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/Core/ArchSpec.h" 13 #include "lldb/Core/DataBuffer.h" 14 #include "lldb/Core/Debugger.h" 15 #include "lldb/Core/Error.h" 16 #include "lldb/Core/FileSpecList.h" 17 #include "lldb/Core/Log.h" 18 #include "lldb/Core/Module.h" 19 #include "lldb/Core/ModuleSpec.h" 20 #include "lldb/Core/PluginManager.h" 21 #include "lldb/Core/RangeMap.h" 22 #include "lldb/Core/Section.h" 23 #include "lldb/Core/StreamFile.h" 24 #include "lldb/Core/StreamString.h" 25 #include "lldb/Core/Timer.h" 26 #include "lldb/Core/UUID.h" 27 #include "lldb/Host/Host.h" 28 #include "lldb/Host/FileSpec.h" 29 #include "lldb/Symbol/ClangNamespaceDecl.h" 30 #include "lldb/Symbol/DWARFCallFrameInfo.h" 31 #include "lldb/Symbol/ObjectFile.h" 32 #include "lldb/Target/MemoryRegionInfo.h" 33 #include "lldb/Target/Platform.h" 34 #include "lldb/Target/Process.h" 35 #include "lldb/Target/SectionLoadList.h" 36 #include "lldb/Target/Target.h" 37 #include "lldb/Target/Thread.h" 38 #include "lldb/Target/ThreadList.h" 39 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 40 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 41 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 42 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 43 44 #include "lldb/Utility/SafeMachO.h" 45 46 #include "ObjectFileMachO.h" 47 48 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 49 // GetLLDBSharedCacheUUID() needs to call dlsym() 50 #include <dlfcn.h> 51 #endif 52 53 #ifndef __APPLE__ 54 #include "Utility/UuidCompatibility.h" 55 #endif 56 57 using namespace lldb; 58 using namespace lldb_private; 59 using namespace llvm::MachO; 60 61 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 62 { 63 public: 64 RegisterContextDarwin_x86_64_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 65 RegisterContextDarwin_x86_64 (thread, 0) 66 { 67 SetRegisterDataFrom_LC_THREAD (data); 68 } 69 70 virtual void 71 InvalidateAllRegisters () 72 { 73 // Do nothing... registers are always valid... 74 } 75 76 void 77 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 78 { 79 lldb::offset_t offset = 0; 80 SetError (GPRRegSet, Read, -1); 81 SetError (FPURegSet, Read, -1); 82 SetError (EXCRegSet, Read, -1); 83 bool done = false; 84 85 while (!done) 86 { 87 int flavor = data.GetU32 (&offset); 88 if (flavor == 0) 89 done = true; 90 else 91 { 92 uint32_t i; 93 uint32_t count = data.GetU32 (&offset); 94 switch (flavor) 95 { 96 case GPRRegSet: 97 for (i=0; i<count; ++i) 98 (&gpr.rax)[i] = data.GetU64(&offset); 99 SetError (GPRRegSet, Read, 0); 100 done = true; 101 102 break; 103 case FPURegSet: 104 // TODO: fill in FPU regs.... 105 //SetError (FPURegSet, Read, -1); 106 done = true; 107 108 break; 109 case EXCRegSet: 110 exc.trapno = data.GetU32(&offset); 111 exc.err = data.GetU32(&offset); 112 exc.faultvaddr = data.GetU64(&offset); 113 SetError (EXCRegSet, Read, 0); 114 done = true; 115 break; 116 case 7: 117 case 8: 118 case 9: 119 // fancy flavors that encapsulate of the above 120 // flavors... 121 break; 122 123 default: 124 done = true; 125 break; 126 } 127 } 128 } 129 } 130 131 132 static size_t 133 WriteRegister (RegisterContext *reg_ctx, const char *name, const char *alt_name, size_t reg_byte_size, Stream &data) 134 { 135 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 136 if (reg_info == NULL) 137 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 138 if (reg_info) 139 { 140 lldb_private::RegisterValue reg_value; 141 if (reg_ctx->ReadRegister(reg_info, reg_value)) 142 { 143 if (reg_info->byte_size >= reg_byte_size) 144 data.Write(reg_value.GetBytes(), reg_byte_size); 145 else 146 { 147 data.Write(reg_value.GetBytes(), reg_info->byte_size); 148 for (size_t i=0, n = reg_byte_size - reg_info->byte_size; i<n; ++ i) 149 data.PutChar(0); 150 } 151 return reg_byte_size; 152 } 153 } 154 // Just write zeros if all else fails 155 for (size_t i=0; i<reg_byte_size; ++ i) 156 data.PutChar(0); 157 return reg_byte_size; 158 } 159 160 static bool 161 Create_LC_THREAD (Thread *thread, Stream &data) 162 { 163 RegisterContextSP reg_ctx_sp (thread->GetRegisterContext()); 164 if (reg_ctx_sp) 165 { 166 RegisterContext *reg_ctx = reg_ctx_sp.get(); 167 168 data.PutHex32 (GPRRegSet); // Flavor 169 data.PutHex32 (GPRWordCount); 170 WriteRegister (reg_ctx, "rax", NULL, 8, data); 171 WriteRegister (reg_ctx, "rbx", NULL, 8, data); 172 WriteRegister (reg_ctx, "rcx", NULL, 8, data); 173 WriteRegister (reg_ctx, "rdx", NULL, 8, data); 174 WriteRegister (reg_ctx, "rdi", NULL, 8, data); 175 WriteRegister (reg_ctx, "rsi", NULL, 8, data); 176 WriteRegister (reg_ctx, "rbp", NULL, 8, data); 177 WriteRegister (reg_ctx, "rsp", NULL, 8, data); 178 WriteRegister (reg_ctx, "r8", NULL, 8, data); 179 WriteRegister (reg_ctx, "r9", NULL, 8, data); 180 WriteRegister (reg_ctx, "r10", NULL, 8, data); 181 WriteRegister (reg_ctx, "r11", NULL, 8, data); 182 WriteRegister (reg_ctx, "r12", NULL, 8, data); 183 WriteRegister (reg_ctx, "r13", NULL, 8, data); 184 WriteRegister (reg_ctx, "r14", NULL, 8, data); 185 WriteRegister (reg_ctx, "r15", NULL, 8, data); 186 WriteRegister (reg_ctx, "rip", NULL, 8, data); 187 WriteRegister (reg_ctx, "rflags", NULL, 8, data); 188 WriteRegister (reg_ctx, "cs", NULL, 8, data); 189 WriteRegister (reg_ctx, "fs", NULL, 8, data); 190 WriteRegister (reg_ctx, "gs", NULL, 8, data); 191 192 // // Write out the FPU registers 193 // const size_t fpu_byte_size = sizeof(FPU); 194 // size_t bytes_written = 0; 195 // data.PutHex32 (FPURegSet); 196 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 197 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 198 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 199 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, data); // uint16_t fcw; // "fctrl" 200 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, data); // uint16_t fsw; // "fstat" 201 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, data); // uint8_t ftw; // "ftag" 202 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 203 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, data); // uint16_t fop; // "fop" 204 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, data); // uint32_t ip; // "fioff" 205 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, data); // uint16_t cs; // "fiseg" 206 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 207 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, data); // uint32_t dp; // "fooff" 208 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, data); // uint16_t ds; // "foseg" 209 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 210 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, data); // uint32_t mxcsr; 211 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 4, data);// uint32_t mxcsrmask; 212 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, sizeof(MMSReg), data); 213 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, sizeof(MMSReg), data); 214 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, sizeof(MMSReg), data); 215 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, sizeof(MMSReg), data); 216 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, sizeof(MMSReg), data); 217 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, sizeof(MMSReg), data); 218 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, sizeof(MMSReg), data); 219 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, sizeof(MMSReg), data); 220 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, sizeof(XMMReg), data); 221 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, sizeof(XMMReg), data); 222 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, sizeof(XMMReg), data); 223 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, sizeof(XMMReg), data); 224 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, sizeof(XMMReg), data); 225 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, sizeof(XMMReg), data); 226 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, sizeof(XMMReg), data); 227 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, sizeof(XMMReg), data); 228 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, sizeof(XMMReg), data); 229 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, sizeof(XMMReg), data); 230 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, sizeof(XMMReg), data); 231 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, sizeof(XMMReg), data); 232 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, sizeof(XMMReg), data); 233 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, sizeof(XMMReg), data); 234 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, sizeof(XMMReg), data); 235 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, sizeof(XMMReg), data); 236 // 237 // // Fill rest with zeros 238 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ i) 239 // data.PutChar(0); 240 241 // Write out the EXC registers 242 data.PutHex32 (EXCRegSet); 243 data.PutHex32 (EXCWordCount); 244 WriteRegister (reg_ctx, "trapno", NULL, 4, data); 245 WriteRegister (reg_ctx, "err", NULL, 4, data); 246 WriteRegister (reg_ctx, "faultvaddr", NULL, 8, data); 247 return true; 248 } 249 return false; 250 } 251 252 protected: 253 virtual int 254 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 255 { 256 return 0; 257 } 258 259 virtual int 260 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 261 { 262 return 0; 263 } 264 265 virtual int 266 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 267 { 268 return 0; 269 } 270 271 virtual int 272 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 273 { 274 return 0; 275 } 276 277 virtual int 278 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 279 { 280 return 0; 281 } 282 283 virtual int 284 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 285 { 286 return 0; 287 } 288 }; 289 290 291 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 292 { 293 public: 294 RegisterContextDarwin_i386_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 295 RegisterContextDarwin_i386 (thread, 0) 296 { 297 SetRegisterDataFrom_LC_THREAD (data); 298 } 299 300 virtual void 301 InvalidateAllRegisters () 302 { 303 // Do nothing... registers are always valid... 304 } 305 306 void 307 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 308 { 309 lldb::offset_t offset = 0; 310 SetError (GPRRegSet, Read, -1); 311 SetError (FPURegSet, Read, -1); 312 SetError (EXCRegSet, Read, -1); 313 bool done = false; 314 315 while (!done) 316 { 317 int flavor = data.GetU32 (&offset); 318 if (flavor == 0) 319 done = true; 320 else 321 { 322 uint32_t i; 323 uint32_t count = data.GetU32 (&offset); 324 switch (flavor) 325 { 326 case GPRRegSet: 327 for (i=0; i<count; ++i) 328 (&gpr.eax)[i] = data.GetU32(&offset); 329 SetError (GPRRegSet, Read, 0); 330 done = true; 331 332 break; 333 case FPURegSet: 334 // TODO: fill in FPU regs.... 335 //SetError (FPURegSet, Read, -1); 336 done = true; 337 338 break; 339 case EXCRegSet: 340 exc.trapno = data.GetU32(&offset); 341 exc.err = data.GetU32(&offset); 342 exc.faultvaddr = data.GetU32(&offset); 343 SetError (EXCRegSet, Read, 0); 344 done = true; 345 break; 346 case 7: 347 case 8: 348 case 9: 349 // fancy flavors that encapsulate of the above 350 // flavors... 351 break; 352 353 default: 354 done = true; 355 break; 356 } 357 } 358 } 359 } 360 361 static size_t 362 WriteRegister (RegisterContext *reg_ctx, const char *name, const char *alt_name, size_t reg_byte_size, Stream &data) 363 { 364 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 365 if (reg_info == NULL) 366 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 367 if (reg_info) 368 { 369 lldb_private::RegisterValue reg_value; 370 if (reg_ctx->ReadRegister(reg_info, reg_value)) 371 { 372 if (reg_info->byte_size >= reg_byte_size) 373 data.Write(reg_value.GetBytes(), reg_byte_size); 374 else 375 { 376 data.Write(reg_value.GetBytes(), reg_info->byte_size); 377 for (size_t i=0, n = reg_byte_size - reg_info->byte_size; i<n; ++ i) 378 data.PutChar(0); 379 } 380 return reg_byte_size; 381 } 382 } 383 // Just write zeros if all else fails 384 for (size_t i=0; i<reg_byte_size; ++ i) 385 data.PutChar(0); 386 return reg_byte_size; 387 } 388 389 static bool 390 Create_LC_THREAD (Thread *thread, Stream &data) 391 { 392 RegisterContextSP reg_ctx_sp (thread->GetRegisterContext()); 393 if (reg_ctx_sp) 394 { 395 RegisterContext *reg_ctx = reg_ctx_sp.get(); 396 397 data.PutHex32 (GPRRegSet); // Flavor 398 data.PutHex32 (GPRWordCount); 399 WriteRegister (reg_ctx, "eax", NULL, 4, data); 400 WriteRegister (reg_ctx, "ebx", NULL, 4, data); 401 WriteRegister (reg_ctx, "ecx", NULL, 4, data); 402 WriteRegister (reg_ctx, "edx", NULL, 4, data); 403 WriteRegister (reg_ctx, "edi", NULL, 4, data); 404 WriteRegister (reg_ctx, "esi", NULL, 4, data); 405 WriteRegister (reg_ctx, "ebp", NULL, 4, data); 406 WriteRegister (reg_ctx, "esp", NULL, 4, data); 407 WriteRegister (reg_ctx, "ss", NULL, 4, data); 408 WriteRegister (reg_ctx, "eflags", NULL, 4, data); 409 WriteRegister (reg_ctx, "eip", NULL, 4, data); 410 WriteRegister (reg_ctx, "cs", NULL, 4, data); 411 WriteRegister (reg_ctx, "ds", NULL, 4, data); 412 WriteRegister (reg_ctx, "es", NULL, 4, data); 413 WriteRegister (reg_ctx, "fs", NULL, 4, data); 414 WriteRegister (reg_ctx, "gs", NULL, 4, data); 415 416 // Write out the EXC registers 417 data.PutHex32 (EXCRegSet); 418 data.PutHex32 (EXCWordCount); 419 WriteRegister (reg_ctx, "trapno", NULL, 4, data); 420 WriteRegister (reg_ctx, "err", NULL, 4, data); 421 WriteRegister (reg_ctx, "faultvaddr", NULL, 4, data); 422 return true; 423 } 424 return false; 425 } 426 427 protected: 428 virtual int 429 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 430 { 431 return 0; 432 } 433 434 virtual int 435 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 436 { 437 return 0; 438 } 439 440 virtual int 441 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 442 { 443 return 0; 444 } 445 446 virtual int 447 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 448 { 449 return 0; 450 } 451 452 virtual int 453 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 454 { 455 return 0; 456 } 457 458 virtual int 459 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 460 { 461 return 0; 462 } 463 }; 464 465 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm 466 { 467 public: 468 RegisterContextDarwin_arm_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 469 RegisterContextDarwin_arm (thread, 0) 470 { 471 SetRegisterDataFrom_LC_THREAD (data); 472 } 473 474 virtual void 475 InvalidateAllRegisters () 476 { 477 // Do nothing... registers are always valid... 478 } 479 480 void 481 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 482 { 483 lldb::offset_t offset = 0; 484 SetError (GPRRegSet, Read, -1); 485 SetError (FPURegSet, Read, -1); 486 SetError (EXCRegSet, Read, -1); 487 bool done = false; 488 489 while (!done) 490 { 491 int flavor = data.GetU32 (&offset); 492 uint32_t count = data.GetU32 (&offset); 493 lldb::offset_t next_thread_state = offset + (count * 4); 494 switch (flavor) 495 { 496 case GPRRegSet: 497 for (uint32_t i=0; i<count; ++i) 498 { 499 gpr.r[i] = data.GetU32(&offset); 500 } 501 502 // Note that gpr.cpsr is also copied by the above loop; this loop technically extends 503 // one element past the end of the gpr.r[] array. 504 505 SetError (GPRRegSet, Read, 0); 506 offset = next_thread_state; 507 break; 508 509 case FPURegSet: 510 { 511 uint8_t *fpu_reg_buf = (uint8_t*) &fpu.floats.s[0]; 512 const int fpu_reg_buf_size = sizeof (fpu.floats); 513 if (data.ExtractBytes (offset, fpu_reg_buf_size, eByteOrderLittle, fpu_reg_buf) == fpu_reg_buf_size) 514 { 515 offset += fpu_reg_buf_size; 516 fpu.fpscr = data.GetU32(&offset); 517 SetError (FPURegSet, Read, 0); 518 } 519 else 520 { 521 done = true; 522 } 523 } 524 offset = next_thread_state; 525 break; 526 527 case EXCRegSet: 528 if (count == 3) 529 { 530 exc.exception = data.GetU32(&offset); 531 exc.fsr = data.GetU32(&offset); 532 exc.far = data.GetU32(&offset); 533 SetError (EXCRegSet, Read, 0); 534 } 535 done = true; 536 offset = next_thread_state; 537 break; 538 539 // Unknown register set flavor, stop trying to parse. 540 default: 541 done = true; 542 } 543 } 544 } 545 546 static size_t 547 WriteRegister (RegisterContext *reg_ctx, const char *name, const char *alt_name, size_t reg_byte_size, Stream &data) 548 { 549 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 550 if (reg_info == NULL) 551 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 552 if (reg_info) 553 { 554 lldb_private::RegisterValue reg_value; 555 if (reg_ctx->ReadRegister(reg_info, reg_value)) 556 { 557 if (reg_info->byte_size >= reg_byte_size) 558 data.Write(reg_value.GetBytes(), reg_byte_size); 559 else 560 { 561 data.Write(reg_value.GetBytes(), reg_info->byte_size); 562 for (size_t i=0, n = reg_byte_size - reg_info->byte_size; i<n; ++ i) 563 data.PutChar(0); 564 } 565 return reg_byte_size; 566 } 567 } 568 // Just write zeros if all else fails 569 for (size_t i=0; i<reg_byte_size; ++ i) 570 data.PutChar(0); 571 return reg_byte_size; 572 } 573 574 static bool 575 Create_LC_THREAD (Thread *thread, Stream &data) 576 { 577 RegisterContextSP reg_ctx_sp (thread->GetRegisterContext()); 578 if (reg_ctx_sp) 579 { 580 RegisterContext *reg_ctx = reg_ctx_sp.get(); 581 582 data.PutHex32 (GPRRegSet); // Flavor 583 data.PutHex32 (GPRWordCount); 584 WriteRegister (reg_ctx, "r0", NULL, 4, data); 585 WriteRegister (reg_ctx, "r1", NULL, 4, data); 586 WriteRegister (reg_ctx, "r2", NULL, 4, data); 587 WriteRegister (reg_ctx, "r3", NULL, 4, data); 588 WriteRegister (reg_ctx, "r4", NULL, 4, data); 589 WriteRegister (reg_ctx, "r5", NULL, 4, data); 590 WriteRegister (reg_ctx, "r6", NULL, 4, data); 591 WriteRegister (reg_ctx, "r7", NULL, 4, data); 592 WriteRegister (reg_ctx, "r8", NULL, 4, data); 593 WriteRegister (reg_ctx, "r9", NULL, 4, data); 594 WriteRegister (reg_ctx, "r10", NULL, 4, data); 595 WriteRegister (reg_ctx, "r11", NULL, 4, data); 596 WriteRegister (reg_ctx, "r12", NULL, 4, data); 597 WriteRegister (reg_ctx, "sp", NULL, 4, data); 598 WriteRegister (reg_ctx, "lr", NULL, 4, data); 599 WriteRegister (reg_ctx, "pc", NULL, 4, data); 600 WriteRegister (reg_ctx, "cpsr", NULL, 4, data); 601 602 // Write out the EXC registers 603 // data.PutHex32 (EXCRegSet); 604 // data.PutHex32 (EXCWordCount); 605 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 606 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 607 // WriteRegister (reg_ctx, "far", NULL, 4, data); 608 return true; 609 } 610 return false; 611 } 612 613 protected: 614 virtual int 615 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 616 { 617 return -1; 618 } 619 620 virtual int 621 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 622 { 623 return -1; 624 } 625 626 virtual int 627 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 628 { 629 return -1; 630 } 631 632 virtual int 633 DoReadDBG (lldb::tid_t tid, int flavor, DBG &dbg) 634 { 635 return -1; 636 } 637 638 virtual int 639 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 640 { 641 return 0; 642 } 643 644 virtual int 645 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 646 { 647 return 0; 648 } 649 650 virtual int 651 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 652 { 653 return 0; 654 } 655 656 virtual int 657 DoWriteDBG (lldb::tid_t tid, int flavor, const DBG &dbg) 658 { 659 return -1; 660 } 661 }; 662 663 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 664 { 665 public: 666 RegisterContextDarwin_arm64_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 667 RegisterContextDarwin_arm64 (thread, 0) 668 { 669 SetRegisterDataFrom_LC_THREAD (data); 670 } 671 672 virtual void 673 InvalidateAllRegisters () 674 { 675 // Do nothing... registers are always valid... 676 } 677 678 void 679 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 680 { 681 lldb::offset_t offset = 0; 682 SetError (GPRRegSet, Read, -1); 683 SetError (FPURegSet, Read, -1); 684 SetError (EXCRegSet, Read, -1); 685 bool done = false; 686 while (!done) 687 { 688 int flavor = data.GetU32 (&offset); 689 uint32_t count = data.GetU32 (&offset); 690 lldb::offset_t next_thread_state = offset + (count * 4); 691 switch (flavor) 692 { 693 case GPRRegSet: 694 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 32-bit register) 695 if (count >= (33 * 2) + 1) 696 { 697 for (uint32_t i=0; i<33; ++i) 698 gpr.x[i] = data.GetU64(&offset); 699 gpr.cpsr = data.GetU32(&offset); 700 SetError (GPRRegSet, Read, 0); 701 } 702 offset = next_thread_state; 703 break; 704 case FPURegSet: 705 { 706 uint8_t *fpu_reg_buf = (uint8_t*) &fpu.v[0]; 707 const int fpu_reg_buf_size = sizeof (fpu); 708 if (fpu_reg_buf_size == count 709 && data.ExtractBytes (offset, fpu_reg_buf_size, eByteOrderLittle, fpu_reg_buf) == fpu_reg_buf_size) 710 { 711 SetError (FPURegSet, Read, 0); 712 } 713 else 714 { 715 done = true; 716 } 717 } 718 offset = next_thread_state; 719 break; 720 case EXCRegSet: 721 if (count == 4) 722 { 723 exc.far = data.GetU64(&offset); 724 exc.esr = data.GetU32(&offset); 725 exc.exception = data.GetU32(&offset); 726 SetError (EXCRegSet, Read, 0); 727 } 728 offset = next_thread_state; 729 break; 730 default: 731 done = true; 732 break; 733 } 734 } 735 } 736 737 static size_t 738 WriteRegister (RegisterContext *reg_ctx, const char *name, const char *alt_name, size_t reg_byte_size, Stream &data) 739 { 740 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 741 if (reg_info == NULL) 742 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 743 if (reg_info) 744 { 745 lldb_private::RegisterValue reg_value; 746 if (reg_ctx->ReadRegister(reg_info, reg_value)) 747 { 748 if (reg_info->byte_size >= reg_byte_size) 749 data.Write(reg_value.GetBytes(), reg_byte_size); 750 else 751 { 752 data.Write(reg_value.GetBytes(), reg_info->byte_size); 753 for (size_t i=0, n = reg_byte_size - reg_info->byte_size; i<n; ++ i) 754 data.PutChar(0); 755 } 756 return reg_byte_size; 757 } 758 } 759 // Just write zeros if all else fails 760 for (size_t i=0; i<reg_byte_size; ++ i) 761 data.PutChar(0); 762 return reg_byte_size; 763 } 764 765 static bool 766 Create_LC_THREAD (Thread *thread, Stream &data) 767 { 768 RegisterContextSP reg_ctx_sp (thread->GetRegisterContext()); 769 if (reg_ctx_sp) 770 { 771 RegisterContext *reg_ctx = reg_ctx_sp.get(); 772 773 data.PutHex32 (GPRRegSet); // Flavor 774 data.PutHex32 (GPRWordCount); 775 WriteRegister (reg_ctx, "x0", NULL, 8, data); 776 WriteRegister (reg_ctx, "x1", NULL, 8, data); 777 WriteRegister (reg_ctx, "x2", NULL, 8, data); 778 WriteRegister (reg_ctx, "x3", NULL, 8, data); 779 WriteRegister (reg_ctx, "x4", NULL, 8, data); 780 WriteRegister (reg_ctx, "x5", NULL, 8, data); 781 WriteRegister (reg_ctx, "x6", NULL, 8, data); 782 WriteRegister (reg_ctx, "x7", NULL, 8, data); 783 WriteRegister (reg_ctx, "x8", NULL, 8, data); 784 WriteRegister (reg_ctx, "x9", NULL, 8, data); 785 WriteRegister (reg_ctx, "x10", NULL, 8, data); 786 WriteRegister (reg_ctx, "x11", NULL, 8, data); 787 WriteRegister (reg_ctx, "x12", NULL, 8, data); 788 WriteRegister (reg_ctx, "x13", NULL, 8, data); 789 WriteRegister (reg_ctx, "x14", NULL, 8, data); 790 WriteRegister (reg_ctx, "x15", NULL, 8, data); 791 WriteRegister (reg_ctx, "x16", NULL, 8, data); 792 WriteRegister (reg_ctx, "x17", NULL, 8, data); 793 WriteRegister (reg_ctx, "x18", NULL, 8, data); 794 WriteRegister (reg_ctx, "x19", NULL, 8, data); 795 WriteRegister (reg_ctx, "x20", NULL, 8, data); 796 WriteRegister (reg_ctx, "x21", NULL, 8, data); 797 WriteRegister (reg_ctx, "x22", NULL, 8, data); 798 WriteRegister (reg_ctx, "x23", NULL, 8, data); 799 WriteRegister (reg_ctx, "x24", NULL, 8, data); 800 WriteRegister (reg_ctx, "x25", NULL, 8, data); 801 WriteRegister (reg_ctx, "x26", NULL, 8, data); 802 WriteRegister (reg_ctx, "x27", NULL, 8, data); 803 WriteRegister (reg_ctx, "x28", NULL, 8, data); 804 WriteRegister (reg_ctx, "fp", NULL, 8, data); 805 WriteRegister (reg_ctx, "lr", NULL, 8, data); 806 WriteRegister (reg_ctx, "sp", NULL, 8, data); 807 WriteRegister (reg_ctx, "pc", NULL, 8, data); 808 WriteRegister (reg_ctx, "cpsr", NULL, 4, data); 809 810 // Write out the EXC registers 811 // data.PutHex32 (EXCRegSet); 812 // data.PutHex32 (EXCWordCount); 813 // WriteRegister (reg_ctx, "far", NULL, 8, data); 814 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 815 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 816 return true; 817 } 818 return false; 819 } 820 821 protected: 822 virtual int 823 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 824 { 825 return -1; 826 } 827 828 virtual int 829 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 830 { 831 return -1; 832 } 833 834 virtual int 835 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 836 { 837 return -1; 838 } 839 840 virtual int 841 DoReadDBG (lldb::tid_t tid, int flavor, DBG &dbg) 842 { 843 return -1; 844 } 845 846 virtual int 847 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 848 { 849 return 0; 850 } 851 852 virtual int 853 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 854 { 855 return 0; 856 } 857 858 virtual int 859 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 860 { 861 return 0; 862 } 863 864 virtual int 865 DoWriteDBG (lldb::tid_t tid, int flavor, const DBG &dbg) 866 { 867 return -1; 868 } 869 }; 870 871 static uint32_t 872 MachHeaderSizeFromMagic(uint32_t magic) 873 { 874 switch (magic) 875 { 876 case MH_MAGIC: 877 case MH_CIGAM: 878 return sizeof(struct mach_header); 879 880 case MH_MAGIC_64: 881 case MH_CIGAM_64: 882 return sizeof(struct mach_header_64); 883 break; 884 885 default: 886 break; 887 } 888 return 0; 889 } 890 891 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 892 893 void 894 ObjectFileMachO::Initialize() 895 { 896 PluginManager::RegisterPlugin (GetPluginNameStatic(), 897 GetPluginDescriptionStatic(), 898 CreateInstance, 899 CreateMemoryInstance, 900 GetModuleSpecifications, 901 SaveCore); 902 } 903 904 void 905 ObjectFileMachO::Terminate() 906 { 907 PluginManager::UnregisterPlugin (CreateInstance); 908 } 909 910 911 lldb_private::ConstString 912 ObjectFileMachO::GetPluginNameStatic() 913 { 914 static ConstString g_name("mach-o"); 915 return g_name; 916 } 917 918 const char * 919 ObjectFileMachO::GetPluginDescriptionStatic() 920 { 921 return "Mach-o object file reader (32 and 64 bit)"; 922 } 923 924 ObjectFile * 925 ObjectFileMachO::CreateInstance (const lldb::ModuleSP &module_sp, 926 DataBufferSP& data_sp, 927 lldb::offset_t data_offset, 928 const FileSpec* file, 929 lldb::offset_t file_offset, 930 lldb::offset_t length) 931 { 932 if (!data_sp) 933 { 934 data_sp = file->MemoryMapFileContentsIfLocal(file_offset, length); 935 data_offset = 0; 936 } 937 938 if (ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 939 { 940 // Update the data to contain the entire file if it doesn't already 941 if (data_sp->GetByteSize() < length) 942 { 943 data_sp = file->MemoryMapFileContentsIfLocal(file_offset, length); 944 data_offset = 0; 945 } 946 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, data_offset, file, file_offset, length)); 947 if (objfile_ap.get() && objfile_ap->ParseHeader()) 948 return objfile_ap.release(); 949 } 950 return NULL; 951 } 952 953 ObjectFile * 954 ObjectFileMachO::CreateMemoryInstance (const lldb::ModuleSP &module_sp, 955 DataBufferSP& data_sp, 956 const ProcessSP &process_sp, 957 lldb::addr_t header_addr) 958 { 959 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) 960 { 961 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, process_sp, header_addr)); 962 if (objfile_ap.get() && objfile_ap->ParseHeader()) 963 return objfile_ap.release(); 964 } 965 return NULL; 966 } 967 968 size_t 969 ObjectFileMachO::GetModuleSpecifications (const lldb_private::FileSpec& file, 970 lldb::DataBufferSP& data_sp, 971 lldb::offset_t data_offset, 972 lldb::offset_t file_offset, 973 lldb::offset_t length, 974 lldb_private::ModuleSpecList &specs) 975 { 976 const size_t initial_count = specs.GetSize(); 977 978 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) 979 { 980 DataExtractor data; 981 data.SetData(data_sp); 982 llvm::MachO::mach_header header; 983 if (ParseHeader (data, &data_offset, header)) 984 { 985 size_t header_and_load_cmds = header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 986 if (header_and_load_cmds >= data_sp->GetByteSize()) 987 { 988 data_sp = file.ReadFileContents(file_offset, header_and_load_cmds); 989 data.SetData(data_sp); 990 data_offset = MachHeaderSizeFromMagic(header.magic); 991 } 992 if (data_sp) 993 { 994 ModuleSpec spec; 995 spec.GetFileSpec() = file; 996 spec.SetObjectOffset(file_offset); 997 spec.SetObjectSize(length); 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 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for 1468 // the 3 fields we care about, so treat them the same. 1469 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) 1470 { 1471 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) 1472 { 1473 if (encryption_cmd.cryptid != 0) 1474 { 1475 EncryptedFileRanges::Entry entry; 1476 entry.SetRangeBase(encryption_cmd.cryptoff); 1477 entry.SetByteSize(encryption_cmd.cryptsize); 1478 encrypted_file_ranges.Append(entry); 1479 } 1480 } 1481 } 1482 offset = load_cmd_offset + encryption_cmd.cmdsize; 1483 } 1484 1485 bool section_file_addresses_changed = false; 1486 1487 offset = MachHeaderSizeFromMagic(m_header.magic); 1488 1489 struct segment_command_64 load_cmd; 1490 for (i=0; i<m_header.ncmds; ++i) 1491 { 1492 const lldb::offset_t load_cmd_offset = offset; 1493 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1494 break; 1495 1496 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1497 { 1498 if (m_data.GetU8(&offset, (uint8_t*)load_cmd.segname, 16)) 1499 { 1500 bool add_section = true; 1501 bool add_to_unified = true; 1502 ConstString const_segname (load_cmd.segname, std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 1503 1504 SectionSP unified_section_sp(unified_section_list.FindSectionByName(const_segname)); 1505 if (is_dsym && unified_section_sp) 1506 { 1507 if (const_segname == GetSegmentNameLINKEDIT()) 1508 { 1509 // We need to keep the __LINKEDIT segment private to this object file only 1510 add_to_unified = false; 1511 } 1512 else 1513 { 1514 // This is the dSYM file and this section has already been created by 1515 // the object file, no need to create it. 1516 add_section = false; 1517 } 1518 } 1519 load_cmd.vmaddr = m_data.GetAddress(&offset); 1520 load_cmd.vmsize = m_data.GetAddress(&offset); 1521 load_cmd.fileoff = m_data.GetAddress(&offset); 1522 load_cmd.filesize = m_data.GetAddress(&offset); 1523 if (m_length != 0 && load_cmd.filesize != 0) 1524 { 1525 if (load_cmd.fileoff > m_length) 1526 { 1527 // We have a load command that says it extends past the end of the file. This is likely 1528 // a corrupt file. We don't have any way to return an error condition here (this method 1529 // was likely invoked from something like ObjectFile::GetSectionList()) -- all we can do 1530 // is null out the SectionList vector and if a process has been set up, dump a message 1531 // to stdout. The most common case here is core file debugging with a truncated file. 1532 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1533 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", 1534 i, 1535 lc_segment_name, 1536 load_cmd.fileoff, 1537 m_length); 1538 1539 load_cmd.fileoff = 0; 1540 load_cmd.filesize = 0; 1541 } 1542 1543 if (load_cmd.fileoff + load_cmd.filesize > m_length) 1544 { 1545 // We have a load command that says it extends past the end of the file. This is likely 1546 // a corrupt file. We don't have any way to return an error condition here (this method 1547 // was likely invoked from something like ObjectFile::GetSectionList()) -- all we can do 1548 // is null out the SectionList vector and if a process has been set up, dump a message 1549 // to stdout. The most common case here is core file debugging with a truncated file. 1550 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1551 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", 1552 i, 1553 lc_segment_name, 1554 load_cmd.fileoff + load_cmd.filesize, 1555 m_length); 1556 1557 // Tuncase the length 1558 load_cmd.filesize = m_length - load_cmd.fileoff; 1559 } 1560 } 1561 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1562 { 1563 1564 const bool segment_is_encrypted = (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1565 1566 // Keep a list of mach segments around in case we need to 1567 // get at data that isn't stored in the abstracted Sections. 1568 m_mach_segments.push_back (load_cmd); 1569 1570 // Use a segment ID of the segment index shifted left by 8 so they 1571 // never conflict with any of the sections. 1572 SectionSP segment_sp; 1573 if (add_section && (const_segname || is_core)) 1574 { 1575 segment_sp.reset(new Section (module_sp, // Module to which this section belongs 1576 this, // Object file to which this sections belongs 1577 ++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 1578 const_segname, // Name of this section 1579 eSectionTypeContainer, // This section is a container of other sections. 1580 load_cmd.vmaddr, // File VM address == addresses as they are found in the object file 1581 load_cmd.vmsize, // VM size in bytes of this section 1582 load_cmd.fileoff, // Offset to the data for this section in the file 1583 load_cmd.filesize, // Size in bytes of this section as found in the file 1584 0, // Segments have no alignment information 1585 load_cmd.flags)); // Flags for this section 1586 1587 segment_sp->SetIsEncrypted (segment_is_encrypted); 1588 m_sections_ap->AddSection(segment_sp); 1589 if (add_to_unified) 1590 unified_section_list.AddSection(segment_sp); 1591 } 1592 else if (unified_section_sp) 1593 { 1594 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) 1595 { 1596 // Check to see if the module was read from memory? 1597 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) 1598 { 1599 // We have a module that is in memory and needs to have its 1600 // file address adjusted. We need to do this because when we 1601 // load a file from memory, its addresses will be slid already, 1602 // yet the addresses in the new symbol file will still be unslid. 1603 // Since everything is stored as section offset, this shouldn't 1604 // cause any problems. 1605 1606 // Make sure we've parsed the symbol table from the 1607 // ObjectFile before we go around changing its Sections. 1608 module_sp->GetObjectFile()->GetSymtab(); 1609 // eh_frame would present the same problems but we parse that on 1610 // a per-function basis as-needed so it's more difficult to 1611 // remove its use of the Sections. Realistically, the environments 1612 // where this code path will be taken will not have eh_frame sections. 1613 1614 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1615 1616 // Notify the module that the section addresses have been changed once 1617 // we're done so any file-address caches can be updated. 1618 section_file_addresses_changed = true; 1619 } 1620 } 1621 m_sections_ap->AddSection(unified_section_sp); 1622 } 1623 1624 struct section_64 sect64; 1625 ::memset (§64, 0, sizeof(sect64)); 1626 // Push a section into our mach sections for the section at 1627 // index zero (NO_SECT) if we don't have any mach sections yet... 1628 if (m_mach_sections.empty()) 1629 m_mach_sections.push_back(sect64); 1630 uint32_t segment_sect_idx; 1631 const lldb::user_id_t first_segment_sectID = sectID + 1; 1632 1633 1634 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1635 for (segment_sect_idx=0; segment_sect_idx<load_cmd.nsects; ++segment_sect_idx) 1636 { 1637 if (m_data.GetU8(&offset, (uint8_t*)sect64.sectname, sizeof(sect64.sectname)) == NULL) 1638 break; 1639 if (m_data.GetU8(&offset, (uint8_t*)sect64.segname, sizeof(sect64.segname)) == NULL) 1640 break; 1641 sect64.addr = m_data.GetAddress(&offset); 1642 sect64.size = m_data.GetAddress(&offset); 1643 1644 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1645 break; 1646 1647 // Keep a list of mach sections around in case we need to 1648 // get at data that isn't stored in the abstracted Sections. 1649 m_mach_sections.push_back (sect64); 1650 1651 if (add_section) 1652 { 1653 ConstString section_name (sect64.sectname, std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1654 if (!const_segname) 1655 { 1656 // We have a segment with no name so we need to conjure up 1657 // segments that correspond to the section's segname if there 1658 // isn't already such a section. If there is such a section, 1659 // we resize the section so that it spans all sections. 1660 // We also mark these sections as fake so address matches don't 1661 // hit if they land in the gaps between the child sections. 1662 const_segname.SetTrimmedCStringWithLength(sect64.segname, sizeof(sect64.segname)); 1663 segment_sp = unified_section_list.FindSectionByName (const_segname); 1664 if (segment_sp.get()) 1665 { 1666 Section *segment = segment_sp.get(); 1667 // Grow the section size as needed. 1668 const lldb::addr_t sect64_min_addr = sect64.addr; 1669 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1670 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1671 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1672 const lldb::addr_t curr_seg_max_addr = curr_seg_min_addr + curr_seg_byte_size; 1673 if (sect64_min_addr >= curr_seg_min_addr) 1674 { 1675 const lldb::addr_t new_seg_byte_size = sect64_max_addr - curr_seg_min_addr; 1676 // Only grow the section size if needed 1677 if (new_seg_byte_size > curr_seg_byte_size) 1678 segment->SetByteSize (new_seg_byte_size); 1679 } 1680 else 1681 { 1682 // We need to change the base address of the segment and 1683 // adjust the child section offsets for all existing children. 1684 const lldb::addr_t slide_amount = sect64_min_addr - curr_seg_min_addr; 1685 segment->Slide(slide_amount, false); 1686 segment->GetChildren().Slide(-slide_amount, false); 1687 segment->SetByteSize (curr_seg_max_addr - sect64_min_addr); 1688 } 1689 1690 // Grow the section size as needed. 1691 if (sect64.offset) 1692 { 1693 const lldb::addr_t segment_min_file_offset = segment->GetFileOffset(); 1694 const lldb::addr_t segment_max_file_offset = segment_min_file_offset + segment->GetFileSize(); 1695 1696 const lldb::addr_t section_min_file_offset = sect64.offset; 1697 const lldb::addr_t section_max_file_offset = section_min_file_offset + sect64.size; 1698 const lldb::addr_t new_file_offset = std::min (section_min_file_offset, segment_min_file_offset); 1699 const lldb::addr_t new_file_size = std::max (section_max_file_offset, segment_max_file_offset) - new_file_offset; 1700 segment->SetFileOffset (new_file_offset); 1701 segment->SetFileSize (new_file_size); 1702 } 1703 } 1704 else 1705 { 1706 // Create a fake section for the section's named segment 1707 segment_sp.reset(new Section (segment_sp, // Parent section 1708 module_sp, // Module to which this section belongs 1709 this, // Object file to which this section belongs 1710 ++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 1711 const_segname, // Name of this section 1712 eSectionTypeContainer, // This section is a container of other sections. 1713 sect64.addr, // File VM address == addresses as they are found in the object file 1714 sect64.size, // VM size in bytes of this section 1715 sect64.offset, // Offset to the data for this section in the file 1716 sect64.offset ? sect64.size : 0, // Size in bytes of this section as found in the file 1717 sect64.align, 1718 load_cmd.flags)); // Flags for this section 1719 segment_sp->SetIsFake(true); 1720 1721 m_sections_ap->AddSection(segment_sp); 1722 if (add_to_unified) 1723 unified_section_list.AddSection(segment_sp); 1724 segment_sp->SetIsEncrypted (segment_is_encrypted); 1725 } 1726 } 1727 assert (segment_sp.get()); 1728 1729 lldb::SectionType sect_type = eSectionTypeOther; 1730 1731 if (sect64.flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1732 sect_type = eSectionTypeCode; 1733 else 1734 { 1735 uint32_t mach_sect_type = sect64.flags & SECTION_TYPE; 1736 static ConstString g_sect_name_objc_data ("__objc_data"); 1737 static ConstString g_sect_name_objc_msgrefs ("__objc_msgrefs"); 1738 static ConstString g_sect_name_objc_selrefs ("__objc_selrefs"); 1739 static ConstString g_sect_name_objc_classrefs ("__objc_classrefs"); 1740 static ConstString g_sect_name_objc_superrefs ("__objc_superrefs"); 1741 static ConstString g_sect_name_objc_const ("__objc_const"); 1742 static ConstString g_sect_name_objc_classlist ("__objc_classlist"); 1743 static ConstString g_sect_name_cfstring ("__cfstring"); 1744 1745 static ConstString g_sect_name_dwarf_debug_abbrev ("__debug_abbrev"); 1746 static ConstString g_sect_name_dwarf_debug_aranges ("__debug_aranges"); 1747 static ConstString g_sect_name_dwarf_debug_frame ("__debug_frame"); 1748 static ConstString g_sect_name_dwarf_debug_info ("__debug_info"); 1749 static ConstString g_sect_name_dwarf_debug_line ("__debug_line"); 1750 static ConstString g_sect_name_dwarf_debug_loc ("__debug_loc"); 1751 static ConstString g_sect_name_dwarf_debug_macinfo ("__debug_macinfo"); 1752 static ConstString g_sect_name_dwarf_debug_pubnames ("__debug_pubnames"); 1753 static ConstString g_sect_name_dwarf_debug_pubtypes ("__debug_pubtypes"); 1754 static ConstString g_sect_name_dwarf_debug_ranges ("__debug_ranges"); 1755 static ConstString g_sect_name_dwarf_debug_str ("__debug_str"); 1756 static ConstString g_sect_name_dwarf_apple_names ("__apple_names"); 1757 static ConstString g_sect_name_dwarf_apple_types ("__apple_types"); 1758 static ConstString g_sect_name_dwarf_apple_namespaces ("__apple_namespac"); 1759 static ConstString g_sect_name_dwarf_apple_objc ("__apple_objc"); 1760 static ConstString g_sect_name_eh_frame ("__eh_frame"); 1761 static ConstString g_sect_name_compact_unwind ("__unwind_info"); 1762 static ConstString g_sect_name_text ("__text"); 1763 static ConstString g_sect_name_data ("__data"); 1764 1765 1766 if (section_name == g_sect_name_dwarf_debug_abbrev) 1767 sect_type = eSectionTypeDWARFDebugAbbrev; 1768 else if (section_name == g_sect_name_dwarf_debug_aranges) 1769 sect_type = eSectionTypeDWARFDebugAranges; 1770 else if (section_name == g_sect_name_dwarf_debug_frame) 1771 sect_type = eSectionTypeDWARFDebugFrame; 1772 else if (section_name == g_sect_name_dwarf_debug_info) 1773 sect_type = eSectionTypeDWARFDebugInfo; 1774 else if (section_name == g_sect_name_dwarf_debug_line) 1775 sect_type = eSectionTypeDWARFDebugLine; 1776 else if (section_name == g_sect_name_dwarf_debug_loc) 1777 sect_type = eSectionTypeDWARFDebugLoc; 1778 else if (section_name == g_sect_name_dwarf_debug_macinfo) 1779 sect_type = eSectionTypeDWARFDebugMacInfo; 1780 else if (section_name == g_sect_name_dwarf_debug_pubnames) 1781 sect_type = eSectionTypeDWARFDebugPubNames; 1782 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 1783 sect_type = eSectionTypeDWARFDebugPubTypes; 1784 else if (section_name == g_sect_name_dwarf_debug_ranges) 1785 sect_type = eSectionTypeDWARFDebugRanges; 1786 else if (section_name == g_sect_name_dwarf_debug_str) 1787 sect_type = eSectionTypeDWARFDebugStr; 1788 else if (section_name == g_sect_name_dwarf_apple_names) 1789 sect_type = eSectionTypeDWARFAppleNames; 1790 else if (section_name == g_sect_name_dwarf_apple_types) 1791 sect_type = eSectionTypeDWARFAppleTypes; 1792 else if (section_name == g_sect_name_dwarf_apple_namespaces) 1793 sect_type = eSectionTypeDWARFAppleNamespaces; 1794 else if (section_name == g_sect_name_dwarf_apple_objc) 1795 sect_type = eSectionTypeDWARFAppleObjC; 1796 else if (section_name == g_sect_name_objc_selrefs) 1797 sect_type = eSectionTypeDataCStringPointers; 1798 else if (section_name == g_sect_name_objc_msgrefs) 1799 sect_type = eSectionTypeDataObjCMessageRefs; 1800 else if (section_name == g_sect_name_eh_frame) 1801 sect_type = eSectionTypeEHFrame; 1802 else if (section_name == g_sect_name_compact_unwind) 1803 sect_type = eSectionTypeCompactUnwind; 1804 else if (section_name == g_sect_name_cfstring) 1805 sect_type = eSectionTypeDataObjCCFStrings; 1806 else if (section_name == g_sect_name_objc_data || 1807 section_name == g_sect_name_objc_classrefs || 1808 section_name == g_sect_name_objc_superrefs || 1809 section_name == g_sect_name_objc_const || 1810 section_name == g_sect_name_objc_classlist) 1811 { 1812 sect_type = eSectionTypeDataPointers; 1813 } 1814 1815 if (sect_type == eSectionTypeOther) 1816 { 1817 switch (mach_sect_type) 1818 { 1819 // TODO: categorize sections by other flags for regular sections 1820 case S_REGULAR: 1821 if (section_name == g_sect_name_text) 1822 sect_type = eSectionTypeCode; 1823 else if (section_name == g_sect_name_data) 1824 sect_type = eSectionTypeData; 1825 else 1826 sect_type = eSectionTypeOther; 1827 break; 1828 case S_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1829 case S_CSTRING_LITERALS: sect_type = eSectionTypeDataCString; break; // section with only literal C strings 1830 case S_4BYTE_LITERALS: sect_type = eSectionTypeData4; break; // section with only 4 byte literals 1831 case S_8BYTE_LITERALS: sect_type = eSectionTypeData8; break; // section with only 8 byte literals 1832 case S_LITERAL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only pointers to literals 1833 case S_NON_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only non-lazy symbol pointers 1834 case S_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only lazy symbol pointers 1835 case S_SYMBOL_STUBS: sect_type = eSectionTypeCode; break; // section with only symbol stubs, byte size of stub in the reserved2 field 1836 case S_MOD_INIT_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for initialization 1837 case S_MOD_TERM_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for termination 1838 case S_COALESCED: sect_type = eSectionTypeOther; break; 1839 case S_GB_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1840 case S_INTERPOSING: sect_type = eSectionTypeCode; break; // section with only pairs of function pointers for interposing 1841 case S_16BYTE_LITERALS: sect_type = eSectionTypeData16; break; // section with only 16 byte literals 1842 case S_DTRACE_DOF: sect_type = eSectionTypeDebug; break; 1843 case S_LAZY_DYLIB_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; 1844 default: break; 1845 } 1846 } 1847 } 1848 1849 SectionSP section_sp(new Section (segment_sp, 1850 module_sp, 1851 this, 1852 ++sectID, 1853 section_name, 1854 sect_type, 1855 sect64.addr - segment_sp->GetFileAddress(), 1856 sect64.size, 1857 sect64.offset, 1858 sect64.offset == 0 ? 0 : sect64.size, 1859 sect64.align, 1860 sect64.flags)); 1861 // Set the section to be encrypted to match the segment 1862 1863 bool section_is_encrypted = false; 1864 if (!segment_is_encrypted && load_cmd.filesize != 0) 1865 section_is_encrypted = encrypted_file_ranges.FindEntryThatContains(sect64.offset) != NULL; 1866 1867 section_sp->SetIsEncrypted (segment_is_encrypted || section_is_encrypted); 1868 segment_sp->GetChildren().AddSection(section_sp); 1869 1870 if (segment_sp->IsFake()) 1871 { 1872 segment_sp.reset(); 1873 const_segname.Clear(); 1874 } 1875 } 1876 } 1877 if (segment_sp && is_dsym) 1878 { 1879 if (first_segment_sectID <= sectID) 1880 { 1881 lldb::user_id_t sect_uid; 1882 for (sect_uid = first_segment_sectID; sect_uid <= sectID; ++sect_uid) 1883 { 1884 SectionSP curr_section_sp(segment_sp->GetChildren().FindSectionByID (sect_uid)); 1885 SectionSP next_section_sp; 1886 if (sect_uid + 1 <= sectID) 1887 next_section_sp = segment_sp->GetChildren().FindSectionByID (sect_uid+1); 1888 1889 if (curr_section_sp.get()) 1890 { 1891 if (curr_section_sp->GetByteSize() == 0) 1892 { 1893 if (next_section_sp.get() != NULL) 1894 curr_section_sp->SetByteSize ( next_section_sp->GetFileAddress() - curr_section_sp->GetFileAddress() ); 1895 else 1896 curr_section_sp->SetByteSize ( load_cmd.vmsize ); 1897 } 1898 } 1899 } 1900 } 1901 } 1902 } 1903 } 1904 } 1905 else if (load_cmd.cmd == LC_DYSYMTAB) 1906 { 1907 m_dysymtab.cmd = load_cmd.cmd; 1908 m_dysymtab.cmdsize = load_cmd.cmdsize; 1909 m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1910 } 1911 1912 offset = load_cmd_offset + load_cmd.cmdsize; 1913 } 1914 1915 1916 if (section_file_addresses_changed && module_sp.get()) 1917 { 1918 module_sp->SectionFileAddressesChanged(); 1919 } 1920 } 1921 } 1922 1923 class MachSymtabSectionInfo 1924 { 1925 public: 1926 1927 MachSymtabSectionInfo (SectionList *section_list) : 1928 m_section_list (section_list), 1929 m_section_infos() 1930 { 1931 // Get the number of sections down to a depth of 1 to include 1932 // all segments and their sections, but no other sections that 1933 // may be added for debug map or 1934 m_section_infos.resize(section_list->GetNumSections(1)); 1935 } 1936 1937 1938 SectionSP 1939 GetSection (uint8_t n_sect, addr_t file_addr) 1940 { 1941 if (n_sect == 0) 1942 return SectionSP(); 1943 if (n_sect < m_section_infos.size()) 1944 { 1945 if (!m_section_infos[n_sect].section_sp) 1946 { 1947 SectionSP section_sp (m_section_list->FindSectionByID (n_sect)); 1948 m_section_infos[n_sect].section_sp = section_sp; 1949 if (section_sp) 1950 { 1951 m_section_infos[n_sect].vm_range.SetBaseAddress (section_sp->GetFileAddress()); 1952 m_section_infos[n_sect].vm_range.SetByteSize (section_sp->GetByteSize()); 1953 } 1954 else 1955 { 1956 Host::SystemLog (Host::eSystemLogError, "error: unable to find section for section %u\n", n_sect); 1957 } 1958 } 1959 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) 1960 { 1961 // Symbol is in section. 1962 return m_section_infos[n_sect].section_sp; 1963 } 1964 else if (m_section_infos[n_sect].vm_range.GetByteSize () == 0 && 1965 m_section_infos[n_sect].vm_range.GetBaseAddress() == file_addr) 1966 { 1967 // Symbol is in section with zero size, but has the same start 1968 // address as the section. This can happen with linker symbols 1969 // (symbols that start with the letter 'l' or 'L'. 1970 return m_section_infos[n_sect].section_sp; 1971 } 1972 } 1973 return m_section_list->FindSectionContainingFileAddress(file_addr); 1974 } 1975 1976 protected: 1977 struct SectionInfo 1978 { 1979 SectionInfo () : 1980 vm_range(), 1981 section_sp () 1982 { 1983 } 1984 1985 VMRange vm_range; 1986 SectionSP section_sp; 1987 }; 1988 SectionList *m_section_list; 1989 std::vector<SectionInfo> m_section_infos; 1990 }; 1991 1992 struct TrieEntry 1993 { 1994 TrieEntry () : 1995 name(), 1996 address(LLDB_INVALID_ADDRESS), 1997 flags (0), 1998 other(0), 1999 import_name() 2000 { 2001 } 2002 2003 void 2004 Clear () 2005 { 2006 name.Clear(); 2007 address = LLDB_INVALID_ADDRESS; 2008 flags = 0; 2009 other = 0; 2010 import_name.Clear(); 2011 } 2012 2013 void 2014 Dump () const 2015 { 2016 printf ("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 2017 static_cast<unsigned long long>(address), 2018 static_cast<unsigned long long>(flags), 2019 static_cast<unsigned long long>(other), name.GetCString()); 2020 if (import_name) 2021 printf (" -> \"%s\"\n", import_name.GetCString()); 2022 else 2023 printf ("\n"); 2024 } 2025 ConstString name; 2026 uint64_t address; 2027 uint64_t flags; 2028 uint64_t other; 2029 ConstString import_name; 2030 }; 2031 2032 struct TrieEntryWithOffset 2033 { 2034 lldb::offset_t nodeOffset; 2035 TrieEntry entry; 2036 2037 TrieEntryWithOffset (lldb::offset_t offset) : 2038 nodeOffset (offset), 2039 entry() 2040 { 2041 } 2042 2043 void 2044 Dump (uint32_t idx) const 2045 { 2046 printf ("[%3u] 0x%16.16llx: ", idx, 2047 static_cast<unsigned long long>(nodeOffset)); 2048 entry.Dump(); 2049 } 2050 2051 bool 2052 operator<(const TrieEntryWithOffset& other) const 2053 { 2054 return ( nodeOffset < other.nodeOffset ); 2055 } 2056 }; 2057 2058 static void 2059 ParseTrieEntries (DataExtractor &data, 2060 lldb::offset_t offset, 2061 std::vector<llvm::StringRef> &nameSlices, 2062 std::set<lldb::addr_t> &resolver_addresses, 2063 std::vector<TrieEntryWithOffset>& output) 2064 { 2065 if (!data.ValidOffset(offset)) 2066 return; 2067 2068 const uint64_t terminalSize = data.GetULEB128(&offset); 2069 lldb::offset_t children_offset = offset + terminalSize; 2070 if ( terminalSize != 0 ) { 2071 TrieEntryWithOffset e (offset); 2072 e.entry.flags = data.GetULEB128(&offset); 2073 const char *import_name = NULL; 2074 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT ) { 2075 e.entry.address = 0; 2076 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2077 import_name = data.GetCStr(&offset); 2078 } 2079 else { 2080 e.entry.address = data.GetULEB128(&offset); 2081 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER ) 2082 { 2083 //resolver_addresses.insert(e.entry.address); 2084 e.entry.other = data.GetULEB128(&offset); 2085 resolver_addresses.insert(e.entry.other); 2086 } 2087 else 2088 e.entry.other = 0; 2089 } 2090 // Only add symbols that are reexport symbols with a valid import name 2091 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && import_name[0]) 2092 { 2093 std::string name; 2094 if (!nameSlices.empty()) 2095 { 2096 for (auto name_slice: nameSlices) 2097 name.append(name_slice.data(), name_slice.size()); 2098 } 2099 if (name.size() > 1) 2100 { 2101 // Skip the leading '_' 2102 e.entry.name.SetCStringWithLength(name.c_str() + 1,name.size() - 1); 2103 } 2104 if (import_name) 2105 { 2106 // Skip the leading '_' 2107 e.entry.import_name.SetCString(import_name+1); 2108 } 2109 output.push_back(e); 2110 } 2111 } 2112 2113 const uint8_t childrenCount = data.GetU8(&children_offset); 2114 for (uint8_t i=0; i < childrenCount; ++i) { 2115 nameSlices.push_back(data.GetCStr(&children_offset)); 2116 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2117 if (childNodeOffset) 2118 { 2119 ParseTrieEntries(data, 2120 childNodeOffset, 2121 nameSlices, 2122 resolver_addresses, 2123 output); 2124 } 2125 nameSlices.pop_back(); 2126 } 2127 } 2128 2129 size_t 2130 ObjectFileMachO::ParseSymtab () 2131 { 2132 Timer scoped_timer(__PRETTY_FUNCTION__, 2133 "ObjectFileMachO::ParseSymtab () module = %s", 2134 m_file.GetFilename().AsCString("")); 2135 ModuleSP module_sp (GetModule()); 2136 if (!module_sp) 2137 return 0; 2138 2139 struct symtab_command symtab_load_command = { 0, 0, 0, 0, 0, 0 }; 2140 struct linkedit_data_command function_starts_load_command = { 0, 0, 0, 0 }; 2141 struct dyld_info_command dyld_info = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; 2142 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2143 FunctionStarts function_starts; 2144 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2145 uint32_t i; 2146 FileSpecList dylib_files; 2147 Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_SYMBOLS)); 2148 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 2149 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 2150 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 2151 2152 for (i=0; i<m_header.ncmds; ++i) 2153 { 2154 const lldb::offset_t cmd_offset = offset; 2155 // Read in the load command and load command size 2156 struct load_command lc; 2157 if (m_data.GetU32(&offset, &lc, 2) == NULL) 2158 break; 2159 // Watch for the symbol table load command 2160 switch (lc.cmd) 2161 { 2162 case LC_SYMTAB: 2163 symtab_load_command.cmd = lc.cmd; 2164 symtab_load_command.cmdsize = lc.cmdsize; 2165 // Read in the rest of the symtab load command 2166 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 0) // fill in symoff, nsyms, stroff, strsize fields 2167 return 0; 2168 if (symtab_load_command.symoff == 0) 2169 { 2170 if (log) 2171 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2172 return 0; 2173 } 2174 2175 if (symtab_load_command.stroff == 0) 2176 { 2177 if (log) 2178 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2179 return 0; 2180 } 2181 2182 if (symtab_load_command.nsyms == 0) 2183 { 2184 if (log) 2185 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2186 return 0; 2187 } 2188 2189 if (symtab_load_command.strsize == 0) 2190 { 2191 if (log) 2192 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2193 return 0; 2194 } 2195 break; 2196 2197 case LC_DYLD_INFO: 2198 case LC_DYLD_INFO_ONLY: 2199 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) 2200 { 2201 dyld_info.cmd = lc.cmd; 2202 dyld_info.cmdsize = lc.cmdsize; 2203 } 2204 else 2205 { 2206 memset (&dyld_info, 0, sizeof(dyld_info)); 2207 } 2208 break; 2209 2210 case LC_LOAD_DYLIB: 2211 case LC_LOAD_WEAK_DYLIB: 2212 case LC_REEXPORT_DYLIB: 2213 case LC_LOADFVMLIB: 2214 case LC_LOAD_UPWARD_DYLIB: 2215 { 2216 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2217 const char *path = m_data.PeekCStr(name_offset); 2218 if (path) 2219 { 2220 FileSpec file_spec(path, false); 2221 // Strip the path if there is @rpath, @executable, etc so we just use the basename 2222 if (path[0] == '@') 2223 file_spec.GetDirectory().Clear(); 2224 2225 if (lc.cmd == LC_REEXPORT_DYLIB) 2226 { 2227 m_reexported_dylibs.AppendIfUnique(file_spec); 2228 } 2229 2230 dylib_files.Append(file_spec); 2231 } 2232 } 2233 break; 2234 2235 case LC_FUNCTION_STARTS: 2236 function_starts_load_command.cmd = lc.cmd; 2237 function_starts_load_command.cmdsize = lc.cmdsize; 2238 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == NULL) // fill in symoff, nsyms, stroff, strsize fields 2239 memset (&function_starts_load_command, 0, sizeof(function_starts_load_command)); 2240 break; 2241 2242 default: 2243 break; 2244 } 2245 offset = cmd_offset + lc.cmdsize; 2246 } 2247 2248 if (symtab_load_command.cmd) 2249 { 2250 Symtab *symtab = m_symtab_ap.get(); 2251 SectionList *section_list = GetSectionList(); 2252 if (section_list == NULL) 2253 return 0; 2254 2255 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2256 const ByteOrder byte_order = m_data.GetByteOrder(); 2257 bool bit_width_32 = addr_byte_size == 4; 2258 const size_t nlist_byte_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2259 2260 DataExtractor nlist_data (NULL, 0, byte_order, addr_byte_size); 2261 DataExtractor strtab_data (NULL, 0, byte_order, addr_byte_size); 2262 DataExtractor function_starts_data (NULL, 0, byte_order, addr_byte_size); 2263 DataExtractor indirect_symbol_index_data (NULL, 0, byte_order, addr_byte_size); 2264 DataExtractor dyld_trie_data (NULL, 0, byte_order, addr_byte_size); 2265 2266 const addr_t nlist_data_byte_size = symtab_load_command.nsyms * nlist_byte_size; 2267 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2268 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2269 2270 ProcessSP process_sp (m_process_wp.lock()); 2271 Process *process = process_sp.get(); 2272 2273 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2274 2275 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) 2276 { 2277 Target &target = process->GetTarget(); 2278 2279 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2280 2281 SectionSP linkedit_section_sp(section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2282 // Reading mach file from memory in a process or core file... 2283 2284 if (linkedit_section_sp) 2285 { 2286 const addr_t linkedit_load_addr = linkedit_section_sp->GetLoadBaseAddress(&target); 2287 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2288 const addr_t symoff_addr = linkedit_load_addr + symtab_load_command.symoff - linkedit_file_offset; 2289 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - linkedit_file_offset; 2290 2291 bool data_was_read = false; 2292 2293 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 2294 if (m_header.flags & 0x80000000u && process->GetAddressByteSize() == sizeof (void*)) 2295 { 2296 // This mach-o memory file is in the dyld shared cache. If this 2297 // program is not remote and this is iOS, then this process will 2298 // share the same shared cache as the process we are debugging and 2299 // we can read the entire __LINKEDIT from the address space in this 2300 // process. This is a needed optimization that is used for local iOS 2301 // debugging only since all shared libraries in the shared cache do 2302 // not have corresponding files that exist in the file system of the 2303 // device. They have been combined into a single file. This means we 2304 // always have to load these files from memory. All of the symbol and 2305 // string tables from all of the __LINKEDIT sections from the shared 2306 // libraries in the shared cache have been merged into a single large 2307 // symbol and string table. Reading all of this symbol and string table 2308 // data across can slow down debug launch times, so we optimize this by 2309 // reading the memory for the __LINKEDIT section from this process. 2310 2311 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 2312 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 2313 bool use_lldb_cache = true; 2314 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && lldb_shared_cache != process_shared_cache) 2315 { 2316 use_lldb_cache = false; 2317 ModuleSP module_sp (GetModule()); 2318 if (module_sp) 2319 module_sp->ReportWarning ("shared cache in process does not match lldb's own shared cache, startup will be slow."); 2320 2321 } 2322 2323 PlatformSP platform_sp (target.GetPlatform()); 2324 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) 2325 { 2326 data_was_read = true; 2327 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, eByteOrderLittle); 2328 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, eByteOrderLittle); 2329 if (function_starts_load_command.cmd) 2330 { 2331 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 2332 function_starts_data.SetData ((void *)func_start_addr, function_starts_load_command.datasize, eByteOrderLittle); 2333 } 2334 } 2335 } 2336 #endif 2337 2338 if (!data_was_read) 2339 { 2340 if (memory_module_load_level == eMemoryModuleLoadLevelComplete) 2341 { 2342 DataBufferSP nlist_data_sp (ReadMemory (process_sp, symoff_addr, nlist_data_byte_size)); 2343 if (nlist_data_sp) 2344 nlist_data.SetData (nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2345 // Load strings individually from memory when loading from memory since shared cache 2346 // string tables contain strings for all symbols from all shared cached libraries 2347 //DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, strtab_data_byte_size)); 2348 //if (strtab_data_sp) 2349 // strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 2350 if (m_dysymtab.nindirectsyms != 0) 2351 { 2352 const addr_t indirect_syms_addr = linkedit_load_addr + m_dysymtab.indirectsymoff - linkedit_file_offset; 2353 DataBufferSP indirect_syms_data_sp (ReadMemory (process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 2354 if (indirect_syms_data_sp) 2355 indirect_symbol_index_data.SetData (indirect_syms_data_sp, 0, indirect_syms_data_sp->GetByteSize()); 2356 } 2357 } 2358 2359 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) 2360 { 2361 if (function_starts_load_command.cmd) 2362 { 2363 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 2364 DataBufferSP func_start_data_sp (ReadMemory (process_sp, func_start_addr, function_starts_load_command.datasize)); 2365 if (func_start_data_sp) 2366 function_starts_data.SetData (func_start_data_sp, 0, func_start_data_sp->GetByteSize()); 2367 } 2368 } 2369 } 2370 } 2371 } 2372 else 2373 { 2374 nlist_data.SetData (m_data, 2375 symtab_load_command.symoff, 2376 nlist_data_byte_size); 2377 strtab_data.SetData (m_data, 2378 symtab_load_command.stroff, 2379 strtab_data_byte_size); 2380 2381 if (dyld_info.export_size > 0) 2382 { 2383 dyld_trie_data.SetData (m_data, 2384 dyld_info.export_off, 2385 dyld_info.export_size); 2386 } 2387 2388 if (m_dysymtab.nindirectsyms != 0) 2389 { 2390 indirect_symbol_index_data.SetData (m_data, 2391 m_dysymtab.indirectsymoff, 2392 m_dysymtab.nindirectsyms * 4); 2393 } 2394 if (function_starts_load_command.cmd) 2395 { 2396 function_starts_data.SetData (m_data, 2397 function_starts_load_command.dataoff, 2398 function_starts_load_command.datasize); 2399 } 2400 } 2401 2402 if (nlist_data.GetByteSize() == 0 && memory_module_load_level == eMemoryModuleLoadLevelComplete) 2403 { 2404 if (log) 2405 module_sp->LogMessage(log, "failed to read nlist data"); 2406 return 0; 2407 } 2408 2409 2410 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2411 if (!have_strtab_data) 2412 { 2413 if (process) 2414 { 2415 if (strtab_addr == LLDB_INVALID_ADDRESS) 2416 { 2417 if (log) 2418 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2419 return 0; 2420 } 2421 } 2422 else 2423 { 2424 if (log) 2425 module_sp->LogMessage(log, "failed to read strtab data"); 2426 return 0; 2427 } 2428 } 2429 2430 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2431 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2432 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2433 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2434 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 2435 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 2436 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 2437 SectionSP eh_frame_section_sp; 2438 if (text_section_sp.get()) 2439 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 2440 else 2441 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 2442 2443 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2444 2445 // lldb works best if it knows the start address of all functions in a module. 2446 // Linker symbols or debug info are normally the best source of information for start addr / size but 2447 // they may be stripped in a released binary. 2448 // Two additional sources of information exist in Mach-O binaries: 2449 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each function's start address in the 2450 // binary, relative to the text section. 2451 // eh_frame - the eh_frame FDEs have the start addr & size of each function 2452 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on all modern binaries. 2453 // Binaries built to run on older releases may need to use eh_frame information. 2454 2455 if (text_section_sp && function_starts_data.GetByteSize()) 2456 { 2457 FunctionStarts::Entry function_start_entry; 2458 function_start_entry.data = false; 2459 lldb::offset_t function_start_offset = 0; 2460 function_start_entry.addr = text_section_sp->GetFileAddress(); 2461 uint64_t delta; 2462 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 0) 2463 { 2464 // Now append the current entry 2465 function_start_entry.addr += delta; 2466 function_starts.Append(function_start_entry); 2467 } 2468 } 2469 else 2470 { 2471 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the load command claiming an eh_frame 2472 // but it doesn't actually have the eh_frame content. And if we have a dSYM, we don't need to do any 2473 // of this fill-in-the-missing-symbols works anyway - the debug info should give us all the functions in 2474 // the module. 2475 if (text_section_sp.get() && eh_frame_section_sp.get() && m_type != eTypeDebugInfo) 2476 { 2477 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, eRegisterKindGCC, true); 2478 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2479 eh_frame.GetFunctionAddressAndSizeVector (functions); 2480 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2481 size_t count = functions.GetSize(); 2482 for (size_t i = 0; i < count; ++i) 2483 { 2484 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = functions.GetEntryAtIndex (i); 2485 if (func) 2486 { 2487 FunctionStarts::Entry function_start_entry; 2488 function_start_entry.addr = func->base - text_base_addr; 2489 function_starts.Append(function_start_entry); 2490 } 2491 } 2492 } 2493 } 2494 2495 const size_t function_starts_count = function_starts.GetSize(); 2496 2497 const user_id_t TEXT_eh_frame_sectID = 2498 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2499 : static_cast<user_id_t>(NO_SECT); 2500 2501 lldb::offset_t nlist_data_offset = 0; 2502 2503 uint32_t N_SO_index = UINT32_MAX; 2504 2505 MachSymtabSectionInfo section_info (section_list); 2506 std::vector<uint32_t> N_FUN_indexes; 2507 std::vector<uint32_t> N_NSYM_indexes; 2508 std::vector<uint32_t> N_INCL_indexes; 2509 std::vector<uint32_t> N_BRAC_indexes; 2510 std::vector<uint32_t> N_COMM_indexes; 2511 typedef std::multimap <uint64_t, uint32_t> ValueToSymbolIndexMap; 2512 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2513 typedef std::map <const char *, uint32_t> ConstNameToSymbolIndexMap; 2514 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2515 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2516 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2517 // Any symbols that get merged into another will get an entry 2518 // in this map so we know 2519 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2520 uint32_t nlist_idx = 0; 2521 Symbol *symbol_ptr = NULL; 2522 2523 uint32_t sym_idx = 0; 2524 Symbol *sym = NULL; 2525 size_t num_syms = 0; 2526 std::string memory_symbol_name; 2527 uint32_t unmapped_local_symbols_found = 0; 2528 2529 std::vector<TrieEntryWithOffset> trie_entries; 2530 std::set<lldb::addr_t> resolver_addresses; 2531 2532 if (dyld_trie_data.GetByteSize() > 0) 2533 { 2534 std::vector<llvm::StringRef> nameSlices; 2535 ParseTrieEntries (dyld_trie_data, 2536 0, 2537 nameSlices, 2538 resolver_addresses, 2539 trie_entries); 2540 2541 ConstString text_segment_name ("__TEXT"); 2542 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 2543 if (text_segment_sp) 2544 { 2545 const lldb::addr_t text_segment_file_addr = text_segment_sp->GetFileAddress(); 2546 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) 2547 { 2548 for (auto &e : trie_entries) 2549 e.entry.address += text_segment_file_addr; 2550 } 2551 } 2552 } 2553 2554 typedef std::set<ConstString> IndirectSymbols; 2555 IndirectSymbols indirect_symbol_names; 2556 2557 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 2558 2559 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been optimized by moving LOCAL 2560 // symbols out of the memory mapped portion of the DSC. The symbol information has all been retained, 2561 // but it isn't available in the normal nlist data. However, there *are* duplicate entries of *some* 2562 // LOCAL symbols in the normal nlist data. To handle this situation correctly, we must first attempt 2563 // to parse any DSC unmapped symbol information. If we find any, we set a flag that tells the normal 2564 // nlist parser to ignore all LOCAL symbols. 2565 2566 if (m_header.flags & 0x80000000u) 2567 { 2568 // Before we can start mapping the DSC, we need to make certain the target process is actually 2569 // using the cache we can find. 2570 2571 // Next we need to determine the correct path for the dyld shared cache. 2572 2573 ArchSpec header_arch; 2574 GetArchitecture(header_arch); 2575 char dsc_path[PATH_MAX]; 2576 2577 snprintf(dsc_path, sizeof(dsc_path), "%s%s%s", 2578 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR */ 2579 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2580 header_arch.GetArchitectureName()); 2581 2582 FileSpec dsc_filespec(dsc_path, false); 2583 2584 // We need definitions of two structures in the on-disk DSC, copy them here manually 2585 struct lldb_copy_dyld_cache_header_v0 2586 { 2587 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2588 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2589 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2590 uint32_t imagesOffset; 2591 uint32_t imagesCount; 2592 uint64_t dyldBaseAddress; 2593 uint64_t codeSignatureOffset; 2594 uint64_t codeSignatureSize; 2595 uint64_t slideInfoOffset; 2596 uint64_t slideInfoSize; 2597 uint64_t localSymbolsOffset; // file offset of where local symbols are stored 2598 uint64_t localSymbolsSize; // size of local symbols information 2599 }; 2600 struct lldb_copy_dyld_cache_header_v1 2601 { 2602 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2603 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2604 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2605 uint32_t imagesOffset; 2606 uint32_t imagesCount; 2607 uint64_t dyldBaseAddress; 2608 uint64_t codeSignatureOffset; 2609 uint64_t codeSignatureSize; 2610 uint64_t slideInfoOffset; 2611 uint64_t slideInfoSize; 2612 uint64_t localSymbolsOffset; 2613 uint64_t localSymbolsSize; 2614 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 and later 2615 }; 2616 2617 struct lldb_copy_dyld_cache_mapping_info 2618 { 2619 uint64_t address; 2620 uint64_t size; 2621 uint64_t fileOffset; 2622 uint32_t maxProt; 2623 uint32_t initProt; 2624 }; 2625 2626 struct lldb_copy_dyld_cache_local_symbols_info 2627 { 2628 uint32_t nlistOffset; 2629 uint32_t nlistCount; 2630 uint32_t stringsOffset; 2631 uint32_t stringsSize; 2632 uint32_t entriesOffset; 2633 uint32_t entriesCount; 2634 }; 2635 struct lldb_copy_dyld_cache_local_symbols_entry 2636 { 2637 uint32_t dylibOffset; 2638 uint32_t nlistStartIndex; 2639 uint32_t nlistCount; 2640 }; 2641 2642 /* The dyld_cache_header has a pointer to the dyld_cache_local_symbols_info structure (localSymbolsOffset). 2643 The dyld_cache_local_symbols_info structure gives us three things: 2644 1. The start and count of the nlist records in the dyld_shared_cache file 2645 2. The start and size of the strings for these nlist records 2646 3. The start and count of dyld_cache_local_symbols_entry entries 2647 2648 There is one dyld_cache_local_symbols_entry per dylib/framework in the dyld shared cache. 2649 The "dylibOffset" field is the Mach-O header of this dylib/framework in the dyld shared cache. 2650 The dyld_cache_local_symbols_entry also lists the start of this dylib/framework's nlist records 2651 and the count of how many nlist records there are for this dylib/framework. 2652 */ 2653 2654 // Process the dsc header to find the unmapped symbols 2655 // 2656 // Save some VM space, do not map the entire cache in one shot. 2657 2658 DataBufferSP dsc_data_sp; 2659 dsc_data_sp = dsc_filespec.MemoryMapFileContentsIfLocal(0, sizeof(struct lldb_copy_dyld_cache_header_v1)); 2660 2661 if (dsc_data_sp) 2662 { 2663 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2664 2665 char version_str[17]; 2666 int version = -1; 2667 lldb::offset_t offset = 0; 2668 memcpy (version_str, dsc_header_data.GetData (&offset, 16), 16); 2669 version_str[16] = '\0'; 2670 if (strncmp (version_str, "dyld_v", 6) == 0 && isdigit (version_str[6])) 2671 { 2672 int v; 2673 if (::sscanf (version_str + 6, "%d", &v) == 1) 2674 { 2675 version = v; 2676 } 2677 } 2678 2679 UUID dsc_uuid; 2680 if (version >= 1) 2681 { 2682 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, uuid); 2683 uint8_t uuid_bytes[sizeof (uuid_t)]; 2684 memcpy (uuid_bytes, dsc_header_data.GetData (&offset, sizeof (uuid_t)), sizeof (uuid_t)); 2685 dsc_uuid.SetBytes (uuid_bytes); 2686 } 2687 2688 bool uuid_match = true; 2689 if (dsc_uuid.IsValid() && process) 2690 { 2691 UUID shared_cache_uuid(GetProcessSharedCacheUUID(process)); 2692 2693 if (shared_cache_uuid.IsValid() && dsc_uuid != shared_cache_uuid) 2694 { 2695 // The on-disk dyld_shared_cache file is not the same as the one in this 2696 // process' memory, don't use it. 2697 uuid_match = false; 2698 ModuleSP module_sp (GetModule()); 2699 if (module_sp) 2700 module_sp->ReportWarning ("process shared cache does not match on-disk dyld_shared_cache file, some symbol names will be missing."); 2701 } 2702 } 2703 2704 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2705 2706 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2707 2708 // If the mappingOffset points to a location inside the header, we've 2709 // opened an old dyld shared cache, and should not proceed further. 2710 if (uuid_match && mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v0)) 2711 { 2712 2713 DataBufferSP dsc_mapping_info_data_sp = dsc_filespec.MemoryMapFileContentsIfLocal(mappingOffset, sizeof (struct lldb_copy_dyld_cache_mapping_info)); 2714 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, byte_order, addr_byte_size); 2715 offset = 0; 2716 2717 // The File addresses (from the in-memory Mach-O load commands) for the shared libraries 2718 // in the shared library cache need to be adjusted by an offset to match up with the 2719 // dylibOffset identifying field in the dyld_cache_local_symbol_entry's. This offset is 2720 // recorded in mapping_offset_value. 2721 const uint64_t mapping_offset_value = dsc_mapping_info_data.GetU64(&offset); 2722 2723 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2724 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2725 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2726 2727 if (localSymbolsOffset && localSymbolsSize) 2728 { 2729 // Map the local symbols 2730 if (DataBufferSP dsc_local_symbols_data_sp = dsc_filespec.MemoryMapFileContentsIfLocal(localSymbolsOffset, localSymbolsSize)) 2731 { 2732 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, byte_order, addr_byte_size); 2733 2734 offset = 0; 2735 2736 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2737 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2738 UndefinedNameToDescMap undefined_name_to_desc; 2739 SymbolIndexToName reexport_shlib_needs_fixup; 2740 2741 2742 // Read the local_symbols_infos struct in one shot 2743 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2744 dsc_local_symbols_data.GetU32(&offset, &local_symbols_info.nlistOffset, 6); 2745 2746 SectionSP text_section_sp(section_list->FindSectionByName(GetSegmentNameTEXT())); 2747 2748 uint32_t header_file_offset = (text_section_sp->GetFileAddress() - mapping_offset_value); 2749 2750 offset = local_symbols_info.entriesOffset; 2751 for (uint32_t entry_index = 0; entry_index < local_symbols_info.entriesCount; entry_index++) 2752 { 2753 struct lldb_copy_dyld_cache_local_symbols_entry local_symbols_entry; 2754 local_symbols_entry.dylibOffset = dsc_local_symbols_data.GetU32(&offset); 2755 local_symbols_entry.nlistStartIndex = dsc_local_symbols_data.GetU32(&offset); 2756 local_symbols_entry.nlistCount = dsc_local_symbols_data.GetU32(&offset); 2757 2758 if (header_file_offset == local_symbols_entry.dylibOffset) 2759 { 2760 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2761 2762 // The normal nlist code cannot correctly size the Symbols array, we need to allocate it here. 2763 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms + unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2764 num_syms = symtab->GetNumSymbols(); 2765 2766 nlist_data_offset = local_symbols_info.nlistOffset + (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2767 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2768 2769 for (uint32_t nlist_index = 0; nlist_index < local_symbols_entry.nlistCount; nlist_index++) 2770 { 2771 ///////////////////////////// 2772 { 2773 struct nlist_64 nlist; 2774 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2775 break; 2776 2777 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked(&nlist_data_offset); 2778 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2779 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2780 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked (&nlist_data_offset); 2781 nlist.n_value = dsc_local_symbols_data.GetAddress_unchecked (&nlist_data_offset); 2782 2783 SymbolType type = eSymbolTypeInvalid; 2784 const char *symbol_name = dsc_local_symbols_data.PeekCStr(string_table_offset + nlist.n_strx); 2785 2786 if (symbol_name == NULL) 2787 { 2788 // No symbol should be NULL, even the symbols with no 2789 // string values should have an offset zero which points 2790 // to an empty C-string 2791 Host::SystemLog (Host::eSystemLogError, 2792 "error: DSC unmapped local symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 2793 entry_index, 2794 nlist.n_strx, 2795 module_sp->GetFileSpec().GetPath().c_str()); 2796 continue; 2797 } 2798 if (symbol_name[0] == '\0') 2799 symbol_name = NULL; 2800 2801 const char *symbol_name_non_abi_mangled = NULL; 2802 2803 SectionSP symbol_section; 2804 uint32_t symbol_byte_size = 0; 2805 bool add_nlist = true; 2806 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2807 bool demangled_is_synthesized = false; 2808 bool is_gsym = false; 2809 bool set_value = true; 2810 2811 assert (sym_idx < num_syms); 2812 2813 sym[sym_idx].SetDebug (is_debug); 2814 2815 if (is_debug) 2816 { 2817 switch (nlist.n_type) 2818 { 2819 case N_GSYM: 2820 // global symbol: name,,NO_SECT,type,0 2821 // Sometimes the N_GSYM value contains the address. 2822 2823 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 2824 // have the same address, but we want to ensure that we always find only the real symbol, 2825 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 2826 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 2827 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 2828 // same address. 2829 2830 is_gsym = true; 2831 sym[sym_idx].SetExternal(true); 2832 2833 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') 2834 { 2835 llvm::StringRef symbol_name_ref(symbol_name); 2836 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 2837 { 2838 symbol_name_non_abi_mangled = symbol_name + 1; 2839 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2840 type = eSymbolTypeObjCClass; 2841 demangled_is_synthesized = true; 2842 2843 } 2844 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 2845 { 2846 symbol_name_non_abi_mangled = symbol_name + 1; 2847 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2848 type = eSymbolTypeObjCMetaClass; 2849 demangled_is_synthesized = true; 2850 } 2851 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 2852 { 2853 symbol_name_non_abi_mangled = symbol_name + 1; 2854 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2855 type = eSymbolTypeObjCIVar; 2856 demangled_is_synthesized = true; 2857 } 2858 } 2859 else 2860 { 2861 if (nlist.n_value != 0) 2862 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2863 type = eSymbolTypeData; 2864 } 2865 break; 2866 2867 case N_FNAME: 2868 // procedure name (f77 kludge): name,,NO_SECT,0,0 2869 type = eSymbolTypeCompiler; 2870 break; 2871 2872 case N_FUN: 2873 // procedure: name,,n_sect,linenumber,address 2874 if (symbol_name) 2875 { 2876 type = eSymbolTypeCode; 2877 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2878 2879 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 2880 // We use the current number of symbols in the symbol table in lieu of 2881 // using nlist_idx in case we ever start trimming entries out 2882 N_FUN_indexes.push_back(sym_idx); 2883 } 2884 else 2885 { 2886 type = eSymbolTypeCompiler; 2887 2888 if ( !N_FUN_indexes.empty() ) 2889 { 2890 // Copy the size of the function into the original STAB entry so we don't have 2891 // to hunt for it later 2892 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 2893 N_FUN_indexes.pop_back(); 2894 // We don't really need the end function STAB as it contains the size which 2895 // we already placed with the original symbol, so don't add it if we want a 2896 // minimal symbol table 2897 add_nlist = false; 2898 } 2899 } 2900 break; 2901 2902 case N_STSYM: 2903 // static symbol: name,,n_sect,type,address 2904 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 2905 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2906 type = eSymbolTypeData; 2907 break; 2908 2909 case N_LCSYM: 2910 // .lcomm symbol: name,,n_sect,type,address 2911 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2912 type = eSymbolTypeCommonBlock; 2913 break; 2914 2915 case N_BNSYM: 2916 // We use the current number of symbols in the symbol table in lieu of 2917 // using nlist_idx in case we ever start trimming entries out 2918 // Skip these if we want minimal symbol tables 2919 add_nlist = false; 2920 break; 2921 2922 case N_ENSYM: 2923 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 2924 // so that we can always skip the entire symbol if we need to navigate 2925 // more quickly at the source level when parsing STABS 2926 // Skip these if we want minimal symbol tables 2927 add_nlist = false; 2928 break; 2929 2930 2931 case N_OPT: 2932 // emitted with gcc2_compiled and in gcc source 2933 type = eSymbolTypeCompiler; 2934 break; 2935 2936 case N_RSYM: 2937 // register sym: name,,NO_SECT,type,register 2938 type = eSymbolTypeVariable; 2939 break; 2940 2941 case N_SLINE: 2942 // src line: 0,,n_sect,linenumber,address 2943 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2944 type = eSymbolTypeLineEntry; 2945 break; 2946 2947 case N_SSYM: 2948 // structure elt: name,,NO_SECT,type,struct_offset 2949 type = eSymbolTypeVariableType; 2950 break; 2951 2952 case N_SO: 2953 // source file name 2954 type = eSymbolTypeSourceFile; 2955 if (symbol_name == NULL) 2956 { 2957 add_nlist = false; 2958 if (N_SO_index != UINT32_MAX) 2959 { 2960 // Set the size of the N_SO to the terminating index of this N_SO 2961 // so that we can always skip the entire N_SO if we need to navigate 2962 // more quickly at the source level when parsing STABS 2963 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2964 symbol_ptr->SetByteSize(sym_idx); 2965 symbol_ptr->SetSizeIsSibling(true); 2966 } 2967 N_NSYM_indexes.clear(); 2968 N_INCL_indexes.clear(); 2969 N_BRAC_indexes.clear(); 2970 N_COMM_indexes.clear(); 2971 N_FUN_indexes.clear(); 2972 N_SO_index = UINT32_MAX; 2973 } 2974 else 2975 { 2976 // We use the current number of symbols in the symbol table in lieu of 2977 // using nlist_idx in case we ever start trimming entries out 2978 const bool N_SO_has_full_path = symbol_name[0] == '/'; 2979 if (N_SO_has_full_path) 2980 { 2981 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2982 { 2983 // We have two consecutive N_SO entries where the first contains a directory 2984 // and the second contains a full path. 2985 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 2986 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2987 add_nlist = false; 2988 } 2989 else 2990 { 2991 // This is the first entry in a N_SO that contains a directory or 2992 // a full path to the source file 2993 N_SO_index = sym_idx; 2994 } 2995 } 2996 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2997 { 2998 // This is usually the second N_SO entry that contains just the filename, 2999 // so here we combine it with the first one if we are minimizing the symbol table 3000 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3001 if (so_path && so_path[0]) 3002 { 3003 std::string full_so_path (so_path); 3004 const size_t double_slash_pos = full_so_path.find("//"); 3005 if (double_slash_pos != std::string::npos) 3006 { 3007 // The linker has been generating bad N_SO entries with doubled up paths 3008 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 3009 // and the second is the directory for the source file so you end up with 3010 // a path that looks like "/tmp/src//tmp/src/" 3011 FileSpec so_dir(so_path, false); 3012 if (!so_dir.Exists()) 3013 { 3014 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 3015 if (so_dir.Exists()) 3016 { 3017 // Trim off the incorrect path 3018 full_so_path.erase(0, double_slash_pos + 1); 3019 } 3020 } 3021 } 3022 if (*full_so_path.rbegin() != '/') 3023 full_so_path += '/'; 3024 full_so_path += symbol_name; 3025 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 3026 add_nlist = false; 3027 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3028 } 3029 } 3030 else 3031 { 3032 // This could be a relative path to a N_SO 3033 N_SO_index = sym_idx; 3034 } 3035 } 3036 break; 3037 3038 case N_OSO: 3039 // object file name: name,,0,0,st_mtime 3040 type = eSymbolTypeObjectFile; 3041 break; 3042 3043 case N_LSYM: 3044 // local sym: name,,NO_SECT,type,offset 3045 type = eSymbolTypeLocal; 3046 break; 3047 3048 //---------------------------------------------------------------------- 3049 // INCL scopes 3050 //---------------------------------------------------------------------- 3051 case N_BINCL: 3052 // include file beginning: name,,NO_SECT,0,sum 3053 // We use the current number of symbols in the symbol table in lieu of 3054 // using nlist_idx in case we ever start trimming entries out 3055 N_INCL_indexes.push_back(sym_idx); 3056 type = eSymbolTypeScopeBegin; 3057 break; 3058 3059 case N_EINCL: 3060 // include file end: name,,NO_SECT,0,0 3061 // Set the size of the N_BINCL to the terminating index of this N_EINCL 3062 // so that we can always skip the entire symbol if we need to navigate 3063 // more quickly at the source level when parsing STABS 3064 if ( !N_INCL_indexes.empty() ) 3065 { 3066 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3067 symbol_ptr->SetByteSize(sym_idx + 1); 3068 symbol_ptr->SetSizeIsSibling(true); 3069 N_INCL_indexes.pop_back(); 3070 } 3071 type = eSymbolTypeScopeEnd; 3072 break; 3073 3074 case N_SOL: 3075 // #included file name: name,,n_sect,0,address 3076 type = eSymbolTypeHeaderFile; 3077 3078 // We currently don't use the header files on darwin 3079 add_nlist = false; 3080 break; 3081 3082 case N_PARAMS: 3083 // compiler parameters: name,,NO_SECT,0,0 3084 type = eSymbolTypeCompiler; 3085 break; 3086 3087 case N_VERSION: 3088 // compiler version: name,,NO_SECT,0,0 3089 type = eSymbolTypeCompiler; 3090 break; 3091 3092 case N_OLEVEL: 3093 // compiler -O level: name,,NO_SECT,0,0 3094 type = eSymbolTypeCompiler; 3095 break; 3096 3097 case N_PSYM: 3098 // parameter: name,,NO_SECT,type,offset 3099 type = eSymbolTypeVariable; 3100 break; 3101 3102 case N_ENTRY: 3103 // alternate entry: name,,n_sect,linenumber,address 3104 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3105 type = eSymbolTypeLineEntry; 3106 break; 3107 3108 //---------------------------------------------------------------------- 3109 // Left and Right Braces 3110 //---------------------------------------------------------------------- 3111 case N_LBRAC: 3112 // left bracket: 0,,NO_SECT,nesting level,address 3113 // We use the current number of symbols in the symbol table in lieu of 3114 // using nlist_idx in case we ever start trimming entries out 3115 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3116 N_BRAC_indexes.push_back(sym_idx); 3117 type = eSymbolTypeScopeBegin; 3118 break; 3119 3120 case N_RBRAC: 3121 // right bracket: 0,,NO_SECT,nesting level,address 3122 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 3123 // so that we can always skip the entire symbol if we need to navigate 3124 // more quickly at the source level when parsing STABS 3125 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3126 if ( !N_BRAC_indexes.empty() ) 3127 { 3128 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3129 symbol_ptr->SetByteSize(sym_idx + 1); 3130 symbol_ptr->SetSizeIsSibling(true); 3131 N_BRAC_indexes.pop_back(); 3132 } 3133 type = eSymbolTypeScopeEnd; 3134 break; 3135 3136 case N_EXCL: 3137 // deleted include file: name,,NO_SECT,0,sum 3138 type = eSymbolTypeHeaderFile; 3139 break; 3140 3141 //---------------------------------------------------------------------- 3142 // COMM scopes 3143 //---------------------------------------------------------------------- 3144 case N_BCOMM: 3145 // begin common: name,,NO_SECT,0,0 3146 // We use the current number of symbols in the symbol table in lieu of 3147 // using nlist_idx in case we ever start trimming entries out 3148 type = eSymbolTypeScopeBegin; 3149 N_COMM_indexes.push_back(sym_idx); 3150 break; 3151 3152 case N_ECOML: 3153 // end common (local name): 0,,n_sect,0,address 3154 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3155 // Fall through 3156 3157 case N_ECOMM: 3158 // end common: name,,n_sect,0,0 3159 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 3160 // so that we can always skip the entire symbol if we need to navigate 3161 // more quickly at the source level when parsing STABS 3162 if ( !N_COMM_indexes.empty() ) 3163 { 3164 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 3165 symbol_ptr->SetByteSize(sym_idx + 1); 3166 symbol_ptr->SetSizeIsSibling(true); 3167 N_COMM_indexes.pop_back(); 3168 } 3169 type = eSymbolTypeScopeEnd; 3170 break; 3171 3172 case N_LENG: 3173 // second stab entry with length information 3174 type = eSymbolTypeAdditional; 3175 break; 3176 3177 default: break; 3178 } 3179 } 3180 else 3181 { 3182 //uint8_t n_pext = N_PEXT & nlist.n_type; 3183 uint8_t n_type = N_TYPE & nlist.n_type; 3184 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3185 3186 switch (n_type) 3187 { 3188 case N_INDR: 3189 { 3190 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 3191 if (reexport_name_cstr && reexport_name_cstr[0]) 3192 { 3193 type = eSymbolTypeReExported; 3194 ConstString reexport_name(reexport_name_cstr + ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3195 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3196 set_value = false; 3197 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3198 indirect_symbol_names.insert(ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3199 } 3200 else 3201 type = eSymbolTypeUndefined; 3202 } 3203 break; 3204 3205 case N_UNDF: 3206 if (symbol_name && symbol_name[0]) 3207 { 3208 ConstString undefined_name(symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3209 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3210 } 3211 // Fall through 3212 case N_PBUD: 3213 type = eSymbolTypeUndefined; 3214 break; 3215 3216 case N_ABS: 3217 type = eSymbolTypeAbsolute; 3218 break; 3219 3220 case N_SECT: 3221 { 3222 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3223 3224 if (symbol_section == NULL) 3225 { 3226 // TODO: warn about this? 3227 add_nlist = false; 3228 break; 3229 } 3230 3231 if (TEXT_eh_frame_sectID == nlist.n_sect) 3232 { 3233 type = eSymbolTypeException; 3234 } 3235 else 3236 { 3237 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 3238 3239 switch (section_type) 3240 { 3241 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 3242 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 3243 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 3244 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 3245 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 3246 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 3247 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 3248 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 3249 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 3250 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 3251 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 3252 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 3253 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 3254 default: 3255 switch (symbol_section->GetType()) 3256 { 3257 case lldb::eSectionTypeCode: 3258 type = eSymbolTypeCode; 3259 break; 3260 case eSectionTypeData: 3261 case eSectionTypeDataCString: // Inlined C string data 3262 case eSectionTypeDataCStringPointers: // Pointers to C string data 3263 case eSectionTypeDataSymbolAddress: // Address of a symbol in the symbol table 3264 case eSectionTypeData4: 3265 case eSectionTypeData8: 3266 case eSectionTypeData16: 3267 type = eSymbolTypeData; 3268 break; 3269 default: 3270 break; 3271 } 3272 break; 3273 } 3274 3275 if (type == eSymbolTypeInvalid) 3276 { 3277 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 3278 if (symbol_section->IsDescendant (text_section_sp.get())) 3279 { 3280 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 3281 S_ATTR_SELF_MODIFYING_CODE | 3282 S_ATTR_SOME_INSTRUCTIONS)) 3283 type = eSymbolTypeData; 3284 else 3285 type = eSymbolTypeCode; 3286 } 3287 else if (symbol_section->IsDescendant(data_section_sp.get())) 3288 { 3289 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 3290 { 3291 type = eSymbolTypeRuntime; 3292 3293 if (symbol_name && 3294 symbol_name[0] == '_' && 3295 symbol_name[1] == 'O' && 3296 symbol_name[2] == 'B') 3297 { 3298 llvm::StringRef symbol_name_ref(symbol_name); 3299 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 3300 { 3301 symbol_name_non_abi_mangled = symbol_name + 1; 3302 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3303 type = eSymbolTypeObjCClass; 3304 demangled_is_synthesized = true; 3305 } 3306 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 3307 { 3308 symbol_name_non_abi_mangled = symbol_name + 1; 3309 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3310 type = eSymbolTypeObjCMetaClass; 3311 demangled_is_synthesized = true; 3312 } 3313 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 3314 { 3315 symbol_name_non_abi_mangled = symbol_name + 1; 3316 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3317 type = eSymbolTypeObjCIVar; 3318 demangled_is_synthesized = true; 3319 } 3320 } 3321 } 3322 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 3323 { 3324 type = eSymbolTypeException; 3325 } 3326 else 3327 { 3328 type = eSymbolTypeData; 3329 } 3330 } 3331 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 3332 { 3333 type = eSymbolTypeTrampoline; 3334 } 3335 else if (symbol_section->IsDescendant(objc_section_sp.get())) 3336 { 3337 type = eSymbolTypeRuntime; 3338 if (symbol_name && symbol_name[0] == '.') 3339 { 3340 llvm::StringRef symbol_name_ref(symbol_name); 3341 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 3342 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 3343 { 3344 symbol_name_non_abi_mangled = symbol_name; 3345 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 3346 type = eSymbolTypeObjCClass; 3347 demangled_is_synthesized = true; 3348 } 3349 } 3350 } 3351 } 3352 } 3353 } 3354 break; 3355 } 3356 } 3357 3358 if (add_nlist) 3359 { 3360 uint64_t symbol_value = nlist.n_value; 3361 if (symbol_name_non_abi_mangled) 3362 { 3363 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 3364 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 3365 } 3366 else 3367 { 3368 bool symbol_name_is_mangled = false; 3369 3370 if (symbol_name && symbol_name[0] == '_') 3371 { 3372 symbol_name_is_mangled = symbol_name[1] == '_'; 3373 symbol_name++; // Skip the leading underscore 3374 } 3375 3376 if (symbol_name) 3377 { 3378 ConstString const_symbol_name(symbol_name); 3379 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 3380 if (is_gsym && is_debug) 3381 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 3382 } 3383 } 3384 if (symbol_section) 3385 { 3386 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3387 if (symbol_byte_size == 0 && function_starts_count > 0) 3388 { 3389 addr_t symbol_lookup_file_addr = nlist.n_value; 3390 // Do an exact address match for non-ARM addresses, else get the closest since 3391 // the symbol might be a thumb symbol which has an address with bit zero set 3392 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 3393 if (is_arm && func_start_entry) 3394 { 3395 // Verify that the function start address is the symbol address (ARM) 3396 // or the symbol address + 1 (thumb) 3397 if (func_start_entry->addr != symbol_lookup_file_addr && 3398 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 3399 { 3400 // Not the right entry, NULL it out... 3401 func_start_entry = NULL; 3402 } 3403 } 3404 if (func_start_entry) 3405 { 3406 func_start_entry->data = true; 3407 3408 addr_t symbol_file_addr = func_start_entry->addr; 3409 uint32_t symbol_flags = 0; 3410 if (is_arm) 3411 { 3412 if (symbol_file_addr & 1) 3413 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3414 symbol_file_addr &= 0xfffffffffffffffeull; 3415 } 3416 3417 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3418 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3419 if (next_func_start_entry) 3420 { 3421 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3422 // Be sure the clear the Thumb address bit when we calculate the size 3423 // from the current and next address 3424 if (is_arm) 3425 next_symbol_file_addr &= 0xfffffffffffffffeull; 3426 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3427 } 3428 else 3429 { 3430 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3431 } 3432 } 3433 } 3434 symbol_value -= section_file_addr; 3435 } 3436 3437 if (is_debug == false) 3438 { 3439 if (type == eSymbolTypeCode) 3440 { 3441 // See if we can find a N_FUN entry for any code symbols. 3442 // If we do find a match, and the name matches, then we 3443 // can merge the two into just the function symbol to avoid 3444 // duplicate entries in the symbol table 3445 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3446 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3447 if (range.first != range.second) 3448 { 3449 bool found_it = false; 3450 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3451 { 3452 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3453 { 3454 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3455 // We just need the flags from the linker symbol, so put these flags 3456 // into the N_FUN flags to avoid duplicate symbols in the symbol table 3457 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3458 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3459 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3460 sym[pos->second].SetType (eSymbolTypeResolver); 3461 sym[sym_idx].Clear(); 3462 found_it = true; 3463 break; 3464 } 3465 } 3466 if (found_it) 3467 continue; 3468 } 3469 else 3470 { 3471 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3472 type = eSymbolTypeResolver; 3473 } 3474 } 3475 else if (type == eSymbolTypeData || 3476 type == eSymbolTypeObjCClass || 3477 type == eSymbolTypeObjCMetaClass || 3478 type == eSymbolTypeObjCIVar ) 3479 { 3480 // See if we can find a N_STSYM entry for any data symbols. 3481 // If we do find a match, and the name matches, then we 3482 // can merge the two into just the Static symbol to avoid 3483 // duplicate entries in the symbol table 3484 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3485 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 3486 if (range.first != range.second) 3487 { 3488 bool found_it = false; 3489 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3490 { 3491 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3492 { 3493 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3494 // We just need the flags from the linker symbol, so put these flags 3495 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3496 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3497 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3498 sym[sym_idx].Clear(); 3499 found_it = true; 3500 break; 3501 } 3502 } 3503 if (found_it) 3504 continue; 3505 } 3506 else 3507 { 3508 // Combine N_GSYM stab entries with the non stab symbol 3509 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 3510 if (pos != N_GSYM_name_to_sym_idx.end()) 3511 { 3512 const uint32_t GSYM_sym_idx = pos->second; 3513 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 3514 // Copy the address, because often the N_GSYM address has an invalid address of zero 3515 // when the global is a common symbol 3516 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 3517 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 3518 // We just need the flags from the linker symbol, so put these flags 3519 // into the N_GSYM flags to avoid duplicate symbols in the symbol table 3520 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3521 sym[sym_idx].Clear(); 3522 continue; 3523 } 3524 } 3525 } 3526 } 3527 3528 sym[sym_idx].SetID (nlist_idx); 3529 sym[sym_idx].SetType (type); 3530 if (set_value) 3531 { 3532 sym[sym_idx].GetAddress().SetSection (symbol_section); 3533 sym[sym_idx].GetAddress().SetOffset (symbol_value); 3534 } 3535 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3536 3537 if (symbol_byte_size > 0) 3538 sym[sym_idx].SetByteSize(symbol_byte_size); 3539 3540 if (demangled_is_synthesized) 3541 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3542 ++sym_idx; 3543 } 3544 else 3545 { 3546 sym[sym_idx].Clear(); 3547 } 3548 3549 } 3550 ///////////////////////////// 3551 } 3552 break; // No more entries to consider 3553 } 3554 } 3555 3556 for (const auto &pos :reexport_shlib_needs_fixup) 3557 { 3558 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3559 if (undef_pos != undefined_name_to_desc.end()) 3560 { 3561 const uint8_t dylib_ordinal = llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3562 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3563 sym[pos.first].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(dylib_ordinal-1)); 3564 } 3565 } 3566 } 3567 } 3568 } 3569 } 3570 } 3571 3572 // Must reset this in case it was mutated above! 3573 nlist_data_offset = 0; 3574 #endif 3575 3576 if (nlist_data.GetByteSize() > 0) 3577 { 3578 3579 // If the sym array was not created while parsing the DSC unmapped 3580 // symbols, create it now. 3581 if (sym == NULL) 3582 { 3583 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3584 num_syms = symtab->GetNumSymbols(); 3585 } 3586 3587 if (unmapped_local_symbols_found) 3588 { 3589 assert(m_dysymtab.ilocalsym == 0); 3590 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3591 nlist_idx = m_dysymtab.nlocalsym; 3592 } 3593 else 3594 { 3595 nlist_idx = 0; 3596 } 3597 3598 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3599 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3600 UndefinedNameToDescMap undefined_name_to_desc; 3601 SymbolIndexToName reexport_shlib_needs_fixup; 3602 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 3603 { 3604 struct nlist_64 nlist; 3605 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 3606 break; 3607 3608 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3609 nlist.n_type = nlist_data.GetU8_unchecked (&nlist_data_offset); 3610 nlist.n_sect = nlist_data.GetU8_unchecked (&nlist_data_offset); 3611 nlist.n_desc = nlist_data.GetU16_unchecked (&nlist_data_offset); 3612 nlist.n_value = nlist_data.GetAddress_unchecked (&nlist_data_offset); 3613 3614 SymbolType type = eSymbolTypeInvalid; 3615 const char *symbol_name = NULL; 3616 3617 if (have_strtab_data) 3618 { 3619 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3620 3621 if (symbol_name == NULL) 3622 { 3623 // No symbol should be NULL, even the symbols with no 3624 // string values should have an offset zero which points 3625 // to an empty C-string 3626 Host::SystemLog (Host::eSystemLogError, 3627 "error: symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 3628 nlist_idx, 3629 nlist.n_strx, 3630 module_sp->GetFileSpec().GetPath().c_str()); 3631 continue; 3632 } 3633 if (symbol_name[0] == '\0') 3634 symbol_name = NULL; 3635 } 3636 else 3637 { 3638 const addr_t str_addr = strtab_addr + nlist.n_strx; 3639 Error str_error; 3640 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, str_error)) 3641 symbol_name = memory_symbol_name.c_str(); 3642 } 3643 const char *symbol_name_non_abi_mangled = NULL; 3644 3645 SectionSP symbol_section; 3646 lldb::addr_t symbol_byte_size = 0; 3647 bool add_nlist = true; 3648 bool is_gsym = false; 3649 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3650 bool demangled_is_synthesized = false; 3651 bool set_value = true; 3652 assert (sym_idx < num_syms); 3653 3654 sym[sym_idx].SetDebug (is_debug); 3655 3656 if (is_debug) 3657 { 3658 switch (nlist.n_type) 3659 { 3660 case N_GSYM: 3661 // global symbol: name,,NO_SECT,type,0 3662 // Sometimes the N_GSYM value contains the address. 3663 3664 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 3665 // have the same address, but we want to ensure that we always find only the real symbol, 3666 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 3667 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 3668 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 3669 // same address. 3670 is_gsym = true; 3671 sym[sym_idx].SetExternal(true); 3672 3673 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') 3674 { 3675 llvm::StringRef symbol_name_ref(symbol_name); 3676 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 3677 { 3678 symbol_name_non_abi_mangled = symbol_name + 1; 3679 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3680 type = eSymbolTypeObjCClass; 3681 demangled_is_synthesized = true; 3682 3683 } 3684 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 3685 { 3686 symbol_name_non_abi_mangled = symbol_name + 1; 3687 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3688 type = eSymbolTypeObjCMetaClass; 3689 demangled_is_synthesized = true; 3690 } 3691 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 3692 { 3693 symbol_name_non_abi_mangled = symbol_name + 1; 3694 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3695 type = eSymbolTypeObjCIVar; 3696 demangled_is_synthesized = true; 3697 } 3698 } 3699 else 3700 { 3701 if (nlist.n_value != 0) 3702 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3703 type = eSymbolTypeData; 3704 } 3705 break; 3706 3707 case N_FNAME: 3708 // procedure name (f77 kludge): name,,NO_SECT,0,0 3709 type = eSymbolTypeCompiler; 3710 break; 3711 3712 case N_FUN: 3713 // procedure: name,,n_sect,linenumber,address 3714 if (symbol_name) 3715 { 3716 type = eSymbolTypeCode; 3717 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3718 3719 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3720 // We use the current number of symbols in the symbol table in lieu of 3721 // using nlist_idx in case we ever start trimming entries out 3722 N_FUN_indexes.push_back(sym_idx); 3723 } 3724 else 3725 { 3726 type = eSymbolTypeCompiler; 3727 3728 if ( !N_FUN_indexes.empty() ) 3729 { 3730 // Copy the size of the function into the original STAB entry so we don't have 3731 // to hunt for it later 3732 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 3733 N_FUN_indexes.pop_back(); 3734 // We don't really need the end function STAB as it contains the size which 3735 // we already placed with the original symbol, so don't add it if we want a 3736 // minimal symbol table 3737 add_nlist = false; 3738 } 3739 } 3740 break; 3741 3742 case N_STSYM: 3743 // static symbol: name,,n_sect,type,address 3744 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3745 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3746 type = eSymbolTypeData; 3747 break; 3748 3749 case N_LCSYM: 3750 // .lcomm symbol: name,,n_sect,type,address 3751 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3752 type = eSymbolTypeCommonBlock; 3753 break; 3754 3755 case N_BNSYM: 3756 // We use the current number of symbols in the symbol table in lieu of 3757 // using nlist_idx in case we ever start trimming entries out 3758 // Skip these if we want minimal symbol tables 3759 add_nlist = false; 3760 break; 3761 3762 case N_ENSYM: 3763 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 3764 // so that we can always skip the entire symbol if we need to navigate 3765 // more quickly at the source level when parsing STABS 3766 // Skip these if we want minimal symbol tables 3767 add_nlist = false; 3768 break; 3769 3770 3771 case N_OPT: 3772 // emitted with gcc2_compiled and in gcc source 3773 type = eSymbolTypeCompiler; 3774 break; 3775 3776 case N_RSYM: 3777 // register sym: name,,NO_SECT,type,register 3778 type = eSymbolTypeVariable; 3779 break; 3780 3781 case N_SLINE: 3782 // src line: 0,,n_sect,linenumber,address 3783 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3784 type = eSymbolTypeLineEntry; 3785 break; 3786 3787 case N_SSYM: 3788 // structure elt: name,,NO_SECT,type,struct_offset 3789 type = eSymbolTypeVariableType; 3790 break; 3791 3792 case N_SO: 3793 // source file name 3794 type = eSymbolTypeSourceFile; 3795 if (symbol_name == NULL) 3796 { 3797 add_nlist = false; 3798 if (N_SO_index != UINT32_MAX) 3799 { 3800 // Set the size of the N_SO to the terminating index of this N_SO 3801 // so that we can always skip the entire N_SO if we need to navigate 3802 // more quickly at the source level when parsing STABS 3803 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3804 symbol_ptr->SetByteSize(sym_idx); 3805 symbol_ptr->SetSizeIsSibling(true); 3806 } 3807 N_NSYM_indexes.clear(); 3808 N_INCL_indexes.clear(); 3809 N_BRAC_indexes.clear(); 3810 N_COMM_indexes.clear(); 3811 N_FUN_indexes.clear(); 3812 N_SO_index = UINT32_MAX; 3813 } 3814 else 3815 { 3816 // We use the current number of symbols in the symbol table in lieu of 3817 // using nlist_idx in case we ever start trimming entries out 3818 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3819 if (N_SO_has_full_path) 3820 { 3821 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3822 { 3823 // We have two consecutive N_SO entries where the first contains a directory 3824 // and the second contains a full path. 3825 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 3826 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3827 add_nlist = false; 3828 } 3829 else 3830 { 3831 // This is the first entry in a N_SO that contains a directory or 3832 // a full path to the source file 3833 N_SO_index = sym_idx; 3834 } 3835 } 3836 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3837 { 3838 // This is usually the second N_SO entry that contains just the filename, 3839 // so here we combine it with the first one if we are minimizing the symbol table 3840 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3841 if (so_path && so_path[0]) 3842 { 3843 std::string full_so_path (so_path); 3844 const size_t double_slash_pos = full_so_path.find("//"); 3845 if (double_slash_pos != std::string::npos) 3846 { 3847 // The linker has been generating bad N_SO entries with doubled up paths 3848 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 3849 // and the second is the directory for the source file so you end up with 3850 // a path that looks like "/tmp/src//tmp/src/" 3851 FileSpec so_dir(so_path, false); 3852 if (!so_dir.Exists()) 3853 { 3854 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 3855 if (so_dir.Exists()) 3856 { 3857 // Trim off the incorrect path 3858 full_so_path.erase(0, double_slash_pos + 1); 3859 } 3860 } 3861 } 3862 if (*full_so_path.rbegin() != '/') 3863 full_so_path += '/'; 3864 full_so_path += symbol_name; 3865 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 3866 add_nlist = false; 3867 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3868 } 3869 } 3870 else 3871 { 3872 // This could be a relative path to a N_SO 3873 N_SO_index = sym_idx; 3874 } 3875 } 3876 3877 break; 3878 3879 case N_OSO: 3880 // object file name: name,,0,0,st_mtime 3881 type = eSymbolTypeObjectFile; 3882 break; 3883 3884 case N_LSYM: 3885 // local sym: name,,NO_SECT,type,offset 3886 type = eSymbolTypeLocal; 3887 break; 3888 3889 //---------------------------------------------------------------------- 3890 // INCL scopes 3891 //---------------------------------------------------------------------- 3892 case N_BINCL: 3893 // include file beginning: name,,NO_SECT,0,sum 3894 // We use the current number of symbols in the symbol table in lieu of 3895 // using nlist_idx in case we ever start trimming entries out 3896 N_INCL_indexes.push_back(sym_idx); 3897 type = eSymbolTypeScopeBegin; 3898 break; 3899 3900 case N_EINCL: 3901 // include file end: name,,NO_SECT,0,0 3902 // Set the size of the N_BINCL to the terminating index of this N_EINCL 3903 // so that we can always skip the entire symbol if we need to navigate 3904 // more quickly at the source level when parsing STABS 3905 if ( !N_INCL_indexes.empty() ) 3906 { 3907 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3908 symbol_ptr->SetByteSize(sym_idx + 1); 3909 symbol_ptr->SetSizeIsSibling(true); 3910 N_INCL_indexes.pop_back(); 3911 } 3912 type = eSymbolTypeScopeEnd; 3913 break; 3914 3915 case N_SOL: 3916 // #included file name: name,,n_sect,0,address 3917 type = eSymbolTypeHeaderFile; 3918 3919 // We currently don't use the header files on darwin 3920 add_nlist = false; 3921 break; 3922 3923 case N_PARAMS: 3924 // compiler parameters: name,,NO_SECT,0,0 3925 type = eSymbolTypeCompiler; 3926 break; 3927 3928 case N_VERSION: 3929 // compiler version: name,,NO_SECT,0,0 3930 type = eSymbolTypeCompiler; 3931 break; 3932 3933 case N_OLEVEL: 3934 // compiler -O level: name,,NO_SECT,0,0 3935 type = eSymbolTypeCompiler; 3936 break; 3937 3938 case N_PSYM: 3939 // parameter: name,,NO_SECT,type,offset 3940 type = eSymbolTypeVariable; 3941 break; 3942 3943 case N_ENTRY: 3944 // alternate entry: name,,n_sect,linenumber,address 3945 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3946 type = eSymbolTypeLineEntry; 3947 break; 3948 3949 //---------------------------------------------------------------------- 3950 // Left and Right Braces 3951 //---------------------------------------------------------------------- 3952 case N_LBRAC: 3953 // left bracket: 0,,NO_SECT,nesting level,address 3954 // We use the current number of symbols in the symbol table in lieu of 3955 // using nlist_idx in case we ever start trimming entries out 3956 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3957 N_BRAC_indexes.push_back(sym_idx); 3958 type = eSymbolTypeScopeBegin; 3959 break; 3960 3961 case N_RBRAC: 3962 // right bracket: 0,,NO_SECT,nesting level,address 3963 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 3964 // so that we can always skip the entire symbol if we need to navigate 3965 // more quickly at the source level when parsing STABS 3966 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3967 if ( !N_BRAC_indexes.empty() ) 3968 { 3969 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3970 symbol_ptr->SetByteSize(sym_idx + 1); 3971 symbol_ptr->SetSizeIsSibling(true); 3972 N_BRAC_indexes.pop_back(); 3973 } 3974 type = eSymbolTypeScopeEnd; 3975 break; 3976 3977 case N_EXCL: 3978 // deleted include file: name,,NO_SECT,0,sum 3979 type = eSymbolTypeHeaderFile; 3980 break; 3981 3982 //---------------------------------------------------------------------- 3983 // COMM scopes 3984 //---------------------------------------------------------------------- 3985 case N_BCOMM: 3986 // begin common: name,,NO_SECT,0,0 3987 // We use the current number of symbols in the symbol table in lieu of 3988 // using nlist_idx in case we ever start trimming entries out 3989 type = eSymbolTypeScopeBegin; 3990 N_COMM_indexes.push_back(sym_idx); 3991 break; 3992 3993 case N_ECOML: 3994 // end common (local name): 0,,n_sect,0,address 3995 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3996 // Fall through 3997 3998 case N_ECOMM: 3999 // end common: name,,n_sect,0,0 4000 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 4001 // so that we can always skip the entire symbol if we need to navigate 4002 // more quickly at the source level when parsing STABS 4003 if ( !N_COMM_indexes.empty() ) 4004 { 4005 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4006 symbol_ptr->SetByteSize(sym_idx + 1); 4007 symbol_ptr->SetSizeIsSibling(true); 4008 N_COMM_indexes.pop_back(); 4009 } 4010 type = eSymbolTypeScopeEnd; 4011 break; 4012 4013 case N_LENG: 4014 // second stab entry with length information 4015 type = eSymbolTypeAdditional; 4016 break; 4017 4018 default: break; 4019 } 4020 } 4021 else 4022 { 4023 //uint8_t n_pext = N_PEXT & nlist.n_type; 4024 uint8_t n_type = N_TYPE & nlist.n_type; 4025 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4026 4027 switch (n_type) 4028 { 4029 case N_INDR: 4030 { 4031 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4032 if (reexport_name_cstr && reexport_name_cstr[0]) 4033 { 4034 type = eSymbolTypeReExported; 4035 ConstString reexport_name(reexport_name_cstr + ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4036 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4037 set_value = false; 4038 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4039 indirect_symbol_names.insert(ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4040 } 4041 else 4042 type = eSymbolTypeUndefined; 4043 } 4044 break; 4045 4046 case N_UNDF: 4047 if (symbol_name && symbol_name[0]) 4048 { 4049 ConstString undefined_name(symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 4050 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4051 } 4052 // Fall through 4053 case N_PBUD: 4054 type = eSymbolTypeUndefined; 4055 break; 4056 4057 case N_ABS: 4058 type = eSymbolTypeAbsolute; 4059 break; 4060 4061 case N_SECT: 4062 { 4063 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 4064 4065 if (!symbol_section) 4066 { 4067 // TODO: warn about this? 4068 add_nlist = false; 4069 break; 4070 } 4071 4072 if (TEXT_eh_frame_sectID == nlist.n_sect) 4073 { 4074 type = eSymbolTypeException; 4075 } 4076 else 4077 { 4078 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4079 4080 switch (section_type) 4081 { 4082 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 4083 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 4084 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 4085 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 4086 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 4087 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 4088 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 4089 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 4090 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 4091 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 4092 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 4093 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 4094 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 4095 default: 4096 switch (symbol_section->GetType()) 4097 { 4098 case lldb::eSectionTypeCode: 4099 type = eSymbolTypeCode; 4100 break; 4101 case eSectionTypeData: 4102 case eSectionTypeDataCString: // Inlined C string data 4103 case eSectionTypeDataCStringPointers: // Pointers to C string data 4104 case eSectionTypeDataSymbolAddress: // Address of a symbol in the symbol table 4105 case eSectionTypeData4: 4106 case eSectionTypeData8: 4107 case eSectionTypeData16: 4108 type = eSymbolTypeData; 4109 break; 4110 default: 4111 break; 4112 } 4113 break; 4114 } 4115 4116 if (type == eSymbolTypeInvalid) 4117 { 4118 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 4119 if (symbol_section->IsDescendant (text_section_sp.get())) 4120 { 4121 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4122 S_ATTR_SELF_MODIFYING_CODE | 4123 S_ATTR_SOME_INSTRUCTIONS)) 4124 type = eSymbolTypeData; 4125 else 4126 type = eSymbolTypeCode; 4127 } 4128 else 4129 if (symbol_section->IsDescendant(data_section_sp.get())) 4130 { 4131 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 4132 { 4133 type = eSymbolTypeRuntime; 4134 4135 if (symbol_name && 4136 symbol_name[0] == '_' && 4137 symbol_name[1] == 'O' && 4138 symbol_name[2] == 'B') 4139 { 4140 llvm::StringRef symbol_name_ref(symbol_name); 4141 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 4142 { 4143 symbol_name_non_abi_mangled = symbol_name + 1; 4144 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 4145 type = eSymbolTypeObjCClass; 4146 demangled_is_synthesized = true; 4147 } 4148 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 4149 { 4150 symbol_name_non_abi_mangled = symbol_name + 1; 4151 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 4152 type = eSymbolTypeObjCMetaClass; 4153 demangled_is_synthesized = true; 4154 } 4155 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 4156 { 4157 symbol_name_non_abi_mangled = symbol_name + 1; 4158 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 4159 type = eSymbolTypeObjCIVar; 4160 demangled_is_synthesized = true; 4161 } 4162 } 4163 } 4164 else 4165 if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 4166 { 4167 type = eSymbolTypeException; 4168 } 4169 else 4170 { 4171 type = eSymbolTypeData; 4172 } 4173 } 4174 else 4175 if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 4176 { 4177 type = eSymbolTypeTrampoline; 4178 } 4179 else 4180 if (symbol_section->IsDescendant(objc_section_sp.get())) 4181 { 4182 type = eSymbolTypeRuntime; 4183 if (symbol_name && symbol_name[0] == '.') 4184 { 4185 llvm::StringRef symbol_name_ref(symbol_name); 4186 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 4187 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 4188 { 4189 symbol_name_non_abi_mangled = symbol_name; 4190 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4191 type = eSymbolTypeObjCClass; 4192 demangled_is_synthesized = true; 4193 } 4194 } 4195 } 4196 } 4197 } 4198 } 4199 break; 4200 } 4201 } 4202 4203 if (add_nlist) 4204 { 4205 uint64_t symbol_value = nlist.n_value; 4206 4207 if (symbol_name_non_abi_mangled) 4208 { 4209 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 4210 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 4211 } 4212 else 4213 { 4214 bool symbol_name_is_mangled = false; 4215 4216 if (symbol_name && symbol_name[0] == '_') 4217 { 4218 symbol_name_is_mangled = symbol_name[1] == '_'; 4219 symbol_name++; // Skip the leading underscore 4220 } 4221 4222 if (symbol_name) 4223 { 4224 ConstString const_symbol_name(symbol_name); 4225 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 4226 } 4227 } 4228 4229 if (is_gsym) 4230 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 4231 4232 if (symbol_section) 4233 { 4234 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4235 if (symbol_byte_size == 0 && function_starts_count > 0) 4236 { 4237 addr_t symbol_lookup_file_addr = nlist.n_value; 4238 // Do an exact address match for non-ARM addresses, else get the closest since 4239 // the symbol might be a thumb symbol which has an address with bit zero set 4240 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 4241 if (is_arm && func_start_entry) 4242 { 4243 // Verify that the function start address is the symbol address (ARM) 4244 // or the symbol address + 1 (thumb) 4245 if (func_start_entry->addr != symbol_lookup_file_addr && 4246 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 4247 { 4248 // Not the right entry, NULL it out... 4249 func_start_entry = NULL; 4250 } 4251 } 4252 if (func_start_entry) 4253 { 4254 func_start_entry->data = true; 4255 4256 addr_t symbol_file_addr = func_start_entry->addr; 4257 if (is_arm) 4258 symbol_file_addr &= 0xfffffffffffffffeull; 4259 4260 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 4261 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 4262 if (next_func_start_entry) 4263 { 4264 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4265 // Be sure the clear the Thumb address bit when we calculate the size 4266 // from the current and next address 4267 if (is_arm) 4268 next_symbol_file_addr &= 0xfffffffffffffffeull; 4269 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 4270 } 4271 else 4272 { 4273 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4274 } 4275 } 4276 } 4277 symbol_value -= section_file_addr; 4278 } 4279 4280 if (is_debug == false) 4281 { 4282 if (type == eSymbolTypeCode) 4283 { 4284 // See if we can find a N_FUN entry for any code symbols. 4285 // If we do find a match, and the name matches, then we 4286 // can merge the two into just the function symbol to avoid 4287 // duplicate entries in the symbol table 4288 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 4289 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4290 if (range.first != range.second) 4291 { 4292 bool found_it = false; 4293 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 4294 { 4295 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 4296 { 4297 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4298 // We just need the flags from the linker symbol, so put these flags 4299 // into the N_FUN flags to avoid duplicate symbols in the symbol table 4300 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4301 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4302 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 4303 sym[pos->second].SetType (eSymbolTypeResolver); 4304 sym[sym_idx].Clear(); 4305 found_it = true; 4306 break; 4307 } 4308 } 4309 if (found_it) 4310 continue; 4311 } 4312 else 4313 { 4314 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 4315 type = eSymbolTypeResolver; 4316 } 4317 } 4318 else if (type == eSymbolTypeData || 4319 type == eSymbolTypeObjCClass || 4320 type == eSymbolTypeObjCMetaClass || 4321 type == eSymbolTypeObjCIVar ) 4322 { 4323 // See if we can find a N_STSYM entry for any data symbols. 4324 // If we do find a match, and the name matches, then we 4325 // can merge the two into just the Static symbol to avoid 4326 // duplicate entries in the symbol table 4327 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 4328 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4329 if (range.first != range.second) 4330 { 4331 bool found_it = false; 4332 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 4333 { 4334 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 4335 { 4336 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4337 // We just need the flags from the linker symbol, so put these flags 4338 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 4339 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4340 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4341 sym[sym_idx].Clear(); 4342 found_it = true; 4343 break; 4344 } 4345 } 4346 if (found_it) 4347 continue; 4348 } 4349 else 4350 { 4351 // Combine N_GSYM stab entries with the non stab symbol 4352 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 4353 if (pos != N_GSYM_name_to_sym_idx.end()) 4354 { 4355 const uint32_t GSYM_sym_idx = pos->second; 4356 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4357 // Copy the address, because often the N_GSYM address has an invalid address of zero 4358 // when the global is a common symbol 4359 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 4360 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 4361 // We just need the flags from the linker symbol, so put these flags 4362 // into the N_GSYM flags to avoid duplicate symbols in the symbol table 4363 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4364 sym[sym_idx].Clear(); 4365 continue; 4366 } 4367 } 4368 } 4369 } 4370 4371 sym[sym_idx].SetID (nlist_idx); 4372 sym[sym_idx].SetType (type); 4373 if (set_value) 4374 { 4375 sym[sym_idx].GetAddress().SetSection (symbol_section); 4376 sym[sym_idx].GetAddress().SetOffset (symbol_value); 4377 } 4378 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4379 4380 if (symbol_byte_size > 0) 4381 sym[sym_idx].SetByteSize(symbol_byte_size); 4382 4383 if (demangled_is_synthesized) 4384 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4385 4386 ++sym_idx; 4387 } 4388 else 4389 { 4390 sym[sym_idx].Clear(); 4391 } 4392 } 4393 4394 for (const auto &pos :reexport_shlib_needs_fixup) 4395 { 4396 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4397 if (undef_pos != undefined_name_to_desc.end()) 4398 { 4399 const uint8_t dylib_ordinal = llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4400 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4401 sym[pos.first].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(dylib_ordinal-1)); 4402 } 4403 } 4404 4405 } 4406 4407 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4408 4409 if (function_starts_count > 0) 4410 { 4411 char synthetic_function_symbol[PATH_MAX]; 4412 uint32_t num_synthetic_function_symbols = 0; 4413 for (i=0; i<function_starts_count; ++i) 4414 { 4415 if (function_starts.GetEntryRef (i).data == false) 4416 ++num_synthetic_function_symbols; 4417 } 4418 4419 if (num_synthetic_function_symbols > 0) 4420 { 4421 if (num_syms < sym_idx + num_synthetic_function_symbols) 4422 { 4423 num_syms = sym_idx + num_synthetic_function_symbols; 4424 sym = symtab->Resize (num_syms); 4425 } 4426 uint32_t synthetic_function_symbol_idx = 0; 4427 for (i=0; i<function_starts_count; ++i) 4428 { 4429 const FunctionStarts::Entry *func_start_entry = function_starts.GetEntryAtIndex (i); 4430 if (func_start_entry->data == false) 4431 { 4432 addr_t symbol_file_addr = func_start_entry->addr; 4433 uint32_t symbol_flags = 0; 4434 if (is_arm) 4435 { 4436 if (symbol_file_addr & 1) 4437 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4438 symbol_file_addr &= 0xfffffffffffffffeull; 4439 } 4440 Address symbol_addr; 4441 if (module_sp->ResolveFileAddress (symbol_file_addr, symbol_addr)) 4442 { 4443 SectionSP symbol_section (symbol_addr.GetSection()); 4444 uint32_t symbol_byte_size = 0; 4445 if (symbol_section) 4446 { 4447 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4448 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 4449 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 4450 if (next_func_start_entry) 4451 { 4452 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4453 if (is_arm) 4454 next_symbol_file_addr &= 0xfffffffffffffffeull; 4455 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 4456 } 4457 else 4458 { 4459 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4460 } 4461 snprintf (synthetic_function_symbol, 4462 sizeof(synthetic_function_symbol), 4463 "___lldb_unnamed_function%u$$%s", 4464 ++synthetic_function_symbol_idx, 4465 module_sp->GetFileSpec().GetFilename().GetCString()); 4466 sym[sym_idx].SetID (synthetic_sym_id++); 4467 sym[sym_idx].GetMangled().SetDemangledName(ConstString(synthetic_function_symbol)); 4468 sym[sym_idx].SetType (eSymbolTypeCode); 4469 sym[sym_idx].SetIsSynthetic (true); 4470 sym[sym_idx].GetAddress() = symbol_addr; 4471 if (symbol_flags) 4472 sym[sym_idx].SetFlags (symbol_flags); 4473 if (symbol_byte_size) 4474 sym[sym_idx].SetByteSize (symbol_byte_size); 4475 ++sym_idx; 4476 } 4477 } 4478 } 4479 } 4480 } 4481 } 4482 4483 // Trim our symbols down to just what we ended up with after 4484 // removing any symbols. 4485 if (sym_idx < num_syms) 4486 { 4487 num_syms = sym_idx; 4488 sym = symtab->Resize (num_syms); 4489 } 4490 4491 // Now synthesize indirect symbols 4492 if (m_dysymtab.nindirectsyms != 0) 4493 { 4494 if (indirect_symbol_index_data.GetByteSize()) 4495 { 4496 NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end(); 4497 4498 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx) 4499 { 4500 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == S_SYMBOL_STUBS) 4501 { 4502 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4503 if (symbol_stub_byte_size == 0) 4504 continue; 4505 4506 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4507 4508 if (num_symbol_stubs == 0) 4509 continue; 4510 4511 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1; 4512 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) 4513 { 4514 const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx; 4515 const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size); 4516 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4517 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4)) 4518 { 4519 const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset); 4520 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4521 continue; 4522 4523 NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id); 4524 Symbol *stub_symbol = NULL; 4525 if (index_pos != end_index_pos) 4526 { 4527 // We have a remapping from the original nlist index to 4528 // a current symbol index, so just look this up by index 4529 stub_symbol = symtab->SymbolAtIndex (index_pos->second); 4530 } 4531 else 4532 { 4533 // We need to lookup a symbol using the original nlist 4534 // symbol index since this index is coming from the 4535 // S_SYMBOL_STUBS 4536 stub_symbol = symtab->FindSymbolByID (stub_sym_id); 4537 } 4538 4539 if (stub_symbol) 4540 { 4541 Address so_addr(symbol_stub_addr, section_list); 4542 4543 if (stub_symbol->GetType() == eSymbolTypeUndefined) 4544 { 4545 // Change the external symbol into a trampoline that makes sense 4546 // These symbols were N_UNDF N_EXT, and are useless to us, so we 4547 // can re-use them so we don't have to make up a synthetic symbol 4548 // for no good reason. 4549 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 4550 stub_symbol->SetType (eSymbolTypeTrampoline); 4551 else 4552 stub_symbol->SetType (eSymbolTypeResolver); 4553 stub_symbol->SetExternal (false); 4554 stub_symbol->GetAddress() = so_addr; 4555 stub_symbol->SetByteSize (symbol_stub_byte_size); 4556 } 4557 else 4558 { 4559 // Make a synthetic symbol to describe the trampoline stub 4560 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4561 if (sym_idx >= num_syms) 4562 { 4563 sym = symtab->Resize (++num_syms); 4564 stub_symbol = NULL; // this pointer no longer valid 4565 } 4566 sym[sym_idx].SetID (synthetic_sym_id++); 4567 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4568 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 4569 sym[sym_idx].SetType (eSymbolTypeTrampoline); 4570 else 4571 sym[sym_idx].SetType (eSymbolTypeResolver); 4572 sym[sym_idx].SetIsSynthetic (true); 4573 sym[sym_idx].GetAddress() = so_addr; 4574 sym[sym_idx].SetByteSize (symbol_stub_byte_size); 4575 ++sym_idx; 4576 } 4577 } 4578 else 4579 { 4580 if (log) 4581 log->Warning ("symbol stub referencing symbol table symbol %u that isn't in our minimal symbol table, fix this!!!", stub_sym_id); 4582 } 4583 } 4584 } 4585 } 4586 } 4587 } 4588 } 4589 4590 4591 if (!trie_entries.empty()) 4592 { 4593 for (const auto &e : trie_entries) 4594 { 4595 if (e.entry.import_name) 4596 { 4597 // Only add indirect symbols from the Trie entries if we 4598 // didn't have a N_INDR nlist entry for this already 4599 if (indirect_symbol_names.find(e.entry.name) == indirect_symbol_names.end()) 4600 { 4601 // Make a synthetic symbol to describe re-exported symbol. 4602 if (sym_idx >= num_syms) 4603 sym = symtab->Resize (++num_syms); 4604 sym[sym_idx].SetID (synthetic_sym_id++); 4605 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4606 sym[sym_idx].SetType (eSymbolTypeReExported); 4607 sym[sym_idx].SetIsSynthetic (true); 4608 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4609 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) 4610 { 4611 sym[sym_idx].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(e.entry.other-1)); 4612 } 4613 ++sym_idx; 4614 } 4615 } 4616 } 4617 } 4618 4619 4620 4621 // StreamFile s(stdout, false); 4622 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4623 // symtab->Dump(&s, NULL, eSortOrderNone); 4624 // Set symbol byte sizes correctly since mach-o nlist entries don't have sizes 4625 symtab->CalculateSymbolSizes(); 4626 4627 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4628 // symtab->Dump(&s, NULL, eSortOrderNone); 4629 4630 return symtab->GetNumSymbols(); 4631 } 4632 return 0; 4633 } 4634 4635 4636 void 4637 ObjectFileMachO::Dump (Stream *s) 4638 { 4639 ModuleSP module_sp(GetModule()); 4640 if (module_sp) 4641 { 4642 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4643 s->Printf("%p: ", static_cast<void*>(this)); 4644 s->Indent(); 4645 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4646 s->PutCString("ObjectFileMachO64"); 4647 else 4648 s->PutCString("ObjectFileMachO32"); 4649 4650 ArchSpec header_arch; 4651 GetArchitecture(header_arch); 4652 4653 *s << ", file = '" << m_file << "', arch = " << header_arch.GetArchitectureName() << "\n"; 4654 4655 SectionList *sections = GetSectionList(); 4656 if (sections) 4657 sections->Dump(s, NULL, true, UINT32_MAX); 4658 4659 if (m_symtab_ap.get()) 4660 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4661 } 4662 } 4663 4664 bool 4665 ObjectFileMachO::GetUUID (const llvm::MachO::mach_header &header, 4666 const lldb_private::DataExtractor &data, 4667 lldb::offset_t lc_offset, 4668 lldb_private::UUID& uuid) 4669 { 4670 uint32_t i; 4671 struct uuid_command load_cmd; 4672 4673 lldb::offset_t offset = lc_offset; 4674 for (i=0; i<header.ncmds; ++i) 4675 { 4676 const lldb::offset_t cmd_offset = offset; 4677 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4678 break; 4679 4680 if (load_cmd.cmd == LC_UUID) 4681 { 4682 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4683 4684 if (uuid_bytes) 4685 { 4686 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4687 // We pretend these object files have no UUID to prevent crashing. 4688 4689 const uint8_t opencl_uuid[] = { 0x8c, 0x8e, 0xb3, 0x9b, 4690 0x3b, 0xa8, 4691 0x4b, 0x16, 4692 0xb6, 0xa4, 4693 0x27, 0x63, 0xbb, 0x14, 0xf0, 0x0d }; 4694 4695 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4696 return false; 4697 4698 uuid.SetBytes (uuid_bytes); 4699 return true; 4700 } 4701 return false; 4702 } 4703 offset = cmd_offset + load_cmd.cmdsize; 4704 } 4705 return false; 4706 } 4707 4708 4709 bool 4710 ObjectFileMachO::GetArchitecture (const llvm::MachO::mach_header &header, 4711 const lldb_private::DataExtractor &data, 4712 lldb::offset_t lc_offset, 4713 ArchSpec &arch) 4714 { 4715 arch.SetArchitecture (eArchTypeMachO, header.cputype, header.cpusubtype); 4716 4717 if (arch.IsValid()) 4718 { 4719 llvm::Triple &triple = arch.GetTriple(); 4720 if (header.filetype == MH_PRELOAD) 4721 { 4722 // Set OS to "unknown" - this is a standalone binary with no dyld et al 4723 triple.setOS(llvm::Triple::UnknownOS); 4724 return true; 4725 } 4726 else 4727 { 4728 struct load_command load_cmd; 4729 4730 lldb::offset_t offset = lc_offset; 4731 for (uint32_t i=0; i<header.ncmds; ++i) 4732 { 4733 const lldb::offset_t cmd_offset = offset; 4734 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4735 break; 4736 4737 switch (load_cmd.cmd) 4738 { 4739 case LC_VERSION_MIN_IPHONEOS: 4740 triple.setOS (llvm::Triple::IOS); 4741 return true; 4742 4743 case LC_VERSION_MIN_MACOSX: 4744 triple.setOS (llvm::Triple::MacOSX); 4745 return true; 4746 4747 default: 4748 break; 4749 } 4750 4751 offset = cmd_offset + load_cmd.cmdsize; 4752 } 4753 4754 // Only set the OS to iOS for ARM, we don't want to set it for x86 and x86_64. 4755 // We do this because we now have MacOSX or iOS as the OS value for x86 and 4756 // x86_64 for normal desktop (MacOSX) and simulator (iOS) binaries. And if 4757 // we compare a "x86_64-apple-ios" to a "x86_64-apple-" triple, it will say 4758 // it is compatible (because the OS is unspecified in the second one and will 4759 // match anything in the first 4760 if (header.cputype == CPU_TYPE_ARM || header.cputype == CPU_TYPE_ARM64) 4761 triple.setOS (llvm::Triple::IOS); 4762 } 4763 } 4764 return arch.IsValid(); 4765 } 4766 4767 bool 4768 ObjectFileMachO::GetUUID (lldb_private::UUID* uuid) 4769 { 4770 ModuleSP module_sp(GetModule()); 4771 if (module_sp) 4772 { 4773 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4774 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4775 return GetUUID (m_header, m_data, offset, *uuid); 4776 } 4777 return false; 4778 } 4779 4780 4781 uint32_t 4782 ObjectFileMachO::GetDependentModules (FileSpecList& files) 4783 { 4784 uint32_t count = 0; 4785 ModuleSP module_sp(GetModule()); 4786 if (module_sp) 4787 { 4788 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4789 struct load_command load_cmd; 4790 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4791 const bool resolve_path = false; // Don't resolve the dependent file paths since they may not reside on this system 4792 uint32_t i; 4793 for (i=0; i<m_header.ncmds; ++i) 4794 { 4795 const uint32_t cmd_offset = offset; 4796 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4797 break; 4798 4799 switch (load_cmd.cmd) 4800 { 4801 case LC_LOAD_DYLIB: 4802 case LC_LOAD_WEAK_DYLIB: 4803 case LC_REEXPORT_DYLIB: 4804 case LC_LOAD_DYLINKER: 4805 case LC_LOADFVMLIB: 4806 case LC_LOAD_UPWARD_DYLIB: 4807 { 4808 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 4809 const char *path = m_data.PeekCStr(name_offset); 4810 // Skip any path that starts with '@' since these are usually: 4811 // @executable_path/.../file 4812 // @rpath/.../file 4813 if (path && path[0] != '@') 4814 { 4815 FileSpec file_spec(path, resolve_path); 4816 if (files.AppendIfUnique(file_spec)) 4817 count++; 4818 } 4819 } 4820 break; 4821 4822 default: 4823 break; 4824 } 4825 offset = cmd_offset + load_cmd.cmdsize; 4826 } 4827 } 4828 return count; 4829 } 4830 4831 lldb_private::Address 4832 ObjectFileMachO::GetEntryPointAddress () 4833 { 4834 // If the object file is not an executable it can't hold the entry point. m_entry_point_address 4835 // is initialized to an invalid address, so we can just return that. 4836 // If m_entry_point_address is valid it means we've found it already, so return the cached value. 4837 4838 if (!IsExecutable() || m_entry_point_address.IsValid()) 4839 return m_entry_point_address; 4840 4841 // Otherwise, look for the UnixThread or Thread command. The data for the Thread command is given in 4842 // /usr/include/mach-o.h, but it is basically: 4843 // 4844 // uint32_t flavor - this is the flavor argument you would pass to thread_get_state 4845 // uint32_t count - this is the count of longs in the thread state data 4846 // struct XXX_thread_state state - this is the structure from <machine/thread_status.h> corresponding to the flavor. 4847 // <repeat this trio> 4848 // 4849 // So we just keep reading the various register flavors till we find the GPR one, then read the PC out of there. 4850 // FIXME: We will need to have a "RegisterContext data provider" class at some point that can get all the registers 4851 // out of data in this form & attach them to a given thread. That should underlie the MacOS X User process plugin, 4852 // and we'll also need it for the MacOS X Core File process plugin. When we have that we can also use it here. 4853 // 4854 // For now we hard-code the offsets and flavors we need: 4855 // 4856 // 4857 4858 ModuleSP module_sp(GetModule()); 4859 if (module_sp) 4860 { 4861 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4862 struct load_command load_cmd; 4863 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4864 uint32_t i; 4865 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 4866 bool done = false; 4867 4868 for (i=0; i<m_header.ncmds; ++i) 4869 { 4870 const lldb::offset_t cmd_offset = offset; 4871 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4872 break; 4873 4874 switch (load_cmd.cmd) 4875 { 4876 case LC_UNIXTHREAD: 4877 case LC_THREAD: 4878 { 4879 while (offset < cmd_offset + load_cmd.cmdsize) 4880 { 4881 uint32_t flavor = m_data.GetU32(&offset); 4882 uint32_t count = m_data.GetU32(&offset); 4883 if (count == 0) 4884 { 4885 // We've gotten off somehow, log and exit; 4886 return m_entry_point_address; 4887 } 4888 4889 switch (m_header.cputype) 4890 { 4891 case llvm::MachO::CPU_TYPE_ARM: 4892 if (flavor == 1) // ARM_THREAD_STATE from mach/arm/thread_status.h 4893 { 4894 offset += 60; // This is the offset of pc in the GPR thread state data structure. 4895 start_address = m_data.GetU32(&offset); 4896 done = true; 4897 } 4898 break; 4899 case llvm::MachO::CPU_TYPE_ARM64: 4900 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 4901 { 4902 offset += 256; // This is the offset of pc in the GPR thread state data structure. 4903 start_address = m_data.GetU64(&offset); 4904 done = true; 4905 } 4906 break; 4907 case llvm::MachO::CPU_TYPE_I386: 4908 if (flavor == 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 4909 { 4910 offset += 40; // This is the offset of eip in the GPR thread state data structure. 4911 start_address = m_data.GetU32(&offset); 4912 done = true; 4913 } 4914 break; 4915 case llvm::MachO::CPU_TYPE_X86_64: 4916 if (flavor == 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 4917 { 4918 offset += 16 * 8; // This is the offset of rip in the GPR thread state data structure. 4919 start_address = m_data.GetU64(&offset); 4920 done = true; 4921 } 4922 break; 4923 default: 4924 return m_entry_point_address; 4925 } 4926 // Haven't found the GPR flavor yet, skip over the data for this flavor: 4927 if (done) 4928 break; 4929 offset += count * 4; 4930 } 4931 } 4932 break; 4933 case LC_MAIN: 4934 { 4935 ConstString text_segment_name ("__TEXT"); 4936 uint64_t entryoffset = m_data.GetU64(&offset); 4937 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 4938 if (text_segment_sp) 4939 { 4940 done = true; 4941 start_address = text_segment_sp->GetFileAddress() + entryoffset; 4942 } 4943 } 4944 4945 default: 4946 break; 4947 } 4948 if (done) 4949 break; 4950 4951 // Go to the next load command: 4952 offset = cmd_offset + load_cmd.cmdsize; 4953 } 4954 4955 if (start_address != LLDB_INVALID_ADDRESS) 4956 { 4957 // We got the start address from the load commands, so now resolve that address in the sections 4958 // of this ObjectFile: 4959 if (!m_entry_point_address.ResolveAddressUsingFileSections (start_address, GetSectionList())) 4960 { 4961 m_entry_point_address.Clear(); 4962 } 4963 } 4964 else 4965 { 4966 // We couldn't read the UnixThread load command - maybe it wasn't there. As a fallback look for the 4967 // "start" symbol in the main executable. 4968 4969 ModuleSP module_sp (GetModule()); 4970 4971 if (module_sp) 4972 { 4973 SymbolContextList contexts; 4974 SymbolContext context; 4975 if (module_sp->FindSymbolsWithNameAndType(ConstString ("start"), eSymbolTypeCode, contexts)) 4976 { 4977 if (contexts.GetContextAtIndex(0, context)) 4978 m_entry_point_address = context.symbol->GetAddress(); 4979 } 4980 } 4981 } 4982 } 4983 4984 return m_entry_point_address; 4985 4986 } 4987 4988 lldb_private::Address 4989 ObjectFileMachO::GetHeaderAddress () 4990 { 4991 lldb_private::Address header_addr; 4992 SectionList *section_list = GetSectionList(); 4993 if (section_list) 4994 { 4995 SectionSP text_segment_sp (section_list->FindSectionByName (GetSegmentNameTEXT())); 4996 if (text_segment_sp) 4997 { 4998 header_addr.SetSection (text_segment_sp); 4999 header_addr.SetOffset (0); 5000 } 5001 } 5002 return header_addr; 5003 } 5004 5005 uint32_t 5006 ObjectFileMachO::GetNumThreadContexts () 5007 { 5008 ModuleSP module_sp(GetModule()); 5009 if (module_sp) 5010 { 5011 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 5012 if (!m_thread_context_offsets_valid) 5013 { 5014 m_thread_context_offsets_valid = true; 5015 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5016 FileRangeArray::Entry file_range; 5017 thread_command thread_cmd; 5018 for (uint32_t i=0; i<m_header.ncmds; ++i) 5019 { 5020 const uint32_t cmd_offset = offset; 5021 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 5022 break; 5023 5024 if (thread_cmd.cmd == LC_THREAD) 5025 { 5026 file_range.SetRangeBase (offset); 5027 file_range.SetByteSize (thread_cmd.cmdsize - 8); 5028 m_thread_context_offsets.Append (file_range); 5029 } 5030 offset = cmd_offset + thread_cmd.cmdsize; 5031 } 5032 } 5033 } 5034 return m_thread_context_offsets.GetSize(); 5035 } 5036 5037 lldb::RegisterContextSP 5038 ObjectFileMachO::GetThreadContextAtIndex (uint32_t idx, lldb_private::Thread &thread) 5039 { 5040 lldb::RegisterContextSP reg_ctx_sp; 5041 5042 ModuleSP module_sp(GetModule()); 5043 if (module_sp) 5044 { 5045 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 5046 if (!m_thread_context_offsets_valid) 5047 GetNumThreadContexts (); 5048 5049 const FileRangeArray::Entry *thread_context_file_range = m_thread_context_offsets.GetEntryAtIndex (idx); 5050 if (thread_context_file_range) 5051 { 5052 5053 DataExtractor data (m_data, 5054 thread_context_file_range->GetRangeBase(), 5055 thread_context_file_range->GetByteSize()); 5056 5057 switch (m_header.cputype) 5058 { 5059 case llvm::MachO::CPU_TYPE_ARM64: 5060 reg_ctx_sp.reset (new RegisterContextDarwin_arm64_Mach (thread, data)); 5061 break; 5062 5063 case llvm::MachO::CPU_TYPE_ARM: 5064 reg_ctx_sp.reset (new RegisterContextDarwin_arm_Mach (thread, data)); 5065 break; 5066 5067 case llvm::MachO::CPU_TYPE_I386: 5068 reg_ctx_sp.reset (new RegisterContextDarwin_i386_Mach (thread, data)); 5069 break; 5070 5071 case llvm::MachO::CPU_TYPE_X86_64: 5072 reg_ctx_sp.reset (new RegisterContextDarwin_x86_64_Mach (thread, data)); 5073 break; 5074 } 5075 } 5076 } 5077 return reg_ctx_sp; 5078 } 5079 5080 5081 ObjectFile::Type 5082 ObjectFileMachO::CalculateType() 5083 { 5084 switch (m_header.filetype) 5085 { 5086 case MH_OBJECT: // 0x1u 5087 if (GetAddressByteSize () == 4) 5088 { 5089 // 32 bit kexts are just object files, but they do have a valid 5090 // UUID load command. 5091 UUID uuid; 5092 if (GetUUID(&uuid)) 5093 { 5094 // this checking for the UUID load command is not enough 5095 // we could eventually look for the symbol named 5096 // "OSKextGetCurrentIdentifier" as this is required of kexts 5097 if (m_strata == eStrataInvalid) 5098 m_strata = eStrataKernel; 5099 return eTypeSharedLibrary; 5100 } 5101 } 5102 return eTypeObjectFile; 5103 5104 case MH_EXECUTE: return eTypeExecutable; // 0x2u 5105 case MH_FVMLIB: return eTypeSharedLibrary; // 0x3u 5106 case MH_CORE: return eTypeCoreFile; // 0x4u 5107 case MH_PRELOAD: return eTypeSharedLibrary; // 0x5u 5108 case MH_DYLIB: return eTypeSharedLibrary; // 0x6u 5109 case MH_DYLINKER: return eTypeDynamicLinker; // 0x7u 5110 case MH_BUNDLE: return eTypeSharedLibrary; // 0x8u 5111 case MH_DYLIB_STUB: return eTypeStubLibrary; // 0x9u 5112 case MH_DSYM: return eTypeDebugInfo; // 0xAu 5113 case MH_KEXT_BUNDLE: return eTypeSharedLibrary; // 0xBu 5114 default: 5115 break; 5116 } 5117 return eTypeUnknown; 5118 } 5119 5120 ObjectFile::Strata 5121 ObjectFileMachO::CalculateStrata() 5122 { 5123 switch (m_header.filetype) 5124 { 5125 case MH_OBJECT: // 0x1u 5126 { 5127 // 32 bit kexts are just object files, but they do have a valid 5128 // UUID load command. 5129 UUID uuid; 5130 if (GetUUID(&uuid)) 5131 { 5132 // this checking for the UUID load command is not enough 5133 // we could eventually look for the symbol named 5134 // "OSKextGetCurrentIdentifier" as this is required of kexts 5135 if (m_type == eTypeInvalid) 5136 m_type = eTypeSharedLibrary; 5137 5138 return eStrataKernel; 5139 } 5140 } 5141 return eStrataUnknown; 5142 5143 case MH_EXECUTE: // 0x2u 5144 // Check for the MH_DYLDLINK bit in the flags 5145 if (m_header.flags & MH_DYLDLINK) 5146 { 5147 return eStrataUser; 5148 } 5149 else 5150 { 5151 SectionList *section_list = GetSectionList(); 5152 if (section_list) 5153 { 5154 static ConstString g_kld_section_name ("__KLD"); 5155 if (section_list->FindSectionByName(g_kld_section_name)) 5156 return eStrataKernel; 5157 } 5158 } 5159 return eStrataRawImage; 5160 5161 case MH_FVMLIB: return eStrataUser; // 0x3u 5162 case MH_CORE: return eStrataUnknown; // 0x4u 5163 case MH_PRELOAD: return eStrataRawImage; // 0x5u 5164 case MH_DYLIB: return eStrataUser; // 0x6u 5165 case MH_DYLINKER: return eStrataUser; // 0x7u 5166 case MH_BUNDLE: return eStrataUser; // 0x8u 5167 case MH_DYLIB_STUB: return eStrataUser; // 0x9u 5168 case MH_DSYM: return eStrataUnknown; // 0xAu 5169 case MH_KEXT_BUNDLE: return eStrataKernel; // 0xBu 5170 default: 5171 break; 5172 } 5173 return eStrataUnknown; 5174 } 5175 5176 5177 uint32_t 5178 ObjectFileMachO::GetVersion (uint32_t *versions, uint32_t num_versions) 5179 { 5180 ModuleSP module_sp(GetModule()); 5181 if (module_sp) 5182 { 5183 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 5184 struct dylib_command load_cmd; 5185 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5186 uint32_t version_cmd = 0; 5187 uint64_t version = 0; 5188 uint32_t i; 5189 for (i=0; i<m_header.ncmds; ++i) 5190 { 5191 const lldb::offset_t cmd_offset = offset; 5192 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5193 break; 5194 5195 if (load_cmd.cmd == LC_ID_DYLIB) 5196 { 5197 if (version_cmd == 0) 5198 { 5199 version_cmd = load_cmd.cmd; 5200 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5201 break; 5202 version = load_cmd.dylib.current_version; 5203 } 5204 break; // Break for now unless there is another more complete version 5205 // number load command in the future. 5206 } 5207 offset = cmd_offset + load_cmd.cmdsize; 5208 } 5209 5210 if (version_cmd == LC_ID_DYLIB) 5211 { 5212 if (versions != NULL && num_versions > 0) 5213 { 5214 if (num_versions > 0) 5215 versions[0] = (version & 0xFFFF0000ull) >> 16; 5216 if (num_versions > 1) 5217 versions[1] = (version & 0x0000FF00ull) >> 8; 5218 if (num_versions > 2) 5219 versions[2] = (version & 0x000000FFull); 5220 // Fill in an remaining version numbers with invalid values 5221 for (i=3; i<num_versions; ++i) 5222 versions[i] = UINT32_MAX; 5223 } 5224 // The LC_ID_DYLIB load command has a version with 3 version numbers 5225 // in it, so always return 3 5226 return 3; 5227 } 5228 } 5229 return false; 5230 } 5231 5232 bool 5233 ObjectFileMachO::GetArchitecture (ArchSpec &arch) 5234 { 5235 ModuleSP module_sp(GetModule()); 5236 if (module_sp) 5237 { 5238 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 5239 return GetArchitecture (m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), arch); 5240 } 5241 return false; 5242 } 5243 5244 5245 UUID 5246 ObjectFileMachO::GetProcessSharedCacheUUID (Process *process) 5247 { 5248 UUID uuid; 5249 if (process) 5250 { 5251 addr_t all_image_infos = process->GetImageInfoAddress(); 5252 5253 // The address returned by GetImageInfoAddress may be the address of dyld (don't want) 5254 // or it may be the address of the dyld_all_image_infos structure (want). The first four 5255 // bytes will be either the version field (all_image_infos) or a Mach-O file magic constant. 5256 // Version 13 and higher of dyld_all_image_infos is required to get the sharedCacheUUID field. 5257 5258 Error err; 5259 uint32_t version_or_magic = process->ReadUnsignedIntegerFromMemory (all_image_infos, 4, -1, err); 5260 if (version_or_magic != static_cast<uint32_t>(-1) 5261 && version_or_magic != MH_MAGIC 5262 && version_or_magic != MH_CIGAM 5263 && version_or_magic != MH_MAGIC_64 5264 && version_or_magic != MH_CIGAM_64 5265 && version_or_magic >= 13) 5266 { 5267 addr_t sharedCacheUUID_address = LLDB_INVALID_ADDRESS; 5268 int wordsize = process->GetAddressByteSize(); 5269 if (wordsize == 8) 5270 { 5271 sharedCacheUUID_address = all_image_infos + 160; // sharedCacheUUID <mach-o/dyld_images.h> 5272 } 5273 if (wordsize == 4) 5274 { 5275 sharedCacheUUID_address = all_image_infos + 84; // sharedCacheUUID <mach-o/dyld_images.h> 5276 } 5277 if (sharedCacheUUID_address != LLDB_INVALID_ADDRESS) 5278 { 5279 uuid_t shared_cache_uuid; 5280 if (process->ReadMemory (sharedCacheUUID_address, shared_cache_uuid, sizeof (uuid_t), err) == sizeof (uuid_t)) 5281 { 5282 uuid.SetBytes (shared_cache_uuid); 5283 } 5284 } 5285 } 5286 } 5287 return uuid; 5288 } 5289 5290 UUID 5291 ObjectFileMachO::GetLLDBSharedCacheUUID () 5292 { 5293 UUID uuid; 5294 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 5295 uint8_t *(*dyld_get_all_image_infos)(void); 5296 dyld_get_all_image_infos = (uint8_t*(*)()) dlsym (RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5297 if (dyld_get_all_image_infos) 5298 { 5299 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5300 if (dyld_all_image_infos_address) 5301 { 5302 uint32_t *version = (uint32_t*) dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5303 if (*version >= 13) 5304 { 5305 uuid_t *sharedCacheUUID_address = 0; 5306 int wordsize = sizeof (uint8_t *); 5307 if (wordsize == 8) 5308 { 5309 sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 160); // sharedCacheUUID <mach-o/dyld_images.h> 5310 } 5311 else 5312 { 5313 sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 84); // sharedCacheUUID <mach-o/dyld_images.h> 5314 } 5315 uuid.SetBytes (sharedCacheUUID_address); 5316 } 5317 } 5318 } 5319 #endif 5320 return uuid; 5321 } 5322 5323 uint32_t 5324 ObjectFileMachO::GetMinimumOSVersion (uint32_t *versions, uint32_t num_versions) 5325 { 5326 if (m_min_os_versions.empty()) 5327 { 5328 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5329 bool success = false; 5330 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 5331 { 5332 const lldb::offset_t load_cmd_offset = offset; 5333 5334 version_min_command lc; 5335 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5336 break; 5337 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 5338 { 5339 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 5340 { 5341 const uint32_t xxxx = lc.version >> 16; 5342 const uint32_t yy = (lc.version >> 8) & 0xffu; 5343 const uint32_t zz = lc.version & 0xffu; 5344 if (xxxx) 5345 { 5346 m_min_os_versions.push_back(xxxx); 5347 m_min_os_versions.push_back(yy); 5348 m_min_os_versions.push_back(zz); 5349 } 5350 success = true; 5351 } 5352 } 5353 offset = load_cmd_offset + lc.cmdsize; 5354 } 5355 5356 if (success == false) 5357 { 5358 // Push an invalid value so we don't keep trying to 5359 m_min_os_versions.push_back(UINT32_MAX); 5360 } 5361 } 5362 5363 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) 5364 { 5365 if (versions != NULL && num_versions > 0) 5366 { 5367 for (size_t i=0; i<num_versions; ++i) 5368 { 5369 if (i < m_min_os_versions.size()) 5370 versions[i] = m_min_os_versions[i]; 5371 else 5372 versions[i] = 0; 5373 } 5374 } 5375 return m_min_os_versions.size(); 5376 } 5377 // Call the superclasses version that will empty out the data 5378 return ObjectFile::GetMinimumOSVersion (versions, num_versions); 5379 } 5380 5381 uint32_t 5382 ObjectFileMachO::GetSDKVersion(uint32_t *versions, uint32_t num_versions) 5383 { 5384 if (m_sdk_versions.empty()) 5385 { 5386 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5387 bool success = false; 5388 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 5389 { 5390 const lldb::offset_t load_cmd_offset = offset; 5391 5392 version_min_command lc; 5393 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5394 break; 5395 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 5396 { 5397 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 5398 { 5399 const uint32_t xxxx = lc.sdk >> 16; 5400 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5401 const uint32_t zz = lc.sdk & 0xffu; 5402 if (xxxx) 5403 { 5404 m_sdk_versions.push_back(xxxx); 5405 m_sdk_versions.push_back(yy); 5406 m_sdk_versions.push_back(zz); 5407 } 5408 success = true; 5409 } 5410 } 5411 offset = load_cmd_offset + lc.cmdsize; 5412 } 5413 5414 if (success == false) 5415 { 5416 // Push an invalid value so we don't keep trying to 5417 m_sdk_versions.push_back(UINT32_MAX); 5418 } 5419 } 5420 5421 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) 5422 { 5423 if (versions != NULL && num_versions > 0) 5424 { 5425 for (size_t i=0; i<num_versions; ++i) 5426 { 5427 if (i < m_sdk_versions.size()) 5428 versions[i] = m_sdk_versions[i]; 5429 else 5430 versions[i] = 0; 5431 } 5432 } 5433 return m_sdk_versions.size(); 5434 } 5435 // Call the superclasses version that will empty out the data 5436 return ObjectFile::GetSDKVersion (versions, num_versions); 5437 } 5438 5439 5440 bool 5441 ObjectFileMachO::GetIsDynamicLinkEditor() 5442 { 5443 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5444 } 5445 5446 //------------------------------------------------------------------ 5447 // PluginInterface protocol 5448 //------------------------------------------------------------------ 5449 lldb_private::ConstString 5450 ObjectFileMachO::GetPluginName() 5451 { 5452 return GetPluginNameStatic(); 5453 } 5454 5455 uint32_t 5456 ObjectFileMachO::GetPluginVersion() 5457 { 5458 return 1; 5459 } 5460 5461 5462 bool 5463 ObjectFileMachO::SetLoadAddress (Target &target, 5464 lldb::addr_t value, 5465 bool value_is_offset) 5466 { 5467 ModuleSP module_sp = GetModule(); 5468 if (module_sp) 5469 { 5470 size_t num_loaded_sections = 0; 5471 SectionList *section_list = GetSectionList (); 5472 if (section_list) 5473 { 5474 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5475 const size_t num_sections = section_list->GetSize(); 5476 5477 const bool is_memory_image = (bool)m_process_wp.lock(); 5478 const Strata strata = GetStrata(); 5479 static ConstString g_linkedit_segname ("__LINKEDIT"); 5480 if (value_is_offset) 5481 { 5482 // "value" is an offset to apply to each top level segment 5483 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) 5484 { 5485 // Iterate through the object file sections to find all 5486 // of the sections that size on disk (to avoid __PAGEZERO) 5487 // and load them 5488 SectionSP section_sp (section_list->GetSectionAtIndex (sect_idx)); 5489 if (section_sp && 5490 section_sp->GetFileSize() > 0 && 5491 section_sp->IsThreadSpecific() == false && 5492 module_sp.get() == section_sp->GetModule().get()) 5493 { 5494 // Ignore __LINKEDIT and __DWARF segments 5495 if (section_sp->GetName() == g_linkedit_segname) 5496 { 5497 // Only map __LINKEDIT if we have an in memory image and this isn't 5498 // a kernel binary like a kext or mach_kernel. 5499 if (is_memory_image == false || strata == eStrataKernel) 5500 continue; 5501 } 5502 if (target.GetSectionLoadList().SetSectionLoadAddress (section_sp, section_sp->GetFileAddress() + value)) 5503 ++num_loaded_sections; 5504 } 5505 } 5506 } 5507 else 5508 { 5509 // "value" is the new base address of the mach_header, adjust each 5510 // section accordingly 5511 5512 // First find the address of the mach header which is the first non-zero 5513 // file sized section whose file offset is zero as this will be subtracted 5514 // from each other valid section's vmaddr and then get "base_addr" added to 5515 // it when loading the module in the target 5516 for (size_t sect_idx = 0; 5517 sect_idx < num_sections && mach_base_file_addr == LLDB_INVALID_ADDRESS; 5518 ++sect_idx) 5519 { 5520 // Iterate through the object file sections to find all 5521 // of the sections that size on disk (to avoid __PAGEZERO) 5522 // and load them 5523 Section *section = section_list->GetSectionAtIndex (sect_idx).get(); 5524 if (section && 5525 section->GetFileSize() > 0 && 5526 section->GetFileOffset() == 0 && 5527 section->IsThreadSpecific() == false && 5528 module_sp.get() == section->GetModule().get()) 5529 { 5530 // Ignore __LINKEDIT and __DWARF segments 5531 if (section->GetName() == g_linkedit_segname) 5532 { 5533 // Only map __LINKEDIT if we have an in memory image and this isn't 5534 // a kernel binary like a kext or mach_kernel. 5535 if (is_memory_image == false || strata == eStrataKernel) 5536 continue; 5537 } 5538 mach_base_file_addr = section->GetFileAddress(); 5539 } 5540 } 5541 5542 if (mach_base_file_addr != LLDB_INVALID_ADDRESS) 5543 { 5544 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) 5545 { 5546 // Iterate through the object file sections to find all 5547 // of the sections that size on disk (to avoid __PAGEZERO) 5548 // and load them 5549 SectionSP section_sp (section_list->GetSectionAtIndex (sect_idx)); 5550 if (section_sp && 5551 section_sp->GetFileSize() > 0 && 5552 section_sp->IsThreadSpecific() == false && 5553 module_sp.get() == section_sp->GetModule().get()) 5554 { 5555 // Ignore __LINKEDIT and __DWARF segments 5556 if (section_sp->GetName() == g_linkedit_segname) 5557 { 5558 // Only map __LINKEDIT if we have an in memory image and this isn't 5559 // a kernel binary like a kext or mach_kernel. 5560 if (is_memory_image == false || strata == eStrataKernel) 5561 continue; 5562 } 5563 if (target.GetSectionLoadList().SetSectionLoadAddress (section_sp, section_sp->GetFileAddress() - mach_base_file_addr + value)) 5564 ++num_loaded_sections; 5565 } 5566 } 5567 } 5568 } 5569 } 5570 return num_loaded_sections > 0; 5571 } 5572 return false; 5573 } 5574 5575 bool 5576 ObjectFileMachO::SaveCore (const lldb::ProcessSP &process_sp, 5577 const FileSpec &outfile, 5578 Error &error) 5579 { 5580 if (process_sp) 5581 { 5582 Target &target = process_sp->GetTarget(); 5583 const ArchSpec target_arch = target.GetArchitecture(); 5584 const llvm::Triple &target_triple = target_arch.GetTriple(); 5585 if (target_triple.getVendor() == llvm::Triple::Apple && 5586 (target_triple.getOS() == llvm::Triple::MacOSX || 5587 target_triple.getOS() == llvm::Triple::IOS)) 5588 { 5589 bool make_core = false; 5590 switch (target_arch.GetMachine()) 5591 { 5592 // arm64 core file writing is having some problem with writing down the 5593 // dyld shared images info struct and/or the main executable binary. May 5594 // turn out to be a debugserver problem, not sure yet. 5595 // case llvm::Triple::aarch64: 5596 5597 case llvm::Triple::arm: 5598 case llvm::Triple::x86: 5599 case llvm::Triple::x86_64: 5600 make_core = true; 5601 break; 5602 default: 5603 error.SetErrorStringWithFormat ("unsupported core architecture: %s", target_triple.str().c_str()); 5604 break; 5605 } 5606 5607 if (make_core) 5608 { 5609 std::vector<segment_command_64> segment_load_commands; 5610 // uint32_t range_info_idx = 0; 5611 MemoryRegionInfo range_info; 5612 Error range_error = process_sp->GetMemoryRegionInfo(0, range_info); 5613 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 5614 const ByteOrder byte_order = target_arch.GetByteOrder(); 5615 if (range_error.Success()) 5616 { 5617 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) 5618 { 5619 const addr_t addr = range_info.GetRange().GetRangeBase(); 5620 const addr_t size = range_info.GetRange().GetByteSize(); 5621 5622 if (size == 0) 5623 break; 5624 5625 // Calculate correct protections 5626 uint32_t prot = 0; 5627 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 5628 prot |= VM_PROT_READ; 5629 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 5630 prot |= VM_PROT_WRITE; 5631 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 5632 prot |= VM_PROT_EXECUTE; 5633 5634 // printf ("[%3u] [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 ") %c%c%c\n", 5635 // range_info_idx, 5636 // addr, 5637 // size, 5638 // (prot & VM_PROT_READ ) ? 'r' : '-', 5639 // (prot & VM_PROT_WRITE ) ? 'w' : '-', 5640 // (prot & VM_PROT_EXECUTE) ? 'x' : '-'); 5641 5642 if (prot != 0) 5643 { 5644 uint32_t cmd_type = LC_SEGMENT_64; 5645 uint32_t segment_size = sizeof (segment_command_64); 5646 if (addr_byte_size == 4) 5647 { 5648 cmd_type = LC_SEGMENT; 5649 segment_size = sizeof (segment_command); 5650 } 5651 segment_command_64 segment = { 5652 cmd_type, // uint32_t cmd; 5653 segment_size, // uint32_t cmdsize; 5654 {0}, // char segname[16]; 5655 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 5656 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 5657 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 5658 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 5659 prot, // uint32_t maxprot; 5660 prot, // uint32_t initprot; 5661 0, // uint32_t nsects; 5662 0 }; // uint32_t flags; 5663 segment_load_commands.push_back(segment); 5664 } 5665 else 5666 { 5667 // No protections and a size of 1 used to be returned from old 5668 // debugservers when we asked about a region that was past the 5669 // last memory region and it indicates the end... 5670 if (size == 1) 5671 break; 5672 } 5673 5674 range_error = process_sp->GetMemoryRegionInfo(range_info.GetRange().GetRangeEnd(), range_info); 5675 if (range_error.Fail()) 5676 break; 5677 } 5678 5679 StreamString buffer (Stream::eBinary, 5680 addr_byte_size, 5681 byte_order); 5682 5683 mach_header_64 mach_header; 5684 if (addr_byte_size == 8) 5685 { 5686 mach_header.magic = MH_MAGIC_64; 5687 } 5688 else 5689 { 5690 mach_header.magic = MH_MAGIC; 5691 } 5692 mach_header.cputype = target_arch.GetMachOCPUType(); 5693 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 5694 mach_header.filetype = MH_CORE; 5695 mach_header.ncmds = segment_load_commands.size(); 5696 mach_header.flags = 0; 5697 mach_header.reserved = 0; 5698 ThreadList &thread_list = process_sp->GetThreadList(); 5699 const uint32_t num_threads = thread_list.GetSize(); 5700 5701 // Make an array of LC_THREAD data items. Each one contains 5702 // the contents of the LC_THREAD load command. The data doesn't 5703 // contain the load command + load command size, we will 5704 // add the load command and load command size as we emit the data. 5705 std::vector<StreamString> LC_THREAD_datas(num_threads); 5706 for (auto &LC_THREAD_data : LC_THREAD_datas) 5707 { 5708 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 5709 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 5710 LC_THREAD_data.SetByteOrder(byte_order); 5711 } 5712 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) 5713 { 5714 ThreadSP thread_sp (thread_list.GetThreadAtIndex(thread_idx)); 5715 if (thread_sp) 5716 { 5717 switch (mach_header.cputype) 5718 { 5719 case llvm::MachO::CPU_TYPE_ARM64: 5720 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5721 break; 5722 5723 case llvm::MachO::CPU_TYPE_ARM: 5724 RegisterContextDarwin_arm_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5725 break; 5726 5727 case llvm::MachO::CPU_TYPE_I386: 5728 RegisterContextDarwin_i386_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5729 break; 5730 5731 case llvm::MachO::CPU_TYPE_X86_64: 5732 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5733 break; 5734 } 5735 5736 } 5737 } 5738 5739 // The size of the load command is the size of the segments... 5740 if (addr_byte_size == 8) 5741 { 5742 mach_header.sizeofcmds = segment_load_commands.size() * sizeof (struct segment_command_64); 5743 } 5744 else 5745 { 5746 mach_header.sizeofcmds = segment_load_commands.size() * sizeof (struct segment_command); 5747 } 5748 5749 // and the size of all LC_THREAD load command 5750 for (const auto &LC_THREAD_data : LC_THREAD_datas) 5751 { 5752 ++mach_header.ncmds; 5753 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 5754 } 5755 5756 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", 5757 mach_header.magic, 5758 mach_header.cputype, 5759 mach_header.cpusubtype, 5760 mach_header.filetype, 5761 mach_header.ncmds, 5762 mach_header.sizeofcmds, 5763 mach_header.flags, 5764 mach_header.reserved); 5765 5766 // Write the mach header 5767 buffer.PutHex32(mach_header.magic); 5768 buffer.PutHex32(mach_header.cputype); 5769 buffer.PutHex32(mach_header.cpusubtype); 5770 buffer.PutHex32(mach_header.filetype); 5771 buffer.PutHex32(mach_header.ncmds); 5772 buffer.PutHex32(mach_header.sizeofcmds); 5773 buffer.PutHex32(mach_header.flags); 5774 if (addr_byte_size == 8) 5775 { 5776 buffer.PutHex32(mach_header.reserved); 5777 } 5778 5779 // Skip the mach header and all load commands and align to the next 5780 // 0x1000 byte boundary 5781 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 5782 if (file_offset & 0x00000fff) 5783 { 5784 file_offset += 0x00001000ull; 5785 file_offset &= (~0x00001000ull + 1); 5786 } 5787 5788 for (auto &segment : segment_load_commands) 5789 { 5790 segment.fileoff = file_offset; 5791 file_offset += segment.filesize; 5792 } 5793 5794 // Write out all of the LC_THREAD load commands 5795 for (const auto &LC_THREAD_data : LC_THREAD_datas) 5796 { 5797 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 5798 buffer.PutHex32(LC_THREAD); 5799 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 5800 buffer.Write(LC_THREAD_data.GetData(), LC_THREAD_data_size); 5801 } 5802 5803 // Write out all of the segment load commands 5804 for (const auto &segment : segment_load_commands) 5805 { 5806 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", 5807 segment.cmd, 5808 segment.cmdsize, 5809 segment.vmaddr, 5810 segment.vmaddr + segment.vmsize, 5811 segment.fileoff, 5812 segment.filesize, 5813 segment.maxprot, 5814 segment.initprot, 5815 segment.nsects, 5816 segment.flags); 5817 5818 buffer.PutHex32(segment.cmd); 5819 buffer.PutHex32(segment.cmdsize); 5820 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 5821 if (addr_byte_size == 8) 5822 { 5823 buffer.PutHex64(segment.vmaddr); 5824 buffer.PutHex64(segment.vmsize); 5825 buffer.PutHex64(segment.fileoff); 5826 buffer.PutHex64(segment.filesize); 5827 } 5828 else 5829 { 5830 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 5831 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 5832 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 5833 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 5834 } 5835 buffer.PutHex32(segment.maxprot); 5836 buffer.PutHex32(segment.initprot); 5837 buffer.PutHex32(segment.nsects); 5838 buffer.PutHex32(segment.flags); 5839 } 5840 5841 File core_file; 5842 std::string core_file_path(outfile.GetPath()); 5843 error = core_file.Open(core_file_path.c_str(), 5844 File::eOpenOptionWrite | 5845 File::eOpenOptionTruncate | 5846 File::eOpenOptionCanCreate); 5847 if (error.Success()) 5848 { 5849 // Read 1 page at a time 5850 uint8_t bytes[0x1000]; 5851 // Write the mach header and load commands out to the core file 5852 size_t bytes_written = buffer.GetString().size(); 5853 error = core_file.Write(buffer.GetString().data(), bytes_written); 5854 if (error.Success()) 5855 { 5856 // Now write the file data for all memory segments in the process 5857 for (const auto &segment : segment_load_commands) 5858 { 5859 if (core_file.SeekFromStart(segment.fileoff) == -1) 5860 { 5861 error.SetErrorStringWithFormat("unable to seek to offset 0x%" PRIx64 " in '%s'", segment.fileoff, core_file_path.c_str()); 5862 break; 5863 } 5864 5865 printf ("Saving %" PRId64 " bytes of data for memory region at 0x%" PRIx64 "\n", segment.vmsize, segment.vmaddr); 5866 addr_t bytes_left = segment.vmsize; 5867 addr_t addr = segment.vmaddr; 5868 Error memory_read_error; 5869 while (bytes_left > 0 && error.Success()) 5870 { 5871 const size_t bytes_to_read = bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 5872 const size_t bytes_read = process_sp->ReadMemory(addr, bytes, bytes_to_read, memory_read_error); 5873 if (bytes_read == bytes_to_read) 5874 { 5875 size_t bytes_written = bytes_read; 5876 error = core_file.Write(bytes, bytes_written); 5877 bytes_left -= bytes_read; 5878 addr += bytes_read; 5879 } 5880 else 5881 { 5882 // Some pages within regions are not readable, those 5883 // should be zero filled 5884 memset (bytes, 0, bytes_to_read); 5885 size_t bytes_written = bytes_to_read; 5886 error = core_file.Write(bytes, bytes_written); 5887 bytes_left -= bytes_to_read; 5888 addr += bytes_to_read; 5889 } 5890 } 5891 } 5892 } 5893 } 5894 } 5895 else 5896 { 5897 error.SetErrorString("process doesn't support getting memory region info"); 5898 } 5899 } 5900 return true; // This is the right plug to handle saving core files for this process 5901 } 5902 } 5903 return false; 5904 } 5905 5906