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 #include "llvm/ADT/StringRef.h" 11 12 #include "lldb/lldb-private-log.h" 13 #include "lldb/Core/ArchSpec.h" 14 #include "lldb/Core/DataBuffer.h" 15 #include "lldb/Core/Debugger.h" 16 #include "lldb/Core/Error.h" 17 #include "lldb/Core/FileSpecList.h" 18 #include "lldb/Core/Log.h" 19 #include "lldb/Core/Module.h" 20 #include "lldb/Core/ModuleSpec.h" 21 #include "lldb/Core/PluginManager.h" 22 #include "lldb/Core/RangeMap.h" 23 #include "lldb/Core/Section.h" 24 #include "lldb/Core/StreamFile.h" 25 #include "lldb/Core/StreamString.h" 26 #include "lldb/Core/Timer.h" 27 #include "lldb/Core/UUID.h" 28 #include "lldb/Host/Host.h" 29 #include "lldb/Host/FileSpec.h" 30 #include "lldb/Symbol/ClangNamespaceDecl.h" 31 #include "lldb/Symbol/DWARFCallFrameInfo.h" 32 #include "lldb/Symbol/ObjectFile.h" 33 #include "lldb/Target/MemoryRegionInfo.h" 34 #include "lldb/Target/Platform.h" 35 #include "lldb/Target/Process.h" 36 #include "lldb/Target/SectionLoadList.h" 37 #include "lldb/Target/Target.h" 38 #include "lldb/Target/Thread.h" 39 #include "lldb/Target/ThreadList.h" 40 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 41 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 42 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 43 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 44 45 #include "lldb/Utility/SafeMachO.h" 46 47 #include "ObjectFileMachO.h" 48 49 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 50 // GetLLDBSharedCacheUUID() needs to call dlsym() 51 #include <dlfcn.h> 52 #endif 53 54 #ifndef __APPLE__ 55 #include "Utility/UuidCompatibility.h" 56 #endif 57 58 using namespace lldb; 59 using namespace lldb_private; 60 using namespace llvm::MachO; 61 62 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 63 { 64 public: 65 RegisterContextDarwin_x86_64_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 66 RegisterContextDarwin_x86_64 (thread, 0) 67 { 68 SetRegisterDataFrom_LC_THREAD (data); 69 } 70 71 virtual void 72 InvalidateAllRegisters () 73 { 74 // Do nothing... registers are always valid... 75 } 76 77 void 78 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 79 { 80 lldb::offset_t offset = 0; 81 SetError (GPRRegSet, Read, -1); 82 SetError (FPURegSet, Read, -1); 83 SetError (EXCRegSet, Read, -1); 84 bool done = false; 85 86 while (!done) 87 { 88 int flavor = data.GetU32 (&offset); 89 if (flavor == 0) 90 done = true; 91 else 92 { 93 uint32_t i; 94 uint32_t count = data.GetU32 (&offset); 95 switch (flavor) 96 { 97 case GPRRegSet: 98 for (i=0; i<count; ++i) 99 (&gpr.rax)[i] = data.GetU64(&offset); 100 SetError (GPRRegSet, Read, 0); 101 done = true; 102 103 break; 104 case FPURegSet: 105 // TODO: fill in FPU regs.... 106 //SetError (FPURegSet, Read, -1); 107 done = true; 108 109 break; 110 case EXCRegSet: 111 exc.trapno = data.GetU32(&offset); 112 exc.err = data.GetU32(&offset); 113 exc.faultvaddr = data.GetU64(&offset); 114 SetError (EXCRegSet, Read, 0); 115 done = true; 116 break; 117 case 7: 118 case 8: 119 case 9: 120 // fancy flavors that encapsulate of the above 121 // flavors... 122 break; 123 124 default: 125 done = true; 126 break; 127 } 128 } 129 } 130 } 131 132 133 static size_t 134 WriteRegister (RegisterContext *reg_ctx, const char *name, const char *alt_name, size_t reg_byte_size, Stream &data) 135 { 136 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 137 if (reg_info == NULL) 138 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 139 if (reg_info) 140 { 141 lldb_private::RegisterValue reg_value; 142 if (reg_ctx->ReadRegister(reg_info, reg_value)) 143 { 144 if (reg_info->byte_size >= reg_byte_size) 145 data.Write(reg_value.GetBytes(), reg_byte_size); 146 else 147 { 148 data.Write(reg_value.GetBytes(), reg_info->byte_size); 149 for (size_t i=0, n = reg_byte_size - reg_info->byte_size; i<n; ++ i) 150 data.PutChar(0); 151 } 152 return reg_byte_size; 153 } 154 } 155 // Just write zeros if all else fails 156 for (size_t i=0; i<reg_byte_size; ++ i) 157 data.PutChar(0); 158 return reg_byte_size; 159 } 160 161 static bool 162 Create_LC_THREAD (Thread *thread, Stream &data) 163 { 164 RegisterContextSP reg_ctx_sp (thread->GetRegisterContext()); 165 if (reg_ctx_sp) 166 { 167 RegisterContext *reg_ctx = reg_ctx_sp.get(); 168 169 data.PutHex32 (GPRRegSet); // Flavor 170 data.PutHex32 (GPRWordCount); 171 WriteRegister (reg_ctx, "rax", NULL, 8, data); 172 WriteRegister (reg_ctx, "rbx", NULL, 8, data); 173 WriteRegister (reg_ctx, "rcx", NULL, 8, data); 174 WriteRegister (reg_ctx, "rdx", NULL, 8, data); 175 WriteRegister (reg_ctx, "rdi", NULL, 8, data); 176 WriteRegister (reg_ctx, "rsi", NULL, 8, data); 177 WriteRegister (reg_ctx, "rbp", NULL, 8, data); 178 WriteRegister (reg_ctx, "rsp", NULL, 8, data); 179 WriteRegister (reg_ctx, "r8", NULL, 8, data); 180 WriteRegister (reg_ctx, "r9", NULL, 8, data); 181 WriteRegister (reg_ctx, "r10", NULL, 8, data); 182 WriteRegister (reg_ctx, "r11", NULL, 8, data); 183 WriteRegister (reg_ctx, "r12", NULL, 8, data); 184 WriteRegister (reg_ctx, "r13", NULL, 8, data); 185 WriteRegister (reg_ctx, "r14", NULL, 8, data); 186 WriteRegister (reg_ctx, "r15", NULL, 8, data); 187 WriteRegister (reg_ctx, "rip", NULL, 8, data); 188 WriteRegister (reg_ctx, "rflags", NULL, 8, data); 189 WriteRegister (reg_ctx, "cs", NULL, 8, data); 190 WriteRegister (reg_ctx, "fs", NULL, 8, data); 191 WriteRegister (reg_ctx, "gs", NULL, 8, data); 192 193 // // Write out the FPU registers 194 // const size_t fpu_byte_size = sizeof(FPU); 195 // size_t bytes_written = 0; 196 // data.PutHex32 (FPURegSet); 197 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 198 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 199 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 200 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, data); // uint16_t fcw; // "fctrl" 201 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, data); // uint16_t fsw; // "fstat" 202 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, data); // uint8_t ftw; // "ftag" 203 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 204 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, data); // uint16_t fop; // "fop" 205 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, data); // uint32_t ip; // "fioff" 206 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, data); // uint16_t cs; // "fiseg" 207 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 208 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, data); // uint32_t dp; // "fooff" 209 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, data); // uint16_t ds; // "foseg" 210 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 211 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, data); // uint32_t mxcsr; 212 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 4, data);// uint32_t mxcsrmask; 213 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, sizeof(MMSReg), data); 214 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, sizeof(MMSReg), data); 215 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, sizeof(MMSReg), data); 216 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, sizeof(MMSReg), data); 217 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, sizeof(MMSReg), data); 218 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, sizeof(MMSReg), data); 219 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, sizeof(MMSReg), data); 220 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, sizeof(MMSReg), data); 221 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, sizeof(XMMReg), data); 222 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, sizeof(XMMReg), data); 223 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, sizeof(XMMReg), data); 224 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, sizeof(XMMReg), data); 225 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, sizeof(XMMReg), data); 226 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, sizeof(XMMReg), data); 227 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, sizeof(XMMReg), data); 228 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, sizeof(XMMReg), data); 229 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, sizeof(XMMReg), data); 230 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, sizeof(XMMReg), data); 231 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, sizeof(XMMReg), data); 232 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, sizeof(XMMReg), data); 233 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, sizeof(XMMReg), data); 234 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, sizeof(XMMReg), data); 235 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, sizeof(XMMReg), data); 236 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, sizeof(XMMReg), data); 237 // 238 // // Fill rest with zeros 239 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ i) 240 // data.PutChar(0); 241 242 // Write out the EXC registers 243 data.PutHex32 (EXCRegSet); 244 data.PutHex32 (EXCWordCount); 245 WriteRegister (reg_ctx, "trapno", NULL, 4, data); 246 WriteRegister (reg_ctx, "err", NULL, 4, data); 247 WriteRegister (reg_ctx, "faultvaddr", NULL, 8, data); 248 return true; 249 } 250 return false; 251 } 252 253 protected: 254 virtual int 255 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 256 { 257 return 0; 258 } 259 260 virtual int 261 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 262 { 263 return 0; 264 } 265 266 virtual int 267 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 268 { 269 return 0; 270 } 271 272 virtual int 273 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 274 { 275 return 0; 276 } 277 278 virtual int 279 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 280 { 281 return 0; 282 } 283 284 virtual int 285 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 286 { 287 return 0; 288 } 289 }; 290 291 292 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 293 { 294 public: 295 RegisterContextDarwin_i386_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 296 RegisterContextDarwin_i386 (thread, 0) 297 { 298 SetRegisterDataFrom_LC_THREAD (data); 299 } 300 301 virtual void 302 InvalidateAllRegisters () 303 { 304 // Do nothing... registers are always valid... 305 } 306 307 void 308 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 309 { 310 lldb::offset_t offset = 0; 311 SetError (GPRRegSet, Read, -1); 312 SetError (FPURegSet, Read, -1); 313 SetError (EXCRegSet, Read, -1); 314 bool done = false; 315 316 while (!done) 317 { 318 int flavor = data.GetU32 (&offset); 319 if (flavor == 0) 320 done = true; 321 else 322 { 323 uint32_t i; 324 uint32_t count = data.GetU32 (&offset); 325 switch (flavor) 326 { 327 case GPRRegSet: 328 for (i=0; i<count; ++i) 329 (&gpr.eax)[i] = data.GetU32(&offset); 330 SetError (GPRRegSet, Read, 0); 331 done = true; 332 333 break; 334 case FPURegSet: 335 // TODO: fill in FPU regs.... 336 //SetError (FPURegSet, Read, -1); 337 done = true; 338 339 break; 340 case EXCRegSet: 341 exc.trapno = data.GetU32(&offset); 342 exc.err = data.GetU32(&offset); 343 exc.faultvaddr = data.GetU32(&offset); 344 SetError (EXCRegSet, Read, 0); 345 done = true; 346 break; 347 case 7: 348 case 8: 349 case 9: 350 // fancy flavors that encapsulate of the above 351 // flavors... 352 break; 353 354 default: 355 done = true; 356 break; 357 } 358 } 359 } 360 } 361 362 static size_t 363 WriteRegister (RegisterContext *reg_ctx, const char *name, const char *alt_name, size_t reg_byte_size, Stream &data) 364 { 365 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 366 if (reg_info == NULL) 367 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 368 if (reg_info) 369 { 370 lldb_private::RegisterValue reg_value; 371 if (reg_ctx->ReadRegister(reg_info, reg_value)) 372 { 373 if (reg_info->byte_size >= reg_byte_size) 374 data.Write(reg_value.GetBytes(), reg_byte_size); 375 else 376 { 377 data.Write(reg_value.GetBytes(), reg_info->byte_size); 378 for (size_t i=0, n = reg_byte_size - reg_info->byte_size; i<n; ++ i) 379 data.PutChar(0); 380 } 381 return reg_byte_size; 382 } 383 } 384 // Just write zeros if all else fails 385 for (size_t i=0; i<reg_byte_size; ++ i) 386 data.PutChar(0); 387 return reg_byte_size; 388 } 389 390 static bool 391 Create_LC_THREAD (Thread *thread, Stream &data) 392 { 393 RegisterContextSP reg_ctx_sp (thread->GetRegisterContext()); 394 if (reg_ctx_sp) 395 { 396 RegisterContext *reg_ctx = reg_ctx_sp.get(); 397 398 data.PutHex32 (GPRRegSet); // Flavor 399 data.PutHex32 (GPRWordCount); 400 WriteRegister (reg_ctx, "eax", NULL, 4, data); 401 WriteRegister (reg_ctx, "ebx", NULL, 4, data); 402 WriteRegister (reg_ctx, "ecx", NULL, 4, data); 403 WriteRegister (reg_ctx, "edx", NULL, 4, data); 404 WriteRegister (reg_ctx, "edi", NULL, 4, data); 405 WriteRegister (reg_ctx, "esi", NULL, 4, data); 406 WriteRegister (reg_ctx, "ebp", NULL, 4, data); 407 WriteRegister (reg_ctx, "esp", NULL, 4, data); 408 WriteRegister (reg_ctx, "ss", NULL, 4, data); 409 WriteRegister (reg_ctx, "eflags", NULL, 4, data); 410 WriteRegister (reg_ctx, "eip", NULL, 4, data); 411 WriteRegister (reg_ctx, "cs", NULL, 4, data); 412 WriteRegister (reg_ctx, "ds", NULL, 4, data); 413 WriteRegister (reg_ctx, "es", NULL, 4, data); 414 WriteRegister (reg_ctx, "fs", NULL, 4, data); 415 WriteRegister (reg_ctx, "gs", NULL, 4, data); 416 417 // Write out the EXC registers 418 data.PutHex32 (EXCRegSet); 419 data.PutHex32 (EXCWordCount); 420 WriteRegister (reg_ctx, "trapno", NULL, 4, data); 421 WriteRegister (reg_ctx, "err", NULL, 4, data); 422 WriteRegister (reg_ctx, "faultvaddr", NULL, 4, data); 423 return true; 424 } 425 return false; 426 } 427 428 protected: 429 virtual int 430 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 431 { 432 return 0; 433 } 434 435 virtual int 436 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 437 { 438 return 0; 439 } 440 441 virtual int 442 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 443 { 444 return 0; 445 } 446 447 virtual int 448 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 449 { 450 return 0; 451 } 452 453 virtual int 454 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 455 { 456 return 0; 457 } 458 459 virtual int 460 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 461 { 462 return 0; 463 } 464 }; 465 466 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm 467 { 468 public: 469 RegisterContextDarwin_arm_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 470 RegisterContextDarwin_arm (thread, 0) 471 { 472 SetRegisterDataFrom_LC_THREAD (data); 473 } 474 475 virtual void 476 InvalidateAllRegisters () 477 { 478 // Do nothing... registers are always valid... 479 } 480 481 void 482 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 483 { 484 lldb::offset_t offset = 0; 485 SetError (GPRRegSet, Read, -1); 486 SetError (FPURegSet, Read, -1); 487 SetError (EXCRegSet, Read, -1); 488 bool done = false; 489 490 while (!done) 491 { 492 int flavor = data.GetU32 (&offset); 493 uint32_t count = data.GetU32 (&offset); 494 lldb::offset_t next_thread_state = offset + (count * 4); 495 switch (flavor) 496 { 497 case GPRRegSet: 498 for (uint32_t i=0; i<count; ++i) 499 { 500 gpr.r[i] = data.GetU32(&offset); 501 } 502 503 // Note that gpr.cpsr is also copied by the above loop; this loop technically extends 504 // one element past the end of the gpr.r[] array. 505 506 SetError (GPRRegSet, Read, 0); 507 offset = next_thread_state; 508 break; 509 510 case FPURegSet: 511 { 512 uint8_t *fpu_reg_buf = (uint8_t*) &fpu.floats.s[0]; 513 const int fpu_reg_buf_size = sizeof (fpu.floats); 514 if (data.ExtractBytes (offset, fpu_reg_buf_size, eByteOrderLittle, fpu_reg_buf) == fpu_reg_buf_size) 515 { 516 offset += fpu_reg_buf_size; 517 fpu.fpscr = data.GetU32(&offset); 518 SetError (FPURegSet, Read, 0); 519 } 520 else 521 { 522 done = true; 523 } 524 } 525 offset = next_thread_state; 526 break; 527 528 case EXCRegSet: 529 if (count == 3) 530 { 531 exc.exception = data.GetU32(&offset); 532 exc.fsr = data.GetU32(&offset); 533 exc.far = data.GetU32(&offset); 534 SetError (EXCRegSet, Read, 0); 535 } 536 done = true; 537 offset = next_thread_state; 538 break; 539 540 // Unknown register set flavor, stop trying to parse. 541 default: 542 done = true; 543 } 544 } 545 } 546 547 static size_t 548 WriteRegister (RegisterContext *reg_ctx, const char *name, const char *alt_name, size_t reg_byte_size, Stream &data) 549 { 550 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 551 if (reg_info == NULL) 552 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 553 if (reg_info) 554 { 555 lldb_private::RegisterValue reg_value; 556 if (reg_ctx->ReadRegister(reg_info, reg_value)) 557 { 558 if (reg_info->byte_size >= reg_byte_size) 559 data.Write(reg_value.GetBytes(), reg_byte_size); 560 else 561 { 562 data.Write(reg_value.GetBytes(), reg_info->byte_size); 563 for (size_t i=0, n = reg_byte_size - reg_info->byte_size; i<n; ++ i) 564 data.PutChar(0); 565 } 566 return reg_byte_size; 567 } 568 } 569 // Just write zeros if all else fails 570 for (size_t i=0; i<reg_byte_size; ++ i) 571 data.PutChar(0); 572 return reg_byte_size; 573 } 574 575 static bool 576 Create_LC_THREAD (Thread *thread, Stream &data) 577 { 578 RegisterContextSP reg_ctx_sp (thread->GetRegisterContext()); 579 if (reg_ctx_sp) 580 { 581 RegisterContext *reg_ctx = reg_ctx_sp.get(); 582 583 data.PutHex32 (GPRRegSet); // Flavor 584 data.PutHex32 (GPRWordCount); 585 WriteRegister (reg_ctx, "r0", NULL, 4, data); 586 WriteRegister (reg_ctx, "r1", NULL, 4, data); 587 WriteRegister (reg_ctx, "r2", NULL, 4, data); 588 WriteRegister (reg_ctx, "r3", NULL, 4, data); 589 WriteRegister (reg_ctx, "r4", NULL, 4, data); 590 WriteRegister (reg_ctx, "r5", NULL, 4, data); 591 WriteRegister (reg_ctx, "r6", NULL, 4, data); 592 WriteRegister (reg_ctx, "r7", NULL, 4, data); 593 WriteRegister (reg_ctx, "r8", NULL, 4, data); 594 WriteRegister (reg_ctx, "r9", NULL, 4, data); 595 WriteRegister (reg_ctx, "r10", NULL, 4, data); 596 WriteRegister (reg_ctx, "r11", NULL, 4, data); 597 WriteRegister (reg_ctx, "r12", NULL, 4, data); 598 WriteRegister (reg_ctx, "sp", NULL, 4, data); 599 WriteRegister (reg_ctx, "lr", NULL, 4, data); 600 WriteRegister (reg_ctx, "pc", NULL, 4, data); 601 WriteRegister (reg_ctx, "cpsr", NULL, 4, data); 602 603 // Write out the EXC registers 604 // data.PutHex32 (EXCRegSet); 605 // data.PutHex32 (EXCWordCount); 606 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 607 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 608 // WriteRegister (reg_ctx, "far", NULL, 4, data); 609 return true; 610 } 611 return false; 612 } 613 614 protected: 615 virtual int 616 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 617 { 618 return -1; 619 } 620 621 virtual int 622 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 623 { 624 return -1; 625 } 626 627 virtual int 628 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 629 { 630 return -1; 631 } 632 633 virtual int 634 DoReadDBG (lldb::tid_t tid, int flavor, DBG &dbg) 635 { 636 return -1; 637 } 638 639 virtual int 640 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 641 { 642 return 0; 643 } 644 645 virtual int 646 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 647 { 648 return 0; 649 } 650 651 virtual int 652 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 653 { 654 return 0; 655 } 656 657 virtual int 658 DoWriteDBG (lldb::tid_t tid, int flavor, const DBG &dbg) 659 { 660 return -1; 661 } 662 }; 663 664 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 665 { 666 public: 667 RegisterContextDarwin_arm64_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 668 RegisterContextDarwin_arm64 (thread, 0) 669 { 670 SetRegisterDataFrom_LC_THREAD (data); 671 } 672 673 virtual void 674 InvalidateAllRegisters () 675 { 676 // Do nothing... registers are always valid... 677 } 678 679 void 680 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 681 { 682 lldb::offset_t offset = 0; 683 SetError (GPRRegSet, Read, -1); 684 SetError (FPURegSet, Read, -1); 685 SetError (EXCRegSet, Read, -1); 686 bool done = false; 687 while (!done) 688 { 689 int flavor = data.GetU32 (&offset); 690 uint32_t count = data.GetU32 (&offset); 691 lldb::offset_t next_thread_state = offset + (count * 4); 692 switch (flavor) 693 { 694 case GPRRegSet: 695 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 32-bit register) 696 if (count >= (33 * 2) + 1) 697 { 698 for (uint32_t i=0; i<33; ++i) 699 gpr.x[i] = data.GetU64(&offset); 700 gpr.cpsr = data.GetU32(&offset); 701 SetError (GPRRegSet, Read, 0); 702 } 703 offset = next_thread_state; 704 break; 705 case FPURegSet: 706 { 707 uint8_t *fpu_reg_buf = (uint8_t*) &fpu.v[0]; 708 const int fpu_reg_buf_size = sizeof (fpu); 709 if (fpu_reg_buf_size == count 710 && data.ExtractBytes (offset, fpu_reg_buf_size, eByteOrderLittle, fpu_reg_buf) == fpu_reg_buf_size) 711 { 712 SetError (FPURegSet, Read, 0); 713 } 714 else 715 { 716 done = true; 717 } 718 } 719 offset = next_thread_state; 720 break; 721 case EXCRegSet: 722 if (count == 4) 723 { 724 exc.far = data.GetU64(&offset); 725 exc.esr = data.GetU32(&offset); 726 exc.exception = data.GetU32(&offset); 727 SetError (EXCRegSet, Read, 0); 728 } 729 offset = next_thread_state; 730 break; 731 default: 732 done = true; 733 break; 734 } 735 } 736 } 737 738 static size_t 739 WriteRegister (RegisterContext *reg_ctx, const char *name, const char *alt_name, size_t reg_byte_size, Stream &data) 740 { 741 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 742 if (reg_info == NULL) 743 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 744 if (reg_info) 745 { 746 lldb_private::RegisterValue reg_value; 747 if (reg_ctx->ReadRegister(reg_info, reg_value)) 748 { 749 if (reg_info->byte_size >= reg_byte_size) 750 data.Write(reg_value.GetBytes(), reg_byte_size); 751 else 752 { 753 data.Write(reg_value.GetBytes(), reg_info->byte_size); 754 for (size_t i=0, n = reg_byte_size - reg_info->byte_size; i<n; ++ i) 755 data.PutChar(0); 756 } 757 return reg_byte_size; 758 } 759 } 760 // Just write zeros if all else fails 761 for (size_t i=0; i<reg_byte_size; ++ i) 762 data.PutChar(0); 763 return reg_byte_size; 764 } 765 766 static bool 767 Create_LC_THREAD (Thread *thread, Stream &data) 768 { 769 RegisterContextSP reg_ctx_sp (thread->GetRegisterContext()); 770 if (reg_ctx_sp) 771 { 772 RegisterContext *reg_ctx = reg_ctx_sp.get(); 773 774 data.PutHex32 (GPRRegSet); // Flavor 775 data.PutHex32 (GPRWordCount); 776 WriteRegister (reg_ctx, "x0", NULL, 8, data); 777 WriteRegister (reg_ctx, "x1", NULL, 8, data); 778 WriteRegister (reg_ctx, "x2", NULL, 8, data); 779 WriteRegister (reg_ctx, "x3", NULL, 8, data); 780 WriteRegister (reg_ctx, "x4", NULL, 8, data); 781 WriteRegister (reg_ctx, "x5", NULL, 8, data); 782 WriteRegister (reg_ctx, "x6", NULL, 8, data); 783 WriteRegister (reg_ctx, "x7", NULL, 8, data); 784 WriteRegister (reg_ctx, "x8", NULL, 8, data); 785 WriteRegister (reg_ctx, "x9", NULL, 8, data); 786 WriteRegister (reg_ctx, "x10", NULL, 8, data); 787 WriteRegister (reg_ctx, "x11", NULL, 8, data); 788 WriteRegister (reg_ctx, "x12", NULL, 8, data); 789 WriteRegister (reg_ctx, "x13", NULL, 8, data); 790 WriteRegister (reg_ctx, "x14", NULL, 8, data); 791 WriteRegister (reg_ctx, "x15", NULL, 8, data); 792 WriteRegister (reg_ctx, "x16", NULL, 8, data); 793 WriteRegister (reg_ctx, "x17", NULL, 8, data); 794 WriteRegister (reg_ctx, "x18", NULL, 8, data); 795 WriteRegister (reg_ctx, "x19", NULL, 8, data); 796 WriteRegister (reg_ctx, "x20", NULL, 8, data); 797 WriteRegister (reg_ctx, "x21", NULL, 8, data); 798 WriteRegister (reg_ctx, "x22", NULL, 8, data); 799 WriteRegister (reg_ctx, "x23", NULL, 8, data); 800 WriteRegister (reg_ctx, "x24", NULL, 8, data); 801 WriteRegister (reg_ctx, "x25", NULL, 8, data); 802 WriteRegister (reg_ctx, "x26", NULL, 8, data); 803 WriteRegister (reg_ctx, "x27", NULL, 8, data); 804 WriteRegister (reg_ctx, "x28", NULL, 8, data); 805 WriteRegister (reg_ctx, "fp", NULL, 8, data); 806 WriteRegister (reg_ctx, "lr", NULL, 8, data); 807 WriteRegister (reg_ctx, "sp", NULL, 8, data); 808 WriteRegister (reg_ctx, "pc", NULL, 8, data); 809 WriteRegister (reg_ctx, "cpsr", NULL, 4, data); 810 811 // Write out the EXC registers 812 // data.PutHex32 (EXCRegSet); 813 // data.PutHex32 (EXCWordCount); 814 // WriteRegister (reg_ctx, "far", NULL, 8, data); 815 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 816 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 817 return true; 818 } 819 return false; 820 } 821 822 protected: 823 virtual int 824 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 825 { 826 return -1; 827 } 828 829 virtual int 830 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 831 { 832 return -1; 833 } 834 835 virtual int 836 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 837 { 838 return -1; 839 } 840 841 virtual int 842 DoReadDBG (lldb::tid_t tid, int flavor, DBG &dbg) 843 { 844 return -1; 845 } 846 847 virtual int 848 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 849 { 850 return 0; 851 } 852 853 virtual int 854 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 855 { 856 return 0; 857 } 858 859 virtual int 860 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 861 { 862 return 0; 863 } 864 865 virtual int 866 DoWriteDBG (lldb::tid_t tid, int flavor, const DBG &dbg) 867 { 868 return -1; 869 } 870 }; 871 872 static uint32_t 873 MachHeaderSizeFromMagic(uint32_t magic) 874 { 875 switch (magic) 876 { 877 case MH_MAGIC: 878 case MH_CIGAM: 879 return sizeof(struct mach_header); 880 881 case MH_MAGIC_64: 882 case MH_CIGAM_64: 883 return sizeof(struct mach_header_64); 884 break; 885 886 default: 887 break; 888 } 889 return 0; 890 } 891 892 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 893 894 void 895 ObjectFileMachO::Initialize() 896 { 897 PluginManager::RegisterPlugin (GetPluginNameStatic(), 898 GetPluginDescriptionStatic(), 899 CreateInstance, 900 CreateMemoryInstance, 901 GetModuleSpecifications, 902 SaveCore); 903 } 904 905 void 906 ObjectFileMachO::Terminate() 907 { 908 PluginManager::UnregisterPlugin (CreateInstance); 909 } 910 911 912 lldb_private::ConstString 913 ObjectFileMachO::GetPluginNameStatic() 914 { 915 static ConstString g_name("mach-o"); 916 return g_name; 917 } 918 919 const char * 920 ObjectFileMachO::GetPluginDescriptionStatic() 921 { 922 return "Mach-o object file reader (32 and 64 bit)"; 923 } 924 925 ObjectFile * 926 ObjectFileMachO::CreateInstance (const lldb::ModuleSP &module_sp, 927 DataBufferSP& data_sp, 928 lldb::offset_t data_offset, 929 const FileSpec* file, 930 lldb::offset_t file_offset, 931 lldb::offset_t length) 932 { 933 if (!data_sp) 934 { 935 data_sp = file->MemoryMapFileContentsIfLocal(file_offset, length); 936 data_offset = 0; 937 } 938 939 if (ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 940 { 941 // Update the data to contain the entire file if it doesn't already 942 if (data_sp->GetByteSize() < length) 943 { 944 data_sp = file->MemoryMapFileContentsIfLocal(file_offset, length); 945 data_offset = 0; 946 } 947 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, data_offset, file, file_offset, length)); 948 if (objfile_ap.get() && objfile_ap->ParseHeader()) 949 return objfile_ap.release(); 950 } 951 return NULL; 952 } 953 954 ObjectFile * 955 ObjectFileMachO::CreateMemoryInstance (const lldb::ModuleSP &module_sp, 956 DataBufferSP& data_sp, 957 const ProcessSP &process_sp, 958 lldb::addr_t header_addr) 959 { 960 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) 961 { 962 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, process_sp, header_addr)); 963 if (objfile_ap.get() && objfile_ap->ParseHeader()) 964 return objfile_ap.release(); 965 } 966 return NULL; 967 } 968 969 size_t 970 ObjectFileMachO::GetModuleSpecifications (const lldb_private::FileSpec& file, 971 lldb::DataBufferSP& data_sp, 972 lldb::offset_t data_offset, 973 lldb::offset_t file_offset, 974 lldb::offset_t length, 975 lldb_private::ModuleSpecList &specs) 976 { 977 const size_t initial_count = specs.GetSize(); 978 979 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) 980 { 981 DataExtractor data; 982 data.SetData(data_sp); 983 llvm::MachO::mach_header header; 984 if (ParseHeader (data, &data_offset, header)) 985 { 986 size_t header_and_load_cmds = header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 987 if (header_and_load_cmds >= data_sp->GetByteSize()) 988 { 989 data_sp = file.ReadFileContents(file_offset, header_and_load_cmds); 990 data.SetData(data_sp); 991 data_offset = MachHeaderSizeFromMagic(header.magic); 992 } 993 if (data_sp) 994 { 995 ModuleSpec spec; 996 spec.GetFileSpec() = file; 997 spec.SetObjectOffset(file_offset); 998 999 if (GetArchitecture (header, data, data_offset, spec.GetArchitecture())) 1000 { 1001 if (spec.GetArchitecture().IsValid()) 1002 { 1003 GetUUID (header, data, data_offset, spec.GetUUID()); 1004 specs.Append(spec); 1005 } 1006 } 1007 } 1008 } 1009 } 1010 return specs.GetSize() - initial_count; 1011 } 1012 1013 1014 1015 const ConstString & 1016 ObjectFileMachO::GetSegmentNameTEXT() 1017 { 1018 static ConstString g_segment_name_TEXT ("__TEXT"); 1019 return g_segment_name_TEXT; 1020 } 1021 1022 const ConstString & 1023 ObjectFileMachO::GetSegmentNameDATA() 1024 { 1025 static ConstString g_segment_name_DATA ("__DATA"); 1026 return g_segment_name_DATA; 1027 } 1028 1029 const ConstString & 1030 ObjectFileMachO::GetSegmentNameOBJC() 1031 { 1032 static ConstString g_segment_name_OBJC ("__OBJC"); 1033 return g_segment_name_OBJC; 1034 } 1035 1036 const ConstString & 1037 ObjectFileMachO::GetSegmentNameLINKEDIT() 1038 { 1039 static ConstString g_section_name_LINKEDIT ("__LINKEDIT"); 1040 return g_section_name_LINKEDIT; 1041 } 1042 1043 const ConstString & 1044 ObjectFileMachO::GetSectionNameEHFrame() 1045 { 1046 static ConstString g_section_name_eh_frame ("__eh_frame"); 1047 return g_section_name_eh_frame; 1048 } 1049 1050 bool 1051 ObjectFileMachO::MagicBytesMatch (DataBufferSP& data_sp, 1052 lldb::addr_t data_offset, 1053 lldb::addr_t data_length) 1054 { 1055 DataExtractor data; 1056 data.SetData (data_sp, data_offset, data_length); 1057 lldb::offset_t offset = 0; 1058 uint32_t magic = data.GetU32(&offset); 1059 return MachHeaderSizeFromMagic(magic) != 0; 1060 } 1061 1062 1063 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 1064 DataBufferSP& data_sp, 1065 lldb::offset_t data_offset, 1066 const FileSpec* file, 1067 lldb::offset_t file_offset, 1068 lldb::offset_t length) : 1069 ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 1070 m_mach_segments(), 1071 m_mach_sections(), 1072 m_entry_point_address(), 1073 m_thread_context_offsets(), 1074 m_thread_context_offsets_valid(false) 1075 { 1076 ::memset (&m_header, 0, sizeof(m_header)); 1077 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 1078 } 1079 1080 ObjectFileMachO::ObjectFileMachO (const lldb::ModuleSP &module_sp, 1081 lldb::DataBufferSP& header_data_sp, 1082 const lldb::ProcessSP &process_sp, 1083 lldb::addr_t header_addr) : 1084 ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 1085 m_mach_segments(), 1086 m_mach_sections(), 1087 m_entry_point_address(), 1088 m_thread_context_offsets(), 1089 m_thread_context_offsets_valid(false) 1090 { 1091 ::memset (&m_header, 0, sizeof(m_header)); 1092 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 1093 } 1094 1095 ObjectFileMachO::~ObjectFileMachO() 1096 { 1097 } 1098 1099 bool 1100 ObjectFileMachO::ParseHeader (DataExtractor &data, 1101 lldb::offset_t *data_offset_ptr, 1102 llvm::MachO::mach_header &header) 1103 { 1104 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1105 // Leave magic in the original byte order 1106 header.magic = data.GetU32(data_offset_ptr); 1107 bool can_parse = false; 1108 bool is_64_bit = false; 1109 switch (header.magic) 1110 { 1111 case MH_MAGIC: 1112 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1113 data.SetAddressByteSize(4); 1114 can_parse = true; 1115 break; 1116 1117 case MH_MAGIC_64: 1118 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1119 data.SetAddressByteSize(8); 1120 can_parse = true; 1121 is_64_bit = true; 1122 break; 1123 1124 case MH_CIGAM: 1125 data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 1126 data.SetAddressByteSize(4); 1127 can_parse = true; 1128 break; 1129 1130 case MH_CIGAM_64: 1131 data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 1132 data.SetAddressByteSize(8); 1133 is_64_bit = true; 1134 can_parse = true; 1135 break; 1136 1137 default: 1138 break; 1139 } 1140 1141 if (can_parse) 1142 { 1143 data.GetU32(data_offset_ptr, &header.cputype, 6); 1144 if (is_64_bit) 1145 *data_offset_ptr += 4; 1146 return true; 1147 } 1148 else 1149 { 1150 memset(&header, 0, sizeof(header)); 1151 } 1152 return false; 1153 } 1154 1155 bool 1156 ObjectFileMachO::ParseHeader () 1157 { 1158 ModuleSP module_sp(GetModule()); 1159 if (module_sp) 1160 { 1161 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 1162 bool can_parse = false; 1163 lldb::offset_t offset = 0; 1164 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1165 // Leave magic in the original byte order 1166 m_header.magic = m_data.GetU32(&offset); 1167 switch (m_header.magic) 1168 { 1169 case MH_MAGIC: 1170 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1171 m_data.SetAddressByteSize(4); 1172 can_parse = true; 1173 break; 1174 1175 case MH_MAGIC_64: 1176 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1177 m_data.SetAddressByteSize(8); 1178 can_parse = true; 1179 break; 1180 1181 case MH_CIGAM: 1182 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 1183 m_data.SetAddressByteSize(4); 1184 can_parse = true; 1185 break; 1186 1187 case MH_CIGAM_64: 1188 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 1189 m_data.SetAddressByteSize(8); 1190 can_parse = true; 1191 break; 1192 1193 default: 1194 break; 1195 } 1196 1197 if (can_parse) 1198 { 1199 m_data.GetU32(&offset, &m_header.cputype, 6); 1200 1201 1202 ArchSpec mach_arch; 1203 1204 if (GetArchitecture (mach_arch)) 1205 { 1206 // Check if the module has a required architecture 1207 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1208 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1209 return false; 1210 1211 if (SetModulesArchitecture (mach_arch)) 1212 { 1213 const size_t header_and_lc_size = m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1214 if (m_data.GetByteSize() < header_and_lc_size) 1215 { 1216 DataBufferSP data_sp; 1217 ProcessSP process_sp (m_process_wp.lock()); 1218 if (process_sp) 1219 { 1220 data_sp = ReadMemory (process_sp, m_memory_addr, header_and_lc_size); 1221 } 1222 else 1223 { 1224 // Read in all only the load command data from the file on disk 1225 data_sp = m_file.ReadFileContents(m_file_offset, header_and_lc_size); 1226 if (data_sp->GetByteSize() != header_and_lc_size) 1227 return false; 1228 } 1229 if (data_sp) 1230 m_data.SetData (data_sp); 1231 } 1232 } 1233 return true; 1234 } 1235 } 1236 else 1237 { 1238 memset(&m_header, 0, sizeof(struct mach_header)); 1239 } 1240 } 1241 return false; 1242 } 1243 1244 1245 ByteOrder 1246 ObjectFileMachO::GetByteOrder () const 1247 { 1248 return m_data.GetByteOrder (); 1249 } 1250 1251 bool 1252 ObjectFileMachO::IsExecutable() const 1253 { 1254 return m_header.filetype == MH_EXECUTE; 1255 } 1256 1257 uint32_t 1258 ObjectFileMachO::GetAddressByteSize () const 1259 { 1260 return m_data.GetAddressByteSize (); 1261 } 1262 1263 AddressClass 1264 ObjectFileMachO::GetAddressClass (lldb::addr_t file_addr) 1265 { 1266 Symtab *symtab = GetSymtab(); 1267 if (symtab) 1268 { 1269 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1270 if (symbol) 1271 { 1272 if (symbol->ValueIsAddress()) 1273 { 1274 SectionSP section_sp (symbol->GetAddress().GetSection()); 1275 if (section_sp) 1276 { 1277 const lldb::SectionType section_type = section_sp->GetType(); 1278 switch (section_type) 1279 { 1280 case eSectionTypeInvalid: 1281 return eAddressClassUnknown; 1282 1283 case eSectionTypeCode: 1284 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) 1285 { 1286 // For ARM we have a bit in the n_desc field of the symbol 1287 // that tells us ARM/Thumb which is bit 0x0008. 1288 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1289 return eAddressClassCodeAlternateISA; 1290 } 1291 return eAddressClassCode; 1292 1293 case eSectionTypeContainer: 1294 return eAddressClassUnknown; 1295 1296 case eSectionTypeData: 1297 case eSectionTypeDataCString: 1298 case eSectionTypeDataCStringPointers: 1299 case eSectionTypeDataSymbolAddress: 1300 case eSectionTypeData4: 1301 case eSectionTypeData8: 1302 case eSectionTypeData16: 1303 case eSectionTypeDataPointers: 1304 case eSectionTypeZeroFill: 1305 case eSectionTypeDataObjCMessageRefs: 1306 case eSectionTypeDataObjCCFStrings: 1307 return eAddressClassData; 1308 1309 case eSectionTypeDebug: 1310 case eSectionTypeDWARFDebugAbbrev: 1311 case eSectionTypeDWARFDebugAranges: 1312 case eSectionTypeDWARFDebugFrame: 1313 case eSectionTypeDWARFDebugInfo: 1314 case eSectionTypeDWARFDebugLine: 1315 case eSectionTypeDWARFDebugLoc: 1316 case eSectionTypeDWARFDebugMacInfo: 1317 case eSectionTypeDWARFDebugPubNames: 1318 case eSectionTypeDWARFDebugPubTypes: 1319 case eSectionTypeDWARFDebugRanges: 1320 case eSectionTypeDWARFDebugStr: 1321 case eSectionTypeDWARFAppleNames: 1322 case eSectionTypeDWARFAppleTypes: 1323 case eSectionTypeDWARFAppleNamespaces: 1324 case eSectionTypeDWARFAppleObjC: 1325 return eAddressClassDebug; 1326 1327 case eSectionTypeEHFrame: 1328 case eSectionTypeCompactUnwind: 1329 return eAddressClassRuntime; 1330 1331 case eSectionTypeELFSymbolTable: 1332 case eSectionTypeELFDynamicSymbols: 1333 case eSectionTypeELFRelocationEntries: 1334 case eSectionTypeELFDynamicLinkInfo: 1335 case eSectionTypeOther: 1336 return eAddressClassUnknown; 1337 } 1338 } 1339 } 1340 1341 const SymbolType symbol_type = symbol->GetType(); 1342 switch (symbol_type) 1343 { 1344 case eSymbolTypeAny: return eAddressClassUnknown; 1345 case eSymbolTypeAbsolute: return eAddressClassUnknown; 1346 1347 case eSymbolTypeCode: 1348 case eSymbolTypeTrampoline: 1349 case eSymbolTypeResolver: 1350 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) 1351 { 1352 // For ARM we have a bit in the n_desc field of the symbol 1353 // that tells us ARM/Thumb which is bit 0x0008. 1354 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1355 return eAddressClassCodeAlternateISA; 1356 } 1357 return eAddressClassCode; 1358 1359 case eSymbolTypeData: return eAddressClassData; 1360 case eSymbolTypeRuntime: return eAddressClassRuntime; 1361 case eSymbolTypeException: return eAddressClassRuntime; 1362 case eSymbolTypeSourceFile: return eAddressClassDebug; 1363 case eSymbolTypeHeaderFile: return eAddressClassDebug; 1364 case eSymbolTypeObjectFile: return eAddressClassDebug; 1365 case eSymbolTypeCommonBlock: return eAddressClassDebug; 1366 case eSymbolTypeBlock: return eAddressClassDebug; 1367 case eSymbolTypeLocal: return eAddressClassData; 1368 case eSymbolTypeParam: return eAddressClassData; 1369 case eSymbolTypeVariable: return eAddressClassData; 1370 case eSymbolTypeVariableType: return eAddressClassDebug; 1371 case eSymbolTypeLineEntry: return eAddressClassDebug; 1372 case eSymbolTypeLineHeader: return eAddressClassDebug; 1373 case eSymbolTypeScopeBegin: return eAddressClassDebug; 1374 case eSymbolTypeScopeEnd: return eAddressClassDebug; 1375 case eSymbolTypeAdditional: return eAddressClassUnknown; 1376 case eSymbolTypeCompiler: return eAddressClassDebug; 1377 case eSymbolTypeInstrumentation:return eAddressClassDebug; 1378 case eSymbolTypeUndefined: return eAddressClassUnknown; 1379 case eSymbolTypeObjCClass: return eAddressClassRuntime; 1380 case eSymbolTypeObjCMetaClass: return eAddressClassRuntime; 1381 case eSymbolTypeObjCIVar: return eAddressClassRuntime; 1382 case eSymbolTypeReExported: return eAddressClassRuntime; 1383 } 1384 } 1385 } 1386 return eAddressClassUnknown; 1387 } 1388 1389 Symtab * 1390 ObjectFileMachO::GetSymtab() 1391 { 1392 ModuleSP module_sp(GetModule()); 1393 if (module_sp) 1394 { 1395 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 1396 if (m_symtab_ap.get() == NULL) 1397 { 1398 m_symtab_ap.reset(new Symtab(this)); 1399 Mutex::Locker symtab_locker (m_symtab_ap->GetMutex()); 1400 ParseSymtab (); 1401 m_symtab_ap->Finalize (); 1402 } 1403 } 1404 return m_symtab_ap.get(); 1405 } 1406 1407 bool 1408 ObjectFileMachO::IsStripped () 1409 { 1410 if (m_dysymtab.cmd == 0) 1411 { 1412 ModuleSP module_sp(GetModule()); 1413 if (module_sp) 1414 { 1415 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1416 for (uint32_t i=0; i<m_header.ncmds; ++i) 1417 { 1418 const lldb::offset_t load_cmd_offset = offset; 1419 1420 load_command lc; 1421 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1422 break; 1423 if (lc.cmd == LC_DYSYMTAB) 1424 { 1425 m_dysymtab.cmd = lc.cmd; 1426 m_dysymtab.cmdsize = lc.cmdsize; 1427 if (m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == NULL) 1428 { 1429 // Clear m_dysymtab if we were unable to read all items from the load command 1430 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 1431 } 1432 } 1433 offset = load_cmd_offset + lc.cmdsize; 1434 } 1435 } 1436 } 1437 if (m_dysymtab.cmd) 1438 return m_dysymtab.nlocalsym <= 1; 1439 return false; 1440 } 1441 1442 void 1443 ObjectFileMachO::CreateSections (SectionList &unified_section_list) 1444 { 1445 if (!m_sections_ap.get()) 1446 { 1447 m_sections_ap.reset(new SectionList()); 1448 1449 const bool is_dsym = (m_header.filetype == MH_DSYM); 1450 lldb::user_id_t segID = 0; 1451 lldb::user_id_t sectID = 0; 1452 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1453 uint32_t i; 1454 const bool is_core = GetType() == eTypeCoreFile; 1455 //bool dump_sections = false; 1456 ModuleSP module_sp (GetModule()); 1457 // First look up any LC_ENCRYPTION_INFO load commands 1458 typedef RangeArray<uint32_t, uint32_t, 8> EncryptedFileRanges; 1459 EncryptedFileRanges encrypted_file_ranges; 1460 encryption_info_command encryption_cmd; 1461 for (i=0; i<m_header.ncmds; ++i) 1462 { 1463 const lldb::offset_t load_cmd_offset = offset; 1464 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1465 break; 1466 1467 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO) 1468 { 1469 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) 1470 { 1471 if (encryption_cmd.cryptid != 0) 1472 { 1473 EncryptedFileRanges::Entry entry; 1474 entry.SetRangeBase(encryption_cmd.cryptoff); 1475 entry.SetByteSize(encryption_cmd.cryptsize); 1476 encrypted_file_ranges.Append(entry); 1477 } 1478 } 1479 } 1480 offset = load_cmd_offset + encryption_cmd.cmdsize; 1481 } 1482 1483 bool section_file_addresses_changed = false; 1484 1485 offset = MachHeaderSizeFromMagic(m_header.magic); 1486 1487 struct segment_command_64 load_cmd; 1488 for (i=0; i<m_header.ncmds; ++i) 1489 { 1490 const lldb::offset_t load_cmd_offset = offset; 1491 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1492 break; 1493 1494 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1495 { 1496 if (m_data.GetU8(&offset, (uint8_t*)load_cmd.segname, 16)) 1497 { 1498 bool add_section = true; 1499 bool add_to_unified = true; 1500 ConstString const_segname (load_cmd.segname, std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 1501 1502 SectionSP unified_section_sp(unified_section_list.FindSectionByName(const_segname)); 1503 if (is_dsym && unified_section_sp) 1504 { 1505 if (const_segname == GetSegmentNameLINKEDIT()) 1506 { 1507 // We need to keep the __LINKEDIT segment private to this object file only 1508 add_to_unified = false; 1509 } 1510 else 1511 { 1512 // This is the dSYM file and this section has already been created by 1513 // the object file, no need to create it. 1514 add_section = false; 1515 } 1516 } 1517 load_cmd.vmaddr = m_data.GetAddress(&offset); 1518 load_cmd.vmsize = m_data.GetAddress(&offset); 1519 load_cmd.fileoff = m_data.GetAddress(&offset); 1520 load_cmd.filesize = m_data.GetAddress(&offset); 1521 if (m_length != 0 && load_cmd.filesize != 0) 1522 { 1523 if (load_cmd.fileoff > m_length) 1524 { 1525 // We have a load command that says it extends past the end of the file. This is likely 1526 // a corrupt file. We don't have any way to return an error condition here (this method 1527 // was likely invoked from something like ObjectFile::GetSectionList()) -- all we can do 1528 // is null out the SectionList vector and if a process has been set up, dump a message 1529 // to stdout. The most common case here is core file debugging with a truncated file. 1530 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1531 module_sp->ReportWarning("load command %u %s has a fileoff (0x%" PRIx64 ") that extends beyond the end of the file (0x%" PRIx64 "), ignoring this section", 1532 i, 1533 lc_segment_name, 1534 load_cmd.fileoff, 1535 m_length); 1536 1537 load_cmd.fileoff = 0; 1538 load_cmd.filesize = 0; 1539 } 1540 1541 if (load_cmd.fileoff + load_cmd.filesize > m_length) 1542 { 1543 // We have a load command that says it extends past the end of the file. This is likely 1544 // a corrupt file. We don't have any way to return an error condition here (this method 1545 // was likely invoked from something like ObjectFile::GetSectionList()) -- all we can do 1546 // is null out the SectionList vector and if a process has been set up, dump a message 1547 // to stdout. The most common case here is core file debugging with a truncated file. 1548 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1549 GetModule()->ReportWarning("load command %u %s has a fileoff + filesize (0x%" PRIx64 ") that extends beyond the end of the file (0x%" PRIx64 "), the segment will be truncated to match", 1550 i, 1551 lc_segment_name, 1552 load_cmd.fileoff + load_cmd.filesize, 1553 m_length); 1554 1555 // Tuncase the length 1556 load_cmd.filesize = m_length - load_cmd.fileoff; 1557 } 1558 } 1559 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1560 { 1561 1562 const bool segment_is_encrypted = (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1563 1564 // Keep a list of mach segments around in case we need to 1565 // get at data that isn't stored in the abstracted Sections. 1566 m_mach_segments.push_back (load_cmd); 1567 1568 // Use a segment ID of the segment index shifted left by 8 so they 1569 // never conflict with any of the sections. 1570 SectionSP segment_sp; 1571 if (add_section && (const_segname || is_core)) 1572 { 1573 segment_sp.reset(new Section (module_sp, // Module to which this section belongs 1574 this, // Object file to which this sections belongs 1575 ++segID << 8, // Section ID is the 1 based segment index shifted right by 8 bits as not to collide with any of the 256 section IDs that are possible 1576 const_segname, // Name of this section 1577 eSectionTypeContainer, // This section is a container of other sections. 1578 load_cmd.vmaddr, // File VM address == addresses as they are found in the object file 1579 load_cmd.vmsize, // VM size in bytes of this section 1580 load_cmd.fileoff, // Offset to the data for this section in the file 1581 load_cmd.filesize, // Size in bytes of this section as found in the file 1582 0, // Segments have no alignment information 1583 load_cmd.flags)); // Flags for this section 1584 1585 segment_sp->SetIsEncrypted (segment_is_encrypted); 1586 m_sections_ap->AddSection(segment_sp); 1587 if (add_to_unified) 1588 unified_section_list.AddSection(segment_sp); 1589 } 1590 else if (unified_section_sp) 1591 { 1592 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) 1593 { 1594 // Check to see if the module was read from memory? 1595 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) 1596 { 1597 // We have a module that is in memory and needs to have its 1598 // file address adjusted. We need to do this because when we 1599 // load a file from memory, its addresses will be slid already, 1600 // yet the addresses in the new symbol file will still be unslid. 1601 // Since everything is stored as section offset, this shouldn't 1602 // cause any problems. 1603 1604 // Make sure we've parsed the symbol table from the 1605 // ObjectFile before we go around changing its Sections. 1606 module_sp->GetObjectFile()->GetSymtab(); 1607 // eh_frame would present the same problems but we parse that on 1608 // a per-function basis as-needed so it's more difficult to 1609 // remove its use of the Sections. Realistically, the environments 1610 // where this code path will be taken will not have eh_frame sections. 1611 1612 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1613 1614 // Notify the module that the section addresses have been changed once 1615 // we're done so any file-address caches can be updated. 1616 section_file_addresses_changed = true; 1617 } 1618 } 1619 m_sections_ap->AddSection(unified_section_sp); 1620 } 1621 1622 struct section_64 sect64; 1623 ::memset (§64, 0, sizeof(sect64)); 1624 // Push a section into our mach sections for the section at 1625 // index zero (NO_SECT) if we don't have any mach sections yet... 1626 if (m_mach_sections.empty()) 1627 m_mach_sections.push_back(sect64); 1628 uint32_t segment_sect_idx; 1629 const lldb::user_id_t first_segment_sectID = sectID + 1; 1630 1631 1632 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1633 for (segment_sect_idx=0; segment_sect_idx<load_cmd.nsects; ++segment_sect_idx) 1634 { 1635 if (m_data.GetU8(&offset, (uint8_t*)sect64.sectname, sizeof(sect64.sectname)) == NULL) 1636 break; 1637 if (m_data.GetU8(&offset, (uint8_t*)sect64.segname, sizeof(sect64.segname)) == NULL) 1638 break; 1639 sect64.addr = m_data.GetAddress(&offset); 1640 sect64.size = m_data.GetAddress(&offset); 1641 1642 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1643 break; 1644 1645 // Keep a list of mach sections around in case we need to 1646 // get at data that isn't stored in the abstracted Sections. 1647 m_mach_sections.push_back (sect64); 1648 1649 if (add_section) 1650 { 1651 ConstString section_name (sect64.sectname, std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1652 if (!const_segname) 1653 { 1654 // We have a segment with no name so we need to conjure up 1655 // segments that correspond to the section's segname if there 1656 // isn't already such a section. If there is such a section, 1657 // we resize the section so that it spans all sections. 1658 // We also mark these sections as fake so address matches don't 1659 // hit if they land in the gaps between the child sections. 1660 const_segname.SetTrimmedCStringWithLength(sect64.segname, sizeof(sect64.segname)); 1661 segment_sp = unified_section_list.FindSectionByName (const_segname); 1662 if (segment_sp.get()) 1663 { 1664 Section *segment = segment_sp.get(); 1665 // Grow the section size as needed. 1666 const lldb::addr_t sect64_min_addr = sect64.addr; 1667 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1668 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1669 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1670 const lldb::addr_t curr_seg_max_addr = curr_seg_min_addr + curr_seg_byte_size; 1671 if (sect64_min_addr >= curr_seg_min_addr) 1672 { 1673 const lldb::addr_t new_seg_byte_size = sect64_max_addr - curr_seg_min_addr; 1674 // Only grow the section size if needed 1675 if (new_seg_byte_size > curr_seg_byte_size) 1676 segment->SetByteSize (new_seg_byte_size); 1677 } 1678 else 1679 { 1680 // We need to change the base address of the segment and 1681 // adjust the child section offsets for all existing children. 1682 const lldb::addr_t slide_amount = sect64_min_addr - curr_seg_min_addr; 1683 segment->Slide(slide_amount, false); 1684 segment->GetChildren().Slide(-slide_amount, false); 1685 segment->SetByteSize (curr_seg_max_addr - sect64_min_addr); 1686 } 1687 1688 // Grow the section size as needed. 1689 if (sect64.offset) 1690 { 1691 const lldb::addr_t segment_min_file_offset = segment->GetFileOffset(); 1692 const lldb::addr_t segment_max_file_offset = segment_min_file_offset + segment->GetFileSize(); 1693 1694 const lldb::addr_t section_min_file_offset = sect64.offset; 1695 const lldb::addr_t section_max_file_offset = section_min_file_offset + sect64.size; 1696 const lldb::addr_t new_file_offset = std::min (section_min_file_offset, segment_min_file_offset); 1697 const lldb::addr_t new_file_size = std::max (section_max_file_offset, segment_max_file_offset) - new_file_offset; 1698 segment->SetFileOffset (new_file_offset); 1699 segment->SetFileSize (new_file_size); 1700 } 1701 } 1702 else 1703 { 1704 // Create a fake section for the section's named segment 1705 segment_sp.reset(new Section (segment_sp, // Parent section 1706 module_sp, // Module to which this section belongs 1707 this, // Object file to which this section belongs 1708 ++segID << 8, // Section ID is the 1 based segment index shifted right by 8 bits as not to collide with any of the 256 section IDs that are possible 1709 const_segname, // Name of this section 1710 eSectionTypeContainer, // This section is a container of other sections. 1711 sect64.addr, // File VM address == addresses as they are found in the object file 1712 sect64.size, // VM size in bytes of this section 1713 sect64.offset, // Offset to the data for this section in the file 1714 sect64.offset ? sect64.size : 0, // Size in bytes of this section as found in the file 1715 sect64.align, 1716 load_cmd.flags)); // Flags for this section 1717 segment_sp->SetIsFake(true); 1718 1719 m_sections_ap->AddSection(segment_sp); 1720 if (add_to_unified) 1721 unified_section_list.AddSection(segment_sp); 1722 segment_sp->SetIsEncrypted (segment_is_encrypted); 1723 } 1724 } 1725 assert (segment_sp.get()); 1726 1727 lldb::SectionType sect_type = eSectionTypeOther; 1728 1729 if (sect64.flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1730 sect_type = eSectionTypeCode; 1731 else 1732 { 1733 uint32_t mach_sect_type = sect64.flags & SECTION_TYPE; 1734 static ConstString g_sect_name_objc_data ("__objc_data"); 1735 static ConstString g_sect_name_objc_msgrefs ("__objc_msgrefs"); 1736 static ConstString g_sect_name_objc_selrefs ("__objc_selrefs"); 1737 static ConstString g_sect_name_objc_classrefs ("__objc_classrefs"); 1738 static ConstString g_sect_name_objc_superrefs ("__objc_superrefs"); 1739 static ConstString g_sect_name_objc_const ("__objc_const"); 1740 static ConstString g_sect_name_objc_classlist ("__objc_classlist"); 1741 static ConstString g_sect_name_cfstring ("__cfstring"); 1742 1743 static ConstString g_sect_name_dwarf_debug_abbrev ("__debug_abbrev"); 1744 static ConstString g_sect_name_dwarf_debug_aranges ("__debug_aranges"); 1745 static ConstString g_sect_name_dwarf_debug_frame ("__debug_frame"); 1746 static ConstString g_sect_name_dwarf_debug_info ("__debug_info"); 1747 static ConstString g_sect_name_dwarf_debug_line ("__debug_line"); 1748 static ConstString g_sect_name_dwarf_debug_loc ("__debug_loc"); 1749 static ConstString g_sect_name_dwarf_debug_macinfo ("__debug_macinfo"); 1750 static ConstString g_sect_name_dwarf_debug_pubnames ("__debug_pubnames"); 1751 static ConstString g_sect_name_dwarf_debug_pubtypes ("__debug_pubtypes"); 1752 static ConstString g_sect_name_dwarf_debug_ranges ("__debug_ranges"); 1753 static ConstString g_sect_name_dwarf_debug_str ("__debug_str"); 1754 static ConstString g_sect_name_dwarf_apple_names ("__apple_names"); 1755 static ConstString g_sect_name_dwarf_apple_types ("__apple_types"); 1756 static ConstString g_sect_name_dwarf_apple_namespaces ("__apple_namespac"); 1757 static ConstString g_sect_name_dwarf_apple_objc ("__apple_objc"); 1758 static ConstString g_sect_name_eh_frame ("__eh_frame"); 1759 static ConstString g_sect_name_compact_unwind ("__unwind_info"); 1760 static ConstString g_sect_name_text ("__text"); 1761 static ConstString g_sect_name_data ("__data"); 1762 1763 1764 if (section_name == g_sect_name_dwarf_debug_abbrev) 1765 sect_type = eSectionTypeDWARFDebugAbbrev; 1766 else if (section_name == g_sect_name_dwarf_debug_aranges) 1767 sect_type = eSectionTypeDWARFDebugAranges; 1768 else if (section_name == g_sect_name_dwarf_debug_frame) 1769 sect_type = eSectionTypeDWARFDebugFrame; 1770 else if (section_name == g_sect_name_dwarf_debug_info) 1771 sect_type = eSectionTypeDWARFDebugInfo; 1772 else if (section_name == g_sect_name_dwarf_debug_line) 1773 sect_type = eSectionTypeDWARFDebugLine; 1774 else if (section_name == g_sect_name_dwarf_debug_loc) 1775 sect_type = eSectionTypeDWARFDebugLoc; 1776 else if (section_name == g_sect_name_dwarf_debug_macinfo) 1777 sect_type = eSectionTypeDWARFDebugMacInfo; 1778 else if (section_name == g_sect_name_dwarf_debug_pubnames) 1779 sect_type = eSectionTypeDWARFDebugPubNames; 1780 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 1781 sect_type = eSectionTypeDWARFDebugPubTypes; 1782 else if (section_name == g_sect_name_dwarf_debug_ranges) 1783 sect_type = eSectionTypeDWARFDebugRanges; 1784 else if (section_name == g_sect_name_dwarf_debug_str) 1785 sect_type = eSectionTypeDWARFDebugStr; 1786 else if (section_name == g_sect_name_dwarf_apple_names) 1787 sect_type = eSectionTypeDWARFAppleNames; 1788 else if (section_name == g_sect_name_dwarf_apple_types) 1789 sect_type = eSectionTypeDWARFAppleTypes; 1790 else if (section_name == g_sect_name_dwarf_apple_namespaces) 1791 sect_type = eSectionTypeDWARFAppleNamespaces; 1792 else if (section_name == g_sect_name_dwarf_apple_objc) 1793 sect_type = eSectionTypeDWARFAppleObjC; 1794 else if (section_name == g_sect_name_objc_selrefs) 1795 sect_type = eSectionTypeDataCStringPointers; 1796 else if (section_name == g_sect_name_objc_msgrefs) 1797 sect_type = eSectionTypeDataObjCMessageRefs; 1798 else if (section_name == g_sect_name_eh_frame) 1799 sect_type = eSectionTypeEHFrame; 1800 else if (section_name == g_sect_name_compact_unwind) 1801 sect_type = eSectionTypeCompactUnwind; 1802 else if (section_name == g_sect_name_cfstring) 1803 sect_type = eSectionTypeDataObjCCFStrings; 1804 else if (section_name == g_sect_name_objc_data || 1805 section_name == g_sect_name_objc_classrefs || 1806 section_name == g_sect_name_objc_superrefs || 1807 section_name == g_sect_name_objc_const || 1808 section_name == g_sect_name_objc_classlist) 1809 { 1810 sect_type = eSectionTypeDataPointers; 1811 } 1812 1813 if (sect_type == eSectionTypeOther) 1814 { 1815 switch (mach_sect_type) 1816 { 1817 // TODO: categorize sections by other flags for regular sections 1818 case S_REGULAR: 1819 if (section_name == g_sect_name_text) 1820 sect_type = eSectionTypeCode; 1821 else if (section_name == g_sect_name_data) 1822 sect_type = eSectionTypeData; 1823 else 1824 sect_type = eSectionTypeOther; 1825 break; 1826 case S_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1827 case S_CSTRING_LITERALS: sect_type = eSectionTypeDataCString; break; // section with only literal C strings 1828 case S_4BYTE_LITERALS: sect_type = eSectionTypeData4; break; // section with only 4 byte literals 1829 case S_8BYTE_LITERALS: sect_type = eSectionTypeData8; break; // section with only 8 byte literals 1830 case S_LITERAL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only pointers to literals 1831 case S_NON_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only non-lazy symbol pointers 1832 case S_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only lazy symbol pointers 1833 case S_SYMBOL_STUBS: sect_type = eSectionTypeCode; break; // section with only symbol stubs, byte size of stub in the reserved2 field 1834 case S_MOD_INIT_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for initialization 1835 case S_MOD_TERM_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for termination 1836 case S_COALESCED: sect_type = eSectionTypeOther; break; 1837 case S_GB_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1838 case S_INTERPOSING: sect_type = eSectionTypeCode; break; // section with only pairs of function pointers for interposing 1839 case S_16BYTE_LITERALS: sect_type = eSectionTypeData16; break; // section with only 16 byte literals 1840 case S_DTRACE_DOF: sect_type = eSectionTypeDebug; break; 1841 case S_LAZY_DYLIB_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; 1842 default: break; 1843 } 1844 } 1845 } 1846 1847 SectionSP section_sp(new Section (segment_sp, 1848 module_sp, 1849 this, 1850 ++sectID, 1851 section_name, 1852 sect_type, 1853 sect64.addr - segment_sp->GetFileAddress(), 1854 sect64.size, 1855 sect64.offset, 1856 sect64.offset == 0 ? 0 : sect64.size, 1857 sect64.align, 1858 sect64.flags)); 1859 // Set the section to be encrypted to match the segment 1860 1861 bool section_is_encrypted = false; 1862 if (!segment_is_encrypted && load_cmd.filesize != 0) 1863 section_is_encrypted = encrypted_file_ranges.FindEntryThatContains(sect64.offset) != NULL; 1864 1865 section_sp->SetIsEncrypted (segment_is_encrypted || section_is_encrypted); 1866 segment_sp->GetChildren().AddSection(section_sp); 1867 1868 if (segment_sp->IsFake()) 1869 { 1870 segment_sp.reset(); 1871 const_segname.Clear(); 1872 } 1873 } 1874 } 1875 if (segment_sp && is_dsym) 1876 { 1877 if (first_segment_sectID <= sectID) 1878 { 1879 lldb::user_id_t sect_uid; 1880 for (sect_uid = first_segment_sectID; sect_uid <= sectID; ++sect_uid) 1881 { 1882 SectionSP curr_section_sp(segment_sp->GetChildren().FindSectionByID (sect_uid)); 1883 SectionSP next_section_sp; 1884 if (sect_uid + 1 <= sectID) 1885 next_section_sp = segment_sp->GetChildren().FindSectionByID (sect_uid+1); 1886 1887 if (curr_section_sp.get()) 1888 { 1889 if (curr_section_sp->GetByteSize() == 0) 1890 { 1891 if (next_section_sp.get() != NULL) 1892 curr_section_sp->SetByteSize ( next_section_sp->GetFileAddress() - curr_section_sp->GetFileAddress() ); 1893 else 1894 curr_section_sp->SetByteSize ( load_cmd.vmsize ); 1895 } 1896 } 1897 } 1898 } 1899 } 1900 } 1901 } 1902 } 1903 else if (load_cmd.cmd == LC_DYSYMTAB) 1904 { 1905 m_dysymtab.cmd = load_cmd.cmd; 1906 m_dysymtab.cmdsize = load_cmd.cmdsize; 1907 m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1908 } 1909 1910 offset = load_cmd_offset + load_cmd.cmdsize; 1911 } 1912 1913 1914 if (section_file_addresses_changed && module_sp.get()) 1915 { 1916 module_sp->SectionFileAddressesChanged(); 1917 } 1918 } 1919 } 1920 1921 class MachSymtabSectionInfo 1922 { 1923 public: 1924 1925 MachSymtabSectionInfo (SectionList *section_list) : 1926 m_section_list (section_list), 1927 m_section_infos() 1928 { 1929 // Get the number of sections down to a depth of 1 to include 1930 // all segments and their sections, but no other sections that 1931 // may be added for debug map or 1932 m_section_infos.resize(section_list->GetNumSections(1)); 1933 } 1934 1935 1936 SectionSP 1937 GetSection (uint8_t n_sect, addr_t file_addr) 1938 { 1939 if (n_sect == 0) 1940 return SectionSP(); 1941 if (n_sect < m_section_infos.size()) 1942 { 1943 if (!m_section_infos[n_sect].section_sp) 1944 { 1945 SectionSP section_sp (m_section_list->FindSectionByID (n_sect)); 1946 m_section_infos[n_sect].section_sp = section_sp; 1947 if (section_sp) 1948 { 1949 m_section_infos[n_sect].vm_range.SetBaseAddress (section_sp->GetFileAddress()); 1950 m_section_infos[n_sect].vm_range.SetByteSize (section_sp->GetByteSize()); 1951 } 1952 else 1953 { 1954 Host::SystemLog (Host::eSystemLogError, "error: unable to find section for section %u\n", n_sect); 1955 } 1956 } 1957 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) 1958 { 1959 // Symbol is in section. 1960 return m_section_infos[n_sect].section_sp; 1961 } 1962 else if (m_section_infos[n_sect].vm_range.GetByteSize () == 0 && 1963 m_section_infos[n_sect].vm_range.GetBaseAddress() == file_addr) 1964 { 1965 // Symbol is in section with zero size, but has the same start 1966 // address as the section. This can happen with linker symbols 1967 // (symbols that start with the letter 'l' or 'L'. 1968 return m_section_infos[n_sect].section_sp; 1969 } 1970 } 1971 return m_section_list->FindSectionContainingFileAddress(file_addr); 1972 } 1973 1974 protected: 1975 struct SectionInfo 1976 { 1977 SectionInfo () : 1978 vm_range(), 1979 section_sp () 1980 { 1981 } 1982 1983 VMRange vm_range; 1984 SectionSP section_sp; 1985 }; 1986 SectionList *m_section_list; 1987 std::vector<SectionInfo> m_section_infos; 1988 }; 1989 1990 struct TrieEntry 1991 { 1992 TrieEntry () : 1993 name(), 1994 address(LLDB_INVALID_ADDRESS), 1995 flags (0), 1996 other(0), 1997 import_name() 1998 { 1999 } 2000 2001 void 2002 Clear () 2003 { 2004 name.Clear(); 2005 address = LLDB_INVALID_ADDRESS; 2006 flags = 0; 2007 other = 0; 2008 import_name.Clear(); 2009 } 2010 2011 void 2012 Dump () const 2013 { 2014 printf ("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 2015 static_cast<unsigned long long>(address), 2016 static_cast<unsigned long long>(flags), 2017 static_cast<unsigned long long>(other), name.GetCString()); 2018 if (import_name) 2019 printf (" -> \"%s\"\n", import_name.GetCString()); 2020 else 2021 printf ("\n"); 2022 } 2023 ConstString name; 2024 uint64_t address; 2025 uint64_t flags; 2026 uint64_t other; 2027 ConstString import_name; 2028 }; 2029 2030 struct TrieEntryWithOffset 2031 { 2032 lldb::offset_t nodeOffset; 2033 TrieEntry entry; 2034 2035 TrieEntryWithOffset (lldb::offset_t offset) : 2036 nodeOffset (offset), 2037 entry() 2038 { 2039 } 2040 2041 void 2042 Dump (uint32_t idx) const 2043 { 2044 printf ("[%3u] 0x%16.16llx: ", idx, 2045 static_cast<unsigned long long>(nodeOffset)); 2046 entry.Dump(); 2047 } 2048 2049 bool 2050 operator<(const TrieEntryWithOffset& other) const 2051 { 2052 return ( nodeOffset < other.nodeOffset ); 2053 } 2054 }; 2055 2056 static void 2057 ParseTrieEntries (DataExtractor &data, 2058 lldb::offset_t offset, 2059 std::vector<llvm::StringRef> &nameSlices, 2060 std::set<lldb::addr_t> &resolver_addresses, 2061 std::vector<TrieEntryWithOffset>& output) 2062 { 2063 if (!data.ValidOffset(offset)) 2064 return; 2065 2066 const uint64_t terminalSize = data.GetULEB128(&offset); 2067 lldb::offset_t children_offset = offset + terminalSize; 2068 if ( terminalSize != 0 ) { 2069 TrieEntryWithOffset e (offset); 2070 e.entry.flags = data.GetULEB128(&offset); 2071 const char *import_name = NULL; 2072 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT ) { 2073 e.entry.address = 0; 2074 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2075 import_name = data.GetCStr(&offset); 2076 } 2077 else { 2078 e.entry.address = data.GetULEB128(&offset); 2079 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER ) 2080 { 2081 //resolver_addresses.insert(e.entry.address); 2082 e.entry.other = data.GetULEB128(&offset); 2083 resolver_addresses.insert(e.entry.other); 2084 } 2085 else 2086 e.entry.other = 0; 2087 } 2088 // Only add symbols that are reexport symbols with a valid import name 2089 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && import_name[0]) 2090 { 2091 std::string name; 2092 if (!nameSlices.empty()) 2093 { 2094 for (auto name_slice: nameSlices) 2095 name.append(name_slice.data(), name_slice.size()); 2096 } 2097 if (name.size() > 1) 2098 { 2099 // Skip the leading '_' 2100 e.entry.name.SetCStringWithLength(name.c_str() + 1,name.size() - 1); 2101 } 2102 if (import_name) 2103 { 2104 // Skip the leading '_' 2105 e.entry.import_name.SetCString(import_name+1); 2106 } 2107 output.push_back(e); 2108 } 2109 } 2110 2111 const uint8_t childrenCount = data.GetU8(&children_offset); 2112 for (uint8_t i=0; i < childrenCount; ++i) { 2113 nameSlices.push_back(data.GetCStr(&children_offset)); 2114 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2115 if (childNodeOffset) 2116 { 2117 ParseTrieEntries(data, 2118 childNodeOffset, 2119 nameSlices, 2120 resolver_addresses, 2121 output); 2122 } 2123 nameSlices.pop_back(); 2124 } 2125 } 2126 2127 size_t 2128 ObjectFileMachO::ParseSymtab () 2129 { 2130 Timer scoped_timer(__PRETTY_FUNCTION__, 2131 "ObjectFileMachO::ParseSymtab () module = %s", 2132 m_file.GetFilename().AsCString("")); 2133 ModuleSP module_sp (GetModule()); 2134 if (!module_sp) 2135 return 0; 2136 2137 struct symtab_command symtab_load_command = { 0, 0, 0, 0, 0, 0 }; 2138 struct linkedit_data_command function_starts_load_command = { 0, 0, 0, 0 }; 2139 struct dyld_info_command dyld_info = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; 2140 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2141 FunctionStarts function_starts; 2142 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2143 uint32_t i; 2144 FileSpecList dylib_files; 2145 Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_SYMBOLS)); 2146 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 2147 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 2148 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 2149 2150 for (i=0; i<m_header.ncmds; ++i) 2151 { 2152 const lldb::offset_t cmd_offset = offset; 2153 // Read in the load command and load command size 2154 struct load_command lc; 2155 if (m_data.GetU32(&offset, &lc, 2) == NULL) 2156 break; 2157 // Watch for the symbol table load command 2158 switch (lc.cmd) 2159 { 2160 case LC_SYMTAB: 2161 symtab_load_command.cmd = lc.cmd; 2162 symtab_load_command.cmdsize = lc.cmdsize; 2163 // Read in the rest of the symtab load command 2164 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 0) // fill in symoff, nsyms, stroff, strsize fields 2165 return 0; 2166 if (symtab_load_command.symoff == 0) 2167 { 2168 if (log) 2169 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2170 return 0; 2171 } 2172 2173 if (symtab_load_command.stroff == 0) 2174 { 2175 if (log) 2176 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2177 return 0; 2178 } 2179 2180 if (symtab_load_command.nsyms == 0) 2181 { 2182 if (log) 2183 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2184 return 0; 2185 } 2186 2187 if (symtab_load_command.strsize == 0) 2188 { 2189 if (log) 2190 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2191 return 0; 2192 } 2193 break; 2194 2195 case LC_DYLD_INFO: 2196 case LC_DYLD_INFO_ONLY: 2197 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) 2198 { 2199 dyld_info.cmd = lc.cmd; 2200 dyld_info.cmdsize = lc.cmdsize; 2201 } 2202 else 2203 { 2204 memset (&dyld_info, 0, sizeof(dyld_info)); 2205 } 2206 break; 2207 2208 case LC_LOAD_DYLIB: 2209 case LC_LOAD_WEAK_DYLIB: 2210 case LC_REEXPORT_DYLIB: 2211 case LC_LOADFVMLIB: 2212 case LC_LOAD_UPWARD_DYLIB: 2213 { 2214 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2215 const char *path = m_data.PeekCStr(name_offset); 2216 if (path) 2217 { 2218 FileSpec file_spec(path, false); 2219 // Strip the path if there is @rpath, @executable, etc so we just use the basename 2220 if (path[0] == '@') 2221 file_spec.GetDirectory().Clear(); 2222 2223 if (lc.cmd == LC_REEXPORT_DYLIB) 2224 { 2225 m_reexported_dylibs.AppendIfUnique(file_spec); 2226 } 2227 2228 dylib_files.Append(file_spec); 2229 } 2230 } 2231 break; 2232 2233 case LC_FUNCTION_STARTS: 2234 function_starts_load_command.cmd = lc.cmd; 2235 function_starts_load_command.cmdsize = lc.cmdsize; 2236 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == NULL) // fill in symoff, nsyms, stroff, strsize fields 2237 memset (&function_starts_load_command, 0, sizeof(function_starts_load_command)); 2238 break; 2239 2240 default: 2241 break; 2242 } 2243 offset = cmd_offset + lc.cmdsize; 2244 } 2245 2246 if (symtab_load_command.cmd) 2247 { 2248 Symtab *symtab = m_symtab_ap.get(); 2249 SectionList *section_list = GetSectionList(); 2250 if (section_list == NULL) 2251 return 0; 2252 2253 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2254 const ByteOrder byte_order = m_data.GetByteOrder(); 2255 bool bit_width_32 = addr_byte_size == 4; 2256 const size_t nlist_byte_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2257 2258 DataExtractor nlist_data (NULL, 0, byte_order, addr_byte_size); 2259 DataExtractor strtab_data (NULL, 0, byte_order, addr_byte_size); 2260 DataExtractor function_starts_data (NULL, 0, byte_order, addr_byte_size); 2261 DataExtractor indirect_symbol_index_data (NULL, 0, byte_order, addr_byte_size); 2262 DataExtractor dyld_trie_data (NULL, 0, byte_order, addr_byte_size); 2263 2264 const addr_t nlist_data_byte_size = symtab_load_command.nsyms * nlist_byte_size; 2265 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2266 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2267 2268 ProcessSP process_sp (m_process_wp.lock()); 2269 Process *process = process_sp.get(); 2270 2271 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2272 2273 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) 2274 { 2275 Target &target = process->GetTarget(); 2276 2277 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2278 2279 SectionSP linkedit_section_sp(section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2280 // Reading mach file from memory in a process or core file... 2281 2282 if (linkedit_section_sp) 2283 { 2284 const addr_t linkedit_load_addr = linkedit_section_sp->GetLoadBaseAddress(&target); 2285 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2286 const addr_t symoff_addr = linkedit_load_addr + symtab_load_command.symoff - linkedit_file_offset; 2287 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - linkedit_file_offset; 2288 2289 bool data_was_read = false; 2290 2291 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 2292 if (m_header.flags & 0x80000000u && process->GetAddressByteSize() == sizeof (void*)) 2293 { 2294 // This mach-o memory file is in the dyld shared cache. If this 2295 // program is not remote and this is iOS, then this process will 2296 // share the same shared cache as the process we are debugging and 2297 // we can read the entire __LINKEDIT from the address space in this 2298 // process. This is a needed optimization that is used for local iOS 2299 // debugging only since all shared libraries in the shared cache do 2300 // not have corresponding files that exist in the file system of the 2301 // device. They have been combined into a single file. This means we 2302 // always have to load these files from memory. All of the symbol and 2303 // string tables from all of the __LINKEDIT sections from the shared 2304 // libraries in the shared cache have been merged into a single large 2305 // symbol and string table. Reading all of this symbol and string table 2306 // data across can slow down debug launch times, so we optimize this by 2307 // reading the memory for the __LINKEDIT section from this process. 2308 2309 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 2310 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 2311 bool use_lldb_cache = true; 2312 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && lldb_shared_cache != process_shared_cache) 2313 { 2314 use_lldb_cache = false; 2315 ModuleSP module_sp (GetModule()); 2316 if (module_sp) 2317 module_sp->ReportWarning ("shared cache in process does not match lldb's own shared cache, startup will be slow."); 2318 2319 } 2320 2321 PlatformSP platform_sp (target.GetPlatform()); 2322 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) 2323 { 2324 data_was_read = true; 2325 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, eByteOrderLittle); 2326 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, eByteOrderLittle); 2327 if (function_starts_load_command.cmd) 2328 { 2329 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 2330 function_starts_data.SetData ((void *)func_start_addr, function_starts_load_command.datasize, eByteOrderLittle); 2331 } 2332 } 2333 } 2334 #endif 2335 2336 if (!data_was_read) 2337 { 2338 if (memory_module_load_level == eMemoryModuleLoadLevelComplete) 2339 { 2340 DataBufferSP nlist_data_sp (ReadMemory (process_sp, symoff_addr, nlist_data_byte_size)); 2341 if (nlist_data_sp) 2342 nlist_data.SetData (nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2343 // Load strings individually from memory when loading from memory since shared cache 2344 // string tables contain strings for all symbols from all shared cached libraries 2345 //DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, strtab_data_byte_size)); 2346 //if (strtab_data_sp) 2347 // strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 2348 if (m_dysymtab.nindirectsyms != 0) 2349 { 2350 const addr_t indirect_syms_addr = linkedit_load_addr + m_dysymtab.indirectsymoff - linkedit_file_offset; 2351 DataBufferSP indirect_syms_data_sp (ReadMemory (process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 2352 if (indirect_syms_data_sp) 2353 indirect_symbol_index_data.SetData (indirect_syms_data_sp, 0, indirect_syms_data_sp->GetByteSize()); 2354 } 2355 } 2356 2357 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) 2358 { 2359 if (function_starts_load_command.cmd) 2360 { 2361 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 2362 DataBufferSP func_start_data_sp (ReadMemory (process_sp, func_start_addr, function_starts_load_command.datasize)); 2363 if (func_start_data_sp) 2364 function_starts_data.SetData (func_start_data_sp, 0, func_start_data_sp->GetByteSize()); 2365 } 2366 } 2367 } 2368 } 2369 } 2370 else 2371 { 2372 nlist_data.SetData (m_data, 2373 symtab_load_command.symoff, 2374 nlist_data_byte_size); 2375 strtab_data.SetData (m_data, 2376 symtab_load_command.stroff, 2377 strtab_data_byte_size); 2378 2379 if (dyld_info.export_size > 0) 2380 { 2381 dyld_trie_data.SetData (m_data, 2382 dyld_info.export_off, 2383 dyld_info.export_size); 2384 } 2385 2386 if (m_dysymtab.nindirectsyms != 0) 2387 { 2388 indirect_symbol_index_data.SetData (m_data, 2389 m_dysymtab.indirectsymoff, 2390 m_dysymtab.nindirectsyms * 4); 2391 } 2392 if (function_starts_load_command.cmd) 2393 { 2394 function_starts_data.SetData (m_data, 2395 function_starts_load_command.dataoff, 2396 function_starts_load_command.datasize); 2397 } 2398 } 2399 2400 if (nlist_data.GetByteSize() == 0 && memory_module_load_level == eMemoryModuleLoadLevelComplete) 2401 { 2402 if (log) 2403 module_sp->LogMessage(log, "failed to read nlist data"); 2404 return 0; 2405 } 2406 2407 2408 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2409 if (!have_strtab_data) 2410 { 2411 if (process) 2412 { 2413 if (strtab_addr == LLDB_INVALID_ADDRESS) 2414 { 2415 if (log) 2416 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2417 return 0; 2418 } 2419 } 2420 else 2421 { 2422 if (log) 2423 module_sp->LogMessage(log, "failed to read strtab data"); 2424 return 0; 2425 } 2426 } 2427 2428 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2429 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2430 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2431 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2432 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 2433 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 2434 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 2435 SectionSP eh_frame_section_sp; 2436 if (text_section_sp.get()) 2437 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 2438 else 2439 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 2440 2441 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2442 2443 // lldb works best if it knows the start address of all functions in a module. 2444 // Linker symbols or debug info are normally the best source of information for start addr / size but 2445 // they may be stripped in a released binary. 2446 // Two additional sources of information exist in Mach-O binaries: 2447 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each function's start address in the 2448 // binary, relative to the text section. 2449 // eh_frame - the eh_frame FDEs have the start addr & size of each function 2450 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on all modern binaries. 2451 // Binaries built to run on older releases may need to use eh_frame information. 2452 2453 if (text_section_sp && function_starts_data.GetByteSize()) 2454 { 2455 FunctionStarts::Entry function_start_entry; 2456 function_start_entry.data = false; 2457 lldb::offset_t function_start_offset = 0; 2458 function_start_entry.addr = text_section_sp->GetFileAddress(); 2459 uint64_t delta; 2460 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 0) 2461 { 2462 // Now append the current entry 2463 function_start_entry.addr += delta; 2464 function_starts.Append(function_start_entry); 2465 } 2466 } 2467 else 2468 { 2469 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the load command claiming an eh_frame 2470 // but it doesn't actually have the eh_frame content. And if we have a dSYM, we don't need to do any 2471 // of this fill-in-the-missing-symbols works anyway - the debug info should give us all the functions in 2472 // the module. 2473 if (text_section_sp.get() && eh_frame_section_sp.get() && m_type != eTypeDebugInfo) 2474 { 2475 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, eRegisterKindGCC, true); 2476 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2477 eh_frame.GetFunctionAddressAndSizeVector (functions); 2478 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2479 size_t count = functions.GetSize(); 2480 for (size_t i = 0; i < count; ++i) 2481 { 2482 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = functions.GetEntryAtIndex (i); 2483 if (func) 2484 { 2485 FunctionStarts::Entry function_start_entry; 2486 function_start_entry.addr = func->base - text_base_addr; 2487 function_starts.Append(function_start_entry); 2488 } 2489 } 2490 } 2491 } 2492 2493 const size_t function_starts_count = function_starts.GetSize(); 2494 2495 const user_id_t TEXT_eh_frame_sectID = 2496 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2497 : static_cast<user_id_t>(NO_SECT); 2498 2499 lldb::offset_t nlist_data_offset = 0; 2500 2501 uint32_t N_SO_index = UINT32_MAX; 2502 2503 MachSymtabSectionInfo section_info (section_list); 2504 std::vector<uint32_t> N_FUN_indexes; 2505 std::vector<uint32_t> N_NSYM_indexes; 2506 std::vector<uint32_t> N_INCL_indexes; 2507 std::vector<uint32_t> N_BRAC_indexes; 2508 std::vector<uint32_t> N_COMM_indexes; 2509 typedef std::multimap <uint64_t, uint32_t> ValueToSymbolIndexMap; 2510 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2511 typedef std::map <const char *, uint32_t> ConstNameToSymbolIndexMap; 2512 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2513 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2514 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2515 // Any symbols that get merged into another will get an entry 2516 // in this map so we know 2517 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2518 uint32_t nlist_idx = 0; 2519 Symbol *symbol_ptr = NULL; 2520 2521 uint32_t sym_idx = 0; 2522 Symbol *sym = NULL; 2523 size_t num_syms = 0; 2524 std::string memory_symbol_name; 2525 uint32_t unmapped_local_symbols_found = 0; 2526 2527 std::vector<TrieEntryWithOffset> trie_entries; 2528 std::set<lldb::addr_t> resolver_addresses; 2529 2530 if (dyld_trie_data.GetByteSize() > 0) 2531 { 2532 std::vector<llvm::StringRef> nameSlices; 2533 ParseTrieEntries (dyld_trie_data, 2534 0, 2535 nameSlices, 2536 resolver_addresses, 2537 trie_entries); 2538 2539 ConstString text_segment_name ("__TEXT"); 2540 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 2541 if (text_segment_sp) 2542 { 2543 const lldb::addr_t text_segment_file_addr = text_segment_sp->GetFileAddress(); 2544 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) 2545 { 2546 for (auto &e : trie_entries) 2547 e.entry.address += text_segment_file_addr; 2548 } 2549 } 2550 } 2551 2552 typedef std::set<ConstString> IndirectSymbols; 2553 IndirectSymbols indirect_symbol_names; 2554 2555 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 2556 2557 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been optimized by moving LOCAL 2558 // symbols out of the memory mapped portion of the DSC. The symbol information has all been retained, 2559 // but it isn't available in the normal nlist data. However, there *are* duplicate entries of *some* 2560 // LOCAL symbols in the normal nlist data. To handle this situation correctly, we must first attempt 2561 // to parse any DSC unmapped symbol information. If we find any, we set a flag that tells the normal 2562 // nlist parser to ignore all LOCAL symbols. 2563 2564 if (m_header.flags & 0x80000000u) 2565 { 2566 // Before we can start mapping the DSC, we need to make certain the target process is actually 2567 // using the cache we can find. 2568 2569 // Next we need to determine the correct path for the dyld shared cache. 2570 2571 ArchSpec header_arch; 2572 GetArchitecture(header_arch); 2573 char dsc_path[PATH_MAX]; 2574 2575 snprintf(dsc_path, sizeof(dsc_path), "%s%s%s", 2576 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR */ 2577 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2578 header_arch.GetArchitectureName()); 2579 2580 FileSpec dsc_filespec(dsc_path, false); 2581 2582 // We need definitions of two structures in the on-disk DSC, copy them here manually 2583 struct lldb_copy_dyld_cache_header_v0 2584 { 2585 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2586 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2587 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2588 uint32_t imagesOffset; 2589 uint32_t imagesCount; 2590 uint64_t dyldBaseAddress; 2591 uint64_t codeSignatureOffset; 2592 uint64_t codeSignatureSize; 2593 uint64_t slideInfoOffset; 2594 uint64_t slideInfoSize; 2595 uint64_t localSymbolsOffset; // file offset of where local symbols are stored 2596 uint64_t localSymbolsSize; // size of local symbols information 2597 }; 2598 struct lldb_copy_dyld_cache_header_v1 2599 { 2600 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2601 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2602 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2603 uint32_t imagesOffset; 2604 uint32_t imagesCount; 2605 uint64_t dyldBaseAddress; 2606 uint64_t codeSignatureOffset; 2607 uint64_t codeSignatureSize; 2608 uint64_t slideInfoOffset; 2609 uint64_t slideInfoSize; 2610 uint64_t localSymbolsOffset; 2611 uint64_t localSymbolsSize; 2612 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 and later 2613 }; 2614 2615 struct lldb_copy_dyld_cache_mapping_info 2616 { 2617 uint64_t address; 2618 uint64_t size; 2619 uint64_t fileOffset; 2620 uint32_t maxProt; 2621 uint32_t initProt; 2622 }; 2623 2624 struct lldb_copy_dyld_cache_local_symbols_info 2625 { 2626 uint32_t nlistOffset; 2627 uint32_t nlistCount; 2628 uint32_t stringsOffset; 2629 uint32_t stringsSize; 2630 uint32_t entriesOffset; 2631 uint32_t entriesCount; 2632 }; 2633 struct lldb_copy_dyld_cache_local_symbols_entry 2634 { 2635 uint32_t dylibOffset; 2636 uint32_t nlistStartIndex; 2637 uint32_t nlistCount; 2638 }; 2639 2640 /* The dyld_cache_header has a pointer to the dyld_cache_local_symbols_info structure (localSymbolsOffset). 2641 The dyld_cache_local_symbols_info structure gives us three things: 2642 1. The start and count of the nlist records in the dyld_shared_cache file 2643 2. The start and size of the strings for these nlist records 2644 3. The start and count of dyld_cache_local_symbols_entry entries 2645 2646 There is one dyld_cache_local_symbols_entry per dylib/framework in the dyld shared cache. 2647 The "dylibOffset" field is the Mach-O header of this dylib/framework in the dyld shared cache. 2648 The dyld_cache_local_symbols_entry also lists the start of this dylib/framework's nlist records 2649 and the count of how many nlist records there are for this dylib/framework. 2650 */ 2651 2652 // Process the dsc header to find the unmapped symbols 2653 // 2654 // Save some VM space, do not map the entire cache in one shot. 2655 2656 DataBufferSP dsc_data_sp; 2657 dsc_data_sp = dsc_filespec.MemoryMapFileContentsIfLocal(0, sizeof(struct lldb_copy_dyld_cache_header_v1)); 2658 2659 if (dsc_data_sp) 2660 { 2661 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2662 2663 char version_str[17]; 2664 int version = -1; 2665 lldb::offset_t offset = 0; 2666 memcpy (version_str, dsc_header_data.GetData (&offset, 16), 16); 2667 version_str[16] = '\0'; 2668 if (strncmp (version_str, "dyld_v", 6) == 0 && isdigit (version_str[6])) 2669 { 2670 int v; 2671 if (::sscanf (version_str + 6, "%d", &v) == 1) 2672 { 2673 version = v; 2674 } 2675 } 2676 2677 UUID dsc_uuid; 2678 if (version >= 1) 2679 { 2680 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, uuid); 2681 uint8_t uuid_bytes[sizeof (uuid_t)]; 2682 memcpy (uuid_bytes, dsc_header_data.GetData (&offset, sizeof (uuid_t)), sizeof (uuid_t)); 2683 dsc_uuid.SetBytes (uuid_bytes); 2684 } 2685 2686 bool uuid_match = true; 2687 if (dsc_uuid.IsValid() && process) 2688 { 2689 UUID shared_cache_uuid(GetProcessSharedCacheUUID(process)); 2690 2691 if (shared_cache_uuid.IsValid() && dsc_uuid != shared_cache_uuid) 2692 { 2693 // The on-disk dyld_shared_cache file is not the same as the one in this 2694 // process' memory, don't use it. 2695 uuid_match = false; 2696 ModuleSP module_sp (GetModule()); 2697 if (module_sp) 2698 module_sp->ReportWarning ("process shared cache does not match on-disk dyld_shared_cache file, some symbol names will be missing."); 2699 } 2700 } 2701 2702 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2703 2704 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2705 2706 // If the mappingOffset points to a location inside the header, we've 2707 // opened an old dyld shared cache, and should not proceed further. 2708 if (uuid_match && mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v0)) 2709 { 2710 2711 DataBufferSP dsc_mapping_info_data_sp = dsc_filespec.MemoryMapFileContentsIfLocal(mappingOffset, sizeof (struct lldb_copy_dyld_cache_mapping_info)); 2712 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, byte_order, addr_byte_size); 2713 offset = 0; 2714 2715 // The File addresses (from the in-memory Mach-O load commands) for the shared libraries 2716 // in the shared library cache need to be adjusted by an offset to match up with the 2717 // dylibOffset identifying field in the dyld_cache_local_symbol_entry's. This offset is 2718 // recorded in mapping_offset_value. 2719 const uint64_t mapping_offset_value = dsc_mapping_info_data.GetU64(&offset); 2720 2721 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2722 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2723 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2724 2725 if (localSymbolsOffset && localSymbolsSize) 2726 { 2727 // Map the local symbols 2728 if (DataBufferSP dsc_local_symbols_data_sp = dsc_filespec.MemoryMapFileContentsIfLocal(localSymbolsOffset, localSymbolsSize)) 2729 { 2730 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, byte_order, addr_byte_size); 2731 2732 offset = 0; 2733 2734 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2735 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2736 UndefinedNameToDescMap undefined_name_to_desc; 2737 SymbolIndexToName reexport_shlib_needs_fixup; 2738 2739 2740 // Read the local_symbols_infos struct in one shot 2741 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2742 dsc_local_symbols_data.GetU32(&offset, &local_symbols_info.nlistOffset, 6); 2743 2744 SectionSP text_section_sp(section_list->FindSectionByName(GetSegmentNameTEXT())); 2745 2746 uint32_t header_file_offset = (text_section_sp->GetFileAddress() - mapping_offset_value); 2747 2748 offset = local_symbols_info.entriesOffset; 2749 for (uint32_t entry_index = 0; entry_index < local_symbols_info.entriesCount; entry_index++) 2750 { 2751 struct lldb_copy_dyld_cache_local_symbols_entry local_symbols_entry; 2752 local_symbols_entry.dylibOffset = dsc_local_symbols_data.GetU32(&offset); 2753 local_symbols_entry.nlistStartIndex = dsc_local_symbols_data.GetU32(&offset); 2754 local_symbols_entry.nlistCount = dsc_local_symbols_data.GetU32(&offset); 2755 2756 if (header_file_offset == local_symbols_entry.dylibOffset) 2757 { 2758 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2759 2760 // The normal nlist code cannot correctly size the Symbols array, we need to allocate it here. 2761 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms + unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2762 num_syms = symtab->GetNumSymbols(); 2763 2764 nlist_data_offset = local_symbols_info.nlistOffset + (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2765 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2766 2767 for (uint32_t nlist_index = 0; nlist_index < local_symbols_entry.nlistCount; nlist_index++) 2768 { 2769 ///////////////////////////// 2770 { 2771 struct nlist_64 nlist; 2772 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2773 break; 2774 2775 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked(&nlist_data_offset); 2776 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2777 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2778 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked (&nlist_data_offset); 2779 nlist.n_value = dsc_local_symbols_data.GetAddress_unchecked (&nlist_data_offset); 2780 2781 SymbolType type = eSymbolTypeInvalid; 2782 const char *symbol_name = dsc_local_symbols_data.PeekCStr(string_table_offset + nlist.n_strx); 2783 2784 if (symbol_name == NULL) 2785 { 2786 // No symbol should be NULL, even the symbols with no 2787 // string values should have an offset zero which points 2788 // to an empty C-string 2789 Host::SystemLog (Host::eSystemLogError, 2790 "error: DSC unmapped local symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 2791 entry_index, 2792 nlist.n_strx, 2793 module_sp->GetFileSpec().GetPath().c_str()); 2794 continue; 2795 } 2796 if (symbol_name[0] == '\0') 2797 symbol_name = NULL; 2798 2799 const char *symbol_name_non_abi_mangled = NULL; 2800 2801 SectionSP symbol_section; 2802 uint32_t symbol_byte_size = 0; 2803 bool add_nlist = true; 2804 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2805 bool demangled_is_synthesized = false; 2806 bool is_gsym = false; 2807 bool set_value = true; 2808 2809 assert (sym_idx < num_syms); 2810 2811 sym[sym_idx].SetDebug (is_debug); 2812 2813 if (is_debug) 2814 { 2815 switch (nlist.n_type) 2816 { 2817 case N_GSYM: 2818 // global symbol: name,,NO_SECT,type,0 2819 // Sometimes the N_GSYM value contains the address. 2820 2821 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 2822 // have the same address, but we want to ensure that we always find only the real symbol, 2823 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 2824 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 2825 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 2826 // same address. 2827 2828 is_gsym = true; 2829 sym[sym_idx].SetExternal(true); 2830 2831 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') 2832 { 2833 llvm::StringRef symbol_name_ref(symbol_name); 2834 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 2835 { 2836 symbol_name_non_abi_mangled = symbol_name + 1; 2837 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2838 type = eSymbolTypeObjCClass; 2839 demangled_is_synthesized = true; 2840 2841 } 2842 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 2843 { 2844 symbol_name_non_abi_mangled = symbol_name + 1; 2845 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2846 type = eSymbolTypeObjCMetaClass; 2847 demangled_is_synthesized = true; 2848 } 2849 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 2850 { 2851 symbol_name_non_abi_mangled = symbol_name + 1; 2852 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2853 type = eSymbolTypeObjCIVar; 2854 demangled_is_synthesized = true; 2855 } 2856 } 2857 else 2858 { 2859 if (nlist.n_value != 0) 2860 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2861 type = eSymbolTypeData; 2862 } 2863 break; 2864 2865 case N_FNAME: 2866 // procedure name (f77 kludge): name,,NO_SECT,0,0 2867 type = eSymbolTypeCompiler; 2868 break; 2869 2870 case N_FUN: 2871 // procedure: name,,n_sect,linenumber,address 2872 if (symbol_name) 2873 { 2874 type = eSymbolTypeCode; 2875 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2876 2877 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 2878 // We use the current number of symbols in the symbol table in lieu of 2879 // using nlist_idx in case we ever start trimming entries out 2880 N_FUN_indexes.push_back(sym_idx); 2881 } 2882 else 2883 { 2884 type = eSymbolTypeCompiler; 2885 2886 if ( !N_FUN_indexes.empty() ) 2887 { 2888 // Copy the size of the function into the original STAB entry so we don't have 2889 // to hunt for it later 2890 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 2891 N_FUN_indexes.pop_back(); 2892 // We don't really need the end function STAB as it contains the size which 2893 // we already placed with the original symbol, so don't add it if we want a 2894 // minimal symbol table 2895 add_nlist = false; 2896 } 2897 } 2898 break; 2899 2900 case N_STSYM: 2901 // static symbol: name,,n_sect,type,address 2902 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 2903 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2904 type = eSymbolTypeData; 2905 break; 2906 2907 case N_LCSYM: 2908 // .lcomm symbol: name,,n_sect,type,address 2909 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2910 type = eSymbolTypeCommonBlock; 2911 break; 2912 2913 case N_BNSYM: 2914 // We use the current number of symbols in the symbol table in lieu of 2915 // using nlist_idx in case we ever start trimming entries out 2916 // Skip these if we want minimal symbol tables 2917 add_nlist = false; 2918 break; 2919 2920 case N_ENSYM: 2921 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 2922 // so that we can always skip the entire symbol if we need to navigate 2923 // more quickly at the source level when parsing STABS 2924 // Skip these if we want minimal symbol tables 2925 add_nlist = false; 2926 break; 2927 2928 2929 case N_OPT: 2930 // emitted with gcc2_compiled and in gcc source 2931 type = eSymbolTypeCompiler; 2932 break; 2933 2934 case N_RSYM: 2935 // register sym: name,,NO_SECT,type,register 2936 type = eSymbolTypeVariable; 2937 break; 2938 2939 case N_SLINE: 2940 // src line: 0,,n_sect,linenumber,address 2941 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2942 type = eSymbolTypeLineEntry; 2943 break; 2944 2945 case N_SSYM: 2946 // structure elt: name,,NO_SECT,type,struct_offset 2947 type = eSymbolTypeVariableType; 2948 break; 2949 2950 case N_SO: 2951 // source file name 2952 type = eSymbolTypeSourceFile; 2953 if (symbol_name == NULL) 2954 { 2955 add_nlist = false; 2956 if (N_SO_index != UINT32_MAX) 2957 { 2958 // Set the size of the N_SO to the terminating index of this N_SO 2959 // so that we can always skip the entire N_SO if we need to navigate 2960 // more quickly at the source level when parsing STABS 2961 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2962 symbol_ptr->SetByteSize(sym_idx); 2963 symbol_ptr->SetSizeIsSibling(true); 2964 } 2965 N_NSYM_indexes.clear(); 2966 N_INCL_indexes.clear(); 2967 N_BRAC_indexes.clear(); 2968 N_COMM_indexes.clear(); 2969 N_FUN_indexes.clear(); 2970 N_SO_index = UINT32_MAX; 2971 } 2972 else 2973 { 2974 // We use the current number of symbols in the symbol table in lieu of 2975 // using nlist_idx in case we ever start trimming entries out 2976 const bool N_SO_has_full_path = symbol_name[0] == '/'; 2977 if (N_SO_has_full_path) 2978 { 2979 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2980 { 2981 // We have two consecutive N_SO entries where the first contains a directory 2982 // and the second contains a full path. 2983 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 2984 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2985 add_nlist = false; 2986 } 2987 else 2988 { 2989 // This is the first entry in a N_SO that contains a directory or 2990 // a full path to the source file 2991 N_SO_index = sym_idx; 2992 } 2993 } 2994 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2995 { 2996 // This is usually the second N_SO entry that contains just the filename, 2997 // so here we combine it with the first one if we are minimizing the symbol table 2998 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 2999 if (so_path && so_path[0]) 3000 { 3001 std::string full_so_path (so_path); 3002 const size_t double_slash_pos = full_so_path.find("//"); 3003 if (double_slash_pos != std::string::npos) 3004 { 3005 // The linker has been generating bad N_SO entries with doubled up paths 3006 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 3007 // and the second is the directory for the source file so you end up with 3008 // a path that looks like "/tmp/src//tmp/src/" 3009 FileSpec so_dir(so_path, false); 3010 if (!so_dir.Exists()) 3011 { 3012 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 3013 if (so_dir.Exists()) 3014 { 3015 // Trim off the incorrect path 3016 full_so_path.erase(0, double_slash_pos + 1); 3017 } 3018 } 3019 } 3020 if (*full_so_path.rbegin() != '/') 3021 full_so_path += '/'; 3022 full_so_path += symbol_name; 3023 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 3024 add_nlist = false; 3025 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3026 } 3027 } 3028 else 3029 { 3030 // This could be a relative path to a N_SO 3031 N_SO_index = sym_idx; 3032 } 3033 } 3034 break; 3035 3036 case N_OSO: 3037 // object file name: name,,0,0,st_mtime 3038 type = eSymbolTypeObjectFile; 3039 break; 3040 3041 case N_LSYM: 3042 // local sym: name,,NO_SECT,type,offset 3043 type = eSymbolTypeLocal; 3044 break; 3045 3046 //---------------------------------------------------------------------- 3047 // INCL scopes 3048 //---------------------------------------------------------------------- 3049 case N_BINCL: 3050 // include file beginning: name,,NO_SECT,0,sum 3051 // We use the current number of symbols in the symbol table in lieu of 3052 // using nlist_idx in case we ever start trimming entries out 3053 N_INCL_indexes.push_back(sym_idx); 3054 type = eSymbolTypeScopeBegin; 3055 break; 3056 3057 case N_EINCL: 3058 // include file end: name,,NO_SECT,0,0 3059 // Set the size of the N_BINCL to the terminating index of this N_EINCL 3060 // so that we can always skip the entire symbol if we need to navigate 3061 // more quickly at the source level when parsing STABS 3062 if ( !N_INCL_indexes.empty() ) 3063 { 3064 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3065 symbol_ptr->SetByteSize(sym_idx + 1); 3066 symbol_ptr->SetSizeIsSibling(true); 3067 N_INCL_indexes.pop_back(); 3068 } 3069 type = eSymbolTypeScopeEnd; 3070 break; 3071 3072 case N_SOL: 3073 // #included file name: name,,n_sect,0,address 3074 type = eSymbolTypeHeaderFile; 3075 3076 // We currently don't use the header files on darwin 3077 add_nlist = false; 3078 break; 3079 3080 case N_PARAMS: 3081 // compiler parameters: name,,NO_SECT,0,0 3082 type = eSymbolTypeCompiler; 3083 break; 3084 3085 case N_VERSION: 3086 // compiler version: name,,NO_SECT,0,0 3087 type = eSymbolTypeCompiler; 3088 break; 3089 3090 case N_OLEVEL: 3091 // compiler -O level: name,,NO_SECT,0,0 3092 type = eSymbolTypeCompiler; 3093 break; 3094 3095 case N_PSYM: 3096 // parameter: name,,NO_SECT,type,offset 3097 type = eSymbolTypeVariable; 3098 break; 3099 3100 case N_ENTRY: 3101 // alternate entry: name,,n_sect,linenumber,address 3102 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3103 type = eSymbolTypeLineEntry; 3104 break; 3105 3106 //---------------------------------------------------------------------- 3107 // Left and Right Braces 3108 //---------------------------------------------------------------------- 3109 case N_LBRAC: 3110 // left bracket: 0,,NO_SECT,nesting level,address 3111 // We use the current number of symbols in the symbol table in lieu of 3112 // using nlist_idx in case we ever start trimming entries out 3113 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3114 N_BRAC_indexes.push_back(sym_idx); 3115 type = eSymbolTypeScopeBegin; 3116 break; 3117 3118 case N_RBRAC: 3119 // right bracket: 0,,NO_SECT,nesting level,address 3120 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 3121 // so that we can always skip the entire symbol if we need to navigate 3122 // more quickly at the source level when parsing STABS 3123 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3124 if ( !N_BRAC_indexes.empty() ) 3125 { 3126 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3127 symbol_ptr->SetByteSize(sym_idx + 1); 3128 symbol_ptr->SetSizeIsSibling(true); 3129 N_BRAC_indexes.pop_back(); 3130 } 3131 type = eSymbolTypeScopeEnd; 3132 break; 3133 3134 case N_EXCL: 3135 // deleted include file: name,,NO_SECT,0,sum 3136 type = eSymbolTypeHeaderFile; 3137 break; 3138 3139 //---------------------------------------------------------------------- 3140 // COMM scopes 3141 //---------------------------------------------------------------------- 3142 case N_BCOMM: 3143 // begin common: name,,NO_SECT,0,0 3144 // We use the current number of symbols in the symbol table in lieu of 3145 // using nlist_idx in case we ever start trimming entries out 3146 type = eSymbolTypeScopeBegin; 3147 N_COMM_indexes.push_back(sym_idx); 3148 break; 3149 3150 case N_ECOML: 3151 // end common (local name): 0,,n_sect,0,address 3152 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3153 // Fall through 3154 3155 case N_ECOMM: 3156 // end common: name,,n_sect,0,0 3157 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 3158 // so that we can always skip the entire symbol if we need to navigate 3159 // more quickly at the source level when parsing STABS 3160 if ( !N_COMM_indexes.empty() ) 3161 { 3162 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 3163 symbol_ptr->SetByteSize(sym_idx + 1); 3164 symbol_ptr->SetSizeIsSibling(true); 3165 N_COMM_indexes.pop_back(); 3166 } 3167 type = eSymbolTypeScopeEnd; 3168 break; 3169 3170 case N_LENG: 3171 // second stab entry with length information 3172 type = eSymbolTypeAdditional; 3173 break; 3174 3175 default: break; 3176 } 3177 } 3178 else 3179 { 3180 //uint8_t n_pext = N_PEXT & nlist.n_type; 3181 uint8_t n_type = N_TYPE & nlist.n_type; 3182 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3183 3184 switch (n_type) 3185 { 3186 case N_INDR: 3187 { 3188 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 3189 if (reexport_name_cstr && reexport_name_cstr[0]) 3190 { 3191 type = eSymbolTypeReExported; 3192 ConstString reexport_name(reexport_name_cstr + ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3193 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3194 set_value = false; 3195 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3196 indirect_symbol_names.insert(ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3197 } 3198 else 3199 type = eSymbolTypeUndefined; 3200 } 3201 break; 3202 3203 case N_UNDF: 3204 if (symbol_name && symbol_name[0]) 3205 { 3206 ConstString undefined_name(symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3207 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3208 } 3209 // Fall through 3210 case N_PBUD: 3211 type = eSymbolTypeUndefined; 3212 break; 3213 3214 case N_ABS: 3215 type = eSymbolTypeAbsolute; 3216 break; 3217 3218 case N_SECT: 3219 { 3220 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3221 3222 if (symbol_section == NULL) 3223 { 3224 // TODO: warn about this? 3225 add_nlist = false; 3226 break; 3227 } 3228 3229 if (TEXT_eh_frame_sectID == nlist.n_sect) 3230 { 3231 type = eSymbolTypeException; 3232 } 3233 else 3234 { 3235 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 3236 3237 switch (section_type) 3238 { 3239 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 3240 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 3241 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 3242 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 3243 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 3244 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 3245 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 3246 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 3247 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 3248 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 3249 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 3250 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 3251 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 3252 default: 3253 switch (symbol_section->GetType()) 3254 { 3255 case lldb::eSectionTypeCode: 3256 type = eSymbolTypeCode; 3257 break; 3258 case eSectionTypeData: 3259 case eSectionTypeDataCString: // Inlined C string data 3260 case eSectionTypeDataCStringPointers: // Pointers to C string data 3261 case eSectionTypeDataSymbolAddress: // Address of a symbol in the symbol table 3262 case eSectionTypeData4: 3263 case eSectionTypeData8: 3264 case eSectionTypeData16: 3265 type = eSymbolTypeData; 3266 break; 3267 default: 3268 break; 3269 } 3270 break; 3271 } 3272 3273 if (type == eSymbolTypeInvalid) 3274 { 3275 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 3276 if (symbol_section->IsDescendant (text_section_sp.get())) 3277 { 3278 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 3279 S_ATTR_SELF_MODIFYING_CODE | 3280 S_ATTR_SOME_INSTRUCTIONS)) 3281 type = eSymbolTypeData; 3282 else 3283 type = eSymbolTypeCode; 3284 } 3285 else if (symbol_section->IsDescendant(data_section_sp.get())) 3286 { 3287 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 3288 { 3289 type = eSymbolTypeRuntime; 3290 3291 if (symbol_name && 3292 symbol_name[0] == '_' && 3293 symbol_name[1] == 'O' && 3294 symbol_name[2] == 'B') 3295 { 3296 llvm::StringRef symbol_name_ref(symbol_name); 3297 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 3298 { 3299 symbol_name_non_abi_mangled = symbol_name + 1; 3300 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3301 type = eSymbolTypeObjCClass; 3302 demangled_is_synthesized = true; 3303 } 3304 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 3305 { 3306 symbol_name_non_abi_mangled = symbol_name + 1; 3307 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3308 type = eSymbolTypeObjCMetaClass; 3309 demangled_is_synthesized = true; 3310 } 3311 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 3312 { 3313 symbol_name_non_abi_mangled = symbol_name + 1; 3314 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3315 type = eSymbolTypeObjCIVar; 3316 demangled_is_synthesized = true; 3317 } 3318 } 3319 } 3320 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 3321 { 3322 type = eSymbolTypeException; 3323 } 3324 else 3325 { 3326 type = eSymbolTypeData; 3327 } 3328 } 3329 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 3330 { 3331 type = eSymbolTypeTrampoline; 3332 } 3333 else if (symbol_section->IsDescendant(objc_section_sp.get())) 3334 { 3335 type = eSymbolTypeRuntime; 3336 if (symbol_name && symbol_name[0] == '.') 3337 { 3338 llvm::StringRef symbol_name_ref(symbol_name); 3339 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 3340 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 3341 { 3342 symbol_name_non_abi_mangled = symbol_name; 3343 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 3344 type = eSymbolTypeObjCClass; 3345 demangled_is_synthesized = true; 3346 } 3347 } 3348 } 3349 } 3350 } 3351 } 3352 break; 3353 } 3354 } 3355 3356 if (add_nlist) 3357 { 3358 uint64_t symbol_value = nlist.n_value; 3359 if (symbol_name_non_abi_mangled) 3360 { 3361 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 3362 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 3363 } 3364 else 3365 { 3366 bool symbol_name_is_mangled = false; 3367 3368 if (symbol_name && symbol_name[0] == '_') 3369 { 3370 symbol_name_is_mangled = symbol_name[1] == '_'; 3371 symbol_name++; // Skip the leading underscore 3372 } 3373 3374 if (symbol_name) 3375 { 3376 ConstString const_symbol_name(symbol_name); 3377 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 3378 if (is_gsym && is_debug) 3379 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 3380 } 3381 } 3382 if (symbol_section) 3383 { 3384 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3385 if (symbol_byte_size == 0 && function_starts_count > 0) 3386 { 3387 addr_t symbol_lookup_file_addr = nlist.n_value; 3388 // Do an exact address match for non-ARM addresses, else get the closest since 3389 // the symbol might be a thumb symbol which has an address with bit zero set 3390 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 3391 if (is_arm && func_start_entry) 3392 { 3393 // Verify that the function start address is the symbol address (ARM) 3394 // or the symbol address + 1 (thumb) 3395 if (func_start_entry->addr != symbol_lookup_file_addr && 3396 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 3397 { 3398 // Not the right entry, NULL it out... 3399 func_start_entry = NULL; 3400 } 3401 } 3402 if (func_start_entry) 3403 { 3404 func_start_entry->data = true; 3405 3406 addr_t symbol_file_addr = func_start_entry->addr; 3407 uint32_t symbol_flags = 0; 3408 if (is_arm) 3409 { 3410 if (symbol_file_addr & 1) 3411 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3412 symbol_file_addr &= 0xfffffffffffffffeull; 3413 } 3414 3415 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3416 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3417 if (next_func_start_entry) 3418 { 3419 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3420 // Be sure the clear the Thumb address bit when we calculate the size 3421 // from the current and next address 3422 if (is_arm) 3423 next_symbol_file_addr &= 0xfffffffffffffffeull; 3424 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3425 } 3426 else 3427 { 3428 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3429 } 3430 } 3431 } 3432 symbol_value -= section_file_addr; 3433 } 3434 3435 if (is_debug == false) 3436 { 3437 if (type == eSymbolTypeCode) 3438 { 3439 // See if we can find a N_FUN entry for any code symbols. 3440 // If we do find a match, and the name matches, then we 3441 // can merge the two into just the function symbol to avoid 3442 // duplicate entries in the symbol table 3443 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3444 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3445 if (range.first != range.second) 3446 { 3447 bool found_it = false; 3448 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3449 { 3450 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3451 { 3452 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3453 // We just need the flags from the linker symbol, so put these flags 3454 // into the N_FUN flags to avoid duplicate symbols in the symbol table 3455 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3456 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3457 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3458 sym[pos->second].SetType (eSymbolTypeResolver); 3459 sym[sym_idx].Clear(); 3460 found_it = true; 3461 break; 3462 } 3463 } 3464 if (found_it) 3465 continue; 3466 } 3467 else 3468 { 3469 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3470 type = eSymbolTypeResolver; 3471 } 3472 } 3473 else if (type == eSymbolTypeData || 3474 type == eSymbolTypeObjCClass || 3475 type == eSymbolTypeObjCMetaClass || 3476 type == eSymbolTypeObjCIVar ) 3477 { 3478 // See if we can find a N_STSYM entry for any data symbols. 3479 // If we do find a match, and the name matches, then we 3480 // can merge the two into just the Static symbol to avoid 3481 // duplicate entries in the symbol table 3482 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3483 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 3484 if (range.first != range.second) 3485 { 3486 bool found_it = false; 3487 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3488 { 3489 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3490 { 3491 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3492 // We just need the flags from the linker symbol, so put these flags 3493 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3494 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3495 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3496 sym[sym_idx].Clear(); 3497 found_it = true; 3498 break; 3499 } 3500 } 3501 if (found_it) 3502 continue; 3503 } 3504 else 3505 { 3506 // Combine N_GSYM stab entries with the non stab symbol 3507 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 3508 if (pos != N_GSYM_name_to_sym_idx.end()) 3509 { 3510 const uint32_t GSYM_sym_idx = pos->second; 3511 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 3512 // Copy the address, because often the N_GSYM address has an invalid address of zero 3513 // when the global is a common symbol 3514 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 3515 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 3516 // We just need the flags from the linker symbol, so put these flags 3517 // into the N_GSYM flags to avoid duplicate symbols in the symbol table 3518 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3519 sym[sym_idx].Clear(); 3520 continue; 3521 } 3522 } 3523 } 3524 } 3525 3526 sym[sym_idx].SetID (nlist_idx); 3527 sym[sym_idx].SetType (type); 3528 if (set_value) 3529 { 3530 sym[sym_idx].GetAddress().SetSection (symbol_section); 3531 sym[sym_idx].GetAddress().SetOffset (symbol_value); 3532 } 3533 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3534 3535 if (symbol_byte_size > 0) 3536 sym[sym_idx].SetByteSize(symbol_byte_size); 3537 3538 if (demangled_is_synthesized) 3539 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3540 ++sym_idx; 3541 } 3542 else 3543 { 3544 sym[sym_idx].Clear(); 3545 } 3546 3547 } 3548 ///////////////////////////// 3549 } 3550 break; // No more entries to consider 3551 } 3552 } 3553 3554 for (const auto &pos :reexport_shlib_needs_fixup) 3555 { 3556 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3557 if (undef_pos != undefined_name_to_desc.end()) 3558 { 3559 const uint8_t dylib_ordinal = llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3560 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3561 sym[pos.first].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(dylib_ordinal-1)); 3562 } 3563 } 3564 } 3565 } 3566 } 3567 } 3568 } 3569 3570 // Must reset this in case it was mutated above! 3571 nlist_data_offset = 0; 3572 #endif 3573 3574 if (nlist_data.GetByteSize() > 0) 3575 { 3576 3577 // If the sym array was not created while parsing the DSC unmapped 3578 // symbols, create it now. 3579 if (sym == NULL) 3580 { 3581 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3582 num_syms = symtab->GetNumSymbols(); 3583 } 3584 3585 if (unmapped_local_symbols_found) 3586 { 3587 assert(m_dysymtab.ilocalsym == 0); 3588 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3589 nlist_idx = m_dysymtab.nlocalsym; 3590 } 3591 else 3592 { 3593 nlist_idx = 0; 3594 } 3595 3596 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3597 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3598 UndefinedNameToDescMap undefined_name_to_desc; 3599 SymbolIndexToName reexport_shlib_needs_fixup; 3600 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 3601 { 3602 struct nlist_64 nlist; 3603 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 3604 break; 3605 3606 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3607 nlist.n_type = nlist_data.GetU8_unchecked (&nlist_data_offset); 3608 nlist.n_sect = nlist_data.GetU8_unchecked (&nlist_data_offset); 3609 nlist.n_desc = nlist_data.GetU16_unchecked (&nlist_data_offset); 3610 nlist.n_value = nlist_data.GetAddress_unchecked (&nlist_data_offset); 3611 3612 SymbolType type = eSymbolTypeInvalid; 3613 const char *symbol_name = NULL; 3614 3615 if (have_strtab_data) 3616 { 3617 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3618 3619 if (symbol_name == NULL) 3620 { 3621 // No symbol should be NULL, even the symbols with no 3622 // string values should have an offset zero which points 3623 // to an empty C-string 3624 Host::SystemLog (Host::eSystemLogError, 3625 "error: symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 3626 nlist_idx, 3627 nlist.n_strx, 3628 module_sp->GetFileSpec().GetPath().c_str()); 3629 continue; 3630 } 3631 if (symbol_name[0] == '\0') 3632 symbol_name = NULL; 3633 } 3634 else 3635 { 3636 const addr_t str_addr = strtab_addr + nlist.n_strx; 3637 Error str_error; 3638 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, str_error)) 3639 symbol_name = memory_symbol_name.c_str(); 3640 } 3641 const char *symbol_name_non_abi_mangled = NULL; 3642 3643 SectionSP symbol_section; 3644 lldb::addr_t symbol_byte_size = 0; 3645 bool add_nlist = true; 3646 bool is_gsym = false; 3647 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3648 bool demangled_is_synthesized = false; 3649 bool set_value = true; 3650 assert (sym_idx < num_syms); 3651 3652 sym[sym_idx].SetDebug (is_debug); 3653 3654 if (is_debug) 3655 { 3656 switch (nlist.n_type) 3657 { 3658 case N_GSYM: 3659 // global symbol: name,,NO_SECT,type,0 3660 // Sometimes the N_GSYM value contains the address. 3661 3662 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 3663 // have the same address, but we want to ensure that we always find only the real symbol, 3664 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 3665 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 3666 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 3667 // same address. 3668 is_gsym = true; 3669 sym[sym_idx].SetExternal(true); 3670 3671 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') 3672 { 3673 llvm::StringRef symbol_name_ref(symbol_name); 3674 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 3675 { 3676 symbol_name_non_abi_mangled = symbol_name + 1; 3677 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3678 type = eSymbolTypeObjCClass; 3679 demangled_is_synthesized = true; 3680 3681 } 3682 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 3683 { 3684 symbol_name_non_abi_mangled = symbol_name + 1; 3685 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3686 type = eSymbolTypeObjCMetaClass; 3687 demangled_is_synthesized = true; 3688 } 3689 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 3690 { 3691 symbol_name_non_abi_mangled = symbol_name + 1; 3692 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3693 type = eSymbolTypeObjCIVar; 3694 demangled_is_synthesized = true; 3695 } 3696 } 3697 else 3698 { 3699 if (nlist.n_value != 0) 3700 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3701 type = eSymbolTypeData; 3702 } 3703 break; 3704 3705 case N_FNAME: 3706 // procedure name (f77 kludge): name,,NO_SECT,0,0 3707 type = eSymbolTypeCompiler; 3708 break; 3709 3710 case N_FUN: 3711 // procedure: name,,n_sect,linenumber,address 3712 if (symbol_name) 3713 { 3714 type = eSymbolTypeCode; 3715 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3716 3717 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3718 // We use the current number of symbols in the symbol table in lieu of 3719 // using nlist_idx in case we ever start trimming entries out 3720 N_FUN_indexes.push_back(sym_idx); 3721 } 3722 else 3723 { 3724 type = eSymbolTypeCompiler; 3725 3726 if ( !N_FUN_indexes.empty() ) 3727 { 3728 // Copy the size of the function into the original STAB entry so we don't have 3729 // to hunt for it later 3730 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 3731 N_FUN_indexes.pop_back(); 3732 // We don't really need the end function STAB as it contains the size which 3733 // we already placed with the original symbol, so don't add it if we want a 3734 // minimal symbol table 3735 add_nlist = false; 3736 } 3737 } 3738 break; 3739 3740 case N_STSYM: 3741 // static symbol: name,,n_sect,type,address 3742 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3743 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3744 type = eSymbolTypeData; 3745 break; 3746 3747 case N_LCSYM: 3748 // .lcomm symbol: name,,n_sect,type,address 3749 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3750 type = eSymbolTypeCommonBlock; 3751 break; 3752 3753 case N_BNSYM: 3754 // We use the current number of symbols in the symbol table in lieu of 3755 // using nlist_idx in case we ever start trimming entries out 3756 // Skip these if we want minimal symbol tables 3757 add_nlist = false; 3758 break; 3759 3760 case N_ENSYM: 3761 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 3762 // so that we can always skip the entire symbol if we need to navigate 3763 // more quickly at the source level when parsing STABS 3764 // Skip these if we want minimal symbol tables 3765 add_nlist = false; 3766 break; 3767 3768 3769 case N_OPT: 3770 // emitted with gcc2_compiled and in gcc source 3771 type = eSymbolTypeCompiler; 3772 break; 3773 3774 case N_RSYM: 3775 // register sym: name,,NO_SECT,type,register 3776 type = eSymbolTypeVariable; 3777 break; 3778 3779 case N_SLINE: 3780 // src line: 0,,n_sect,linenumber,address 3781 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3782 type = eSymbolTypeLineEntry; 3783 break; 3784 3785 case N_SSYM: 3786 // structure elt: name,,NO_SECT,type,struct_offset 3787 type = eSymbolTypeVariableType; 3788 break; 3789 3790 case N_SO: 3791 // source file name 3792 type = eSymbolTypeSourceFile; 3793 if (symbol_name == NULL) 3794 { 3795 add_nlist = false; 3796 if (N_SO_index != UINT32_MAX) 3797 { 3798 // Set the size of the N_SO to the terminating index of this N_SO 3799 // so that we can always skip the entire N_SO if we need to navigate 3800 // more quickly at the source level when parsing STABS 3801 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3802 symbol_ptr->SetByteSize(sym_idx); 3803 symbol_ptr->SetSizeIsSibling(true); 3804 } 3805 N_NSYM_indexes.clear(); 3806 N_INCL_indexes.clear(); 3807 N_BRAC_indexes.clear(); 3808 N_COMM_indexes.clear(); 3809 N_FUN_indexes.clear(); 3810 N_SO_index = UINT32_MAX; 3811 } 3812 else 3813 { 3814 // We use the current number of symbols in the symbol table in lieu of 3815 // using nlist_idx in case we ever start trimming entries out 3816 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3817 if (N_SO_has_full_path) 3818 { 3819 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3820 { 3821 // We have two consecutive N_SO entries where the first contains a directory 3822 // and the second contains a full path. 3823 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 3824 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3825 add_nlist = false; 3826 } 3827 else 3828 { 3829 // This is the first entry in a N_SO that contains a directory or 3830 // a full path to the source file 3831 N_SO_index = sym_idx; 3832 } 3833 } 3834 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3835 { 3836 // This is usually the second N_SO entry that contains just the filename, 3837 // so here we combine it with the first one if we are minimizing the symbol table 3838 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3839 if (so_path && so_path[0]) 3840 { 3841 std::string full_so_path (so_path); 3842 const size_t double_slash_pos = full_so_path.find("//"); 3843 if (double_slash_pos != std::string::npos) 3844 { 3845 // The linker has been generating bad N_SO entries with doubled up paths 3846 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 3847 // and the second is the directory for the source file so you end up with 3848 // a path that looks like "/tmp/src//tmp/src/" 3849 FileSpec so_dir(so_path, false); 3850 if (!so_dir.Exists()) 3851 { 3852 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 3853 if (so_dir.Exists()) 3854 { 3855 // Trim off the incorrect path 3856 full_so_path.erase(0, double_slash_pos + 1); 3857 } 3858 } 3859 } 3860 if (*full_so_path.rbegin() != '/') 3861 full_so_path += '/'; 3862 full_so_path += symbol_name; 3863 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 3864 add_nlist = false; 3865 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3866 } 3867 } 3868 else 3869 { 3870 // This could be a relative path to a N_SO 3871 N_SO_index = sym_idx; 3872 } 3873 } 3874 3875 break; 3876 3877 case N_OSO: 3878 // object file name: name,,0,0,st_mtime 3879 type = eSymbolTypeObjectFile; 3880 break; 3881 3882 case N_LSYM: 3883 // local sym: name,,NO_SECT,type,offset 3884 type = eSymbolTypeLocal; 3885 break; 3886 3887 //---------------------------------------------------------------------- 3888 // INCL scopes 3889 //---------------------------------------------------------------------- 3890 case N_BINCL: 3891 // include file beginning: name,,NO_SECT,0,sum 3892 // We use the current number of symbols in the symbol table in lieu of 3893 // using nlist_idx in case we ever start trimming entries out 3894 N_INCL_indexes.push_back(sym_idx); 3895 type = eSymbolTypeScopeBegin; 3896 break; 3897 3898 case N_EINCL: 3899 // include file end: name,,NO_SECT,0,0 3900 // Set the size of the N_BINCL to the terminating index of this N_EINCL 3901 // so that we can always skip the entire symbol if we need to navigate 3902 // more quickly at the source level when parsing STABS 3903 if ( !N_INCL_indexes.empty() ) 3904 { 3905 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3906 symbol_ptr->SetByteSize(sym_idx + 1); 3907 symbol_ptr->SetSizeIsSibling(true); 3908 N_INCL_indexes.pop_back(); 3909 } 3910 type = eSymbolTypeScopeEnd; 3911 break; 3912 3913 case N_SOL: 3914 // #included file name: name,,n_sect,0,address 3915 type = eSymbolTypeHeaderFile; 3916 3917 // We currently don't use the header files on darwin 3918 add_nlist = false; 3919 break; 3920 3921 case N_PARAMS: 3922 // compiler parameters: name,,NO_SECT,0,0 3923 type = eSymbolTypeCompiler; 3924 break; 3925 3926 case N_VERSION: 3927 // compiler version: name,,NO_SECT,0,0 3928 type = eSymbolTypeCompiler; 3929 break; 3930 3931 case N_OLEVEL: 3932 // compiler -O level: name,,NO_SECT,0,0 3933 type = eSymbolTypeCompiler; 3934 break; 3935 3936 case N_PSYM: 3937 // parameter: name,,NO_SECT,type,offset 3938 type = eSymbolTypeVariable; 3939 break; 3940 3941 case N_ENTRY: 3942 // alternate entry: name,,n_sect,linenumber,address 3943 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3944 type = eSymbolTypeLineEntry; 3945 break; 3946 3947 //---------------------------------------------------------------------- 3948 // Left and Right Braces 3949 //---------------------------------------------------------------------- 3950 case N_LBRAC: 3951 // left bracket: 0,,NO_SECT,nesting level,address 3952 // We use the current number of symbols in the symbol table in lieu of 3953 // using nlist_idx in case we ever start trimming entries out 3954 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3955 N_BRAC_indexes.push_back(sym_idx); 3956 type = eSymbolTypeScopeBegin; 3957 break; 3958 3959 case N_RBRAC: 3960 // right bracket: 0,,NO_SECT,nesting level,address 3961 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 3962 // so that we can always skip the entire symbol if we need to navigate 3963 // more quickly at the source level when parsing STABS 3964 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3965 if ( !N_BRAC_indexes.empty() ) 3966 { 3967 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3968 symbol_ptr->SetByteSize(sym_idx + 1); 3969 symbol_ptr->SetSizeIsSibling(true); 3970 N_BRAC_indexes.pop_back(); 3971 } 3972 type = eSymbolTypeScopeEnd; 3973 break; 3974 3975 case N_EXCL: 3976 // deleted include file: name,,NO_SECT,0,sum 3977 type = eSymbolTypeHeaderFile; 3978 break; 3979 3980 //---------------------------------------------------------------------- 3981 // COMM scopes 3982 //---------------------------------------------------------------------- 3983 case N_BCOMM: 3984 // begin common: name,,NO_SECT,0,0 3985 // We use the current number of symbols in the symbol table in lieu of 3986 // using nlist_idx in case we ever start trimming entries out 3987 type = eSymbolTypeScopeBegin; 3988 N_COMM_indexes.push_back(sym_idx); 3989 break; 3990 3991 case N_ECOML: 3992 // end common (local name): 0,,n_sect,0,address 3993 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3994 // Fall through 3995 3996 case N_ECOMM: 3997 // end common: name,,n_sect,0,0 3998 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 3999 // so that we can always skip the entire symbol if we need to navigate 4000 // more quickly at the source level when parsing STABS 4001 if ( !N_COMM_indexes.empty() ) 4002 { 4003 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4004 symbol_ptr->SetByteSize(sym_idx + 1); 4005 symbol_ptr->SetSizeIsSibling(true); 4006 N_COMM_indexes.pop_back(); 4007 } 4008 type = eSymbolTypeScopeEnd; 4009 break; 4010 4011 case N_LENG: 4012 // second stab entry with length information 4013 type = eSymbolTypeAdditional; 4014 break; 4015 4016 default: break; 4017 } 4018 } 4019 else 4020 { 4021 //uint8_t n_pext = N_PEXT & nlist.n_type; 4022 uint8_t n_type = N_TYPE & nlist.n_type; 4023 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4024 4025 switch (n_type) 4026 { 4027 case N_INDR: 4028 { 4029 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4030 if (reexport_name_cstr && reexport_name_cstr[0]) 4031 { 4032 type = eSymbolTypeReExported; 4033 ConstString reexport_name(reexport_name_cstr + ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4034 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4035 set_value = false; 4036 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4037 indirect_symbol_names.insert(ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4038 } 4039 else 4040 type = eSymbolTypeUndefined; 4041 } 4042 break; 4043 4044 case N_UNDF: 4045 if (symbol_name && symbol_name[0]) 4046 { 4047 ConstString undefined_name(symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 4048 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4049 } 4050 // Fall through 4051 case N_PBUD: 4052 type = eSymbolTypeUndefined; 4053 break; 4054 4055 case N_ABS: 4056 type = eSymbolTypeAbsolute; 4057 break; 4058 4059 case N_SECT: 4060 { 4061 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 4062 4063 if (!symbol_section) 4064 { 4065 // TODO: warn about this? 4066 add_nlist = false; 4067 break; 4068 } 4069 4070 if (TEXT_eh_frame_sectID == nlist.n_sect) 4071 { 4072 type = eSymbolTypeException; 4073 } 4074 else 4075 { 4076 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4077 4078 switch (section_type) 4079 { 4080 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 4081 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 4082 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 4083 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 4084 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 4085 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 4086 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 4087 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 4088 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 4089 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 4090 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 4091 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 4092 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 4093 default: 4094 switch (symbol_section->GetType()) 4095 { 4096 case lldb::eSectionTypeCode: 4097 type = eSymbolTypeCode; 4098 break; 4099 case eSectionTypeData: 4100 case eSectionTypeDataCString: // Inlined C string data 4101 case eSectionTypeDataCStringPointers: // Pointers to C string data 4102 case eSectionTypeDataSymbolAddress: // Address of a symbol in the symbol table 4103 case eSectionTypeData4: 4104 case eSectionTypeData8: 4105 case eSectionTypeData16: 4106 type = eSymbolTypeData; 4107 break; 4108 default: 4109 break; 4110 } 4111 break; 4112 } 4113 4114 if (type == eSymbolTypeInvalid) 4115 { 4116 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 4117 if (symbol_section->IsDescendant (text_section_sp.get())) 4118 { 4119 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4120 S_ATTR_SELF_MODIFYING_CODE | 4121 S_ATTR_SOME_INSTRUCTIONS)) 4122 type = eSymbolTypeData; 4123 else 4124 type = eSymbolTypeCode; 4125 } 4126 else 4127 if (symbol_section->IsDescendant(data_section_sp.get())) 4128 { 4129 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 4130 { 4131 type = eSymbolTypeRuntime; 4132 4133 if (symbol_name && 4134 symbol_name[0] == '_' && 4135 symbol_name[1] == 'O' && 4136 symbol_name[2] == 'B') 4137 { 4138 llvm::StringRef symbol_name_ref(symbol_name); 4139 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 4140 { 4141 symbol_name_non_abi_mangled = symbol_name + 1; 4142 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 4143 type = eSymbolTypeObjCClass; 4144 demangled_is_synthesized = true; 4145 } 4146 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 4147 { 4148 symbol_name_non_abi_mangled = symbol_name + 1; 4149 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 4150 type = eSymbolTypeObjCMetaClass; 4151 demangled_is_synthesized = true; 4152 } 4153 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 4154 { 4155 symbol_name_non_abi_mangled = symbol_name + 1; 4156 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 4157 type = eSymbolTypeObjCIVar; 4158 demangled_is_synthesized = true; 4159 } 4160 } 4161 } 4162 else 4163 if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 4164 { 4165 type = eSymbolTypeException; 4166 } 4167 else 4168 { 4169 type = eSymbolTypeData; 4170 } 4171 } 4172 else 4173 if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 4174 { 4175 type = eSymbolTypeTrampoline; 4176 } 4177 else 4178 if (symbol_section->IsDescendant(objc_section_sp.get())) 4179 { 4180 type = eSymbolTypeRuntime; 4181 if (symbol_name && symbol_name[0] == '.') 4182 { 4183 llvm::StringRef symbol_name_ref(symbol_name); 4184 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 4185 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 4186 { 4187 symbol_name_non_abi_mangled = symbol_name; 4188 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4189 type = eSymbolTypeObjCClass; 4190 demangled_is_synthesized = true; 4191 } 4192 } 4193 } 4194 } 4195 } 4196 } 4197 break; 4198 } 4199 } 4200 4201 if (add_nlist) 4202 { 4203 uint64_t symbol_value = nlist.n_value; 4204 4205 if (symbol_name_non_abi_mangled) 4206 { 4207 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 4208 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 4209 } 4210 else 4211 { 4212 bool symbol_name_is_mangled = false; 4213 4214 if (symbol_name && symbol_name[0] == '_') 4215 { 4216 symbol_name_is_mangled = symbol_name[1] == '_'; 4217 symbol_name++; // Skip the leading underscore 4218 } 4219 4220 if (symbol_name) 4221 { 4222 ConstString const_symbol_name(symbol_name); 4223 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 4224 } 4225 } 4226 4227 if (is_gsym) 4228 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 4229 4230 if (symbol_section) 4231 { 4232 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4233 if (symbol_byte_size == 0 && function_starts_count > 0) 4234 { 4235 addr_t symbol_lookup_file_addr = nlist.n_value; 4236 // Do an exact address match for non-ARM addresses, else get the closest since 4237 // the symbol might be a thumb symbol which has an address with bit zero set 4238 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 4239 if (is_arm && func_start_entry) 4240 { 4241 // Verify that the function start address is the symbol address (ARM) 4242 // or the symbol address + 1 (thumb) 4243 if (func_start_entry->addr != symbol_lookup_file_addr && 4244 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 4245 { 4246 // Not the right entry, NULL it out... 4247 func_start_entry = NULL; 4248 } 4249 } 4250 if (func_start_entry) 4251 { 4252 func_start_entry->data = true; 4253 4254 addr_t symbol_file_addr = func_start_entry->addr; 4255 if (is_arm) 4256 symbol_file_addr &= 0xfffffffffffffffeull; 4257 4258 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 4259 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 4260 if (next_func_start_entry) 4261 { 4262 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4263 // Be sure the clear the Thumb address bit when we calculate the size 4264 // from the current and next address 4265 if (is_arm) 4266 next_symbol_file_addr &= 0xfffffffffffffffeull; 4267 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 4268 } 4269 else 4270 { 4271 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4272 } 4273 } 4274 } 4275 symbol_value -= section_file_addr; 4276 } 4277 4278 if (is_debug == false) 4279 { 4280 if (type == eSymbolTypeCode) 4281 { 4282 // See if we can find a N_FUN entry for any code symbols. 4283 // If we do find a match, and the name matches, then we 4284 // can merge the two into just the function symbol to avoid 4285 // duplicate entries in the symbol table 4286 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 4287 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4288 if (range.first != range.second) 4289 { 4290 bool found_it = false; 4291 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 4292 { 4293 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 4294 { 4295 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4296 // We just need the flags from the linker symbol, so put these flags 4297 // into the N_FUN flags to avoid duplicate symbols in the symbol table 4298 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4299 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4300 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 4301 sym[pos->second].SetType (eSymbolTypeResolver); 4302 sym[sym_idx].Clear(); 4303 found_it = true; 4304 break; 4305 } 4306 } 4307 if (found_it) 4308 continue; 4309 } 4310 else 4311 { 4312 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 4313 type = eSymbolTypeResolver; 4314 } 4315 } 4316 else if (type == eSymbolTypeData || 4317 type == eSymbolTypeObjCClass || 4318 type == eSymbolTypeObjCMetaClass || 4319 type == eSymbolTypeObjCIVar ) 4320 { 4321 // See if we can find a N_STSYM entry for any data symbols. 4322 // If we do find a match, and the name matches, then we 4323 // can merge the two into just the Static symbol to avoid 4324 // duplicate entries in the symbol table 4325 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 4326 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4327 if (range.first != range.second) 4328 { 4329 bool found_it = false; 4330 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 4331 { 4332 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 4333 { 4334 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4335 // We just need the flags from the linker symbol, so put these flags 4336 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 4337 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4338 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4339 sym[sym_idx].Clear(); 4340 found_it = true; 4341 break; 4342 } 4343 } 4344 if (found_it) 4345 continue; 4346 } 4347 else 4348 { 4349 // Combine N_GSYM stab entries with the non stab symbol 4350 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 4351 if (pos != N_GSYM_name_to_sym_idx.end()) 4352 { 4353 const uint32_t GSYM_sym_idx = pos->second; 4354 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4355 // Copy the address, because often the N_GSYM address has an invalid address of zero 4356 // when the global is a common symbol 4357 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 4358 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 4359 // We just need the flags from the linker symbol, so put these flags 4360 // into the N_GSYM flags to avoid duplicate symbols in the symbol table 4361 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4362 sym[sym_idx].Clear(); 4363 continue; 4364 } 4365 } 4366 } 4367 } 4368 4369 sym[sym_idx].SetID (nlist_idx); 4370 sym[sym_idx].SetType (type); 4371 if (set_value) 4372 { 4373 sym[sym_idx].GetAddress().SetSection (symbol_section); 4374 sym[sym_idx].GetAddress().SetOffset (symbol_value); 4375 } 4376 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4377 4378 if (symbol_byte_size > 0) 4379 sym[sym_idx].SetByteSize(symbol_byte_size); 4380 4381 if (demangled_is_synthesized) 4382 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4383 4384 ++sym_idx; 4385 } 4386 else 4387 { 4388 sym[sym_idx].Clear(); 4389 } 4390 } 4391 4392 for (const auto &pos :reexport_shlib_needs_fixup) 4393 { 4394 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4395 if (undef_pos != undefined_name_to_desc.end()) 4396 { 4397 const uint8_t dylib_ordinal = llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4398 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4399 sym[pos.first].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(dylib_ordinal-1)); 4400 } 4401 } 4402 4403 } 4404 4405 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4406 4407 if (function_starts_count > 0) 4408 { 4409 char synthetic_function_symbol[PATH_MAX]; 4410 uint32_t num_synthetic_function_symbols = 0; 4411 for (i=0; i<function_starts_count; ++i) 4412 { 4413 if (function_starts.GetEntryRef (i).data == false) 4414 ++num_synthetic_function_symbols; 4415 } 4416 4417 if (num_synthetic_function_symbols > 0) 4418 { 4419 if (num_syms < sym_idx + num_synthetic_function_symbols) 4420 { 4421 num_syms = sym_idx + num_synthetic_function_symbols; 4422 sym = symtab->Resize (num_syms); 4423 } 4424 uint32_t synthetic_function_symbol_idx = 0; 4425 for (i=0; i<function_starts_count; ++i) 4426 { 4427 const FunctionStarts::Entry *func_start_entry = function_starts.GetEntryAtIndex (i); 4428 if (func_start_entry->data == false) 4429 { 4430 addr_t symbol_file_addr = func_start_entry->addr; 4431 uint32_t symbol_flags = 0; 4432 if (is_arm) 4433 { 4434 if (symbol_file_addr & 1) 4435 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4436 symbol_file_addr &= 0xfffffffffffffffeull; 4437 } 4438 Address symbol_addr; 4439 if (module_sp->ResolveFileAddress (symbol_file_addr, symbol_addr)) 4440 { 4441 SectionSP symbol_section (symbol_addr.GetSection()); 4442 uint32_t symbol_byte_size = 0; 4443 if (symbol_section) 4444 { 4445 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4446 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 4447 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 4448 if (next_func_start_entry) 4449 { 4450 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4451 if (is_arm) 4452 next_symbol_file_addr &= 0xfffffffffffffffeull; 4453 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 4454 } 4455 else 4456 { 4457 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4458 } 4459 snprintf (synthetic_function_symbol, 4460 sizeof(synthetic_function_symbol), 4461 "___lldb_unnamed_function%u$$%s", 4462 ++synthetic_function_symbol_idx, 4463 module_sp->GetFileSpec().GetFilename().GetCString()); 4464 sym[sym_idx].SetID (synthetic_sym_id++); 4465 sym[sym_idx].GetMangled().SetDemangledName(ConstString(synthetic_function_symbol)); 4466 sym[sym_idx].SetType (eSymbolTypeCode); 4467 sym[sym_idx].SetIsSynthetic (true); 4468 sym[sym_idx].GetAddress() = symbol_addr; 4469 if (symbol_flags) 4470 sym[sym_idx].SetFlags (symbol_flags); 4471 if (symbol_byte_size) 4472 sym[sym_idx].SetByteSize (symbol_byte_size); 4473 ++sym_idx; 4474 } 4475 } 4476 } 4477 } 4478 } 4479 } 4480 4481 // Trim our symbols down to just what we ended up with after 4482 // removing any symbols. 4483 if (sym_idx < num_syms) 4484 { 4485 num_syms = sym_idx; 4486 sym = symtab->Resize (num_syms); 4487 } 4488 4489 // Now synthesize indirect symbols 4490 if (m_dysymtab.nindirectsyms != 0) 4491 { 4492 if (indirect_symbol_index_data.GetByteSize()) 4493 { 4494 NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end(); 4495 4496 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx) 4497 { 4498 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == S_SYMBOL_STUBS) 4499 { 4500 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4501 if (symbol_stub_byte_size == 0) 4502 continue; 4503 4504 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4505 4506 if (num_symbol_stubs == 0) 4507 continue; 4508 4509 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1; 4510 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) 4511 { 4512 const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx; 4513 const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size); 4514 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4515 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4)) 4516 { 4517 const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset); 4518 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4519 continue; 4520 4521 NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id); 4522 Symbol *stub_symbol = NULL; 4523 if (index_pos != end_index_pos) 4524 { 4525 // We have a remapping from the original nlist index to 4526 // a current symbol index, so just look this up by index 4527 stub_symbol = symtab->SymbolAtIndex (index_pos->second); 4528 } 4529 else 4530 { 4531 // We need to lookup a symbol using the original nlist 4532 // symbol index since this index is coming from the 4533 // S_SYMBOL_STUBS 4534 stub_symbol = symtab->FindSymbolByID (stub_sym_id); 4535 } 4536 4537 if (stub_symbol) 4538 { 4539 Address so_addr(symbol_stub_addr, section_list); 4540 4541 if (stub_symbol->GetType() == eSymbolTypeUndefined) 4542 { 4543 // Change the external symbol into a trampoline that makes sense 4544 // These symbols were N_UNDF N_EXT, and are useless to us, so we 4545 // can re-use them so we don't have to make up a synthetic symbol 4546 // for no good reason. 4547 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 4548 stub_symbol->SetType (eSymbolTypeTrampoline); 4549 else 4550 stub_symbol->SetType (eSymbolTypeResolver); 4551 stub_symbol->SetExternal (false); 4552 stub_symbol->GetAddress() = so_addr; 4553 stub_symbol->SetByteSize (symbol_stub_byte_size); 4554 } 4555 else 4556 { 4557 // Make a synthetic symbol to describe the trampoline stub 4558 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4559 if (sym_idx >= num_syms) 4560 { 4561 sym = symtab->Resize (++num_syms); 4562 stub_symbol = NULL; // this pointer no longer valid 4563 } 4564 sym[sym_idx].SetID (synthetic_sym_id++); 4565 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4566 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 4567 sym[sym_idx].SetType (eSymbolTypeTrampoline); 4568 else 4569 sym[sym_idx].SetType (eSymbolTypeResolver); 4570 sym[sym_idx].SetIsSynthetic (true); 4571 sym[sym_idx].GetAddress() = so_addr; 4572 sym[sym_idx].SetByteSize (symbol_stub_byte_size); 4573 ++sym_idx; 4574 } 4575 } 4576 else 4577 { 4578 if (log) 4579 log->Warning ("symbol stub referencing symbol table symbol %u that isn't in our minimal symbol table, fix this!!!", stub_sym_id); 4580 } 4581 } 4582 } 4583 } 4584 } 4585 } 4586 } 4587 4588 4589 if (!trie_entries.empty()) 4590 { 4591 for (const auto &e : trie_entries) 4592 { 4593 if (e.entry.import_name) 4594 { 4595 // Only add indirect symbols from the Trie entries if we 4596 // didn't have a N_INDR nlist entry for this already 4597 if (indirect_symbol_names.find(e.entry.name) == indirect_symbol_names.end()) 4598 { 4599 // Make a synthetic symbol to describe re-exported symbol. 4600 if (sym_idx >= num_syms) 4601 sym = symtab->Resize (++num_syms); 4602 sym[sym_idx].SetID (synthetic_sym_id++); 4603 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4604 sym[sym_idx].SetType (eSymbolTypeReExported); 4605 sym[sym_idx].SetIsSynthetic (true); 4606 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4607 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) 4608 { 4609 sym[sym_idx].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(e.entry.other-1)); 4610 } 4611 ++sym_idx; 4612 } 4613 } 4614 } 4615 } 4616 4617 4618 4619 // StreamFile s(stdout, false); 4620 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4621 // symtab->Dump(&s, NULL, eSortOrderNone); 4622 // Set symbol byte sizes correctly since mach-o nlist entries don't have sizes 4623 symtab->CalculateSymbolSizes(); 4624 4625 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4626 // symtab->Dump(&s, NULL, eSortOrderNone); 4627 4628 return symtab->GetNumSymbols(); 4629 } 4630 return 0; 4631 } 4632 4633 4634 void 4635 ObjectFileMachO::Dump (Stream *s) 4636 { 4637 ModuleSP module_sp(GetModule()); 4638 if (module_sp) 4639 { 4640 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4641 s->Printf("%p: ", static_cast<void*>(this)); 4642 s->Indent(); 4643 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4644 s->PutCString("ObjectFileMachO64"); 4645 else 4646 s->PutCString("ObjectFileMachO32"); 4647 4648 ArchSpec header_arch; 4649 GetArchitecture(header_arch); 4650 4651 *s << ", file = '" << m_file << "', arch = " << header_arch.GetArchitectureName() << "\n"; 4652 4653 SectionList *sections = GetSectionList(); 4654 if (sections) 4655 sections->Dump(s, NULL, true, UINT32_MAX); 4656 4657 if (m_symtab_ap.get()) 4658 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4659 } 4660 } 4661 4662 bool 4663 ObjectFileMachO::GetUUID (const llvm::MachO::mach_header &header, 4664 const lldb_private::DataExtractor &data, 4665 lldb::offset_t lc_offset, 4666 lldb_private::UUID& uuid) 4667 { 4668 uint32_t i; 4669 struct uuid_command load_cmd; 4670 4671 lldb::offset_t offset = lc_offset; 4672 for (i=0; i<header.ncmds; ++i) 4673 { 4674 const lldb::offset_t cmd_offset = offset; 4675 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4676 break; 4677 4678 if (load_cmd.cmd == LC_UUID) 4679 { 4680 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4681 4682 if (uuid_bytes) 4683 { 4684 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4685 // We pretend these object files have no UUID to prevent crashing. 4686 4687 const uint8_t opencl_uuid[] = { 0x8c, 0x8e, 0xb3, 0x9b, 4688 0x3b, 0xa8, 4689 0x4b, 0x16, 4690 0xb6, 0xa4, 4691 0x27, 0x63, 0xbb, 0x14, 0xf0, 0x0d }; 4692 4693 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4694 return false; 4695 4696 uuid.SetBytes (uuid_bytes); 4697 return true; 4698 } 4699 return false; 4700 } 4701 offset = cmd_offset + load_cmd.cmdsize; 4702 } 4703 return false; 4704 } 4705 4706 4707 bool 4708 ObjectFileMachO::GetArchitecture (const llvm::MachO::mach_header &header, 4709 const lldb_private::DataExtractor &data, 4710 lldb::offset_t lc_offset, 4711 ArchSpec &arch) 4712 { 4713 arch.SetArchitecture (eArchTypeMachO, header.cputype, header.cpusubtype); 4714 4715 if (arch.IsValid()) 4716 { 4717 llvm::Triple &triple = arch.GetTriple(); 4718 if (header.filetype == MH_PRELOAD) 4719 { 4720 // Set OS to "unknown" - this is a standalone binary with no dyld et al 4721 triple.setOS(llvm::Triple::UnknownOS); 4722 return true; 4723 } 4724 else 4725 { 4726 struct load_command load_cmd; 4727 4728 lldb::offset_t offset = lc_offset; 4729 for (uint32_t i=0; i<header.ncmds; ++i) 4730 { 4731 const lldb::offset_t cmd_offset = offset; 4732 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4733 break; 4734 4735 switch (load_cmd.cmd) 4736 { 4737 case LC_VERSION_MIN_IPHONEOS: 4738 triple.setOS (llvm::Triple::IOS); 4739 return true; 4740 4741 case LC_VERSION_MIN_MACOSX: 4742 triple.setOS (llvm::Triple::MacOSX); 4743 return true; 4744 4745 default: 4746 break; 4747 } 4748 4749 offset = cmd_offset + load_cmd.cmdsize; 4750 } 4751 4752 // Only set the OS to iOS for ARM, we don't want to set it for x86 and x86_64. 4753 // We do this because we now have MacOSX or iOS as the OS value for x86 and 4754 // x86_64 for normal desktop (MacOSX) and simulator (iOS) binaries. And if 4755 // we compare a "x86_64-apple-ios" to a "x86_64-apple-" triple, it will say 4756 // it is compatible (because the OS is unspecified in the second one and will 4757 // match anything in the first 4758 if (header.cputype == CPU_TYPE_ARM || header.cputype == CPU_TYPE_ARM64) 4759 triple.setOS (llvm::Triple::IOS); 4760 } 4761 } 4762 return arch.IsValid(); 4763 } 4764 4765 bool 4766 ObjectFileMachO::GetUUID (lldb_private::UUID* uuid) 4767 { 4768 ModuleSP module_sp(GetModule()); 4769 if (module_sp) 4770 { 4771 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4772 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4773 return GetUUID (m_header, m_data, offset, *uuid); 4774 } 4775 return false; 4776 } 4777 4778 4779 uint32_t 4780 ObjectFileMachO::GetDependentModules (FileSpecList& files) 4781 { 4782 uint32_t count = 0; 4783 ModuleSP module_sp(GetModule()); 4784 if (module_sp) 4785 { 4786 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4787 struct load_command load_cmd; 4788 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4789 const bool resolve_path = false; // Don't resolve the dependent file paths since they may not reside on this system 4790 uint32_t i; 4791 for (i=0; i<m_header.ncmds; ++i) 4792 { 4793 const uint32_t cmd_offset = offset; 4794 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4795 break; 4796 4797 switch (load_cmd.cmd) 4798 { 4799 case LC_LOAD_DYLIB: 4800 case LC_LOAD_WEAK_DYLIB: 4801 case LC_REEXPORT_DYLIB: 4802 case LC_LOAD_DYLINKER: 4803 case LC_LOADFVMLIB: 4804 case LC_LOAD_UPWARD_DYLIB: 4805 { 4806 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 4807 const char *path = m_data.PeekCStr(name_offset); 4808 // Skip any path that starts with '@' since these are usually: 4809 // @executable_path/.../file 4810 // @rpath/.../file 4811 if (path && path[0] != '@') 4812 { 4813 FileSpec file_spec(path, resolve_path); 4814 if (files.AppendIfUnique(file_spec)) 4815 count++; 4816 } 4817 } 4818 break; 4819 4820 default: 4821 break; 4822 } 4823 offset = cmd_offset + load_cmd.cmdsize; 4824 } 4825 } 4826 return count; 4827 } 4828 4829 lldb_private::Address 4830 ObjectFileMachO::GetEntryPointAddress () 4831 { 4832 // If the object file is not an executable it can't hold the entry point. m_entry_point_address 4833 // is initialized to an invalid address, so we can just return that. 4834 // If m_entry_point_address is valid it means we've found it already, so return the cached value. 4835 4836 if (!IsExecutable() || m_entry_point_address.IsValid()) 4837 return m_entry_point_address; 4838 4839 // Otherwise, look for the UnixThread or Thread command. The data for the Thread command is given in 4840 // /usr/include/mach-o.h, but it is basically: 4841 // 4842 // uint32_t flavor - this is the flavor argument you would pass to thread_get_state 4843 // uint32_t count - this is the count of longs in the thread state data 4844 // struct XXX_thread_state state - this is the structure from <machine/thread_status.h> corresponding to the flavor. 4845 // <repeat this trio> 4846 // 4847 // So we just keep reading the various register flavors till we find the GPR one, then read the PC out of there. 4848 // FIXME: We will need to have a "RegisterContext data provider" class at some point that can get all the registers 4849 // out of data in this form & attach them to a given thread. That should underlie the MacOS X User process plugin, 4850 // and we'll also need it for the MacOS X Core File process plugin. When we have that we can also use it here. 4851 // 4852 // For now we hard-code the offsets and flavors we need: 4853 // 4854 // 4855 4856 ModuleSP module_sp(GetModule()); 4857 if (module_sp) 4858 { 4859 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4860 struct load_command load_cmd; 4861 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4862 uint32_t i; 4863 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 4864 bool done = false; 4865 4866 for (i=0; i<m_header.ncmds; ++i) 4867 { 4868 const lldb::offset_t cmd_offset = offset; 4869 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4870 break; 4871 4872 switch (load_cmd.cmd) 4873 { 4874 case LC_UNIXTHREAD: 4875 case LC_THREAD: 4876 { 4877 while (offset < cmd_offset + load_cmd.cmdsize) 4878 { 4879 uint32_t flavor = m_data.GetU32(&offset); 4880 uint32_t count = m_data.GetU32(&offset); 4881 if (count == 0) 4882 { 4883 // We've gotten off somehow, log and exit; 4884 return m_entry_point_address; 4885 } 4886 4887 switch (m_header.cputype) 4888 { 4889 case llvm::MachO::CPU_TYPE_ARM: 4890 if (flavor == 1) // ARM_THREAD_STATE from mach/arm/thread_status.h 4891 { 4892 offset += 60; // This is the offset of pc in the GPR thread state data structure. 4893 start_address = m_data.GetU32(&offset); 4894 done = true; 4895 } 4896 break; 4897 case llvm::MachO::CPU_TYPE_ARM64: 4898 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 4899 { 4900 offset += 256; // This is the offset of pc in the GPR thread state data structure. 4901 start_address = m_data.GetU64(&offset); 4902 done = true; 4903 } 4904 break; 4905 case llvm::MachO::CPU_TYPE_I386: 4906 if (flavor == 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 4907 { 4908 offset += 40; // This is the offset of eip in the GPR thread state data structure. 4909 start_address = m_data.GetU32(&offset); 4910 done = true; 4911 } 4912 break; 4913 case llvm::MachO::CPU_TYPE_X86_64: 4914 if (flavor == 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 4915 { 4916 offset += 16 * 8; // This is the offset of rip in the GPR thread state data structure. 4917 start_address = m_data.GetU64(&offset); 4918 done = true; 4919 } 4920 break; 4921 default: 4922 return m_entry_point_address; 4923 } 4924 // Haven't found the GPR flavor yet, skip over the data for this flavor: 4925 if (done) 4926 break; 4927 offset += count * 4; 4928 } 4929 } 4930 break; 4931 case LC_MAIN: 4932 { 4933 ConstString text_segment_name ("__TEXT"); 4934 uint64_t entryoffset = m_data.GetU64(&offset); 4935 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 4936 if (text_segment_sp) 4937 { 4938 done = true; 4939 start_address = text_segment_sp->GetFileAddress() + entryoffset; 4940 } 4941 } 4942 4943 default: 4944 break; 4945 } 4946 if (done) 4947 break; 4948 4949 // Go to the next load command: 4950 offset = cmd_offset + load_cmd.cmdsize; 4951 } 4952 4953 if (start_address != LLDB_INVALID_ADDRESS) 4954 { 4955 // We got the start address from the load commands, so now resolve that address in the sections 4956 // of this ObjectFile: 4957 if (!m_entry_point_address.ResolveAddressUsingFileSections (start_address, GetSectionList())) 4958 { 4959 m_entry_point_address.Clear(); 4960 } 4961 } 4962 else 4963 { 4964 // We couldn't read the UnixThread load command - maybe it wasn't there. As a fallback look for the 4965 // "start" symbol in the main executable. 4966 4967 ModuleSP module_sp (GetModule()); 4968 4969 if (module_sp) 4970 { 4971 SymbolContextList contexts; 4972 SymbolContext context; 4973 if (module_sp->FindSymbolsWithNameAndType(ConstString ("start"), eSymbolTypeCode, contexts)) 4974 { 4975 if (contexts.GetContextAtIndex(0, context)) 4976 m_entry_point_address = context.symbol->GetAddress(); 4977 } 4978 } 4979 } 4980 } 4981 4982 return m_entry_point_address; 4983 4984 } 4985 4986 lldb_private::Address 4987 ObjectFileMachO::GetHeaderAddress () 4988 { 4989 lldb_private::Address header_addr; 4990 SectionList *section_list = GetSectionList(); 4991 if (section_list) 4992 { 4993 SectionSP text_segment_sp (section_list->FindSectionByName (GetSegmentNameTEXT())); 4994 if (text_segment_sp) 4995 { 4996 header_addr.SetSection (text_segment_sp); 4997 header_addr.SetOffset (0); 4998 } 4999 } 5000 return header_addr; 5001 } 5002 5003 uint32_t 5004 ObjectFileMachO::GetNumThreadContexts () 5005 { 5006 ModuleSP module_sp(GetModule()); 5007 if (module_sp) 5008 { 5009 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 5010 if (!m_thread_context_offsets_valid) 5011 { 5012 m_thread_context_offsets_valid = true; 5013 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5014 FileRangeArray::Entry file_range; 5015 thread_command thread_cmd; 5016 for (uint32_t i=0; i<m_header.ncmds; ++i) 5017 { 5018 const uint32_t cmd_offset = offset; 5019 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 5020 break; 5021 5022 if (thread_cmd.cmd == LC_THREAD) 5023 { 5024 file_range.SetRangeBase (offset); 5025 file_range.SetByteSize (thread_cmd.cmdsize - 8); 5026 m_thread_context_offsets.Append (file_range); 5027 } 5028 offset = cmd_offset + thread_cmd.cmdsize; 5029 } 5030 } 5031 } 5032 return m_thread_context_offsets.GetSize(); 5033 } 5034 5035 lldb::RegisterContextSP 5036 ObjectFileMachO::GetThreadContextAtIndex (uint32_t idx, lldb_private::Thread &thread) 5037 { 5038 lldb::RegisterContextSP reg_ctx_sp; 5039 5040 ModuleSP module_sp(GetModule()); 5041 if (module_sp) 5042 { 5043 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 5044 if (!m_thread_context_offsets_valid) 5045 GetNumThreadContexts (); 5046 5047 const FileRangeArray::Entry *thread_context_file_range = m_thread_context_offsets.GetEntryAtIndex (idx); 5048 if (thread_context_file_range) 5049 { 5050 5051 DataExtractor data (m_data, 5052 thread_context_file_range->GetRangeBase(), 5053 thread_context_file_range->GetByteSize()); 5054 5055 switch (m_header.cputype) 5056 { 5057 case llvm::MachO::CPU_TYPE_ARM64: 5058 reg_ctx_sp.reset (new RegisterContextDarwin_arm64_Mach (thread, data)); 5059 break; 5060 5061 case llvm::MachO::CPU_TYPE_ARM: 5062 reg_ctx_sp.reset (new RegisterContextDarwin_arm_Mach (thread, data)); 5063 break; 5064 5065 case llvm::MachO::CPU_TYPE_I386: 5066 reg_ctx_sp.reset (new RegisterContextDarwin_i386_Mach (thread, data)); 5067 break; 5068 5069 case llvm::MachO::CPU_TYPE_X86_64: 5070 reg_ctx_sp.reset (new RegisterContextDarwin_x86_64_Mach (thread, data)); 5071 break; 5072 } 5073 } 5074 } 5075 return reg_ctx_sp; 5076 } 5077 5078 5079 ObjectFile::Type 5080 ObjectFileMachO::CalculateType() 5081 { 5082 switch (m_header.filetype) 5083 { 5084 case MH_OBJECT: // 0x1u 5085 if (GetAddressByteSize () == 4) 5086 { 5087 // 32 bit kexts are just object files, but they do have a valid 5088 // UUID load command. 5089 UUID uuid; 5090 if (GetUUID(&uuid)) 5091 { 5092 // this checking for the UUID load command is not enough 5093 // we could eventually look for the symbol named 5094 // "OSKextGetCurrentIdentifier" as this is required of kexts 5095 if (m_strata == eStrataInvalid) 5096 m_strata = eStrataKernel; 5097 return eTypeSharedLibrary; 5098 } 5099 } 5100 return eTypeObjectFile; 5101 5102 case MH_EXECUTE: return eTypeExecutable; // 0x2u 5103 case MH_FVMLIB: return eTypeSharedLibrary; // 0x3u 5104 case MH_CORE: return eTypeCoreFile; // 0x4u 5105 case MH_PRELOAD: return eTypeSharedLibrary; // 0x5u 5106 case MH_DYLIB: return eTypeSharedLibrary; // 0x6u 5107 case MH_DYLINKER: return eTypeDynamicLinker; // 0x7u 5108 case MH_BUNDLE: return eTypeSharedLibrary; // 0x8u 5109 case MH_DYLIB_STUB: return eTypeStubLibrary; // 0x9u 5110 case MH_DSYM: return eTypeDebugInfo; // 0xAu 5111 case MH_KEXT_BUNDLE: return eTypeSharedLibrary; // 0xBu 5112 default: 5113 break; 5114 } 5115 return eTypeUnknown; 5116 } 5117 5118 ObjectFile::Strata 5119 ObjectFileMachO::CalculateStrata() 5120 { 5121 switch (m_header.filetype) 5122 { 5123 case MH_OBJECT: // 0x1u 5124 { 5125 // 32 bit kexts are just object files, but they do have a valid 5126 // UUID load command. 5127 UUID uuid; 5128 if (GetUUID(&uuid)) 5129 { 5130 // this checking for the UUID load command is not enough 5131 // we could eventually look for the symbol named 5132 // "OSKextGetCurrentIdentifier" as this is required of kexts 5133 if (m_type == eTypeInvalid) 5134 m_type = eTypeSharedLibrary; 5135 5136 return eStrataKernel; 5137 } 5138 } 5139 return eStrataUnknown; 5140 5141 case MH_EXECUTE: // 0x2u 5142 // Check for the MH_DYLDLINK bit in the flags 5143 if (m_header.flags & MH_DYLDLINK) 5144 { 5145 return eStrataUser; 5146 } 5147 else 5148 { 5149 SectionList *section_list = GetSectionList(); 5150 if (section_list) 5151 { 5152 static ConstString g_kld_section_name ("__KLD"); 5153 if (section_list->FindSectionByName(g_kld_section_name)) 5154 return eStrataKernel; 5155 } 5156 } 5157 return eStrataRawImage; 5158 5159 case MH_FVMLIB: return eStrataUser; // 0x3u 5160 case MH_CORE: return eStrataUnknown; // 0x4u 5161 case MH_PRELOAD: return eStrataRawImage; // 0x5u 5162 case MH_DYLIB: return eStrataUser; // 0x6u 5163 case MH_DYLINKER: return eStrataUser; // 0x7u 5164 case MH_BUNDLE: return eStrataUser; // 0x8u 5165 case MH_DYLIB_STUB: return eStrataUser; // 0x9u 5166 case MH_DSYM: return eStrataUnknown; // 0xAu 5167 case MH_KEXT_BUNDLE: return eStrataKernel; // 0xBu 5168 default: 5169 break; 5170 } 5171 return eStrataUnknown; 5172 } 5173 5174 5175 uint32_t 5176 ObjectFileMachO::GetVersion (uint32_t *versions, uint32_t num_versions) 5177 { 5178 ModuleSP module_sp(GetModule()); 5179 if (module_sp) 5180 { 5181 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 5182 struct dylib_command load_cmd; 5183 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5184 uint32_t version_cmd = 0; 5185 uint64_t version = 0; 5186 uint32_t i; 5187 for (i=0; i<m_header.ncmds; ++i) 5188 { 5189 const lldb::offset_t cmd_offset = offset; 5190 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5191 break; 5192 5193 if (load_cmd.cmd == LC_ID_DYLIB) 5194 { 5195 if (version_cmd == 0) 5196 { 5197 version_cmd = load_cmd.cmd; 5198 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5199 break; 5200 version = load_cmd.dylib.current_version; 5201 } 5202 break; // Break for now unless there is another more complete version 5203 // number load command in the future. 5204 } 5205 offset = cmd_offset + load_cmd.cmdsize; 5206 } 5207 5208 if (version_cmd == LC_ID_DYLIB) 5209 { 5210 if (versions != NULL && num_versions > 0) 5211 { 5212 if (num_versions > 0) 5213 versions[0] = (version & 0xFFFF0000ull) >> 16; 5214 if (num_versions > 1) 5215 versions[1] = (version & 0x0000FF00ull) >> 8; 5216 if (num_versions > 2) 5217 versions[2] = (version & 0x000000FFull); 5218 // Fill in an remaining version numbers with invalid values 5219 for (i=3; i<num_versions; ++i) 5220 versions[i] = UINT32_MAX; 5221 } 5222 // The LC_ID_DYLIB load command has a version with 3 version numbers 5223 // in it, so always return 3 5224 return 3; 5225 } 5226 } 5227 return false; 5228 } 5229 5230 bool 5231 ObjectFileMachO::GetArchitecture (ArchSpec &arch) 5232 { 5233 ModuleSP module_sp(GetModule()); 5234 if (module_sp) 5235 { 5236 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 5237 return GetArchitecture (m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), arch); 5238 } 5239 return false; 5240 } 5241 5242 5243 UUID 5244 ObjectFileMachO::GetProcessSharedCacheUUID (Process *process) 5245 { 5246 UUID uuid; 5247 if (process) 5248 { 5249 addr_t all_image_infos = process->GetImageInfoAddress(); 5250 5251 // The address returned by GetImageInfoAddress may be the address of dyld (don't want) 5252 // or it may be the address of the dyld_all_image_infos structure (want). The first four 5253 // bytes will be either the version field (all_image_infos) or a Mach-O file magic constant. 5254 // Version 13 and higher of dyld_all_image_infos is required to get the sharedCacheUUID field. 5255 5256 Error err; 5257 uint32_t version_or_magic = process->ReadUnsignedIntegerFromMemory (all_image_infos, 4, -1, err); 5258 if (version_or_magic != static_cast<uint32_t>(-1) 5259 && version_or_magic != MH_MAGIC 5260 && version_or_magic != MH_CIGAM 5261 && version_or_magic != MH_MAGIC_64 5262 && version_or_magic != MH_CIGAM_64 5263 && version_or_magic >= 13) 5264 { 5265 addr_t sharedCacheUUID_address = LLDB_INVALID_ADDRESS; 5266 int wordsize = process->GetAddressByteSize(); 5267 if (wordsize == 8) 5268 { 5269 sharedCacheUUID_address = all_image_infos + 160; // sharedCacheUUID <mach-o/dyld_images.h> 5270 } 5271 if (wordsize == 4) 5272 { 5273 sharedCacheUUID_address = all_image_infos + 84; // sharedCacheUUID <mach-o/dyld_images.h> 5274 } 5275 if (sharedCacheUUID_address != LLDB_INVALID_ADDRESS) 5276 { 5277 uuid_t shared_cache_uuid; 5278 if (process->ReadMemory (sharedCacheUUID_address, shared_cache_uuid, sizeof (uuid_t), err) == sizeof (uuid_t)) 5279 { 5280 uuid.SetBytes (shared_cache_uuid); 5281 } 5282 } 5283 } 5284 } 5285 return uuid; 5286 } 5287 5288 UUID 5289 ObjectFileMachO::GetLLDBSharedCacheUUID () 5290 { 5291 UUID uuid; 5292 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 5293 uint8_t *(*dyld_get_all_image_infos)(void); 5294 dyld_get_all_image_infos = (uint8_t*(*)()) dlsym (RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5295 if (dyld_get_all_image_infos) 5296 { 5297 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5298 if (dyld_all_image_infos_address) 5299 { 5300 uint32_t *version = (uint32_t*) dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5301 if (*version >= 13) 5302 { 5303 uuid_t *sharedCacheUUID_address = 0; 5304 int wordsize = sizeof (uint8_t *); 5305 if (wordsize == 8) 5306 { 5307 sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 160); // sharedCacheUUID <mach-o/dyld_images.h> 5308 } 5309 else 5310 { 5311 sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 84); // sharedCacheUUID <mach-o/dyld_images.h> 5312 } 5313 uuid.SetBytes (sharedCacheUUID_address); 5314 } 5315 } 5316 } 5317 #endif 5318 return uuid; 5319 } 5320 5321 uint32_t 5322 ObjectFileMachO::GetMinimumOSVersion (uint32_t *versions, uint32_t num_versions) 5323 { 5324 if (m_min_os_versions.empty()) 5325 { 5326 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5327 bool success = false; 5328 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 5329 { 5330 const lldb::offset_t load_cmd_offset = offset; 5331 5332 version_min_command lc; 5333 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5334 break; 5335 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 5336 { 5337 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 5338 { 5339 const uint32_t xxxx = lc.version >> 16; 5340 const uint32_t yy = (lc.version >> 8) & 0xffu; 5341 const uint32_t zz = lc.version & 0xffu; 5342 if (xxxx) 5343 { 5344 m_min_os_versions.push_back(xxxx); 5345 m_min_os_versions.push_back(yy); 5346 m_min_os_versions.push_back(zz); 5347 } 5348 success = true; 5349 } 5350 } 5351 offset = load_cmd_offset + lc.cmdsize; 5352 } 5353 5354 if (success == false) 5355 { 5356 // Push an invalid value so we don't keep trying to 5357 m_min_os_versions.push_back(UINT32_MAX); 5358 } 5359 } 5360 5361 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) 5362 { 5363 if (versions != NULL && num_versions > 0) 5364 { 5365 for (size_t i=0; i<num_versions; ++i) 5366 { 5367 if (i < m_min_os_versions.size()) 5368 versions[i] = m_min_os_versions[i]; 5369 else 5370 versions[i] = 0; 5371 } 5372 } 5373 return m_min_os_versions.size(); 5374 } 5375 // Call the superclasses version that will empty out the data 5376 return ObjectFile::GetMinimumOSVersion (versions, num_versions); 5377 } 5378 5379 uint32_t 5380 ObjectFileMachO::GetSDKVersion(uint32_t *versions, uint32_t num_versions) 5381 { 5382 if (m_sdk_versions.empty()) 5383 { 5384 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5385 bool success = false; 5386 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 5387 { 5388 const lldb::offset_t load_cmd_offset = offset; 5389 5390 version_min_command lc; 5391 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5392 break; 5393 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 5394 { 5395 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 5396 { 5397 const uint32_t xxxx = lc.sdk >> 16; 5398 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5399 const uint32_t zz = lc.sdk & 0xffu; 5400 if (xxxx) 5401 { 5402 m_sdk_versions.push_back(xxxx); 5403 m_sdk_versions.push_back(yy); 5404 m_sdk_versions.push_back(zz); 5405 } 5406 success = true; 5407 } 5408 } 5409 offset = load_cmd_offset + lc.cmdsize; 5410 } 5411 5412 if (success == false) 5413 { 5414 // Push an invalid value so we don't keep trying to 5415 m_sdk_versions.push_back(UINT32_MAX); 5416 } 5417 } 5418 5419 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) 5420 { 5421 if (versions != NULL && num_versions > 0) 5422 { 5423 for (size_t i=0; i<num_versions; ++i) 5424 { 5425 if (i < m_sdk_versions.size()) 5426 versions[i] = m_sdk_versions[i]; 5427 else 5428 versions[i] = 0; 5429 } 5430 } 5431 return m_sdk_versions.size(); 5432 } 5433 // Call the superclasses version that will empty out the data 5434 return ObjectFile::GetSDKVersion (versions, num_versions); 5435 } 5436 5437 5438 bool 5439 ObjectFileMachO::GetIsDynamicLinkEditor() 5440 { 5441 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5442 } 5443 5444 //------------------------------------------------------------------ 5445 // PluginInterface protocol 5446 //------------------------------------------------------------------ 5447 lldb_private::ConstString 5448 ObjectFileMachO::GetPluginName() 5449 { 5450 return GetPluginNameStatic(); 5451 } 5452 5453 uint32_t 5454 ObjectFileMachO::GetPluginVersion() 5455 { 5456 return 1; 5457 } 5458 5459 5460 bool 5461 ObjectFileMachO::SetLoadAddress (Target &target, 5462 lldb::addr_t value, 5463 bool value_is_offset) 5464 { 5465 ModuleSP module_sp = GetModule(); 5466 if (module_sp) 5467 { 5468 size_t num_loaded_sections = 0; 5469 SectionList *section_list = GetSectionList (); 5470 if (section_list) 5471 { 5472 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5473 const size_t num_sections = section_list->GetSize(); 5474 5475 const bool is_memory_image = (bool)m_process_wp.lock(); 5476 const Strata strata = GetStrata(); 5477 static ConstString g_linkedit_segname ("__LINKEDIT"); 5478 if (value_is_offset) 5479 { 5480 // "value" is an offset to apply to each top level segment 5481 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) 5482 { 5483 // Iterate through the object file sections to find all 5484 // of the sections that size on disk (to avoid __PAGEZERO) 5485 // and load them 5486 SectionSP section_sp (section_list->GetSectionAtIndex (sect_idx)); 5487 if (section_sp && 5488 section_sp->GetFileSize() > 0 && 5489 section_sp->IsThreadSpecific() == false && 5490 module_sp.get() == section_sp->GetModule().get()) 5491 { 5492 // Ignore __LINKEDIT and __DWARF segments 5493 if (section_sp->GetName() == g_linkedit_segname) 5494 { 5495 // Only map __LINKEDIT if we have an in memory image and this isn't 5496 // a kernel binary like a kext or mach_kernel. 5497 if (is_memory_image == false || strata == eStrataKernel) 5498 continue; 5499 } 5500 if (target.GetSectionLoadList().SetSectionLoadAddress (section_sp, section_sp->GetFileAddress() + value)) 5501 ++num_loaded_sections; 5502 } 5503 } 5504 } 5505 else 5506 { 5507 // "value" is the new base address of the mach_header, adjust each 5508 // section accordingly 5509 5510 // First find the address of the mach header which is the first non-zero 5511 // file sized section whose file offset is zero as this will be subtracted 5512 // from each other valid section's vmaddr and then get "base_addr" added to 5513 // it when loading the module in the target 5514 for (size_t sect_idx = 0; 5515 sect_idx < num_sections && mach_base_file_addr == LLDB_INVALID_ADDRESS; 5516 ++sect_idx) 5517 { 5518 // Iterate through the object file sections to find all 5519 // of the sections that size on disk (to avoid __PAGEZERO) 5520 // and load them 5521 Section *section = section_list->GetSectionAtIndex (sect_idx).get(); 5522 if (section && 5523 section->GetFileSize() > 0 && 5524 section->GetFileOffset() == 0 && 5525 section->IsThreadSpecific() == false && 5526 module_sp.get() == section->GetModule().get()) 5527 { 5528 // Ignore __LINKEDIT and __DWARF segments 5529 if (section->GetName() == g_linkedit_segname) 5530 { 5531 // Only map __LINKEDIT if we have an in memory image and this isn't 5532 // a kernel binary like a kext or mach_kernel. 5533 if (is_memory_image == false || strata == eStrataKernel) 5534 continue; 5535 } 5536 mach_base_file_addr = section->GetFileAddress(); 5537 } 5538 } 5539 5540 if (mach_base_file_addr != LLDB_INVALID_ADDRESS) 5541 { 5542 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) 5543 { 5544 // Iterate through the object file sections to find all 5545 // of the sections that size on disk (to avoid __PAGEZERO) 5546 // and load them 5547 SectionSP section_sp (section_list->GetSectionAtIndex (sect_idx)); 5548 if (section_sp && 5549 section_sp->GetFileSize() > 0 && 5550 section_sp->IsThreadSpecific() == false && 5551 module_sp.get() == section_sp->GetModule().get()) 5552 { 5553 // Ignore __LINKEDIT and __DWARF segments 5554 if (section_sp->GetName() == g_linkedit_segname) 5555 { 5556 // Only map __LINKEDIT if we have an in memory image and this isn't 5557 // a kernel binary like a kext or mach_kernel. 5558 if (is_memory_image == false || strata == eStrataKernel) 5559 continue; 5560 } 5561 if (target.GetSectionLoadList().SetSectionLoadAddress (section_sp, section_sp->GetFileAddress() - mach_base_file_addr + value)) 5562 ++num_loaded_sections; 5563 } 5564 } 5565 } 5566 } 5567 } 5568 return num_loaded_sections > 0; 5569 } 5570 return false; 5571 } 5572 5573 bool 5574 ObjectFileMachO::SaveCore (const lldb::ProcessSP &process_sp, 5575 const FileSpec &outfile, 5576 Error &error) 5577 { 5578 if (process_sp) 5579 { 5580 Target &target = process_sp->GetTarget(); 5581 const ArchSpec target_arch = target.GetArchitecture(); 5582 const llvm::Triple &target_triple = target_arch.GetTriple(); 5583 if (target_triple.getVendor() == llvm::Triple::Apple && 5584 (target_triple.getOS() == llvm::Triple::MacOSX || 5585 target_triple.getOS() == llvm::Triple::IOS)) 5586 { 5587 bool make_core = false; 5588 switch (target_arch.GetMachine()) 5589 { 5590 // arm64 core file writing is having some problem with writing down the 5591 // dyld shared images info struct and/or the main executable binary. May 5592 // turn out to be a debugserver problem, not sure yet. 5593 // case llvm::Triple::aarch64: 5594 5595 case llvm::Triple::arm: 5596 case llvm::Triple::x86: 5597 case llvm::Triple::x86_64: 5598 make_core = true; 5599 break; 5600 default: 5601 error.SetErrorStringWithFormat ("unsupported core architecture: %s", target_triple.str().c_str()); 5602 break; 5603 } 5604 5605 if (make_core) 5606 { 5607 std::vector<segment_command_64> segment_load_commands; 5608 // uint32_t range_info_idx = 0; 5609 MemoryRegionInfo range_info; 5610 Error range_error = process_sp->GetMemoryRegionInfo(0, range_info); 5611 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 5612 const ByteOrder byte_order = target_arch.GetByteOrder(); 5613 if (range_error.Success()) 5614 { 5615 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) 5616 { 5617 const addr_t addr = range_info.GetRange().GetRangeBase(); 5618 const addr_t size = range_info.GetRange().GetByteSize(); 5619 5620 if (size == 0) 5621 break; 5622 5623 // Calculate correct protections 5624 uint32_t prot = 0; 5625 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 5626 prot |= VM_PROT_READ; 5627 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 5628 prot |= VM_PROT_WRITE; 5629 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 5630 prot |= VM_PROT_EXECUTE; 5631 5632 // printf ("[%3u] [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 ") %c%c%c\n", 5633 // range_info_idx, 5634 // addr, 5635 // size, 5636 // (prot & VM_PROT_READ ) ? 'r' : '-', 5637 // (prot & VM_PROT_WRITE ) ? 'w' : '-', 5638 // (prot & VM_PROT_EXECUTE) ? 'x' : '-'); 5639 5640 if (prot != 0) 5641 { 5642 uint32_t cmd_type = LC_SEGMENT_64; 5643 uint32_t segment_size = sizeof (segment_command_64); 5644 if (addr_byte_size == 4) 5645 { 5646 cmd_type = LC_SEGMENT; 5647 segment_size = sizeof (segment_command); 5648 } 5649 segment_command_64 segment = { 5650 cmd_type, // uint32_t cmd; 5651 segment_size, // uint32_t cmdsize; 5652 {0}, // char segname[16]; 5653 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 5654 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 5655 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 5656 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 5657 prot, // uint32_t maxprot; 5658 prot, // uint32_t initprot; 5659 0, // uint32_t nsects; 5660 0 }; // uint32_t flags; 5661 segment_load_commands.push_back(segment); 5662 } 5663 else 5664 { 5665 // No protections and a size of 1 used to be returned from old 5666 // debugservers when we asked about a region that was past the 5667 // last memory region and it indicates the end... 5668 if (size == 1) 5669 break; 5670 } 5671 5672 range_error = process_sp->GetMemoryRegionInfo(range_info.GetRange().GetRangeEnd(), range_info); 5673 if (range_error.Fail()) 5674 break; 5675 } 5676 5677 StreamString buffer (Stream::eBinary, 5678 addr_byte_size, 5679 byte_order); 5680 5681 mach_header_64 mach_header; 5682 if (addr_byte_size == 8) 5683 { 5684 mach_header.magic = MH_MAGIC_64; 5685 } 5686 else 5687 { 5688 mach_header.magic = MH_MAGIC; 5689 } 5690 mach_header.cputype = target_arch.GetMachOCPUType(); 5691 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 5692 mach_header.filetype = MH_CORE; 5693 mach_header.ncmds = segment_load_commands.size(); 5694 mach_header.flags = 0; 5695 mach_header.reserved = 0; 5696 ThreadList &thread_list = process_sp->GetThreadList(); 5697 const uint32_t num_threads = thread_list.GetSize(); 5698 5699 // Make an array of LC_THREAD data items. Each one contains 5700 // the contents of the LC_THREAD load command. The data doesn't 5701 // contain the load command + load command size, we will 5702 // add the load command and load command size as we emit the data. 5703 std::vector<StreamString> LC_THREAD_datas(num_threads); 5704 for (auto &LC_THREAD_data : LC_THREAD_datas) 5705 { 5706 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 5707 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 5708 LC_THREAD_data.SetByteOrder(byte_order); 5709 } 5710 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) 5711 { 5712 ThreadSP thread_sp (thread_list.GetThreadAtIndex(thread_idx)); 5713 if (thread_sp) 5714 { 5715 switch (mach_header.cputype) 5716 { 5717 case llvm::MachO::CPU_TYPE_ARM64: 5718 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5719 break; 5720 5721 case llvm::MachO::CPU_TYPE_ARM: 5722 RegisterContextDarwin_arm_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5723 break; 5724 5725 case llvm::MachO::CPU_TYPE_I386: 5726 RegisterContextDarwin_i386_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5727 break; 5728 5729 case llvm::MachO::CPU_TYPE_X86_64: 5730 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5731 break; 5732 } 5733 5734 } 5735 } 5736 5737 // The size of the load command is the size of the segments... 5738 if (addr_byte_size == 8) 5739 { 5740 mach_header.sizeofcmds = segment_load_commands.size() * sizeof (struct segment_command_64); 5741 } 5742 else 5743 { 5744 mach_header.sizeofcmds = segment_load_commands.size() * sizeof (struct segment_command); 5745 } 5746 5747 // and the size of all LC_THREAD load command 5748 for (const auto &LC_THREAD_data : LC_THREAD_datas) 5749 { 5750 ++mach_header.ncmds; 5751 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 5752 } 5753 5754 printf ("mach_header: 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x\n", 5755 mach_header.magic, 5756 mach_header.cputype, 5757 mach_header.cpusubtype, 5758 mach_header.filetype, 5759 mach_header.ncmds, 5760 mach_header.sizeofcmds, 5761 mach_header.flags, 5762 mach_header.reserved); 5763 5764 // Write the mach header 5765 buffer.PutHex32(mach_header.magic); 5766 buffer.PutHex32(mach_header.cputype); 5767 buffer.PutHex32(mach_header.cpusubtype); 5768 buffer.PutHex32(mach_header.filetype); 5769 buffer.PutHex32(mach_header.ncmds); 5770 buffer.PutHex32(mach_header.sizeofcmds); 5771 buffer.PutHex32(mach_header.flags); 5772 if (addr_byte_size == 8) 5773 { 5774 buffer.PutHex32(mach_header.reserved); 5775 } 5776 5777 // Skip the mach header and all load commands and align to the next 5778 // 0x1000 byte boundary 5779 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 5780 if (file_offset & 0x00000fff) 5781 { 5782 file_offset += 0x00001000ull; 5783 file_offset &= (~0x00001000ull + 1); 5784 } 5785 5786 for (auto &segment : segment_load_commands) 5787 { 5788 segment.fileoff = file_offset; 5789 file_offset += segment.filesize; 5790 } 5791 5792 // Write out all of the LC_THREAD load commands 5793 for (const auto &LC_THREAD_data : LC_THREAD_datas) 5794 { 5795 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 5796 buffer.PutHex32(LC_THREAD); 5797 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 5798 buffer.Write(LC_THREAD_data.GetData(), LC_THREAD_data_size); 5799 } 5800 5801 // Write out all of the segment load commands 5802 for (const auto &segment : segment_load_commands) 5803 { 5804 printf ("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 5805 segment.cmd, 5806 segment.cmdsize, 5807 segment.vmaddr, 5808 segment.vmaddr + segment.vmsize, 5809 segment.fileoff, 5810 segment.filesize, 5811 segment.maxprot, 5812 segment.initprot, 5813 segment.nsects, 5814 segment.flags); 5815 5816 buffer.PutHex32(segment.cmd); 5817 buffer.PutHex32(segment.cmdsize); 5818 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 5819 if (addr_byte_size == 8) 5820 { 5821 buffer.PutHex64(segment.vmaddr); 5822 buffer.PutHex64(segment.vmsize); 5823 buffer.PutHex64(segment.fileoff); 5824 buffer.PutHex64(segment.filesize); 5825 } 5826 else 5827 { 5828 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 5829 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 5830 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 5831 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 5832 } 5833 buffer.PutHex32(segment.maxprot); 5834 buffer.PutHex32(segment.initprot); 5835 buffer.PutHex32(segment.nsects); 5836 buffer.PutHex32(segment.flags); 5837 } 5838 5839 File core_file; 5840 std::string core_file_path(outfile.GetPath()); 5841 error = core_file.Open(core_file_path.c_str(), 5842 File::eOpenOptionWrite | 5843 File::eOpenOptionTruncate | 5844 File::eOpenOptionCanCreate); 5845 if (error.Success()) 5846 { 5847 // Read 1 page at a time 5848 uint8_t bytes[0x1000]; 5849 // Write the mach header and load commands out to the core file 5850 size_t bytes_written = buffer.GetString().size(); 5851 error = core_file.Write(buffer.GetString().data(), bytes_written); 5852 if (error.Success()) 5853 { 5854 // Now write the file data for all memory segments in the process 5855 for (const auto &segment : segment_load_commands) 5856 { 5857 if (core_file.SeekFromStart(segment.fileoff) == -1) 5858 { 5859 error.SetErrorStringWithFormat("unable to seek to offset 0x%" PRIx64 " in '%s'", segment.fileoff, core_file_path.c_str()); 5860 break; 5861 } 5862 5863 printf ("Saving %" PRId64 " bytes of data for memory region at 0x%" PRIx64 "\n", segment.vmsize, segment.vmaddr); 5864 addr_t bytes_left = segment.vmsize; 5865 addr_t addr = segment.vmaddr; 5866 Error memory_read_error; 5867 while (bytes_left > 0 && error.Success()) 5868 { 5869 const size_t bytes_to_read = bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 5870 const size_t bytes_read = process_sp->ReadMemory(addr, bytes, bytes_to_read, memory_read_error); 5871 if (bytes_read == bytes_to_read) 5872 { 5873 size_t bytes_written = bytes_read; 5874 error = core_file.Write(bytes, bytes_written); 5875 bytes_left -= bytes_read; 5876 addr += bytes_read; 5877 } 5878 else 5879 { 5880 // Some pages within regions are not readable, those 5881 // should be zero filled 5882 memset (bytes, 0, bytes_to_read); 5883 size_t bytes_written = bytes_to_read; 5884 error = core_file.Write(bytes, bytes_written); 5885 bytes_left -= bytes_to_read; 5886 addr += bytes_to_read; 5887 } 5888 } 5889 } 5890 } 5891 } 5892 } 5893 else 5894 { 5895 error.SetErrorString("process doesn't support getting memory region info"); 5896 } 5897 } 5898 return true; // This is the right plug to handle saving core files for this process 5899 } 5900 } 5901 return false; 5902 } 5903 5904