesized = true; 3879 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3880 symbol_name_non_abi_mangled = symbol_name + 1; 3881 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3882 type = eSymbolTypeObjCIVar; 3883 demangled_is_synthesized = true; 3884 } 3885 } else { 3886 if (nlist.n_value != 0) 3887 symbol_section = 3888 section_info.GetSection(nlist.n_sect, nlist.n_value); 3889 type = eSymbolTypeData; 3890 } 3891 break; 3892 3893 case N_FNAME: 3894 // procedure name (f77 kludge): name,,NO_SECT,0,0 3895 type = eSymbolTypeCompiler; 3896 break; 3897 3898 case N_FUN: 3899 // procedure: name,,n_sect,linenumber,address 3900 if (symbol_name) { 3901 type = eSymbolTypeCode; 3902 symbol_section = 3903 section_info.GetSection(nlist.n_sect, nlist.n_value); 3904 3905 N_FUN_addr_to_sym_idx.insert( 3906 std::make_pair(nlist.n_value, sym_idx)); 3907 // We use the current number of symbols in the symbol table in 3908 // lieu of 3909 // using nlist_idx in case we ever start trimming entries out 3910 N_FUN_indexes.push_back(sym_idx); 3911 } else { 3912 type = eSymbolTypeCompiler; 3913 3914 if (!N_FUN_indexes.empty()) { 3915 // Copy the size of the function into the original STAB entry so 3916 // we don't have 3917 // to hunt for it later 3918 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3919 ->SetByteSize(nlist.n_value); 3920 N_FUN_indexes.pop_back(); 3921 // We don't really need the end function STAB as it contains the 3922 // size which 3923 // we already placed with the original symbol, so don't add it 3924 // if we want a 3925 // minimal symbol table 3926 add_nlist = false; 3927 } 3928 } 3929 break; 3930 3931 case N_STSYM: 3932 // static symbol: name,,n_sect,type,address 3933 N_STSYM_addr_to_sym_idx.insert( 3934 std::make_pair(nlist.n_value, sym_idx)); 3935 symbol_section = 3936 section_info.GetSection(nlist.n_sect, nlist.n_value); 3937 if (symbol_name && symbol_name[0]) { 3938 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3939 eSymbolTypeData); 3940 } 3941 break; 3942 3943 case N_LCSYM: 3944 // .lcomm symbol: name,,n_sect,type,address 3945 symbol_section = 3946 section_info.GetSection(nlist.n_sect, nlist.n_value); 3947 type = eSymbolTypeCommonBlock; 3948 break; 3949 3950 case N_BNSYM: 3951 // We use the current number of symbols in the symbol table in lieu 3952 // of 3953 // using nlist_idx in case we ever start trimming entries out 3954 // Skip these if we want minimal symbol tables 3955 add_nlist = false; 3956 break; 3957 3958 case N_ENSYM: 3959 // Set the size of the N_BNSYM to the terminating index of this 3960 // N_ENSYM 3961 // so that we can always skip the entire symbol if we need to 3962 // navigate 3963 // more quickly at the source level when parsing STABS 3964 // Skip these if we want minimal symbol tables 3965 add_nlist = false; 3966 break; 3967 3968 case N_OPT: 3969 // emitted with gcc2_compiled and in gcc source 3970 type = eSymbolTypeCompiler; 3971 break; 3972 3973 case N_RSYM: 3974 // register sym: name,,NO_SECT,type,register 3975 type = eSymbolTypeVariable; 3976 break; 3977 3978 case N_SLINE: 3979 // src line: 0,,n_sect,linenumber,address 3980 symbol_section = 3981 section_info.GetSection(nlist.n_sect, nlist.n_value); 3982 type = eSymbolTypeLineEntry; 3983 break; 3984 3985 case N_SSYM: 3986 // structure elt: name,,NO_SECT,type,struct_offset 3987 type = eSymbolTypeVariableType; 3988 break; 3989 3990 case N_SO: 3991 // source file name 3992 type = eSymbolTypeSourceFile; 3993 if (symbol_name == NULL) { 3994 add_nlist = false; 3995 if (N_SO_index != UINT32_MAX) { 3996 // Set the size of the N_SO to the terminating index of this 3997 // N_SO 3998 // so that we can always skip the entire N_SO if we need to 3999 // navigate 4000 // more quickly at the source level when parsing STABS 4001 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 4002 symbol_ptr->SetByteSize(sym_idx); 4003 symbol_ptr->SetSizeIsSibling(true); 4004 } 4005 N_NSYM_indexes.clear(); 4006 N_INCL_indexes.clear(); 4007 N_BRAC_indexes.clear(); 4008 N_COMM_indexes.clear(); 4009 N_FUN_indexes.clear(); 4010 N_SO_index = UINT32_MAX; 4011 } else { 4012 // We use the current number of symbols in the symbol table in 4013 // lieu of 4014 // using nlist_idx in case we ever start trimming entries out 4015 const bool N_SO_has_full_path = symbol_name[0] == '/'; 4016 if (N_SO_has_full_path) { 4017 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 4018 // We have two consecutive N_SO entries where the first 4019 // contains a directory 4020 // and the second contains a full path. 4021 sym[sym_idx - 1].GetMangled().SetValue( 4022 ConstString(symbol_name), false); 4023 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4024 add_nlist = false; 4025 } else { 4026 // This is the first entry in a N_SO that contains a directory 4027 // or 4028 // a full path to the source file 4029 N_SO_index = sym_idx; 4030 } 4031 } else if ((N_SO_index == sym_idx - 1) && 4032 ((sym_idx - 1) < num_syms)) { 4033 // This is usually the second N_SO entry that contains just the 4034 // filename, 4035 // so here we combine it with the first one if we are minimizing 4036 // the symbol table 4037 const char *so_path = 4038 sym[sym_idx - 1] 4039 .GetMangled() 4040 .GetDemangledName(lldb::eLanguageTypeUnknown) 4041 .AsCString(); 4042 if (so_path && so_path[0]) { 4043 std::string full_so_path(so_path); 4044 const size_t double_slash_pos = full_so_path.find("//"); 4045 if (double_slash_pos != std::string::npos) { 4046 // The linker has been generating bad N_SO entries with 4047 // doubled up paths 4048 // in the format "%s%s" where the first string in the 4049 // DW_AT_comp_dir, 4050 // and the second is the directory for the source file so 4051 // you end up with 4052 // a path that looks like "/tmp/src//tmp/src/" 4053 FileSpec so_dir(so_path, false); 4054 if (!so_dir.Exists()) { 4055 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 4056 false); 4057 if (so_dir.Exists()) { 4058 // Trim off the incorrect path 4059 full_so_path.erase(0, double_slash_pos + 1); 4060 } 4061 } 4062 } 4063 if (*full_so_path.rbegin() != '/') 4064 full_so_path += '/'; 4065 full_so_path += symbol_name; 4066 sym[sym_idx - 1].GetMangled().SetValue( 4067 ConstString(full_so_path.c_str()), false); 4068 add_nlist = false; 4069 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4070 } 4071 } else { 4072 // This could be a relative path to a N_SO 4073 N_SO_index = sym_idx; 4074 } 4075 } 4076 break; 4077 4078 case N_OSO: 4079 // object file name: name,,0,0,st_mtime 4080 type = eSymbolTypeObjectFile; 4081 break; 4082 4083 case N_LSYM: 4084 // local sym: name,,NO_SECT,type,offset 4085 type = eSymbolTypeLocal; 4086 break; 4087 4088 //---------------------------------------------------------------------- 4089 // INCL scopes 4090 //---------------------------------------------------------------------- 4091 case N_BINCL: 4092 // include file beginning: name,,NO_SECT,0,sum 4093 // We use the current number of symbols in the symbol table in lieu 4094 // of 4095 // using nlist_idx in case we ever start trimming entries out 4096 N_INCL_indexes.push_back(sym_idx); 4097 type = eSymbolTypeScopeBegin; 4098 break; 4099 4100 case N_EINCL: 4101 // include file end: name,,NO_SECT,0,0 4102 // Set the size of the N_BINCL to the terminating index of this 4103 // N_EINCL 4104 // so that we can always skip the entire symbol if we need to 4105 // navigate 4106 // more quickly at the source level when parsing STABS 4107 if (!N_INCL_indexes.empty()) { 4108 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4109 symbol_ptr->SetByteSize(sym_idx + 1); 4110 symbol_ptr->SetSizeIsSibling(true); 4111 N_INCL_indexes.pop_back(); 4112 } 4113 type = eSymbolTypeScopeEnd; 4114 break; 4115 4116 case N_SOL: 4117 // #included file name: name,,n_sect,0,address 4118 type = eSymbolTypeHeaderFile; 4119 4120 // We currently don't use the header files on darwin 4121 add_nlist = false; 4122 break; 4123 4124 case N_PARAMS: 4125 // compiler parameters: name,,NO_SECT,0,0 4126 type = eSymbolTypeCompiler; 4127 break; 4128 4129 case N_VERSION: 4130 // compiler version: name,,NO_SECT,0,0 4131 type = eSymbolTypeCompiler; 4132 break; 4133 4134 case N_OLEVEL: 4135 // compiler -O level: name,,NO_SECT,0,0 4136 type = eSymbolTypeCompiler; 4137 break; 4138 4139 case N_PSYM: 4140 // parameter: name,,NO_SECT,type,offset 4141 type = eSymbolTypeVariable; 4142 break; 4143 4144 case N_ENTRY: 4145 // alternate entry: name,,n_sect,linenumber,address 4146 symbol_section = 4147 section_info.GetSection(nlist.n_sect, nlist.n_value); 4148 type = eSymbolTypeLineEntry; 4149 break; 4150 4151 //---------------------------------------------------------------------- 4152 // Left and Right Braces 4153 //---------------------------------------------------------------------- 4154 case N_LBRAC: 4155 // left bracket: 0,,NO_SECT,nesting level,address 4156 // We use the current number of symbols in the symbol table in lieu 4157 // of 4158 // using nlist_idx in case we ever start trimming entries out 4159 symbol_section = 4160 section_info.GetSection(nlist.n_sect, nlist.n_value); 4161 N_BRAC_indexes.push_back(sym_idx); 4162 type = eSymbolTypeScopeBegin; 4163 break; 4164 4165 case N_RBRAC: 4166 // right bracket: 0,,NO_SECT,nesting level,address 4167 // Set the size of the N_LBRAC to the terminating index of this 4168 // N_RBRAC 4169 // so that we can always skip the entire symbol if we need to 4170 // navigate 4171 // more quickly at the source level when parsing STABS 4172 symbol_section = 4173 section_info.GetSection(nlist.n_sect, nlist.n_value); 4174 if (!N_BRAC_indexes.empty()) { 4175 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4176 symbol_ptr->SetByteSize(sym_idx + 1); 4177 symbol_ptr->SetSizeIsSibling(true); 4178 N_BRAC_indexes.pop_back(); 4179 } 4180 type = eSymbolTypeScopeEnd; 4181 break; 4182 4183 case N_EXCL: 4184 // deleted include file: name,,NO_SECT,0,sum 4185 type = eSymbolTypeHeaderFile; 4186 break; 4187 4188 //---------------------------------------------------------------------- 4189 // COMM scopes 4190 //---------------------------------------------------------------------- 4191 case N_BCOMM: 4192 // begin common: name,,NO_SECT,0,0 4193 // We use the current number of symbols in the symbol table in lieu 4194 // of 4195 // using nlist_idx in case we ever start trimming entries out 4196 type = eSymbolTypeScopeBegin; 4197 N_COMM_indexes.push_back(sym_idx); 4198 break; 4199 4200 case N_ECOML: 4201 // end common (local name): 0,,n_sect,0,address 4202 symbol_section = 4203 section_info.GetSection(nlist.n_sect, nlist.n_value); 4204 LLVM_FALLTHROUGH; 4205 4206 case N_ECOMM: 4207 // end common: name,,n_sect,0,0 4208 // Set the size of the N_BCOMM to the terminating index of this 4209 // N_ECOMM/N_ECOML 4210 // so that we can always skip the entire symbol if we need to 4211 // navigate 4212 // more quickly at the source level when parsing STABS 4213 if (!N_COMM_indexes.empty()) { 4214 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4215 symbol_ptr->SetByteSize(sym_idx + 1); 4216 symbol_ptr->SetSizeIsSibling(true); 4217 N_COMM_indexes.pop_back(); 4218 } 4219 type = eSymbolTypeScopeEnd; 4220 break; 4221 4222 case N_LENG: 4223 // second stab entry with length information 4224 type = eSymbolTypeAdditional; 4225 break; 4226 4227 default: 4228 break; 4229 } 4230 } else { 4231 // uint8_t n_pext = N_PEXT & nlist.n_type; 4232 uint8_t n_type = N_TYPE & nlist.n_type; 4233 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4234 4235 switch (n_type) { 4236 case N_INDR: { 4237 const char *reexport_name_cstr = 4238 strtab_data.PeekCStr(nlist.n_value); 4239 if (reexport_name_cstr && reexport_name_cstr[0]) { 4240 type = eSymbolTypeReExported; 4241 ConstString reexport_name( 4242 reexport_name_cstr + 4243 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4244 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4245 set_value = false; 4246 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4247 indirect_symbol_names.insert( 4248 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4249 } else 4250 type = eSymbolTypeUndefined; 4251 } break; 4252 4253 case N_UNDF: 4254 if (symbol_name && symbol_name[0]) { 4255 ConstString undefined_name(symbol_name + 4256 ((symbol_name[0] == '_') ? 1 : 0)); 4257 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4258 } 4259 LLVM_FALLTHROUGH; 4260 4261 case N_PBUD: 4262 type = eSymbolTypeUndefined; 4263 break; 4264 4265 case N_ABS: 4266 type = eSymbolTypeAbsolute; 4267 break; 4268 4269 case N_SECT: { 4270 symbol_section = 4271 section_info.GetSection(nlist.n_sect, nlist.n_value); 4272 4273 if (!symbol_section) { 4274 // TODO: warn about this? 4275 add_nlist = false; 4276 break; 4277 } 4278 4279 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4280 type = eSymbolTypeException; 4281 } else { 4282 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4283 4284 switch (section_type) { 4285 case S_CSTRING_LITERALS: 4286 type = eSymbolTypeData; 4287 break; // section with only literal C strings 4288 case S_4BYTE_LITERALS: 4289 type = eSymbolTypeData; 4290 break; // section with only 4 byte literals 4291 case S_8BYTE_LITERALS: 4292 type = eSymbolTypeData; 4293 break; // section with only 8 byte literals 4294 case S_LITERAL_POINTERS: 4295 type = eSymbolTypeTrampoline; 4296 break; // section with only pointers to literals 4297 case S_NON_LAZY_SYMBOL_POINTERS: 4298 type = eSymbolTypeTrampoline; 4299 break; // section with only non-lazy symbol pointers 4300 case S_LAZY_SYMBOL_POINTERS: 4301 type = eSymbolTypeTrampoline; 4302 break; // section with only lazy symbol pointers 4303 case S_SYMBOL_STUBS: 4304 type = eSymbolTypeTrampoline; 4305 break; // section with only symbol stubs, byte size of stub in 4306 // the reserved2 field 4307 case S_MOD_INIT_FUNC_POINTERS: 4308 type = eSymbolTypeCode; 4309 break; // section with only function pointers for initialization 4310 case S_MOD_TERM_FUNC_POINTERS: 4311 type = eSymbolTypeCode; 4312 break; // section with only function pointers for termination 4313 case S_INTERPOSING: 4314 type = eSymbolTypeTrampoline; 4315 break; // section with only pairs of function pointers for 4316 // interposing 4317 case S_16BYTE_LITERALS: 4318 type = eSymbolTypeData; 4319 break; // section with only 16 byte literals 4320 case S_DTRACE_DOF: 4321 type = eSymbolTypeInstrumentation; 4322 break; 4323 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4324 type = eSymbolTypeTrampoline; 4325 break; 4326 default: 4327 switch (symbol_section->GetType()) { 4328 case lldb::eSectionTypeCode: 4329 type = eSymbolTypeCode; 4330 break; 4331 case eSectionTypeData: 4332 case eSectionTypeDataCString: // Inlined C string data 4333 case eSectionTypeDataCStringPointers: // Pointers to C string 4334 // data 4335 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4336 // the symbol table 4337 case eSectionTypeData4: 4338 case eSectionTypeData8: 4339 case eSectionTypeData16: 4340 type = eSymbolTypeData; 4341 break; 4342 default: 4343 break; 4344 } 4345 break; 4346 } 4347 4348 if (type == eSymbolTypeInvalid) { 4349 const char *symbol_sect_name = 4350 symbol_section->GetName().AsCString(); 4351 if (symbol_section->IsDescendant(text_section_sp.get())) { 4352 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4353 S_ATTR_SELF_MODIFYING_CODE | 4354 S_ATTR_SOME_INSTRUCTIONS)) 4355 type = eSymbolTypeData; 4356 else 4357 type = eSymbolTypeCode; 4358 } else if (symbol_section->IsDescendant( 4359 data_section_sp.get()) || 4360 symbol_section->IsDescendant( 4361 data_dirty_section_sp.get()) || 4362 symbol_section->IsDescendant( 4363 data_const_section_sp.get())) { 4364 if (symbol_sect_name && 4365 ::strstr(symbol_sect_name, "__objc") == 4366 symbol_sect_name) { 4367 type = eSymbolTypeRuntime; 4368 4369 if (symbol_name) { 4370 llvm::StringRef symbol_name_ref(symbol_name); 4371 if (symbol_name_ref.startswith("_OBJC_")) { 4372 static const llvm::StringRef g_objc_v2_prefix_class( 4373 "_OBJC_CLASS_$_"); 4374 static const llvm::StringRef g_objc_v2_prefix_metaclass( 4375 "_OBJC_METACLASS_$_"); 4376 static const llvm::StringRef g_objc_v2_prefix_ivar( 4377 "_OBJC_IVAR_$_"); 4378 if (symbol_name_ref.startswith( 4379 g_objc_v2_prefix_class)) { 4380 symbol_name_non_abi_mangled = symbol_name + 1; 4381 symbol_name = 4382 symbol_name + g_objc_v2_prefix_class.size(); 4383 type = eSymbolTypeObjCClass; 4384 demangled_is_synthesized = true; 4385 } else if (symbol_name_ref.startswith( 4386 g_objc_v2_prefix_metaclass)) { 4387 symbol_name_non_abi_mangled = symbol_name + 1; 4388 symbol_name = 4389 symbol_name + g_objc_v2_prefix_metaclass.size(); 4390 type = eSymbolTypeObjCMetaClass; 4391 demangled_is_synthesized = true; 4392 } else if (symbol_name_ref.startswith( 4393 g_objc_v2_prefix_ivar)) { 4394 symbol_name_non_abi_mangled = symbol_name + 1; 4395 symbol_name = 4396 symbol_name + g_objc_v2_prefix_ivar.size(); 4397 type = eSymbolTypeObjCIVar; 4398 demangled_is_synthesized = true; 4399 } 4400 } 4401 } 4402 } else if (symbol_sect_name && 4403 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4404 symbol_sect_name) { 4405 type = eSymbolTypeException; 4406 } else { 4407 type = eSymbolTypeData; 4408 } 4409 } else if (symbol_sect_name && 4410 ::strstr(symbol_sect_name, "__IMPORT") == 4411 symbol_sect_name) { 4412 type = eSymbolTypeTrampoline; 4413 } else if (symbol_section->IsDescendant( 4414 objc_section_sp.get())) { 4415 type = eSymbolTypeRuntime; 4416 if (symbol_name && symbol_name[0] == '.') { 4417 llvm::StringRef symbol_name_ref(symbol_name); 4418 static const llvm::StringRef g_objc_v1_prefix_class( 4419 ".objc_class_name_"); 4420 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4421 symbol_name_non_abi_mangled = symbol_name; 4422 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4423 type = eSymbolTypeObjCClass; 4424 demangled_is_synthesized = true; 4425 } 4426 } 4427 } 4428 } 4429 } 4430 } break; 4431 } 4432 } 4433 4434 if (add_nlist) { 4435 uint64_t symbol_value = nlist.n_value; 4436 4437 if (symbol_name_non_abi_mangled) { 4438 sym[sym_idx].GetMangled().SetMangledName( 4439 ConstString(symbol_name_non_abi_mangled)); 4440 sym[sym_idx].GetMangled().SetDemangledName( 4441 ConstString(symbol_name)); 4442 } else { 4443 bool symbol_name_is_mangled = false; 4444 4445 if (symbol_name && symbol_name[0] == '_') { 4446 symbol_name_is_mangled = symbol_name[1] == '_'; 4447 symbol_name++; // Skip the leading underscore 4448 } 4449 4450 if (symbol_name) { 4451 ConstString const_symbol_name(symbol_name); 4452 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4453 symbol_name_is_mangled); 4454 } 4455 } 4456 4457 if (is_gsym) { 4458 const char *gsym_name = sym[sym_idx] 4459 .GetMangled() 4460 .GetName(lldb::eLanguageTypeUnknown, 4461 Mangled::ePreferMangled) 4462 .GetCString(); 4463 if (gsym_name) 4464 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4465 } 4466 4467 if (symbol_section) { 4468 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4469 if (symbol_byte_size == 0 && function_starts_count > 0) { 4470 addr_t symbol_lookup_file_addr = nlist.n_value; 4471 // Do an exact address match for non-ARM addresses, else get the 4472 // closest since 4473 // the symbol might be a thumb symbol which has an address with 4474 // bit zero set 4475 FunctionStarts::Entry *func_start_entry = 4476 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4477 if (is_arm && func_start_entry) { 4478 // Verify that the function start address is the symbol address 4479 // (ARM) 4480 // or the symbol address + 1 (thumb) 4481 if (func_start_entry->addr != symbol_lookup_file_addr && 4482 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4483 // Not the right entry, NULL it out... 4484 func_start_entry = NULL; 4485 } 4486 } 4487 if (func_start_entry) { 4488 func_start_entry->data = true; 4489 4490 addr_t symbol_file_addr = func_start_entry->addr; 4491 if (is_arm) 4492 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4493 4494 const FunctionStarts::Entry *next_func_start_entry = 4495 function_starts.FindNextEntry(func_start_entry); 4496 const addr_t section_end_file_addr = 4497 section_file_addr + symbol_section->GetByteSize(); 4498 if (next_func_start_entry) { 4499 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4500 // Be sure the clear the Thumb address bit when we calculate 4501 // the size 4502 // from the current and next address 4503 if (is_arm) 4504 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4505 symbol_byte_size = std::min<lldb::addr_t>( 4506 next_symbol_file_addr - symbol_file_addr, 4507 section_end_file_addr - symbol_file_addr); 4508 } else { 4509 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4510 } 4511 } 4512 } 4513 symbol_value -= section_file_addr; 4514 } 4515 4516 if (is_debug == false) { 4517 if (type == eSymbolTypeCode) { 4518 // See if we can find a N_FUN entry for any code symbols. 4519 // If we do find a match, and the name matches, then we 4520 // can merge the two into just the function symbol to avoid 4521 // duplicate entries in the symbol table 4522 std::pair<ValueToSymbolIndexMap::const_iterator, 4523 ValueToSymbolIndexMap::const_iterator> 4524 range; 4525 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4526 if (range.first != range.second) { 4527 bool found_it = false; 4528 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4529 pos != range.second; ++pos) { 4530 if (sym[sym_idx].GetMangled().GetName( 4531 lldb::eLanguageTypeUnknown, 4532 Mangled::ePreferMangled) == 4533 sym[pos->second].GetMangled().GetName( 4534 lldb::eLanguageTypeUnknown, 4535 Mangled::ePreferMangled)) { 4536 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4537 // We just need the flags from the linker symbol, so put 4538 // these flags 4539 // into the N_FUN flags to avoid duplicate symbols in the 4540 // symbol table 4541 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4542 sym[pos->second].SetFlags(nlist.n_type << 16 | 4543 nlist.n_desc); 4544 if (resolver_addresses.find(nlist.n_value) != 4545 resolver_addresses.end()) 4546 sym[pos->second].SetType(eSymbolTypeResolver); 4547 sym[sym_idx].Clear(); 4548 found_it = true; 4549 break; 4550 } 4551 } 4552 if (found_it) 4553 continue; 4554 } else { 4555 if (resolver_addresses.find(nlist.n_value) != 4556 resolver_addresses.end()) 4557 type = eSymbolTypeResolver; 4558 } 4559 } else if (type == eSymbolTypeData || 4560 type == eSymbolTypeObjCClass || 4561 type == eSymbolTypeObjCMetaClass || 4562 type == eSymbolTypeObjCIVar) { 4563 // See if we can find a N_STSYM entry for any data symbols. 4564 // If we do find a match, and the name matches, then we 4565 // can merge the two into just the Static symbol to avoid 4566 // duplicate entries in the symbol table 4567 std::pair<ValueToSymbolIndexMap::const_iterator, 4568 ValueToSymbolIndexMap::const_iterator> 4569 range; 4570 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4571 if (range.first != range.second) { 4572 bool found_it = false; 4573 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4574 pos != range.second; ++pos) { 4575 if (sym[sym_idx].GetMangled().GetName( 4576 lldb::eLanguageTypeUnknown, 4577 Mangled::ePreferMangled) == 4578 sym[pos->second].GetMangled().GetName( 4579 lldb::eLanguageTypeUnknown, 4580 Mangled::ePreferMangled)) { 4581 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4582 // We just need the flags from the linker symbol, so put 4583 // these flags 4584 // into the N_STSYM flags to avoid duplicate symbols in the 4585 // symbol table 4586 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4587 sym[pos->second].SetFlags(nlist.n_type << 16 | 4588 nlist.n_desc); 4589 sym[sym_idx].Clear(); 4590 found_it = true; 4591 break; 4592 } 4593 } 4594 if (found_it) 4595 continue; 4596 } else { 4597 // Combine N_GSYM stab entries with the non stab symbol 4598 const char *gsym_name = sym[sym_idx] 4599 .GetMangled() 4600 .GetName(lldb::eLanguageTypeUnknown, 4601 Mangled::ePreferMangled) 4602 .GetCString(); 4603 if (gsym_name) { 4604 ConstNameToSymbolIndexMap::const_iterator pos = 4605 N_GSYM_name_to_sym_idx.find(gsym_name); 4606 if (pos != N_GSYM_name_to_sym_idx.end()) { 4607 const uint32_t GSYM_sym_idx = pos->second; 4608 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4609 // Copy the address, because often the N_GSYM address has an 4610 // invalid address of zero 4611 // when the global is a common symbol 4612 sym[GSYM_sym_idx].GetAddressRef().SetSection( 4613 symbol_section); 4614 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4615 // We just need the flags from the linker symbol, so put 4616 // these flags 4617 // into the N_GSYM flags to avoid duplicate symbols in the 4618 // symbol table 4619 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 4620 nlist.n_desc); 4621 sym[sym_idx].Clear(); 4622 continue; 4623 } 4624 } 4625 } 4626 } 4627 } 4628 4629 sym[sym_idx].SetID(nlist_idx); 4630 sym[sym_idx].SetType(type); 4631 if (set_value) { 4632 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4633 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4634 } 4635 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4636 4637 if (symbol_byte_size > 0) 4638 sym[sym_idx].SetByteSize(symbol_byte_size); 4639 4640 if (demangled_is_synthesized) 4641 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4642 4643 ++sym_idx; 4644 } else { 4645 sym[sym_idx].Clear(); 4646 } 4647 } 4648 4649 for (const auto &pos : reexport_shlib_needs_fixup) { 4650 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4651 if (undef_pos != undefined_name_to_desc.end()) { 4652 const uint8_t dylib_ordinal = 4653 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4654 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4655 sym[pos.first].SetReExportedSymbolSharedLibrary( 4656 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4657 } 4658 } 4659 } 4660 4661 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4662 4663 if (function_starts_count > 0) { 4664 uint32_t num_synthetic_function_symbols = 0; 4665 for (i = 0; i < function_starts_count; ++i) { 4666 if (function_starts.GetEntryRef(i).data == false) 4667 ++num_synthetic_function_symbols; 4668 } 4669 4670 if (num_synthetic_function_symbols > 0) { 4671 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4672 num_syms = sym_idx + num_synthetic_function_symbols; 4673 sym = symtab->Resize(num_syms); 4674 } 4675 for (i = 0; i < function_starts_count; ++i) { 4676 const FunctionStarts::Entry *func_start_entry = 4677 function_starts.GetEntryAtIndex(i); 4678 if (func_start_entry->data == false) { 4679 addr_t symbol_file_addr = func_start_entry->addr; 4680 uint32_t symbol_flags = 0; 4681 if (is_arm) { 4682 if (symbol_file_addr & 1) 4683 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4684 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4685 } 4686 Address symbol_addr; 4687 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4688 SectionSP symbol_section(symbol_addr.GetSection()); 4689 uint32_t symbol_byte_size = 0; 4690 if (symbol_section) { 4691 const addr_t section_file_addr = 4692 symbol_section->GetFileAddress(); 4693 const FunctionStarts::Entry *next_func_start_entry = 4694 function_starts.FindNextEntry(func_start_entry); 4695 const addr_t section_end_file_addr = 4696 section_file_addr + symbol_section->GetByteSize(); 4697 if (next_func_start_entry) { 4698 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4699 if (is_arm) 4700 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4701 symbol_byte_size = std::min<lldb::addr_t>( 4702 next_symbol_file_addr - symbol_file_addr, 4703 section_end_file_addr - symbol_file_addr); 4704 } else { 4705 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4706 } 4707 sym[sym_idx].SetID(synthetic_sym_id++); 4708 sym[sym_idx].GetMangled().SetDemangledName( 4709 GetNextSyntheticSymbolName()); 4710 sym[sym_idx].SetType(eSymbolTypeCode); 4711 sym[sym_idx].SetIsSynthetic(true); 4712 sym[sym_idx].GetAddressRef() = symbol_addr; 4713 if (symbol_flags) 4714 sym[sym_idx].SetFlags(symbol_flags); 4715 if (symbol_byte_size) 4716 sym[sym_idx].SetByteSize(symbol_byte_size); 4717 ++sym_idx; 4718 } 4719 } 4720 } 4721 } 4722 } 4723 } 4724 4725 // Trim our symbols down to just what we ended up with after 4726 // removing any symbols. 4727 if (sym_idx < num_syms) { 4728 num_syms = sym_idx; 4729 sym = symtab->Resize(num_syms); 4730 } 4731 4732 // Now synthesize indirect symbols 4733 if (m_dysymtab.nindirectsyms != 0) { 4734 if (indirect_symbol_index_data.GetByteSize()) { 4735 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4736 m_nlist_idx_to_sym_idx.end(); 4737 4738 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4739 ++sect_idx) { 4740 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4741 S_SYMBOL_STUBS) { 4742 uint32_t symbol_stub_byte_size = 4743 m_mach_sections[sect_idx].reserved2; 4744 if (symbol_stub_byte_size == 0) 4745 continue; 4746 4747 const uint32_t num_symbol_stubs = 4748 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4749 4750 if (num_symbol_stubs == 0) 4751 continue; 4752 4753 const uint32_t symbol_stub_index_offset = 4754 m_mach_sections[sect_idx].reserved1; 4755 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; 4756 ++stub_idx) { 4757 const uint32_t symbol_stub_index = 4758 symbol_stub_index_offset + stub_idx; 4759 const lldb::addr_t symbol_stub_addr = 4760 m_mach_sections[sect_idx].addr + 4761 (stub_idx * symbol_stub_byte_size); 4762 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4763 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4764 symbol_stub_offset, 4)) { 4765 const uint32_t stub_sym_id = 4766 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4767 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4768 continue; 4769 4770 NListIndexToSymbolIndexMap::const_iterator index_pos = 4771 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4772 Symbol *stub_symbol = NULL; 4773 if (index_pos != end_index_pos) { 4774 // We have a remapping from the original nlist index to 4775 // a current symbol index, so just look this up by index 4776 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4777 } else { 4778 // We need to lookup a symbol using the original nlist 4779 // symbol index since this index is coming from the 4780 // S_SYMBOL_STUBS 4781 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4782 } 4783 4784 if (stub_symbol) { 4785 Address so_addr(symbol_stub_addr, section_list); 4786 4787 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4788 // Change the external symbol into a trampoline that makes 4789 // sense 4790 // These symbols were N_UNDF N_EXT, and are useless to us, 4791 // so we 4792 // can re-use them so we don't have to make up a synthetic 4793 // symbol 4794 // for no good reason. 4795 if (resolver_addresses.find(symbol_stub_addr) == 4796 resolver_addresses.end()) 4797 stub_symbol->SetType(eSymbolTypeTrampoline); 4798 else 4799 stub_symbol->SetType(eSymbolTypeResolver); 4800 stub_symbol->SetExternal(false); 4801 stub_symbol->GetAddressRef() = so_addr; 4802 stub_symbol->SetByteSize(symbol_stub_byte_size); 4803 } else { 4804 // Make a synthetic symbol to describe the trampoline stub 4805 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4806 if (sym_idx >= num_syms) { 4807 sym = symtab->Resize(++num_syms); 4808 stub_symbol = NULL; // this pointer no longer valid 4809 } 4810 sym[sym_idx].SetID(synthetic_sym_id++); 4811 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4812 if (resolver_addresses.find(symbol_stub_addr) == 4813 resolver_addresses.end()) 4814 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4815 else 4816 sym[sym_idx].SetType(eSymbolTypeResolver); 4817 sym[sym_idx].SetIsSynthetic(true); 4818 sym[sym_idx].GetAddressRef() = so_addr; 4819 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4820 ++sym_idx; 4821 } 4822 } else { 4823 if (log) 4824 log->Warning("symbol stub referencing symbol table symbol " 4825 "%u that isn't in our minimal symbol table, " 4826 "fix this!!!", 4827 stub_sym_id); 4828 } 4829 } 4830 } 4831 } 4832 } 4833 } 4834 } 4835 4836 if (!trie_entries.empty()) { 4837 for (const auto &e : trie_entries) { 4838 if (e.entry.import_name) { 4839 // Only add indirect symbols from the Trie entries if we 4840 // didn't have a N_INDR nlist entry for this already 4841 if (indirect_symbol_names.find(e.entry.name) == 4842 indirect_symbol_names.end()) { 4843 // Make a synthetic symbol to describe re-exported symbol. 4844 if (sym_idx >= num_syms) 4845 sym = symtab->Resize(++num_syms); 4846 sym[sym_idx].SetID(synthetic_sym_id++); 4847 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4848 sym[sym_idx].SetType(eSymbolTypeReExported); 4849 sym[sym_idx].SetIsSynthetic(true); 4850 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4851 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4852 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4853 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4854 } 4855 ++sym_idx; 4856 } 4857 } 4858 } 4859 } 4860 4861 // StreamFile s(stdout, false); 4862 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4863 // symtab->Dump(&s, NULL, eSortOrderNone); 4864 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4865 // sizes 4866 symtab->CalculateSymbolSizes(); 4867 4868 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4869 // symtab->Dump(&s, NULL, eSortOrderNone); 4870 4871 return symtab->GetNumSymbols(); 4872 } 4873 return 0; 4874 } 4875 4876 void ObjectFileMachO::Dump(Stream *s) { 4877 ModuleSP module_sp(GetModule()); 4878 if (module_sp) { 4879 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4880 s->Printf("%p: ", static_cast<void *>(this)); 4881 s->Indent(); 4882 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4883 s->PutCString("ObjectFileMachO64"); 4884 else 4885 s->PutCString("ObjectFileMachO32"); 4886 4887 ArchSpec header_arch; 4888 GetArchitecture(header_arch); 4889 4890 *s << ", file = '" << m_file 4891 << "', arch = " << header_arch.GetArchitectureName() << "\n"; 4892 4893 SectionList *sections = GetSectionList(); 4894 if (sections) 4895 sections->Dump(s, NULL, true, UINT32_MAX); 4896 4897 if (m_symtab_ap.get()) 4898 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4899 } 4900 } 4901 4902 bool ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4903 const lldb_private::DataExtractor &data, 4904 lldb::offset_t lc_offset, 4905 lldb_private::UUID &uuid) { 4906 uint32_t i; 4907 struct uuid_command load_cmd; 4908 4909 lldb::offset_t offset = lc_offset; 4910 for (i = 0; i < header.ncmds; ++i) { 4911 const lldb::offset_t cmd_offset = offset; 4912 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4913 break; 4914 4915 if (load_cmd.cmd == LC_UUID) { 4916 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4917 4918 if (uuid_bytes) { 4919 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4920 // We pretend these object files have no UUID to prevent crashing. 4921 4922 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4923 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4924 0xbb, 0x14, 0xf0, 0x0d}; 4925 4926 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4927 return false; 4928 4929 uuid.SetBytes(uuid_bytes); 4930 return true; 4931 } 4932 return false; 4933 } 4934 offset = cmd_offset + load_cmd.cmdsize; 4935 } 4936 return false; 4937 } 4938 4939 bool ObjectFileMachO::GetArchitecture(const llvm::MachO::mach_header &header, 4940 const lldb_private::DataExtractor &data, 4941 lldb::offset_t lc_offset, 4942 ArchSpec &arch) { 4943 arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4944 4945 if (arch.IsValid()) { 4946 llvm::Triple &triple = arch.GetTriple(); 4947 4948 // Set OS to an unspecified unknown or a "*" so it can match any OS 4949 triple.setOS(llvm::Triple::UnknownOS); 4950 triple.setOSName(llvm::StringRef()); 4951 4952 if (header.filetype == MH_PRELOAD) { 4953 if (header.cputype == CPU_TYPE_ARM) { 4954 // If this is a 32-bit arm binary, and it's a standalone binary, 4955 // force the Vendor to Apple so we don't accidentally pick up 4956 // the generic armv7 ABI at runtime. Apple's armv7 ABI always uses 4957 // r7 for the frame pointer register; most other armv7 ABIs use a 4958 // combination of r7 and r11. 4959 triple.setVendor(llvm::Triple::Apple); 4960 } else { 4961 // Set vendor to an unspecified unknown or a "*" so it can match any 4962 // vendor 4963 // This is required for correct behavior of EFI debugging on x86_64 4964 triple.setVendor(llvm::Triple::UnknownVendor); 4965 triple.setVendorName(llvm::StringRef()); 4966 } 4967 return true; 4968 } else { 4969 struct load_command load_cmd; 4970 4971 lldb::offset_t offset = lc_offset; 4972 for (uint32_t i = 0; i < header.ncmds; ++i) { 4973 const lldb::offset_t cmd_offset = offset; 4974 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4975 break; 4976 4977 switch (load_cmd.cmd) { 4978 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4979 triple.setOS(llvm::Triple::IOS); 4980 return true; 4981 4982 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4983 triple.setOS(llvm::Triple::MacOSX); 4984 return true; 4985 4986 case llvm::MachO::LC_VERSION_MIN_TVOS: 4987 triple.setOS(llvm::Triple::TvOS); 4988 return true; 4989 4990 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4991 triple.setOS(llvm::Triple::WatchOS); 4992 return true; 4993 4994 default: 4995 break; 4996 } 4997 4998 offset = cmd_offset + load_cmd.cmdsize; 4999 } 5000 5001 if (header.filetype != MH_KEXT_BUNDLE) { 5002 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 5003 // so lets not say our Vendor is Apple, leave it as an unspecified 5004 // unknown 5005 triple.setVendor(llvm::Triple::UnknownVendor); 5006 triple.setVendorName(llvm::StringRef()); 5007 } 5008 } 5009 } 5010 return arch.IsValid(); 5011 } 5012 5013 bool ObjectFileMachO::GetUUID(lldb_private::UUID *uuid) { 5014 ModuleSP module_sp(GetModule()); 5015 if (module_sp) { 5016 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5017 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5018 return GetUUID(m_header, m_data, offset, *uuid); 5019 } 5020 return false; 5021 } 5022 5023 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5024 uint32_t count = 0; 5025 ModuleSP module_sp(GetModule()); 5026 if (module_sp) { 5027 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5028 struct load_command load_cmd; 5029 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5030 std::vector<std::string> rpath_paths; 5031 std::vector<std::string> rpath_relative_paths; 5032 std::vector<std::string> at_exec_relative_paths; 5033 const bool resolve_path = false; // Don't resolve the dependent file paths 5034 // since they may not reside on this system 5035 uint32_t i; 5036 for (i = 0; i < m_header.ncmds; ++i) { 5037 const uint32_t cmd_offset = offset; 5038 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5039 break; 5040 5041 switch (load_cmd.cmd) { 5042 case LC_RPATH: 5043 case LC_LOAD_DYLIB: 5044 case LC_LOAD_WEAK_DYLIB: 5045 case LC_REEXPORT_DYLIB: 5046 case LC_LOAD_DYLINKER: 5047 case LC_LOADFVMLIB: 5048 case LC_LOAD_UPWARD_DYLIB: { 5049 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5050 const char *path = m_data.PeekCStr(name_offset); 5051 if (path) { 5052 if (load_cmd.cmd == LC_RPATH) 5053 rpath_paths.push_back(path); 5054 else { 5055 if (path[0] == '@') { 5056 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5057 rpath_relative_paths.push_back(path + strlen("@rpath")); 5058 else if (strncmp(path, "@executable_path", 5059 strlen("@executable_path")) == 0) 5060 at_exec_relative_paths.push_back(path 5061 + strlen("@executable_path")); 5062 } else { 5063 FileSpec file_spec(path, resolve_path); 5064 if (files.AppendIfUnique(file_spec)) 5065 count++; 5066 } 5067 } 5068 } 5069 } break; 5070 5071 default: 5072 break; 5073 } 5074 offset = cmd_offset + load_cmd.cmdsize; 5075 } 5076 5077 FileSpec this_file_spec(m_file); 5078 this_file_spec.ResolvePath(); 5079 5080 if (!rpath_paths.empty()) { 5081 // Fixup all LC_RPATH values to be absolute paths 5082 std::string loader_path("@loader_path"); 5083 std::string executable_path("@executable_path"); 5084 for (auto &rpath : rpath_paths) { 5085 if (rpath.find(loader_path) == 0) { 5086 rpath.erase(0, loader_path.size()); 5087 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5088 } else if (rpath.find(executable_path) == 0) { 5089 rpath.erase(0, executable_path.size()); 5090 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5091 } 5092 } 5093 5094 for (const auto &rpath_relative_path : rpath_relative_paths) { 5095 for (const auto &rpath : rpath_paths) { 5096 std::string path = rpath; 5097 path += rpath_relative_path; 5098 // It is OK to resolve this path because we must find a file on 5099 // disk for us to accept it anyway if it is rpath relative. 5100 FileSpec file_spec(path, true); 5101 // Remove any redundant parts of the path (like "../foo") since 5102 // LC_RPATH values often contain "..". 5103 file_spec = file_spec.GetNormalizedPath(); 5104 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) { 5105 count++; 5106 break; 5107 } 5108 } 5109 } 5110 } 5111 5112 // We may have @executable_paths but no RPATHS. Figure those out here. 5113 // Only do this if this object file is the executable. We have no way to 5114 // get back to the actual executable otherwise, so we won't get the right 5115 // path. 5116 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5117 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5118 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5119 FileSpec file_spec = 5120 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5121 file_spec = file_spec.GetNormalizedPath(); 5122 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) 5123 count++; 5124 } 5125 } 5126 } 5127 return count; 5128 } 5129 5130 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5131 // If the object file is not an executable it can't hold the entry point. 5132 // m_entry_point_address 5133 // is initialized to an invalid address, so we can just return that. 5134 // If m_entry_point_address is valid it means we've found it already, so 5135 // return the cached value. 5136 5137 if (!IsExecutable() || m_entry_point_address.IsValid()) 5138 return m_entry_point_address; 5139 5140 // Otherwise, look for the UnixThread or Thread command. The data for the 5141 // Thread command is given in 5142 // /usr/include/mach-o.h, but it is basically: 5143 // 5144 // uint32_t flavor - this is the flavor argument you would pass to 5145 // thread_get_state 5146 // uint32_t count - this is the count of longs in the thread state data 5147 // struct XXX_thread_state state - this is the structure from 5148 // <machine/thread_status.h> corresponding to the flavor. 5149 // <repeat this trio> 5150 // 5151 // So we just keep reading the various register flavors till we find the GPR 5152 // one, then read the PC out of there. 5153 // FIXME: We will need to have a "RegisterContext data provider" class at some 5154 // point that can get all the registers 5155 // out of data in this form & attach them to a given thread. That should 5156 // underlie the MacOS X User process plugin, 5157 // and we'll also need it for the MacOS X Core File process plugin. When we 5158 // have that we can also use it here. 5159 // 5160 // For now we hard-code the offsets and flavors we need: 5161 // 5162 // 5163 5164 ModuleSP module_sp(GetModule()); 5165 if (module_sp) { 5166 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5167 struct load_command load_cmd; 5168 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5169 uint32_t i; 5170 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5171 bool done = false; 5172 5173 for (i = 0; i < m_header.ncmds; ++i) { 5174 const lldb::offset_t cmd_offset = offset; 5175 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5176 break; 5177 5178 switch (load_cmd.cmd) { 5179 case LC_UNIXTHREAD: 5180 case LC_THREAD: { 5181 while (offset < cmd_offset + load_cmd.cmdsize) { 5182 uint32_t flavor = m_data.GetU32(&offset); 5183 uint32_t count = m_data.GetU32(&offset); 5184 if (count == 0) { 5185 // We've gotten off somehow, log and exit; 5186 return m_entry_point_address; 5187 } 5188 5189 switch (m_header.cputype) { 5190 case llvm::MachO::CPU_TYPE_ARM: 5191 if (flavor == 1 || 5192 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 from 5193 // mach/arm/thread_status.h 5194 { 5195 offset += 60; // This is the offset of pc in the GPR thread state 5196 // data structure. 5197 start_address = m_data.GetU32(&offset); 5198 done = true; 5199 } 5200 break; 5201 case llvm::MachO::CPU_TYPE_ARM64: 5202 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5203 { 5204 offset += 256; // This is the offset of pc in the GPR thread state 5205 // data structure. 5206 start_address = m_data.GetU64(&offset); 5207 done = true; 5208 } 5209 break; 5210 case llvm::MachO::CPU_TYPE_I386: 5211 if (flavor == 5212 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5213 { 5214 offset += 40; // This is the offset of eip in the GPR thread state 5215 // data structure. 5216 start_address = m_data.GetU32(&offset); 5217 done = true; 5218 } 5219 break; 5220 case llvm::MachO::CPU_TYPE_X86_64: 5221 if (flavor == 5222 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5223 { 5224 offset += 16 * 8; // This is the offset of rip in the GPR thread 5225 // state data structure. 5226 start_address = m_data.GetU64(&offset); 5227 done = true; 5228 } 5229 break; 5230 default: 5231 return m_entry_point_address; 5232 } 5233 // Haven't found the GPR flavor yet, skip over the data for this 5234 // flavor: 5235 if (done) 5236 break; 5237 offset += count * 4; 5238 } 5239 } break; 5240 case LC_MAIN: { 5241 ConstString text_segment_name("__TEXT"); 5242 uint64_t entryoffset = m_data.GetU64(&offset); 5243 SectionSP text_segment_sp = 5244 GetSectionList()->FindSectionByName(text_segment_name); 5245 if (text_segment_sp) { 5246 done = true; 5247 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5248 } 5249 } break; 5250 5251 default: 5252 break; 5253 } 5254 if (done) 5255 break; 5256 5257 // Go to the next load command: 5258 offset = cmd_offset + load_cmd.cmdsize; 5259 } 5260 5261 if (start_address != LLDB_INVALID_ADDRESS) { 5262 // We got the start address from the load commands, so now resolve that 5263 // address in the sections 5264 // of this ObjectFile: 5265 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5266 start_address, GetSectionList())) { 5267 m_entry_point_address.Clear(); 5268 } 5269 } else { 5270 // We couldn't read the UnixThread load command - maybe it wasn't there. 5271 // As a fallback look for the 5272 // "start" symbol in the main executable. 5273 5274 ModuleSP module_sp(GetModule()); 5275 5276 if (module_sp) { 5277 SymbolContextList contexts; 5278 SymbolContext context; 5279 if (module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5280 eSymbolTypeCode, contexts)) { 5281 if (contexts.GetContextAtIndex(0, context)) 5282 m_entry_point_address = context.symbol->GetAddress(); 5283 } 5284 } 5285 } 5286 } 5287 5288 return m_entry_point_address; 5289 } 5290 5291 lldb_private::Address ObjectFileMachO::GetHeaderAddress() { 5292 lldb_private::Address header_addr; 5293 SectionList *section_list = GetSectionList(); 5294 if (section_list) { 5295 SectionSP text_segment_sp( 5296 section_list->FindSectionByName(GetSegmentNameTEXT())); 5297 if (text_segment_sp) { 5298 header_addr.SetSection(text_segment_sp); 5299 header_addr.SetOffset(0); 5300 } 5301 } 5302 return header_addr; 5303 } 5304 5305 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5306 ModuleSP module_sp(GetModule()); 5307 if (module_sp) { 5308 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5309 if (!m_thread_context_offsets_valid) { 5310 m_thread_context_offsets_valid = true; 5311 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5312 FileRangeArray::Entry file_range; 5313 thread_command thread_cmd; 5314 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5315 const uint32_t cmd_offset = offset; 5316 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 5317 break; 5318 5319 if (thread_cmd.cmd == LC_THREAD) { 5320 file_range.SetRangeBase(offset); 5321 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5322 m_thread_context_offsets.Append(file_range); 5323 } 5324 offset = cmd_offset + thread_cmd.cmdsize; 5325 } 5326 } 5327 } 5328 return m_thread_context_offsets.GetSize(); 5329 } 5330 5331 std::string ObjectFileMachO::GetIdentifierString() { 5332 std::string result; 5333 ModuleSP module_sp(GetModule()); 5334 if (module_sp) { 5335 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5336 5337 // First, look over the load commands for an LC_NOTE load command 5338 // with data_owner string "kern ver str" & use that if found. 5339 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5340 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5341 const uint32_t cmd_offset = offset; 5342 load_command lc; 5343 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5344 break; 5345 if (lc.cmd == LC_NOTE) 5346 { 5347 char data_owner[17]; 5348 m_data.CopyData (offset, 16, data_owner); 5349 data_owner[16] = '\0'; 5350 offset += 16; 5351 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5352 uint64_t size = m_data.GetU64_unchecked (&offset); 5353 5354 // "kern ver str" has a uint32_t version and then a 5355 // nul terminated c-string. 5356 if (strcmp ("kern ver str", data_owner) == 0) 5357 { 5358 offset = fileoff; 5359 uint32_t version; 5360 if (m_data.GetU32 (&offset, &version, 1) != nullptr) 5361 { 5362 if (version == 1) 5363 { 5364 uint32_t strsize = size - sizeof (uint32_t); 5365 char *buf = (char*) malloc (strsize); 5366 if (buf) 5367 { 5368 m_data.CopyData (offset, strsize, buf); 5369 buf[strsize - 1] = '\0'; 5370 result = buf; 5371 if (buf) 5372 free (buf); 5373 return result; 5374 } 5375 } 5376 } 5377 } 5378 } 5379 offset = cmd_offset + lc.cmdsize; 5380 } 5381 5382 // Second, make a pass over the load commands looking for an 5383 // obsolete LC_IDENT load command. 5384 offset = MachHeaderSizeFromMagic(m_header.magic); 5385 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5386 const uint32_t cmd_offset = offset; 5387 struct ident_command ident_command; 5388 if (m_data.GetU32(&offset, &ident_command, 2) == NULL) 5389 break; 5390 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5391 char *buf = (char *) malloc (ident_command.cmdsize); 5392 if (buf != nullptr 5393 && m_data.CopyData (offset, ident_command.cmdsize, buf) == ident_command.cmdsize) { 5394 buf[ident_command.cmdsize - 1] = '\0'; 5395 result = buf; 5396 } 5397 if (buf) 5398 free (buf); 5399 } 5400 offset = cmd_offset + ident_command.cmdsize; 5401 } 5402 5403 } 5404 return result; 5405 } 5406 5407 bool ObjectFileMachO::GetCorefileMainBinaryInfo (addr_t &address, UUID &uuid) { 5408 address = LLDB_INVALID_ADDRESS; 5409 uuid.Clear(); 5410 ModuleSP module_sp(GetModule()); 5411 if (module_sp) { 5412 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5413 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5414 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5415 const uint32_t cmd_offset = offset; 5416 load_command lc; 5417 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5418 break; 5419 if (lc.cmd == LC_NOTE) 5420 { 5421 char data_owner[17]; 5422 memset (data_owner, 0, sizeof (data_owner)); 5423 m_data.CopyData (offset, 16, data_owner); 5424 offset += 16; 5425 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5426 uint64_t size = m_data.GetU64_unchecked (&offset); 5427 5428 // "main bin spec" (main binary specification) data payload is formatted: 5429 // uint32_t version [currently 1] 5430 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user process] 5431 // uint64_t address [ UINT64_MAX if address not specified ] 5432 // uuid_t uuid [ all zero's if uuid not specified ] 5433 // uint32_t log2_pagesize [ process page size in log base 2, e.g. 4k pages are 12. 0 for unspecified ] 5434 5435 if (strcmp ("main bin spec", data_owner) == 0 && size >= 32) 5436 { 5437 offset = fileoff; 5438 uint32_t version; 5439 if (m_data.GetU32 (&offset, &version, 1) != nullptr && version == 1) 5440 { 5441 uint32_t type = 0; 5442 uuid_t raw_uuid; 5443 memset (raw_uuid, 0, sizeof (uuid_t)); 5444 5445 if (m_data.GetU32 (&offset, &type, 1) 5446 && m_data.GetU64 (&offset, &address, 1) 5447 && m_data.CopyData (offset, sizeof (uuid_t), raw_uuid) != 0 5448 && uuid.SetBytes (raw_uuid, sizeof (uuid_t))) 5449 { 5450 return true; 5451 } 5452 } 5453 } 5454 } 5455 offset = cmd_offset + lc.cmdsize; 5456 } 5457 } 5458 return false; 5459 } 5460 5461 lldb::RegisterContextSP 5462 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5463 lldb_private::Thread &thread) { 5464 lldb::RegisterContextSP reg_ctx_sp; 5465 5466 ModuleSP module_sp(GetModule()); 5467 if (module_sp) { 5468 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5469 if (!m_thread_context_offsets_valid) 5470 GetNumThreadContexts(); 5471 5472 const FileRangeArray::Entry *thread_context_file_range = 5473 m_thread_context_offsets.GetEntryAtIndex(idx); 5474 if (thread_context_file_range) { 5475 5476 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5477 thread_context_file_range->GetByteSize()); 5478 5479 switch (m_header.cputype) { 5480 case llvm::MachO::CPU_TYPE_ARM64: 5481 reg_ctx_sp.reset(new RegisterContextDarwin_arm64_Mach(thread, data)); 5482 break; 5483 5484 case llvm::MachO::CPU_TYPE_ARM: 5485 reg_ctx_sp.reset(new RegisterContextDarwin_arm_Mach(thread, data)); 5486 break; 5487 5488 case llvm::MachO::CPU_TYPE_I386: 5489 reg_ctx_sp.reset(new RegisterContextDarwin_i386_Mach(thread, data)); 5490 break; 5491 5492 case llvm::MachO::CPU_TYPE_X86_64: 5493 reg_ctx_sp.reset(new RegisterContextDarwin_x86_64_Mach(thread, data)); 5494 break; 5495 } 5496 } 5497 } 5498 return reg_ctx_sp; 5499 } 5500 5501 ObjectFile::Type ObjectFileMachO::CalculateType() { 5502 switch (m_header.filetype) { 5503 case MH_OBJECT: // 0x1u 5504 if (GetAddressByteSize() == 4) { 5505 // 32 bit kexts are just object files, but they do have a valid 5506 // UUID load command. 5507 UUID uuid; 5508 if (GetUUID(&uuid)) { 5509 // this checking for the UUID load command is not enough 5510 // we could eventually look for the symbol named 5511 // "OSKextGetCurrentIdentifier" as this is required of kexts 5512 if (m_strata == eStrataInvalid) 5513 m_strata = eStrataKernel; 5514 return eTypeSharedLibrary; 5515 } 5516 } 5517 return eTypeObjectFile; 5518 5519 case MH_EXECUTE: 5520 return eTypeExecutable; // 0x2u 5521 case MH_FVMLIB: 5522 return eTypeSharedLibrary; // 0x3u 5523 case MH_CORE: 5524 return eTypeCoreFile; // 0x4u 5525 case MH_PRELOAD: 5526 return eTypeSharedLibrary; // 0x5u 5527 case MH_DYLIB: 5528 return eTypeSharedLibrary; // 0x6u 5529 case MH_DYLINKER: 5530 return eTypeDynamicLinker; // 0x7u 5531 case MH_BUNDLE: 5532 return eTypeSharedLibrary; // 0x8u 5533 case MH_DYLIB_STUB: 5534 return eTypeStubLibrary; // 0x9u 5535 case MH_DSYM: 5536 return eTypeDebugInfo; // 0xAu 5537 case MH_KEXT_BUNDLE: 5538 return eTypeSharedLibrary; // 0xBu 5539 default: 5540 break; 5541 } 5542 return eTypeUnknown; 5543 } 5544 5545 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5546 switch (m_header.filetype) { 5547 case MH_OBJECT: // 0x1u 5548 { 5549 // 32 bit kexts are just object files, but they do have a valid 5550 // UUID load command. 5551 UUID uuid; 5552 if (GetUUID(&uuid)) { 5553 // this checking for the UUID load command is not enough 5554 // we could eventually look for the symbol named 5555 // "OSKextGetCurrentIdentifier" as this is required of kexts 5556 if (m_type == eTypeInvalid) 5557 m_type = eTypeSharedLibrary; 5558 5559 return eStrataKernel; 5560 } 5561 } 5562 return eStrataUnknown; 5563 5564 case MH_EXECUTE: // 0x2u 5565 // Check for the MH_DYLDLINK bit in the flags 5566 if (m_header.flags & MH_DYLDLINK) { 5567 return eStrataUser; 5568 } else { 5569 SectionList *section_list = GetSectionList(); 5570 if (section_list) { 5571 static ConstString g_kld_section_name("__KLD"); 5572 if (section_list->FindSectionByName(g_kld_section_name)) 5573 return eStrataKernel; 5574 } 5575 } 5576 return eStrataRawImage; 5577 5578 case MH_FVMLIB: 5579 return eStrataUser; // 0x3u 5580 case MH_CORE: 5581 return eStrataUnknown; // 0x4u 5582 case MH_PRELOAD: 5583 return eStrataRawImage; // 0x5u 5584 case MH_DYLIB: 5585 return eStrataUser; // 0x6u 5586 case MH_DYLINKER: 5587 return eStrataUser; // 0x7u 5588 case MH_BUNDLE: 5589 return eStrataUser; // 0x8u 5590 case MH_DYLIB_STUB: 5591 return eStrataUser; // 0x9u 5592 case MH_DSYM: 5593 return eStrataUnknown; // 0xAu 5594 case MH_KEXT_BUNDLE: 5595 return eStrataKernel; // 0xBu 5596 default: 5597 break; 5598 } 5599 return eStrataUnknown; 5600 } 5601 5602 uint32_t ObjectFileMachO::GetVersion(uint32_t *versions, 5603 uint32_t num_versions) { 5604 ModuleSP module_sp(GetModule()); 5605 if (module_sp) { 5606 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5607 struct dylib_command load_cmd; 5608 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5609 uint32_t version_cmd = 0; 5610 uint64_t version = 0; 5611 uint32_t i; 5612 for (i = 0; i < m_header.ncmds; ++i) { 5613 const lldb::offset_t cmd_offset = offset; 5614 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5615 break; 5616 5617 if (load_cmd.cmd == LC_ID_DYLIB) { 5618 if (version_cmd == 0) { 5619 version_cmd = load_cmd.cmd; 5620 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5621 break; 5622 version = load_cmd.dylib.current_version; 5623 } 5624 break; // Break for now unless there is another more complete version 5625 // number load command in the future. 5626 } 5627 offset = cmd_offset + load_cmd.cmdsize; 5628 } 5629 5630 if (version_cmd == LC_ID_DYLIB) { 5631 if (versions != NULL && num_versions > 0) { 5632 if (num_versions > 0) 5633 versions[0] = (version & 0xFFFF0000ull) >> 16; 5634 if (num_versions > 1) 5635 versions[1] = (version & 0x0000FF00ull) >> 8; 5636 if (num_versions > 2) 5637 versions[2] = (version & 0x000000FFull); 5638 // Fill in an remaining version numbers with invalid values 5639 for (i = 3; i < num_versions; ++i) 5640 versions[i] = UINT32_MAX; 5641 } 5642 // The LC_ID_DYLIB load command has a version with 3 version numbers 5643 // in it, so always return 3 5644 return 3; 5645 } 5646 } 5647 return false; 5648 } 5649 5650 bool ObjectFileMachO::GetArchitecture(ArchSpec &arch) { 5651 ModuleSP module_sp(GetModule()); 5652 if (module_sp) { 5653 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5654 return GetArchitecture(m_header, m_data, 5655 MachHeaderSizeFromMagic(m_header.magic), arch); 5656 } 5657 return false; 5658 } 5659 5660 UUID ObjectFileMachO::GetProcessSharedCacheUUID(Process *process) { 5661 UUID uuid; 5662 if (process && process->GetDynamicLoader()) { 5663 DynamicLoader *dl = process->GetDynamicLoader(); 5664 addr_t load_address; 5665 LazyBool using_shared_cache; 5666 LazyBool private_shared_cache; 5667 dl->GetSharedCacheInformation(load_address, uuid, using_shared_cache, 5668 private_shared_cache); 5669 } 5670 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5671 if (log) 5672 log->Printf("inferior process shared cache has a UUID of %s", uuid.GetAsString().c_str()); 5673 return uuid; 5674 } 5675 5676 UUID ObjectFileMachO::GetLLDBSharedCacheUUID() { 5677 UUID uuid; 5678 #if defined(__APPLE__) && \ 5679 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5680 uint8_t *(*dyld_get_all_image_infos)(void); 5681 dyld_get_all_image_infos = 5682 (uint8_t * (*)())dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5683 if (dyld_get_all_image_infos) { 5684 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5685 if (dyld_all_image_infos_address) { 5686 uint32_t *version = (uint32_t *) 5687 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5688 if (*version >= 13) { 5689 uuid_t *sharedCacheUUID_address = 0; 5690 int wordsize = sizeof(uint8_t *); 5691 if (wordsize == 8) { 5692 sharedCacheUUID_address = 5693 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5694 160); // sharedCacheUUID <mach-o/dyld_images.h> 5695 } else { 5696 sharedCacheUUID_address = 5697 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5698 84); // sharedCacheUUID <mach-o/dyld_images.h> 5699 } 5700 uuid.SetBytes(sharedCacheUUID_address); 5701 } 5702 } 5703 } else { 5704 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5705 bool *(*dyld_get_shared_cache_uuid)(uuid_t); 5706 dyld_get_shared_cache_uuid = (bool *(*)(uuid_t)) 5707 dlsym(RTLD_DEFAULT, "_dyld_get_shared_cache_uuid"); 5708 if (dyld_get_shared_cache_uuid) { 5709 uuid_t tmp_uuid; 5710 if (dyld_get_shared_cache_uuid(tmp_uuid)) 5711 uuid.SetBytes(tmp_uuid); 5712 } 5713 } 5714 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5715 if (log && uuid.IsValid()) 5716 log->Printf("lldb's in-memory shared cache has a UUID of %s", uuid.GetAsString().c_str()); 5717 #endif 5718 return uuid; 5719 } 5720 5721 uint32_t ObjectFileMachO::GetMinimumOSVersion(uint32_t *versions, 5722 uint32_t num_versions) { 5723 if (m_min_os_versions.empty()) { 5724 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5725 bool success = false; 5726 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5727 const lldb::offset_t load_cmd_offset = offset; 5728 5729 version_min_command lc; 5730 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5731 break; 5732 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5733 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5734 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5735 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5736 if (m_data.GetU32(&offset, &lc.version, 5737 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5738 const uint32_t xxxx = lc.version >> 16; 5739 const uint32_t yy = (lc.version >> 8) & 0xffu; 5740 const uint32_t zz = lc.version & 0xffu; 5741 if (xxxx) { 5742 m_min_os_versions.push_back(xxxx); 5743 m_min_os_versions.push_back(yy); 5744 m_min_os_versions.push_back(zz); 5745 } 5746 success = true; 5747 } 5748 } 5749 offset = load_cmd_offset + lc.cmdsize; 5750 } 5751 5752 if (success == false) { 5753 // Push an invalid value so we don't keep trying to 5754 m_min_os_versions.push_back(UINT32_MAX); 5755 } 5756 } 5757 5758 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) { 5759 if (versions != NULL && num_versions > 0) { 5760 for (size_t i = 0; i < num_versions; ++i) { 5761 if (i < m_min_os_versions.size()) 5762 versions[i] = m_min_os_versions[i]; 5763 else 5764 versions[i] = 0; 5765 } 5766 } 5767 return m_min_os_versions.size(); 5768 } 5769 // Call the superclasses version that will empty out the data 5770 return ObjectFile::GetMinimumOSVersion(versions, num_versions); 5771 } 5772 5773 uint32_t ObjectFileMachO::GetSDKVersion(uint32_t *versions, 5774 uint32_t num_versions) { 5775 if (m_sdk_versions.empty()) { 5776 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5777 bool success = false; 5778 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5779 const lldb::offset_t load_cmd_offset = offset; 5780 5781 version_min_command lc; 5782 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5783 break; 5784 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5785 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5786 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5787 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5788 if (m_data.GetU32(&offset, &lc.version, 5789 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5790 const uint32_t xxxx = lc.sdk >> 16; 5791 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5792 const uint32_t zz = lc.sdk & 0xffu; 5793 if (xxxx) { 5794 m_sdk_versions.push_back(xxxx); 5795 m_sdk_versions.push_back(yy); 5796 m_sdk_versions.push_back(zz); 5797 } 5798 success = true; 5799 } 5800 } 5801 offset = load_cmd_offset + lc.cmdsize; 5802 } 5803 5804 if (success == false) { 5805 // Push an invalid value so we don't keep trying to 5806 m_sdk_versions.push_back(UINT32_MAX); 5807 } 5808 } 5809 5810 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) { 5811 if (versions != NULL && num_versions > 0) { 5812 for (size_t i = 0; i < num_versions; ++i) { 5813 if (i < m_sdk_versions.size()) 5814 versions[i] = m_sdk_versions[i]; 5815 else 5816 versions[i] = 0; 5817 } 5818 } 5819 return m_sdk_versions.size(); 5820 } 5821 // Call the superclasses version that will empty out the data 5822 return ObjectFile::GetSDKVersion(versions, num_versions); 5823 } 5824 5825 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5826 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5827 } 5828 5829 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5830 return m_allow_assembly_emulation_unwind_plans; 5831 } 5832 5833 //------------------------------------------------------------------ 5834 // PluginInterface protocol 5835 //------------------------------------------------------------------ 5836 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5837 return GetPluginNameStatic(); 5838 } 5839 5840 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5841 5842 Section *ObjectFileMachO::GetMachHeaderSection() { 5843 // Find the first address of the mach header which is the first non-zero 5844 // file sized section whose file offset is zero. This is the base file address 5845 // of the mach-o file which can be subtracted from the vmaddr of the other 5846 // segments found in memory and added to the load address 5847 ModuleSP module_sp = GetModule(); 5848 if (module_sp) { 5849 SectionList *section_list = GetSectionList(); 5850 if (section_list) { 5851 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5852 const size_t num_sections = section_list->GetSize(); 5853 5854 for (size_t sect_idx = 0; sect_idx < num_sections && 5855 mach_base_file_addr == LLDB_INVALID_ADDRESS; 5856 ++sect_idx) { 5857 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5858 if (section && section->GetFileSize() > 0 && 5859 section->GetFileOffset() == 0 && 5860 section->IsThreadSpecific() == false && 5861 module_sp.get() == section->GetModule().get()) { 5862 return section; 5863 } 5864 } 5865 } 5866 } 5867 return nullptr; 5868 } 5869 5870 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 5871 lldb::addr_t mach_header_load_address, const Section *mach_header_section, 5872 const Section *section) { 5873 ModuleSP module_sp = GetModule(); 5874 if (module_sp && mach_header_section && section && 5875 mach_header_load_address != LLDB_INVALID_ADDRESS) { 5876 lldb::addr_t mach_header_file_addr = mach_header_section->GetFileAddress(); 5877 if (mach_header_file_addr != LLDB_INVALID_ADDRESS) { 5878 if (section && section->GetFileSize() > 0 && 5879 section->IsThreadSpecific() == false && 5880 module_sp.get() == section->GetModule().get()) { 5881 // Ignore __LINKEDIT and __DWARF segments 5882 if (section->GetName() == GetSegmentNameLINKEDIT()) { 5883 // Only map __LINKEDIT if we have an in memory image and this isn't 5884 // a kernel binary like a kext or mach_kernel. 5885 const bool is_memory_image = (bool)m_process_wp.lock(); 5886 const Strata strata = GetStrata(); 5887 if (is_memory_image == false || strata == eStrataKernel) 5888 return LLDB_INVALID_ADDRESS; 5889 } 5890 return section->GetFileAddress() - mach_header_file_addr + 5891 mach_header_load_address; 5892 } 5893 } 5894 } 5895 return LLDB_INVALID_ADDRESS; 5896 } 5897 5898 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 5899 bool value_is_offset) { 5900 ModuleSP module_sp = GetModule(); 5901 if (module_sp) { 5902 size_t num_loaded_sections = 0; 5903 SectionList *section_list = GetSectionList(); 5904 if (section_list) { 5905 const size_t num_sections = section_list->GetSize(); 5906 5907 if (value_is_offset) { 5908 // "value" is an offset to apply to each top level segment 5909 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5910 // Iterate through the object file sections to find all 5911 // of the sections that size on disk (to avoid __PAGEZERO) 5912 // and load them 5913 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5914 if (section_sp && section_sp->GetFileSize() > 0 && 5915 section_sp->IsThreadSpecific() == false && 5916 module_sp.get() == section_sp->GetModule().get()) { 5917 // Ignore __LINKEDIT and __DWARF segments 5918 if (section_sp->GetName() == GetSegmentNameLINKEDIT()) { 5919 // Only map __LINKEDIT if we have an in memory image and this 5920 // isn't 5921 // a kernel binary like a kext or mach_kernel. 5922 const bool is_memory_image = (bool)m_process_wp.lock(); 5923 const Strata strata = GetStrata(); 5924 if (is_memory_image == false || strata == eStrataKernel) 5925 continue; 5926 } 5927 if (target.GetSectionLoadList().SetSectionLoadAddress( 5928 section_sp, section_sp->GetFileAddress() + value)) 5929 ++num_loaded_sections; 5930 } 5931 } 5932 } else { 5933 // "value" is the new base address of the mach_header, adjust each 5934 // section accordingly 5935 5936 Section *mach_header_section = GetMachHeaderSection(); 5937 if (mach_header_section) { 5938 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5939 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5940 5941 lldb::addr_t section_load_addr = 5942 CalculateSectionLoadAddressForMemoryImage( 5943 value, mach_header_section, section_sp.get()); 5944 if (section_load_addr != LLDB_INVALID_ADDRESS) { 5945 if (target.GetSectionLoadList().SetSectionLoadAddress( 5946 section_sp, section_load_addr)) 5947 ++num_loaded_sections; 5948 } 5949 } 5950 } 5951 } 5952 } 5953 return num_loaded_sections > 0; 5954 } 5955 return false; 5956 } 5957 5958 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 5959 const FileSpec &outfile, Status &error) { 5960 if (process_sp) { 5961 Target &target = process_sp->GetTarget(); 5962 const ArchSpec target_arch = target.GetArchitecture(); 5963 const llvm::Triple &target_triple = target_arch.GetTriple(); 5964 if (target_triple.getVendor() == llvm::Triple::Apple && 5965 (target_triple.getOS() == llvm::Triple::MacOSX || 5966 target_triple.getOS() == llvm::Triple::IOS || 5967 target_triple.getOS() == llvm::Triple::WatchOS || 5968 target_triple.getOS() == llvm::Triple::TvOS)) { 5969 bool make_core = false; 5970 switch (target_arch.GetMachine()) { 5971 case llvm::Triple::aarch64: 5972 case llvm::Triple::arm: 5973 case llvm::Triple::thumb: 5974 case llvm::Triple::x86: 5975 case llvm::Triple::x86_64: 5976 make_core = true; 5977 break; 5978 default: 5979 error.SetErrorStringWithFormat("unsupported core architecture: %s", 5980 target_triple.str().c_str()); 5981 break; 5982 } 5983 5984 if (make_core) { 5985 std::vector<segment_command_64> segment_load_commands; 5986 // uint32_t range_info_idx = 0; 5987 MemoryRegionInfo range_info; 5988 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 5989 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 5990 const ByteOrder byte_order = target_arch.GetByteOrder(); 5991 if (range_error.Success()) { 5992 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 5993 const addr_t addr = range_info.GetRange().GetRangeBase(); 5994 const addr_t size = range_info.GetRange().GetByteSize(); 5995 5996 if (size == 0) 5997 break; 5998 5999 // Calculate correct protections 6000 uint32_t prot = 0; 6001 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6002 prot |= VM_PROT_READ; 6003 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6004 prot |= VM_PROT_WRITE; 6005 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6006 prot |= VM_PROT_EXECUTE; 6007 6008 // printf ("[%3u] [0x%16.16" PRIx64 " - 6009 // 0x%16.16" PRIx64 ") %c%c%c\n", 6010 // range_info_idx, 6011 // addr, 6012 // size, 6013 // (prot & VM_PROT_READ ) ? 'r' : 6014 // '-', 6015 // (prot & VM_PROT_WRITE ) ? 'w' : 6016 // '-', 6017 // (prot & VM_PROT_EXECUTE) ? 'x' : 6018 // '-'); 6019 6020 if (prot != 0) { 6021 uint32_t cmd_type = LC_SEGMENT_64; 6022 uint32_t segment_size = sizeof(segment_command_64); 6023 if (addr_byte_size == 4) { 6024 cmd_type = LC_SEGMENT; 6025 segment_size = sizeof(segment_command); 6026 } 6027 segment_command_64 segment = { 6028 cmd_type, // uint32_t cmd; 6029 segment_size, // uint32_t cmdsize; 6030 {0}, // char segname[16]; 6031 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6032 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6033 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6034 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6035 prot, // uint32_t maxprot; 6036 prot, // uint32_t initprot; 6037 0, // uint32_t nsects; 6038 0}; // uint32_t flags; 6039 segment_load_commands.push_back(segment); 6040 } else { 6041 // No protections and a size of 1 used to be returned from old 6042 // debugservers when we asked about a region that was past the 6043 // last memory region and it indicates the end... 6044 if (size == 1) 6045 break; 6046 } 6047 6048 range_error = process_sp->GetMemoryRegionInfo( 6049 range_info.GetRange().GetRangeEnd(), range_info); 6050 if (range_error.Fail()) 6051 break; 6052 } 6053 6054 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6055 6056 mach_header_64 mach_header; 6057 if (addr_byte_size == 8) { 6058 mach_header.magic = MH_MAGIC_64; 6059 } else { 6060 mach_header.magic = MH_MAGIC; 6061 } 6062 mach_header.cputype = target_arch.GetMachOCPUType(); 6063 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6064 mach_header.filetype = MH_CORE; 6065 mach_header.ncmds = segment_load_commands.size(); 6066 mach_header.flags = 0; 6067 mach_header.reserved = 0; 6068 ThreadList &thread_list = process_sp->GetThreadList(); 6069 const uint32_t num_threads = thread_list.GetSize(); 6070 6071 // Make an array of LC_THREAD data items. Each one contains 6072 // the contents of the LC_THREAD load command. The data doesn't 6073 // contain the load command + load command size, we will 6074 // add the load command and load command size as we emit the data. 6075 std::vector<StreamString> LC_THREAD_datas(num_threads); 6076 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6077 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6078 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6079 LC_THREAD_data.SetByteOrder(byte_order); 6080 } 6081 for (uint32_t thread_idx = 0; thread_idx < num_threads; 6082 ++thread_idx) { 6083 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6084 if (thread_sp) { 6085 switch (mach_header.cputype) { 6086 case llvm::MachO::CPU_TYPE_ARM64: 6087 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6088 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6089 break; 6090 6091 case llvm::MachO::CPU_TYPE_ARM: 6092 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6093 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6094 break; 6095 6096 case llvm::MachO::CPU_TYPE_I386: 6097 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6098 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6099 break; 6100 6101 case llvm::MachO::CPU_TYPE_X86_64: 6102 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6103 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6104 break; 6105 } 6106 } 6107 } 6108 6109 // The size of the load command is the size of the segments... 6110 if (addr_byte_size == 8) { 6111 mach_header.sizeofcmds = segment_load_commands.size() * 6112 sizeof(struct segment_command_64); 6113 } else { 6114 mach_header.sizeofcmds = 6115 segment_load_commands.size() * sizeof(struct segment_command); 6116 } 6117 6118 // and the size of all LC_THREAD load command 6119 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6120 ++mach_header.ncmds; 6121 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6122 } 6123 6124 printf("mach_header: 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x " 6125 "0x%8.8x 0x%8.8x\n", 6126 mach_header.magic, mach_header.cputype, mach_header.cpusubtype, 6127 mach_header.filetype, mach_header.ncmds, 6128 mach_header.sizeofcmds, mach_header.flags, 6129 mach_header.reserved); 6130 6131 // Write the mach header 6132 buffer.PutHex32(mach_header.magic); 6133 buffer.PutHex32(mach_header.cputype); 6134 buffer.PutHex32(mach_header.cpusubtype); 6135 buffer.PutHex32(mach_header.filetype); 6136 buffer.PutHex32(mach_header.ncmds); 6137 buffer.PutHex32(mach_header.sizeofcmds); 6138 buffer.PutHex32(mach_header.flags); 6139 if (addr_byte_size == 8) { 6140 buffer.PutHex32(mach_header.reserved); 6141 } 6142 6143 // Skip the mach header and all load commands and align to the next 6144 // 0x1000 byte boundary 6145 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6146 if (file_offset & 0x00000fff) { 6147 file_offset += 0x00001000ull; 6148 file_offset &= (~0x00001000ull + 1); 6149 } 6150 6151 for (auto &segment : segment_load_commands) { 6152 segment.fileoff = file_offset; 6153 file_offset += segment.filesize; 6154 } 6155 6156 // Write out all of the LC_THREAD load commands 6157 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6158 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6159 buffer.PutHex32(LC_THREAD); 6160 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6161 buffer.Write(LC_THREAD_data.GetString().data(), 6162 LC_THREAD_data_size); 6163 } 6164 6165 // Write out all of the segment load commands 6166 for (const auto &segment : segment_load_commands) { 6167 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6168 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6169 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6170 segment.cmd, segment.cmdsize, segment.vmaddr, 6171 segment.vmaddr + segment.vmsize, segment.fileoff, 6172 segment.filesize, segment.maxprot, segment.initprot, 6173 segment.nsects, segment.flags); 6174 6175 buffer.PutHex32(segment.cmd); 6176 buffer.PutHex32(segment.cmdsize); 6177 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6178 if (addr_byte_size == 8) { 6179 buffer.PutHex64(segment.vmaddr); 6180 buffer.PutHex64(segment.vmsize); 6181 buffer.PutHex64(segment.fileoff); 6182 buffer.PutHex64(segment.filesize); 6183 } else { 6184 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6185 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6186 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6187 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6188 } 6189 buffer.PutHex32(segment.maxprot); 6190 buffer.PutHex32(segment.initprot); 6191 buffer.PutHex32(segment.nsects); 6192 buffer.PutHex32(segment.flags); 6193 } 6194 6195 File core_file; 6196 std::string core_file_path(outfile.GetPath()); 6197 error = core_file.Open(core_file_path.c_str(), 6198 File::eOpenOptionWrite | 6199 File::eOpenOptionTruncate | 6200 File::eOpenOptionCanCreate); 6201 if (error.Success()) { 6202 // Read 1 page at a time 6203 uint8_t bytes[0x1000]; 6204 // Write the mach header and load commands out to the core file 6205 size_t bytes_written = buffer.GetString().size(); 6206 error = core_file.Write(buffer.GetString().data(), bytes_written); 6207 if (error.Success()) { 6208 // Now write the file data for all memory segments in the process 6209 for (const auto &segment : segment_load_commands) { 6210 if (core_file.SeekFromStart(segment.fileoff) == -1) { 6211 error.SetErrorStringWithFormat( 6212 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6213 segment.fileoff, core_file_path.c_str()); 6214 break; 6215 } 6216 6217 printf("Saving %" PRId64 6218 " bytes of data for memory region at 0x%" PRIx64 "\n", 6219 segment.vmsize, segment.vmaddr); 6220 addr_t bytes_left = segment.vmsize; 6221 addr_t addr = segment.vmaddr; 6222 Status memory_read_error; 6223 while (bytes_left > 0 && error.Success()) { 6224 const size_t bytes_to_read = 6225 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6226 const size_t bytes_read = process_sp->ReadMemory( 6227 addr, bytes, bytes_to_read, memory_read_error); 6228 if (bytes_read == bytes_to_read) { 6229 size_t bytes_written = bytes_read; 6230 error = core_file.Write(bytes, bytes_written); 6231 bytes_left -= bytes_read; 6232 addr += bytes_read; 6233 } else { 6234 // Some pages within regions are not readable, those 6235 // should be zero filled 6236 memset(bytes, 0, bytes_to_read); 6237 size_t bytes_written = bytes_to_read; 6238 error = core_file.Write(bytes, bytes_written); 6239 bytes_left -= bytes_to_read; 6240 addr += bytes_to_read; 6241 } 6242 } 6243 } 6244 } 6245 } 6246 } else { 6247 error.SetErrorString( 6248 "process doesn't support getting memory region info"); 6249 } 6250 } 6251 return true; // This is the right plug to handle saving core files for 6252 // this process 6253 } 6254 } 6255 return false; 6256 } 6257