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/Section.h" 27 #include "lldb/Core/StreamFile.h" 28 #include "lldb/Host/Host.h" 29 #include "lldb/Symbol/DWARFCallFrameInfo.h" 30 #include "lldb/Symbol/ObjectFile.h" 31 #include "lldb/Target/DynamicLoader.h" 32 #include "lldb/Target/MemoryRegionInfo.h" 33 #include "lldb/Target/Platform.h" 34 #include "lldb/Target/Process.h" 35 #include "lldb/Target/SectionLoadList.h" 36 #include "lldb/Target/Target.h" 37 #include "lldb/Target/Thread.h" 38 #include "lldb/Target/ThreadList.h" 39 #include "lldb/Utility/ArchSpec.h" 40 #include "lldb/Utility/DataBuffer.h" 41 #include "lldb/Utility/FileSpec.h" 42 #include "lldb/Utility/Log.h" 43 #include "lldb/Utility/RegisterValue.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/Host/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 flavors... 165 break; 166 167 default: 168 done = true; 169 break; 170 } 171 } 172 } 173 } 174 175 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 176 const char *alt_name, size_t reg_byte_size, 177 Stream &data) { 178 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 179 if (reg_info == NULL) 180 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 181 if (reg_info) { 182 lldb_private::RegisterValue reg_value; 183 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 184 if (reg_info->byte_size >= reg_byte_size) 185 data.Write(reg_value.GetBytes(), reg_byte_size); 186 else { 187 data.Write(reg_value.GetBytes(), reg_info->byte_size); 188 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 189 ++i) 190 data.PutChar(0); 191 } 192 return reg_byte_size; 193 } 194 } 195 // Just write zeros if all else fails 196 for (size_t i = 0; i < reg_byte_size; ++i) 197 data.PutChar(0); 198 return reg_byte_size; 199 } 200 201 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 202 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 203 if (reg_ctx_sp) { 204 RegisterContext *reg_ctx = reg_ctx_sp.get(); 205 206 data.PutHex32(GPRRegSet); // Flavor 207 data.PutHex32(GPRWordCount); 208 WriteRegister(reg_ctx, "rax", NULL, 8, data); 209 WriteRegister(reg_ctx, "rbx", NULL, 8, data); 210 WriteRegister(reg_ctx, "rcx", NULL, 8, data); 211 WriteRegister(reg_ctx, "rdx", NULL, 8, data); 212 WriteRegister(reg_ctx, "rdi", NULL, 8, data); 213 WriteRegister(reg_ctx, "rsi", NULL, 8, data); 214 WriteRegister(reg_ctx, "rbp", NULL, 8, data); 215 WriteRegister(reg_ctx, "rsp", NULL, 8, data); 216 WriteRegister(reg_ctx, "r8", NULL, 8, data); 217 WriteRegister(reg_ctx, "r9", NULL, 8, data); 218 WriteRegister(reg_ctx, "r10", NULL, 8, data); 219 WriteRegister(reg_ctx, "r11", NULL, 8, data); 220 WriteRegister(reg_ctx, "r12", NULL, 8, data); 221 WriteRegister(reg_ctx, "r13", NULL, 8, data); 222 WriteRegister(reg_ctx, "r14", NULL, 8, data); 223 WriteRegister(reg_ctx, "r15", NULL, 8, data); 224 WriteRegister(reg_ctx, "rip", NULL, 8, data); 225 WriteRegister(reg_ctx, "rflags", NULL, 8, data); 226 WriteRegister(reg_ctx, "cs", NULL, 8, data); 227 WriteRegister(reg_ctx, "fs", NULL, 8, data); 228 WriteRegister(reg_ctx, "gs", NULL, 8, data); 229 230 // // Write out the FPU registers 231 // const size_t fpu_byte_size = sizeof(FPU); 232 // size_t bytes_written = 0; 233 // data.PutHex32 (FPURegSet); 234 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 235 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 236 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 237 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 238 // data); // uint16_t fcw; // "fctrl" 239 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 240 // data); // uint16_t fsw; // "fstat" 241 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 242 // data); // uint8_t ftw; // "ftag" 243 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 244 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 245 // data); // uint16_t fop; // "fop" 246 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 247 // data); // uint32_t ip; // "fioff" 248 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 249 // data); // uint16_t cs; // "fiseg" 250 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 251 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 252 // data); // uint32_t dp; // "fooff" 253 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 254 // data); // uint16_t ds; // "foseg" 255 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 256 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 257 // data); // uint32_t mxcsr; 258 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 259 // 4, data);// uint32_t mxcsrmask; 260 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 261 // sizeof(MMSReg), data); 262 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 263 // sizeof(MMSReg), data); 264 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 265 // sizeof(MMSReg), data); 266 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 267 // sizeof(MMSReg), data); 268 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 269 // sizeof(MMSReg), data); 270 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 271 // sizeof(MMSReg), data); 272 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 273 // sizeof(MMSReg), data); 274 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 275 // sizeof(MMSReg), data); 276 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 277 // sizeof(XMMReg), data); 278 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 279 // sizeof(XMMReg), data); 280 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 281 // sizeof(XMMReg), data); 282 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 283 // sizeof(XMMReg), data); 284 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 285 // sizeof(XMMReg), data); 286 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 287 // sizeof(XMMReg), data); 288 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 289 // sizeof(XMMReg), data); 290 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 291 // sizeof(XMMReg), data); 292 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 293 // sizeof(XMMReg), data); 294 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 295 // sizeof(XMMReg), data); 296 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 297 // sizeof(XMMReg), data); 298 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 299 // sizeof(XMMReg), data); 300 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 301 // sizeof(XMMReg), data); 302 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 303 // sizeof(XMMReg), data); 304 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 305 // sizeof(XMMReg), data); 306 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 307 // sizeof(XMMReg), data); 308 // 309 // // Fill rest with zeros 310 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 311 // i) 312 // data.PutChar(0); 313 314 // Write out the EXC registers 315 data.PutHex32(EXCRegSet); 316 data.PutHex32(EXCWordCount); 317 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 318 WriteRegister(reg_ctx, "err", NULL, 4, data); 319 WriteRegister(reg_ctx, "faultvaddr", NULL, 8, data); 320 return true; 321 } 322 return false; 323 } 324 325 protected: 326 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 327 328 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 329 330 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 331 332 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 333 return 0; 334 } 335 336 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 337 return 0; 338 } 339 340 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 341 return 0; 342 } 343 }; 344 345 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 346 public: 347 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 348 const DataExtractor &data) 349 : RegisterContextDarwin_i386(thread, 0) { 350 SetRegisterDataFrom_LC_THREAD(data); 351 } 352 353 void InvalidateAllRegisters() override { 354 // Do nothing... registers are always valid... 355 } 356 357 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 358 lldb::offset_t offset = 0; 359 SetError(GPRRegSet, Read, -1); 360 SetError(FPURegSet, Read, -1); 361 SetError(EXCRegSet, Read, -1); 362 bool done = false; 363 364 while (!done) { 365 int flavor = data.GetU32(&offset); 366 if (flavor == 0) 367 done = true; 368 else { 369 uint32_t i; 370 uint32_t count = data.GetU32(&offset); 371 switch (flavor) { 372 case GPRRegSet: 373 for (i = 0; i < count; ++i) 374 (&gpr.eax)[i] = data.GetU32(&offset); 375 SetError(GPRRegSet, Read, 0); 376 done = true; 377 378 break; 379 case FPURegSet: 380 // TODO: fill in FPU regs.... 381 // SetError (FPURegSet, Read, -1); 382 done = true; 383 384 break; 385 case EXCRegSet: 386 exc.trapno = data.GetU32(&offset); 387 exc.err = data.GetU32(&offset); 388 exc.faultvaddr = data.GetU32(&offset); 389 SetError(EXCRegSet, Read, 0); 390 done = true; 391 break; 392 case 7: 393 case 8: 394 case 9: 395 // fancy flavors that encapsulate of the above flavors... 396 break; 397 398 default: 399 done = true; 400 break; 401 } 402 } 403 } 404 } 405 406 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 407 const char *alt_name, size_t reg_byte_size, 408 Stream &data) { 409 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 410 if (reg_info == NULL) 411 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 412 if (reg_info) { 413 lldb_private::RegisterValue reg_value; 414 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 415 if (reg_info->byte_size >= reg_byte_size) 416 data.Write(reg_value.GetBytes(), reg_byte_size); 417 else { 418 data.Write(reg_value.GetBytes(), reg_info->byte_size); 419 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 420 ++i) 421 data.PutChar(0); 422 } 423 return reg_byte_size; 424 } 425 } 426 // Just write zeros if all else fails 427 for (size_t i = 0; i < reg_byte_size; ++i) 428 data.PutChar(0); 429 return reg_byte_size; 430 } 431 432 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 433 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 434 if (reg_ctx_sp) { 435 RegisterContext *reg_ctx = reg_ctx_sp.get(); 436 437 data.PutHex32(GPRRegSet); // Flavor 438 data.PutHex32(GPRWordCount); 439 WriteRegister(reg_ctx, "eax", NULL, 4, data); 440 WriteRegister(reg_ctx, "ebx", NULL, 4, data); 441 WriteRegister(reg_ctx, "ecx", NULL, 4, data); 442 WriteRegister(reg_ctx, "edx", NULL, 4, data); 443 WriteRegister(reg_ctx, "edi", NULL, 4, data); 444 WriteRegister(reg_ctx, "esi", NULL, 4, data); 445 WriteRegister(reg_ctx, "ebp", NULL, 4, data); 446 WriteRegister(reg_ctx, "esp", NULL, 4, data); 447 WriteRegister(reg_ctx, "ss", NULL, 4, data); 448 WriteRegister(reg_ctx, "eflags", NULL, 4, data); 449 WriteRegister(reg_ctx, "eip", NULL, 4, data); 450 WriteRegister(reg_ctx, "cs", NULL, 4, data); 451 WriteRegister(reg_ctx, "ds", NULL, 4, data); 452 WriteRegister(reg_ctx, "es", NULL, 4, data); 453 WriteRegister(reg_ctx, "fs", NULL, 4, data); 454 WriteRegister(reg_ctx, "gs", NULL, 4, data); 455 456 // Write out the EXC registers 457 data.PutHex32(EXCRegSet); 458 data.PutHex32(EXCWordCount); 459 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 460 WriteRegister(reg_ctx, "err", NULL, 4, data); 461 WriteRegister(reg_ctx, "faultvaddr", NULL, 4, data); 462 return true; 463 } 464 return false; 465 } 466 467 protected: 468 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 469 470 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 471 472 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 473 474 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 475 return 0; 476 } 477 478 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 479 return 0; 480 } 481 482 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 483 return 0; 484 } 485 }; 486 487 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 488 public: 489 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 490 const DataExtractor &data) 491 : RegisterContextDarwin_arm(thread, 0) { 492 SetRegisterDataFrom_LC_THREAD(data); 493 } 494 495 void InvalidateAllRegisters() override { 496 // Do nothing... registers are always valid... 497 } 498 499 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 500 lldb::offset_t offset = 0; 501 SetError(GPRRegSet, Read, -1); 502 SetError(FPURegSet, Read, -1); 503 SetError(EXCRegSet, Read, -1); 504 bool done = false; 505 506 while (!done) { 507 int flavor = data.GetU32(&offset); 508 uint32_t count = data.GetU32(&offset); 509 lldb::offset_t next_thread_state = offset + (count * 4); 510 switch (flavor) { 511 case GPRAltRegSet: 512 case GPRRegSet: 513 for (uint32_t i = 0; i < count; ++i) { 514 gpr.r[i] = data.GetU32(&offset); 515 } 516 517 // Note that gpr.cpsr is also copied by the above loop; this loop 518 // technically extends one element past the end of the gpr.r[] array. 519 520 SetError(GPRRegSet, Read, 0); 521 offset = next_thread_state; 522 break; 523 524 case FPURegSet: { 525 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 526 const int fpu_reg_buf_size = sizeof(fpu.floats); 527 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 528 fpu_reg_buf) == fpu_reg_buf_size) { 529 offset += fpu_reg_buf_size; 530 fpu.fpscr = data.GetU32(&offset); 531 SetError(FPURegSet, Read, 0); 532 } else { 533 done = true; 534 } 535 } 536 offset = next_thread_state; 537 break; 538 539 case EXCRegSet: 540 if (count == 3) { 541 exc.exception = data.GetU32(&offset); 542 exc.fsr = data.GetU32(&offset); 543 exc.far = data.GetU32(&offset); 544 SetError(EXCRegSet, Read, 0); 545 } 546 done = true; 547 offset = next_thread_state; 548 break; 549 550 // Unknown register set flavor, stop trying to parse. 551 default: 552 done = true; 553 } 554 } 555 } 556 557 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 558 const char *alt_name, size_t reg_byte_size, 559 Stream &data) { 560 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 561 if (reg_info == NULL) 562 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 563 if (reg_info) { 564 lldb_private::RegisterValue reg_value; 565 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 566 if (reg_info->byte_size >= reg_byte_size) 567 data.Write(reg_value.GetBytes(), reg_byte_size); 568 else { 569 data.Write(reg_value.GetBytes(), reg_info->byte_size); 570 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 571 ++i) 572 data.PutChar(0); 573 } 574 return reg_byte_size; 575 } 576 } 577 // Just write zeros if all else fails 578 for (size_t i = 0; i < reg_byte_size; ++i) 579 data.PutChar(0); 580 return reg_byte_size; 581 } 582 583 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 584 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 585 if (reg_ctx_sp) { 586 RegisterContext *reg_ctx = reg_ctx_sp.get(); 587 588 data.PutHex32(GPRRegSet); // Flavor 589 data.PutHex32(GPRWordCount); 590 WriteRegister(reg_ctx, "r0", NULL, 4, data); 591 WriteRegister(reg_ctx, "r1", NULL, 4, data); 592 WriteRegister(reg_ctx, "r2", NULL, 4, data); 593 WriteRegister(reg_ctx, "r3", NULL, 4, data); 594 WriteRegister(reg_ctx, "r4", NULL, 4, data); 595 WriteRegister(reg_ctx, "r5", NULL, 4, data); 596 WriteRegister(reg_ctx, "r6", NULL, 4, data); 597 WriteRegister(reg_ctx, "r7", NULL, 4, data); 598 WriteRegister(reg_ctx, "r8", NULL, 4, data); 599 WriteRegister(reg_ctx, "r9", NULL, 4, data); 600 WriteRegister(reg_ctx, "r10", NULL, 4, data); 601 WriteRegister(reg_ctx, "r11", NULL, 4, data); 602 WriteRegister(reg_ctx, "r12", NULL, 4, data); 603 WriteRegister(reg_ctx, "sp", NULL, 4, data); 604 WriteRegister(reg_ctx, "lr", NULL, 4, data); 605 WriteRegister(reg_ctx, "pc", NULL, 4, data); 606 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 607 608 // Write out the EXC registers 609 // data.PutHex32 (EXCRegSet); 610 // data.PutHex32 (EXCWordCount); 611 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 612 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 613 // WriteRegister (reg_ctx, "far", NULL, 4, data); 614 return true; 615 } 616 return false; 617 } 618 619 protected: 620 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 621 622 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 623 624 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 625 626 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 627 628 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 629 return 0; 630 } 631 632 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 633 return 0; 634 } 635 636 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 637 return 0; 638 } 639 640 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 641 return -1; 642 } 643 }; 644 645 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 646 public: 647 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 648 const DataExtractor &data) 649 : RegisterContextDarwin_arm64(thread, 0) { 650 SetRegisterDataFrom_LC_THREAD(data); 651 } 652 653 void InvalidateAllRegisters() override { 654 // Do nothing... registers are always valid... 655 } 656 657 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 658 lldb::offset_t offset = 0; 659 SetError(GPRRegSet, Read, -1); 660 SetError(FPURegSet, Read, -1); 661 SetError(EXCRegSet, Read, -1); 662 bool done = false; 663 while (!done) { 664 int flavor = data.GetU32(&offset); 665 uint32_t count = data.GetU32(&offset); 666 lldb::offset_t next_thread_state = offset + (count * 4); 667 switch (flavor) { 668 case GPRRegSet: 669 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 670 // 32-bit register) 671 if (count >= (33 * 2) + 1) { 672 for (uint32_t i = 0; i < 29; ++i) 673 gpr.x[i] = data.GetU64(&offset); 674 gpr.fp = data.GetU64(&offset); 675 gpr.lr = data.GetU64(&offset); 676 gpr.sp = data.GetU64(&offset); 677 gpr.pc = data.GetU64(&offset); 678 gpr.cpsr = data.GetU32(&offset); 679 SetError(GPRRegSet, Read, 0); 680 } 681 offset = next_thread_state; 682 break; 683 case FPURegSet: { 684 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 685 const int fpu_reg_buf_size = sizeof(fpu); 686 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 687 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 688 fpu_reg_buf) == fpu_reg_buf_size) { 689 SetError(FPURegSet, Read, 0); 690 } else { 691 done = true; 692 } 693 } 694 offset = next_thread_state; 695 break; 696 case EXCRegSet: 697 if (count == 4) { 698 exc.far = data.GetU64(&offset); 699 exc.esr = data.GetU32(&offset); 700 exc.exception = data.GetU32(&offset); 701 SetError(EXCRegSet, Read, 0); 702 } 703 offset = next_thread_state; 704 break; 705 default: 706 done = true; 707 break; 708 } 709 } 710 } 711 712 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 713 const char *alt_name, size_t reg_byte_size, 714 Stream &data) { 715 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 716 if (reg_info == NULL) 717 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 718 if (reg_info) { 719 lldb_private::RegisterValue reg_value; 720 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 721 if (reg_info->byte_size >= reg_byte_size) 722 data.Write(reg_value.GetBytes(), reg_byte_size); 723 else { 724 data.Write(reg_value.GetBytes(), reg_info->byte_size); 725 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 726 ++i) 727 data.PutChar(0); 728 } 729 return reg_byte_size; 730 } 731 } 732 // Just write zeros if all else fails 733 for (size_t i = 0; i < reg_byte_size; ++i) 734 data.PutChar(0); 735 return reg_byte_size; 736 } 737 738 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 739 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 740 if (reg_ctx_sp) { 741 RegisterContext *reg_ctx = reg_ctx_sp.get(); 742 743 data.PutHex32(GPRRegSet); // Flavor 744 data.PutHex32(GPRWordCount); 745 WriteRegister(reg_ctx, "x0", NULL, 8, data); 746 WriteRegister(reg_ctx, "x1", NULL, 8, data); 747 WriteRegister(reg_ctx, "x2", NULL, 8, data); 748 WriteRegister(reg_ctx, "x3", NULL, 8, data); 749 WriteRegister(reg_ctx, "x4", NULL, 8, data); 750 WriteRegister(reg_ctx, "x5", NULL, 8, data); 751 WriteRegister(reg_ctx, "x6", NULL, 8, data); 752 WriteRegister(reg_ctx, "x7", NULL, 8, data); 753 WriteRegister(reg_ctx, "x8", NULL, 8, data); 754 WriteRegister(reg_ctx, "x9", NULL, 8, data); 755 WriteRegister(reg_ctx, "x10", NULL, 8, data); 756 WriteRegister(reg_ctx, "x11", NULL, 8, data); 757 WriteRegister(reg_ctx, "x12", NULL, 8, data); 758 WriteRegister(reg_ctx, "x13", NULL, 8, data); 759 WriteRegister(reg_ctx, "x14", NULL, 8, data); 760 WriteRegister(reg_ctx, "x15", NULL, 8, data); 761 WriteRegister(reg_ctx, "x16", NULL, 8, data); 762 WriteRegister(reg_ctx, "x17", NULL, 8, data); 763 WriteRegister(reg_ctx, "x18", NULL, 8, data); 764 WriteRegister(reg_ctx, "x19", NULL, 8, data); 765 WriteRegister(reg_ctx, "x20", NULL, 8, data); 766 WriteRegister(reg_ctx, "x21", NULL, 8, data); 767 WriteRegister(reg_ctx, "x22", NULL, 8, data); 768 WriteRegister(reg_ctx, "x23", NULL, 8, data); 769 WriteRegister(reg_ctx, "x24", NULL, 8, data); 770 WriteRegister(reg_ctx, "x25", NULL, 8, data); 771 WriteRegister(reg_ctx, "x26", NULL, 8, data); 772 WriteRegister(reg_ctx, "x27", NULL, 8, data); 773 WriteRegister(reg_ctx, "x28", NULL, 8, data); 774 WriteRegister(reg_ctx, "fp", NULL, 8, data); 775 WriteRegister(reg_ctx, "lr", NULL, 8, data); 776 WriteRegister(reg_ctx, "sp", NULL, 8, data); 777 WriteRegister(reg_ctx, "pc", NULL, 8, data); 778 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 779 780 // Write out the EXC registers 781 // data.PutHex32 (EXCRegSet); 782 // data.PutHex32 (EXCWordCount); 783 // WriteRegister (reg_ctx, "far", NULL, 8, data); 784 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 785 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 786 return true; 787 } 788 return false; 789 } 790 791 protected: 792 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 793 794 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 795 796 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 797 798 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 799 800 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 801 return 0; 802 } 803 804 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 805 return 0; 806 } 807 808 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 809 return 0; 810 } 811 812 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 813 return -1; 814 } 815 }; 816 817 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 818 switch (magic) { 819 case MH_MAGIC: 820 case MH_CIGAM: 821 return sizeof(struct mach_header); 822 823 case MH_MAGIC_64: 824 case MH_CIGAM_64: 825 return sizeof(struct mach_header_64); 826 break; 827 828 default: 829 break; 830 } 831 return 0; 832 } 833 834 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 835 836 void ObjectFileMachO::Initialize() { 837 PluginManager::RegisterPlugin( 838 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 839 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 840 } 841 842 void ObjectFileMachO::Terminate() { 843 PluginManager::UnregisterPlugin(CreateInstance); 844 } 845 846 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 847 static ConstString g_name("mach-o"); 848 return g_name; 849 } 850 851 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 852 return "Mach-o object file reader (32 and 64 bit)"; 853 } 854 855 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 856 DataBufferSP &data_sp, 857 lldb::offset_t data_offset, 858 const FileSpec *file, 859 lldb::offset_t file_offset, 860 lldb::offset_t length) { 861 if (!data_sp) { 862 data_sp = MapFileData(*file, length, file_offset); 863 if (!data_sp) 864 return nullptr; 865 data_offset = 0; 866 } 867 868 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 869 return nullptr; 870 871 // Update the data to contain the entire file if it doesn't already 872 if (data_sp->GetByteSize() < length) { 873 data_sp = MapFileData(*file, length, file_offset); 874 if (!data_sp) 875 return nullptr; 876 data_offset = 0; 877 } 878 auto objfile_ap = llvm::make_unique<ObjectFileMachO>( 879 module_sp, data_sp, data_offset, file, file_offset, length); 880 if (!objfile_ap || !objfile_ap->ParseHeader()) 881 return nullptr; 882 883 return objfile_ap.release(); 884 } 885 886 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 887 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 888 const ProcessSP &process_sp, lldb::addr_t header_addr) { 889 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 890 std::unique_ptr<ObjectFile> objfile_ap( 891 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 892 if (objfile_ap.get() && objfile_ap->ParseHeader()) 893 return objfile_ap.release(); 894 } 895 return NULL; 896 } 897 898 size_t ObjectFileMachO::GetModuleSpecifications( 899 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 900 lldb::offset_t data_offset, lldb::offset_t file_offset, 901 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 902 const size_t initial_count = specs.GetSize(); 903 904 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 905 DataExtractor data; 906 data.SetData(data_sp); 907 llvm::MachO::mach_header header; 908 if (ParseHeader(data, &data_offset, header)) { 909 size_t header_and_load_cmds = 910 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 911 if (header_and_load_cmds >= data_sp->GetByteSize()) { 912 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 913 data.SetData(data_sp); 914 data_offset = MachHeaderSizeFromMagic(header.magic); 915 } 916 if (data_sp) { 917 ModuleSpec spec; 918 spec.GetFileSpec() = file; 919 spec.SetObjectOffset(file_offset); 920 spec.SetObjectSize(length); 921 922 if (GetArchitecture(header, data, data_offset, 923 spec.GetArchitecture())) { 924 if (spec.GetArchitecture().IsValid()) { 925 GetUUID(header, data, data_offset, spec.GetUUID()); 926 specs.Append(spec); 927 } 928 } 929 } 930 } 931 } 932 return specs.GetSize() - initial_count; 933 } 934 935 const ConstString &ObjectFileMachO::GetSegmentNameTEXT() { 936 static ConstString g_segment_name_TEXT("__TEXT"); 937 return g_segment_name_TEXT; 938 } 939 940 const ConstString &ObjectFileMachO::GetSegmentNameDATA() { 941 static ConstString g_segment_name_DATA("__DATA"); 942 return g_segment_name_DATA; 943 } 944 945 const ConstString &ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 946 static ConstString g_segment_name("__DATA_DIRTY"); 947 return g_segment_name; 948 } 949 950 const ConstString &ObjectFileMachO::GetSegmentNameDATA_CONST() { 951 static ConstString g_segment_name("__DATA_CONST"); 952 return g_segment_name; 953 } 954 955 const ConstString &ObjectFileMachO::GetSegmentNameOBJC() { 956 static ConstString g_segment_name_OBJC("__OBJC"); 957 return g_segment_name_OBJC; 958 } 959 960 const ConstString &ObjectFileMachO::GetSegmentNameLINKEDIT() { 961 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 962 return g_section_name_LINKEDIT; 963 } 964 965 const ConstString &ObjectFileMachO::GetSectionNameEHFrame() { 966 static ConstString g_section_name_eh_frame("__eh_frame"); 967 return g_section_name_eh_frame; 968 } 969 970 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 971 lldb::addr_t data_offset, 972 lldb::addr_t data_length) { 973 DataExtractor data; 974 data.SetData(data_sp, data_offset, data_length); 975 lldb::offset_t offset = 0; 976 uint32_t magic = data.GetU32(&offset); 977 return MachHeaderSizeFromMagic(magic) != 0; 978 } 979 980 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 981 DataBufferSP &data_sp, 982 lldb::offset_t data_offset, 983 const FileSpec *file, 984 lldb::offset_t file_offset, 985 lldb::offset_t length) 986 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 987 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 988 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 989 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 990 ::memset(&m_header, 0, sizeof(m_header)); 991 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 992 } 993 994 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 995 lldb::DataBufferSP &header_data_sp, 996 const lldb::ProcessSP &process_sp, 997 lldb::addr_t header_addr) 998 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 999 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 1000 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 1001 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 1002 ::memset(&m_header, 0, sizeof(m_header)); 1003 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1004 } 1005 1006 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 1007 lldb::offset_t *data_offset_ptr, 1008 llvm::MachO::mach_header &header) { 1009 data.SetByteOrder(endian::InlHostByteOrder()); 1010 // Leave magic in the original byte order 1011 header.magic = data.GetU32(data_offset_ptr); 1012 bool can_parse = false; 1013 bool is_64_bit = false; 1014 switch (header.magic) { 1015 case MH_MAGIC: 1016 data.SetByteOrder(endian::InlHostByteOrder()); 1017 data.SetAddressByteSize(4); 1018 can_parse = true; 1019 break; 1020 1021 case MH_MAGIC_64: 1022 data.SetByteOrder(endian::InlHostByteOrder()); 1023 data.SetAddressByteSize(8); 1024 can_parse = true; 1025 is_64_bit = true; 1026 break; 1027 1028 case MH_CIGAM: 1029 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1030 ? eByteOrderLittle 1031 : eByteOrderBig); 1032 data.SetAddressByteSize(4); 1033 can_parse = true; 1034 break; 1035 1036 case MH_CIGAM_64: 1037 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1038 ? eByteOrderLittle 1039 : eByteOrderBig); 1040 data.SetAddressByteSize(8); 1041 is_64_bit = true; 1042 can_parse = true; 1043 break; 1044 1045 default: 1046 break; 1047 } 1048 1049 if (can_parse) { 1050 data.GetU32(data_offset_ptr, &header.cputype, 6); 1051 if (is_64_bit) 1052 *data_offset_ptr += 4; 1053 return true; 1054 } else { 1055 memset(&header, 0, sizeof(header)); 1056 } 1057 return false; 1058 } 1059 1060 bool ObjectFileMachO::ParseHeader() { 1061 ModuleSP module_sp(GetModule()); 1062 if (module_sp) { 1063 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1064 bool can_parse = false; 1065 lldb::offset_t offset = 0; 1066 m_data.SetByteOrder(endian::InlHostByteOrder()); 1067 // Leave magic in the original byte order 1068 m_header.magic = m_data.GetU32(&offset); 1069 switch (m_header.magic) { 1070 case MH_MAGIC: 1071 m_data.SetByteOrder(endian::InlHostByteOrder()); 1072 m_data.SetAddressByteSize(4); 1073 can_parse = true; 1074 break; 1075 1076 case MH_MAGIC_64: 1077 m_data.SetByteOrder(endian::InlHostByteOrder()); 1078 m_data.SetAddressByteSize(8); 1079 can_parse = true; 1080 break; 1081 1082 case MH_CIGAM: 1083 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1084 ? eByteOrderLittle 1085 : eByteOrderBig); 1086 m_data.SetAddressByteSize(4); 1087 can_parse = true; 1088 break; 1089 1090 case MH_CIGAM_64: 1091 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1092 ? eByteOrderLittle 1093 : eByteOrderBig); 1094 m_data.SetAddressByteSize(8); 1095 can_parse = true; 1096 break; 1097 1098 default: 1099 break; 1100 } 1101 1102 if (can_parse) { 1103 m_data.GetU32(&offset, &m_header.cputype, 6); 1104 1105 ArchSpec mach_arch; 1106 1107 if (GetArchitecture(mach_arch)) { 1108 // Check if the module has a required architecture 1109 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1110 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1111 return false; 1112 1113 if (SetModulesArchitecture(mach_arch)) { 1114 const size_t header_and_lc_size = 1115 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1116 if (m_data.GetByteSize() < header_and_lc_size) { 1117 DataBufferSP data_sp; 1118 ProcessSP process_sp(m_process_wp.lock()); 1119 if (process_sp) { 1120 data_sp = 1121 ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1122 } else { 1123 // Read in all only the load command data from the file on disk 1124 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset); 1125 if (data_sp->GetByteSize() != header_and_lc_size) 1126 return false; 1127 } 1128 if (data_sp) 1129 m_data.SetData(data_sp); 1130 } 1131 } 1132 return true; 1133 } 1134 } else { 1135 memset(&m_header, 0, sizeof(struct mach_header)); 1136 } 1137 } 1138 return false; 1139 } 1140 1141 ByteOrder ObjectFileMachO::GetByteOrder() const { 1142 return m_data.GetByteOrder(); 1143 } 1144 1145 bool ObjectFileMachO::IsExecutable() const { 1146 return m_header.filetype == MH_EXECUTE; 1147 } 1148 1149 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1150 return m_data.GetAddressByteSize(); 1151 } 1152 1153 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1154 Symtab *symtab = GetSymtab(); 1155 if (symtab) { 1156 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1157 if (symbol) { 1158 if (symbol->ValueIsAddress()) { 1159 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1160 if (section_sp) { 1161 const lldb::SectionType section_type = section_sp->GetType(); 1162 switch (section_type) { 1163 case eSectionTypeInvalid: 1164 return AddressClass::eUnknown; 1165 1166 case eSectionTypeCode: 1167 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1168 // For ARM we have a bit in the n_desc field of the symbol that 1169 // tells us ARM/Thumb which is bit 0x0008. 1170 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1171 return AddressClass::eCodeAlternateISA; 1172 } 1173 return AddressClass::eCode; 1174 1175 case eSectionTypeContainer: 1176 return AddressClass::eUnknown; 1177 1178 case eSectionTypeData: 1179 case eSectionTypeDataCString: 1180 case eSectionTypeDataCStringPointers: 1181 case eSectionTypeDataSymbolAddress: 1182 case eSectionTypeData4: 1183 case eSectionTypeData8: 1184 case eSectionTypeData16: 1185 case eSectionTypeDataPointers: 1186 case eSectionTypeZeroFill: 1187 case eSectionTypeDataObjCMessageRefs: 1188 case eSectionTypeDataObjCCFStrings: 1189 case eSectionTypeGoSymtab: 1190 return AddressClass::eData; 1191 1192 case eSectionTypeDebug: 1193 case eSectionTypeDWARFDebugAbbrev: 1194 case eSectionTypeDWARFDebugAddr: 1195 case eSectionTypeDWARFDebugAranges: 1196 case eSectionTypeDWARFDebugCuIndex: 1197 case eSectionTypeDWARFDebugFrame: 1198 case eSectionTypeDWARFDebugInfo: 1199 case eSectionTypeDWARFDebugLine: 1200 case eSectionTypeDWARFDebugLineStr: 1201 case eSectionTypeDWARFDebugLoc: 1202 case eSectionTypeDWARFDebugMacInfo: 1203 case eSectionTypeDWARFDebugMacro: 1204 case eSectionTypeDWARFDebugNames: 1205 case eSectionTypeDWARFDebugPubNames: 1206 case eSectionTypeDWARFDebugPubTypes: 1207 case eSectionTypeDWARFDebugRanges: 1208 case eSectionTypeDWARFDebugStr: 1209 case eSectionTypeDWARFDebugStrOffsets: 1210 case eSectionTypeDWARFDebugTypes: 1211 case eSectionTypeDWARFAppleNames: 1212 case eSectionTypeDWARFAppleTypes: 1213 case eSectionTypeDWARFAppleNamespaces: 1214 case eSectionTypeDWARFAppleObjC: 1215 case eSectionTypeDWARFGNUDebugAltLink: 1216 return AddressClass::eDebug; 1217 1218 case eSectionTypeEHFrame: 1219 case eSectionTypeARMexidx: 1220 case eSectionTypeARMextab: 1221 case eSectionTypeCompactUnwind: 1222 return AddressClass::eRuntime; 1223 1224 case eSectionTypeAbsoluteAddress: 1225 case eSectionTypeELFSymbolTable: 1226 case eSectionTypeELFDynamicSymbols: 1227 case eSectionTypeELFRelocationEntries: 1228 case eSectionTypeELFDynamicLinkInfo: 1229 case eSectionTypeOther: 1230 return AddressClass::eUnknown; 1231 } 1232 } 1233 } 1234 1235 const SymbolType symbol_type = symbol->GetType(); 1236 switch (symbol_type) { 1237 case eSymbolTypeAny: 1238 return AddressClass::eUnknown; 1239 case eSymbolTypeAbsolute: 1240 return AddressClass::eUnknown; 1241 1242 case eSymbolTypeCode: 1243 case eSymbolTypeTrampoline: 1244 case eSymbolTypeResolver: 1245 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1246 // For ARM we have a bit in the n_desc field of the symbol that tells 1247 // us ARM/Thumb which is bit 0x0008. 1248 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1249 return AddressClass::eCodeAlternateISA; 1250 } 1251 return AddressClass::eCode; 1252 1253 case eSymbolTypeData: 1254 return AddressClass::eData; 1255 case eSymbolTypeRuntime: 1256 return AddressClass::eRuntime; 1257 case eSymbolTypeException: 1258 return AddressClass::eRuntime; 1259 case eSymbolTypeSourceFile: 1260 return AddressClass::eDebug; 1261 case eSymbolTypeHeaderFile: 1262 return AddressClass::eDebug; 1263 case eSymbolTypeObjectFile: 1264 return AddressClass::eDebug; 1265 case eSymbolTypeCommonBlock: 1266 return AddressClass::eDebug; 1267 case eSymbolTypeBlock: 1268 return AddressClass::eDebug; 1269 case eSymbolTypeLocal: 1270 return AddressClass::eData; 1271 case eSymbolTypeParam: 1272 return AddressClass::eData; 1273 case eSymbolTypeVariable: 1274 return AddressClass::eData; 1275 case eSymbolTypeVariableType: 1276 return AddressClass::eDebug; 1277 case eSymbolTypeLineEntry: 1278 return AddressClass::eDebug; 1279 case eSymbolTypeLineHeader: 1280 return AddressClass::eDebug; 1281 case eSymbolTypeScopeBegin: 1282 return AddressClass::eDebug; 1283 case eSymbolTypeScopeEnd: 1284 return AddressClass::eDebug; 1285 case eSymbolTypeAdditional: 1286 return AddressClass::eUnknown; 1287 case eSymbolTypeCompiler: 1288 return AddressClass::eDebug; 1289 case eSymbolTypeInstrumentation: 1290 return AddressClass::eDebug; 1291 case eSymbolTypeUndefined: 1292 return AddressClass::eUnknown; 1293 case eSymbolTypeObjCClass: 1294 return AddressClass::eRuntime; 1295 case eSymbolTypeObjCMetaClass: 1296 return AddressClass::eRuntime; 1297 case eSymbolTypeObjCIVar: 1298 return AddressClass::eRuntime; 1299 case eSymbolTypeReExported: 1300 return AddressClass::eRuntime; 1301 } 1302 } 1303 } 1304 return AddressClass::eUnknown; 1305 } 1306 1307 Symtab *ObjectFileMachO::GetSymtab() { 1308 ModuleSP module_sp(GetModule()); 1309 if (module_sp) { 1310 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1311 if (m_symtab_ap.get() == NULL) { 1312 m_symtab_ap.reset(new Symtab(this)); 1313 std::lock_guard<std::recursive_mutex> symtab_guard( 1314 m_symtab_ap->GetMutex()); 1315 ParseSymtab(); 1316 m_symtab_ap->Finalize(); 1317 } 1318 } 1319 return m_symtab_ap.get(); 1320 } 1321 1322 bool ObjectFileMachO::IsStripped() { 1323 if (m_dysymtab.cmd == 0) { 1324 ModuleSP module_sp(GetModule()); 1325 if (module_sp) { 1326 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1327 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1328 const lldb::offset_t load_cmd_offset = offset; 1329 1330 load_command lc; 1331 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1332 break; 1333 if (lc.cmd == LC_DYSYMTAB) { 1334 m_dysymtab.cmd = lc.cmd; 1335 m_dysymtab.cmdsize = lc.cmdsize; 1336 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1337 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1338 NULL) { 1339 // Clear m_dysymtab if we were unable to read all items from the 1340 // load command 1341 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1342 } 1343 } 1344 offset = load_cmd_offset + lc.cmdsize; 1345 } 1346 } 1347 } 1348 if (m_dysymtab.cmd) 1349 return m_dysymtab.nlocalsym <= 1; 1350 return false; 1351 } 1352 1353 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1354 EncryptedFileRanges result; 1355 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1356 1357 encryption_info_command encryption_cmd; 1358 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1359 const lldb::offset_t load_cmd_offset = offset; 1360 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1361 break; 1362 1363 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the 1364 // 3 fields we care about, so treat them the same. 1365 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1366 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1367 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1368 if (encryption_cmd.cryptid != 0) { 1369 EncryptedFileRanges::Entry entry; 1370 entry.SetRangeBase(encryption_cmd.cryptoff); 1371 entry.SetByteSize(encryption_cmd.cryptsize); 1372 result.Append(entry); 1373 } 1374 } 1375 } 1376 offset = load_cmd_offset + encryption_cmd.cmdsize; 1377 } 1378 1379 return result; 1380 } 1381 1382 void ObjectFileMachO::SanitizeSegmentCommand(segment_command_64 &seg_cmd, 1383 uint32_t cmd_idx) { 1384 if (m_length == 0 || seg_cmd.filesize == 0) 1385 return; 1386 1387 if (seg_cmd.fileoff > m_length) { 1388 // We have a load command that says it extends past the end of the file. 1389 // This is likely a corrupt file. We don't have any way to return an error 1390 // condition here (this method was likely invoked from something like 1391 // ObjectFile::GetSectionList()), so we just null out the section contents, 1392 // and dump a message to stdout. The most common case here is core file 1393 // debugging with a truncated file. 1394 const char *lc_segment_name = 1395 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1396 GetModule()->ReportWarning( 1397 "load command %u %s has a fileoff (0x%" PRIx64 1398 ") that extends beyond the end of the file (0x%" PRIx64 1399 "), ignoring this section", 1400 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1401 1402 seg_cmd.fileoff = 0; 1403 seg_cmd.filesize = 0; 1404 } 1405 1406 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1407 // We have a load command that says it extends past the end of the file. 1408 // This is likely a corrupt file. We don't have any way to return an error 1409 // condition here (this method was likely invoked from something like 1410 // ObjectFile::GetSectionList()), so we just null out the section contents, 1411 // and dump a message to stdout. The most common case here is core file 1412 // debugging with a truncated file. 1413 const char *lc_segment_name = 1414 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1415 GetModule()->ReportWarning( 1416 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1417 ") that extends beyond the end of the file (0x%" PRIx64 1418 "), the segment will be truncated to match", 1419 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1420 1421 // Truncate the length 1422 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1423 } 1424 } 1425 1426 static uint32_t GetSegmentPermissions(const segment_command_64 &seg_cmd) { 1427 uint32_t result = 0; 1428 if (seg_cmd.initprot & VM_PROT_READ) 1429 result |= ePermissionsReadable; 1430 if (seg_cmd.initprot & VM_PROT_WRITE) 1431 result |= ePermissionsWritable; 1432 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1433 result |= ePermissionsExecutable; 1434 return result; 1435 } 1436 1437 static lldb::SectionType GetSectionType(uint32_t flags, 1438 ConstString section_name) { 1439 1440 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1441 return eSectionTypeCode; 1442 1443 uint32_t mach_sect_type = flags & SECTION_TYPE; 1444 static ConstString g_sect_name_objc_data("__objc_data"); 1445 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1446 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1447 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1448 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1449 static ConstString g_sect_name_objc_const("__objc_const"); 1450 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1451 static ConstString g_sect_name_cfstring("__cfstring"); 1452 1453 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1454 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1455 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1456 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1457 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1458 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1459 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1460 static ConstString g_sect_name_dwarf_debug_names("__debug_names"); 1461 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1462 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1463 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1464 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1465 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1466 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1467 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1468 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1469 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1470 static ConstString g_sect_name_eh_frame("__eh_frame"); 1471 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1472 static ConstString g_sect_name_text("__text"); 1473 static ConstString g_sect_name_data("__data"); 1474 static ConstString g_sect_name_go_symtab("__gosymtab"); 1475 1476 if (section_name == g_sect_name_dwarf_debug_abbrev) 1477 return eSectionTypeDWARFDebugAbbrev; 1478 if (section_name == g_sect_name_dwarf_debug_aranges) 1479 return eSectionTypeDWARFDebugAranges; 1480 if (section_name == g_sect_name_dwarf_debug_frame) 1481 return eSectionTypeDWARFDebugFrame; 1482 if (section_name == g_sect_name_dwarf_debug_info) 1483 return eSectionTypeDWARFDebugInfo; 1484 if (section_name == g_sect_name_dwarf_debug_line) 1485 return eSectionTypeDWARFDebugLine; 1486 if (section_name == g_sect_name_dwarf_debug_loc) 1487 return eSectionTypeDWARFDebugLoc; 1488 if (section_name == g_sect_name_dwarf_debug_macinfo) 1489 return eSectionTypeDWARFDebugMacInfo; 1490 if (section_name == g_sect_name_dwarf_debug_names) 1491 return eSectionTypeDWARFDebugNames; 1492 if (section_name == g_sect_name_dwarf_debug_pubnames) 1493 return eSectionTypeDWARFDebugPubNames; 1494 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1495 return eSectionTypeDWARFDebugPubTypes; 1496 if (section_name == g_sect_name_dwarf_debug_ranges) 1497 return eSectionTypeDWARFDebugRanges; 1498 if (section_name == g_sect_name_dwarf_debug_str) 1499 return eSectionTypeDWARFDebugStr; 1500 if (section_name == g_sect_name_dwarf_debug_types) 1501 return eSectionTypeDWARFDebugTypes; 1502 if (section_name == g_sect_name_dwarf_apple_names) 1503 return eSectionTypeDWARFAppleNames; 1504 if (section_name == g_sect_name_dwarf_apple_types) 1505 return eSectionTypeDWARFAppleTypes; 1506 if (section_name == g_sect_name_dwarf_apple_namespaces) 1507 return eSectionTypeDWARFAppleNamespaces; 1508 if (section_name == g_sect_name_dwarf_apple_objc) 1509 return eSectionTypeDWARFAppleObjC; 1510 if (section_name == g_sect_name_objc_selrefs) 1511 return eSectionTypeDataCStringPointers; 1512 if (section_name == g_sect_name_objc_msgrefs) 1513 return eSectionTypeDataObjCMessageRefs; 1514 if (section_name == g_sect_name_eh_frame) 1515 return eSectionTypeEHFrame; 1516 if (section_name == g_sect_name_compact_unwind) 1517 return eSectionTypeCompactUnwind; 1518 if (section_name == g_sect_name_cfstring) 1519 return eSectionTypeDataObjCCFStrings; 1520 if (section_name == g_sect_name_go_symtab) 1521 return eSectionTypeGoSymtab; 1522 if (section_name == g_sect_name_objc_data || 1523 section_name == g_sect_name_objc_classrefs || 1524 section_name == g_sect_name_objc_superrefs || 1525 section_name == g_sect_name_objc_const || 1526 section_name == g_sect_name_objc_classlist) { 1527 return eSectionTypeDataPointers; 1528 } 1529 1530 switch (mach_sect_type) { 1531 // TODO: categorize sections by other flags for regular sections 1532 case S_REGULAR: 1533 if (section_name == g_sect_name_text) 1534 return eSectionTypeCode; 1535 if (section_name == g_sect_name_data) 1536 return eSectionTypeData; 1537 return eSectionTypeOther; 1538 case S_ZEROFILL: 1539 return eSectionTypeZeroFill; 1540 case S_CSTRING_LITERALS: // section with only literal C strings 1541 return eSectionTypeDataCString; 1542 case S_4BYTE_LITERALS: // section with only 4 byte literals 1543 return eSectionTypeData4; 1544 case S_8BYTE_LITERALS: // section with only 8 byte literals 1545 return eSectionTypeData8; 1546 case S_LITERAL_POINTERS: // section with only pointers to literals 1547 return eSectionTypeDataPointers; 1548 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1549 return eSectionTypeDataPointers; 1550 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1551 return eSectionTypeDataPointers; 1552 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1553 // the reserved2 field 1554 return eSectionTypeCode; 1555 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1556 // initialization 1557 return eSectionTypeDataPointers; 1558 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1559 // termination 1560 return eSectionTypeDataPointers; 1561 case S_COALESCED: 1562 return eSectionTypeOther; 1563 case S_GB_ZEROFILL: 1564 return eSectionTypeZeroFill; 1565 case S_INTERPOSING: // section with only pairs of function pointers for 1566 // interposing 1567 return eSectionTypeCode; 1568 case S_16BYTE_LITERALS: // section with only 16 byte literals 1569 return eSectionTypeData16; 1570 case S_DTRACE_DOF: 1571 return eSectionTypeDebug; 1572 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1573 return eSectionTypeDataPointers; 1574 default: 1575 return eSectionTypeOther; 1576 } 1577 } 1578 1579 struct ObjectFileMachO::SegmentParsingContext { 1580 const EncryptedFileRanges EncryptedRanges; 1581 lldb_private::SectionList &UnifiedList; 1582 uint32_t NextSegmentIdx = 0; 1583 uint32_t NextSectionIdx = 0; 1584 bool FileAddressesChanged = false; 1585 1586 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1587 lldb_private::SectionList &UnifiedList) 1588 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1589 }; 1590 1591 void ObjectFileMachO::ProcessSegmentCommand(const load_command &load_cmd_, 1592 lldb::offset_t offset, 1593 uint32_t cmd_idx, 1594 SegmentParsingContext &context) { 1595 segment_command_64 load_cmd; 1596 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1597 1598 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1599 return; 1600 1601 ModuleSP module_sp = GetModule(); 1602 const bool is_core = GetType() == eTypeCoreFile; 1603 const bool is_dsym = (m_header.filetype == MH_DSYM); 1604 bool add_section = true; 1605 bool add_to_unified = true; 1606 ConstString const_segname( 1607 load_cmd.segname, 1608 std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 1609 1610 SectionSP unified_section_sp( 1611 context.UnifiedList.FindSectionByName(const_segname)); 1612 if (is_dsym && unified_section_sp) { 1613 if (const_segname == GetSegmentNameLINKEDIT()) { 1614 // We need to keep the __LINKEDIT segment private to this object file 1615 // only 1616 add_to_unified = false; 1617 } else { 1618 // This is the dSYM file and this section has already been created by the 1619 // object file, no need to create it. 1620 add_section = false; 1621 } 1622 } 1623 load_cmd.vmaddr = m_data.GetAddress(&offset); 1624 load_cmd.vmsize = m_data.GetAddress(&offset); 1625 load_cmd.fileoff = m_data.GetAddress(&offset); 1626 load_cmd.filesize = m_data.GetAddress(&offset); 1627 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1628 return; 1629 1630 SanitizeSegmentCommand(load_cmd, cmd_idx); 1631 1632 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1633 const bool segment_is_encrypted = 1634 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1635 1636 // Keep a list of mach segments around in case we need to get at data that 1637 // isn't stored in the abstracted Sections. 1638 m_mach_segments.push_back(load_cmd); 1639 1640 // Use a segment ID of the segment index shifted left by 8 so they never 1641 // conflict with any of the sections. 1642 SectionSP segment_sp; 1643 if (add_section && (const_segname || is_core)) { 1644 segment_sp.reset(new Section( 1645 module_sp, // Module to which this section belongs 1646 this, // Object file to which this sections belongs 1647 ++context.NextSegmentIdx 1648 << 8, // Section ID is the 1 based segment index 1649 // shifted right by 8 bits as not to collide with any of the 256 1650 // section IDs that are possible 1651 const_segname, // Name of this section 1652 eSectionTypeContainer, // This section is a container of other 1653 // sections. 1654 load_cmd.vmaddr, // File VM address == addresses as they are 1655 // found in the object file 1656 load_cmd.vmsize, // VM size in bytes of this section 1657 load_cmd.fileoff, // Offset to the data for this section in 1658 // the file 1659 load_cmd.filesize, // Size in bytes of this section as found 1660 // in the file 1661 0, // Segments have no alignment information 1662 load_cmd.flags)); // Flags for this section 1663 1664 segment_sp->SetIsEncrypted(segment_is_encrypted); 1665 m_sections_ap->AddSection(segment_sp); 1666 segment_sp->SetPermissions(segment_permissions); 1667 if (add_to_unified) 1668 context.UnifiedList.AddSection(segment_sp); 1669 } else if (unified_section_sp) { 1670 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1671 // Check to see if the module was read from memory? 1672 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) { 1673 // We have a module that is in memory and needs to have its file 1674 // address adjusted. We need to do this because when we load a file 1675 // from memory, its addresses will be slid already, yet the addresses 1676 // in the new symbol file will still be unslid. Since everything is 1677 // stored as section offset, this shouldn't cause any problems. 1678 1679 // Make sure we've parsed the symbol table from the ObjectFile before 1680 // we go around changing its Sections. 1681 module_sp->GetObjectFile()->GetSymtab(); 1682 // eh_frame would present the same problems but we parse that on a per- 1683 // function basis as-needed so it's more difficult to remove its use of 1684 // the Sections. Realistically, the environments where this code path 1685 // will be taken will not have eh_frame sections. 1686 1687 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1688 1689 // Notify the module that the section addresses have been changed once 1690 // we're done so any file-address caches can be updated. 1691 context.FileAddressesChanged = true; 1692 } 1693 } 1694 m_sections_ap->AddSection(unified_section_sp); 1695 } 1696 1697 struct section_64 sect64; 1698 ::memset(§64, 0, sizeof(sect64)); 1699 // Push a section into our mach sections for the section at index zero 1700 // (NO_SECT) if we don't have any mach sections yet... 1701 if (m_mach_sections.empty()) 1702 m_mach_sections.push_back(sect64); 1703 uint32_t segment_sect_idx; 1704 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1705 1706 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1707 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1708 ++segment_sect_idx) { 1709 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1710 sizeof(sect64.sectname)) == NULL) 1711 break; 1712 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1713 sizeof(sect64.segname)) == NULL) 1714 break; 1715 sect64.addr = m_data.GetAddress(&offset); 1716 sect64.size = m_data.GetAddress(&offset); 1717 1718 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1719 break; 1720 1721 // Keep a list of mach sections around in case we need to get at data that 1722 // isn't stored in the abstracted Sections. 1723 m_mach_sections.push_back(sect64); 1724 1725 if (add_section) { 1726 ConstString section_name( 1727 sect64.sectname, 1728 std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1729 if (!const_segname) { 1730 // We have a segment with no name so we need to conjure up segments 1731 // that correspond to the section's segname if there isn't already such 1732 // a section. If there is such a section, we resize the section so that 1733 // it spans all sections. We also mark these sections as fake so 1734 // address matches don't hit if they land in the gaps between the child 1735 // sections. 1736 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1737 sizeof(sect64.segname)); 1738 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1739 if (segment_sp.get()) { 1740 Section *segment = segment_sp.get(); 1741 // Grow the section size as needed. 1742 const lldb::addr_t sect64_min_addr = sect64.addr; 1743 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1744 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1745 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1746 const lldb::addr_t curr_seg_max_addr = 1747 curr_seg_min_addr + curr_seg_byte_size; 1748 if (sect64_min_addr >= curr_seg_min_addr) { 1749 const lldb::addr_t new_seg_byte_size = 1750 sect64_max_addr - curr_seg_min_addr; 1751 // Only grow the section size if needed 1752 if (new_seg_byte_size > curr_seg_byte_size) 1753 segment->SetByteSize(new_seg_byte_size); 1754 } else { 1755 // We need to change the base address of the segment and adjust the 1756 // child section offsets for all existing children. 1757 const lldb::addr_t slide_amount = 1758 sect64_min_addr - curr_seg_min_addr; 1759 segment->Slide(slide_amount, false); 1760 segment->GetChildren().Slide(-slide_amount, false); 1761 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1762 } 1763 1764 // Grow the section size as needed. 1765 if (sect64.offset) { 1766 const lldb::addr_t segment_min_file_offset = 1767 segment->GetFileOffset(); 1768 const lldb::addr_t segment_max_file_offset = 1769 segment_min_file_offset + segment->GetFileSize(); 1770 1771 const lldb::addr_t section_min_file_offset = sect64.offset; 1772 const lldb::addr_t section_max_file_offset = 1773 section_min_file_offset + sect64.size; 1774 const lldb::addr_t new_file_offset = 1775 std::min(section_min_file_offset, segment_min_file_offset); 1776 const lldb::addr_t new_file_size = 1777 std::max(section_max_file_offset, segment_max_file_offset) - 1778 new_file_offset; 1779 segment->SetFileOffset(new_file_offset); 1780 segment->SetFileSize(new_file_size); 1781 } 1782 } else { 1783 // Create a fake section for the section's named segment 1784 segment_sp.reset(new Section( 1785 segment_sp, // Parent section 1786 module_sp, // Module to which this section belongs 1787 this, // Object file to which this section belongs 1788 ++context.NextSegmentIdx 1789 << 8, // Section ID is the 1 based segment index 1790 // shifted right by 8 bits as not to 1791 // collide with any of the 256 section IDs 1792 // that are possible 1793 const_segname, // Name of this section 1794 eSectionTypeContainer, // This section is a container of 1795 // other sections. 1796 sect64.addr, // File VM address == addresses as they are 1797 // found in the object file 1798 sect64.size, // VM size in bytes of this section 1799 sect64.offset, // Offset to the data for this section in 1800 // the file 1801 sect64.offset ? sect64.size : 0, // Size in bytes of 1802 // this section as 1803 // found in the file 1804 sect64.align, 1805 load_cmd.flags)); // Flags for this section 1806 segment_sp->SetIsFake(true); 1807 segment_sp->SetPermissions(segment_permissions); 1808 m_sections_ap->AddSection(segment_sp); 1809 if (add_to_unified) 1810 context.UnifiedList.AddSection(segment_sp); 1811 segment_sp->SetIsEncrypted(segment_is_encrypted); 1812 } 1813 } 1814 assert(segment_sp.get()); 1815 1816 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1817 1818 SectionSP section_sp(new Section( 1819 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1820 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1821 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1822 sect64.flags)); 1823 // Set the section to be encrypted to match the segment 1824 1825 bool section_is_encrypted = false; 1826 if (!segment_is_encrypted && load_cmd.filesize != 0) 1827 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1828 sect64.offset) != NULL; 1829 1830 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1831 section_sp->SetPermissions(segment_permissions); 1832 segment_sp->GetChildren().AddSection(section_sp); 1833 1834 if (segment_sp->IsFake()) { 1835 segment_sp.reset(); 1836 const_segname.Clear(); 1837 } 1838 } 1839 } 1840 if (segment_sp && is_dsym) { 1841 if (first_segment_sectID <= context.NextSectionIdx) { 1842 lldb::user_id_t sect_uid; 1843 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1844 ++sect_uid) { 1845 SectionSP curr_section_sp( 1846 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1847 SectionSP next_section_sp; 1848 if (sect_uid + 1 <= context.NextSectionIdx) 1849 next_section_sp = 1850 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1851 1852 if (curr_section_sp.get()) { 1853 if (curr_section_sp->GetByteSize() == 0) { 1854 if (next_section_sp.get() != NULL) 1855 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1856 curr_section_sp->GetFileAddress()); 1857 else 1858 curr_section_sp->SetByteSize(load_cmd.vmsize); 1859 } 1860 } 1861 } 1862 } 1863 } 1864 } 1865 1866 void ObjectFileMachO::ProcessDysymtabCommand(const load_command &load_cmd, 1867 lldb::offset_t offset) { 1868 m_dysymtab.cmd = load_cmd.cmd; 1869 m_dysymtab.cmdsize = load_cmd.cmdsize; 1870 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1871 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1872 } 1873 1874 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1875 if (m_sections_ap) 1876 return; 1877 1878 m_sections_ap.reset(new SectionList()); 1879 1880 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1881 // bool dump_sections = false; 1882 ModuleSP module_sp(GetModule()); 1883 1884 offset = MachHeaderSizeFromMagic(m_header.magic); 1885 1886 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1887 struct load_command load_cmd; 1888 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1889 const lldb::offset_t load_cmd_offset = offset; 1890 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1891 break; 1892 1893 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1894 ProcessSegmentCommand(load_cmd, offset, i, context); 1895 else if (load_cmd.cmd == LC_DYSYMTAB) 1896 ProcessDysymtabCommand(load_cmd, offset); 1897 1898 offset = load_cmd_offset + load_cmd.cmdsize; 1899 } 1900 1901 if (context.FileAddressesChanged && module_sp) 1902 module_sp->SectionFileAddressesChanged(); 1903 } 1904 1905 class MachSymtabSectionInfo { 1906 public: 1907 MachSymtabSectionInfo(SectionList *section_list) 1908 : m_section_list(section_list), m_section_infos() { 1909 // Get the number of sections down to a depth of 1 to include all segments 1910 // and their sections, but no other sections that may be added for debug 1911 // map or 1912 m_section_infos.resize(section_list->GetNumSections(1)); 1913 } 1914 1915 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1916 if (n_sect == 0) 1917 return SectionSP(); 1918 if (n_sect < m_section_infos.size()) { 1919 if (!m_section_infos[n_sect].section_sp) { 1920 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1921 m_section_infos[n_sect].section_sp = section_sp; 1922 if (section_sp) { 1923 m_section_infos[n_sect].vm_range.SetBaseAddress( 1924 section_sp->GetFileAddress()); 1925 m_section_infos[n_sect].vm_range.SetByteSize( 1926 section_sp->GetByteSize()); 1927 } else { 1928 Host::SystemLog(Host::eSystemLogError, 1929 "error: unable to find section for section %u\n", 1930 n_sect); 1931 } 1932 } 1933 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1934 // Symbol is in section. 1935 return m_section_infos[n_sect].section_sp; 1936 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1937 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1938 file_addr) { 1939 // Symbol is in section with zero size, but has the same start address 1940 // as the section. This can happen with linker symbols (symbols that 1941 // start with the letter 'l' or 'L'. 1942 return m_section_infos[n_sect].section_sp; 1943 } 1944 } 1945 return m_section_list->FindSectionContainingFileAddress(file_addr); 1946 } 1947 1948 protected: 1949 struct SectionInfo { 1950 SectionInfo() : vm_range(), section_sp() {} 1951 1952 VMRange vm_range; 1953 SectionSP section_sp; 1954 }; 1955 SectionList *m_section_list; 1956 std::vector<SectionInfo> m_section_infos; 1957 }; 1958 1959 struct TrieEntry { 1960 TrieEntry() 1961 : name(), address(LLDB_INVALID_ADDRESS), flags(0), other(0), 1962 import_name() {} 1963 1964 void Clear() { 1965 name.Clear(); 1966 address = LLDB_INVALID_ADDRESS; 1967 flags = 0; 1968 other = 0; 1969 import_name.Clear(); 1970 } 1971 1972 void Dump() const { 1973 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1974 static_cast<unsigned long long>(address), 1975 static_cast<unsigned long long>(flags), 1976 static_cast<unsigned long long>(other), name.GetCString()); 1977 if (import_name) 1978 printf(" -> \"%s\"\n", import_name.GetCString()); 1979 else 1980 printf("\n"); 1981 } 1982 ConstString name; 1983 uint64_t address; 1984 uint64_t flags; 1985 uint64_t other; 1986 ConstString import_name; 1987 }; 1988 1989 struct TrieEntryWithOffset { 1990 lldb::offset_t nodeOffset; 1991 TrieEntry entry; 1992 1993 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1994 1995 void Dump(uint32_t idx) const { 1996 printf("[%3u] 0x%16.16llx: ", idx, 1997 static_cast<unsigned long long>(nodeOffset)); 1998 entry.Dump(); 1999 } 2000 2001 bool operator<(const TrieEntryWithOffset &other) const { 2002 return (nodeOffset < other.nodeOffset); 2003 } 2004 }; 2005 2006 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2007 const bool is_arm, 2008 std::vector<llvm::StringRef> &nameSlices, 2009 std::set<lldb::addr_t> &resolver_addresses, 2010 std::vector<TrieEntryWithOffset> &output) { 2011 if (!data.ValidOffset(offset)) 2012 return true; 2013 2014 const uint64_t terminalSize = data.GetULEB128(&offset); 2015 lldb::offset_t children_offset = offset + terminalSize; 2016 if (terminalSize != 0) { 2017 TrieEntryWithOffset e(offset); 2018 e.entry.flags = data.GetULEB128(&offset); 2019 const char *import_name = NULL; 2020 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2021 e.entry.address = 0; 2022 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2023 import_name = data.GetCStr(&offset); 2024 } else { 2025 e.entry.address = data.GetULEB128(&offset); 2026 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2027 e.entry.other = data.GetULEB128(&offset); 2028 uint64_t resolver_addr = e.entry.other; 2029 if (is_arm) 2030 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2031 resolver_addresses.insert(resolver_addr); 2032 } else 2033 e.entry.other = 0; 2034 } 2035 // Only add symbols that are reexport symbols with a valid import name 2036 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && 2037 import_name[0]) { 2038 std::string name; 2039 if (!nameSlices.empty()) { 2040 for (auto name_slice : nameSlices) 2041 name.append(name_slice.data(), name_slice.size()); 2042 } 2043 if (name.size() > 1) { 2044 // Skip the leading '_' 2045 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2046 } 2047 if (import_name) { 2048 // Skip the leading '_' 2049 e.entry.import_name.SetCString(import_name + 1); 2050 } 2051 output.push_back(e); 2052 } 2053 } 2054 2055 const uint8_t childrenCount = data.GetU8(&children_offset); 2056 for (uint8_t i = 0; i < childrenCount; ++i) { 2057 const char *cstr = data.GetCStr(&children_offset); 2058 if (cstr) 2059 nameSlices.push_back(llvm::StringRef(cstr)); 2060 else 2061 return false; // Corrupt data 2062 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2063 if (childNodeOffset) { 2064 if (!ParseTrieEntries(data, childNodeOffset, is_arm, nameSlices, 2065 resolver_addresses, output)) { 2066 return false; 2067 } 2068 } 2069 nameSlices.pop_back(); 2070 } 2071 return true; 2072 } 2073 2074 // Read the UUID out of a dyld_shared_cache file on-disk. 2075 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2076 const ByteOrder byte_order, 2077 const uint32_t addr_byte_size) { 2078 UUID dsc_uuid; 2079 DataBufferSP DscData = MapFileData( 2080 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2081 if (!DscData) 2082 return dsc_uuid; 2083 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2084 2085 char version_str[7]; 2086 lldb::offset_t offset = 0; 2087 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2088 version_str[6] = '\0'; 2089 if (strcmp(version_str, "dyld_v") == 0) { 2090 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2091 dsc_uuid = UUID::fromOptionalData( 2092 dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t)); 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 address has been set in the target. We can't 2254 // fail to read the symbol table, so calculate the right address 2255 // manually 2256 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2257 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2258 } 2259 2260 const addr_t linkedit_file_offset = 2261 linkedit_section_sp->GetFileOffset(); 2262 const addr_t symoff_addr = linkedit_load_addr + 2263 symtab_load_command.symoff - 2264 linkedit_file_offset; 2265 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2266 linkedit_file_offset; 2267 2268 bool data_was_read = false; 2269 2270 #if defined(__APPLE__) && \ 2271 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2272 if (m_header.flags & 0x80000000u && 2273 process->GetAddressByteSize() == sizeof(void *)) { 2274 // This mach-o memory file is in the dyld shared cache. If this 2275 // program is not remote and this is iOS, then this process will 2276 // share the same shared cache as the process we are debugging and we 2277 // can read the entire __LINKEDIT from the address space in this 2278 // process. This is a needed optimization that is used for local iOS 2279 // debugging only since all shared libraries in the shared cache do 2280 // not have corresponding files that exist in the file system of the 2281 // device. They have been combined into a single file. This means we 2282 // always have to load these files from memory. All of the symbol and 2283 // string tables from all of the __LINKEDIT sections from the shared 2284 // libraries in the shared cache have been merged into a single large 2285 // symbol and string table. Reading all of this symbol and string 2286 // table data across can slow down debug launch times, so we optimize 2287 // this by reading the memory for the __LINKEDIT section from this 2288 // process. 2289 2290 UUID lldb_shared_cache; 2291 addr_t lldb_shared_cache_addr; 2292 GetLLDBSharedCacheUUID (lldb_shared_cache_addr, lldb_shared_cache); 2293 UUID process_shared_cache; 2294 addr_t process_shared_cache_addr; 2295 GetProcessSharedCacheUUID(process, process_shared_cache_addr, process_shared_cache); 2296 bool use_lldb_cache = true; 2297 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2298 (lldb_shared_cache != process_shared_cache 2299 || process_shared_cache_addr != lldb_shared_cache_addr)) { 2300 use_lldb_cache = false; 2301 } 2302 2303 PlatformSP platform_sp(target.GetPlatform()); 2304 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2305 data_was_read = true; 2306 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2307 eByteOrderLittle); 2308 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2309 eByteOrderLittle); 2310 if (function_starts_load_command.cmd) { 2311 const addr_t func_start_addr = 2312 linkedit_load_addr + function_starts_load_command.dataoff - 2313 linkedit_file_offset; 2314 function_starts_data.SetData( 2315 (void *)func_start_addr, 2316 function_starts_load_command.datasize, eByteOrderLittle); 2317 } 2318 } 2319 } 2320 #endif 2321 2322 if (!data_was_read) { 2323 // Always load dyld - the dynamic linker - from memory if we didn't 2324 // find a binary anywhere else. lldb will not register 2325 // dylib/framework/bundle loads/unloads if we don't have the dyld 2326 // symbols, we force dyld to load from memory despite the user's 2327 // target.memory-module-load-level setting. 2328 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2329 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2330 DataBufferSP nlist_data_sp( 2331 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2332 if (nlist_data_sp) 2333 nlist_data.SetData(nlist_data_sp, 0, 2334 nlist_data_sp->GetByteSize()); 2335 if (m_dysymtab.nindirectsyms != 0) { 2336 const addr_t indirect_syms_addr = linkedit_load_addr + 2337 m_dysymtab.indirectsymoff - 2338 linkedit_file_offset; 2339 DataBufferSP indirect_syms_data_sp( 2340 ReadMemory(process_sp, indirect_syms_addr, 2341 m_dysymtab.nindirectsyms * 4)); 2342 if (indirect_syms_data_sp) 2343 indirect_symbol_index_data.SetData( 2344 indirect_syms_data_sp, 0, 2345 indirect_syms_data_sp->GetByteSize()); 2346 // If this binary is outside the shared cache, 2347 // cache the string table. 2348 // Binaries in the shared cache all share a giant string table, and 2349 // we can't share the string tables across multiple ObjectFileMachO's, 2350 // so we'd end up re-reading this mega-strtab for every binary 2351 // in the shared cache - it would be a big perf problem. 2352 // For binaries outside the shared cache, it's faster to read the 2353 // entire strtab at once instead of piece-by-piece as we process 2354 // the nlist records. 2355 if ((m_header.flags & 0x80000000u) == 0) { 2356 DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, 2357 strtab_data_byte_size)); 2358 if (strtab_data_sp) { 2359 strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 2360 } 2361 } 2362 } 2363 } 2364 if (memory_module_load_level >= 2365 eMemoryModuleLoadLevelPartial) { 2366 if (function_starts_load_command.cmd) { 2367 const addr_t func_start_addr = 2368 linkedit_load_addr + function_starts_load_command.dataoff - 2369 linkedit_file_offset; 2370 DataBufferSP func_start_data_sp( 2371 ReadMemory(process_sp, func_start_addr, 2372 function_starts_load_command.datasize)); 2373 if (func_start_data_sp) 2374 function_starts_data.SetData(func_start_data_sp, 0, 2375 func_start_data_sp->GetByteSize()); 2376 } 2377 } 2378 } 2379 } 2380 } else { 2381 nlist_data.SetData(m_data, symtab_load_command.symoff, 2382 nlist_data_byte_size); 2383 strtab_data.SetData(m_data, symtab_load_command.stroff, 2384 strtab_data_byte_size); 2385 2386 if (dyld_info.export_size > 0) { 2387 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2388 dyld_info.export_size); 2389 } 2390 2391 if (m_dysymtab.nindirectsyms != 0) { 2392 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2393 m_dysymtab.nindirectsyms * 4); 2394 } 2395 if (function_starts_load_command.cmd) { 2396 function_starts_data.SetData(m_data, 2397 function_starts_load_command.dataoff, 2398 function_starts_load_command.datasize); 2399 } 2400 } 2401 2402 if (nlist_data.GetByteSize() == 0 && 2403 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2404 if (log) 2405 module_sp->LogMessage(log, "failed to read nlist data"); 2406 return 0; 2407 } 2408 2409 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2410 if (!have_strtab_data) { 2411 if (process) { 2412 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2413 if (log) 2414 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2415 return 0; 2416 } 2417 } else { 2418 if (log) 2419 module_sp->LogMessage(log, "failed to read strtab data"); 2420 return 0; 2421 } 2422 } 2423 2424 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2425 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2426 const ConstString &g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2427 const ConstString &g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2428 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2429 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2430 SectionSP text_section_sp( 2431 section_list->FindSectionByName(g_segment_name_TEXT)); 2432 SectionSP data_section_sp( 2433 section_list->FindSectionByName(g_segment_name_DATA)); 2434 SectionSP data_dirty_section_sp( 2435 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2436 SectionSP data_const_section_sp( 2437 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2438 SectionSP objc_section_sp( 2439 section_list->FindSectionByName(g_segment_name_OBJC)); 2440 SectionSP eh_frame_section_sp; 2441 if (text_section_sp.get()) 2442 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2443 g_section_name_eh_frame); 2444 else 2445 eh_frame_section_sp = 2446 section_list->FindSectionByName(g_section_name_eh_frame); 2447 2448 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2449 2450 // lldb works best if it knows the start address of all functions in a 2451 // module. Linker symbols or debug info are normally the best source of 2452 // information for start addr / size but they may be stripped in a released 2453 // binary. Two additional sources of information exist in Mach-O binaries: 2454 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2455 // function's start address in the 2456 // binary, relative to the text section. 2457 // eh_frame - the eh_frame FDEs have the start addr & size of 2458 // each function 2459 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2460 // all modern binaries. 2461 // Binaries built to run on older releases may need to use eh_frame 2462 // information. 2463 2464 if (text_section_sp && function_starts_data.GetByteSize()) { 2465 FunctionStarts::Entry function_start_entry; 2466 function_start_entry.data = false; 2467 lldb::offset_t function_start_offset = 0; 2468 function_start_entry.addr = text_section_sp->GetFileAddress(); 2469 uint64_t delta; 2470 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2471 0) { 2472 // Now append the current entry 2473 function_start_entry.addr += delta; 2474 function_starts.Append(function_start_entry); 2475 } 2476 } else { 2477 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2478 // load command claiming an eh_frame but it doesn't actually have the 2479 // eh_frame content. And if we have a dSYM, we don't need to do any of 2480 // this fill-in-the-missing-symbols works anyway - the debug info should 2481 // give us all the functions in the module. 2482 if (text_section_sp.get() && eh_frame_section_sp.get() && 2483 m_type != eTypeDebugInfo) { 2484 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2485 DWARFCallFrameInfo::EH); 2486 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2487 eh_frame.GetFunctionAddressAndSizeVector(functions); 2488 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2489 size_t count = functions.GetSize(); 2490 for (size_t i = 0; i < count; ++i) { 2491 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2492 functions.GetEntryAtIndex(i); 2493 if (func) { 2494 FunctionStarts::Entry function_start_entry; 2495 function_start_entry.addr = func->base - text_base_addr; 2496 function_starts.Append(function_start_entry); 2497 } 2498 } 2499 } 2500 } 2501 2502 const size_t function_starts_count = function_starts.GetSize(); 2503 2504 // For user process binaries (executables, dylibs, frameworks, bundles), if 2505 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2506 // going to assume the binary has been stripped. Don't allow assembly 2507 // language instruction emulation because we don't know proper function 2508 // start boundaries. 2509 // 2510 // For all other types of binaries (kernels, stand-alone bare board 2511 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2512 // sections - we should not make any assumptions about them based on that. 2513 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2514 m_allow_assembly_emulation_unwind_plans = false; 2515 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2516 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2517 2518 if (unwind_or_symbol_log) 2519 module_sp->LogMessage( 2520 unwind_or_symbol_log, 2521 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2522 } 2523 2524 const user_id_t TEXT_eh_frame_sectID = 2525 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2526 : static_cast<user_id_t>(NO_SECT); 2527 2528 lldb::offset_t nlist_data_offset = 0; 2529 2530 uint32_t N_SO_index = UINT32_MAX; 2531 2532 MachSymtabSectionInfo section_info(section_list); 2533 std::vector<uint32_t> N_FUN_indexes; 2534 std::vector<uint32_t> N_NSYM_indexes; 2535 std::vector<uint32_t> N_INCL_indexes; 2536 std::vector<uint32_t> N_BRAC_indexes; 2537 std::vector<uint32_t> N_COMM_indexes; 2538 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2539 typedef std::map<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2540 typedef std::map<const char *, uint32_t> ConstNameToSymbolIndexMap; 2541 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2542 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2543 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2544 // Any symbols that get merged into another will get an entry in this map 2545 // so we know 2546 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2547 uint32_t nlist_idx = 0; 2548 Symbol *symbol_ptr = NULL; 2549 2550 uint32_t sym_idx = 0; 2551 Symbol *sym = NULL; 2552 size_t num_syms = 0; 2553 std::string memory_symbol_name; 2554 uint32_t unmapped_local_symbols_found = 0; 2555 2556 std::vector<TrieEntryWithOffset> trie_entries; 2557 std::set<lldb::addr_t> resolver_addresses; 2558 2559 if (dyld_trie_data.GetByteSize() > 0) { 2560 std::vector<llvm::StringRef> nameSlices; 2561 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, 2562 resolver_addresses, trie_entries); 2563 2564 ConstString text_segment_name("__TEXT"); 2565 SectionSP text_segment_sp = 2566 GetSectionList()->FindSectionByName(text_segment_name); 2567 if (text_segment_sp) { 2568 const lldb::addr_t text_segment_file_addr = 2569 text_segment_sp->GetFileAddress(); 2570 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2571 for (auto &e : trie_entries) 2572 e.entry.address += text_segment_file_addr; 2573 } 2574 } 2575 } 2576 2577 typedef std::set<ConstString> IndirectSymbols; 2578 IndirectSymbols indirect_symbol_names; 2579 2580 #if defined(__APPLE__) && \ 2581 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2582 2583 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2584 // optimized by moving LOCAL symbols out of the memory mapped portion of 2585 // the DSC. The symbol information has all been retained, but it isn't 2586 // available in the normal nlist data. However, there *are* duplicate 2587 // entries of *some* 2588 // LOCAL symbols in the normal nlist data. To handle this situation 2589 // correctly, we must first attempt 2590 // to parse any DSC unmapped symbol information. If we find any, we set a 2591 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2592 2593 if (m_header.flags & 0x80000000u) { 2594 // Before we can start mapping the DSC, we need to make certain the 2595 // target process is actually using the cache we can find. 2596 2597 // Next we need to determine the correct path for the dyld shared cache. 2598 2599 ArchSpec header_arch; 2600 GetArchitecture(header_arch); 2601 char dsc_path[PATH_MAX]; 2602 char dsc_path_development[PATH_MAX]; 2603 2604 snprintf( 2605 dsc_path, sizeof(dsc_path), "%s%s%s", 2606 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2607 */ 2608 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2609 header_arch.GetArchitectureName()); 2610 2611 snprintf( 2612 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2613 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2614 */ 2615 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2616 header_arch.GetArchitectureName(), ".development"); 2617 2618 FileSpec dsc_nondevelopment_filespec(dsc_path, false); 2619 FileSpec dsc_development_filespec(dsc_path_development, false); 2620 FileSpec dsc_filespec; 2621 2622 UUID dsc_uuid; 2623 UUID process_shared_cache_uuid; 2624 addr_t process_shared_cache_base_addr; 2625 2626 if (process) { 2627 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, process_shared_cache_uuid); 2628 } 2629 2630 // First see if we can find an exact match for the inferior process 2631 // shared cache UUID in the development or non-development shared caches 2632 // on disk. 2633 if (process_shared_cache_uuid.IsValid()) { 2634 if (dsc_development_filespec.Exists()) { 2635 UUID dsc_development_uuid = GetSharedCacheUUID( 2636 dsc_development_filespec, byte_order, addr_byte_size); 2637 if (dsc_development_uuid.IsValid() && 2638 dsc_development_uuid == process_shared_cache_uuid) { 2639 dsc_filespec = dsc_development_filespec; 2640 dsc_uuid = dsc_development_uuid; 2641 } 2642 } 2643 if (!dsc_uuid.IsValid() && dsc_nondevelopment_filespec.Exists()) { 2644 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2645 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2646 if (dsc_nondevelopment_uuid.IsValid() && 2647 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2648 dsc_filespec = dsc_nondevelopment_filespec; 2649 dsc_uuid = dsc_nondevelopment_uuid; 2650 } 2651 } 2652 } 2653 2654 // Failing a UUID match, prefer the development dyld_shared cache if both 2655 // are present. 2656 if (!dsc_filespec.Exists()) { 2657 if (dsc_development_filespec.Exists()) { 2658 dsc_filespec = dsc_development_filespec; 2659 } else { 2660 dsc_filespec = dsc_nondevelopment_filespec; 2661 } 2662 } 2663 2664 /* The dyld_cache_header has a pointer to the 2665 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2666 The dyld_cache_local_symbols_info structure gives us three things: 2667 1. The start and count of the nlist records in the dyld_shared_cache 2668 file 2669 2. The start and size of the strings for these nlist records 2670 3. The start and count of dyld_cache_local_symbols_entry entries 2671 2672 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2673 dyld shared cache. 2674 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2675 the dyld shared cache. 2676 The dyld_cache_local_symbols_entry also lists the start of this 2677 dylib/framework's nlist records 2678 and the count of how many nlist records there are for this 2679 dylib/framework. 2680 */ 2681 2682 // Process the dyld shared cache header to find the unmapped symbols 2683 2684 DataBufferSP dsc_data_sp = MapFileData( 2685 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2686 if (!dsc_uuid.IsValid()) { 2687 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2688 } 2689 if (dsc_data_sp) { 2690 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2691 2692 bool uuid_match = true; 2693 if (dsc_uuid.IsValid() && process) { 2694 if (process_shared_cache_uuid.IsValid() && 2695 dsc_uuid != process_shared_cache_uuid) { 2696 // The on-disk dyld_shared_cache file is not the same as the one in 2697 // this process' memory, don't use it. 2698 uuid_match = false; 2699 ModuleSP module_sp(GetModule()); 2700 if (module_sp) 2701 module_sp->ReportWarning("process shared cache does not match " 2702 "on-disk dyld_shared_cache file, some " 2703 "symbol names will be missing."); 2704 } 2705 } 2706 2707 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2708 2709 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2710 2711 // If the mappingOffset points to a location inside the header, we've 2712 // opened an old dyld shared cache, and should not proceed further. 2713 if (uuid_match && 2714 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2715 2716 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2717 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2718 mappingOffset); 2719 2720 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2721 byte_order, addr_byte_size); 2722 offset = 0; 2723 2724 // The File addresses (from the in-memory Mach-O load commands) for 2725 // the shared libraries in the shared library cache need to be 2726 // adjusted by an offset to match up with the dylibOffset identifying 2727 // field in the dyld_cache_local_symbol_entry's. This offset is 2728 // recorded in mapping_offset_value. 2729 const uint64_t mapping_offset_value = 2730 dsc_mapping_info_data.GetU64(&offset); 2731 2732 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, 2733 localSymbolsOffset); 2734 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2735 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2736 2737 if (localSymbolsOffset && localSymbolsSize) { 2738 // Map the local symbols 2739 DataBufferSP dsc_local_symbols_data_sp = 2740 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2741 2742 if (dsc_local_symbols_data_sp) { 2743 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2744 byte_order, addr_byte_size); 2745 2746 offset = 0; 2747 2748 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2749 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2750 UndefinedNameToDescMap undefined_name_to_desc; 2751 SymbolIndexToName reexport_shlib_needs_fixup; 2752 2753 // Read the local_symbols_infos struct in one shot 2754 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2755 dsc_local_symbols_data.GetU32(&offset, 2756 &local_symbols_info.nlistOffset, 6); 2757 2758 SectionSP text_section_sp( 2759 section_list->FindSectionByName(GetSegmentNameTEXT())); 2760 2761 uint32_t header_file_offset = 2762 (text_section_sp->GetFileAddress() - mapping_offset_value); 2763 2764 offset = local_symbols_info.entriesOffset; 2765 for (uint32_t entry_index = 0; 2766 entry_index < local_symbols_info.entriesCount; 2767 entry_index++) { 2768 struct lldb_copy_dyld_cache_local_symbols_entry 2769 local_symbols_entry; 2770 local_symbols_entry.dylibOffset = 2771 dsc_local_symbols_data.GetU32(&offset); 2772 local_symbols_entry.nlistStartIndex = 2773 dsc_local_symbols_data.GetU32(&offset); 2774 local_symbols_entry.nlistCount = 2775 dsc_local_symbols_data.GetU32(&offset); 2776 2777 if (header_file_offset == local_symbols_entry.dylibOffset) { 2778 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2779 2780 // The normal nlist code cannot correctly size the Symbols 2781 // array, we need to allocate it here. 2782 sym = symtab->Resize( 2783 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2784 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2785 num_syms = symtab->GetNumSymbols(); 2786 2787 nlist_data_offset = 2788 local_symbols_info.nlistOffset + 2789 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2790 uint32_t string_table_offset = 2791 local_symbols_info.stringsOffset; 2792 2793 for (uint32_t nlist_index = 0; 2794 nlist_index < local_symbols_entry.nlistCount; 2795 nlist_index++) { 2796 ///////////////////////////// 2797 { 2798 struct nlist_64 nlist; 2799 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize( 2800 nlist_data_offset, nlist_byte_size)) 2801 break; 2802 2803 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked( 2804 &nlist_data_offset); 2805 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked( 2806 &nlist_data_offset); 2807 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked( 2808 &nlist_data_offset); 2809 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked( 2810 &nlist_data_offset); 2811 nlist.n_value = 2812 dsc_local_symbols_data.GetAddress_unchecked( 2813 &nlist_data_offset); 2814 2815 SymbolType type = eSymbolTypeInvalid; 2816 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2817 string_table_offset + nlist.n_strx); 2818 2819 if (symbol_name == NULL) { 2820 // No symbol should be NULL, even the symbols with no 2821 // string values should have an offset zero which 2822 // points to an empty C-string 2823 Host::SystemLog( 2824 Host::eSystemLogError, 2825 "error: DSC unmapped local symbol[%u] has invalid " 2826 "string table offset 0x%x in %s, ignoring symbol\n", 2827 entry_index, nlist.n_strx, 2828 module_sp->GetFileSpec().GetPath().c_str()); 2829 continue; 2830 } 2831 if (symbol_name[0] == '\0') 2832 symbol_name = NULL; 2833 2834 const char *symbol_name_non_abi_mangled = NULL; 2835 2836 SectionSP symbol_section; 2837 uint32_t symbol_byte_size = 0; 2838 bool add_nlist = true; 2839 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2840 bool demangled_is_synthesized = false; 2841 bool is_gsym = false; 2842 bool set_value = true; 2843 2844 assert(sym_idx < num_syms); 2845 2846 sym[sym_idx].SetDebug(is_debug); 2847 2848 if (is_debug) { 2849 switch (nlist.n_type) { 2850 case N_GSYM: 2851 // global symbol: name,,NO_SECT,type,0 2852 // Sometimes the N_GSYM value contains the address. 2853 2854 // FIXME: In the .o files, we have a GSYM and a debug 2855 // symbol for all the ObjC data. They 2856 // have the same address, but we want to ensure that 2857 // we always find only the real symbol, 'cause we 2858 // don't currently correctly attribute the 2859 // GSYM one to the ObjCClass/Ivar/MetaClass 2860 // symbol type. This is a temporary hack to make 2861 // sure the ObjectiveC symbols get treated correctly. 2862 // To do this right, we should coalesce all the GSYM 2863 // & global symbols that have the same address. 2864 2865 is_gsym = true; 2866 sym[sym_idx].SetExternal(true); 2867 2868 if (symbol_name && symbol_name[0] == '_' && 2869 symbol_name[1] == 'O') { 2870 llvm::StringRef symbol_name_ref(symbol_name); 2871 if (symbol_name_ref.startswith( 2872 g_objc_v2_prefix_class)) { 2873 symbol_name_non_abi_mangled = symbol_name + 1; 2874 symbol_name = 2875 symbol_name + g_objc_v2_prefix_class.size(); 2876 type = eSymbolTypeObjCClass; 2877 demangled_is_synthesized = true; 2878 2879 } else if (symbol_name_ref.startswith( 2880 g_objc_v2_prefix_metaclass)) { 2881 symbol_name_non_abi_mangled = symbol_name + 1; 2882 symbol_name = symbol_name + 2883 g_objc_v2_prefix_metaclass.size(); 2884 type = eSymbolTypeObjCMetaClass; 2885 demangled_is_synthesized = true; 2886 } else if (symbol_name_ref.startswith( 2887 g_objc_v2_prefix_ivar)) { 2888 symbol_name_non_abi_mangled = symbol_name + 1; 2889 symbol_name = 2890 symbol_name + g_objc_v2_prefix_ivar.size(); 2891 type = eSymbolTypeObjCIVar; 2892 demangled_is_synthesized = true; 2893 } 2894 } else { 2895 if (nlist.n_value != 0) 2896 symbol_section = section_info.GetSection( 2897 nlist.n_sect, nlist.n_value); 2898 type = eSymbolTypeData; 2899 } 2900 break; 2901 2902 case N_FNAME: 2903 // procedure name (f77 kludge): name,,NO_SECT,0,0 2904 type = eSymbolTypeCompiler; 2905 break; 2906 2907 case N_FUN: 2908 // procedure: name,,n_sect,linenumber,address 2909 if (symbol_name) { 2910 type = eSymbolTypeCode; 2911 symbol_section = section_info.GetSection( 2912 nlist.n_sect, nlist.n_value); 2913 2914 N_FUN_addr_to_sym_idx.insert( 2915 std::make_pair(nlist.n_value, sym_idx)); 2916 // We use the current number of symbols in the 2917 // symbol table in lieu of using nlist_idx in case 2918 // we ever start trimming entries out 2919 N_FUN_indexes.push_back(sym_idx); 2920 } else { 2921 type = eSymbolTypeCompiler; 2922 2923 if (!N_FUN_indexes.empty()) { 2924 // Copy the size of the function into the 2925 // original 2926 // STAB entry so we don't have 2927 // to hunt for it later 2928 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2929 ->SetByteSize(nlist.n_value); 2930 N_FUN_indexes.pop_back(); 2931 // We don't really need the end function STAB as 2932 // it contains the size which we already placed 2933 // with the original symbol, so don't add it if 2934 // we want a minimal symbol table 2935 add_nlist = false; 2936 } 2937 } 2938 break; 2939 2940 case N_STSYM: 2941 // static symbol: name,,n_sect,type,address 2942 N_STSYM_addr_to_sym_idx.insert( 2943 std::make_pair(nlist.n_value, sym_idx)); 2944 symbol_section = section_info.GetSection( 2945 nlist.n_sect, nlist.n_value); 2946 if (symbol_name && symbol_name[0]) { 2947 type = ObjectFile::GetSymbolTypeFromName( 2948 symbol_name + 1, eSymbolTypeData); 2949 } 2950 break; 2951 2952 case N_LCSYM: 2953 // .lcomm symbol: name,,n_sect,type,address 2954 symbol_section = section_info.GetSection( 2955 nlist.n_sect, nlist.n_value); 2956 type = eSymbolTypeCommonBlock; 2957 break; 2958 2959 case N_BNSYM: 2960 // We use the current number of symbols in the symbol 2961 // table in lieu of using nlist_idx in case we ever 2962 // start trimming entries out Skip these if we want 2963 // minimal symbol tables 2964 add_nlist = false; 2965 break; 2966 2967 case N_ENSYM: 2968 // Set the size of the N_BNSYM to the terminating 2969 // index of this N_ENSYM so that we can always skip 2970 // the entire symbol if we need to navigate more 2971 // quickly at the source level when parsing STABS 2972 // Skip these if we want minimal symbol tables 2973 add_nlist = false; 2974 break; 2975 2976 case N_OPT: 2977 // emitted with gcc2_compiled and in gcc source 2978 type = eSymbolTypeCompiler; 2979 break; 2980 2981 case N_RSYM: 2982 // register sym: name,,NO_SECT,type,register 2983 type = eSymbolTypeVariable; 2984 break; 2985 2986 case N_SLINE: 2987 // src line: 0,,n_sect,linenumber,address 2988 symbol_section = section_info.GetSection( 2989 nlist.n_sect, nlist.n_value); 2990 type = eSymbolTypeLineEntry; 2991 break; 2992 2993 case N_SSYM: 2994 // structure elt: name,,NO_SECT,type,struct_offset 2995 type = eSymbolTypeVariableType; 2996 break; 2997 2998 case N_SO: 2999 // source file name 3000 type = eSymbolTypeSourceFile; 3001 if (symbol_name == NULL) { 3002 add_nlist = false; 3003 if (N_SO_index != UINT32_MAX) { 3004 // Set the size of the N_SO to the terminating 3005 // index of this N_SO so that we can always skip 3006 // the entire N_SO if we need to navigate more 3007 // quickly at the source level when parsing STABS 3008 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3009 symbol_ptr->SetByteSize(sym_idx); 3010 symbol_ptr->SetSizeIsSibling(true); 3011 } 3012 N_NSYM_indexes.clear(); 3013 N_INCL_indexes.clear(); 3014 N_BRAC_indexes.clear(); 3015 N_COMM_indexes.clear(); 3016 N_FUN_indexes.clear(); 3017 N_SO_index = UINT32_MAX; 3018 } else { 3019 // We use the current number of symbols in the 3020 // symbol table in lieu of using nlist_idx in case 3021 // we ever start trimming entries out 3022 const bool N_SO_has_full_path = 3023 symbol_name[0] == '/'; 3024 if (N_SO_has_full_path) { 3025 if ((N_SO_index == sym_idx - 1) && 3026 ((sym_idx - 1) < num_syms)) { 3027 // We have two consecutive N_SO entries where 3028 // the first contains a directory and the 3029 // second contains a full path. 3030 sym[sym_idx - 1].GetMangled().SetValue( 3031 ConstString(symbol_name), false); 3032 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3033 add_nlist = false; 3034 } else { 3035 // This is the first entry in a N_SO that 3036 // contains a directory or 3037 // a full path to the source file 3038 N_SO_index = sym_idx; 3039 } 3040 } else if ((N_SO_index == sym_idx - 1) && 3041 ((sym_idx - 1) < num_syms)) { 3042 // This is usually the second N_SO entry that 3043 // contains just the filename, so here we combine 3044 // it with the first one if we are minimizing the 3045 // symbol table 3046 const char *so_path = 3047 sym[sym_idx - 1] 3048 .GetMangled() 3049 .GetDemangledName( 3050 lldb::eLanguageTypeUnknown) 3051 .AsCString(); 3052 if (so_path && so_path[0]) { 3053 std::string full_so_path(so_path); 3054 const size_t double_slash_pos = 3055 full_so_path.find("//"); 3056 if (double_slash_pos != std::string::npos) { 3057 // The linker has been generating bad N_SO 3058 // entries with doubled up paths 3059 // in the format "%s%s" where the first 3060 // string in the DW_AT_comp_dir, and the 3061 // second is the directory for the source 3062 // file so you end up with a path that looks 3063 // like "/tmp/src//tmp/src/" 3064 FileSpec so_dir(so_path, false); 3065 if (!so_dir.Exists()) { 3066 so_dir.SetFile( 3067 &full_so_path[double_slash_pos + 1], 3068 false); 3069 if (so_dir.Exists()) { 3070 // Trim off the incorrect path 3071 full_so_path.erase(0, 3072 double_slash_pos + 1); 3073 } 3074 } 3075 } 3076 if (*full_so_path.rbegin() != '/') 3077 full_so_path += '/'; 3078 full_so_path += symbol_name; 3079 sym[sym_idx - 1].GetMangled().SetValue( 3080 ConstString(full_so_path.c_str()), false); 3081 add_nlist = false; 3082 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3083 } 3084 } else { 3085 // This could be a relative path to a N_SO 3086 N_SO_index = sym_idx; 3087 } 3088 } 3089 break; 3090 3091 case N_OSO: 3092 // object file name: name,,0,0,st_mtime 3093 type = eSymbolTypeObjectFile; 3094 break; 3095 3096 case N_LSYM: 3097 // local sym: name,,NO_SECT,type,offset 3098 type = eSymbolTypeLocal; 3099 break; 3100 3101 //---------------------------------------------------------------------- 3102 // INCL scopes 3103 //---------------------------------------------------------------------- 3104 case N_BINCL: 3105 // include file beginning: name,,NO_SECT,0,sum We use 3106 // the current number of symbols in the symbol table 3107 // in lieu of using nlist_idx in case we ever start 3108 // trimming entries out 3109 N_INCL_indexes.push_back(sym_idx); 3110 type = eSymbolTypeScopeBegin; 3111 break; 3112 3113 case N_EINCL: 3114 // include file end: name,,NO_SECT,0,0 3115 // Set the size of the N_BINCL to the terminating 3116 // index of this N_EINCL so that we can always skip 3117 // the entire symbol if we need to navigate more 3118 // quickly at the source level when parsing STABS 3119 if (!N_INCL_indexes.empty()) { 3120 symbol_ptr = 3121 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3122 symbol_ptr->SetByteSize(sym_idx + 1); 3123 symbol_ptr->SetSizeIsSibling(true); 3124 N_INCL_indexes.pop_back(); 3125 } 3126 type = eSymbolTypeScopeEnd; 3127 break; 3128 3129 case N_SOL: 3130 // #included file name: name,,n_sect,0,address 3131 type = eSymbolTypeHeaderFile; 3132 3133 // We currently don't use the header files on darwin 3134 add_nlist = false; 3135 break; 3136 3137 case N_PARAMS: 3138 // compiler parameters: name,,NO_SECT,0,0 3139 type = eSymbolTypeCompiler; 3140 break; 3141 3142 case N_VERSION: 3143 // compiler version: name,,NO_SECT,0,0 3144 type = eSymbolTypeCompiler; 3145 break; 3146 3147 case N_OLEVEL: 3148 // compiler -O level: name,,NO_SECT,0,0 3149 type = eSymbolTypeCompiler; 3150 break; 3151 3152 case N_PSYM: 3153 // parameter: name,,NO_SECT,type,offset 3154 type = eSymbolTypeVariable; 3155 break; 3156 3157 case N_ENTRY: 3158 // alternate entry: name,,n_sect,linenumber,address 3159 symbol_section = section_info.GetSection( 3160 nlist.n_sect, nlist.n_value); 3161 type = eSymbolTypeLineEntry; 3162 break; 3163 3164 //---------------------------------------------------------------------- 3165 // Left and Right Braces 3166 //---------------------------------------------------------------------- 3167 case N_LBRAC: 3168 // left bracket: 0,,NO_SECT,nesting level,address We 3169 // use the current number of symbols in the symbol 3170 // table in lieu of using nlist_idx in case we ever 3171 // start trimming entries out 3172 symbol_section = section_info.GetSection( 3173 nlist.n_sect, nlist.n_value); 3174 N_BRAC_indexes.push_back(sym_idx); 3175 type = eSymbolTypeScopeBegin; 3176 break; 3177 3178 case N_RBRAC: 3179 // right bracket: 0,,NO_SECT,nesting level,address 3180 // Set the size of the N_LBRAC to the terminating 3181 // index of this N_RBRAC so that we can always skip 3182 // the entire symbol if we need to navigate more 3183 // quickly at the source level when parsing STABS 3184 symbol_section = section_info.GetSection( 3185 nlist.n_sect, nlist.n_value); 3186 if (!N_BRAC_indexes.empty()) { 3187 symbol_ptr = 3188 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3189 symbol_ptr->SetByteSize(sym_idx + 1); 3190 symbol_ptr->SetSizeIsSibling(true); 3191 N_BRAC_indexes.pop_back(); 3192 } 3193 type = eSymbolTypeScopeEnd; 3194 break; 3195 3196 case N_EXCL: 3197 // deleted include file: name,,NO_SECT,0,sum 3198 type = eSymbolTypeHeaderFile; 3199 break; 3200 3201 //---------------------------------------------------------------------- 3202 // COMM scopes 3203 //---------------------------------------------------------------------- 3204 case N_BCOMM: 3205 // begin common: name,,NO_SECT,0,0 3206 // We use the current number of symbols in the symbol 3207 // table in lieu of using nlist_idx in case we ever 3208 // start trimming entries out 3209 type = eSymbolTypeScopeBegin; 3210 N_COMM_indexes.push_back(sym_idx); 3211 break; 3212 3213 case N_ECOML: 3214 // end common (local name): 0,,n_sect,0,address 3215 symbol_section = section_info.GetSection( 3216 nlist.n_sect, nlist.n_value); 3217 // Fall through 3218 3219 case N_ECOMM: 3220 // end common: name,,n_sect,0,0 3221 // Set the size of the N_BCOMM to the terminating 3222 // index of this N_ECOMM/N_ECOML so that we can 3223 // always skip the entire symbol if we need to 3224 // navigate more quickly at the source level when 3225 // parsing STABS 3226 if (!N_COMM_indexes.empty()) { 3227 symbol_ptr = 3228 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3229 symbol_ptr->SetByteSize(sym_idx + 1); 3230 symbol_ptr->SetSizeIsSibling(true); 3231 N_COMM_indexes.pop_back(); 3232 } 3233 type = eSymbolTypeScopeEnd; 3234 break; 3235 3236 case N_LENG: 3237 // second stab entry with length information 3238 type = eSymbolTypeAdditional; 3239 break; 3240 3241 default: 3242 break; 3243 } 3244 } else { 3245 // uint8_t n_pext = N_PEXT & nlist.n_type; 3246 uint8_t n_type = N_TYPE & nlist.n_type; 3247 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3248 3249 switch (n_type) { 3250 case N_INDR: { 3251 const char *reexport_name_cstr = 3252 strtab_data.PeekCStr(nlist.n_value); 3253 if (reexport_name_cstr && reexport_name_cstr[0]) { 3254 type = eSymbolTypeReExported; 3255 ConstString reexport_name( 3256 reexport_name_cstr + 3257 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3258 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3259 set_value = false; 3260 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3261 indirect_symbol_names.insert( 3262 ConstString(symbol_name + 3263 ((symbol_name[0] == '_') ? 1 : 0))); 3264 } else 3265 type = eSymbolTypeUndefined; 3266 } break; 3267 3268 case N_UNDF: 3269 if (symbol_name && symbol_name[0]) { 3270 ConstString undefined_name( 3271 symbol_name + 3272 ((symbol_name[0] == '_') ? 1 : 0)); 3273 undefined_name_to_desc[undefined_name] = 3274 nlist.n_desc; 3275 } 3276 // Fall through 3277 case N_PBUD: 3278 type = eSymbolTypeUndefined; 3279 break; 3280 3281 case N_ABS: 3282 type = eSymbolTypeAbsolute; 3283 break; 3284 3285 case N_SECT: { 3286 symbol_section = section_info.GetSection( 3287 nlist.n_sect, nlist.n_value); 3288 3289 if (symbol_section == NULL) { 3290 // TODO: warn about this? 3291 add_nlist = false; 3292 break; 3293 } 3294 3295 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3296 type = eSymbolTypeException; 3297 } else { 3298 uint32_t section_type = 3299 symbol_section->Get() & SECTION_TYPE; 3300 3301 switch (section_type) { 3302 case S_CSTRING_LITERALS: 3303 type = eSymbolTypeData; 3304 break; // section with only literal C strings 3305 case S_4BYTE_LITERALS: 3306 type = eSymbolTypeData; 3307 break; // section with only 4 byte literals 3308 case S_8BYTE_LITERALS: 3309 type = eSymbolTypeData; 3310 break; // section with only 8 byte literals 3311 case S_LITERAL_POINTERS: 3312 type = eSymbolTypeTrampoline; 3313 break; // section with only pointers to literals 3314 case S_NON_LAZY_SYMBOL_POINTERS: 3315 type = eSymbolTypeTrampoline; 3316 break; // section with only non-lazy symbol 3317 // pointers 3318 case S_LAZY_SYMBOL_POINTERS: 3319 type = eSymbolTypeTrampoline; 3320 break; // section with only lazy symbol pointers 3321 case S_SYMBOL_STUBS: 3322 type = eSymbolTypeTrampoline; 3323 break; // section with only symbol stubs, byte 3324 // size of stub in the reserved2 field 3325 case S_MOD_INIT_FUNC_POINTERS: 3326 type = eSymbolTypeCode; 3327 break; // section with only function pointers for 3328 // initialization 3329 case S_MOD_TERM_FUNC_POINTERS: 3330 type = eSymbolTypeCode; 3331 break; // section with only function pointers for 3332 // termination 3333 case S_INTERPOSING: 3334 type = eSymbolTypeTrampoline; 3335 break; // section with only pairs of function 3336 // pointers for interposing 3337 case S_16BYTE_LITERALS: 3338 type = eSymbolTypeData; 3339 break; // section with only 16 byte literals 3340 case S_DTRACE_DOF: 3341 type = eSymbolTypeInstrumentation; 3342 break; 3343 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3344 type = eSymbolTypeTrampoline; 3345 break; 3346 default: 3347 switch (symbol_section->GetType()) { 3348 case lldb::eSectionTypeCode: 3349 type = eSymbolTypeCode; 3350 break; 3351 case eSectionTypeData: 3352 case eSectionTypeDataCString: // Inlined C string 3353 // data 3354 case eSectionTypeDataCStringPointers: // Pointers 3355 // to C 3356 // string 3357 // data 3358 case eSectionTypeDataSymbolAddress: // Address of 3359 // a symbol in 3360 // the symbol 3361 // table 3362 case eSectionTypeData4: 3363 case eSectionTypeData8: 3364 case eSectionTypeData16: 3365 type = eSymbolTypeData; 3366 break; 3367 default: 3368 break; 3369 } 3370 break; 3371 } 3372 3373 if (type == eSymbolTypeInvalid) { 3374 const char *symbol_sect_name = 3375 symbol_section->GetName().AsCString(); 3376 if (symbol_section->IsDescendant( 3377 text_section_sp.get())) { 3378 if (symbol_section->IsClear( 3379 S_ATTR_PURE_INSTRUCTIONS | 3380 S_ATTR_SELF_MODIFYING_CODE | 3381 S_ATTR_SOME_INSTRUCTIONS)) 3382 type = eSymbolTypeData; 3383 else 3384 type = eSymbolTypeCode; 3385 } else if (symbol_section->IsDescendant( 3386 data_section_sp.get()) || 3387 symbol_section->IsDescendant( 3388 data_dirty_section_sp.get()) || 3389 symbol_section->IsDescendant( 3390 data_const_section_sp.get())) { 3391 if (symbol_sect_name && 3392 ::strstr(symbol_sect_name, "__objc") == 3393 symbol_sect_name) { 3394 type = eSymbolTypeRuntime; 3395 3396 if (symbol_name) { 3397 llvm::StringRef symbol_name_ref( 3398 symbol_name); 3399 if (symbol_name_ref.startswith("_OBJC_")) { 3400 static const llvm::StringRef 3401 g_objc_v2_prefix_class( 3402 "_OBJC_CLASS_$_"); 3403 static const llvm::StringRef 3404 g_objc_v2_prefix_metaclass( 3405 "_OBJC_METACLASS_$_"); 3406 static const llvm::StringRef 3407 g_objc_v2_prefix_ivar( 3408 "_OBJC_IVAR_$_"); 3409 if (symbol_name_ref.startswith( 3410 g_objc_v2_prefix_class)) { 3411 symbol_name_non_abi_mangled = 3412 symbol_name + 1; 3413 symbol_name = 3414 symbol_name + 3415 g_objc_v2_prefix_class.size(); 3416 type = eSymbolTypeObjCClass; 3417 demangled_is_synthesized = true; 3418 } else if ( 3419 symbol_name_ref.startswith( 3420 g_objc_v2_prefix_metaclass)) { 3421 symbol_name_non_abi_mangled = 3422 symbol_name + 1; 3423 symbol_name = 3424 symbol_name + 3425 g_objc_v2_prefix_metaclass.size(); 3426 type = eSymbolTypeObjCMetaClass; 3427 demangled_is_synthesized = true; 3428 } else if (symbol_name_ref.startswith( 3429 g_objc_v2_prefix_ivar)) { 3430 symbol_name_non_abi_mangled = 3431 symbol_name + 1; 3432 symbol_name = 3433 symbol_name + 3434 g_objc_v2_prefix_ivar.size(); 3435 type = eSymbolTypeObjCIVar; 3436 demangled_is_synthesized = true; 3437 } 3438 } 3439 } 3440 } else if (symbol_sect_name && 3441 ::strstr(symbol_sect_name, 3442 "__gcc_except_tab") == 3443 symbol_sect_name) { 3444 type = eSymbolTypeException; 3445 } else { 3446 type = eSymbolTypeData; 3447 } 3448 } else if (symbol_sect_name && 3449 ::strstr(symbol_sect_name, 3450 "__IMPORT") == 3451 symbol_sect_name) { 3452 type = eSymbolTypeTrampoline; 3453 } else if (symbol_section->IsDescendant( 3454 objc_section_sp.get())) { 3455 type = eSymbolTypeRuntime; 3456 if (symbol_name && symbol_name[0] == '.') { 3457 llvm::StringRef symbol_name_ref(symbol_name); 3458 static const llvm::StringRef 3459 g_objc_v1_prefix_class( 3460 ".objc_class_name_"); 3461 if (symbol_name_ref.startswith( 3462 g_objc_v1_prefix_class)) { 3463 symbol_name_non_abi_mangled = symbol_name; 3464 symbol_name = symbol_name + 3465 g_objc_v1_prefix_class.size(); 3466 type = eSymbolTypeObjCClass; 3467 demangled_is_synthesized = true; 3468 } 3469 } 3470 } 3471 } 3472 } 3473 } break; 3474 } 3475 } 3476 3477 if (add_nlist) { 3478 uint64_t symbol_value = nlist.n_value; 3479 if (symbol_name_non_abi_mangled) { 3480 sym[sym_idx].GetMangled().SetMangledName( 3481 ConstString(symbol_name_non_abi_mangled)); 3482 sym[sym_idx].GetMangled().SetDemangledName( 3483 ConstString(symbol_name)); 3484 } else { 3485 bool symbol_name_is_mangled = false; 3486 3487 if (symbol_name && symbol_name[0] == '_') { 3488 symbol_name_is_mangled = symbol_name[1] == '_'; 3489 symbol_name++; // Skip the leading underscore 3490 } 3491 3492 if (symbol_name) { 3493 ConstString const_symbol_name(symbol_name); 3494 sym[sym_idx].GetMangled().SetValue( 3495 const_symbol_name, symbol_name_is_mangled); 3496 if (is_gsym && is_debug) { 3497 const char *gsym_name = 3498 sym[sym_idx] 3499 .GetMangled() 3500 .GetName(lldb::eLanguageTypeUnknown, 3501 Mangled::ePreferMangled) 3502 .GetCString(); 3503 if (gsym_name) 3504 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3505 } 3506 } 3507 } 3508 if (symbol_section) { 3509 const addr_t section_file_addr = 3510 symbol_section->GetFileAddress(); 3511 if (symbol_byte_size == 0 && 3512 function_starts_count > 0) { 3513 addr_t symbol_lookup_file_addr = nlist.n_value; 3514 // Do an exact address match for non-ARM addresses, 3515 // else get the closest since the symbol might be a 3516 // thumb symbol which has an address with bit zero 3517 // set 3518 FunctionStarts::Entry *func_start_entry = 3519 function_starts.FindEntry( 3520 symbol_lookup_file_addr, !is_arm); 3521 if (is_arm && func_start_entry) { 3522 // Verify that the function start address is the 3523 // symbol address (ARM) or the symbol address + 1 3524 // (thumb) 3525 if (func_start_entry->addr != 3526 symbol_lookup_file_addr && 3527 func_start_entry->addr != 3528 (symbol_lookup_file_addr + 1)) { 3529 // Not the right entry, NULL it out... 3530 func_start_entry = NULL; 3531 } 3532 } 3533 if (func_start_entry) { 3534 func_start_entry->data = true; 3535 3536 addr_t symbol_file_addr = func_start_entry->addr; 3537 uint32_t symbol_flags = 0; 3538 if (is_arm) { 3539 if (symbol_file_addr & 1) 3540 symbol_flags = 3541 MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3542 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3543 } 3544 3545 const FunctionStarts::Entry 3546 *next_func_start_entry = 3547 function_starts.FindNextEntry( 3548 func_start_entry); 3549 const addr_t section_end_file_addr = 3550 section_file_addr + 3551 symbol_section->GetByteSize(); 3552 if (next_func_start_entry) { 3553 addr_t next_symbol_file_addr = 3554 next_func_start_entry->addr; 3555 // Be sure the clear the Thumb address bit when 3556 // we calculate the size from the current and 3557 // next address 3558 if (is_arm) 3559 next_symbol_file_addr &= 3560 THUMB_ADDRESS_BIT_MASK; 3561 symbol_byte_size = std::min<lldb::addr_t>( 3562 next_symbol_file_addr - symbol_file_addr, 3563 section_end_file_addr - symbol_file_addr); 3564 } else { 3565 symbol_byte_size = 3566 section_end_file_addr - symbol_file_addr; 3567 } 3568 } 3569 } 3570 symbol_value -= section_file_addr; 3571 } 3572 3573 if (is_debug == false) { 3574 if (type == eSymbolTypeCode) { 3575 // See if we can find a N_FUN entry for any code 3576 // symbols. If we do find a match, and the name 3577 // matches, then we can merge the two into just the 3578 // function symbol to avoid duplicate entries in 3579 // the symbol table 3580 std::pair<ValueToSymbolIndexMap::const_iterator, 3581 ValueToSymbolIndexMap::const_iterator> 3582 range; 3583 range = N_FUN_addr_to_sym_idx.equal_range( 3584 nlist.n_value); 3585 if (range.first != range.second) { 3586 bool found_it = false; 3587 for (ValueToSymbolIndexMap::const_iterator pos = 3588 range.first; 3589 pos != range.second; ++pos) { 3590 if (sym[sym_idx].GetMangled().GetName( 3591 lldb::eLanguageTypeUnknown, 3592 Mangled::ePreferMangled) == 3593 sym[pos->second].GetMangled().GetName( 3594 lldb::eLanguageTypeUnknown, 3595 Mangled::ePreferMangled)) { 3596 m_nlist_idx_to_sym_idx[nlist_idx] = 3597 pos->second; 3598 // We just need the flags from the linker 3599 // symbol, so put these flags 3600 // into the N_FUN flags to avoid duplicate 3601 // symbols in the symbol table 3602 sym[pos->second].SetExternal( 3603 sym[sym_idx].IsExternal()); 3604 sym[pos->second].SetFlags(nlist.n_type << 16 | 3605 nlist.n_desc); 3606 if (resolver_addresses.find(nlist.n_value) != 3607 resolver_addresses.end()) 3608 sym[pos->second].SetType( 3609 eSymbolTypeResolver); 3610 sym[sym_idx].Clear(); 3611 found_it = true; 3612 break; 3613 } 3614 } 3615 if (found_it) 3616 continue; 3617 } else { 3618 if (resolver_addresses.find(nlist.n_value) != 3619 resolver_addresses.end()) 3620 type = eSymbolTypeResolver; 3621 } 3622 } else if (type == eSymbolTypeData || 3623 type == eSymbolTypeObjCClass || 3624 type == eSymbolTypeObjCMetaClass || 3625 type == eSymbolTypeObjCIVar) { 3626 // See if we can find a N_STSYM entry for any data 3627 // symbols. If we do find a match, and the name 3628 // matches, then we can merge the two into just the 3629 // Static symbol to avoid duplicate entries in the 3630 // symbol table 3631 std::pair<ValueToSymbolIndexMap::const_iterator, 3632 ValueToSymbolIndexMap::const_iterator> 3633 range; 3634 range = N_STSYM_addr_to_sym_idx.equal_range( 3635 nlist.n_value); 3636 if (range.first != range.second) { 3637 bool found_it = false; 3638 for (ValueToSymbolIndexMap::const_iterator pos = 3639 range.first; 3640 pos != range.second; ++pos) { 3641 if (sym[sym_idx].GetMangled().GetName( 3642 lldb::eLanguageTypeUnknown, 3643 Mangled::ePreferMangled) == 3644 sym[pos->second].GetMangled().GetName( 3645 lldb::eLanguageTypeUnknown, 3646 Mangled::ePreferMangled)) { 3647 m_nlist_idx_to_sym_idx[nlist_idx] = 3648 pos->second; 3649 // We just need the flags from the linker 3650 // symbol, so put these flags 3651 // into the N_STSYM flags to avoid duplicate 3652 // symbols in the symbol table 3653 sym[pos->second].SetExternal( 3654 sym[sym_idx].IsExternal()); 3655 sym[pos->second].SetFlags(nlist.n_type << 16 | 3656 nlist.n_desc); 3657 sym[sym_idx].Clear(); 3658 found_it = true; 3659 break; 3660 } 3661 } 3662 if (found_it) 3663 continue; 3664 } else { 3665 const char *gsym_name = 3666 sym[sym_idx] 3667 .GetMangled() 3668 .GetName(lldb::eLanguageTypeUnknown, 3669 Mangled::ePreferMangled) 3670 .GetCString(); 3671 if (gsym_name) { 3672 // Combine N_GSYM stab entries with the non 3673 // stab symbol 3674 ConstNameToSymbolIndexMap::const_iterator pos = 3675 N_GSYM_name_to_sym_idx.find(gsym_name); 3676 if (pos != N_GSYM_name_to_sym_idx.end()) { 3677 const uint32_t GSYM_sym_idx = pos->second; 3678 m_nlist_idx_to_sym_idx[nlist_idx] = 3679 GSYM_sym_idx; 3680 // Copy the address, because often the N_GSYM 3681 // address has an invalid address of zero 3682 // when the global is a common symbol 3683 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3684 symbol_section); 3685 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3686 symbol_value); 3687 // We just need the flags from the linker 3688 // symbol, so put these flags 3689 // into the N_GSYM flags to avoid duplicate 3690 // symbols in the symbol table 3691 sym[GSYM_sym_idx].SetFlags( 3692 nlist.n_type << 16 | nlist.n_desc); 3693 sym[sym_idx].Clear(); 3694 continue; 3695 } 3696 } 3697 } 3698 } 3699 } 3700 3701 sym[sym_idx].SetID(nlist_idx); 3702 sym[sym_idx].SetType(type); 3703 if (set_value) { 3704 sym[sym_idx].GetAddressRef().SetSection( 3705 symbol_section); 3706 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3707 } 3708 sym[sym_idx].SetFlags(nlist.n_type << 16 | 3709 nlist.n_desc); 3710 3711 if (symbol_byte_size > 0) 3712 sym[sym_idx].SetByteSize(symbol_byte_size); 3713 3714 if (demangled_is_synthesized) 3715 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3716 ++sym_idx; 3717 } else { 3718 sym[sym_idx].Clear(); 3719 } 3720 } 3721 ///////////////////////////// 3722 } 3723 break; // No more entries to consider 3724 } 3725 } 3726 3727 for (const auto &pos : reexport_shlib_needs_fixup) { 3728 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3729 if (undef_pos != undefined_name_to_desc.end()) { 3730 const uint8_t dylib_ordinal = 3731 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3732 if (dylib_ordinal > 0 && 3733 dylib_ordinal < dylib_files.GetSize()) 3734 sym[pos.first].SetReExportedSymbolSharedLibrary( 3735 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3736 } 3737 } 3738 } 3739 } 3740 } 3741 } 3742 } 3743 3744 // Must reset this in case it was mutated above! 3745 nlist_data_offset = 0; 3746 #endif 3747 3748 if (nlist_data.GetByteSize() > 0) { 3749 3750 // If the sym array was not created while parsing the DSC unmapped 3751 // symbols, create it now. 3752 if (sym == NULL) { 3753 sym = symtab->Resize(symtab_load_command.nsyms + 3754 m_dysymtab.nindirectsyms); 3755 num_syms = symtab->GetNumSymbols(); 3756 } 3757 3758 if (unmapped_local_symbols_found) { 3759 assert(m_dysymtab.ilocalsym == 0); 3760 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3761 nlist_idx = m_dysymtab.nlocalsym; 3762 } else { 3763 nlist_idx = 0; 3764 } 3765 3766 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3767 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3768 UndefinedNameToDescMap undefined_name_to_desc; 3769 SymbolIndexToName reexport_shlib_needs_fixup; 3770 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 3771 struct nlist_64 nlist; 3772 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, 3773 nlist_byte_size)) 3774 break; 3775 3776 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3777 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 3778 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 3779 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 3780 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 3781 3782 SymbolType type = eSymbolTypeInvalid; 3783 const char *symbol_name = NULL; 3784 3785 if (have_strtab_data) { 3786 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3787 3788 if (symbol_name == NULL) { 3789 // No symbol should be NULL, even the symbols with no string values 3790 // should have an offset zero which points to an empty C-string 3791 Host::SystemLog(Host::eSystemLogError, 3792 "error: symbol[%u] has invalid string table offset " 3793 "0x%x in %s, ignoring symbol\n", 3794 nlist_idx, nlist.n_strx, 3795 module_sp->GetFileSpec().GetPath().c_str()); 3796 continue; 3797 } 3798 if (symbol_name[0] == '\0') 3799 symbol_name = NULL; 3800 } else { 3801 const addr_t str_addr = strtab_addr + nlist.n_strx; 3802 Status str_error; 3803 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3804 str_error)) 3805 symbol_name = memory_symbol_name.c_str(); 3806 } 3807 const char *symbol_name_non_abi_mangled = NULL; 3808 3809 SectionSP symbol_section; 3810 lldb::addr_t symbol_byte_size = 0; 3811 bool add_nlist = true; 3812 bool is_gsym = false; 3813 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3814 bool demangled_is_synthesized = false; 3815 bool set_value = true; 3816 assert(sym_idx < num_syms); 3817 3818 sym[sym_idx].SetDebug(is_debug); 3819 3820 if (is_debug) { 3821 switch (nlist.n_type) { 3822 case N_GSYM: 3823 // global symbol: name,,NO_SECT,type,0 3824 // Sometimes the N_GSYM value contains the address. 3825 3826 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3827 // the ObjC data. They 3828 // have the same address, but we want to ensure that we always find 3829 // only the real symbol, 'cause we don't currently correctly 3830 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3831 // type. This is a temporary hack to make sure the ObjectiveC 3832 // symbols get treated correctly. To do this right, we should 3833 // coalesce all the GSYM & global symbols that have the same 3834 // address. 3835 is_gsym = true; 3836 sym[sym_idx].SetExternal(true); 3837 3838 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3839 llvm::StringRef symbol_name_ref(symbol_name); 3840 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3841 symbol_name_non_abi_mangled = symbol_name + 1; 3842 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3843 type = eSymbolTypeObjCClass; 3844 demangled_is_synthesized = true; 3845 3846 } else if (symbol_name_ref.startswith( 3847 g_objc_v2_prefix_metaclass)) { 3848 symbol_name_non_abi_mangled = symbol_name + 1; 3849 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3850 type = eSymbolTypeObjCMetaClass; 3851 demangled_is_synthesized = true; 3852 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3853 symbol_name_non_abi_mangled = symbol_name + 1; 3854 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3855 type = eSymbolTypeObjCIVar; 3856 demangled_is_synthesized = true; 3857 } 3858 } else { 3859 if (nlist.n_value != 0) 3860 symbol_section = 3861 section_info.GetSection(nlist.n_sect, nlist.n_value); 3862 type = eSymbolTypeData; 3863 } 3864 break; 3865 3866 case N_FNAME: 3867 // procedure name (f77 kludge): name,,NO_SECT,0,0 3868 type = eSymbolTypeCompiler; 3869 break; 3870 3871 case N_FUN: 3872 // procedure: name,,n_sect,linenumber,address 3873 if (symbol_name) { 3874 type = eSymbolTypeCode; 3875 symbol_section = 3876 section_info.GetSection(nlist.n_sect, nlist.n_value); 3877 3878 N_FUN_addr_to_sym_idx.insert( 3879 std::make_pair(nlist.n_value, sym_idx)); 3880 // We use the current number of symbols in the symbol table in 3881 // lieu of using nlist_idx in case we ever start trimming entries 3882 // out 3883 N_FUN_indexes.push_back(sym_idx); 3884 } else { 3885 type = eSymbolTypeCompiler; 3886 3887 if (!N_FUN_indexes.empty()) { 3888 // Copy the size of the function into the original STAB entry 3889 // so we don't have to hunt for it later 3890 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3891 ->SetByteSize(nlist.n_value); 3892 N_FUN_indexes.pop_back(); 3893 // We don't really need the end function STAB as it contains 3894 // the size which we already placed with the original symbol, 3895 // so don't add it if we want a minimal symbol table 3896 add_nlist = false; 3897 } 3898 } 3899 break; 3900 3901 case N_STSYM: 3902 // static symbol: name,,n_sect,type,address 3903 N_STSYM_addr_to_sym_idx.insert( 3904 std::make_pair(nlist.n_value, sym_idx)); 3905 symbol_section = 3906 section_info.GetSection(nlist.n_sect, nlist.n_value); 3907 if (symbol_name && symbol_name[0]) { 3908 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3909 eSymbolTypeData); 3910 } 3911 break; 3912 3913 case N_LCSYM: 3914 // .lcomm symbol: name,,n_sect,type,address 3915 symbol_section = 3916 section_info.GetSection(nlist.n_sect, nlist.n_value); 3917 type = eSymbolTypeCommonBlock; 3918 break; 3919 3920 case N_BNSYM: 3921 // We use the current number of symbols in the symbol table in lieu 3922 // of using nlist_idx in case we ever start trimming entries out 3923 // Skip these if we want minimal symbol tables 3924 add_nlist = false; 3925 break; 3926 3927 case N_ENSYM: 3928 // Set the size of the N_BNSYM to the terminating index of this 3929 // N_ENSYM so that we can always skip the entire symbol if we need 3930 // to navigate more quickly at the source level when parsing STABS 3931 // Skip these if we want minimal symbol tables 3932 add_nlist = false; 3933 break; 3934 3935 case N_OPT: 3936 // emitted with gcc2_compiled and in gcc source 3937 type = eSymbolTypeCompiler; 3938 break; 3939 3940 case N_RSYM: 3941 // register sym: name,,NO_SECT,type,register 3942 type = eSymbolTypeVariable; 3943 break; 3944 3945 case N_SLINE: 3946 // src line: 0,,n_sect,linenumber,address 3947 symbol_section = 3948 section_info.GetSection(nlist.n_sect, nlist.n_value); 3949 type = eSymbolTypeLineEntry; 3950 break; 3951 3952 case N_SSYM: 3953 // structure elt: name,,NO_SECT,type,struct_offset 3954 type = eSymbolTypeVariableType; 3955 break; 3956 3957 case N_SO: 3958 // source file name 3959 type = eSymbolTypeSourceFile; 3960 if (symbol_name == NULL) { 3961 add_nlist = false; 3962 if (N_SO_index != UINT32_MAX) { 3963 // Set the size of the N_SO to the terminating index of this 3964 // N_SO so that we can always skip the entire N_SO if we need 3965 // to navigate more quickly at the source level when parsing 3966 // STABS 3967 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3968 symbol_ptr->SetByteSize(sym_idx); 3969 symbol_ptr->SetSizeIsSibling(true); 3970 } 3971 N_NSYM_indexes.clear(); 3972 N_INCL_indexes.clear(); 3973 N_BRAC_indexes.clear(); 3974 N_COMM_indexes.clear(); 3975 N_FUN_indexes.clear(); 3976 N_SO_index = UINT32_MAX; 3977 } else { 3978 // We use the current number of symbols in the symbol table in 3979 // lieu of using nlist_idx in case we ever start trimming entries 3980 // out 3981 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3982 if (N_SO_has_full_path) { 3983 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3984 // We have two consecutive N_SO entries where the first 3985 // contains a directory and the second contains a full path. 3986 sym[sym_idx - 1].GetMangled().SetValue( 3987 ConstString(symbol_name), false); 3988 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3989 add_nlist = false; 3990 } else { 3991 // This is the first entry in a N_SO that contains a 3992 // directory or a full path to the source file 3993 N_SO_index = sym_idx; 3994 } 3995 } else if ((N_SO_index == sym_idx - 1) && 3996 ((sym_idx - 1) < num_syms)) { 3997 // This is usually the second N_SO entry that contains just the 3998 // filename, so here we combine it with the first one if we are 3999 // minimizing the symbol table 4000 const char *so_path = 4001 sym[sym_idx - 1] 4002 .GetMangled() 4003 .GetDemangledName(lldb::eLanguageTypeUnknown) 4004 .AsCString(); 4005 if (so_path && so_path[0]) { 4006 std::string full_so_path(so_path); 4007 const size_t double_slash_pos = full_so_path.find("//"); 4008 if (double_slash_pos != std::string::npos) { 4009 // The linker has been generating bad N_SO entries with 4010 // doubled up paths in the format "%s%s" where the first 4011 // string in the DW_AT_comp_dir, and the second is the 4012 // directory for the source file so you end up with a path 4013 // that looks like "/tmp/src//tmp/src/" 4014 FileSpec so_dir(so_path, false); 4015 if (!so_dir.Exists()) { 4016 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false, 4017 FileSpec::Style::native); 4018 if (so_dir.Exists()) { 4019 // Trim off the incorrect path 4020 full_so_path.erase(0, double_slash_pos + 1); 4021 } 4022 } 4023 } 4024 if (*full_so_path.rbegin() != '/') 4025 full_so_path += '/'; 4026 full_so_path += symbol_name; 4027 sym[sym_idx - 1].GetMangled().SetValue( 4028 ConstString(full_so_path.c_str()), false); 4029 add_nlist = false; 4030 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4031 } 4032 } else { 4033 // This could be a relative path to a N_SO 4034 N_SO_index = sym_idx; 4035 } 4036 } 4037 break; 4038 4039 case N_OSO: 4040 // object file name: name,,0,0,st_mtime 4041 type = eSymbolTypeObjectFile; 4042 break; 4043 4044 case N_LSYM: 4045 // local sym: name,,NO_SECT,type,offset 4046 type = eSymbolTypeLocal; 4047 break; 4048 4049 //---------------------------------------------------------------------- 4050 // INCL scopes 4051 //---------------------------------------------------------------------- 4052 case N_BINCL: 4053 // include file beginning: name,,NO_SECT,0,sum We use the current 4054 // number of symbols in the symbol table in lieu of using nlist_idx 4055 // in case we ever start trimming entries out 4056 N_INCL_indexes.push_back(sym_idx); 4057 type = eSymbolTypeScopeBegin; 4058 break; 4059 4060 case N_EINCL: 4061 // include file end: name,,NO_SECT,0,0 4062 // Set the size of the N_BINCL to the terminating index of this 4063 // N_EINCL so that we can always skip the entire symbol if we need 4064 // to navigate more quickly at the source level when parsing STABS 4065 if (!N_INCL_indexes.empty()) { 4066 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4067 symbol_ptr->SetByteSize(sym_idx + 1); 4068 symbol_ptr->SetSizeIsSibling(true); 4069 N_INCL_indexes.pop_back(); 4070 } 4071 type = eSymbolTypeScopeEnd; 4072 break; 4073 4074 case N_SOL: 4075 // #included file name: name,,n_sect,0,address 4076 type = eSymbolTypeHeaderFile; 4077 4078 // We currently don't use the header files on darwin 4079 add_nlist = false; 4080 break; 4081 4082 case N_PARAMS: 4083 // compiler parameters: name,,NO_SECT,0,0 4084 type = eSymbolTypeCompiler; 4085 break; 4086 4087 case N_VERSION: 4088 // compiler version: name,,NO_SECT,0,0 4089 type = eSymbolTypeCompiler; 4090 break; 4091 4092 case N_OLEVEL: 4093 // compiler -O level: name,,NO_SECT,0,0 4094 type = eSymbolTypeCompiler; 4095 break; 4096 4097 case N_PSYM: 4098 // parameter: name,,NO_SECT,type,offset 4099 type = eSymbolTypeVariable; 4100 break; 4101 4102 case N_ENTRY: 4103 // alternate entry: name,,n_sect,linenumber,address 4104 symbol_section = 4105 section_info.GetSection(nlist.n_sect, nlist.n_value); 4106 type = eSymbolTypeLineEntry; 4107 break; 4108 4109 //---------------------------------------------------------------------- 4110 // Left and Right Braces 4111 //---------------------------------------------------------------------- 4112 case N_LBRAC: 4113 // left bracket: 0,,NO_SECT,nesting level,address We use the 4114 // current number of symbols in the symbol table in lieu of using 4115 // nlist_idx in case we ever start trimming entries out 4116 symbol_section = 4117 section_info.GetSection(nlist.n_sect, nlist.n_value); 4118 N_BRAC_indexes.push_back(sym_idx); 4119 type = eSymbolTypeScopeBegin; 4120 break; 4121 4122 case N_RBRAC: 4123 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4124 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4125 // can always skip the entire symbol if we need to navigate more 4126 // quickly at the source level when parsing STABS 4127 symbol_section = 4128 section_info.GetSection(nlist.n_sect, nlist.n_value); 4129 if (!N_BRAC_indexes.empty()) { 4130 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4131 symbol_ptr->SetByteSize(sym_idx + 1); 4132 symbol_ptr->SetSizeIsSibling(true); 4133 N_BRAC_indexes.pop_back(); 4134 } 4135 type = eSymbolTypeScopeEnd; 4136 break; 4137 4138 case N_EXCL: 4139 // deleted include file: name,,NO_SECT,0,sum 4140 type = eSymbolTypeHeaderFile; 4141 break; 4142 4143 //---------------------------------------------------------------------- 4144 // COMM scopes 4145 //---------------------------------------------------------------------- 4146 case N_BCOMM: 4147 // begin common: name,,NO_SECT,0,0 4148 // We use the current number of symbols in the symbol table in lieu 4149 // of using nlist_idx in case we ever start trimming entries out 4150 type = eSymbolTypeScopeBegin; 4151 N_COMM_indexes.push_back(sym_idx); 4152 break; 4153 4154 case N_ECOML: 4155 // end common (local name): 0,,n_sect,0,address 4156 symbol_section = 4157 section_info.GetSection(nlist.n_sect, nlist.n_value); 4158 LLVM_FALLTHROUGH; 4159 4160 case N_ECOMM: 4161 // end common: name,,n_sect,0,0 4162 // Set the size of the N_BCOMM to the terminating index of this 4163 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4164 // we need to navigate more quickly at the source level when 4165 // parsing STABS 4166 if (!N_COMM_indexes.empty()) { 4167 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4168 symbol_ptr->SetByteSize(sym_idx + 1); 4169 symbol_ptr->SetSizeIsSibling(true); 4170 N_COMM_indexes.pop_back(); 4171 } 4172 type = eSymbolTypeScopeEnd; 4173 break; 4174 4175 case N_LENG: 4176 // second stab entry with length information 4177 type = eSymbolTypeAdditional; 4178 break; 4179 4180 default: 4181 break; 4182 } 4183 } else { 4184 // uint8_t n_pext = N_PEXT & nlist.n_type; 4185 uint8_t n_type = N_TYPE & nlist.n_type; 4186 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4187 4188 switch (n_type) { 4189 case N_INDR: { 4190 const char *reexport_name_cstr = 4191 strtab_data.PeekCStr(nlist.n_value); 4192 if (reexport_name_cstr && reexport_name_cstr[0]) { 4193 type = eSymbolTypeReExported; 4194 ConstString reexport_name( 4195 reexport_name_cstr + 4196 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4197 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4198 set_value = false; 4199 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4200 indirect_symbol_names.insert( 4201 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4202 } else 4203 type = eSymbolTypeUndefined; 4204 } break; 4205 4206 case N_UNDF: 4207 if (symbol_name && symbol_name[0]) { 4208 ConstString undefined_name(symbol_name + 4209 ((symbol_name[0] == '_') ? 1 : 0)); 4210 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4211 } 4212 LLVM_FALLTHROUGH; 4213 4214 case N_PBUD: 4215 type = eSymbolTypeUndefined; 4216 break; 4217 4218 case N_ABS: 4219 type = eSymbolTypeAbsolute; 4220 break; 4221 4222 case N_SECT: { 4223 symbol_section = 4224 section_info.GetSection(nlist.n_sect, nlist.n_value); 4225 4226 if (!symbol_section) { 4227 // TODO: warn about this? 4228 add_nlist = false; 4229 break; 4230 } 4231 4232 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4233 type = eSymbolTypeException; 4234 } else { 4235 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4236 4237 switch (section_type) { 4238 case S_CSTRING_LITERALS: 4239 type = eSymbolTypeData; 4240 break; // section with only literal C strings 4241 case S_4BYTE_LITERALS: 4242 type = eSymbolTypeData; 4243 break; // section with only 4 byte literals 4244 case S_8BYTE_LITERALS: 4245 type = eSymbolTypeData; 4246 break; // section with only 8 byte literals 4247 case S_LITERAL_POINTERS: 4248 type = eSymbolTypeTrampoline; 4249 break; // section with only pointers to literals 4250 case S_NON_LAZY_SYMBOL_POINTERS: 4251 type = eSymbolTypeTrampoline; 4252 break; // section with only non-lazy symbol pointers 4253 case S_LAZY_SYMBOL_POINTERS: 4254 type = eSymbolTypeTrampoline; 4255 break; // section with only lazy symbol pointers 4256 case S_SYMBOL_STUBS: 4257 type = eSymbolTypeTrampoline; 4258 break; // section with only symbol stubs, byte size of stub in 4259 // the reserved2 field 4260 case S_MOD_INIT_FUNC_POINTERS: 4261 type = eSymbolTypeCode; 4262 break; // section with only function pointers for initialization 4263 case S_MOD_TERM_FUNC_POINTERS: 4264 type = eSymbolTypeCode; 4265 break; // section with only function pointers for termination 4266 case S_INTERPOSING: 4267 type = eSymbolTypeTrampoline; 4268 break; // section with only pairs of function pointers for 4269 // interposing 4270 case S_16BYTE_LITERALS: 4271 type = eSymbolTypeData; 4272 break; // section with only 16 byte literals 4273 case S_DTRACE_DOF: 4274 type = eSymbolTypeInstrumentation; 4275 break; 4276 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4277 type = eSymbolTypeTrampoline; 4278 break; 4279 default: 4280 switch (symbol_section->GetType()) { 4281 case lldb::eSectionTypeCode: 4282 type = eSymbolTypeCode; 4283 break; 4284 case eSectionTypeData: 4285 case eSectionTypeDataCString: // Inlined C string data 4286 case eSectionTypeDataCStringPointers: // Pointers to C string 4287 // data 4288 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4289 // the symbol table 4290 case eSectionTypeData4: 4291 case eSectionTypeData8: 4292 case eSectionTypeData16: 4293 type = eSymbolTypeData; 4294 break; 4295 default: 4296 break; 4297 } 4298 break; 4299 } 4300 4301 if (type == eSymbolTypeInvalid) { 4302 const char *symbol_sect_name = 4303 symbol_section->GetName().AsCString(); 4304 if (symbol_section->IsDescendant(text_section_sp.get())) { 4305 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4306 S_ATTR_SELF_MODIFYING_CODE | 4307 S_ATTR_SOME_INSTRUCTIONS)) 4308 type = eSymbolTypeData; 4309 else 4310 type = eSymbolTypeCode; 4311 } else if (symbol_section->IsDescendant( 4312 data_section_sp.get()) || 4313 symbol_section->IsDescendant( 4314 data_dirty_section_sp.get()) || 4315 symbol_section->IsDescendant( 4316 data_const_section_sp.get())) { 4317 if (symbol_sect_name && 4318 ::strstr(symbol_sect_name, "__objc") == 4319 symbol_sect_name) { 4320 type = eSymbolTypeRuntime; 4321 4322 if (symbol_name) { 4323 llvm::StringRef symbol_name_ref(symbol_name); 4324 if (symbol_name_ref.startswith("_OBJC_")) { 4325 static const llvm::StringRef g_objc_v2_prefix_class( 4326 "_OBJC_CLASS_$_"); 4327 static const llvm::StringRef g_objc_v2_prefix_metaclass( 4328 "_OBJC_METACLASS_$_"); 4329 static const llvm::StringRef g_objc_v2_prefix_ivar( 4330 "_OBJC_IVAR_$_"); 4331 if (symbol_name_ref.startswith( 4332 g_objc_v2_prefix_class)) { 4333 symbol_name_non_abi_mangled = symbol_name + 1; 4334 symbol_name = 4335 symbol_name + g_objc_v2_prefix_class.size(); 4336 type = eSymbolTypeObjCClass; 4337 demangled_is_synthesized = true; 4338 } else if (symbol_name_ref.startswith( 4339 g_objc_v2_prefix_metaclass)) { 4340 symbol_name_non_abi_mangled = symbol_name + 1; 4341 symbol_name = 4342 symbol_name + g_objc_v2_prefix_metaclass.size(); 4343 type = eSymbolTypeObjCMetaClass; 4344 demangled_is_synthesized = true; 4345 } else if (symbol_name_ref.startswith( 4346 g_objc_v2_prefix_ivar)) { 4347 symbol_name_non_abi_mangled = symbol_name + 1; 4348 symbol_name = 4349 symbol_name + g_objc_v2_prefix_ivar.size(); 4350 type = eSymbolTypeObjCIVar; 4351 demangled_is_synthesized = true; 4352 } 4353 } 4354 } 4355 } else if (symbol_sect_name && 4356 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4357 symbol_sect_name) { 4358 type = eSymbolTypeException; 4359 } else { 4360 type = eSymbolTypeData; 4361 } 4362 } else if (symbol_sect_name && 4363 ::strstr(symbol_sect_name, "__IMPORT") == 4364 symbol_sect_name) { 4365 type = eSymbolTypeTrampoline; 4366 } else if (symbol_section->IsDescendant( 4367 objc_section_sp.get())) { 4368 type = eSymbolTypeRuntime; 4369 if (symbol_name && symbol_name[0] == '.') { 4370 llvm::StringRef symbol_name_ref(symbol_name); 4371 static const llvm::StringRef g_objc_v1_prefix_class( 4372 ".objc_class_name_"); 4373 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4374 symbol_name_non_abi_mangled = symbol_name; 4375 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4376 type = eSymbolTypeObjCClass; 4377 demangled_is_synthesized = true; 4378 } 4379 } 4380 } 4381 } 4382 } 4383 } break; 4384 } 4385 } 4386 4387 if (add_nlist) { 4388 uint64_t symbol_value = nlist.n_value; 4389 4390 if (symbol_name_non_abi_mangled) { 4391 sym[sym_idx].GetMangled().SetMangledName( 4392 ConstString(symbol_name_non_abi_mangled)); 4393 sym[sym_idx].GetMangled().SetDemangledName( 4394 ConstString(symbol_name)); 4395 } else { 4396 bool symbol_name_is_mangled = false; 4397 4398 if (symbol_name && symbol_name[0] == '_') { 4399 symbol_name_is_mangled = symbol_name[1] == '_'; 4400 symbol_name++; // Skip the leading underscore 4401 } 4402 4403 if (symbol_name) { 4404 ConstString const_symbol_name(symbol_name); 4405 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4406 symbol_name_is_mangled); 4407 } 4408 } 4409 4410 if (is_gsym) { 4411 const char *gsym_name = sym[sym_idx] 4412 .GetMangled() 4413 .GetName(lldb::eLanguageTypeUnknown, 4414 Mangled::ePreferMangled) 4415 .GetCString(); 4416 if (gsym_name) 4417 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4418 } 4419 4420 if (symbol_section) { 4421 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4422 if (symbol_byte_size == 0 && function_starts_count > 0) { 4423 addr_t symbol_lookup_file_addr = nlist.n_value; 4424 // Do an exact address match for non-ARM addresses, else get the 4425 // closest since the symbol might be a thumb symbol which has an 4426 // address with bit zero set 4427 FunctionStarts::Entry *func_start_entry = 4428 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4429 if (is_arm && func_start_entry) { 4430 // Verify that the function start address is the symbol address 4431 // (ARM) or the symbol address + 1 (thumb) 4432 if (func_start_entry->addr != symbol_lookup_file_addr && 4433 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4434 // Not the right entry, NULL it out... 4435 func_start_entry = NULL; 4436 } 4437 } 4438 if (func_start_entry) { 4439 func_start_entry->data = true; 4440 4441 addr_t symbol_file_addr = func_start_entry->addr; 4442 if (is_arm) 4443 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4444 4445 const FunctionStarts::Entry *next_func_start_entry = 4446 function_starts.FindNextEntry(func_start_entry); 4447 const addr_t section_end_file_addr = 4448 section_file_addr + symbol_section->GetByteSize(); 4449 if (next_func_start_entry) { 4450 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4451 // Be sure the clear the Thumb address bit when we calculate 4452 // the size from the current and next address 4453 if (is_arm) 4454 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4455 symbol_byte_size = std::min<lldb::addr_t>( 4456 next_symbol_file_addr - symbol_file_addr, 4457 section_end_file_addr - symbol_file_addr); 4458 } else { 4459 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4460 } 4461 } 4462 } 4463 symbol_value -= section_file_addr; 4464 } 4465 4466 if (is_debug == false) { 4467 if (type == eSymbolTypeCode) { 4468 // See if we can find a N_FUN entry for any code symbols. If we 4469 // do find a match, and the name matches, then we can merge the 4470 // two into just the function symbol to avoid duplicate entries 4471 // in the symbol table 4472 std::pair<ValueToSymbolIndexMap::const_iterator, 4473 ValueToSymbolIndexMap::const_iterator> 4474 range; 4475 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4476 if (range.first != range.second) { 4477 bool found_it = false; 4478 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4479 pos != range.second; ++pos) { 4480 if (sym[sym_idx].GetMangled().GetName( 4481 lldb::eLanguageTypeUnknown, 4482 Mangled::ePreferMangled) == 4483 sym[pos->second].GetMangled().GetName( 4484 lldb::eLanguageTypeUnknown, 4485 Mangled::ePreferMangled)) { 4486 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4487 // We just need the flags from the linker symbol, so put 4488 // these flags into the N_FUN flags to avoid duplicate 4489 // symbols in the symbol table 4490 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4491 sym[pos->second].SetFlags(nlist.n_type << 16 | 4492 nlist.n_desc); 4493 if (resolver_addresses.find(nlist.n_value) != 4494 resolver_addresses.end()) 4495 sym[pos->second].SetType(eSymbolTypeResolver); 4496 sym[sym_idx].Clear(); 4497 found_it = true; 4498 break; 4499 } 4500 } 4501 if (found_it) 4502 continue; 4503 } else { 4504 if (resolver_addresses.find(nlist.n_value) != 4505 resolver_addresses.end()) 4506 type = eSymbolTypeResolver; 4507 } 4508 } else if (type == eSymbolTypeData || 4509 type == eSymbolTypeObjCClass || 4510 type == eSymbolTypeObjCMetaClass || 4511 type == eSymbolTypeObjCIVar) { 4512 // See if we can find a N_STSYM entry for any data symbols. If we 4513 // do find a match, and the name matches, then we can merge the 4514 // two into just the Static symbol to avoid duplicate entries in 4515 // the symbol table 4516 std::pair<ValueToSymbolIndexMap::const_iterator, 4517 ValueToSymbolIndexMap::const_iterator> 4518 range; 4519 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4520 if (range.first != range.second) { 4521 bool found_it = false; 4522 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4523 pos != range.second; ++pos) { 4524 if (sym[sym_idx].GetMangled().GetName( 4525 lldb::eLanguageTypeUnknown, 4526 Mangled::ePreferMangled) == 4527 sym[pos->second].GetMangled().GetName( 4528 lldb::eLanguageTypeUnknown, 4529 Mangled::ePreferMangled)) { 4530 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4531 // We just need the flags from the linker symbol, so put 4532 // these flags into the N_STSYM flags to avoid duplicate 4533 // symbols in the symbol table 4534 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4535 sym[pos->second].SetFlags(nlist.n_type << 16 | 4536 nlist.n_desc); 4537 sym[sym_idx].Clear(); 4538 found_it = true; 4539 break; 4540 } 4541 } 4542 if (found_it) 4543 continue; 4544 } else { 4545 // Combine N_GSYM stab entries with the non stab symbol 4546 const char *gsym_name = sym[sym_idx] 4547 .GetMangled() 4548 .GetName(lldb::eLanguageTypeUnknown, 4549 Mangled::ePreferMangled) 4550 .GetCString(); 4551 if (gsym_name) { 4552 ConstNameToSymbolIndexMap::const_iterator pos = 4553 N_GSYM_name_to_sym_idx.find(gsym_name); 4554 if (pos != N_GSYM_name_to_sym_idx.end()) { 4555 const uint32_t GSYM_sym_idx = pos->second; 4556 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4557 // Copy the address, because often the N_GSYM address has 4558 // an invalid address of zero when the global is a common 4559 // symbol 4560 sym[GSYM_sym_idx].GetAddressRef().SetSection( 4561 symbol_section); 4562 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4563 // We just need the flags from the linker symbol, so put 4564 // these flags into the N_GSYM flags to avoid duplicate 4565 // symbols in the symbol table 4566 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 4567 nlist.n_desc); 4568 sym[sym_idx].Clear(); 4569 continue; 4570 } 4571 } 4572 } 4573 } 4574 } 4575 4576 sym[sym_idx].SetID(nlist_idx); 4577 sym[sym_idx].SetType(type); 4578 if (set_value) { 4579 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4580 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4581 } 4582 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4583 4584 if (symbol_byte_size > 0) 4585 sym[sym_idx].SetByteSize(symbol_byte_size); 4586 4587 if (demangled_is_synthesized) 4588 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4589 4590 ++sym_idx; 4591 } else { 4592 sym[sym_idx].Clear(); 4593 } 4594 } 4595 4596 for (const auto &pos : reexport_shlib_needs_fixup) { 4597 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4598 if (undef_pos != undefined_name_to_desc.end()) { 4599 const uint8_t dylib_ordinal = 4600 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4601 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4602 sym[pos.first].SetReExportedSymbolSharedLibrary( 4603 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4604 } 4605 } 4606 } 4607 4608 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4609 4610 if (function_starts_count > 0) { 4611 uint32_t num_synthetic_function_symbols = 0; 4612 for (i = 0; i < function_starts_count; ++i) { 4613 if (function_starts.GetEntryRef(i).data == false) 4614 ++num_synthetic_function_symbols; 4615 } 4616 4617 if (num_synthetic_function_symbols > 0) { 4618 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4619 num_syms = sym_idx + num_synthetic_function_symbols; 4620 sym = symtab->Resize(num_syms); 4621 } 4622 for (i = 0; i < function_starts_count; ++i) { 4623 const FunctionStarts::Entry *func_start_entry = 4624 function_starts.GetEntryAtIndex(i); 4625 if (func_start_entry->data == false) { 4626 addr_t symbol_file_addr = func_start_entry->addr; 4627 uint32_t symbol_flags = 0; 4628 if (is_arm) { 4629 if (symbol_file_addr & 1) 4630 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4631 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4632 } 4633 Address symbol_addr; 4634 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4635 SectionSP symbol_section(symbol_addr.GetSection()); 4636 uint32_t symbol_byte_size = 0; 4637 if (symbol_section) { 4638 const addr_t section_file_addr = 4639 symbol_section->GetFileAddress(); 4640 const FunctionStarts::Entry *next_func_start_entry = 4641 function_starts.FindNextEntry(func_start_entry); 4642 const addr_t section_end_file_addr = 4643 section_file_addr + symbol_section->GetByteSize(); 4644 if (next_func_start_entry) { 4645 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4646 if (is_arm) 4647 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4648 symbol_byte_size = std::min<lldb::addr_t>( 4649 next_symbol_file_addr - symbol_file_addr, 4650 section_end_file_addr - symbol_file_addr); 4651 } else { 4652 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4653 } 4654 sym[sym_idx].SetID(synthetic_sym_id++); 4655 sym[sym_idx].GetMangled().SetDemangledName( 4656 GetNextSyntheticSymbolName()); 4657 sym[sym_idx].SetType(eSymbolTypeCode); 4658 sym[sym_idx].SetIsSynthetic(true); 4659 sym[sym_idx].GetAddressRef() = symbol_addr; 4660 if (symbol_flags) 4661 sym[sym_idx].SetFlags(symbol_flags); 4662 if (symbol_byte_size) 4663 sym[sym_idx].SetByteSize(symbol_byte_size); 4664 ++sym_idx; 4665 } 4666 } 4667 } 4668 } 4669 } 4670 } 4671 4672 // Trim our symbols down to just what we ended up with after removing any 4673 // symbols. 4674 if (sym_idx < num_syms) { 4675 num_syms = sym_idx; 4676 sym = symtab->Resize(num_syms); 4677 } 4678 4679 // Now synthesize indirect symbols 4680 if (m_dysymtab.nindirectsyms != 0) { 4681 if (indirect_symbol_index_data.GetByteSize()) { 4682 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4683 m_nlist_idx_to_sym_idx.end(); 4684 4685 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4686 ++sect_idx) { 4687 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4688 S_SYMBOL_STUBS) { 4689 uint32_t symbol_stub_byte_size = 4690 m_mach_sections[sect_idx].reserved2; 4691 if (symbol_stub_byte_size == 0) 4692 continue; 4693 4694 const uint32_t num_symbol_stubs = 4695 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4696 4697 if (num_symbol_stubs == 0) 4698 continue; 4699 4700 const uint32_t symbol_stub_index_offset = 4701 m_mach_sections[sect_idx].reserved1; 4702 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; 4703 ++stub_idx) { 4704 const uint32_t symbol_stub_index = 4705 symbol_stub_index_offset + stub_idx; 4706 const lldb::addr_t symbol_stub_addr = 4707 m_mach_sections[sect_idx].addr + 4708 (stub_idx * symbol_stub_byte_size); 4709 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4710 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4711 symbol_stub_offset, 4)) { 4712 const uint32_t stub_sym_id = 4713 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4714 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4715 continue; 4716 4717 NListIndexToSymbolIndexMap::const_iterator index_pos = 4718 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4719 Symbol *stub_symbol = NULL; 4720 if (index_pos != end_index_pos) { 4721 // We have a remapping from the original nlist index to a 4722 // current symbol index, so just look this up by index 4723 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4724 } else { 4725 // We need to lookup a symbol using the original nlist symbol 4726 // index since this index is coming from the S_SYMBOL_STUBS 4727 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4728 } 4729 4730 if (stub_symbol) { 4731 Address so_addr(symbol_stub_addr, section_list); 4732 4733 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4734 // Change the external symbol into a trampoline that makes 4735 // sense These symbols were N_UNDF N_EXT, and are useless 4736 // to us, so we can re-use them so we don't have to make up 4737 // a synthetic symbol for no good reason. 4738 if (resolver_addresses.find(symbol_stub_addr) == 4739 resolver_addresses.end()) 4740 stub_symbol->SetType(eSymbolTypeTrampoline); 4741 else 4742 stub_symbol->SetType(eSymbolTypeResolver); 4743 stub_symbol->SetExternal(false); 4744 stub_symbol->GetAddressRef() = so_addr; 4745 stub_symbol->SetByteSize(symbol_stub_byte_size); 4746 } else { 4747 // Make a synthetic symbol to describe the trampoline stub 4748 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4749 if (sym_idx >= num_syms) { 4750 sym = symtab->Resize(++num_syms); 4751 stub_symbol = NULL; // this pointer no longer valid 4752 } 4753 sym[sym_idx].SetID(synthetic_sym_id++); 4754 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4755 if (resolver_addresses.find(symbol_stub_addr) == 4756 resolver_addresses.end()) 4757 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4758 else 4759 sym[sym_idx].SetType(eSymbolTypeResolver); 4760 sym[sym_idx].SetIsSynthetic(true); 4761 sym[sym_idx].GetAddressRef() = so_addr; 4762 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4763 ++sym_idx; 4764 } 4765 } else { 4766 if (log) 4767 log->Warning("symbol stub referencing symbol table symbol " 4768 "%u that isn't in our minimal symbol table, " 4769 "fix this!!!", 4770 stub_sym_id); 4771 } 4772 } 4773 } 4774 } 4775 } 4776 } 4777 } 4778 4779 if (!trie_entries.empty()) { 4780 for (const auto &e : trie_entries) { 4781 if (e.entry.import_name) { 4782 // Only add indirect symbols from the Trie entries if we didn't have 4783 // a N_INDR nlist entry for this already 4784 if (indirect_symbol_names.find(e.entry.name) == 4785 indirect_symbol_names.end()) { 4786 // Make a synthetic symbol to describe re-exported symbol. 4787 if (sym_idx >= num_syms) 4788 sym = symtab->Resize(++num_syms); 4789 sym[sym_idx].SetID(synthetic_sym_id++); 4790 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4791 sym[sym_idx].SetType(eSymbolTypeReExported); 4792 sym[sym_idx].SetIsSynthetic(true); 4793 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4794 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4795 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4796 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4797 } 4798 ++sym_idx; 4799 } 4800 } 4801 } 4802 } 4803 4804 // StreamFile s(stdout, false); 4805 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4806 // symtab->Dump(&s, NULL, eSortOrderNone); 4807 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4808 // sizes 4809 symtab->CalculateSymbolSizes(); 4810 4811 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4812 // symtab->Dump(&s, NULL, eSortOrderNone); 4813 4814 return symtab->GetNumSymbols(); 4815 } 4816 return 0; 4817 } 4818 4819 void ObjectFileMachO::Dump(Stream *s) { 4820 ModuleSP module_sp(GetModule()); 4821 if (module_sp) { 4822 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4823 s->Printf("%p: ", static_cast<void *>(this)); 4824 s->Indent(); 4825 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4826 s->PutCString("ObjectFileMachO64"); 4827 else 4828 s->PutCString("ObjectFileMachO32"); 4829 4830 ArchSpec header_arch; 4831 GetArchitecture(header_arch); 4832 4833 *s << ", file = '" << m_file 4834 << "', triple = " << header_arch.GetTriple().getTriple() << "\n"; 4835 4836 SectionList *sections = GetSectionList(); 4837 if (sections) 4838 sections->Dump(s, NULL, true, UINT32_MAX); 4839 4840 if (m_symtab_ap.get()) 4841 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4842 } 4843 } 4844 4845 bool ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4846 const lldb_private::DataExtractor &data, 4847 lldb::offset_t lc_offset, 4848 lldb_private::UUID &uuid) { 4849 uint32_t i; 4850 struct uuid_command load_cmd; 4851 4852 lldb::offset_t offset = lc_offset; 4853 for (i = 0; i < header.ncmds; ++i) { 4854 const lldb::offset_t cmd_offset = offset; 4855 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4856 break; 4857 4858 if (load_cmd.cmd == LC_UUID) { 4859 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4860 4861 if (uuid_bytes) { 4862 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4863 // We pretend these object files have no UUID to prevent crashing. 4864 4865 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4866 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4867 0xbb, 0x14, 0xf0, 0x0d}; 4868 4869 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4870 return false; 4871 4872 uuid = UUID::fromOptionalData(uuid_bytes, 16); 4873 return true; 4874 } 4875 return false; 4876 } 4877 offset = cmd_offset + load_cmd.cmdsize; 4878 } 4879 return false; 4880 } 4881 4882 static const char *GetOSName(uint32_t cmd) { 4883 switch (cmd) { 4884 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4885 return "ios"; 4886 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4887 return "macosx"; 4888 case llvm::MachO::LC_VERSION_MIN_TVOS: 4889 return "tvos"; 4890 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4891 return "watchos"; 4892 default: 4893 llvm_unreachable("unexpected LC_VERSION load command"); 4894 } 4895 } 4896 4897 bool ObjectFileMachO::GetArchitecture(const llvm::MachO::mach_header &header, 4898 const lldb_private::DataExtractor &data, 4899 lldb::offset_t lc_offset, 4900 ArchSpec &arch) { 4901 arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4902 4903 if (arch.IsValid()) { 4904 llvm::Triple &triple = arch.GetTriple(); 4905 4906 // Set OS to an unspecified unknown or a "*" so it can match any OS 4907 triple.setOS(llvm::Triple::UnknownOS); 4908 triple.setOSName(llvm::StringRef()); 4909 4910 if (header.filetype == MH_PRELOAD) { 4911 if (header.cputype == CPU_TYPE_ARM) { 4912 // If this is a 32-bit arm binary, and it's a standalone binary, force 4913 // the Vendor to Apple so we don't accidentally pick up the generic 4914 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4915 // frame pointer register; most other armv7 ABIs use a combination of 4916 // r7 and r11. 4917 triple.setVendor(llvm::Triple::Apple); 4918 } else { 4919 // Set vendor to an unspecified unknown or a "*" so it can match any 4920 // vendor This is required for correct behavior of EFI debugging on 4921 // x86_64 4922 triple.setVendor(llvm::Triple::UnknownVendor); 4923 triple.setVendorName(llvm::StringRef()); 4924 } 4925 return true; 4926 } else { 4927 struct load_command load_cmd; 4928 4929 lldb::offset_t offset = lc_offset; 4930 for (uint32_t i = 0; i < header.ncmds; ++i) { 4931 const lldb::offset_t cmd_offset = offset; 4932 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4933 break; 4934 4935 uint32_t major, minor, patch; 4936 struct version_min_command version_min; 4937 4938 llvm::SmallString<16> os_name; 4939 llvm::raw_svector_ostream os(os_name); 4940 4941 switch (load_cmd.cmd) { 4942 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4943 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4944 case llvm::MachO::LC_VERSION_MIN_TVOS: 4945 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4946 if (load_cmd.cmdsize != sizeof(version_min)) 4947 break; 4948 data.ExtractBytes(cmd_offset, 4949 sizeof(version_min), data.GetByteOrder(), 4950 &version_min); 4951 major = version_min.version >> 16; 4952 minor = (version_min.version >> 8) & 0xffu; 4953 patch = version_min.version & 0xffu; 4954 os << GetOSName(load_cmd.cmd) << major << '.' << minor << '.' 4955 << patch; 4956 triple.setOSName(os.str()); 4957 return true; 4958 default: 4959 break; 4960 } 4961 4962 offset = cmd_offset + load_cmd.cmdsize; 4963 } 4964 4965 if (header.filetype != MH_KEXT_BUNDLE) { 4966 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 4967 // so lets not say our Vendor is Apple, leave it as an unspecified 4968 // unknown 4969 triple.setVendor(llvm::Triple::UnknownVendor); 4970 triple.setVendorName(llvm::StringRef()); 4971 } 4972 } 4973 } 4974 return arch.IsValid(); 4975 } 4976 4977 bool ObjectFileMachO::GetUUID(lldb_private::UUID *uuid) { 4978 ModuleSP module_sp(GetModule()); 4979 if (module_sp) { 4980 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4981 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4982 return GetUUID(m_header, m_data, offset, *uuid); 4983 } 4984 return false; 4985 } 4986 4987 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 4988 uint32_t count = 0; 4989 ModuleSP module_sp(GetModule()); 4990 if (module_sp) { 4991 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4992 struct load_command load_cmd; 4993 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4994 std::vector<std::string> rpath_paths; 4995 std::vector<std::string> rpath_relative_paths; 4996 std::vector<std::string> at_exec_relative_paths; 4997 const bool resolve_path = false; // Don't resolve the dependent file paths 4998 // since they may not reside on this 4999 // system 5000 uint32_t i; 5001 for (i = 0; i < m_header.ncmds; ++i) { 5002 const uint32_t cmd_offset = offset; 5003 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5004 break; 5005 5006 switch (load_cmd.cmd) { 5007 case LC_RPATH: 5008 case LC_LOAD_DYLIB: 5009 case LC_LOAD_WEAK_DYLIB: 5010 case LC_REEXPORT_DYLIB: 5011 case LC_LOAD_DYLINKER: 5012 case LC_LOADFVMLIB: 5013 case LC_LOAD_UPWARD_DYLIB: { 5014 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5015 const char *path = m_data.PeekCStr(name_offset); 5016 if (path) { 5017 if (load_cmd.cmd == LC_RPATH) 5018 rpath_paths.push_back(path); 5019 else { 5020 if (path[0] == '@') { 5021 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5022 rpath_relative_paths.push_back(path + strlen("@rpath")); 5023 else if (strncmp(path, "@executable_path", 5024 strlen("@executable_path")) == 0) 5025 at_exec_relative_paths.push_back(path 5026 + strlen("@executable_path")); 5027 } else { 5028 FileSpec file_spec(path, resolve_path); 5029 if (files.AppendIfUnique(file_spec)) 5030 count++; 5031 } 5032 } 5033 } 5034 } break; 5035 5036 default: 5037 break; 5038 } 5039 offset = cmd_offset + load_cmd.cmdsize; 5040 } 5041 5042 FileSpec this_file_spec(m_file); 5043 this_file_spec.ResolvePath(); 5044 5045 if (!rpath_paths.empty()) { 5046 // Fixup all LC_RPATH values to be absolute paths 5047 std::string loader_path("@loader_path"); 5048 std::string executable_path("@executable_path"); 5049 for (auto &rpath : rpath_paths) { 5050 if (rpath.find(loader_path) == 0) { 5051 rpath.erase(0, loader_path.size()); 5052 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5053 } else if (rpath.find(executable_path) == 0) { 5054 rpath.erase(0, executable_path.size()); 5055 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5056 } 5057 } 5058 5059 for (const auto &rpath_relative_path : rpath_relative_paths) { 5060 for (const auto &rpath : rpath_paths) { 5061 std::string path = rpath; 5062 path += rpath_relative_path; 5063 // It is OK to resolve this path because we must find a file on disk 5064 // for us to accept it anyway if it is rpath relative. 5065 FileSpec file_spec(path, true); 5066 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) { 5067 count++; 5068 break; 5069 } 5070 } 5071 } 5072 } 5073 5074 // We may have @executable_paths but no RPATHS. Figure those out here. 5075 // Only do this if this object file is the executable. We have no way to 5076 // get back to the actual executable otherwise, so we won't get the right 5077 // path. 5078 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5079 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5080 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5081 FileSpec file_spec = 5082 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5083 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) 5084 count++; 5085 } 5086 } 5087 } 5088 return count; 5089 } 5090 5091 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5092 // If the object file is not an executable it can't hold the entry point. 5093 // m_entry_point_address is initialized to an invalid address, so we can just 5094 // return that. If m_entry_point_address is valid it means we've found it 5095 // already, so return the cached value. 5096 5097 if (!IsExecutable() || m_entry_point_address.IsValid()) 5098 return m_entry_point_address; 5099 5100 // Otherwise, look for the UnixThread or Thread command. The data for the 5101 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5102 // 5103 // uint32_t flavor - this is the flavor argument you would pass to 5104 // thread_get_state 5105 // uint32_t count - this is the count of longs in the thread state data 5106 // struct XXX_thread_state state - this is the structure from 5107 // <machine/thread_status.h> corresponding to the flavor. 5108 // <repeat this trio> 5109 // 5110 // So we just keep reading the various register flavors till we find the GPR 5111 // one, then read the PC out of there. 5112 // FIXME: We will need to have a "RegisterContext data provider" class at some 5113 // point that can get all the registers 5114 // out of data in this form & attach them to a given thread. That should 5115 // underlie the MacOS X User process plugin, and we'll also need it for the 5116 // MacOS X Core File process plugin. When we have that we can also use it 5117 // here. 5118 // 5119 // For now we hard-code the offsets and flavors we need: 5120 // 5121 // 5122 5123 ModuleSP module_sp(GetModule()); 5124 if (module_sp) { 5125 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5126 struct load_command load_cmd; 5127 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5128 uint32_t i; 5129 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5130 bool done = false; 5131 5132 for (i = 0; i < m_header.ncmds; ++i) { 5133 const lldb::offset_t cmd_offset = offset; 5134 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5135 break; 5136 5137 switch (load_cmd.cmd) { 5138 case LC_UNIXTHREAD: 5139 case LC_THREAD: { 5140 while (offset < cmd_offset + load_cmd.cmdsize) { 5141 uint32_t flavor = m_data.GetU32(&offset); 5142 uint32_t count = m_data.GetU32(&offset); 5143 if (count == 0) { 5144 // We've gotten off somehow, log and exit; 5145 return m_entry_point_address; 5146 } 5147 5148 switch (m_header.cputype) { 5149 case llvm::MachO::CPU_TYPE_ARM: 5150 if (flavor == 1 || 5151 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 from 5152 // mach/arm/thread_status.h 5153 { 5154 offset += 60; // This is the offset of pc in the GPR thread state 5155 // data structure. 5156 start_address = m_data.GetU32(&offset); 5157 done = true; 5158 } 5159 break; 5160 case llvm::MachO::CPU_TYPE_ARM64: 5161 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5162 { 5163 offset += 256; // This is the offset of pc in the GPR thread state 5164 // data structure. 5165 start_address = m_data.GetU64(&offset); 5166 done = true; 5167 } 5168 break; 5169 case llvm::MachO::CPU_TYPE_I386: 5170 if (flavor == 5171 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5172 { 5173 offset += 40; // This is the offset of eip in the GPR thread state 5174 // data structure. 5175 start_address = m_data.GetU32(&offset); 5176 done = true; 5177 } 5178 break; 5179 case llvm::MachO::CPU_TYPE_X86_64: 5180 if (flavor == 5181 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5182 { 5183 offset += 16 * 8; // This is the offset of rip in the GPR thread 5184 // state data structure. 5185 start_address = m_data.GetU64(&offset); 5186 done = true; 5187 } 5188 break; 5189 default: 5190 return m_entry_point_address; 5191 } 5192 // Haven't found the GPR flavor yet, skip over the data for this 5193 // flavor: 5194 if (done) 5195 break; 5196 offset += count * 4; 5197 } 5198 } break; 5199 case LC_MAIN: { 5200 ConstString text_segment_name("__TEXT"); 5201 uint64_t entryoffset = m_data.GetU64(&offset); 5202 SectionSP text_segment_sp = 5203 GetSectionList()->FindSectionByName(text_segment_name); 5204 if (text_segment_sp) { 5205 done = true; 5206 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5207 } 5208 } break; 5209 5210 default: 5211 break; 5212 } 5213 if (done) 5214 break; 5215 5216 // Go to the next load command: 5217 offset = cmd_offset + load_cmd.cmdsize; 5218 } 5219 5220 if (start_address != LLDB_INVALID_ADDRESS) { 5221 // We got the start address from the load commands, so now resolve that 5222 // address in the sections of this ObjectFile: 5223 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5224 start_address, GetSectionList())) { 5225 m_entry_point_address.Clear(); 5226 } 5227 } else { 5228 // We couldn't read the UnixThread load command - maybe it wasn't there. 5229 // As a fallback look for the "start" symbol in the main executable. 5230 5231 ModuleSP module_sp(GetModule()); 5232 5233 if (module_sp) { 5234 SymbolContextList contexts; 5235 SymbolContext context; 5236 if (module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5237 eSymbolTypeCode, contexts)) { 5238 if (contexts.GetContextAtIndex(0, context)) 5239 m_entry_point_address = context.symbol->GetAddress(); 5240 } 5241 } 5242 } 5243 } 5244 5245 return m_entry_point_address; 5246 } 5247 5248 lldb_private::Address ObjectFileMachO::GetHeaderAddress() { 5249 lldb_private::Address header_addr; 5250 SectionList *section_list = GetSectionList(); 5251 if (section_list) { 5252 SectionSP text_segment_sp( 5253 section_list->FindSectionByName(GetSegmentNameTEXT())); 5254 if (text_segment_sp) { 5255 header_addr.SetSection(text_segment_sp); 5256 header_addr.SetOffset(0); 5257 } 5258 } 5259 return header_addr; 5260 } 5261 5262 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5263 ModuleSP module_sp(GetModule()); 5264 if (module_sp) { 5265 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5266 if (!m_thread_context_offsets_valid) { 5267 m_thread_context_offsets_valid = true; 5268 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5269 FileRangeArray::Entry file_range; 5270 thread_command thread_cmd; 5271 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5272 const uint32_t cmd_offset = offset; 5273 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 5274 break; 5275 5276 if (thread_cmd.cmd == LC_THREAD) { 5277 file_range.SetRangeBase(offset); 5278 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5279 m_thread_context_offsets.Append(file_range); 5280 } 5281 offset = cmd_offset + thread_cmd.cmdsize; 5282 } 5283 } 5284 } 5285 return m_thread_context_offsets.GetSize(); 5286 } 5287 5288 std::string ObjectFileMachO::GetIdentifierString() { 5289 std::string result; 5290 ModuleSP module_sp(GetModule()); 5291 if (module_sp) { 5292 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5293 5294 // First, look over the load commands for an LC_NOTE load command with 5295 // data_owner string "kern ver str" & use that if found. 5296 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5297 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5298 const uint32_t cmd_offset = offset; 5299 load_command lc; 5300 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5301 break; 5302 if (lc.cmd == LC_NOTE) 5303 { 5304 char data_owner[17]; 5305 m_data.CopyData (offset, 16, data_owner); 5306 data_owner[16] = '\0'; 5307 offset += 16; 5308 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5309 uint64_t size = m_data.GetU64_unchecked (&offset); 5310 5311 // "kern ver str" has a uint32_t version and then a nul terminated 5312 // c-string. 5313 if (strcmp ("kern ver str", data_owner) == 0) 5314 { 5315 offset = fileoff; 5316 uint32_t version; 5317 if (m_data.GetU32 (&offset, &version, 1) != nullptr) 5318 { 5319 if (version == 1) 5320 { 5321 uint32_t strsize = size - sizeof (uint32_t); 5322 char *buf = (char*) malloc (strsize); 5323 if (buf) 5324 { 5325 m_data.CopyData (offset, strsize, buf); 5326 buf[strsize - 1] = '\0'; 5327 result = buf; 5328 if (buf) 5329 free (buf); 5330 return result; 5331 } 5332 } 5333 } 5334 } 5335 } 5336 offset = cmd_offset + lc.cmdsize; 5337 } 5338 5339 // Second, make a pass over the load commands looking for an obsolete 5340 // LC_IDENT load command. 5341 offset = MachHeaderSizeFromMagic(m_header.magic); 5342 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5343 const uint32_t cmd_offset = offset; 5344 struct ident_command ident_command; 5345 if (m_data.GetU32(&offset, &ident_command, 2) == NULL) 5346 break; 5347 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5348 char *buf = (char *) malloc (ident_command.cmdsize); 5349 if (buf != nullptr 5350 && m_data.CopyData (offset, ident_command.cmdsize, buf) == ident_command.cmdsize) { 5351 buf[ident_command.cmdsize - 1] = '\0'; 5352 result = buf; 5353 } 5354 if (buf) 5355 free (buf); 5356 } 5357 offset = cmd_offset + ident_command.cmdsize; 5358 } 5359 5360 } 5361 return result; 5362 } 5363 5364 bool ObjectFileMachO::GetCorefileMainBinaryInfo (addr_t &address, UUID &uuid) { 5365 address = LLDB_INVALID_ADDRESS; 5366 uuid.Clear(); 5367 ModuleSP module_sp(GetModule()); 5368 if (module_sp) { 5369 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5370 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5371 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5372 const uint32_t cmd_offset = offset; 5373 load_command lc; 5374 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5375 break; 5376 if (lc.cmd == LC_NOTE) 5377 { 5378 char data_owner[17]; 5379 memset (data_owner, 0, sizeof (data_owner)); 5380 m_data.CopyData (offset, 16, data_owner); 5381 offset += 16; 5382 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5383 uint64_t size = m_data.GetU64_unchecked (&offset); 5384 5385 // "main bin spec" (main binary specification) data payload is 5386 // formatted: 5387 // uint32_t version [currently 1] 5388 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user process] 5389 // uint64_t address [ UINT64_MAX if address not specified ] 5390 // uuid_t uuid [ all zero's if uuid not specified ] 5391 // uint32_t log2_pagesize [ process page size in log base 2, e.g. 4k pages are 12. 0 for unspecified ] 5392 5393 if (strcmp ("main bin spec", data_owner) == 0 && size >= 32) 5394 { 5395 offset = fileoff; 5396 uint32_t version; 5397 if (m_data.GetU32 (&offset, &version, 1) != nullptr && version == 1) 5398 { 5399 uint32_t type = 0; 5400 uuid_t raw_uuid; 5401 memset (raw_uuid, 0, sizeof (uuid_t)); 5402 5403 if (m_data.GetU32(&offset, &type, 1) && 5404 m_data.GetU64(&offset, &address, 1) && 5405 m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5406 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5407 return true; 5408 } 5409 } 5410 } 5411 } 5412 offset = cmd_offset + lc.cmdsize; 5413 } 5414 } 5415 return false; 5416 } 5417 5418 lldb::RegisterContextSP 5419 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5420 lldb_private::Thread &thread) { 5421 lldb::RegisterContextSP reg_ctx_sp; 5422 5423 ModuleSP module_sp(GetModule()); 5424 if (module_sp) { 5425 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5426 if (!m_thread_context_offsets_valid) 5427 GetNumThreadContexts(); 5428 5429 const FileRangeArray::Entry *thread_context_file_range = 5430 m_thread_context_offsets.GetEntryAtIndex(idx); 5431 if (thread_context_file_range) { 5432 5433 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5434 thread_context_file_range->GetByteSize()); 5435 5436 switch (m_header.cputype) { 5437 case llvm::MachO::CPU_TYPE_ARM64: 5438 reg_ctx_sp.reset(new RegisterContextDarwin_arm64_Mach(thread, data)); 5439 break; 5440 5441 case llvm::MachO::CPU_TYPE_ARM: 5442 reg_ctx_sp.reset(new RegisterContextDarwin_arm_Mach(thread, data)); 5443 break; 5444 5445 case llvm::MachO::CPU_TYPE_I386: 5446 reg_ctx_sp.reset(new RegisterContextDarwin_i386_Mach(thread, data)); 5447 break; 5448 5449 case llvm::MachO::CPU_TYPE_X86_64: 5450 reg_ctx_sp.reset(new RegisterContextDarwin_x86_64_Mach(thread, data)); 5451 break; 5452 } 5453 } 5454 } 5455 return reg_ctx_sp; 5456 } 5457 5458 ObjectFile::Type ObjectFileMachO::CalculateType() { 5459 switch (m_header.filetype) { 5460 case MH_OBJECT: // 0x1u 5461 if (GetAddressByteSize() == 4) { 5462 // 32 bit kexts are just object files, but they do have a valid 5463 // UUID load command. 5464 UUID uuid; 5465 if (GetUUID(&uuid)) { 5466 // this checking for the UUID load command is not enough we could 5467 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5468 // this is required of kexts 5469 if (m_strata == eStrataInvalid) 5470 m_strata = eStrataKernel; 5471 return eTypeSharedLibrary; 5472 } 5473 } 5474 return eTypeObjectFile; 5475 5476 case MH_EXECUTE: 5477 return eTypeExecutable; // 0x2u 5478 case MH_FVMLIB: 5479 return eTypeSharedLibrary; // 0x3u 5480 case MH_CORE: 5481 return eTypeCoreFile; // 0x4u 5482 case MH_PRELOAD: 5483 return eTypeSharedLibrary; // 0x5u 5484 case MH_DYLIB: 5485 return eTypeSharedLibrary; // 0x6u 5486 case MH_DYLINKER: 5487 return eTypeDynamicLinker; // 0x7u 5488 case MH_BUNDLE: 5489 return eTypeSharedLibrary; // 0x8u 5490 case MH_DYLIB_STUB: 5491 return eTypeStubLibrary; // 0x9u 5492 case MH_DSYM: 5493 return eTypeDebugInfo; // 0xAu 5494 case MH_KEXT_BUNDLE: 5495 return eTypeSharedLibrary; // 0xBu 5496 default: 5497 break; 5498 } 5499 return eTypeUnknown; 5500 } 5501 5502 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5503 switch (m_header.filetype) { 5504 case MH_OBJECT: // 0x1u 5505 { 5506 // 32 bit kexts are just object files, but they do have a valid 5507 // UUID load command. 5508 UUID uuid; 5509 if (GetUUID(&uuid)) { 5510 // this checking for the UUID load command is not enough we could 5511 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5512 // this is required of kexts 5513 if (m_type == eTypeInvalid) 5514 m_type = eTypeSharedLibrary; 5515 5516 return eStrataKernel; 5517 } 5518 } 5519 return eStrataUnknown; 5520 5521 case MH_EXECUTE: // 0x2u 5522 // Check for the MH_DYLDLINK bit in the flags 5523 if (m_header.flags & MH_DYLDLINK) { 5524 return eStrataUser; 5525 } else { 5526 SectionList *section_list = GetSectionList(); 5527 if (section_list) { 5528 static ConstString g_kld_section_name("__KLD"); 5529 if (section_list->FindSectionByName(g_kld_section_name)) 5530 return eStrataKernel; 5531 } 5532 } 5533 return eStrataRawImage; 5534 5535 case MH_FVMLIB: 5536 return eStrataUser; // 0x3u 5537 case MH_CORE: 5538 return eStrataUnknown; // 0x4u 5539 case MH_PRELOAD: 5540 return eStrataRawImage; // 0x5u 5541 case MH_DYLIB: 5542 return eStrataUser; // 0x6u 5543 case MH_DYLINKER: 5544 return eStrataUser; // 0x7u 5545 case MH_BUNDLE: 5546 return eStrataUser; // 0x8u 5547 case MH_DYLIB_STUB: 5548 return eStrataUser; // 0x9u 5549 case MH_DSYM: 5550 return eStrataUnknown; // 0xAu 5551 case MH_KEXT_BUNDLE: 5552 return eStrataKernel; // 0xBu 5553 default: 5554 break; 5555 } 5556 return eStrataUnknown; 5557 } 5558 5559 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5560 ModuleSP module_sp(GetModule()); 5561 if (module_sp) { 5562 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5563 struct dylib_command load_cmd; 5564 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5565 uint32_t version_cmd = 0; 5566 uint64_t version = 0; 5567 uint32_t i; 5568 for (i = 0; i < m_header.ncmds; ++i) { 5569 const lldb::offset_t cmd_offset = offset; 5570 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5571 break; 5572 5573 if (load_cmd.cmd == LC_ID_DYLIB) { 5574 if (version_cmd == 0) { 5575 version_cmd = load_cmd.cmd; 5576 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5577 break; 5578 version = load_cmd.dylib.current_version; 5579 } 5580 break; // Break for now unless there is another more complete version 5581 // number load command in the future. 5582 } 5583 offset = cmd_offset + load_cmd.cmdsize; 5584 } 5585 5586 if (version_cmd == LC_ID_DYLIB) { 5587 unsigned major = (version & 0xFFFF0000ull) >> 16; 5588 unsigned minor = (version & 0x0000FF00ull) >> 8; 5589 unsigned subminor = (version & 0x000000FFull); 5590 return llvm::VersionTuple(major, minor, subminor); 5591 } 5592 } 5593 return llvm::VersionTuple(); 5594 } 5595 5596 bool ObjectFileMachO::GetArchitecture(ArchSpec &arch) { 5597 ModuleSP module_sp(GetModule()); 5598 if (module_sp) { 5599 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5600 return GetArchitecture(m_header, m_data, 5601 MachHeaderSizeFromMagic(m_header.magic), arch); 5602 } 5603 return false; 5604 } 5605 5606 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, addr_t &base_addr, UUID &uuid) { 5607 uuid.Clear(); 5608 base_addr = LLDB_INVALID_ADDRESS; 5609 if (process && process->GetDynamicLoader()) { 5610 DynamicLoader *dl = process->GetDynamicLoader(); 5611 LazyBool using_shared_cache; 5612 LazyBool private_shared_cache; 5613 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5614 private_shared_cache); 5615 } 5616 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5617 if (log) 5618 log->Printf("inferior process shared cache has a UUID of %s, base address 0x%" PRIx64 , uuid.GetAsString().c_str(), base_addr); 5619 } 5620 5621 // From dyld SPI header dyld_process_info.h 5622 typedef void *dyld_process_info; 5623 struct lldb_copy__dyld_process_cache_info { 5624 uuid_t cacheUUID; // UUID of cache used by process 5625 uint64_t cacheBaseAddress; // load address of dyld shared cache 5626 bool noCache; // process is running without a dyld cache 5627 bool privateCache; // process is using a private copy of its dyld cache 5628 }; 5629 5630 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with llvm 5631 // enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile errors. 5632 // So we need to use the actual underlying types of task_t and kern_return_t 5633 // below. 5634 extern "C" unsigned int /*task_t*/ mach_task_self(); 5635 5636 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5637 uuid.Clear(); 5638 base_addr = LLDB_INVALID_ADDRESS; 5639 5640 #if defined(__APPLE__) && \ 5641 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5642 uint8_t *(*dyld_get_all_image_infos)(void); 5643 dyld_get_all_image_infos = 5644 (uint8_t * (*)())dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5645 if (dyld_get_all_image_infos) { 5646 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5647 if (dyld_all_image_infos_address) { 5648 uint32_t *version = (uint32_t *) 5649 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5650 if (*version >= 13) { 5651 uuid_t *sharedCacheUUID_address = 0; 5652 int wordsize = sizeof(uint8_t *); 5653 if (wordsize == 8) { 5654 sharedCacheUUID_address = 5655 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5656 160); // sharedCacheUUID <mach-o/dyld_images.h> 5657 if (*version >= 15) 5658 base_addr = *(uint64_t *) ((uint8_t *) dyld_all_image_infos_address 5659 + 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5660 } else { 5661 sharedCacheUUID_address = 5662 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5663 84); // sharedCacheUUID <mach-o/dyld_images.h> 5664 if (*version >= 15) { 5665 base_addr = 0; 5666 base_addr = *(uint32_t *) ((uint8_t *) dyld_all_image_infos_address 5667 + 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5668 } 5669 } 5670 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5671 } 5672 } 5673 } else { 5674 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5675 dyld_process_info (*dyld_process_info_create)(unsigned int /* task_t */ task, uint64_t timestamp, unsigned int /*kern_return_t*/ *kernelError); 5676 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5677 void (*dyld_process_info_release)(dyld_process_info info); 5678 5679 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, unsigned int /*kern_return_t*/ *)) 5680 dlsym (RTLD_DEFAULT, "_dyld_process_info_create"); 5681 dyld_process_info_get_cache = (void (*)(void *, void *)) 5682 dlsym (RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5683 dyld_process_info_release = (void (*)(void *)) 5684 dlsym (RTLD_DEFAULT, "_dyld_process_info_release"); 5685 5686 if (dyld_process_info_create && dyld_process_info_get_cache) { 5687 unsigned int /*kern_return_t */ kern_ret; 5688 dyld_process_info process_info = dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5689 if (process_info) { 5690 struct lldb_copy__dyld_process_cache_info sc_info; 5691 memset (&sc_info, 0, sizeof (struct lldb_copy__dyld_process_cache_info)); 5692 dyld_process_info_get_cache (process_info, &sc_info); 5693 if (sc_info.cacheBaseAddress != 0) { 5694 base_addr = sc_info.cacheBaseAddress; 5695 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 5696 } 5697 dyld_process_info_release (process_info); 5698 } 5699 } 5700 } 5701 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5702 if (log && uuid.IsValid()) 5703 log->Printf("lldb's in-memory shared cache has a UUID of %s base address of 0x%" PRIx64, uuid.GetAsString().c_str(), base_addr); 5704 #endif 5705 } 5706 5707 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5708 if (!m_min_os_version) { 5709 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5710 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5711 const lldb::offset_t load_cmd_offset = offset; 5712 5713 version_min_command lc; 5714 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5715 break; 5716 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5717 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5718 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5719 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5720 if (m_data.GetU32(&offset, &lc.version, 5721 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5722 const uint32_t xxxx = lc.version >> 16; 5723 const uint32_t yy = (lc.version >> 8) & 0xffu; 5724 const uint32_t zz = lc.version & 0xffu; 5725 if (xxxx) { 5726 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5727 break; 5728 } 5729 } 5730 } 5731 offset = load_cmd_offset + lc.cmdsize; 5732 } 5733 5734 if (!m_min_os_version) { 5735 // Set version to an empty value so we don't keep trying to 5736 m_min_os_version = llvm::VersionTuple(); 5737 } 5738 } 5739 5740 return *m_min_os_version; 5741 } 5742 5743 uint32_t ObjectFileMachO::GetSDKVersion(uint32_t *versions, 5744 uint32_t num_versions) { 5745 if (m_sdk_versions.empty()) { 5746 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5747 bool success = false; 5748 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5749 const lldb::offset_t load_cmd_offset = offset; 5750 5751 version_min_command lc; 5752 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5753 break; 5754 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5755 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5756 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5757 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5758 if (m_data.GetU32(&offset, &lc.version, 5759 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5760 const uint32_t xxxx = lc.sdk >> 16; 5761 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5762 const uint32_t zz = lc.sdk & 0xffu; 5763 if (xxxx) { 5764 m_sdk_versions.push_back(xxxx); 5765 m_sdk_versions.push_back(yy); 5766 m_sdk_versions.push_back(zz); 5767 success = true; 5768 } else { 5769 GetModule()->ReportWarning( 5770 "minimum OS version load command with invalid (0) version found."); 5771 } 5772 } 5773 } 5774 offset = load_cmd_offset + lc.cmdsize; 5775 } 5776 5777 if (success == false) { 5778 // Push an invalid value so we don't try to find 5779 // the version # again on the next call to this 5780 // method. 5781 m_sdk_versions.push_back(UINT32_MAX); 5782 } 5783 } 5784 5785 // Legitimate version numbers will have 3 entries pushed 5786 // on to m_sdk_versions. If we only have one value, it's 5787 // the sentinel value indicating that this object file 5788 // does not have a valid minimum os version #. 5789 if (m_sdk_versions.size() > 1) { 5790 if (versions != NULL && num_versions > 0) { 5791 for (size_t i = 0; i < num_versions; ++i) { 5792 if (i < m_sdk_versions.size()) 5793 versions[i] = m_sdk_versions[i]; 5794 else 5795 versions[i] = 0; 5796 } 5797 } 5798 return m_sdk_versions.size(); 5799 } 5800 // Call the superclasses version that will empty out the data 5801 return ObjectFile::GetSDKVersion(versions, num_versions); 5802 } 5803 5804 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5805 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5806 } 5807 5808 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5809 return m_allow_assembly_emulation_unwind_plans; 5810 } 5811 5812 //------------------------------------------------------------------ 5813 // PluginInterface protocol 5814 //------------------------------------------------------------------ 5815 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5816 return GetPluginNameStatic(); 5817 } 5818 5819 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5820 5821 Section *ObjectFileMachO::GetMachHeaderSection() { 5822 // Find the first address of the mach header which is the first non-zero file 5823 // sized section whose file offset is zero. This is the base file address of 5824 // the mach-o file which can be subtracted from the vmaddr of the other 5825 // segments found in memory and added to the load address 5826 ModuleSP module_sp = GetModule(); 5827 if (module_sp) { 5828 SectionList *section_list = GetSectionList(); 5829 if (section_list) { 5830 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5831 const size_t num_sections = section_list->GetSize(); 5832 5833 for (size_t sect_idx = 0; sect_idx < num_sections && 5834 mach_base_file_addr == LLDB_INVALID_ADDRESS; 5835 ++sect_idx) { 5836 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5837 if (section && section->GetFileSize() > 0 && 5838 section->GetFileOffset() == 0 && 5839 section->IsThreadSpecific() == false && 5840 module_sp.get() == section->GetModule().get()) { 5841 return section; 5842 } 5843 } 5844 } 5845 } 5846 return nullptr; 5847 } 5848 5849 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 5850 lldb::addr_t mach_header_load_address, const Section *mach_header_section, 5851 const Section *section) { 5852 ModuleSP module_sp = GetModule(); 5853 if (module_sp && mach_header_section && section && 5854 mach_header_load_address != LLDB_INVALID_ADDRESS) { 5855 lldb::addr_t mach_header_file_addr = mach_header_section->GetFileAddress(); 5856 if (mach_header_file_addr != LLDB_INVALID_ADDRESS) { 5857 if (section && section->GetFileSize() > 0 && 5858 section->IsThreadSpecific() == false && 5859 module_sp.get() == section->GetModule().get()) { 5860 // Ignore __LINKEDIT and __DWARF segments 5861 if (section->GetName() == GetSegmentNameLINKEDIT()) { 5862 // Only map __LINKEDIT if we have an in memory image and this isn't a 5863 // kernel binary like a kext or mach_kernel. 5864 const bool is_memory_image = (bool)m_process_wp.lock(); 5865 const Strata strata = GetStrata(); 5866 if (is_memory_image == false || strata == eStrataKernel) 5867 return LLDB_INVALID_ADDRESS; 5868 } 5869 return section->GetFileAddress() - mach_header_file_addr + 5870 mach_header_load_address; 5871 } 5872 } 5873 } 5874 return LLDB_INVALID_ADDRESS; 5875 } 5876 5877 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 5878 bool value_is_offset) { 5879 ModuleSP module_sp = GetModule(); 5880 if (module_sp) { 5881 size_t num_loaded_sections = 0; 5882 SectionList *section_list = GetSectionList(); 5883 if (section_list) { 5884 const size_t num_sections = section_list->GetSize(); 5885 5886 if (value_is_offset) { 5887 // "value" is an offset to apply to each top level segment 5888 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5889 // Iterate through the object file sections to find all of the 5890 // sections that size on disk (to avoid __PAGEZERO) and load them 5891 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5892 if (section_sp && section_sp->GetFileSize() > 0 && 5893 section_sp->IsThreadSpecific() == false && 5894 module_sp.get() == section_sp->GetModule().get()) { 5895 // Ignore __LINKEDIT and __DWARF segments 5896 if (section_sp->GetName() == GetSegmentNameLINKEDIT()) { 5897 // Only map __LINKEDIT if we have an in memory image and this 5898 // isn't a kernel binary like a kext or mach_kernel. 5899 const bool is_memory_image = (bool)m_process_wp.lock(); 5900 const Strata strata = GetStrata(); 5901 if (is_memory_image == false || strata == eStrataKernel) 5902 continue; 5903 } 5904 if (target.GetSectionLoadList().SetSectionLoadAddress( 5905 section_sp, section_sp->GetFileAddress() + value)) 5906 ++num_loaded_sections; 5907 } 5908 } 5909 } else { 5910 // "value" is the new base address of the mach_header, adjust each 5911 // section accordingly 5912 5913 Section *mach_header_section = GetMachHeaderSection(); 5914 if (mach_header_section) { 5915 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5916 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5917 5918 lldb::addr_t section_load_addr = 5919 CalculateSectionLoadAddressForMemoryImage( 5920 value, mach_header_section, section_sp.get()); 5921 if (section_load_addr != LLDB_INVALID_ADDRESS) { 5922 if (target.GetSectionLoadList().SetSectionLoadAddress( 5923 section_sp, section_load_addr)) 5924 ++num_loaded_sections; 5925 } 5926 } 5927 } 5928 } 5929 } 5930 return num_loaded_sections > 0; 5931 } 5932 return false; 5933 } 5934 5935 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 5936 const FileSpec &outfile, Status &error) { 5937 if (process_sp) { 5938 Target &target = process_sp->GetTarget(); 5939 const ArchSpec target_arch = target.GetArchitecture(); 5940 const llvm::Triple &target_triple = target_arch.GetTriple(); 5941 if (target_triple.getVendor() == llvm::Triple::Apple && 5942 (target_triple.getOS() == llvm::Triple::MacOSX || 5943 target_triple.getOS() == llvm::Triple::IOS || 5944 target_triple.getOS() == llvm::Triple::WatchOS || 5945 target_triple.getOS() == llvm::Triple::TvOS)) { 5946 bool make_core = false; 5947 switch (target_arch.GetMachine()) { 5948 case llvm::Triple::aarch64: 5949 case llvm::Triple::arm: 5950 case llvm::Triple::thumb: 5951 case llvm::Triple::x86: 5952 case llvm::Triple::x86_64: 5953 make_core = true; 5954 break; 5955 default: 5956 error.SetErrorStringWithFormat("unsupported core architecture: %s", 5957 target_triple.str().c_str()); 5958 break; 5959 } 5960 5961 if (make_core) { 5962 std::vector<segment_command_64> segment_load_commands; 5963 // uint32_t range_info_idx = 0; 5964 MemoryRegionInfo range_info; 5965 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 5966 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 5967 const ByteOrder byte_order = target_arch.GetByteOrder(); 5968 if (range_error.Success()) { 5969 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 5970 const addr_t addr = range_info.GetRange().GetRangeBase(); 5971 const addr_t size = range_info.GetRange().GetByteSize(); 5972 5973 if (size == 0) 5974 break; 5975 5976 // Calculate correct protections 5977 uint32_t prot = 0; 5978 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 5979 prot |= VM_PROT_READ; 5980 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 5981 prot |= VM_PROT_WRITE; 5982 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 5983 prot |= VM_PROT_EXECUTE; 5984 5985 // printf ("[%3u] [0x%16.16" PRIx64 " - 5986 // 0x%16.16" PRIx64 ") %c%c%c\n", 5987 // range_info_idx, 5988 // addr, 5989 // size, 5990 // (prot & VM_PROT_READ ) ? 'r' : 5991 // '-', 5992 // (prot & VM_PROT_WRITE ) ? 'w' : 5993 // '-', 5994 // (prot & VM_PROT_EXECUTE) ? 'x' : 5995 // '-'); 5996 5997 if (prot != 0) { 5998 uint32_t cmd_type = LC_SEGMENT_64; 5999 uint32_t segment_size = sizeof(segment_command_64); 6000 if (addr_byte_size == 4) { 6001 cmd_type = LC_SEGMENT; 6002 segment_size = sizeof(segment_command); 6003 } 6004 segment_command_64 segment = { 6005 cmd_type, // uint32_t cmd; 6006 segment_size, // uint32_t cmdsize; 6007 {0}, // char segname[16]; 6008 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6009 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6010 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6011 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6012 prot, // uint32_t maxprot; 6013 prot, // uint32_t initprot; 6014 0, // uint32_t nsects; 6015 0}; // uint32_t flags; 6016 segment_load_commands.push_back(segment); 6017 } else { 6018 // No protections and a size of 1 used to be returned from old 6019 // debugservers when we asked about a region that was past the 6020 // last memory region and it indicates the end... 6021 if (size == 1) 6022 break; 6023 } 6024 6025 range_error = process_sp->GetMemoryRegionInfo( 6026 range_info.GetRange().GetRangeEnd(), range_info); 6027 if (range_error.Fail()) 6028 break; 6029 } 6030 6031 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6032 6033 mach_header_64 mach_header; 6034 if (addr_byte_size == 8) { 6035 mach_header.magic = MH_MAGIC_64; 6036 } else { 6037 mach_header.magic = MH_MAGIC; 6038 } 6039 mach_header.cputype = target_arch.GetMachOCPUType(); 6040 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6041 mach_header.filetype = MH_CORE; 6042 mach_header.ncmds = segment_load_commands.size(); 6043 mach_header.flags = 0; 6044 mach_header.reserved = 0; 6045 ThreadList &thread_list = process_sp->GetThreadList(); 6046 const uint32_t num_threads = thread_list.GetSize(); 6047 6048 // Make an array of LC_THREAD data items. Each one contains the 6049 // contents of the LC_THREAD load command. The data doesn't contain 6050 // the load command + load command size, we will add the load command 6051 // and load command size as we emit the data. 6052 std::vector<StreamString> LC_THREAD_datas(num_threads); 6053 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6054 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6055 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6056 LC_THREAD_data.SetByteOrder(byte_order); 6057 } 6058 for (uint32_t thread_idx = 0; thread_idx < num_threads; 6059 ++thread_idx) { 6060 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6061 if (thread_sp) { 6062 switch (mach_header.cputype) { 6063 case llvm::MachO::CPU_TYPE_ARM64: 6064 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6065 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6066 break; 6067 6068 case llvm::MachO::CPU_TYPE_ARM: 6069 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6070 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6071 break; 6072 6073 case llvm::MachO::CPU_TYPE_I386: 6074 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6075 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6076 break; 6077 6078 case llvm::MachO::CPU_TYPE_X86_64: 6079 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6080 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6081 break; 6082 } 6083 } 6084 } 6085 6086 // The size of the load command is the size of the segments... 6087 if (addr_byte_size == 8) { 6088 mach_header.sizeofcmds = segment_load_commands.size() * 6089 sizeof(struct segment_command_64); 6090 } else { 6091 mach_header.sizeofcmds = 6092 segment_load_commands.size() * sizeof(struct segment_command); 6093 } 6094 6095 // and the size of all LC_THREAD load command 6096 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6097 ++mach_header.ncmds; 6098 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6099 } 6100 6101 printf("mach_header: 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x " 6102 "0x%8.8x 0x%8.8x\n", 6103 mach_header.magic, mach_header.cputype, mach_header.cpusubtype, 6104 mach_header.filetype, mach_header.ncmds, 6105 mach_header.sizeofcmds, mach_header.flags, 6106 mach_header.reserved); 6107 6108 // Write the mach header 6109 buffer.PutHex32(mach_header.magic); 6110 buffer.PutHex32(mach_header.cputype); 6111 buffer.PutHex32(mach_header.cpusubtype); 6112 buffer.PutHex32(mach_header.filetype); 6113 buffer.PutHex32(mach_header.ncmds); 6114 buffer.PutHex32(mach_header.sizeofcmds); 6115 buffer.PutHex32(mach_header.flags); 6116 if (addr_byte_size == 8) { 6117 buffer.PutHex32(mach_header.reserved); 6118 } 6119 6120 // Skip the mach header and all load commands and align to the next 6121 // 0x1000 byte boundary 6122 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6123 if (file_offset & 0x00000fff) { 6124 file_offset += 0x00001000ull; 6125 file_offset &= (~0x00001000ull + 1); 6126 } 6127 6128 for (auto &segment : segment_load_commands) { 6129 segment.fileoff = file_offset; 6130 file_offset += segment.filesize; 6131 } 6132 6133 // Write out all of the LC_THREAD load commands 6134 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6135 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6136 buffer.PutHex32(LC_THREAD); 6137 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6138 buffer.Write(LC_THREAD_data.GetString().data(), 6139 LC_THREAD_data_size); 6140 } 6141 6142 // Write out all of the segment load commands 6143 for (const auto &segment : segment_load_commands) { 6144 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6145 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6146 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6147 segment.cmd, segment.cmdsize, segment.vmaddr, 6148 segment.vmaddr + segment.vmsize, segment.fileoff, 6149 segment.filesize, segment.maxprot, segment.initprot, 6150 segment.nsects, segment.flags); 6151 6152 buffer.PutHex32(segment.cmd); 6153 buffer.PutHex32(segment.cmdsize); 6154 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6155 if (addr_byte_size == 8) { 6156 buffer.PutHex64(segment.vmaddr); 6157 buffer.PutHex64(segment.vmsize); 6158 buffer.PutHex64(segment.fileoff); 6159 buffer.PutHex64(segment.filesize); 6160 } else { 6161 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6162 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6163 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6164 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6165 } 6166 buffer.PutHex32(segment.maxprot); 6167 buffer.PutHex32(segment.initprot); 6168 buffer.PutHex32(segment.nsects); 6169 buffer.PutHex32(segment.flags); 6170 } 6171 6172 File core_file; 6173 std::string core_file_path(outfile.GetPath()); 6174 error = core_file.Open(core_file_path.c_str(), 6175 File::eOpenOptionWrite | 6176 File::eOpenOptionTruncate | 6177 File::eOpenOptionCanCreate); 6178 if (error.Success()) { 6179 // Read 1 page at a time 6180 uint8_t bytes[0x1000]; 6181 // Write the mach header and load commands out to the core file 6182 size_t bytes_written = buffer.GetString().size(); 6183 error = core_file.Write(buffer.GetString().data(), bytes_written); 6184 if (error.Success()) { 6185 // Now write the file data for all memory segments in the process 6186 for (const auto &segment : segment_load_commands) { 6187 if (core_file.SeekFromStart(segment.fileoff) == -1) { 6188 error.SetErrorStringWithFormat( 6189 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6190 segment.fileoff, core_file_path.c_str()); 6191 break; 6192 } 6193 6194 printf("Saving %" PRId64 6195 " bytes of data for memory region at 0x%" PRIx64 "\n", 6196 segment.vmsize, segment.vmaddr); 6197 addr_t bytes_left = segment.vmsize; 6198 addr_t addr = segment.vmaddr; 6199 Status memory_read_error; 6200 while (bytes_left > 0 && error.Success()) { 6201 const size_t bytes_to_read = 6202 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6203 const size_t bytes_read = process_sp->ReadMemory( 6204 addr, bytes, bytes_to_read, memory_read_error); 6205 if (bytes_read == bytes_to_read) { 6206 size_t bytes_written = bytes_read; 6207 error = core_file.Write(bytes, bytes_written); 6208 bytes_left -= bytes_read; 6209 addr += bytes_read; 6210 } else { 6211 // Some pages within regions are not readable, those should 6212 // be zero filled 6213 memset(bytes, 0, bytes_to_read); 6214 size_t bytes_written = bytes_to_read; 6215 error = core_file.Write(bytes, bytes_written); 6216 bytes_left -= bytes_to_read; 6217 addr += bytes_to_read; 6218 } 6219 } 6220 } 6221 } 6222 } 6223 } else { 6224 error.SetErrorString( 6225 "process doesn't support getting memory region info"); 6226 } 6227 } 6228 return true; // This is the right plug to handle saving core files for 6229 // this process 6230 } 6231 } 6232 return false; 6233 } 6234