1 //===-- ObjectFileMachO.cpp -------------------------------------*- C++ -*-===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 10 #include "llvm/ADT/StringRef.h" 11 12 #include "lldb/lldb-private-log.h" 13 #include "lldb/Core/ArchSpec.h" 14 #include "lldb/Core/DataBuffer.h" 15 #include "lldb/Core/Debugger.h" 16 #include "lldb/Core/Error.h" 17 #include "lldb/Core/FileSpecList.h" 18 #include "lldb/Core/Log.h" 19 #include "lldb/Core/Module.h" 20 #include "lldb/Core/ModuleSpec.h" 21 #include "lldb/Core/PluginManager.h" 22 #include "lldb/Core/RangeMap.h" 23 #include "lldb/Core/Section.h" 24 #include "lldb/Core/StreamFile.h" 25 #include "lldb/Core/StreamString.h" 26 #include "lldb/Core/Timer.h" 27 #include "lldb/Core/UUID.h" 28 #include "lldb/Host/Host.h" 29 #include "lldb/Host/FileSpec.h" 30 #include "lldb/Symbol/ClangNamespaceDecl.h" 31 #include "lldb/Symbol/DWARFCallFrameInfo.h" 32 #include "lldb/Symbol/ObjectFile.h" 33 #include "lldb/Target/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->MemoryMapFileContents(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->MemoryMapFileContents(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); 997 998 if (GetArchitecture (header, data, data_offset, spec.GetArchitecture())) 999 { 1000 if (spec.GetArchitecture().IsValid()) 1001 { 1002 GetUUID (header, data, data_offset, spec.GetUUID()); 1003 specs.Append(spec); 1004 } 1005 } 1006 } 1007 } 1008 } 1009 return specs.GetSize() - initial_count; 1010 } 1011 1012 1013 1014 const ConstString & 1015 ObjectFileMachO::GetSegmentNameTEXT() 1016 { 1017 static ConstString g_segment_name_TEXT ("__TEXT"); 1018 return g_segment_name_TEXT; 1019 } 1020 1021 const ConstString & 1022 ObjectFileMachO::GetSegmentNameDATA() 1023 { 1024 static ConstString g_segment_name_DATA ("__DATA"); 1025 return g_segment_name_DATA; 1026 } 1027 1028 const ConstString & 1029 ObjectFileMachO::GetSegmentNameOBJC() 1030 { 1031 static ConstString g_segment_name_OBJC ("__OBJC"); 1032 return g_segment_name_OBJC; 1033 } 1034 1035 const ConstString & 1036 ObjectFileMachO::GetSegmentNameLINKEDIT() 1037 { 1038 static ConstString g_section_name_LINKEDIT ("__LINKEDIT"); 1039 return g_section_name_LINKEDIT; 1040 } 1041 1042 const ConstString & 1043 ObjectFileMachO::GetSectionNameEHFrame() 1044 { 1045 static ConstString g_section_name_eh_frame ("__eh_frame"); 1046 return g_section_name_eh_frame; 1047 } 1048 1049 bool 1050 ObjectFileMachO::MagicBytesMatch (DataBufferSP& data_sp, 1051 lldb::addr_t data_offset, 1052 lldb::addr_t data_length) 1053 { 1054 DataExtractor data; 1055 data.SetData (data_sp, data_offset, data_length); 1056 lldb::offset_t offset = 0; 1057 uint32_t magic = data.GetU32(&offset); 1058 return MachHeaderSizeFromMagic(magic) != 0; 1059 } 1060 1061 1062 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 1063 DataBufferSP& data_sp, 1064 lldb::offset_t data_offset, 1065 const FileSpec* file, 1066 lldb::offset_t file_offset, 1067 lldb::offset_t length) : 1068 ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 1069 m_mach_segments(), 1070 m_mach_sections(), 1071 m_entry_point_address(), 1072 m_thread_context_offsets(), 1073 m_thread_context_offsets_valid(false) 1074 { 1075 ::memset (&m_header, 0, sizeof(m_header)); 1076 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 1077 } 1078 1079 ObjectFileMachO::ObjectFileMachO (const lldb::ModuleSP &module_sp, 1080 lldb::DataBufferSP& header_data_sp, 1081 const lldb::ProcessSP &process_sp, 1082 lldb::addr_t header_addr) : 1083 ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 1084 m_mach_segments(), 1085 m_mach_sections(), 1086 m_entry_point_address(), 1087 m_thread_context_offsets(), 1088 m_thread_context_offsets_valid(false) 1089 { 1090 ::memset (&m_header, 0, sizeof(m_header)); 1091 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 1092 } 1093 1094 ObjectFileMachO::~ObjectFileMachO() 1095 { 1096 } 1097 1098 bool 1099 ObjectFileMachO::ParseHeader (DataExtractor &data, 1100 lldb::offset_t *data_offset_ptr, 1101 llvm::MachO::mach_header &header) 1102 { 1103 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1104 // Leave magic in the original byte order 1105 header.magic = data.GetU32(data_offset_ptr); 1106 bool can_parse = false; 1107 bool is_64_bit = false; 1108 switch (header.magic) 1109 { 1110 case MH_MAGIC: 1111 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1112 data.SetAddressByteSize(4); 1113 can_parse = true; 1114 break; 1115 1116 case MH_MAGIC_64: 1117 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1118 data.SetAddressByteSize(8); 1119 can_parse = true; 1120 is_64_bit = true; 1121 break; 1122 1123 case MH_CIGAM: 1124 data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 1125 data.SetAddressByteSize(4); 1126 can_parse = true; 1127 break; 1128 1129 case MH_CIGAM_64: 1130 data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 1131 data.SetAddressByteSize(8); 1132 is_64_bit = true; 1133 can_parse = true; 1134 break; 1135 1136 default: 1137 break; 1138 } 1139 1140 if (can_parse) 1141 { 1142 data.GetU32(data_offset_ptr, &header.cputype, 6); 1143 if (is_64_bit) 1144 *data_offset_ptr += 4; 1145 return true; 1146 } 1147 else 1148 { 1149 memset(&header, 0, sizeof(header)); 1150 } 1151 return false; 1152 } 1153 1154 bool 1155 ObjectFileMachO::ParseHeader () 1156 { 1157 ModuleSP module_sp(GetModule()); 1158 if (module_sp) 1159 { 1160 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 1161 bool can_parse = false; 1162 lldb::offset_t offset = 0; 1163 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1164 // Leave magic in the original byte order 1165 m_header.magic = m_data.GetU32(&offset); 1166 switch (m_header.magic) 1167 { 1168 case MH_MAGIC: 1169 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1170 m_data.SetAddressByteSize(4); 1171 can_parse = true; 1172 break; 1173 1174 case MH_MAGIC_64: 1175 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1176 m_data.SetAddressByteSize(8); 1177 can_parse = true; 1178 break; 1179 1180 case MH_CIGAM: 1181 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 1182 m_data.SetAddressByteSize(4); 1183 can_parse = true; 1184 break; 1185 1186 case MH_CIGAM_64: 1187 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 1188 m_data.SetAddressByteSize(8); 1189 can_parse = true; 1190 break; 1191 1192 default: 1193 break; 1194 } 1195 1196 if (can_parse) 1197 { 1198 m_data.GetU32(&offset, &m_header.cputype, 6); 1199 1200 1201 ArchSpec mach_arch; 1202 1203 if (GetArchitecture (mach_arch)) 1204 { 1205 // Check if the module has a required architecture 1206 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1207 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1208 return false; 1209 1210 if (SetModulesArchitecture (mach_arch)) 1211 { 1212 const size_t header_and_lc_size = m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1213 if (m_data.GetByteSize() < header_and_lc_size) 1214 { 1215 DataBufferSP data_sp; 1216 ProcessSP process_sp (m_process_wp.lock()); 1217 if (process_sp) 1218 { 1219 data_sp = ReadMemory (process_sp, m_memory_addr, header_and_lc_size); 1220 } 1221 else 1222 { 1223 // Read in all only the load command data from the file on disk 1224 data_sp = m_file.ReadFileContents(m_file_offset, header_and_lc_size); 1225 if (data_sp->GetByteSize() != header_and_lc_size) 1226 return false; 1227 } 1228 if (data_sp) 1229 m_data.SetData (data_sp); 1230 } 1231 } 1232 return true; 1233 } 1234 } 1235 else 1236 { 1237 memset(&m_header, 0, sizeof(struct mach_header)); 1238 } 1239 } 1240 return false; 1241 } 1242 1243 1244 ByteOrder 1245 ObjectFileMachO::GetByteOrder () const 1246 { 1247 return m_data.GetByteOrder (); 1248 } 1249 1250 bool 1251 ObjectFileMachO::IsExecutable() const 1252 { 1253 return m_header.filetype == MH_EXECUTE; 1254 } 1255 1256 uint32_t 1257 ObjectFileMachO::GetAddressByteSize () const 1258 { 1259 return m_data.GetAddressByteSize (); 1260 } 1261 1262 AddressClass 1263 ObjectFileMachO::GetAddressClass (lldb::addr_t file_addr) 1264 { 1265 Symtab *symtab = GetSymtab(); 1266 if (symtab) 1267 { 1268 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1269 if (symbol) 1270 { 1271 if (symbol->ValueIsAddress()) 1272 { 1273 SectionSP section_sp (symbol->GetAddress().GetSection()); 1274 if (section_sp) 1275 { 1276 const lldb::SectionType section_type = section_sp->GetType(); 1277 switch (section_type) 1278 { 1279 case eSectionTypeInvalid: 1280 return eAddressClassUnknown; 1281 1282 case eSectionTypeCode: 1283 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) 1284 { 1285 // For ARM we have a bit in the n_desc field of the symbol 1286 // that tells us ARM/Thumb which is bit 0x0008. 1287 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1288 return eAddressClassCodeAlternateISA; 1289 } 1290 return eAddressClassCode; 1291 1292 case eSectionTypeContainer: 1293 return eAddressClassUnknown; 1294 1295 case eSectionTypeData: 1296 case eSectionTypeDataCString: 1297 case eSectionTypeDataCStringPointers: 1298 case eSectionTypeDataSymbolAddress: 1299 case eSectionTypeData4: 1300 case eSectionTypeData8: 1301 case eSectionTypeData16: 1302 case eSectionTypeDataPointers: 1303 case eSectionTypeZeroFill: 1304 case eSectionTypeDataObjCMessageRefs: 1305 case eSectionTypeDataObjCCFStrings: 1306 return eAddressClassData; 1307 1308 case eSectionTypeDebug: 1309 case eSectionTypeDWARFDebugAbbrev: 1310 case eSectionTypeDWARFDebugAranges: 1311 case eSectionTypeDWARFDebugFrame: 1312 case eSectionTypeDWARFDebugInfo: 1313 case eSectionTypeDWARFDebugLine: 1314 case eSectionTypeDWARFDebugLoc: 1315 case eSectionTypeDWARFDebugMacInfo: 1316 case eSectionTypeDWARFDebugPubNames: 1317 case eSectionTypeDWARFDebugPubTypes: 1318 case eSectionTypeDWARFDebugRanges: 1319 case eSectionTypeDWARFDebugStr: 1320 case eSectionTypeDWARFAppleNames: 1321 case eSectionTypeDWARFAppleTypes: 1322 case eSectionTypeDWARFAppleNamespaces: 1323 case eSectionTypeDWARFAppleObjC: 1324 return eAddressClassDebug; 1325 1326 case eSectionTypeEHFrame: 1327 case eSectionTypeCompactUnwind: 1328 return eAddressClassRuntime; 1329 1330 case eSectionTypeELFSymbolTable: 1331 case eSectionTypeELFDynamicSymbols: 1332 case eSectionTypeELFRelocationEntries: 1333 case eSectionTypeELFDynamicLinkInfo: 1334 case eSectionTypeOther: 1335 return eAddressClassUnknown; 1336 } 1337 } 1338 } 1339 1340 const SymbolType symbol_type = symbol->GetType(); 1341 switch (symbol_type) 1342 { 1343 case eSymbolTypeAny: return eAddressClassUnknown; 1344 case eSymbolTypeAbsolute: return eAddressClassUnknown; 1345 1346 case eSymbolTypeCode: 1347 case eSymbolTypeTrampoline: 1348 case eSymbolTypeResolver: 1349 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) 1350 { 1351 // For ARM we have a bit in the n_desc field of the symbol 1352 // that tells us ARM/Thumb which is bit 0x0008. 1353 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1354 return eAddressClassCodeAlternateISA; 1355 } 1356 return eAddressClassCode; 1357 1358 case eSymbolTypeData: return eAddressClassData; 1359 case eSymbolTypeRuntime: return eAddressClassRuntime; 1360 case eSymbolTypeException: return eAddressClassRuntime; 1361 case eSymbolTypeSourceFile: return eAddressClassDebug; 1362 case eSymbolTypeHeaderFile: return eAddressClassDebug; 1363 case eSymbolTypeObjectFile: return eAddressClassDebug; 1364 case eSymbolTypeCommonBlock: return eAddressClassDebug; 1365 case eSymbolTypeBlock: return eAddressClassDebug; 1366 case eSymbolTypeLocal: return eAddressClassData; 1367 case eSymbolTypeParam: return eAddressClassData; 1368 case eSymbolTypeVariable: return eAddressClassData; 1369 case eSymbolTypeVariableType: return eAddressClassDebug; 1370 case eSymbolTypeLineEntry: return eAddressClassDebug; 1371 case eSymbolTypeLineHeader: return eAddressClassDebug; 1372 case eSymbolTypeScopeBegin: return eAddressClassDebug; 1373 case eSymbolTypeScopeEnd: return eAddressClassDebug; 1374 case eSymbolTypeAdditional: return eAddressClassUnknown; 1375 case eSymbolTypeCompiler: return eAddressClassDebug; 1376 case eSymbolTypeInstrumentation:return eAddressClassDebug; 1377 case eSymbolTypeUndefined: return eAddressClassUnknown; 1378 case eSymbolTypeObjCClass: return eAddressClassRuntime; 1379 case eSymbolTypeObjCMetaClass: return eAddressClassRuntime; 1380 case eSymbolTypeObjCIVar: return eAddressClassRuntime; 1381 case eSymbolTypeReExported: return eAddressClassRuntime; 1382 } 1383 } 1384 } 1385 return eAddressClassUnknown; 1386 } 1387 1388 Symtab * 1389 ObjectFileMachO::GetSymtab() 1390 { 1391 ModuleSP module_sp(GetModule()); 1392 if (module_sp) 1393 { 1394 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 1395 if (m_symtab_ap.get() == NULL) 1396 { 1397 m_symtab_ap.reset(new Symtab(this)); 1398 Mutex::Locker symtab_locker (m_symtab_ap->GetMutex()); 1399 ParseSymtab (); 1400 m_symtab_ap->Finalize (); 1401 } 1402 } 1403 return m_symtab_ap.get(); 1404 } 1405 1406 bool 1407 ObjectFileMachO::IsStripped () 1408 { 1409 if (m_dysymtab.cmd == 0) 1410 { 1411 ModuleSP module_sp(GetModule()); 1412 if (module_sp) 1413 { 1414 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1415 for (uint32_t i=0; i<m_header.ncmds; ++i) 1416 { 1417 const lldb::offset_t load_cmd_offset = offset; 1418 1419 load_command lc; 1420 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1421 break; 1422 if (lc.cmd == LC_DYSYMTAB) 1423 { 1424 m_dysymtab.cmd = lc.cmd; 1425 m_dysymtab.cmdsize = lc.cmdsize; 1426 if (m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == NULL) 1427 { 1428 // Clear m_dysymtab if we were unable to read all items from the load command 1429 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 1430 } 1431 } 1432 offset = load_cmd_offset + lc.cmdsize; 1433 } 1434 } 1435 } 1436 if (m_dysymtab.cmd) 1437 return m_dysymtab.nlocalsym <= 1; 1438 return false; 1439 } 1440 1441 void 1442 ObjectFileMachO::CreateSections (SectionList &unified_section_list) 1443 { 1444 if (!m_sections_ap.get()) 1445 { 1446 m_sections_ap.reset(new SectionList()); 1447 1448 const bool is_dsym = (m_header.filetype == MH_DSYM); 1449 lldb::user_id_t segID = 0; 1450 lldb::user_id_t sectID = 0; 1451 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1452 uint32_t i; 1453 const bool is_core = GetType() == eTypeCoreFile; 1454 //bool dump_sections = false; 1455 ModuleSP module_sp (GetModule()); 1456 // First look up any LC_ENCRYPTION_INFO load commands 1457 typedef RangeArray<uint32_t, uint32_t, 8> EncryptedFileRanges; 1458 EncryptedFileRanges encrypted_file_ranges; 1459 encryption_info_command encryption_cmd; 1460 for (i=0; i<m_header.ncmds; ++i) 1461 { 1462 const lldb::offset_t load_cmd_offset = offset; 1463 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1464 break; 1465 1466 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO) 1467 { 1468 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) 1469 { 1470 if (encryption_cmd.cryptid != 0) 1471 { 1472 EncryptedFileRanges::Entry entry; 1473 entry.SetRangeBase(encryption_cmd.cryptoff); 1474 entry.SetByteSize(encryption_cmd.cryptsize); 1475 encrypted_file_ranges.Append(entry); 1476 } 1477 } 1478 } 1479 offset = load_cmd_offset + encryption_cmd.cmdsize; 1480 } 1481 1482 bool section_file_addresses_changed = false; 1483 1484 offset = MachHeaderSizeFromMagic(m_header.magic); 1485 1486 struct segment_command_64 load_cmd; 1487 for (i=0; i<m_header.ncmds; ++i) 1488 { 1489 const lldb::offset_t load_cmd_offset = offset; 1490 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1491 break; 1492 1493 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1494 { 1495 if (m_data.GetU8(&offset, (uint8_t*)load_cmd.segname, 16)) 1496 { 1497 bool add_section = true; 1498 bool add_to_unified = true; 1499 ConstString const_segname (load_cmd.segname, std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 1500 1501 SectionSP unified_section_sp(unified_section_list.FindSectionByName(const_segname)); 1502 if (is_dsym && unified_section_sp) 1503 { 1504 if (const_segname == GetSegmentNameLINKEDIT()) 1505 { 1506 // We need to keep the __LINKEDIT segment private to this object file only 1507 add_to_unified = false; 1508 } 1509 else 1510 { 1511 // This is the dSYM file and this section has already been created by 1512 // the object file, no need to create it. 1513 add_section = false; 1514 } 1515 } 1516 load_cmd.vmaddr = m_data.GetAddress(&offset); 1517 load_cmd.vmsize = m_data.GetAddress(&offset); 1518 load_cmd.fileoff = m_data.GetAddress(&offset); 1519 load_cmd.filesize = m_data.GetAddress(&offset); 1520 if (m_length != 0 && load_cmd.filesize != 0) 1521 { 1522 if (load_cmd.fileoff > m_length) 1523 { 1524 // We have a load command that says it extends past the end of the file. This is likely 1525 // a corrupt file. We don't have any way to return an error condition here (this method 1526 // was likely invoked from something like ObjectFile::GetSectionList()) -- all we can do 1527 // is null out the SectionList vector and if a process has been set up, dump a message 1528 // to stdout. The most common case here is core file debugging with a truncated file. 1529 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1530 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", 1531 i, 1532 lc_segment_name, 1533 load_cmd.fileoff, 1534 m_length); 1535 1536 load_cmd.fileoff = 0; 1537 load_cmd.filesize = 0; 1538 } 1539 1540 if (load_cmd.fileoff + load_cmd.filesize > m_length) 1541 { 1542 // We have a load command that says it extends past the end of the file. This is likely 1543 // a corrupt file. We don't have any way to return an error condition here (this method 1544 // was likely invoked from something like ObjectFile::GetSectionList()) -- all we can do 1545 // is null out the SectionList vector and if a process has been set up, dump a message 1546 // to stdout. The most common case here is core file debugging with a truncated file. 1547 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1548 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", 1549 i, 1550 lc_segment_name, 1551 load_cmd.fileoff + load_cmd.filesize, 1552 m_length); 1553 1554 // Tuncase the length 1555 load_cmd.filesize = m_length - load_cmd.fileoff; 1556 } 1557 } 1558 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1559 { 1560 1561 const bool segment_is_encrypted = (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1562 1563 // Keep a list of mach segments around in case we need to 1564 // get at data that isn't stored in the abstracted Sections. 1565 m_mach_segments.push_back (load_cmd); 1566 1567 // Use a segment ID of the segment index shifted left by 8 so they 1568 // never conflict with any of the sections. 1569 SectionSP segment_sp; 1570 if (add_section && (const_segname || is_core)) 1571 { 1572 segment_sp.reset(new Section (module_sp, // Module to which this section belongs 1573 this, // Object file to which this sections belongs 1574 ++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 1575 const_segname, // Name of this section 1576 eSectionTypeContainer, // This section is a container of other sections. 1577 load_cmd.vmaddr, // File VM address == addresses as they are found in the object file 1578 load_cmd.vmsize, // VM size in bytes of this section 1579 load_cmd.fileoff, // Offset to the data for this section in the file 1580 load_cmd.filesize, // Size in bytes of this section as found in the file 1581 0, // Segments have no alignment information 1582 load_cmd.flags)); // Flags for this section 1583 1584 segment_sp->SetIsEncrypted (segment_is_encrypted); 1585 m_sections_ap->AddSection(segment_sp); 1586 if (add_to_unified) 1587 unified_section_list.AddSection(segment_sp); 1588 } 1589 else if (unified_section_sp) 1590 { 1591 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) 1592 { 1593 // Check to see if the module was read from memory? 1594 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) 1595 { 1596 // We have a module that is in memory and needs to have its 1597 // file address adjusted. We need to do this because when we 1598 // load a file from memory, its addresses will be slid already, 1599 // yet the addresses in the new symbol file will still be unslid. 1600 // Since everything is stored as section offset, this shouldn't 1601 // cause any problems. 1602 1603 // Make sure we've parsed the symbol table from the 1604 // ObjectFile before we go around changing its Sections. 1605 module_sp->GetObjectFile()->GetSymtab(); 1606 // eh_frame would present the same problems but we parse that on 1607 // a per-function basis as-needed so it's more difficult to 1608 // remove its use of the Sections. Realistically, the environments 1609 // where this code path will be taken will not have eh_frame sections. 1610 1611 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1612 1613 // Notify the module that the section addresses have been changed once 1614 // we're done so any file-address caches can be updated. 1615 section_file_addresses_changed = true; 1616 } 1617 } 1618 m_sections_ap->AddSection(unified_section_sp); 1619 } 1620 1621 struct section_64 sect64; 1622 ::memset (§64, 0, sizeof(sect64)); 1623 // Push a section into our mach sections for the section at 1624 // index zero (NO_SECT) if we don't have any mach sections yet... 1625 if (m_mach_sections.empty()) 1626 m_mach_sections.push_back(sect64); 1627 uint32_t segment_sect_idx; 1628 const lldb::user_id_t first_segment_sectID = sectID + 1; 1629 1630 1631 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1632 for (segment_sect_idx=0; segment_sect_idx<load_cmd.nsects; ++segment_sect_idx) 1633 { 1634 if (m_data.GetU8(&offset, (uint8_t*)sect64.sectname, sizeof(sect64.sectname)) == NULL) 1635 break; 1636 if (m_data.GetU8(&offset, (uint8_t*)sect64.segname, sizeof(sect64.segname)) == NULL) 1637 break; 1638 sect64.addr = m_data.GetAddress(&offset); 1639 sect64.size = m_data.GetAddress(&offset); 1640 1641 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1642 break; 1643 1644 // Keep a list of mach sections around in case we need to 1645 // get at data that isn't stored in the abstracted Sections. 1646 m_mach_sections.push_back (sect64); 1647 1648 if (add_section) 1649 { 1650 ConstString section_name (sect64.sectname, std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1651 if (!const_segname) 1652 { 1653 // We have a segment with no name so we need to conjure up 1654 // segments that correspond to the section's segname if there 1655 // isn't already such a section. If there is such a section, 1656 // we resize the section so that it spans all sections. 1657 // We also mark these sections as fake so address matches don't 1658 // hit if they land in the gaps between the child sections. 1659 const_segname.SetTrimmedCStringWithLength(sect64.segname, sizeof(sect64.segname)); 1660 segment_sp = unified_section_list.FindSectionByName (const_segname); 1661 if (segment_sp.get()) 1662 { 1663 Section *segment = segment_sp.get(); 1664 // Grow the section size as needed. 1665 const lldb::addr_t sect64_min_addr = sect64.addr; 1666 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1667 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1668 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1669 const lldb::addr_t curr_seg_max_addr = curr_seg_min_addr + curr_seg_byte_size; 1670 if (sect64_min_addr >= curr_seg_min_addr) 1671 { 1672 const lldb::addr_t new_seg_byte_size = sect64_max_addr - curr_seg_min_addr; 1673 // Only grow the section size if needed 1674 if (new_seg_byte_size > curr_seg_byte_size) 1675 segment->SetByteSize (new_seg_byte_size); 1676 } 1677 else 1678 { 1679 // We need to change the base address of the segment and 1680 // adjust the child section offsets for all existing children. 1681 const lldb::addr_t slide_amount = sect64_min_addr - curr_seg_min_addr; 1682 segment->Slide(slide_amount, false); 1683 segment->GetChildren().Slide(-slide_amount, false); 1684 segment->SetByteSize (curr_seg_max_addr - sect64_min_addr); 1685 } 1686 1687 // Grow the section size as needed. 1688 if (sect64.offset) 1689 { 1690 const lldb::addr_t segment_min_file_offset = segment->GetFileOffset(); 1691 const lldb::addr_t segment_max_file_offset = segment_min_file_offset + segment->GetFileSize(); 1692 1693 const lldb::addr_t section_min_file_offset = sect64.offset; 1694 const lldb::addr_t section_max_file_offset = section_min_file_offset + sect64.size; 1695 const lldb::addr_t new_file_offset = std::min (section_min_file_offset, segment_min_file_offset); 1696 const lldb::addr_t new_file_size = std::max (section_max_file_offset, segment_max_file_offset) - new_file_offset; 1697 segment->SetFileOffset (new_file_offset); 1698 segment->SetFileSize (new_file_size); 1699 } 1700 } 1701 else 1702 { 1703 // Create a fake section for the section's named segment 1704 segment_sp.reset(new Section (segment_sp, // Parent section 1705 module_sp, // Module to which this section belongs 1706 this, // Object file to which this section belongs 1707 ++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 1708 const_segname, // Name of this section 1709 eSectionTypeContainer, // This section is a container of other sections. 1710 sect64.addr, // File VM address == addresses as they are found in the object file 1711 sect64.size, // VM size in bytes of this section 1712 sect64.offset, // Offset to the data for this section in the file 1713 sect64.offset ? sect64.size : 0, // Size in bytes of this section as found in the file 1714 sect64.align, 1715 load_cmd.flags)); // Flags for this section 1716 segment_sp->SetIsFake(true); 1717 1718 m_sections_ap->AddSection(segment_sp); 1719 if (add_to_unified) 1720 unified_section_list.AddSection(segment_sp); 1721 segment_sp->SetIsEncrypted (segment_is_encrypted); 1722 } 1723 } 1724 assert (segment_sp.get()); 1725 1726 lldb::SectionType sect_type = eSectionTypeOther; 1727 1728 if (sect64.flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1729 sect_type = eSectionTypeCode; 1730 else 1731 { 1732 uint32_t mach_sect_type = sect64.flags & SECTION_TYPE; 1733 static ConstString g_sect_name_objc_data ("__objc_data"); 1734 static ConstString g_sect_name_objc_msgrefs ("__objc_msgrefs"); 1735 static ConstString g_sect_name_objc_selrefs ("__objc_selrefs"); 1736 static ConstString g_sect_name_objc_classrefs ("__objc_classrefs"); 1737 static ConstString g_sect_name_objc_superrefs ("__objc_superrefs"); 1738 static ConstString g_sect_name_objc_const ("__objc_const"); 1739 static ConstString g_sect_name_objc_classlist ("__objc_classlist"); 1740 static ConstString g_sect_name_cfstring ("__cfstring"); 1741 1742 static ConstString g_sect_name_dwarf_debug_abbrev ("__debug_abbrev"); 1743 static ConstString g_sect_name_dwarf_debug_aranges ("__debug_aranges"); 1744 static ConstString g_sect_name_dwarf_debug_frame ("__debug_frame"); 1745 static ConstString g_sect_name_dwarf_debug_info ("__debug_info"); 1746 static ConstString g_sect_name_dwarf_debug_line ("__debug_line"); 1747 static ConstString g_sect_name_dwarf_debug_loc ("__debug_loc"); 1748 static ConstString g_sect_name_dwarf_debug_macinfo ("__debug_macinfo"); 1749 static ConstString g_sect_name_dwarf_debug_pubnames ("__debug_pubnames"); 1750 static ConstString g_sect_name_dwarf_debug_pubtypes ("__debug_pubtypes"); 1751 static ConstString g_sect_name_dwarf_debug_ranges ("__debug_ranges"); 1752 static ConstString g_sect_name_dwarf_debug_str ("__debug_str"); 1753 static ConstString g_sect_name_dwarf_apple_names ("__apple_names"); 1754 static ConstString g_sect_name_dwarf_apple_types ("__apple_types"); 1755 static ConstString g_sect_name_dwarf_apple_namespaces ("__apple_namespac"); 1756 static ConstString g_sect_name_dwarf_apple_objc ("__apple_objc"); 1757 static ConstString g_sect_name_eh_frame ("__eh_frame"); 1758 static ConstString g_sect_name_compact_unwind ("__unwind_info"); 1759 static ConstString g_sect_name_text ("__text"); 1760 static ConstString g_sect_name_data ("__data"); 1761 1762 1763 if (section_name == g_sect_name_dwarf_debug_abbrev) 1764 sect_type = eSectionTypeDWARFDebugAbbrev; 1765 else if (section_name == g_sect_name_dwarf_debug_aranges) 1766 sect_type = eSectionTypeDWARFDebugAranges; 1767 else if (section_name == g_sect_name_dwarf_debug_frame) 1768 sect_type = eSectionTypeDWARFDebugFrame; 1769 else if (section_name == g_sect_name_dwarf_debug_info) 1770 sect_type = eSectionTypeDWARFDebugInfo; 1771 else if (section_name == g_sect_name_dwarf_debug_line) 1772 sect_type = eSectionTypeDWARFDebugLine; 1773 else if (section_name == g_sect_name_dwarf_debug_loc) 1774 sect_type = eSectionTypeDWARFDebugLoc; 1775 else if (section_name == g_sect_name_dwarf_debug_macinfo) 1776 sect_type = eSectionTypeDWARFDebugMacInfo; 1777 else if (section_name == g_sect_name_dwarf_debug_pubnames) 1778 sect_type = eSectionTypeDWARFDebugPubNames; 1779 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 1780 sect_type = eSectionTypeDWARFDebugPubTypes; 1781 else if (section_name == g_sect_name_dwarf_debug_ranges) 1782 sect_type = eSectionTypeDWARFDebugRanges; 1783 else if (section_name == g_sect_name_dwarf_debug_str) 1784 sect_type = eSectionTypeDWARFDebugStr; 1785 else if (section_name == g_sect_name_dwarf_apple_names) 1786 sect_type = eSectionTypeDWARFAppleNames; 1787 else if (section_name == g_sect_name_dwarf_apple_types) 1788 sect_type = eSectionTypeDWARFAppleTypes; 1789 else if (section_name == g_sect_name_dwarf_apple_namespaces) 1790 sect_type = eSectionTypeDWARFAppleNamespaces; 1791 else if (section_name == g_sect_name_dwarf_apple_objc) 1792 sect_type = eSectionTypeDWARFAppleObjC; 1793 else if (section_name == g_sect_name_objc_selrefs) 1794 sect_type = eSectionTypeDataCStringPointers; 1795 else if (section_name == g_sect_name_objc_msgrefs) 1796 sect_type = eSectionTypeDataObjCMessageRefs; 1797 else if (section_name == g_sect_name_eh_frame) 1798 sect_type = eSectionTypeEHFrame; 1799 else if (section_name == g_sect_name_compact_unwind) 1800 sect_type = eSectionTypeCompactUnwind; 1801 else if (section_name == g_sect_name_cfstring) 1802 sect_type = eSectionTypeDataObjCCFStrings; 1803 else if (section_name == g_sect_name_objc_data || 1804 section_name == g_sect_name_objc_classrefs || 1805 section_name == g_sect_name_objc_superrefs || 1806 section_name == g_sect_name_objc_const || 1807 section_name == g_sect_name_objc_classlist) 1808 { 1809 sect_type = eSectionTypeDataPointers; 1810 } 1811 1812 if (sect_type == eSectionTypeOther) 1813 { 1814 switch (mach_sect_type) 1815 { 1816 // TODO: categorize sections by other flags for regular sections 1817 case S_REGULAR: 1818 if (section_name == g_sect_name_text) 1819 sect_type = eSectionTypeCode; 1820 else if (section_name == g_sect_name_data) 1821 sect_type = eSectionTypeData; 1822 else 1823 sect_type = eSectionTypeOther; 1824 break; 1825 case S_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1826 case S_CSTRING_LITERALS: sect_type = eSectionTypeDataCString; break; // section with only literal C strings 1827 case S_4BYTE_LITERALS: sect_type = eSectionTypeData4; break; // section with only 4 byte literals 1828 case S_8BYTE_LITERALS: sect_type = eSectionTypeData8; break; // section with only 8 byte literals 1829 case S_LITERAL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only pointers to literals 1830 case S_NON_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only non-lazy symbol pointers 1831 case S_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only lazy symbol pointers 1832 case S_SYMBOL_STUBS: sect_type = eSectionTypeCode; break; // section with only symbol stubs, byte size of stub in the reserved2 field 1833 case S_MOD_INIT_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for initialization 1834 case S_MOD_TERM_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for termination 1835 case S_COALESCED: sect_type = eSectionTypeOther; break; 1836 case S_GB_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1837 case S_INTERPOSING: sect_type = eSectionTypeCode; break; // section with only pairs of function pointers for interposing 1838 case S_16BYTE_LITERALS: sect_type = eSectionTypeData16; break; // section with only 16 byte literals 1839 case S_DTRACE_DOF: sect_type = eSectionTypeDebug; break; 1840 case S_LAZY_DYLIB_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; 1841 default: break; 1842 } 1843 } 1844 } 1845 1846 SectionSP section_sp(new Section (segment_sp, 1847 module_sp, 1848 this, 1849 ++sectID, 1850 section_name, 1851 sect_type, 1852 sect64.addr - segment_sp->GetFileAddress(), 1853 sect64.size, 1854 sect64.offset, 1855 sect64.offset == 0 ? 0 : sect64.size, 1856 sect64.align, 1857 sect64.flags)); 1858 // Set the section to be encrypted to match the segment 1859 1860 bool section_is_encrypted = false; 1861 if (!segment_is_encrypted && load_cmd.filesize != 0) 1862 section_is_encrypted = encrypted_file_ranges.FindEntryThatContains(sect64.offset) != NULL; 1863 1864 section_sp->SetIsEncrypted (segment_is_encrypted || section_is_encrypted); 1865 segment_sp->GetChildren().AddSection(section_sp); 1866 1867 if (segment_sp->IsFake()) 1868 { 1869 segment_sp.reset(); 1870 const_segname.Clear(); 1871 } 1872 } 1873 } 1874 if (segment_sp && is_dsym) 1875 { 1876 if (first_segment_sectID <= sectID) 1877 { 1878 lldb::user_id_t sect_uid; 1879 for (sect_uid = first_segment_sectID; sect_uid <= sectID; ++sect_uid) 1880 { 1881 SectionSP curr_section_sp(segment_sp->GetChildren().FindSectionByID (sect_uid)); 1882 SectionSP next_section_sp; 1883 if (sect_uid + 1 <= sectID) 1884 next_section_sp = segment_sp->GetChildren().FindSectionByID (sect_uid+1); 1885 1886 if (curr_section_sp.get()) 1887 { 1888 if (curr_section_sp->GetByteSize() == 0) 1889 { 1890 if (next_section_sp.get() != NULL) 1891 curr_section_sp->SetByteSize ( next_section_sp->GetFileAddress() - curr_section_sp->GetFileAddress() ); 1892 else 1893 curr_section_sp->SetByteSize ( load_cmd.vmsize ); 1894 } 1895 } 1896 } 1897 } 1898 } 1899 } 1900 } 1901 } 1902 else if (load_cmd.cmd == LC_DYSYMTAB) 1903 { 1904 m_dysymtab.cmd = load_cmd.cmd; 1905 m_dysymtab.cmdsize = load_cmd.cmdsize; 1906 m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1907 } 1908 1909 offset = load_cmd_offset + load_cmd.cmdsize; 1910 } 1911 1912 1913 if (section_file_addresses_changed && module_sp.get()) 1914 { 1915 module_sp->SectionFileAddressesChanged(); 1916 } 1917 } 1918 } 1919 1920 class MachSymtabSectionInfo 1921 { 1922 public: 1923 1924 MachSymtabSectionInfo (SectionList *section_list) : 1925 m_section_list (section_list), 1926 m_section_infos() 1927 { 1928 // Get the number of sections down to a depth of 1 to include 1929 // all segments and their sections, but no other sections that 1930 // may be added for debug map or 1931 m_section_infos.resize(section_list->GetNumSections(1)); 1932 } 1933 1934 1935 SectionSP 1936 GetSection (uint8_t n_sect, addr_t file_addr) 1937 { 1938 if (n_sect == 0) 1939 return SectionSP(); 1940 if (n_sect < m_section_infos.size()) 1941 { 1942 if (!m_section_infos[n_sect].section_sp) 1943 { 1944 SectionSP section_sp (m_section_list->FindSectionByID (n_sect)); 1945 m_section_infos[n_sect].section_sp = section_sp; 1946 if (section_sp) 1947 { 1948 m_section_infos[n_sect].vm_range.SetBaseAddress (section_sp->GetFileAddress()); 1949 m_section_infos[n_sect].vm_range.SetByteSize (section_sp->GetByteSize()); 1950 } 1951 else 1952 { 1953 Host::SystemLog (Host::eSystemLogError, "error: unable to find section for section %u\n", n_sect); 1954 } 1955 } 1956 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) 1957 { 1958 // Symbol is in section. 1959 return m_section_infos[n_sect].section_sp; 1960 } 1961 else if (m_section_infos[n_sect].vm_range.GetByteSize () == 0 && 1962 m_section_infos[n_sect].vm_range.GetBaseAddress() == file_addr) 1963 { 1964 // Symbol is in section with zero size, but has the same start 1965 // address as the section. This can happen with linker symbols 1966 // (symbols that start with the letter 'l' or 'L'. 1967 return m_section_infos[n_sect].section_sp; 1968 } 1969 } 1970 return m_section_list->FindSectionContainingFileAddress(file_addr); 1971 } 1972 1973 protected: 1974 struct SectionInfo 1975 { 1976 SectionInfo () : 1977 vm_range(), 1978 section_sp () 1979 { 1980 } 1981 1982 VMRange vm_range; 1983 SectionSP section_sp; 1984 }; 1985 SectionList *m_section_list; 1986 std::vector<SectionInfo> m_section_infos; 1987 }; 1988 1989 struct TrieEntry 1990 { 1991 TrieEntry () : 1992 name(), 1993 address(LLDB_INVALID_ADDRESS), 1994 flags (0), 1995 other(0), 1996 import_name() 1997 { 1998 } 1999 2000 void 2001 Clear () 2002 { 2003 name.Clear(); 2004 address = LLDB_INVALID_ADDRESS; 2005 flags = 0; 2006 other = 0; 2007 import_name.Clear(); 2008 } 2009 2010 void 2011 Dump () const 2012 { 2013 printf ("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 2014 static_cast<unsigned long long>(address), 2015 static_cast<unsigned long long>(flags), 2016 static_cast<unsigned long long>(other), name.GetCString()); 2017 if (import_name) 2018 printf (" -> \"%s\"\n", import_name.GetCString()); 2019 else 2020 printf ("\n"); 2021 } 2022 ConstString name; 2023 uint64_t address; 2024 uint64_t flags; 2025 uint64_t other; 2026 ConstString import_name; 2027 }; 2028 2029 struct TrieEntryWithOffset 2030 { 2031 lldb::offset_t nodeOffset; 2032 TrieEntry entry; 2033 2034 TrieEntryWithOffset (lldb::offset_t offset) : 2035 nodeOffset (offset), 2036 entry() 2037 { 2038 } 2039 2040 void 2041 Dump (uint32_t idx) const 2042 { 2043 printf ("[%3u] 0x%16.16llx: ", idx, 2044 static_cast<unsigned long long>(nodeOffset)); 2045 entry.Dump(); 2046 } 2047 2048 bool 2049 operator<(const TrieEntryWithOffset& other) const 2050 { 2051 return ( nodeOffset < other.nodeOffset ); 2052 } 2053 }; 2054 2055 static void 2056 ParseTrieEntries (DataExtractor &data, 2057 lldb::offset_t offset, 2058 std::vector<llvm::StringRef> &nameSlices, 2059 std::set<lldb::addr_t> &resolver_addresses, 2060 std::vector<TrieEntryWithOffset>& output) 2061 { 2062 if (!data.ValidOffset(offset)) 2063 return; 2064 2065 const uint64_t terminalSize = data.GetULEB128(&offset); 2066 lldb::offset_t children_offset = offset + terminalSize; 2067 if ( terminalSize != 0 ) { 2068 TrieEntryWithOffset e (offset); 2069 e.entry.flags = data.GetULEB128(&offset); 2070 const char *import_name = NULL; 2071 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT ) { 2072 e.entry.address = 0; 2073 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2074 import_name = data.GetCStr(&offset); 2075 } 2076 else { 2077 e.entry.address = data.GetULEB128(&offset); 2078 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER ) 2079 { 2080 //resolver_addresses.insert(e.entry.address); 2081 e.entry.other = data.GetULEB128(&offset); 2082 resolver_addresses.insert(e.entry.other); 2083 } 2084 else 2085 e.entry.other = 0; 2086 } 2087 // Only add symbols that are reexport symbols with a valid import name 2088 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && import_name[0]) 2089 { 2090 std::string name; 2091 if (!nameSlices.empty()) 2092 { 2093 for (auto name_slice: nameSlices) 2094 name.append(name_slice.data(), name_slice.size()); 2095 } 2096 if (name.size() > 1) 2097 { 2098 // Skip the leading '_' 2099 e.entry.name.SetCStringWithLength(name.c_str() + 1,name.size() - 1); 2100 } 2101 if (import_name) 2102 { 2103 // Skip the leading '_' 2104 e.entry.import_name.SetCString(import_name+1); 2105 } 2106 output.push_back(e); 2107 } 2108 } 2109 2110 const uint8_t childrenCount = data.GetU8(&children_offset); 2111 for (uint8_t i=0; i < childrenCount; ++i) { 2112 nameSlices.push_back(data.GetCStr(&children_offset)); 2113 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2114 if (childNodeOffset) 2115 { 2116 ParseTrieEntries(data, 2117 childNodeOffset, 2118 nameSlices, 2119 resolver_addresses, 2120 output); 2121 } 2122 nameSlices.pop_back(); 2123 } 2124 } 2125 2126 size_t 2127 ObjectFileMachO::ParseSymtab () 2128 { 2129 Timer scoped_timer(__PRETTY_FUNCTION__, 2130 "ObjectFileMachO::ParseSymtab () module = %s", 2131 m_file.GetFilename().AsCString("")); 2132 ModuleSP module_sp (GetModule()); 2133 if (!module_sp) 2134 return 0; 2135 2136 struct symtab_command symtab_load_command = { 0, 0, 0, 0, 0, 0 }; 2137 struct linkedit_data_command function_starts_load_command = { 0, 0, 0, 0 }; 2138 struct dyld_info_command dyld_info = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; 2139 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2140 FunctionStarts function_starts; 2141 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2142 uint32_t i; 2143 FileSpecList dylib_files; 2144 Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_SYMBOLS)); 2145 2146 for (i=0; i<m_header.ncmds; ++i) 2147 { 2148 const lldb::offset_t cmd_offset = offset; 2149 // Read in the load command and load command size 2150 struct load_command lc; 2151 if (m_data.GetU32(&offset, &lc, 2) == NULL) 2152 break; 2153 // Watch for the symbol table load command 2154 switch (lc.cmd) 2155 { 2156 case LC_SYMTAB: 2157 symtab_load_command.cmd = lc.cmd; 2158 symtab_load_command.cmdsize = lc.cmdsize; 2159 // Read in the rest of the symtab load command 2160 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 0) // fill in symoff, nsyms, stroff, strsize fields 2161 return 0; 2162 if (symtab_load_command.symoff == 0) 2163 { 2164 if (log) 2165 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2166 return 0; 2167 } 2168 2169 if (symtab_load_command.stroff == 0) 2170 { 2171 if (log) 2172 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2173 return 0; 2174 } 2175 2176 if (symtab_load_command.nsyms == 0) 2177 { 2178 if (log) 2179 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2180 return 0; 2181 } 2182 2183 if (symtab_load_command.strsize == 0) 2184 { 2185 if (log) 2186 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2187 return 0; 2188 } 2189 break; 2190 2191 case LC_DYLD_INFO: 2192 case LC_DYLD_INFO_ONLY: 2193 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) 2194 { 2195 dyld_info.cmd = lc.cmd; 2196 dyld_info.cmdsize = lc.cmdsize; 2197 } 2198 else 2199 { 2200 memset (&dyld_info, 0, sizeof(dyld_info)); 2201 } 2202 break; 2203 2204 case LC_LOAD_DYLIB: 2205 case LC_LOAD_WEAK_DYLIB: 2206 case LC_REEXPORT_DYLIB: 2207 case LC_LOADFVMLIB: 2208 case LC_LOAD_UPWARD_DYLIB: 2209 { 2210 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2211 const char *path = m_data.PeekCStr(name_offset); 2212 if (path) 2213 { 2214 FileSpec file_spec(path, false); 2215 // Strip the path if there is @rpath, @executable, etc so we just use the basename 2216 if (path[0] == '@') 2217 file_spec.GetDirectory().Clear(); 2218 2219 if (lc.cmd == LC_REEXPORT_DYLIB) 2220 { 2221 m_reexported_dylibs.AppendIfUnique(file_spec); 2222 } 2223 2224 dylib_files.Append(file_spec); 2225 } 2226 } 2227 break; 2228 2229 case LC_FUNCTION_STARTS: 2230 function_starts_load_command.cmd = lc.cmd; 2231 function_starts_load_command.cmdsize = lc.cmdsize; 2232 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == NULL) // fill in symoff, nsyms, stroff, strsize fields 2233 memset (&function_starts_load_command, 0, sizeof(function_starts_load_command)); 2234 break; 2235 2236 default: 2237 break; 2238 } 2239 offset = cmd_offset + lc.cmdsize; 2240 } 2241 2242 if (symtab_load_command.cmd) 2243 { 2244 Symtab *symtab = m_symtab_ap.get(); 2245 SectionList *section_list = GetSectionList(); 2246 if (section_list == NULL) 2247 return 0; 2248 2249 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2250 const ByteOrder byte_order = m_data.GetByteOrder(); 2251 bool bit_width_32 = addr_byte_size == 4; 2252 const size_t nlist_byte_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2253 2254 DataExtractor nlist_data (NULL, 0, byte_order, addr_byte_size); 2255 DataExtractor strtab_data (NULL, 0, byte_order, addr_byte_size); 2256 DataExtractor function_starts_data (NULL, 0, byte_order, addr_byte_size); 2257 DataExtractor indirect_symbol_index_data (NULL, 0, byte_order, addr_byte_size); 2258 DataExtractor dyld_trie_data (NULL, 0, byte_order, addr_byte_size); 2259 2260 const addr_t nlist_data_byte_size = symtab_load_command.nsyms * nlist_byte_size; 2261 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2262 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2263 2264 ProcessSP process_sp (m_process_wp.lock()); 2265 Process *process = process_sp.get(); 2266 2267 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2268 2269 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) 2270 { 2271 Target &target = process->GetTarget(); 2272 2273 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2274 2275 SectionSP linkedit_section_sp(section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2276 // Reading mach file from memory in a process or core file... 2277 2278 if (linkedit_section_sp) 2279 { 2280 const addr_t linkedit_load_addr = linkedit_section_sp->GetLoadBaseAddress(&target); 2281 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2282 const addr_t symoff_addr = linkedit_load_addr + symtab_load_command.symoff - linkedit_file_offset; 2283 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - linkedit_file_offset; 2284 2285 bool data_was_read = false; 2286 2287 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 2288 if (m_header.flags & 0x80000000u && process->GetAddressByteSize() == sizeof (void*)) 2289 { 2290 // This mach-o memory file is in the dyld shared cache. If this 2291 // program is not remote and this is iOS, then this process will 2292 // share the same shared cache as the process we are debugging and 2293 // we can read the entire __LINKEDIT from the address space in this 2294 // process. This is a needed optimization that is used for local iOS 2295 // debugging only since all shared libraries in the shared cache do 2296 // not have corresponding files that exist in the file system of the 2297 // device. They have been combined into a single file. This means we 2298 // always have to load these files from memory. All of the symbol and 2299 // string tables from all of the __LINKEDIT sections from the shared 2300 // libraries in the shared cache have been merged into a single large 2301 // symbol and string table. Reading all of this symbol and string table 2302 // data across can slow down debug launch times, so we optimize this by 2303 // reading the memory for the __LINKEDIT section from this process. 2304 2305 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 2306 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 2307 bool use_lldb_cache = true; 2308 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && lldb_shared_cache != process_shared_cache) 2309 { 2310 use_lldb_cache = false; 2311 ModuleSP module_sp (GetModule()); 2312 if (module_sp) 2313 module_sp->ReportWarning ("shared cache in process does not match lldb's own shared cache, startup will be slow."); 2314 2315 } 2316 2317 PlatformSP platform_sp (target.GetPlatform()); 2318 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) 2319 { 2320 data_was_read = true; 2321 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, eByteOrderLittle); 2322 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, eByteOrderLittle); 2323 if (function_starts_load_command.cmd) 2324 { 2325 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 2326 function_starts_data.SetData ((void *)func_start_addr, function_starts_load_command.datasize, eByteOrderLittle); 2327 } 2328 } 2329 } 2330 #endif 2331 2332 if (!data_was_read) 2333 { 2334 if (memory_module_load_level == eMemoryModuleLoadLevelComplete) 2335 { 2336 DataBufferSP nlist_data_sp (ReadMemory (process_sp, symoff_addr, nlist_data_byte_size)); 2337 if (nlist_data_sp) 2338 nlist_data.SetData (nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2339 // Load strings individually from memory when loading from memory since shared cache 2340 // string tables contain strings for all symbols from all shared cached libraries 2341 //DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, strtab_data_byte_size)); 2342 //if (strtab_data_sp) 2343 // strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 2344 if (m_dysymtab.nindirectsyms != 0) 2345 { 2346 const addr_t indirect_syms_addr = linkedit_load_addr + m_dysymtab.indirectsymoff - linkedit_file_offset; 2347 DataBufferSP indirect_syms_data_sp (ReadMemory (process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 2348 if (indirect_syms_data_sp) 2349 indirect_symbol_index_data.SetData (indirect_syms_data_sp, 0, indirect_syms_data_sp->GetByteSize()); 2350 } 2351 } 2352 2353 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) 2354 { 2355 if (function_starts_load_command.cmd) 2356 { 2357 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 2358 DataBufferSP func_start_data_sp (ReadMemory (process_sp, func_start_addr, function_starts_load_command.datasize)); 2359 if (func_start_data_sp) 2360 function_starts_data.SetData (func_start_data_sp, 0, func_start_data_sp->GetByteSize()); 2361 } 2362 } 2363 } 2364 } 2365 } 2366 else 2367 { 2368 nlist_data.SetData (m_data, 2369 symtab_load_command.symoff, 2370 nlist_data_byte_size); 2371 strtab_data.SetData (m_data, 2372 symtab_load_command.stroff, 2373 strtab_data_byte_size); 2374 2375 if (dyld_info.export_size > 0) 2376 { 2377 dyld_trie_data.SetData (m_data, 2378 dyld_info.export_off, 2379 dyld_info.export_size); 2380 } 2381 2382 if (m_dysymtab.nindirectsyms != 0) 2383 { 2384 indirect_symbol_index_data.SetData (m_data, 2385 m_dysymtab.indirectsymoff, 2386 m_dysymtab.nindirectsyms * 4); 2387 } 2388 if (function_starts_load_command.cmd) 2389 { 2390 function_starts_data.SetData (m_data, 2391 function_starts_load_command.dataoff, 2392 function_starts_load_command.datasize); 2393 } 2394 } 2395 2396 if (nlist_data.GetByteSize() == 0 && memory_module_load_level == eMemoryModuleLoadLevelComplete) 2397 { 2398 if (log) 2399 module_sp->LogMessage(log, "failed to read nlist data"); 2400 return 0; 2401 } 2402 2403 2404 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2405 if (!have_strtab_data) 2406 { 2407 if (process) 2408 { 2409 if (strtab_addr == LLDB_INVALID_ADDRESS) 2410 { 2411 if (log) 2412 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2413 return 0; 2414 } 2415 } 2416 else 2417 { 2418 if (log) 2419 module_sp->LogMessage(log, "failed to read strtab data"); 2420 return 0; 2421 } 2422 } 2423 2424 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2425 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2426 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2427 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2428 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 2429 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 2430 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 2431 SectionSP eh_frame_section_sp; 2432 if (text_section_sp.get()) 2433 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 2434 else 2435 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 2436 2437 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2438 2439 // lldb works best if it knows the start address of all functions in a module. 2440 // Linker symbols or debug info are normally the best source of information for start addr / size but 2441 // they may be stripped in a released binary. 2442 // Two additional sources of information exist in Mach-O binaries: 2443 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each function's start address in the 2444 // binary, relative to the text section. 2445 // eh_frame - the eh_frame FDEs have the start addr & size of each function 2446 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on all modern binaries. 2447 // Binaries built to run on older releases may need to use eh_frame information. 2448 2449 if (text_section_sp && function_starts_data.GetByteSize()) 2450 { 2451 FunctionStarts::Entry function_start_entry; 2452 function_start_entry.data = false; 2453 lldb::offset_t function_start_offset = 0; 2454 function_start_entry.addr = text_section_sp->GetFileAddress(); 2455 uint64_t delta; 2456 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 0) 2457 { 2458 // Now append the current entry 2459 function_start_entry.addr += delta; 2460 function_starts.Append(function_start_entry); 2461 } 2462 } 2463 else 2464 { 2465 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the load command claiming an eh_frame 2466 // but it doesn't actually have the eh_frame content. And if we have a dSYM, we don't need to do any 2467 // of this fill-in-the-missing-symbols works anyway - the debug info should give us all the functions in 2468 // the module. 2469 if (text_section_sp.get() && eh_frame_section_sp.get() && m_type != eTypeDebugInfo) 2470 { 2471 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, eRegisterKindGCC, true); 2472 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2473 eh_frame.GetFunctionAddressAndSizeVector (functions); 2474 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2475 size_t count = functions.GetSize(); 2476 for (size_t i = 0; i < count; ++i) 2477 { 2478 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = functions.GetEntryAtIndex (i); 2479 if (func) 2480 { 2481 FunctionStarts::Entry function_start_entry; 2482 function_start_entry.addr = func->base - text_base_addr; 2483 function_starts.Append(function_start_entry); 2484 } 2485 } 2486 } 2487 } 2488 2489 const size_t function_starts_count = function_starts.GetSize(); 2490 2491 const user_id_t TEXT_eh_frame_sectID = 2492 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2493 : static_cast<user_id_t>(NO_SECT); 2494 2495 lldb::offset_t nlist_data_offset = 0; 2496 2497 uint32_t N_SO_index = UINT32_MAX; 2498 2499 MachSymtabSectionInfo section_info (section_list); 2500 std::vector<uint32_t> N_FUN_indexes; 2501 std::vector<uint32_t> N_NSYM_indexes; 2502 std::vector<uint32_t> N_INCL_indexes; 2503 std::vector<uint32_t> N_BRAC_indexes; 2504 std::vector<uint32_t> N_COMM_indexes; 2505 typedef std::multimap <uint64_t, uint32_t> ValueToSymbolIndexMap; 2506 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2507 typedef std::map <const char *, uint32_t> ConstNameToSymbolIndexMap; 2508 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2509 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2510 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2511 // Any symbols that get merged into another will get an entry 2512 // in this map so we know 2513 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2514 uint32_t nlist_idx = 0; 2515 Symbol *symbol_ptr = NULL; 2516 2517 uint32_t sym_idx = 0; 2518 Symbol *sym = NULL; 2519 size_t num_syms = 0; 2520 std::string memory_symbol_name; 2521 uint32_t unmapped_local_symbols_found = 0; 2522 2523 std::vector<TrieEntryWithOffset> trie_entries; 2524 std::set<lldb::addr_t> resolver_addresses; 2525 2526 if (dyld_trie_data.GetByteSize() > 0) 2527 { 2528 std::vector<llvm::StringRef> nameSlices; 2529 ParseTrieEntries (dyld_trie_data, 2530 0, 2531 nameSlices, 2532 resolver_addresses, 2533 trie_entries); 2534 2535 ConstString text_segment_name ("__TEXT"); 2536 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 2537 if (text_segment_sp) 2538 { 2539 const lldb::addr_t text_segment_file_addr = text_segment_sp->GetFileAddress(); 2540 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) 2541 { 2542 for (auto &e : trie_entries) 2543 e.entry.address += text_segment_file_addr; 2544 } 2545 } 2546 } 2547 2548 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 2549 2550 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been optimized by moving LOCAL 2551 // symbols out of the memory mapped portion of the DSC. The symbol information has all been retained, 2552 // but it isn't available in the normal nlist data. However, there *are* duplicate entries of *some* 2553 // LOCAL symbols in the normal nlist data. To handle this situation correctly, we must first attempt 2554 // to parse any DSC unmapped symbol information. If we find any, we set a flag that tells the normal 2555 // nlist parser to ignore all LOCAL symbols. 2556 2557 if (m_header.flags & 0x80000000u) 2558 { 2559 // Before we can start mapping the DSC, we need to make certain the target process is actually 2560 // using the cache we can find. 2561 2562 // Next we need to determine the correct path for the dyld shared cache. 2563 2564 ArchSpec header_arch; 2565 GetArchitecture(header_arch); 2566 char dsc_path[PATH_MAX]; 2567 2568 snprintf(dsc_path, sizeof(dsc_path), "%s%s%s", 2569 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR */ 2570 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2571 header_arch.GetArchitectureName()); 2572 2573 FileSpec dsc_filespec(dsc_path, false); 2574 2575 // We need definitions of two structures in the on-disk DSC, copy them here manually 2576 struct lldb_copy_dyld_cache_header_v0 2577 { 2578 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2579 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2580 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2581 uint32_t imagesOffset; 2582 uint32_t imagesCount; 2583 uint64_t dyldBaseAddress; 2584 uint64_t codeSignatureOffset; 2585 uint64_t codeSignatureSize; 2586 uint64_t slideInfoOffset; 2587 uint64_t slideInfoSize; 2588 uint64_t localSymbolsOffset; // file offset of where local symbols are stored 2589 uint64_t localSymbolsSize; // size of local symbols information 2590 }; 2591 struct lldb_copy_dyld_cache_header_v1 2592 { 2593 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2594 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2595 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2596 uint32_t imagesOffset; 2597 uint32_t imagesCount; 2598 uint64_t dyldBaseAddress; 2599 uint64_t codeSignatureOffset; 2600 uint64_t codeSignatureSize; 2601 uint64_t slideInfoOffset; 2602 uint64_t slideInfoSize; 2603 uint64_t localSymbolsOffset; 2604 uint64_t localSymbolsSize; 2605 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 and later 2606 }; 2607 2608 struct lldb_copy_dyld_cache_mapping_info 2609 { 2610 uint64_t address; 2611 uint64_t size; 2612 uint64_t fileOffset; 2613 uint32_t maxProt; 2614 uint32_t initProt; 2615 }; 2616 2617 struct lldb_copy_dyld_cache_local_symbols_info 2618 { 2619 uint32_t nlistOffset; 2620 uint32_t nlistCount; 2621 uint32_t stringsOffset; 2622 uint32_t stringsSize; 2623 uint32_t entriesOffset; 2624 uint32_t entriesCount; 2625 }; 2626 struct lldb_copy_dyld_cache_local_symbols_entry 2627 { 2628 uint32_t dylibOffset; 2629 uint32_t nlistStartIndex; 2630 uint32_t nlistCount; 2631 }; 2632 2633 /* The dyld_cache_header has a pointer to the dyld_cache_local_symbols_info structure (localSymbolsOffset). 2634 The dyld_cache_local_symbols_info structure gives us three things: 2635 1. The start and count of the nlist records in the dyld_shared_cache file 2636 2. The start and size of the strings for these nlist records 2637 3. The start and count of dyld_cache_local_symbols_entry entries 2638 2639 There is one dyld_cache_local_symbols_entry per dylib/framework in the dyld shared cache. 2640 The "dylibOffset" field is the Mach-O header of this dylib/framework in the dyld shared cache. 2641 The dyld_cache_local_symbols_entry also lists the start of this dylib/framework's nlist records 2642 and the count of how many nlist records there are for this dylib/framework. 2643 */ 2644 2645 // Process the dsc header to find the unmapped symbols 2646 // 2647 // Save some VM space, do not map the entire cache in one shot. 2648 2649 DataBufferSP dsc_data_sp; 2650 dsc_data_sp = dsc_filespec.MemoryMapFileContents(0, sizeof(struct lldb_copy_dyld_cache_header_v1)); 2651 2652 if (dsc_data_sp) 2653 { 2654 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2655 2656 char version_str[17]; 2657 int version = -1; 2658 lldb::offset_t offset = 0; 2659 memcpy (version_str, dsc_header_data.GetData (&offset, 16), 16); 2660 version_str[16] = '\0'; 2661 if (strncmp (version_str, "dyld_v", 6) == 0 && isdigit (version_str[6])) 2662 { 2663 int v; 2664 if (::sscanf (version_str + 6, "%d", &v) == 1) 2665 { 2666 version = v; 2667 } 2668 } 2669 2670 UUID dsc_uuid; 2671 if (version >= 1) 2672 { 2673 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, uuid); 2674 uint8_t uuid_bytes[sizeof (uuid_t)]; 2675 memcpy (uuid_bytes, dsc_header_data.GetData (&offset, sizeof (uuid_t)), sizeof (uuid_t)); 2676 dsc_uuid.SetBytes (uuid_bytes); 2677 } 2678 2679 bool uuid_match = true; 2680 if (dsc_uuid.IsValid() && process) 2681 { 2682 UUID shared_cache_uuid(GetProcessSharedCacheUUID(process)); 2683 2684 if (shared_cache_uuid.IsValid() && dsc_uuid != shared_cache_uuid) 2685 { 2686 // The on-disk dyld_shared_cache file is not the same as the one in this 2687 // process' memory, don't use it. 2688 uuid_match = false; 2689 ModuleSP module_sp (GetModule()); 2690 if (module_sp) 2691 module_sp->ReportWarning ("process shared cache does not match on-disk dyld_shared_cache file, some symbol names will be missing."); 2692 } 2693 } 2694 2695 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2696 2697 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2698 2699 // If the mappingOffset points to a location inside the header, we've 2700 // opened an old dyld shared cache, and should not proceed further. 2701 if (uuid_match && mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v0)) 2702 { 2703 2704 DataBufferSP dsc_mapping_info_data_sp = dsc_filespec.MemoryMapFileContents(mappingOffset, sizeof (struct lldb_copy_dyld_cache_mapping_info)); 2705 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, byte_order, addr_byte_size); 2706 offset = 0; 2707 2708 // The File addresses (from the in-memory Mach-O load commands) for the shared libraries 2709 // in the shared library cache need to be adjusted by an offset to match up with the 2710 // dylibOffset identifying field in the dyld_cache_local_symbol_entry's. This offset is 2711 // recorded in mapping_offset_value. 2712 const uint64_t mapping_offset_value = dsc_mapping_info_data.GetU64(&offset); 2713 2714 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2715 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2716 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2717 2718 if (localSymbolsOffset && localSymbolsSize) 2719 { 2720 // Map the local symbols 2721 if (DataBufferSP dsc_local_symbols_data_sp = dsc_filespec.MemoryMapFileContents(localSymbolsOffset, localSymbolsSize)) 2722 { 2723 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, byte_order, addr_byte_size); 2724 2725 offset = 0; 2726 2727 // Read the local_symbols_infos struct in one shot 2728 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2729 dsc_local_symbols_data.GetU32(&offset, &local_symbols_info.nlistOffset, 6); 2730 2731 SectionSP text_section_sp(section_list->FindSectionByName(GetSegmentNameTEXT())); 2732 2733 uint32_t header_file_offset = (text_section_sp->GetFileAddress() - mapping_offset_value); 2734 2735 offset = local_symbols_info.entriesOffset; 2736 for (uint32_t entry_index = 0; entry_index < local_symbols_info.entriesCount; entry_index++) 2737 { 2738 struct lldb_copy_dyld_cache_local_symbols_entry local_symbols_entry; 2739 local_symbols_entry.dylibOffset = dsc_local_symbols_data.GetU32(&offset); 2740 local_symbols_entry.nlistStartIndex = dsc_local_symbols_data.GetU32(&offset); 2741 local_symbols_entry.nlistCount = dsc_local_symbols_data.GetU32(&offset); 2742 2743 if (header_file_offset == local_symbols_entry.dylibOffset) 2744 { 2745 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2746 2747 // The normal nlist code cannot correctly size the Symbols array, we need to allocate it here. 2748 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms + unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2749 num_syms = symtab->GetNumSymbols(); 2750 2751 nlist_data_offset = local_symbols_info.nlistOffset + (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2752 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2753 2754 for (uint32_t nlist_index = 0; nlist_index < local_symbols_entry.nlistCount; nlist_index++) 2755 { 2756 ///////////////////////////// 2757 { 2758 struct nlist_64 nlist; 2759 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2760 break; 2761 2762 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked(&nlist_data_offset); 2763 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2764 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2765 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked (&nlist_data_offset); 2766 nlist.n_value = dsc_local_symbols_data.GetAddress_unchecked (&nlist_data_offset); 2767 2768 SymbolType type = eSymbolTypeInvalid; 2769 const char *symbol_name = dsc_local_symbols_data.PeekCStr(string_table_offset + nlist.n_strx); 2770 2771 if (symbol_name == NULL) 2772 { 2773 // No symbol should be NULL, even the symbols with no 2774 // string values should have an offset zero which points 2775 // to an empty C-string 2776 Host::SystemLog (Host::eSystemLogError, 2777 "error: DSC unmapped local symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 2778 entry_index, 2779 nlist.n_strx, 2780 module_sp->GetFileSpec().GetPath().c_str()); 2781 continue; 2782 } 2783 if (symbol_name[0] == '\0') 2784 symbol_name = NULL; 2785 2786 const char *symbol_name_non_abi_mangled = NULL; 2787 2788 SectionSP symbol_section; 2789 uint32_t symbol_byte_size = 0; 2790 bool add_nlist = true; 2791 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2792 bool demangled_is_synthesized = false; 2793 bool is_gsym = false; 2794 2795 assert (sym_idx < num_syms); 2796 2797 sym[sym_idx].SetDebug (is_debug); 2798 2799 if (is_debug) 2800 { 2801 switch (nlist.n_type) 2802 { 2803 case N_GSYM: 2804 // global symbol: name,,NO_SECT,type,0 2805 // Sometimes the N_GSYM value contains the address. 2806 2807 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 2808 // have the same address, but we want to ensure that we always find only the real symbol, 2809 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 2810 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 2811 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 2812 // same address. 2813 2814 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 2815 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 2816 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 2817 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 2818 add_nlist = false; 2819 else 2820 { 2821 is_gsym = true; 2822 sym[sym_idx].SetExternal(true); 2823 if (nlist.n_value != 0) 2824 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2825 type = eSymbolTypeData; 2826 } 2827 break; 2828 2829 case N_FNAME: 2830 // procedure name (f77 kludge): name,,NO_SECT,0,0 2831 type = eSymbolTypeCompiler; 2832 break; 2833 2834 case N_FUN: 2835 // procedure: name,,n_sect,linenumber,address 2836 if (symbol_name) 2837 { 2838 type = eSymbolTypeCode; 2839 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2840 2841 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 2842 // We use the current number of symbols in the symbol table in lieu of 2843 // using nlist_idx in case we ever start trimming entries out 2844 N_FUN_indexes.push_back(sym_idx); 2845 } 2846 else 2847 { 2848 type = eSymbolTypeCompiler; 2849 2850 if ( !N_FUN_indexes.empty() ) 2851 { 2852 // Copy the size of the function into the original STAB entry so we don't have 2853 // to hunt for it later 2854 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 2855 N_FUN_indexes.pop_back(); 2856 // We don't really need the end function STAB as it contains the size which 2857 // we already placed with the original symbol, so don't add it if we want a 2858 // minimal symbol table 2859 add_nlist = false; 2860 } 2861 } 2862 break; 2863 2864 case N_STSYM: 2865 // static symbol: name,,n_sect,type,address 2866 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 2867 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2868 type = eSymbolTypeData; 2869 break; 2870 2871 case N_LCSYM: 2872 // .lcomm symbol: name,,n_sect,type,address 2873 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2874 type = eSymbolTypeCommonBlock; 2875 break; 2876 2877 case N_BNSYM: 2878 // We use the current number of symbols in the symbol table in lieu of 2879 // using nlist_idx in case we ever start trimming entries out 2880 // Skip these if we want minimal symbol tables 2881 add_nlist = false; 2882 break; 2883 2884 case N_ENSYM: 2885 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 2886 // so that we can always skip the entire symbol if we need to navigate 2887 // more quickly at the source level when parsing STABS 2888 // Skip these if we want minimal symbol tables 2889 add_nlist = false; 2890 break; 2891 2892 2893 case N_OPT: 2894 // emitted with gcc2_compiled and in gcc source 2895 type = eSymbolTypeCompiler; 2896 break; 2897 2898 case N_RSYM: 2899 // register sym: name,,NO_SECT,type,register 2900 type = eSymbolTypeVariable; 2901 break; 2902 2903 case N_SLINE: 2904 // src line: 0,,n_sect,linenumber,address 2905 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2906 type = eSymbolTypeLineEntry; 2907 break; 2908 2909 case N_SSYM: 2910 // structure elt: name,,NO_SECT,type,struct_offset 2911 type = eSymbolTypeVariableType; 2912 break; 2913 2914 case N_SO: 2915 // source file name 2916 type = eSymbolTypeSourceFile; 2917 if (symbol_name == NULL) 2918 { 2919 add_nlist = false; 2920 if (N_SO_index != UINT32_MAX) 2921 { 2922 // Set the size of the N_SO to the terminating index of this N_SO 2923 // so that we can always skip the entire N_SO if we need to navigate 2924 // more quickly at the source level when parsing STABS 2925 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2926 symbol_ptr->SetByteSize(sym_idx); 2927 symbol_ptr->SetSizeIsSibling(true); 2928 } 2929 N_NSYM_indexes.clear(); 2930 N_INCL_indexes.clear(); 2931 N_BRAC_indexes.clear(); 2932 N_COMM_indexes.clear(); 2933 N_FUN_indexes.clear(); 2934 N_SO_index = UINT32_MAX; 2935 } 2936 else 2937 { 2938 // We use the current number of symbols in the symbol table in lieu of 2939 // using nlist_idx in case we ever start trimming entries out 2940 const bool N_SO_has_full_path = symbol_name[0] == '/'; 2941 if (N_SO_has_full_path) 2942 { 2943 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2944 { 2945 // We have two consecutive N_SO entries where the first contains a directory 2946 // and the second contains a full path. 2947 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 2948 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2949 add_nlist = false; 2950 } 2951 else 2952 { 2953 // This is the first entry in a N_SO that contains a directory or 2954 // a full path to the source file 2955 N_SO_index = sym_idx; 2956 } 2957 } 2958 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2959 { 2960 // This is usually the second N_SO entry that contains just the filename, 2961 // so here we combine it with the first one if we are minimizing the symbol table 2962 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 2963 if (so_path && so_path[0]) 2964 { 2965 std::string full_so_path (so_path); 2966 const size_t double_slash_pos = full_so_path.find("//"); 2967 if (double_slash_pos != std::string::npos) 2968 { 2969 // The linker has been generating bad N_SO entries with doubled up paths 2970 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 2971 // and the second is the directory for the source file so you end up with 2972 // a path that looks like "/tmp/src//tmp/src/" 2973 FileSpec so_dir(so_path, false); 2974 if (!so_dir.Exists()) 2975 { 2976 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 2977 if (so_dir.Exists()) 2978 { 2979 // Trim off the incorrect path 2980 full_so_path.erase(0, double_slash_pos + 1); 2981 } 2982 } 2983 } 2984 if (*full_so_path.rbegin() != '/') 2985 full_so_path += '/'; 2986 full_so_path += symbol_name; 2987 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 2988 add_nlist = false; 2989 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2990 } 2991 } 2992 else 2993 { 2994 // This could be a relative path to a N_SO 2995 N_SO_index = sym_idx; 2996 } 2997 } 2998 break; 2999 3000 case N_OSO: 3001 // object file name: name,,0,0,st_mtime 3002 type = eSymbolTypeObjectFile; 3003 break; 3004 3005 case N_LSYM: 3006 // local sym: name,,NO_SECT,type,offset 3007 type = eSymbolTypeLocal; 3008 break; 3009 3010 //---------------------------------------------------------------------- 3011 // INCL scopes 3012 //---------------------------------------------------------------------- 3013 case N_BINCL: 3014 // include file beginning: name,,NO_SECT,0,sum 3015 // We use the current number of symbols in the symbol table in lieu of 3016 // using nlist_idx in case we ever start trimming entries out 3017 N_INCL_indexes.push_back(sym_idx); 3018 type = eSymbolTypeScopeBegin; 3019 break; 3020 3021 case N_EINCL: 3022 // include file end: name,,NO_SECT,0,0 3023 // Set the size of the N_BINCL to the terminating index of this N_EINCL 3024 // so that we can always skip the entire symbol if we need to navigate 3025 // more quickly at the source level when parsing STABS 3026 if ( !N_INCL_indexes.empty() ) 3027 { 3028 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3029 symbol_ptr->SetByteSize(sym_idx + 1); 3030 symbol_ptr->SetSizeIsSibling(true); 3031 N_INCL_indexes.pop_back(); 3032 } 3033 type = eSymbolTypeScopeEnd; 3034 break; 3035 3036 case N_SOL: 3037 // #included file name: name,,n_sect,0,address 3038 type = eSymbolTypeHeaderFile; 3039 3040 // We currently don't use the header files on darwin 3041 add_nlist = false; 3042 break; 3043 3044 case N_PARAMS: 3045 // compiler parameters: name,,NO_SECT,0,0 3046 type = eSymbolTypeCompiler; 3047 break; 3048 3049 case N_VERSION: 3050 // compiler version: name,,NO_SECT,0,0 3051 type = eSymbolTypeCompiler; 3052 break; 3053 3054 case N_OLEVEL: 3055 // compiler -O level: name,,NO_SECT,0,0 3056 type = eSymbolTypeCompiler; 3057 break; 3058 3059 case N_PSYM: 3060 // parameter: name,,NO_SECT,type,offset 3061 type = eSymbolTypeVariable; 3062 break; 3063 3064 case N_ENTRY: 3065 // alternate entry: name,,n_sect,linenumber,address 3066 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3067 type = eSymbolTypeLineEntry; 3068 break; 3069 3070 //---------------------------------------------------------------------- 3071 // Left and Right Braces 3072 //---------------------------------------------------------------------- 3073 case N_LBRAC: 3074 // left bracket: 0,,NO_SECT,nesting level,address 3075 // We use the current number of symbols in the symbol table in lieu of 3076 // using nlist_idx in case we ever start trimming entries out 3077 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3078 N_BRAC_indexes.push_back(sym_idx); 3079 type = eSymbolTypeScopeBegin; 3080 break; 3081 3082 case N_RBRAC: 3083 // right bracket: 0,,NO_SECT,nesting level,address 3084 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 3085 // so that we can always skip the entire symbol if we need to navigate 3086 // more quickly at the source level when parsing STABS 3087 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3088 if ( !N_BRAC_indexes.empty() ) 3089 { 3090 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3091 symbol_ptr->SetByteSize(sym_idx + 1); 3092 symbol_ptr->SetSizeIsSibling(true); 3093 N_BRAC_indexes.pop_back(); 3094 } 3095 type = eSymbolTypeScopeEnd; 3096 break; 3097 3098 case N_EXCL: 3099 // deleted include file: name,,NO_SECT,0,sum 3100 type = eSymbolTypeHeaderFile; 3101 break; 3102 3103 //---------------------------------------------------------------------- 3104 // COMM scopes 3105 //---------------------------------------------------------------------- 3106 case N_BCOMM: 3107 // begin common: name,,NO_SECT,0,0 3108 // We use the current number of symbols in the symbol table in lieu of 3109 // using nlist_idx in case we ever start trimming entries out 3110 type = eSymbolTypeScopeBegin; 3111 N_COMM_indexes.push_back(sym_idx); 3112 break; 3113 3114 case N_ECOML: 3115 // end common (local name): 0,,n_sect,0,address 3116 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3117 // Fall through 3118 3119 case N_ECOMM: 3120 // end common: name,,n_sect,0,0 3121 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 3122 // so that we can always skip the entire symbol if we need to navigate 3123 // more quickly at the source level when parsing STABS 3124 if ( !N_COMM_indexes.empty() ) 3125 { 3126 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 3127 symbol_ptr->SetByteSize(sym_idx + 1); 3128 symbol_ptr->SetSizeIsSibling(true); 3129 N_COMM_indexes.pop_back(); 3130 } 3131 type = eSymbolTypeScopeEnd; 3132 break; 3133 3134 case N_LENG: 3135 // second stab entry with length information 3136 type = eSymbolTypeAdditional; 3137 break; 3138 3139 default: break; 3140 } 3141 } 3142 else 3143 { 3144 //uint8_t n_pext = N_PEXT & nlist.n_type; 3145 uint8_t n_type = N_TYPE & nlist.n_type; 3146 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3147 3148 switch (n_type) 3149 { 3150 case N_INDR: // Fall through 3151 case N_PBUD: // Fall through 3152 case N_UNDF: 3153 type = eSymbolTypeUndefined; 3154 break; 3155 3156 case N_ABS: 3157 type = eSymbolTypeAbsolute; 3158 break; 3159 3160 case N_SECT: 3161 { 3162 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3163 3164 if (symbol_section == NULL) 3165 { 3166 // TODO: warn about this? 3167 add_nlist = false; 3168 break; 3169 } 3170 3171 if (TEXT_eh_frame_sectID == nlist.n_sect) 3172 { 3173 type = eSymbolTypeException; 3174 } 3175 else 3176 { 3177 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 3178 3179 switch (section_type) 3180 { 3181 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 3182 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 3183 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 3184 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 3185 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 3186 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 3187 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 3188 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 3189 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 3190 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 3191 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 3192 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 3193 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 3194 default: 3195 switch (symbol_section->GetType()) 3196 { 3197 case lldb::eSectionTypeCode: 3198 type = eSymbolTypeCode; 3199 break; 3200 case eSectionTypeData: 3201 case eSectionTypeDataCString: // Inlined C string data 3202 case eSectionTypeDataCStringPointers: // Pointers to C string data 3203 case eSectionTypeDataSymbolAddress: // Address of a symbol in the symbol table 3204 case eSectionTypeData4: 3205 case eSectionTypeData8: 3206 case eSectionTypeData16: 3207 type = eSymbolTypeData; 3208 break; 3209 default: 3210 break; 3211 } 3212 break; 3213 } 3214 3215 if (type == eSymbolTypeInvalid) 3216 { 3217 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 3218 if (symbol_section->IsDescendant (text_section_sp.get())) 3219 { 3220 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 3221 S_ATTR_SELF_MODIFYING_CODE | 3222 S_ATTR_SOME_INSTRUCTIONS)) 3223 type = eSymbolTypeData; 3224 else 3225 type = eSymbolTypeCode; 3226 } 3227 else if (symbol_section->IsDescendant(data_section_sp.get())) 3228 { 3229 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 3230 { 3231 type = eSymbolTypeRuntime; 3232 3233 if (symbol_name && 3234 symbol_name[0] == '_' && 3235 symbol_name[1] == 'O' && 3236 symbol_name[2] == 'B') 3237 { 3238 llvm::StringRef symbol_name_ref(symbol_name); 3239 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 3240 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 3241 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 3242 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 3243 { 3244 symbol_name_non_abi_mangled = symbol_name + 1; 3245 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3246 type = eSymbolTypeObjCClass; 3247 demangled_is_synthesized = true; 3248 } 3249 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 3250 { 3251 symbol_name_non_abi_mangled = symbol_name + 1; 3252 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3253 type = eSymbolTypeObjCMetaClass; 3254 demangled_is_synthesized = true; 3255 } 3256 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 3257 { 3258 symbol_name_non_abi_mangled = symbol_name + 1; 3259 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3260 type = eSymbolTypeObjCIVar; 3261 demangled_is_synthesized = true; 3262 } 3263 } 3264 } 3265 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 3266 { 3267 type = eSymbolTypeException; 3268 } 3269 else 3270 { 3271 type = eSymbolTypeData; 3272 } 3273 } 3274 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 3275 { 3276 type = eSymbolTypeTrampoline; 3277 } 3278 else if (symbol_section->IsDescendant(objc_section_sp.get())) 3279 { 3280 type = eSymbolTypeRuntime; 3281 if (symbol_name && symbol_name[0] == '.') 3282 { 3283 llvm::StringRef symbol_name_ref(symbol_name); 3284 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 3285 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 3286 { 3287 symbol_name_non_abi_mangled = symbol_name; 3288 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 3289 type = eSymbolTypeObjCClass; 3290 demangled_is_synthesized = true; 3291 } 3292 } 3293 } 3294 } 3295 } 3296 } 3297 break; 3298 } 3299 } 3300 3301 if (add_nlist) 3302 { 3303 uint64_t symbol_value = nlist.n_value; 3304 if (symbol_name_non_abi_mangled) 3305 { 3306 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 3307 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 3308 } 3309 else 3310 { 3311 bool symbol_name_is_mangled = false; 3312 3313 if (symbol_name && symbol_name[0] == '_') 3314 { 3315 symbol_name_is_mangled = symbol_name[1] == '_'; 3316 symbol_name++; // Skip the leading underscore 3317 } 3318 3319 if (symbol_name) 3320 { 3321 ConstString const_symbol_name(symbol_name); 3322 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 3323 if (is_gsym && is_debug) 3324 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 3325 } 3326 } 3327 if (symbol_section) 3328 { 3329 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3330 if (symbol_byte_size == 0 && function_starts_count > 0) 3331 { 3332 addr_t symbol_lookup_file_addr = nlist.n_value; 3333 // Do an exact address match for non-ARM addresses, else get the closest since 3334 // the symbol might be a thumb symbol which has an address with bit zero set 3335 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 3336 if (is_arm && func_start_entry) 3337 { 3338 // Verify that the function start address is the symbol address (ARM) 3339 // or the symbol address + 1 (thumb) 3340 if (func_start_entry->addr != symbol_lookup_file_addr && 3341 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 3342 { 3343 // Not the right entry, NULL it out... 3344 func_start_entry = NULL; 3345 } 3346 } 3347 if (func_start_entry) 3348 { 3349 func_start_entry->data = true; 3350 3351 addr_t symbol_file_addr = func_start_entry->addr; 3352 uint32_t symbol_flags = 0; 3353 if (is_arm) 3354 { 3355 if (symbol_file_addr & 1) 3356 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3357 symbol_file_addr &= 0xfffffffffffffffeull; 3358 } 3359 3360 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3361 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3362 if (next_func_start_entry) 3363 { 3364 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3365 // Be sure the clear the Thumb address bit when we calculate the size 3366 // from the current and next address 3367 if (is_arm) 3368 next_symbol_file_addr &= 0xfffffffffffffffeull; 3369 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3370 } 3371 else 3372 { 3373 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3374 } 3375 } 3376 } 3377 symbol_value -= section_file_addr; 3378 } 3379 3380 if (is_debug == false) 3381 { 3382 if (type == eSymbolTypeCode) 3383 { 3384 // See if we can find a N_FUN entry for any code symbols. 3385 // If we do find a match, and the name matches, then we 3386 // can merge the two into just the function symbol to avoid 3387 // duplicate entries in the symbol table 3388 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3389 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3390 if (range.first != range.second) 3391 { 3392 bool found_it = false; 3393 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3394 { 3395 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3396 { 3397 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3398 // We just need the flags from the linker symbol, so put these flags 3399 // into the N_FUN flags to avoid duplicate symbols in the symbol table 3400 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3401 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3402 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3403 sym[pos->second].SetType (eSymbolTypeResolver); 3404 sym[sym_idx].Clear(); 3405 found_it = true; 3406 break; 3407 } 3408 } 3409 if (found_it) 3410 continue; 3411 } 3412 else 3413 { 3414 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3415 type = eSymbolTypeResolver; 3416 } 3417 } 3418 else if (type == eSymbolTypeData) 3419 { 3420 // See if we can find a N_STSYM entry for any data symbols. 3421 // If we do find a match, and the name matches, then we 3422 // can merge the two into just the Static symbol to avoid 3423 // duplicate entries in the symbol table 3424 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3425 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 3426 if (range.first != range.second) 3427 { 3428 bool found_it = false; 3429 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3430 { 3431 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3432 { 3433 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3434 // We just need the flags from the linker symbol, so put these flags 3435 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3436 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3437 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3438 sym[sym_idx].Clear(); 3439 found_it = true; 3440 break; 3441 } 3442 } 3443 if (found_it) 3444 continue; 3445 } 3446 else 3447 { 3448 // Combine N_GSYM stab entries with the non stab symbol 3449 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 3450 if (pos != N_GSYM_name_to_sym_idx.end()) 3451 { 3452 const uint32_t GSYM_sym_idx = pos->second; 3453 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 3454 // Copy the address, because often the N_GSYM address has an invalid address of zero 3455 // when the global is a common symbol 3456 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 3457 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 3458 // We just need the flags from the linker symbol, so put these flags 3459 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3460 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3461 sym[sym_idx].Clear(); 3462 continue; 3463 } 3464 } 3465 } 3466 } 3467 3468 sym[sym_idx].SetID (nlist_idx); 3469 sym[sym_idx].SetType (type); 3470 sym[sym_idx].GetAddress().SetSection (symbol_section); 3471 sym[sym_idx].GetAddress().SetOffset (symbol_value); 3472 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3473 3474 if (symbol_byte_size > 0) 3475 sym[sym_idx].SetByteSize(symbol_byte_size); 3476 3477 if (demangled_is_synthesized) 3478 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3479 ++sym_idx; 3480 } 3481 else 3482 { 3483 sym[sym_idx].Clear(); 3484 } 3485 3486 } 3487 ///////////////////////////// 3488 } 3489 break; // No more entries to consider 3490 } 3491 } 3492 } 3493 } 3494 } 3495 } 3496 } 3497 3498 // Must reset this in case it was mutated above! 3499 nlist_data_offset = 0; 3500 #endif 3501 3502 if (nlist_data.GetByteSize() > 0) 3503 { 3504 3505 // If the sym array was not created while parsing the DSC unmapped 3506 // symbols, create it now. 3507 if (sym == NULL) 3508 { 3509 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3510 num_syms = symtab->GetNumSymbols(); 3511 } 3512 3513 if (unmapped_local_symbols_found) 3514 { 3515 assert(m_dysymtab.ilocalsym == 0); 3516 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3517 nlist_idx = m_dysymtab.nlocalsym; 3518 } 3519 else 3520 { 3521 nlist_idx = 0; 3522 } 3523 3524 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 3525 { 3526 struct nlist_64 nlist; 3527 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 3528 break; 3529 3530 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3531 nlist.n_type = nlist_data.GetU8_unchecked (&nlist_data_offset); 3532 nlist.n_sect = nlist_data.GetU8_unchecked (&nlist_data_offset); 3533 nlist.n_desc = nlist_data.GetU16_unchecked (&nlist_data_offset); 3534 nlist.n_value = nlist_data.GetAddress_unchecked (&nlist_data_offset); 3535 3536 SymbolType type = eSymbolTypeInvalid; 3537 const char *symbol_name = NULL; 3538 3539 if (have_strtab_data) 3540 { 3541 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3542 3543 if (symbol_name == NULL) 3544 { 3545 // No symbol should be NULL, even the symbols with no 3546 // string values should have an offset zero which points 3547 // to an empty C-string 3548 Host::SystemLog (Host::eSystemLogError, 3549 "error: symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 3550 nlist_idx, 3551 nlist.n_strx, 3552 module_sp->GetFileSpec().GetPath().c_str()); 3553 continue; 3554 } 3555 if (symbol_name[0] == '\0') 3556 symbol_name = NULL; 3557 } 3558 else 3559 { 3560 const addr_t str_addr = strtab_addr + nlist.n_strx; 3561 Error str_error; 3562 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, str_error)) 3563 symbol_name = memory_symbol_name.c_str(); 3564 } 3565 const char *symbol_name_non_abi_mangled = NULL; 3566 3567 SectionSP symbol_section; 3568 lldb::addr_t symbol_byte_size = 0; 3569 bool add_nlist = true; 3570 bool is_gsym = false; 3571 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3572 bool demangled_is_synthesized = false; 3573 3574 assert (sym_idx < num_syms); 3575 3576 sym[sym_idx].SetDebug (is_debug); 3577 3578 if (is_debug) 3579 { 3580 switch (nlist.n_type) 3581 { 3582 case N_GSYM: 3583 // global symbol: name,,NO_SECT,type,0 3584 // Sometimes the N_GSYM value contains the address. 3585 3586 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 3587 // have the same address, but we want to ensure that we always find only the real symbol, 3588 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 3589 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 3590 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 3591 // same address. 3592 3593 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 3594 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 3595 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 3596 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 3597 add_nlist = false; 3598 else 3599 { 3600 is_gsym = true; 3601 sym[sym_idx].SetExternal(true); 3602 if (nlist.n_value != 0) 3603 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3604 type = eSymbolTypeData; 3605 } 3606 break; 3607 3608 case N_FNAME: 3609 // procedure name (f77 kludge): name,,NO_SECT,0,0 3610 type = eSymbolTypeCompiler; 3611 break; 3612 3613 case N_FUN: 3614 // procedure: name,,n_sect,linenumber,address 3615 if (symbol_name) 3616 { 3617 type = eSymbolTypeCode; 3618 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3619 3620 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3621 // We use the current number of symbols in the symbol table in lieu of 3622 // using nlist_idx in case we ever start trimming entries out 3623 N_FUN_indexes.push_back(sym_idx); 3624 } 3625 else 3626 { 3627 type = eSymbolTypeCompiler; 3628 3629 if ( !N_FUN_indexes.empty() ) 3630 { 3631 // Copy the size of the function into the original STAB entry so we don't have 3632 // to hunt for it later 3633 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 3634 N_FUN_indexes.pop_back(); 3635 // We don't really need the end function STAB as it contains the size which 3636 // we already placed with the original symbol, so don't add it if we want a 3637 // minimal symbol table 3638 add_nlist = false; 3639 } 3640 } 3641 break; 3642 3643 case N_STSYM: 3644 // static symbol: name,,n_sect,type,address 3645 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3646 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3647 type = eSymbolTypeData; 3648 break; 3649 3650 case N_LCSYM: 3651 // .lcomm symbol: name,,n_sect,type,address 3652 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3653 type = eSymbolTypeCommonBlock; 3654 break; 3655 3656 case N_BNSYM: 3657 // We use the current number of symbols in the symbol table in lieu of 3658 // using nlist_idx in case we ever start trimming entries out 3659 // Skip these if we want minimal symbol tables 3660 add_nlist = false; 3661 break; 3662 3663 case N_ENSYM: 3664 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 3665 // so that we can always skip the entire symbol if we need to navigate 3666 // more quickly at the source level when parsing STABS 3667 // Skip these if we want minimal symbol tables 3668 add_nlist = false; 3669 break; 3670 3671 3672 case N_OPT: 3673 // emitted with gcc2_compiled and in gcc source 3674 type = eSymbolTypeCompiler; 3675 break; 3676 3677 case N_RSYM: 3678 // register sym: name,,NO_SECT,type,register 3679 type = eSymbolTypeVariable; 3680 break; 3681 3682 case N_SLINE: 3683 // src line: 0,,n_sect,linenumber,address 3684 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3685 type = eSymbolTypeLineEntry; 3686 break; 3687 3688 case N_SSYM: 3689 // structure elt: name,,NO_SECT,type,struct_offset 3690 type = eSymbolTypeVariableType; 3691 break; 3692 3693 case N_SO: 3694 // source file name 3695 type = eSymbolTypeSourceFile; 3696 if (symbol_name == NULL) 3697 { 3698 add_nlist = false; 3699 if (N_SO_index != UINT32_MAX) 3700 { 3701 // Set the size of the N_SO to the terminating index of this N_SO 3702 // so that we can always skip the entire N_SO if we need to navigate 3703 // more quickly at the source level when parsing STABS 3704 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3705 symbol_ptr->SetByteSize(sym_idx); 3706 symbol_ptr->SetSizeIsSibling(true); 3707 } 3708 N_NSYM_indexes.clear(); 3709 N_INCL_indexes.clear(); 3710 N_BRAC_indexes.clear(); 3711 N_COMM_indexes.clear(); 3712 N_FUN_indexes.clear(); 3713 N_SO_index = UINT32_MAX; 3714 } 3715 else 3716 { 3717 // We use the current number of symbols in the symbol table in lieu of 3718 // using nlist_idx in case we ever start trimming entries out 3719 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3720 if (N_SO_has_full_path) 3721 { 3722 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3723 { 3724 // We have two consecutive N_SO entries where the first contains a directory 3725 // and the second contains a full path. 3726 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 3727 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3728 add_nlist = false; 3729 } 3730 else 3731 { 3732 // This is the first entry in a N_SO that contains a directory or 3733 // a full path to the source file 3734 N_SO_index = sym_idx; 3735 } 3736 } 3737 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3738 { 3739 // This is usually the second N_SO entry that contains just the filename, 3740 // so here we combine it with the first one if we are minimizing the symbol table 3741 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3742 if (so_path && so_path[0]) 3743 { 3744 std::string full_so_path (so_path); 3745 const size_t double_slash_pos = full_so_path.find("//"); 3746 if (double_slash_pos != std::string::npos) 3747 { 3748 // The linker has been generating bad N_SO entries with doubled up paths 3749 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 3750 // and the second is the directory for the source file so you end up with 3751 // a path that looks like "/tmp/src//tmp/src/" 3752 FileSpec so_dir(so_path, false); 3753 if (!so_dir.Exists()) 3754 { 3755 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 3756 if (so_dir.Exists()) 3757 { 3758 // Trim off the incorrect path 3759 full_so_path.erase(0, double_slash_pos + 1); 3760 } 3761 } 3762 } 3763 if (*full_so_path.rbegin() != '/') 3764 full_so_path += '/'; 3765 full_so_path += symbol_name; 3766 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 3767 add_nlist = false; 3768 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3769 } 3770 } 3771 else 3772 { 3773 // This could be a relative path to a N_SO 3774 N_SO_index = sym_idx; 3775 } 3776 } 3777 3778 break; 3779 3780 case N_OSO: 3781 // object file name: name,,0,0,st_mtime 3782 type = eSymbolTypeObjectFile; 3783 break; 3784 3785 case N_LSYM: 3786 // local sym: name,,NO_SECT,type,offset 3787 type = eSymbolTypeLocal; 3788 break; 3789 3790 //---------------------------------------------------------------------- 3791 // INCL scopes 3792 //---------------------------------------------------------------------- 3793 case N_BINCL: 3794 // include file beginning: name,,NO_SECT,0,sum 3795 // We use the current number of symbols in the symbol table in lieu of 3796 // using nlist_idx in case we ever start trimming entries out 3797 N_INCL_indexes.push_back(sym_idx); 3798 type = eSymbolTypeScopeBegin; 3799 break; 3800 3801 case N_EINCL: 3802 // include file end: name,,NO_SECT,0,0 3803 // Set the size of the N_BINCL to the terminating index of this N_EINCL 3804 // so that we can always skip the entire symbol if we need to navigate 3805 // more quickly at the source level when parsing STABS 3806 if ( !N_INCL_indexes.empty() ) 3807 { 3808 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3809 symbol_ptr->SetByteSize(sym_idx + 1); 3810 symbol_ptr->SetSizeIsSibling(true); 3811 N_INCL_indexes.pop_back(); 3812 } 3813 type = eSymbolTypeScopeEnd; 3814 break; 3815 3816 case N_SOL: 3817 // #included file name: name,,n_sect,0,address 3818 type = eSymbolTypeHeaderFile; 3819 3820 // We currently don't use the header files on darwin 3821 add_nlist = false; 3822 break; 3823 3824 case N_PARAMS: 3825 // compiler parameters: name,,NO_SECT,0,0 3826 type = eSymbolTypeCompiler; 3827 break; 3828 3829 case N_VERSION: 3830 // compiler version: name,,NO_SECT,0,0 3831 type = eSymbolTypeCompiler; 3832 break; 3833 3834 case N_OLEVEL: 3835 // compiler -O level: name,,NO_SECT,0,0 3836 type = eSymbolTypeCompiler; 3837 break; 3838 3839 case N_PSYM: 3840 // parameter: name,,NO_SECT,type,offset 3841 type = eSymbolTypeVariable; 3842 break; 3843 3844 case N_ENTRY: 3845 // alternate entry: name,,n_sect,linenumber,address 3846 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3847 type = eSymbolTypeLineEntry; 3848 break; 3849 3850 //---------------------------------------------------------------------- 3851 // Left and Right Braces 3852 //---------------------------------------------------------------------- 3853 case N_LBRAC: 3854 // left bracket: 0,,NO_SECT,nesting level,address 3855 // We use the current number of symbols in the symbol table in lieu of 3856 // using nlist_idx in case we ever start trimming entries out 3857 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3858 N_BRAC_indexes.push_back(sym_idx); 3859 type = eSymbolTypeScopeBegin; 3860 break; 3861 3862 case N_RBRAC: 3863 // right bracket: 0,,NO_SECT,nesting level,address 3864 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 3865 // so that we can always skip the entire symbol if we need to navigate 3866 // more quickly at the source level when parsing STABS 3867 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3868 if ( !N_BRAC_indexes.empty() ) 3869 { 3870 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3871 symbol_ptr->SetByteSize(sym_idx + 1); 3872 symbol_ptr->SetSizeIsSibling(true); 3873 N_BRAC_indexes.pop_back(); 3874 } 3875 type = eSymbolTypeScopeEnd; 3876 break; 3877 3878 case N_EXCL: 3879 // deleted include file: name,,NO_SECT,0,sum 3880 type = eSymbolTypeHeaderFile; 3881 break; 3882 3883 //---------------------------------------------------------------------- 3884 // COMM scopes 3885 //---------------------------------------------------------------------- 3886 case N_BCOMM: 3887 // begin common: name,,NO_SECT,0,0 3888 // We use the current number of symbols in the symbol table in lieu of 3889 // using nlist_idx in case we ever start trimming entries out 3890 type = eSymbolTypeScopeBegin; 3891 N_COMM_indexes.push_back(sym_idx); 3892 break; 3893 3894 case N_ECOML: 3895 // end common (local name): 0,,n_sect,0,address 3896 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3897 // Fall through 3898 3899 case N_ECOMM: 3900 // end common: name,,n_sect,0,0 3901 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 3902 // so that we can always skip the entire symbol if we need to navigate 3903 // more quickly at the source level when parsing STABS 3904 if ( !N_COMM_indexes.empty() ) 3905 { 3906 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 3907 symbol_ptr->SetByteSize(sym_idx + 1); 3908 symbol_ptr->SetSizeIsSibling(true); 3909 N_COMM_indexes.pop_back(); 3910 } 3911 type = eSymbolTypeScopeEnd; 3912 break; 3913 3914 case N_LENG: 3915 // second stab entry with length information 3916 type = eSymbolTypeAdditional; 3917 break; 3918 3919 default: break; 3920 } 3921 } 3922 else 3923 { 3924 //uint8_t n_pext = N_PEXT & nlist.n_type; 3925 uint8_t n_type = N_TYPE & nlist.n_type; 3926 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3927 3928 switch (n_type) 3929 { 3930 case N_INDR:// Fall through 3931 case N_PBUD:// Fall through 3932 case N_UNDF: 3933 type = eSymbolTypeUndefined; 3934 break; 3935 3936 case N_ABS: 3937 type = eSymbolTypeAbsolute; 3938 break; 3939 3940 case N_SECT: 3941 { 3942 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3943 3944 if (!symbol_section) 3945 { 3946 // TODO: warn about this? 3947 add_nlist = false; 3948 break; 3949 } 3950 3951 if (TEXT_eh_frame_sectID == nlist.n_sect) 3952 { 3953 type = eSymbolTypeException; 3954 } 3955 else 3956 { 3957 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 3958 3959 switch (section_type) 3960 { 3961 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 3962 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 3963 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 3964 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 3965 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 3966 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 3967 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 3968 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 3969 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 3970 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 3971 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 3972 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 3973 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 3974 default: 3975 switch (symbol_section->GetType()) 3976 { 3977 case lldb::eSectionTypeCode: 3978 type = eSymbolTypeCode; 3979 break; 3980 case eSectionTypeData: 3981 case eSectionTypeDataCString: // Inlined C string data 3982 case eSectionTypeDataCStringPointers: // Pointers to C string data 3983 case eSectionTypeDataSymbolAddress: // Address of a symbol in the symbol table 3984 case eSectionTypeData4: 3985 case eSectionTypeData8: 3986 case eSectionTypeData16: 3987 type = eSymbolTypeData; 3988 break; 3989 default: 3990 break; 3991 } 3992 break; 3993 } 3994 3995 if (type == eSymbolTypeInvalid) 3996 { 3997 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 3998 if (symbol_section->IsDescendant (text_section_sp.get())) 3999 { 4000 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4001 S_ATTR_SELF_MODIFYING_CODE | 4002 S_ATTR_SOME_INSTRUCTIONS)) 4003 type = eSymbolTypeData; 4004 else 4005 type = eSymbolTypeCode; 4006 } 4007 else 4008 if (symbol_section->IsDescendant(data_section_sp.get())) 4009 { 4010 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 4011 { 4012 type = eSymbolTypeRuntime; 4013 4014 if (symbol_name && 4015 symbol_name[0] == '_' && 4016 symbol_name[1] == 'O' && 4017 symbol_name[2] == 'B') 4018 { 4019 llvm::StringRef symbol_name_ref(symbol_name); 4020 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 4021 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 4022 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 4023 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 4024 { 4025 symbol_name_non_abi_mangled = symbol_name + 1; 4026 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 4027 type = eSymbolTypeObjCClass; 4028 demangled_is_synthesized = true; 4029 } 4030 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 4031 { 4032 symbol_name_non_abi_mangled = symbol_name + 1; 4033 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 4034 type = eSymbolTypeObjCMetaClass; 4035 demangled_is_synthesized = true; 4036 } 4037 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 4038 { 4039 symbol_name_non_abi_mangled = symbol_name + 1; 4040 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 4041 type = eSymbolTypeObjCIVar; 4042 demangled_is_synthesized = true; 4043 } 4044 } 4045 } 4046 else 4047 if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 4048 { 4049 type = eSymbolTypeException; 4050 } 4051 else 4052 { 4053 type = eSymbolTypeData; 4054 } 4055 } 4056 else 4057 if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 4058 { 4059 type = eSymbolTypeTrampoline; 4060 } 4061 else 4062 if (symbol_section->IsDescendant(objc_section_sp.get())) 4063 { 4064 type = eSymbolTypeRuntime; 4065 if (symbol_name && symbol_name[0] == '.') 4066 { 4067 llvm::StringRef symbol_name_ref(symbol_name); 4068 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 4069 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 4070 { 4071 symbol_name_non_abi_mangled = symbol_name; 4072 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4073 type = eSymbolTypeObjCClass; 4074 demangled_is_synthesized = true; 4075 } 4076 } 4077 } 4078 } 4079 } 4080 } 4081 break; 4082 } 4083 } 4084 4085 if (add_nlist) 4086 { 4087 uint64_t symbol_value = nlist.n_value; 4088 4089 if (symbol_name_non_abi_mangled) 4090 { 4091 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 4092 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 4093 } 4094 else 4095 { 4096 bool symbol_name_is_mangled = false; 4097 4098 if (symbol_name && symbol_name[0] == '_') 4099 { 4100 symbol_name_is_mangled = symbol_name[1] == '_'; 4101 symbol_name++; // Skip the leading underscore 4102 } 4103 4104 if (symbol_name) 4105 { 4106 ConstString const_symbol_name(symbol_name); 4107 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 4108 if (is_gsym && is_debug) 4109 { 4110 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 4111 } 4112 } 4113 } 4114 if (symbol_section) 4115 { 4116 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4117 if (symbol_byte_size == 0 && function_starts_count > 0) 4118 { 4119 addr_t symbol_lookup_file_addr = nlist.n_value; 4120 // Do an exact address match for non-ARM addresses, else get the closest since 4121 // the symbol might be a thumb symbol which has an address with bit zero set 4122 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 4123 if (is_arm && func_start_entry) 4124 { 4125 // Verify that the function start address is the symbol address (ARM) 4126 // or the symbol address + 1 (thumb) 4127 if (func_start_entry->addr != symbol_lookup_file_addr && 4128 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 4129 { 4130 // Not the right entry, NULL it out... 4131 func_start_entry = NULL; 4132 } 4133 } 4134 if (func_start_entry) 4135 { 4136 func_start_entry->data = true; 4137 4138 addr_t symbol_file_addr = func_start_entry->addr; 4139 if (is_arm) 4140 symbol_file_addr &= 0xfffffffffffffffeull; 4141 4142 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 4143 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 4144 if (next_func_start_entry) 4145 { 4146 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4147 // Be sure the clear the Thumb address bit when we calculate the size 4148 // from the current and next address 4149 if (is_arm) 4150 next_symbol_file_addr &= 0xfffffffffffffffeull; 4151 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 4152 } 4153 else 4154 { 4155 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4156 } 4157 } 4158 } 4159 symbol_value -= section_file_addr; 4160 } 4161 4162 if (is_debug == false) 4163 { 4164 if (type == eSymbolTypeCode) 4165 { 4166 // See if we can find a N_FUN entry for any code symbols. 4167 // If we do find a match, and the name matches, then we 4168 // can merge the two into just the function symbol to avoid 4169 // duplicate entries in the symbol table 4170 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 4171 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4172 if (range.first != range.second) 4173 { 4174 bool found_it = false; 4175 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 4176 { 4177 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 4178 { 4179 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4180 // We just need the flags from the linker symbol, so put these flags 4181 // into the N_FUN flags to avoid duplicate symbols in the symbol table 4182 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4183 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4184 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 4185 sym[pos->second].SetType (eSymbolTypeResolver); 4186 sym[sym_idx].Clear(); 4187 found_it = true; 4188 break; 4189 } 4190 } 4191 if (found_it) 4192 continue; 4193 } 4194 else 4195 { 4196 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 4197 type = eSymbolTypeResolver; 4198 } 4199 } 4200 else if (type == eSymbolTypeData) 4201 { 4202 // See if we can find a N_STSYM entry for any data symbols. 4203 // If we do find a match, and the name matches, then we 4204 // can merge the two into just the Static symbol to avoid 4205 // duplicate entries in the symbol table 4206 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 4207 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4208 if (range.first != range.second) 4209 { 4210 bool found_it = false; 4211 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 4212 { 4213 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 4214 { 4215 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4216 // We just need the flags from the linker symbol, so put these flags 4217 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 4218 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4219 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4220 sym[sym_idx].Clear(); 4221 found_it = true; 4222 break; 4223 } 4224 } 4225 if (found_it) 4226 continue; 4227 } 4228 else 4229 { 4230 // Combine N_GSYM stab entries with the non stab symbol 4231 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 4232 if (pos != N_GSYM_name_to_sym_idx.end()) 4233 { 4234 const uint32_t GSYM_sym_idx = pos->second; 4235 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4236 // Copy the address, because often the N_GSYM address has an invalid address of zero 4237 // when the global is a common symbol 4238 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 4239 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 4240 // We just need the flags from the linker symbol, so put these flags 4241 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 4242 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4243 sym[sym_idx].Clear(); 4244 continue; 4245 } 4246 } 4247 } 4248 } 4249 4250 sym[sym_idx].SetID (nlist_idx); 4251 sym[sym_idx].SetType (type); 4252 sym[sym_idx].GetAddress().SetSection (symbol_section); 4253 sym[sym_idx].GetAddress().SetOffset (symbol_value); 4254 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4255 4256 if (symbol_byte_size > 0) 4257 sym[sym_idx].SetByteSize(symbol_byte_size); 4258 4259 if (demangled_is_synthesized) 4260 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4261 4262 ++sym_idx; 4263 } 4264 else 4265 { 4266 sym[sym_idx].Clear(); 4267 } 4268 } 4269 } 4270 4271 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4272 4273 if (function_starts_count > 0) 4274 { 4275 char synthetic_function_symbol[PATH_MAX]; 4276 uint32_t num_synthetic_function_symbols = 0; 4277 for (i=0; i<function_starts_count; ++i) 4278 { 4279 if (function_starts.GetEntryRef (i).data == false) 4280 ++num_synthetic_function_symbols; 4281 } 4282 4283 if (num_synthetic_function_symbols > 0) 4284 { 4285 if (num_syms < sym_idx + num_synthetic_function_symbols) 4286 { 4287 num_syms = sym_idx + num_synthetic_function_symbols; 4288 sym = symtab->Resize (num_syms); 4289 } 4290 uint32_t synthetic_function_symbol_idx = 0; 4291 for (i=0; i<function_starts_count; ++i) 4292 { 4293 const FunctionStarts::Entry *func_start_entry = function_starts.GetEntryAtIndex (i); 4294 if (func_start_entry->data == false) 4295 { 4296 addr_t symbol_file_addr = func_start_entry->addr; 4297 uint32_t symbol_flags = 0; 4298 if (is_arm) 4299 { 4300 if (symbol_file_addr & 1) 4301 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4302 symbol_file_addr &= 0xfffffffffffffffeull; 4303 } 4304 Address symbol_addr; 4305 if (module_sp->ResolveFileAddress (symbol_file_addr, symbol_addr)) 4306 { 4307 SectionSP symbol_section (symbol_addr.GetSection()); 4308 uint32_t symbol_byte_size = 0; 4309 if (symbol_section) 4310 { 4311 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4312 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 4313 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 4314 if (next_func_start_entry) 4315 { 4316 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4317 if (is_arm) 4318 next_symbol_file_addr &= 0xfffffffffffffffeull; 4319 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 4320 } 4321 else 4322 { 4323 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4324 } 4325 snprintf (synthetic_function_symbol, 4326 sizeof(synthetic_function_symbol), 4327 "___lldb_unnamed_function%u$$%s", 4328 ++synthetic_function_symbol_idx, 4329 module_sp->GetFileSpec().GetFilename().GetCString()); 4330 sym[sym_idx].SetID (synthetic_sym_id++); 4331 sym[sym_idx].GetMangled().SetDemangledName(ConstString(synthetic_function_symbol)); 4332 sym[sym_idx].SetType (eSymbolTypeCode); 4333 sym[sym_idx].SetIsSynthetic (true); 4334 sym[sym_idx].GetAddress() = symbol_addr; 4335 if (symbol_flags) 4336 sym[sym_idx].SetFlags (symbol_flags); 4337 if (symbol_byte_size) 4338 sym[sym_idx].SetByteSize (symbol_byte_size); 4339 ++sym_idx; 4340 } 4341 } 4342 } 4343 } 4344 } 4345 } 4346 4347 // Trim our symbols down to just what we ended up with after 4348 // removing any symbols. 4349 if (sym_idx < num_syms) 4350 { 4351 num_syms = sym_idx; 4352 sym = symtab->Resize (num_syms); 4353 } 4354 4355 // Now synthesize indirect symbols 4356 if (m_dysymtab.nindirectsyms != 0) 4357 { 4358 if (indirect_symbol_index_data.GetByteSize()) 4359 { 4360 NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end(); 4361 4362 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx) 4363 { 4364 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == S_SYMBOL_STUBS) 4365 { 4366 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4367 if (symbol_stub_byte_size == 0) 4368 continue; 4369 4370 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4371 4372 if (num_symbol_stubs == 0) 4373 continue; 4374 4375 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1; 4376 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) 4377 { 4378 const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx; 4379 const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size); 4380 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4381 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4)) 4382 { 4383 const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset); 4384 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4385 continue; 4386 4387 NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id); 4388 Symbol *stub_symbol = NULL; 4389 if (index_pos != end_index_pos) 4390 { 4391 // We have a remapping from the original nlist index to 4392 // a current symbol index, so just look this up by index 4393 stub_symbol = symtab->SymbolAtIndex (index_pos->second); 4394 } 4395 else 4396 { 4397 // We need to lookup a symbol using the original nlist 4398 // symbol index since this index is coming from the 4399 // S_SYMBOL_STUBS 4400 stub_symbol = symtab->FindSymbolByID (stub_sym_id); 4401 } 4402 4403 if (stub_symbol) 4404 { 4405 Address so_addr(symbol_stub_addr, section_list); 4406 4407 if (stub_symbol->GetType() == eSymbolTypeUndefined) 4408 { 4409 // Change the external symbol into a trampoline that makes sense 4410 // These symbols were N_UNDF N_EXT, and are useless to us, so we 4411 // can re-use them so we don't have to make up a synthetic symbol 4412 // for no good reason. 4413 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 4414 stub_symbol->SetType (eSymbolTypeTrampoline); 4415 else 4416 stub_symbol->SetType (eSymbolTypeResolver); 4417 stub_symbol->SetExternal (false); 4418 stub_symbol->GetAddress() = so_addr; 4419 stub_symbol->SetByteSize (symbol_stub_byte_size); 4420 } 4421 else 4422 { 4423 // Make a synthetic symbol to describe the trampoline stub 4424 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4425 if (sym_idx >= num_syms) 4426 { 4427 sym = symtab->Resize (++num_syms); 4428 stub_symbol = NULL; // this pointer no longer valid 4429 } 4430 sym[sym_idx].SetID (synthetic_sym_id++); 4431 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4432 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 4433 sym[sym_idx].SetType (eSymbolTypeTrampoline); 4434 else 4435 sym[sym_idx].SetType (eSymbolTypeResolver); 4436 sym[sym_idx].SetIsSynthetic (true); 4437 sym[sym_idx].GetAddress() = so_addr; 4438 sym[sym_idx].SetByteSize (symbol_stub_byte_size); 4439 ++sym_idx; 4440 } 4441 } 4442 else 4443 { 4444 if (log) 4445 log->Warning ("symbol stub referencing symbol table symbol %u that isn't in our minimal symbol table, fix this!!!", stub_sym_id); 4446 } 4447 } 4448 } 4449 } 4450 } 4451 } 4452 } 4453 4454 4455 if (!trie_entries.empty()) 4456 { 4457 for (const auto &e : trie_entries) 4458 { 4459 if (e.entry.import_name) 4460 { 4461 // Make a synthetic symbol to describe re-exported symbol. 4462 if (sym_idx >= num_syms) 4463 sym = symtab->Resize (++num_syms); 4464 sym[sym_idx].SetID (synthetic_sym_id++); 4465 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4466 sym[sym_idx].SetType (eSymbolTypeReExported); 4467 sym[sym_idx].SetIsSynthetic (true); 4468 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4469 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) 4470 { 4471 sym[sym_idx].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(e.entry.other-1)); 4472 } 4473 ++sym_idx; 4474 } 4475 } 4476 } 4477 4478 4479 4480 // StreamFile s(stdout, false); 4481 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4482 // symtab->Dump(&s, NULL, eSortOrderNone); 4483 // Set symbol byte sizes correctly since mach-o nlist entries don't have sizes 4484 symtab->CalculateSymbolSizes(); 4485 4486 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4487 // symtab->Dump(&s, NULL, eSortOrderNone); 4488 4489 return symtab->GetNumSymbols(); 4490 } 4491 return 0; 4492 } 4493 4494 4495 void 4496 ObjectFileMachO::Dump (Stream *s) 4497 { 4498 ModuleSP module_sp(GetModule()); 4499 if (module_sp) 4500 { 4501 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4502 s->Printf("%p: ", static_cast<void*>(this)); 4503 s->Indent(); 4504 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4505 s->PutCString("ObjectFileMachO64"); 4506 else 4507 s->PutCString("ObjectFileMachO32"); 4508 4509 ArchSpec header_arch; 4510 GetArchitecture(header_arch); 4511 4512 *s << ", file = '" << m_file << "', arch = " << header_arch.GetArchitectureName() << "\n"; 4513 4514 SectionList *sections = GetSectionList(); 4515 if (sections) 4516 sections->Dump(s, NULL, true, UINT32_MAX); 4517 4518 if (m_symtab_ap.get()) 4519 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4520 } 4521 } 4522 4523 bool 4524 ObjectFileMachO::GetUUID (const llvm::MachO::mach_header &header, 4525 const lldb_private::DataExtractor &data, 4526 lldb::offset_t lc_offset, 4527 lldb_private::UUID& uuid) 4528 { 4529 uint32_t i; 4530 struct uuid_command load_cmd; 4531 4532 lldb::offset_t offset = lc_offset; 4533 for (i=0; i<header.ncmds; ++i) 4534 { 4535 const lldb::offset_t cmd_offset = offset; 4536 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4537 break; 4538 4539 if (load_cmd.cmd == LC_UUID) 4540 { 4541 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4542 4543 if (uuid_bytes) 4544 { 4545 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4546 // We pretend these object files have no UUID to prevent crashing. 4547 4548 const uint8_t opencl_uuid[] = { 0x8c, 0x8e, 0xb3, 0x9b, 4549 0x3b, 0xa8, 4550 0x4b, 0x16, 4551 0xb6, 0xa4, 4552 0x27, 0x63, 0xbb, 0x14, 0xf0, 0x0d }; 4553 4554 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4555 return false; 4556 4557 uuid.SetBytes (uuid_bytes); 4558 return true; 4559 } 4560 return false; 4561 } 4562 offset = cmd_offset + load_cmd.cmdsize; 4563 } 4564 return false; 4565 } 4566 4567 4568 bool 4569 ObjectFileMachO::GetArchitecture (const llvm::MachO::mach_header &header, 4570 const lldb_private::DataExtractor &data, 4571 lldb::offset_t lc_offset, 4572 ArchSpec &arch) 4573 { 4574 arch.SetArchitecture (eArchTypeMachO, header.cputype, header.cpusubtype); 4575 4576 if (arch.IsValid()) 4577 { 4578 llvm::Triple &triple = arch.GetTriple(); 4579 if (header.filetype == MH_PRELOAD) 4580 { 4581 // Set OS to "unknown" - this is a standalone binary with no dyld et al 4582 triple.setOS(llvm::Triple::UnknownOS); 4583 return true; 4584 } 4585 else 4586 { 4587 struct load_command load_cmd; 4588 4589 lldb::offset_t offset = lc_offset; 4590 for (uint32_t i=0; i<header.ncmds; ++i) 4591 { 4592 const lldb::offset_t cmd_offset = offset; 4593 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4594 break; 4595 4596 switch (load_cmd.cmd) 4597 { 4598 case LC_VERSION_MIN_IPHONEOS: 4599 triple.setOS (llvm::Triple::IOS); 4600 return true; 4601 4602 case LC_VERSION_MIN_MACOSX: 4603 triple.setOS (llvm::Triple::MacOSX); 4604 return true; 4605 4606 default: 4607 break; 4608 } 4609 4610 offset = cmd_offset + load_cmd.cmdsize; 4611 } 4612 4613 // Only set the OS to iOS for ARM, we don't want to set it for x86 and x86_64. 4614 // We do this because we now have MacOSX or iOS as the OS value for x86 and 4615 // x86_64 for normal desktop (MacOSX) and simulator (iOS) binaries. And if 4616 // we compare a "x86_64-apple-ios" to a "x86_64-apple-" triple, it will say 4617 // it is compatible (because the OS is unspecified in the second one and will 4618 // match anything in the first 4619 if (header.cputype == CPU_TYPE_ARM || header.cputype == CPU_TYPE_ARM64) 4620 triple.setOS (llvm::Triple::IOS); 4621 } 4622 } 4623 return arch.IsValid(); 4624 } 4625 4626 bool 4627 ObjectFileMachO::GetUUID (lldb_private::UUID* uuid) 4628 { 4629 ModuleSP module_sp(GetModule()); 4630 if (module_sp) 4631 { 4632 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4633 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4634 return GetUUID (m_header, m_data, offset, *uuid); 4635 } 4636 return false; 4637 } 4638 4639 4640 uint32_t 4641 ObjectFileMachO::GetDependentModules (FileSpecList& files) 4642 { 4643 uint32_t count = 0; 4644 ModuleSP module_sp(GetModule()); 4645 if (module_sp) 4646 { 4647 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4648 struct load_command load_cmd; 4649 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4650 const bool resolve_path = false; // Don't resolve the dependent file paths since they may not reside on this system 4651 uint32_t i; 4652 for (i=0; i<m_header.ncmds; ++i) 4653 { 4654 const uint32_t cmd_offset = offset; 4655 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4656 break; 4657 4658 switch (load_cmd.cmd) 4659 { 4660 case LC_LOAD_DYLIB: 4661 case LC_LOAD_WEAK_DYLIB: 4662 case LC_REEXPORT_DYLIB: 4663 case LC_LOAD_DYLINKER: 4664 case LC_LOADFVMLIB: 4665 case LC_LOAD_UPWARD_DYLIB: 4666 { 4667 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 4668 const char *path = m_data.PeekCStr(name_offset); 4669 // Skip any path that starts with '@' since these are usually: 4670 // @executable_path/.../file 4671 // @rpath/.../file 4672 if (path && path[0] != '@') 4673 { 4674 FileSpec file_spec(path, resolve_path); 4675 if (files.AppendIfUnique(file_spec)) 4676 count++; 4677 } 4678 } 4679 break; 4680 4681 default: 4682 break; 4683 } 4684 offset = cmd_offset + load_cmd.cmdsize; 4685 } 4686 } 4687 return count; 4688 } 4689 4690 lldb_private::Address 4691 ObjectFileMachO::GetEntryPointAddress () 4692 { 4693 // If the object file is not an executable it can't hold the entry point. m_entry_point_address 4694 // is initialized to an invalid address, so we can just return that. 4695 // If m_entry_point_address is valid it means we've found it already, so return the cached value. 4696 4697 if (!IsExecutable() || m_entry_point_address.IsValid()) 4698 return m_entry_point_address; 4699 4700 // Otherwise, look for the UnixThread or Thread command. The data for the Thread command is given in 4701 // /usr/include/mach-o.h, but it is basically: 4702 // 4703 // uint32_t flavor - this is the flavor argument you would pass to thread_get_state 4704 // uint32_t count - this is the count of longs in the thread state data 4705 // struct XXX_thread_state state - this is the structure from <machine/thread_status.h> corresponding to the flavor. 4706 // <repeat this trio> 4707 // 4708 // So we just keep reading the various register flavors till we find the GPR one, then read the PC out of there. 4709 // FIXME: We will need to have a "RegisterContext data provider" class at some point that can get all the registers 4710 // out of data in this form & attach them to a given thread. That should underlie the MacOS X User process plugin, 4711 // and we'll also need it for the MacOS X Core File process plugin. When we have that we can also use it here. 4712 // 4713 // For now we hard-code the offsets and flavors we need: 4714 // 4715 // 4716 4717 ModuleSP module_sp(GetModule()); 4718 if (module_sp) 4719 { 4720 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4721 struct load_command load_cmd; 4722 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4723 uint32_t i; 4724 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 4725 bool done = false; 4726 4727 for (i=0; i<m_header.ncmds; ++i) 4728 { 4729 const lldb::offset_t cmd_offset = offset; 4730 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4731 break; 4732 4733 switch (load_cmd.cmd) 4734 { 4735 case LC_UNIXTHREAD: 4736 case LC_THREAD: 4737 { 4738 while (offset < cmd_offset + load_cmd.cmdsize) 4739 { 4740 uint32_t flavor = m_data.GetU32(&offset); 4741 uint32_t count = m_data.GetU32(&offset); 4742 if (count == 0) 4743 { 4744 // We've gotten off somehow, log and exit; 4745 return m_entry_point_address; 4746 } 4747 4748 switch (m_header.cputype) 4749 { 4750 case llvm::MachO::CPU_TYPE_ARM: 4751 if (flavor == 1) // ARM_THREAD_STATE from mach/arm/thread_status.h 4752 { 4753 offset += 60; // This is the offset of pc in the GPR thread state data structure. 4754 start_address = m_data.GetU32(&offset); 4755 done = true; 4756 } 4757 break; 4758 case llvm::MachO::CPU_TYPE_ARM64: 4759 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 4760 { 4761 offset += 256; // This is the offset of pc in the GPR thread state data structure. 4762 start_address = m_data.GetU64(&offset); 4763 done = true; 4764 } 4765 break; 4766 case llvm::MachO::CPU_TYPE_I386: 4767 if (flavor == 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 4768 { 4769 offset += 40; // This is the offset of eip in the GPR thread state data structure. 4770 start_address = m_data.GetU32(&offset); 4771 done = true; 4772 } 4773 break; 4774 case llvm::MachO::CPU_TYPE_X86_64: 4775 if (flavor == 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 4776 { 4777 offset += 16 * 8; // This is the offset of rip in the GPR thread state data structure. 4778 start_address = m_data.GetU64(&offset); 4779 done = true; 4780 } 4781 break; 4782 default: 4783 return m_entry_point_address; 4784 } 4785 // Haven't found the GPR flavor yet, skip over the data for this flavor: 4786 if (done) 4787 break; 4788 offset += count * 4; 4789 } 4790 } 4791 break; 4792 case LC_MAIN: 4793 { 4794 ConstString text_segment_name ("__TEXT"); 4795 uint64_t entryoffset = m_data.GetU64(&offset); 4796 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 4797 if (text_segment_sp) 4798 { 4799 done = true; 4800 start_address = text_segment_sp->GetFileAddress() + entryoffset; 4801 } 4802 } 4803 4804 default: 4805 break; 4806 } 4807 if (done) 4808 break; 4809 4810 // Go to the next load command: 4811 offset = cmd_offset + load_cmd.cmdsize; 4812 } 4813 4814 if (start_address != LLDB_INVALID_ADDRESS) 4815 { 4816 // We got the start address from the load commands, so now resolve that address in the sections 4817 // of this ObjectFile: 4818 if (!m_entry_point_address.ResolveAddressUsingFileSections (start_address, GetSectionList())) 4819 { 4820 m_entry_point_address.Clear(); 4821 } 4822 } 4823 else 4824 { 4825 // We couldn't read the UnixThread load command - maybe it wasn't there. As a fallback look for the 4826 // "start" symbol in the main executable. 4827 4828 ModuleSP module_sp (GetModule()); 4829 4830 if (module_sp) 4831 { 4832 SymbolContextList contexts; 4833 SymbolContext context; 4834 if (module_sp->FindSymbolsWithNameAndType(ConstString ("start"), eSymbolTypeCode, contexts)) 4835 { 4836 if (contexts.GetContextAtIndex(0, context)) 4837 m_entry_point_address = context.symbol->GetAddress(); 4838 } 4839 } 4840 } 4841 } 4842 4843 return m_entry_point_address; 4844 4845 } 4846 4847 lldb_private::Address 4848 ObjectFileMachO::GetHeaderAddress () 4849 { 4850 lldb_private::Address header_addr; 4851 SectionList *section_list = GetSectionList(); 4852 if (section_list) 4853 { 4854 SectionSP text_segment_sp (section_list->FindSectionByName (GetSegmentNameTEXT())); 4855 if (text_segment_sp) 4856 { 4857 header_addr.SetSection (text_segment_sp); 4858 header_addr.SetOffset (0); 4859 } 4860 } 4861 return header_addr; 4862 } 4863 4864 uint32_t 4865 ObjectFileMachO::GetNumThreadContexts () 4866 { 4867 ModuleSP module_sp(GetModule()); 4868 if (module_sp) 4869 { 4870 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4871 if (!m_thread_context_offsets_valid) 4872 { 4873 m_thread_context_offsets_valid = true; 4874 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4875 FileRangeArray::Entry file_range; 4876 thread_command thread_cmd; 4877 for (uint32_t i=0; i<m_header.ncmds; ++i) 4878 { 4879 const uint32_t cmd_offset = offset; 4880 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 4881 break; 4882 4883 if (thread_cmd.cmd == LC_THREAD) 4884 { 4885 file_range.SetRangeBase (offset); 4886 file_range.SetByteSize (thread_cmd.cmdsize - 8); 4887 m_thread_context_offsets.Append (file_range); 4888 } 4889 offset = cmd_offset + thread_cmd.cmdsize; 4890 } 4891 } 4892 } 4893 return m_thread_context_offsets.GetSize(); 4894 } 4895 4896 lldb::RegisterContextSP 4897 ObjectFileMachO::GetThreadContextAtIndex (uint32_t idx, lldb_private::Thread &thread) 4898 { 4899 lldb::RegisterContextSP reg_ctx_sp; 4900 4901 ModuleSP module_sp(GetModule()); 4902 if (module_sp) 4903 { 4904 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4905 if (!m_thread_context_offsets_valid) 4906 GetNumThreadContexts (); 4907 4908 const FileRangeArray::Entry *thread_context_file_range = m_thread_context_offsets.GetEntryAtIndex (idx); 4909 if (thread_context_file_range) 4910 { 4911 4912 DataExtractor data (m_data, 4913 thread_context_file_range->GetRangeBase(), 4914 thread_context_file_range->GetByteSize()); 4915 4916 switch (m_header.cputype) 4917 { 4918 case llvm::MachO::CPU_TYPE_ARM64: 4919 reg_ctx_sp.reset (new RegisterContextDarwin_arm64_Mach (thread, data)); 4920 break; 4921 4922 case llvm::MachO::CPU_TYPE_ARM: 4923 reg_ctx_sp.reset (new RegisterContextDarwin_arm_Mach (thread, data)); 4924 break; 4925 4926 case llvm::MachO::CPU_TYPE_I386: 4927 reg_ctx_sp.reset (new RegisterContextDarwin_i386_Mach (thread, data)); 4928 break; 4929 4930 case llvm::MachO::CPU_TYPE_X86_64: 4931 reg_ctx_sp.reset (new RegisterContextDarwin_x86_64_Mach (thread, data)); 4932 break; 4933 } 4934 } 4935 } 4936 return reg_ctx_sp; 4937 } 4938 4939 4940 ObjectFile::Type 4941 ObjectFileMachO::CalculateType() 4942 { 4943 switch (m_header.filetype) 4944 { 4945 case MH_OBJECT: // 0x1u 4946 if (GetAddressByteSize () == 4) 4947 { 4948 // 32 bit kexts are just object files, but they do have a valid 4949 // UUID load command. 4950 UUID uuid; 4951 if (GetUUID(&uuid)) 4952 { 4953 // this checking for the UUID load command is not enough 4954 // we could eventually look for the symbol named 4955 // "OSKextGetCurrentIdentifier" as this is required of kexts 4956 if (m_strata == eStrataInvalid) 4957 m_strata = eStrataKernel; 4958 return eTypeSharedLibrary; 4959 } 4960 } 4961 return eTypeObjectFile; 4962 4963 case MH_EXECUTE: return eTypeExecutable; // 0x2u 4964 case MH_FVMLIB: return eTypeSharedLibrary; // 0x3u 4965 case MH_CORE: return eTypeCoreFile; // 0x4u 4966 case MH_PRELOAD: return eTypeSharedLibrary; // 0x5u 4967 case MH_DYLIB: return eTypeSharedLibrary; // 0x6u 4968 case MH_DYLINKER: return eTypeDynamicLinker; // 0x7u 4969 case MH_BUNDLE: return eTypeSharedLibrary; // 0x8u 4970 case MH_DYLIB_STUB: return eTypeStubLibrary; // 0x9u 4971 case MH_DSYM: return eTypeDebugInfo; // 0xAu 4972 case MH_KEXT_BUNDLE: return eTypeSharedLibrary; // 0xBu 4973 default: 4974 break; 4975 } 4976 return eTypeUnknown; 4977 } 4978 4979 ObjectFile::Strata 4980 ObjectFileMachO::CalculateStrata() 4981 { 4982 switch (m_header.filetype) 4983 { 4984 case MH_OBJECT: // 0x1u 4985 { 4986 // 32 bit kexts are just object files, but they do have a valid 4987 // UUID load command. 4988 UUID uuid; 4989 if (GetUUID(&uuid)) 4990 { 4991 // this checking for the UUID load command is not enough 4992 // we could eventually look for the symbol named 4993 // "OSKextGetCurrentIdentifier" as this is required of kexts 4994 if (m_type == eTypeInvalid) 4995 m_type = eTypeSharedLibrary; 4996 4997 return eStrataKernel; 4998 } 4999 } 5000 return eStrataUnknown; 5001 5002 case MH_EXECUTE: // 0x2u 5003 // Check for the MH_DYLDLINK bit in the flags 5004 if (m_header.flags & MH_DYLDLINK) 5005 { 5006 return eStrataUser; 5007 } 5008 else 5009 { 5010 SectionList *section_list = GetSectionList(); 5011 if (section_list) 5012 { 5013 static ConstString g_kld_section_name ("__KLD"); 5014 if (section_list->FindSectionByName(g_kld_section_name)) 5015 return eStrataKernel; 5016 } 5017 } 5018 return eStrataRawImage; 5019 5020 case MH_FVMLIB: return eStrataUser; // 0x3u 5021 case MH_CORE: return eStrataUnknown; // 0x4u 5022 case MH_PRELOAD: return eStrataRawImage; // 0x5u 5023 case MH_DYLIB: return eStrataUser; // 0x6u 5024 case MH_DYLINKER: return eStrataUser; // 0x7u 5025 case MH_BUNDLE: return eStrataUser; // 0x8u 5026 case MH_DYLIB_STUB: return eStrataUser; // 0x9u 5027 case MH_DSYM: return eStrataUnknown; // 0xAu 5028 case MH_KEXT_BUNDLE: return eStrataKernel; // 0xBu 5029 default: 5030 break; 5031 } 5032 return eStrataUnknown; 5033 } 5034 5035 5036 uint32_t 5037 ObjectFileMachO::GetVersion (uint32_t *versions, uint32_t num_versions) 5038 { 5039 ModuleSP module_sp(GetModule()); 5040 if (module_sp) 5041 { 5042 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 5043 struct dylib_command load_cmd; 5044 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5045 uint32_t version_cmd = 0; 5046 uint64_t version = 0; 5047 uint32_t i; 5048 for (i=0; i<m_header.ncmds; ++i) 5049 { 5050 const lldb::offset_t cmd_offset = offset; 5051 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5052 break; 5053 5054 if (load_cmd.cmd == LC_ID_DYLIB) 5055 { 5056 if (version_cmd == 0) 5057 { 5058 version_cmd = load_cmd.cmd; 5059 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5060 break; 5061 version = load_cmd.dylib.current_version; 5062 } 5063 break; // Break for now unless there is another more complete version 5064 // number load command in the future. 5065 } 5066 offset = cmd_offset + load_cmd.cmdsize; 5067 } 5068 5069 if (version_cmd == LC_ID_DYLIB) 5070 { 5071 if (versions != NULL && num_versions > 0) 5072 { 5073 if (num_versions > 0) 5074 versions[0] = (version & 0xFFFF0000ull) >> 16; 5075 if (num_versions > 1) 5076 versions[1] = (version & 0x0000FF00ull) >> 8; 5077 if (num_versions > 2) 5078 versions[2] = (version & 0x000000FFull); 5079 // Fill in an remaining version numbers with invalid values 5080 for (i=3; i<num_versions; ++i) 5081 versions[i] = UINT32_MAX; 5082 } 5083 // The LC_ID_DYLIB load command has a version with 3 version numbers 5084 // in it, so always return 3 5085 return 3; 5086 } 5087 } 5088 return false; 5089 } 5090 5091 bool 5092 ObjectFileMachO::GetArchitecture (ArchSpec &arch) 5093 { 5094 ModuleSP module_sp(GetModule()); 5095 if (module_sp) 5096 { 5097 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 5098 return GetArchitecture (m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), arch); 5099 } 5100 return false; 5101 } 5102 5103 5104 UUID 5105 ObjectFileMachO::GetProcessSharedCacheUUID (Process *process) 5106 { 5107 UUID uuid; 5108 if (process) 5109 { 5110 addr_t all_image_infos = process->GetImageInfoAddress(); 5111 5112 // The address returned by GetImageInfoAddress may be the address of dyld (don't want) 5113 // or it may be the address of the dyld_all_image_infos structure (want). The first four 5114 // bytes will be either the version field (all_image_infos) or a Mach-O file magic constant. 5115 // Version 13 and higher of dyld_all_image_infos is required to get the sharedCacheUUID field. 5116 5117 Error err; 5118 uint32_t version_or_magic = process->ReadUnsignedIntegerFromMemory (all_image_infos, 4, -1, err); 5119 if (version_or_magic != static_cast<uint32_t>(-1) 5120 && version_or_magic != MH_MAGIC 5121 && version_or_magic != MH_CIGAM 5122 && version_or_magic != MH_MAGIC_64 5123 && version_or_magic != MH_CIGAM_64 5124 && version_or_magic >= 13) 5125 { 5126 addr_t sharedCacheUUID_address = LLDB_INVALID_ADDRESS; 5127 int wordsize = process->GetAddressByteSize(); 5128 if (wordsize == 8) 5129 { 5130 sharedCacheUUID_address = all_image_infos + 160; // sharedCacheUUID <mach-o/dyld_images.h> 5131 } 5132 if (wordsize == 4) 5133 { 5134 sharedCacheUUID_address = all_image_infos + 84; // sharedCacheUUID <mach-o/dyld_images.h> 5135 } 5136 if (sharedCacheUUID_address != LLDB_INVALID_ADDRESS) 5137 { 5138 uuid_t shared_cache_uuid; 5139 if (process->ReadMemory (sharedCacheUUID_address, shared_cache_uuid, sizeof (uuid_t), err) == sizeof (uuid_t)) 5140 { 5141 uuid.SetBytes (shared_cache_uuid); 5142 } 5143 } 5144 } 5145 } 5146 return uuid; 5147 } 5148 5149 UUID 5150 ObjectFileMachO::GetLLDBSharedCacheUUID () 5151 { 5152 UUID uuid; 5153 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 5154 uint8_t *(*dyld_get_all_image_infos)(void); 5155 dyld_get_all_image_infos = (uint8_t*(*)()) dlsym (RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5156 if (dyld_get_all_image_infos) 5157 { 5158 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5159 if (dyld_all_image_infos_address) 5160 { 5161 uint32_t *version = (uint32_t*) dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5162 if (*version >= 13) 5163 { 5164 uuid_t *sharedCacheUUID_address = 0; 5165 int wordsize = sizeof (uint8_t *); 5166 if (wordsize == 8) 5167 { 5168 sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 160); // sharedCacheUUID <mach-o/dyld_images.h> 5169 } 5170 else 5171 { 5172 sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 84); // sharedCacheUUID <mach-o/dyld_images.h> 5173 } 5174 uuid.SetBytes (sharedCacheUUID_address); 5175 } 5176 } 5177 } 5178 #endif 5179 return uuid; 5180 } 5181 5182 uint32_t 5183 ObjectFileMachO::GetMinimumOSVersion (uint32_t *versions, uint32_t num_versions) 5184 { 5185 if (m_min_os_versions.empty()) 5186 { 5187 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5188 bool success = false; 5189 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 5190 { 5191 const lldb::offset_t load_cmd_offset = offset; 5192 5193 version_min_command lc; 5194 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5195 break; 5196 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 5197 { 5198 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 5199 { 5200 const uint32_t xxxx = lc.version >> 16; 5201 const uint32_t yy = (lc.version >> 8) & 0xffu; 5202 const uint32_t zz = lc.version & 0xffu; 5203 if (xxxx) 5204 { 5205 m_min_os_versions.push_back(xxxx); 5206 m_min_os_versions.push_back(yy); 5207 m_min_os_versions.push_back(zz); 5208 } 5209 success = true; 5210 } 5211 } 5212 offset = load_cmd_offset + lc.cmdsize; 5213 } 5214 5215 if (success == false) 5216 { 5217 // Push an invalid value so we don't keep trying to 5218 m_min_os_versions.push_back(UINT32_MAX); 5219 } 5220 } 5221 5222 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) 5223 { 5224 if (versions != NULL && num_versions > 0) 5225 { 5226 for (size_t i=0; i<num_versions; ++i) 5227 { 5228 if (i < m_min_os_versions.size()) 5229 versions[i] = m_min_os_versions[i]; 5230 else 5231 versions[i] = 0; 5232 } 5233 } 5234 return m_min_os_versions.size(); 5235 } 5236 // Call the superclasses version that will empty out the data 5237 return ObjectFile::GetMinimumOSVersion (versions, num_versions); 5238 } 5239 5240 uint32_t 5241 ObjectFileMachO::GetSDKVersion(uint32_t *versions, uint32_t num_versions) 5242 { 5243 if (m_sdk_versions.empty()) 5244 { 5245 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5246 bool success = false; 5247 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 5248 { 5249 const lldb::offset_t load_cmd_offset = offset; 5250 5251 version_min_command lc; 5252 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5253 break; 5254 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 5255 { 5256 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 5257 { 5258 const uint32_t xxxx = lc.sdk >> 16; 5259 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5260 const uint32_t zz = lc.sdk & 0xffu; 5261 if (xxxx) 5262 { 5263 m_sdk_versions.push_back(xxxx); 5264 m_sdk_versions.push_back(yy); 5265 m_sdk_versions.push_back(zz); 5266 } 5267 success = true; 5268 } 5269 } 5270 offset = load_cmd_offset + lc.cmdsize; 5271 } 5272 5273 if (success == false) 5274 { 5275 // Push an invalid value so we don't keep trying to 5276 m_sdk_versions.push_back(UINT32_MAX); 5277 } 5278 } 5279 5280 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) 5281 { 5282 if (versions != NULL && num_versions > 0) 5283 { 5284 for (size_t i=0; i<num_versions; ++i) 5285 { 5286 if (i < m_sdk_versions.size()) 5287 versions[i] = m_sdk_versions[i]; 5288 else 5289 versions[i] = 0; 5290 } 5291 } 5292 return m_sdk_versions.size(); 5293 } 5294 // Call the superclasses version that will empty out the data 5295 return ObjectFile::GetSDKVersion (versions, num_versions); 5296 } 5297 5298 5299 bool 5300 ObjectFileMachO::GetIsDynamicLinkEditor() 5301 { 5302 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5303 } 5304 5305 //------------------------------------------------------------------ 5306 // PluginInterface protocol 5307 //------------------------------------------------------------------ 5308 lldb_private::ConstString 5309 ObjectFileMachO::GetPluginName() 5310 { 5311 return GetPluginNameStatic(); 5312 } 5313 5314 uint32_t 5315 ObjectFileMachO::GetPluginVersion() 5316 { 5317 return 1; 5318 } 5319 5320 5321 bool 5322 ObjectFileMachO::SetLoadAddress (Target &target, 5323 lldb::addr_t value, 5324 bool value_is_offset) 5325 { 5326 ModuleSP module_sp = GetModule(); 5327 if (module_sp) 5328 { 5329 size_t num_loaded_sections = 0; 5330 SectionList *section_list = GetSectionList (); 5331 if (section_list) 5332 { 5333 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5334 const size_t num_sections = section_list->GetSize(); 5335 5336 const bool is_memory_image = (bool)m_process_wp.lock(); 5337 const Strata strata = GetStrata(); 5338 static ConstString g_linkedit_segname ("__LINKEDIT"); 5339 if (value_is_offset) 5340 { 5341 // "value" is an offset to apply to each top level segment 5342 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) 5343 { 5344 // Iterate through the object file sections to find all 5345 // of the sections that size on disk (to avoid __PAGEZERO) 5346 // and load them 5347 SectionSP section_sp (section_list->GetSectionAtIndex (sect_idx)); 5348 if (section_sp && 5349 section_sp->GetFileSize() > 0 && 5350 section_sp->IsThreadSpecific() == false && 5351 module_sp.get() == section_sp->GetModule().get()) 5352 { 5353 // Ignore __LINKEDIT and __DWARF segments 5354 if (section_sp->GetName() == g_linkedit_segname) 5355 { 5356 // Only map __LINKEDIT if we have an in memory image and this isn't 5357 // a kernel binary like a kext or mach_kernel. 5358 if (is_memory_image == false || strata == eStrataKernel) 5359 continue; 5360 } 5361 if (target.GetSectionLoadList().SetSectionLoadAddress (section_sp, section_sp->GetFileAddress() + value)) 5362 ++num_loaded_sections; 5363 } 5364 } 5365 } 5366 else 5367 { 5368 // "value" is the new base address of the mach_header, adjust each 5369 // section accordingly 5370 5371 // First find the address of the mach header which is the first non-zero 5372 // file sized section whose file offset is zero as this will be subtracted 5373 // from each other valid section's vmaddr and then get "base_addr" added to 5374 // it when loading the module in the target 5375 for (size_t sect_idx = 0; 5376 sect_idx < num_sections && mach_base_file_addr == LLDB_INVALID_ADDRESS; 5377 ++sect_idx) 5378 { 5379 // Iterate through the object file sections to find all 5380 // of the sections that size on disk (to avoid __PAGEZERO) 5381 // and load them 5382 Section *section = section_list->GetSectionAtIndex (sect_idx).get(); 5383 if (section && 5384 section->GetFileSize() > 0 && 5385 section->GetFileOffset() == 0 && 5386 section->IsThreadSpecific() == false && 5387 module_sp.get() == section->GetModule().get()) 5388 { 5389 // Ignore __LINKEDIT and __DWARF segments 5390 if (section->GetName() == g_linkedit_segname) 5391 { 5392 // Only map __LINKEDIT if we have an in memory image and this isn't 5393 // a kernel binary like a kext or mach_kernel. 5394 if (is_memory_image == false || strata == eStrataKernel) 5395 continue; 5396 } 5397 mach_base_file_addr = section->GetFileAddress(); 5398 } 5399 } 5400 5401 if (mach_base_file_addr != LLDB_INVALID_ADDRESS) 5402 { 5403 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) 5404 { 5405 // Iterate through the object file sections to find all 5406 // of the sections that size on disk (to avoid __PAGEZERO) 5407 // and load them 5408 SectionSP section_sp (section_list->GetSectionAtIndex (sect_idx)); 5409 if (section_sp && 5410 section_sp->GetFileSize() > 0 && 5411 section_sp->IsThreadSpecific() == false && 5412 module_sp.get() == section_sp->GetModule().get()) 5413 { 5414 // Ignore __LINKEDIT and __DWARF segments 5415 if (section_sp->GetName() == g_linkedit_segname) 5416 { 5417 // Only map __LINKEDIT if we have an in memory image and this isn't 5418 // a kernel binary like a kext or mach_kernel. 5419 if (is_memory_image == false || strata == eStrataKernel) 5420 continue; 5421 } 5422 if (target.GetSectionLoadList().SetSectionLoadAddress (section_sp, section_sp->GetFileAddress() - mach_base_file_addr + value)) 5423 ++num_loaded_sections; 5424 } 5425 } 5426 } 5427 } 5428 } 5429 return num_loaded_sections > 0; 5430 } 5431 return false; 5432 } 5433 5434 bool 5435 ObjectFileMachO::SaveCore (const lldb::ProcessSP &process_sp, 5436 const FileSpec &outfile, 5437 Error &error) 5438 { 5439 if (process_sp) 5440 { 5441 Target &target = process_sp->GetTarget(); 5442 const ArchSpec target_arch = target.GetArchitecture(); 5443 const llvm::Triple &target_triple = target_arch.GetTriple(); 5444 if (target_triple.getVendor() == llvm::Triple::Apple && 5445 (target_triple.getOS() == llvm::Triple::MacOSX || 5446 target_triple.getOS() == llvm::Triple::IOS)) 5447 { 5448 bool make_core = false; 5449 switch (target_arch.GetMachine()) 5450 { 5451 // arm64 core file writing is having some problem with writing down the 5452 // dyld shared images info struct and/or the main executable binary. May 5453 // turn out to be a debugserver problem, not sure yet. 5454 // case llvm::Triple::aarch64: 5455 5456 case llvm::Triple::arm: 5457 case llvm::Triple::x86: 5458 case llvm::Triple::x86_64: 5459 make_core = true; 5460 break; 5461 default: 5462 error.SetErrorStringWithFormat ("unsupported core architecture: %s", target_triple.str().c_str()); 5463 break; 5464 } 5465 5466 if (make_core) 5467 { 5468 std::vector<segment_command_64> segment_load_commands; 5469 // uint32_t range_info_idx = 0; 5470 MemoryRegionInfo range_info; 5471 Error range_error = process_sp->GetMemoryRegionInfo(0, range_info); 5472 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 5473 const ByteOrder byte_order = target_arch.GetByteOrder(); 5474 if (range_error.Success()) 5475 { 5476 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) 5477 { 5478 const addr_t addr = range_info.GetRange().GetRangeBase(); 5479 const addr_t size = range_info.GetRange().GetByteSize(); 5480 5481 if (size == 0) 5482 break; 5483 5484 // Calculate correct protections 5485 uint32_t prot = 0; 5486 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 5487 prot |= VM_PROT_READ; 5488 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 5489 prot |= VM_PROT_WRITE; 5490 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 5491 prot |= VM_PROT_EXECUTE; 5492 5493 // printf ("[%3u] [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 ") %c%c%c\n", 5494 // range_info_idx, 5495 // addr, 5496 // size, 5497 // (prot & VM_PROT_READ ) ? 'r' : '-', 5498 // (prot & VM_PROT_WRITE ) ? 'w' : '-', 5499 // (prot & VM_PROT_EXECUTE) ? 'x' : '-'); 5500 5501 if (prot != 0) 5502 { 5503 uint32_t cmd_type = LC_SEGMENT_64; 5504 uint32_t segment_size = sizeof (segment_command_64); 5505 if (addr_byte_size == 4) 5506 { 5507 cmd_type = LC_SEGMENT; 5508 segment_size = sizeof (segment_command); 5509 } 5510 segment_command_64 segment = { 5511 cmd_type, // uint32_t cmd; 5512 segment_size, // uint32_t cmdsize; 5513 {0}, // char segname[16]; 5514 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 5515 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 5516 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 5517 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 5518 prot, // uint32_t maxprot; 5519 prot, // uint32_t initprot; 5520 0, // uint32_t nsects; 5521 0 }; // uint32_t flags; 5522 segment_load_commands.push_back(segment); 5523 } 5524 else 5525 { 5526 // No protections and a size of 1 used to be returned from old 5527 // debugservers when we asked about a region that was past the 5528 // last memory region and it indicates the end... 5529 if (size == 1) 5530 break; 5531 } 5532 5533 range_error = process_sp->GetMemoryRegionInfo(range_info.GetRange().GetRangeEnd(), range_info); 5534 if (range_error.Fail()) 5535 break; 5536 } 5537 5538 StreamString buffer (Stream::eBinary, 5539 addr_byte_size, 5540 byte_order); 5541 5542 mach_header_64 mach_header; 5543 if (addr_byte_size == 8) 5544 { 5545 mach_header.magic = MH_MAGIC_64; 5546 } 5547 else 5548 { 5549 mach_header.magic = MH_MAGIC; 5550 } 5551 mach_header.cputype = target_arch.GetMachOCPUType(); 5552 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 5553 mach_header.filetype = MH_CORE; 5554 mach_header.ncmds = segment_load_commands.size(); 5555 mach_header.flags = 0; 5556 mach_header.reserved = 0; 5557 ThreadList &thread_list = process_sp->GetThreadList(); 5558 const uint32_t num_threads = thread_list.GetSize(); 5559 5560 // Make an array of LC_THREAD data items. Each one contains 5561 // the contents of the LC_THREAD load command. The data doesn't 5562 // contain the load command + load command size, we will 5563 // add the load command and load command size as we emit the data. 5564 std::vector<StreamString> LC_THREAD_datas(num_threads); 5565 for (auto &LC_THREAD_data : LC_THREAD_datas) 5566 { 5567 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 5568 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 5569 LC_THREAD_data.SetByteOrder(byte_order); 5570 } 5571 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) 5572 { 5573 ThreadSP thread_sp (thread_list.GetThreadAtIndex(thread_idx)); 5574 if (thread_sp) 5575 { 5576 switch (mach_header.cputype) 5577 { 5578 case llvm::MachO::CPU_TYPE_ARM64: 5579 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5580 break; 5581 5582 case llvm::MachO::CPU_TYPE_ARM: 5583 RegisterContextDarwin_arm_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5584 break; 5585 5586 case llvm::MachO::CPU_TYPE_I386: 5587 RegisterContextDarwin_i386_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5588 break; 5589 5590 case llvm::MachO::CPU_TYPE_X86_64: 5591 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5592 break; 5593 } 5594 5595 } 5596 } 5597 5598 // The size of the load command is the size of the segments... 5599 if (addr_byte_size == 8) 5600 { 5601 mach_header.sizeofcmds = segment_load_commands.size() * sizeof (struct segment_command_64); 5602 } 5603 else 5604 { 5605 mach_header.sizeofcmds = segment_load_commands.size() * sizeof (struct segment_command); 5606 } 5607 5608 // and the size of all LC_THREAD load command 5609 for (const auto &LC_THREAD_data : LC_THREAD_datas) 5610 { 5611 ++mach_header.ncmds; 5612 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 5613 } 5614 5615 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", 5616 mach_header.magic, 5617 mach_header.cputype, 5618 mach_header.cpusubtype, 5619 mach_header.filetype, 5620 mach_header.ncmds, 5621 mach_header.sizeofcmds, 5622 mach_header.flags, 5623 mach_header.reserved); 5624 5625 // Write the mach header 5626 buffer.PutHex32(mach_header.magic); 5627 buffer.PutHex32(mach_header.cputype); 5628 buffer.PutHex32(mach_header.cpusubtype); 5629 buffer.PutHex32(mach_header.filetype); 5630 buffer.PutHex32(mach_header.ncmds); 5631 buffer.PutHex32(mach_header.sizeofcmds); 5632 buffer.PutHex32(mach_header.flags); 5633 if (addr_byte_size == 8) 5634 { 5635 buffer.PutHex32(mach_header.reserved); 5636 } 5637 5638 // Skip the mach header and all load commands and align to the next 5639 // 0x1000 byte boundary 5640 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 5641 if (file_offset & 0x00000fff) 5642 { 5643 file_offset += 0x00001000ull; 5644 file_offset &= (~0x00001000ull + 1); 5645 } 5646 5647 for (auto &segment : segment_load_commands) 5648 { 5649 segment.fileoff = file_offset; 5650 file_offset += segment.filesize; 5651 } 5652 5653 // Write out all of the LC_THREAD load commands 5654 for (const auto &LC_THREAD_data : LC_THREAD_datas) 5655 { 5656 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 5657 buffer.PutHex32(LC_THREAD); 5658 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 5659 buffer.Write(LC_THREAD_data.GetData(), LC_THREAD_data_size); 5660 } 5661 5662 // Write out all of the segment load commands 5663 for (const auto &segment : segment_load_commands) 5664 { 5665 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", 5666 segment.cmd, 5667 segment.cmdsize, 5668 segment.vmaddr, 5669 segment.vmaddr + segment.vmsize, 5670 segment.fileoff, 5671 segment.filesize, 5672 segment.maxprot, 5673 segment.initprot, 5674 segment.nsects, 5675 segment.flags); 5676 5677 buffer.PutHex32(segment.cmd); 5678 buffer.PutHex32(segment.cmdsize); 5679 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 5680 if (addr_byte_size == 8) 5681 { 5682 buffer.PutHex64(segment.vmaddr); 5683 buffer.PutHex64(segment.vmsize); 5684 buffer.PutHex64(segment.fileoff); 5685 buffer.PutHex64(segment.filesize); 5686 } 5687 else 5688 { 5689 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 5690 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 5691 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 5692 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 5693 } 5694 buffer.PutHex32(segment.maxprot); 5695 buffer.PutHex32(segment.initprot); 5696 buffer.PutHex32(segment.nsects); 5697 buffer.PutHex32(segment.flags); 5698 } 5699 5700 File core_file; 5701 std::string core_file_path(outfile.GetPath()); 5702 error = core_file.Open(core_file_path.c_str(), 5703 File::eOpenOptionWrite | 5704 File::eOpenOptionTruncate | 5705 File::eOpenOptionCanCreate); 5706 if (error.Success()) 5707 { 5708 // Read 1 page at a time 5709 uint8_t bytes[0x1000]; 5710 // Write the mach header and load commands out to the core file 5711 size_t bytes_written = buffer.GetString().size(); 5712 error = core_file.Write(buffer.GetString().data(), bytes_written); 5713 if (error.Success()) 5714 { 5715 // Now write the file data for all memory segments in the process 5716 for (const auto &segment : segment_load_commands) 5717 { 5718 if (core_file.SeekFromStart(segment.fileoff) == -1) 5719 { 5720 error.SetErrorStringWithFormat("unable to seek to offset 0x%" PRIx64 " in '%s'", segment.fileoff, core_file_path.c_str()); 5721 break; 5722 } 5723 5724 printf ("Saving %" PRId64 " bytes of data for memory region at 0x%" PRIx64 "\n", segment.vmsize, segment.vmaddr); 5725 addr_t bytes_left = segment.vmsize; 5726 addr_t addr = segment.vmaddr; 5727 Error memory_read_error; 5728 while (bytes_left > 0 && error.Success()) 5729 { 5730 const size_t bytes_to_read = bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 5731 const size_t bytes_read = process_sp->ReadMemory(addr, bytes, bytes_to_read, memory_read_error); 5732 if (bytes_read == bytes_to_read) 5733 { 5734 size_t bytes_written = bytes_read; 5735 error = core_file.Write(bytes, bytes_written); 5736 bytes_left -= bytes_read; 5737 addr += bytes_read; 5738 } 5739 else 5740 { 5741 // Some pages within regions are not readable, those 5742 // should be zero filled 5743 memset (bytes, 0, bytes_to_read); 5744 size_t bytes_written = bytes_to_read; 5745 error = core_file.Write(bytes, bytes_written); 5746 bytes_left -= bytes_to_read; 5747 addr += bytes_to_read; 5748 } 5749 } 5750 } 5751 } 5752 } 5753 } 5754 else 5755 { 5756 error.SetErrorString("process doesn't support getting memory region info"); 5757 } 5758 } 5759 return true; // This is the right plug to handle saving core files for this process 5760 } 5761 } 5762 return false; 5763 } 5764 5765