1 //===-- ObjectFileMachO.cpp -------------------------------------*- C++ -*-===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 10 #include "llvm/ADT/StringRef.h" 11 12 #include "lldb/Core/ArchSpec.h" 13 #include "lldb/Core/DataBuffer.h" 14 #include "lldb/Core/Debugger.h" 15 #include "lldb/Core/Error.h" 16 #include "lldb/Core/FileSpecList.h" 17 #include "lldb/Core/Log.h" 18 #include "lldb/Core/Module.h" 19 #include "lldb/Core/ModuleSpec.h" 20 #include "lldb/Core/PluginManager.h" 21 #include "lldb/Core/RangeMap.h" 22 #include "lldb/Core/Section.h" 23 #include "lldb/Core/StreamFile.h" 24 #include "lldb/Core/StreamString.h" 25 #include "lldb/Core/Timer.h" 26 #include "lldb/Core/UUID.h" 27 #include "lldb/Host/Host.h" 28 #include "lldb/Host/FileSpec.h" 29 #include "lldb/Symbol/DWARFCallFrameInfo.h" 30 #include "lldb/Symbol/ObjectFile.h" 31 #include "lldb/Target/MemoryRegionInfo.h" 32 #include "lldb/Target/Platform.h" 33 #include "lldb/Target/Process.h" 34 #include "lldb/Target/SectionLoadList.h" 35 #include "lldb/Target/Target.h" 36 #include "lldb/Target/Thread.h" 37 #include "lldb/Target/ThreadList.h" 38 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 39 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 40 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 41 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 42 43 #include "lldb/Utility/SafeMachO.h" 44 45 #include "ObjectFileMachO.h" 46 47 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 48 // GetLLDBSharedCacheUUID() needs to call dlsym() 49 #include <dlfcn.h> 50 #endif 51 52 #ifndef __APPLE__ 53 #include "Utility/UuidCompatibility.h" 54 #endif 55 56 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 57 using namespace lldb; 58 using namespace lldb_private; 59 using namespace llvm::MachO; 60 61 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 62 { 63 public: 64 RegisterContextDarwin_x86_64_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 65 RegisterContextDarwin_x86_64 (thread, 0) 66 { 67 SetRegisterDataFrom_LC_THREAD (data); 68 } 69 70 virtual void 71 InvalidateAllRegisters () 72 { 73 // Do nothing... registers are always valid... 74 } 75 76 void 77 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 78 { 79 lldb::offset_t offset = 0; 80 SetError (GPRRegSet, Read, -1); 81 SetError (FPURegSet, Read, -1); 82 SetError (EXCRegSet, Read, -1); 83 bool done = false; 84 85 while (!done) 86 { 87 int flavor = data.GetU32 (&offset); 88 if (flavor == 0) 89 done = true; 90 else 91 { 92 uint32_t i; 93 uint32_t count = data.GetU32 (&offset); 94 switch (flavor) 95 { 96 case GPRRegSet: 97 for (i=0; i<count; ++i) 98 (&gpr.rax)[i] = data.GetU64(&offset); 99 SetError (GPRRegSet, Read, 0); 100 done = true; 101 102 break; 103 case FPURegSet: 104 // TODO: fill in FPU regs.... 105 //SetError (FPURegSet, Read, -1); 106 done = true; 107 108 break; 109 case EXCRegSet: 110 exc.trapno = data.GetU32(&offset); 111 exc.err = data.GetU32(&offset); 112 exc.faultvaddr = data.GetU64(&offset); 113 SetError (EXCRegSet, Read, 0); 114 done = true; 115 break; 116 case 7: 117 case 8: 118 case 9: 119 // fancy flavors that encapsulate of the above 120 // flavors... 121 break; 122 123 default: 124 done = true; 125 break; 126 } 127 } 128 } 129 } 130 131 132 static size_t 133 WriteRegister (RegisterContext *reg_ctx, const char *name, const char *alt_name, size_t reg_byte_size, Stream &data) 134 { 135 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 136 if (reg_info == NULL) 137 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 138 if (reg_info) 139 { 140 lldb_private::RegisterValue reg_value; 141 if (reg_ctx->ReadRegister(reg_info, reg_value)) 142 { 143 if (reg_info->byte_size >= reg_byte_size) 144 data.Write(reg_value.GetBytes(), reg_byte_size); 145 else 146 { 147 data.Write(reg_value.GetBytes(), reg_info->byte_size); 148 for (size_t i=0, n = reg_byte_size - reg_info->byte_size; i<n; ++ i) 149 data.PutChar(0); 150 } 151 return reg_byte_size; 152 } 153 } 154 // Just write zeros if all else fails 155 for (size_t i=0; i<reg_byte_size; ++ i) 156 data.PutChar(0); 157 return reg_byte_size; 158 } 159 160 static bool 161 Create_LC_THREAD (Thread *thread, Stream &data) 162 { 163 RegisterContextSP reg_ctx_sp (thread->GetRegisterContext()); 164 if (reg_ctx_sp) 165 { 166 RegisterContext *reg_ctx = reg_ctx_sp.get(); 167 168 data.PutHex32 (GPRRegSet); // Flavor 169 data.PutHex32 (GPRWordCount); 170 WriteRegister (reg_ctx, "rax", NULL, 8, data); 171 WriteRegister (reg_ctx, "rbx", NULL, 8, data); 172 WriteRegister (reg_ctx, "rcx", NULL, 8, data); 173 WriteRegister (reg_ctx, "rdx", NULL, 8, data); 174 WriteRegister (reg_ctx, "rdi", NULL, 8, data); 175 WriteRegister (reg_ctx, "rsi", NULL, 8, data); 176 WriteRegister (reg_ctx, "rbp", NULL, 8, data); 177 WriteRegister (reg_ctx, "rsp", NULL, 8, data); 178 WriteRegister (reg_ctx, "r8", NULL, 8, data); 179 WriteRegister (reg_ctx, "r9", NULL, 8, data); 180 WriteRegister (reg_ctx, "r10", NULL, 8, data); 181 WriteRegister (reg_ctx, "r11", NULL, 8, data); 182 WriteRegister (reg_ctx, "r12", NULL, 8, data); 183 WriteRegister (reg_ctx, "r13", NULL, 8, data); 184 WriteRegister (reg_ctx, "r14", NULL, 8, data); 185 WriteRegister (reg_ctx, "r15", NULL, 8, data); 186 WriteRegister (reg_ctx, "rip", NULL, 8, data); 187 WriteRegister (reg_ctx, "rflags", NULL, 8, data); 188 WriteRegister (reg_ctx, "cs", NULL, 8, data); 189 WriteRegister (reg_ctx, "fs", NULL, 8, data); 190 WriteRegister (reg_ctx, "gs", NULL, 8, data); 191 192 // // Write out the FPU registers 193 // const size_t fpu_byte_size = sizeof(FPU); 194 // size_t bytes_written = 0; 195 // data.PutHex32 (FPURegSet); 196 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 197 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 198 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 199 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, data); // uint16_t fcw; // "fctrl" 200 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, data); // uint16_t fsw; // "fstat" 201 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, data); // uint8_t ftw; // "ftag" 202 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 203 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, data); // uint16_t fop; // "fop" 204 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, data); // uint32_t ip; // "fioff" 205 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, data); // uint16_t cs; // "fiseg" 206 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 207 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, data); // uint32_t dp; // "fooff" 208 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, data); // uint16_t ds; // "foseg" 209 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 210 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, data); // uint32_t mxcsr; 211 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 4, data);// uint32_t mxcsrmask; 212 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, sizeof(MMSReg), data); 213 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, sizeof(MMSReg), data); 214 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, sizeof(MMSReg), data); 215 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, sizeof(MMSReg), data); 216 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, sizeof(MMSReg), data); 217 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, sizeof(MMSReg), data); 218 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, sizeof(MMSReg), data); 219 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, sizeof(MMSReg), data); 220 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, sizeof(XMMReg), data); 221 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, sizeof(XMMReg), data); 222 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, sizeof(XMMReg), data); 223 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, sizeof(XMMReg), data); 224 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, sizeof(XMMReg), data); 225 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, sizeof(XMMReg), data); 226 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, sizeof(XMMReg), data); 227 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, sizeof(XMMReg), data); 228 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, sizeof(XMMReg), data); 229 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, sizeof(XMMReg), data); 230 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, sizeof(XMMReg), data); 231 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, sizeof(XMMReg), data); 232 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, sizeof(XMMReg), data); 233 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, sizeof(XMMReg), data); 234 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, sizeof(XMMReg), data); 235 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, sizeof(XMMReg), data); 236 // 237 // // Fill rest with zeros 238 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ i) 239 // data.PutChar(0); 240 241 // Write out the EXC registers 242 data.PutHex32 (EXCRegSet); 243 data.PutHex32 (EXCWordCount); 244 WriteRegister (reg_ctx, "trapno", NULL, 4, data); 245 WriteRegister (reg_ctx, "err", NULL, 4, data); 246 WriteRegister (reg_ctx, "faultvaddr", NULL, 8, data); 247 return true; 248 } 249 return false; 250 } 251 252 protected: 253 virtual int 254 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 255 { 256 return 0; 257 } 258 259 virtual int 260 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 261 { 262 return 0; 263 } 264 265 virtual int 266 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 267 { 268 return 0; 269 } 270 271 virtual int 272 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 273 { 274 return 0; 275 } 276 277 virtual int 278 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 279 { 280 return 0; 281 } 282 283 virtual int 284 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 285 { 286 return 0; 287 } 288 }; 289 290 291 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 292 { 293 public: 294 RegisterContextDarwin_i386_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 295 RegisterContextDarwin_i386 (thread, 0) 296 { 297 SetRegisterDataFrom_LC_THREAD (data); 298 } 299 300 virtual void 301 InvalidateAllRegisters () 302 { 303 // Do nothing... registers are always valid... 304 } 305 306 void 307 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 308 { 309 lldb::offset_t offset = 0; 310 SetError (GPRRegSet, Read, -1); 311 SetError (FPURegSet, Read, -1); 312 SetError (EXCRegSet, Read, -1); 313 bool done = false; 314 315 while (!done) 316 { 317 int flavor = data.GetU32 (&offset); 318 if (flavor == 0) 319 done = true; 320 else 321 { 322 uint32_t i; 323 uint32_t count = data.GetU32 (&offset); 324 switch (flavor) 325 { 326 case GPRRegSet: 327 for (i=0; i<count; ++i) 328 (&gpr.eax)[i] = data.GetU32(&offset); 329 SetError (GPRRegSet, Read, 0); 330 done = true; 331 332 break; 333 case FPURegSet: 334 // TODO: fill in FPU regs.... 335 //SetError (FPURegSet, Read, -1); 336 done = true; 337 338 break; 339 case EXCRegSet: 340 exc.trapno = data.GetU32(&offset); 341 exc.err = data.GetU32(&offset); 342 exc.faultvaddr = data.GetU32(&offset); 343 SetError (EXCRegSet, Read, 0); 344 done = true; 345 break; 346 case 7: 347 case 8: 348 case 9: 349 // fancy flavors that encapsulate of the above 350 // flavors... 351 break; 352 353 default: 354 done = true; 355 break; 356 } 357 } 358 } 359 } 360 361 static size_t 362 WriteRegister (RegisterContext *reg_ctx, const char *name, const char *alt_name, size_t reg_byte_size, Stream &data) 363 { 364 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 365 if (reg_info == NULL) 366 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 367 if (reg_info) 368 { 369 lldb_private::RegisterValue reg_value; 370 if (reg_ctx->ReadRegister(reg_info, reg_value)) 371 { 372 if (reg_info->byte_size >= reg_byte_size) 373 data.Write(reg_value.GetBytes(), reg_byte_size); 374 else 375 { 376 data.Write(reg_value.GetBytes(), reg_info->byte_size); 377 for (size_t i=0, n = reg_byte_size - reg_info->byte_size; i<n; ++ i) 378 data.PutChar(0); 379 } 380 return reg_byte_size; 381 } 382 } 383 // Just write zeros if all else fails 384 for (size_t i=0; i<reg_byte_size; ++ i) 385 data.PutChar(0); 386 return reg_byte_size; 387 } 388 389 static bool 390 Create_LC_THREAD (Thread *thread, Stream &data) 391 { 392 RegisterContextSP reg_ctx_sp (thread->GetRegisterContext()); 393 if (reg_ctx_sp) 394 { 395 RegisterContext *reg_ctx = reg_ctx_sp.get(); 396 397 data.PutHex32 (GPRRegSet); // Flavor 398 data.PutHex32 (GPRWordCount); 399 WriteRegister (reg_ctx, "eax", NULL, 4, data); 400 WriteRegister (reg_ctx, "ebx", NULL, 4, data); 401 WriteRegister (reg_ctx, "ecx", NULL, 4, data); 402 WriteRegister (reg_ctx, "edx", NULL, 4, data); 403 WriteRegister (reg_ctx, "edi", NULL, 4, data); 404 WriteRegister (reg_ctx, "esi", NULL, 4, data); 405 WriteRegister (reg_ctx, "ebp", NULL, 4, data); 406 WriteRegister (reg_ctx, "esp", NULL, 4, data); 407 WriteRegister (reg_ctx, "ss", NULL, 4, data); 408 WriteRegister (reg_ctx, "eflags", NULL, 4, data); 409 WriteRegister (reg_ctx, "eip", NULL, 4, data); 410 WriteRegister (reg_ctx, "cs", NULL, 4, data); 411 WriteRegister (reg_ctx, "ds", NULL, 4, data); 412 WriteRegister (reg_ctx, "es", NULL, 4, data); 413 WriteRegister (reg_ctx, "fs", NULL, 4, data); 414 WriteRegister (reg_ctx, "gs", NULL, 4, data); 415 416 // Write out the EXC registers 417 data.PutHex32 (EXCRegSet); 418 data.PutHex32 (EXCWordCount); 419 WriteRegister (reg_ctx, "trapno", NULL, 4, data); 420 WriteRegister (reg_ctx, "err", NULL, 4, data); 421 WriteRegister (reg_ctx, "faultvaddr", NULL, 4, data); 422 return true; 423 } 424 return false; 425 } 426 427 protected: 428 virtual int 429 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 430 { 431 return 0; 432 } 433 434 virtual int 435 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 436 { 437 return 0; 438 } 439 440 virtual int 441 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 442 { 443 return 0; 444 } 445 446 virtual int 447 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 448 { 449 return 0; 450 } 451 452 virtual int 453 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 454 { 455 return 0; 456 } 457 458 virtual int 459 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 460 { 461 return 0; 462 } 463 }; 464 465 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm 466 { 467 public: 468 RegisterContextDarwin_arm_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 469 RegisterContextDarwin_arm (thread, 0) 470 { 471 SetRegisterDataFrom_LC_THREAD (data); 472 } 473 474 virtual void 475 InvalidateAllRegisters () 476 { 477 // Do nothing... registers are always valid... 478 } 479 480 void 481 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 482 { 483 lldb::offset_t offset = 0; 484 SetError (GPRRegSet, Read, -1); 485 SetError (FPURegSet, Read, -1); 486 SetError (EXCRegSet, Read, -1); 487 bool done = false; 488 489 while (!done) 490 { 491 int flavor = data.GetU32 (&offset); 492 uint32_t count = data.GetU32 (&offset); 493 lldb::offset_t next_thread_state = offset + (count * 4); 494 switch (flavor) 495 { 496 case GPRRegSet: 497 for (uint32_t i=0; i<count; ++i) 498 { 499 gpr.r[i] = data.GetU32(&offset); 500 } 501 502 // Note that gpr.cpsr is also copied by the above loop; this loop technically extends 503 // one element past the end of the gpr.r[] array. 504 505 SetError (GPRRegSet, Read, 0); 506 offset = next_thread_state; 507 break; 508 509 case FPURegSet: 510 { 511 uint8_t *fpu_reg_buf = (uint8_t*) &fpu.floats.s[0]; 512 const int fpu_reg_buf_size = sizeof (fpu.floats); 513 if (data.ExtractBytes (offset, fpu_reg_buf_size, eByteOrderLittle, fpu_reg_buf) == fpu_reg_buf_size) 514 { 515 offset += fpu_reg_buf_size; 516 fpu.fpscr = data.GetU32(&offset); 517 SetError (FPURegSet, Read, 0); 518 } 519 else 520 { 521 done = true; 522 } 523 } 524 offset = next_thread_state; 525 break; 526 527 case EXCRegSet: 528 if (count == 3) 529 { 530 exc.exception = data.GetU32(&offset); 531 exc.fsr = data.GetU32(&offset); 532 exc.far = data.GetU32(&offset); 533 SetError (EXCRegSet, Read, 0); 534 } 535 done = true; 536 offset = next_thread_state; 537 break; 538 539 // Unknown register set flavor, stop trying to parse. 540 default: 541 done = true; 542 } 543 } 544 } 545 546 static size_t 547 WriteRegister (RegisterContext *reg_ctx, const char *name, const char *alt_name, size_t reg_byte_size, Stream &data) 548 { 549 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 550 if (reg_info == NULL) 551 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 552 if (reg_info) 553 { 554 lldb_private::RegisterValue reg_value; 555 if (reg_ctx->ReadRegister(reg_info, reg_value)) 556 { 557 if (reg_info->byte_size >= reg_byte_size) 558 data.Write(reg_value.GetBytes(), reg_byte_size); 559 else 560 { 561 data.Write(reg_value.GetBytes(), reg_info->byte_size); 562 for (size_t i=0, n = reg_byte_size - reg_info->byte_size; i<n; ++ i) 563 data.PutChar(0); 564 } 565 return reg_byte_size; 566 } 567 } 568 // Just write zeros if all else fails 569 for (size_t i=0; i<reg_byte_size; ++ i) 570 data.PutChar(0); 571 return reg_byte_size; 572 } 573 574 static bool 575 Create_LC_THREAD (Thread *thread, Stream &data) 576 { 577 RegisterContextSP reg_ctx_sp (thread->GetRegisterContext()); 578 if (reg_ctx_sp) 579 { 580 RegisterContext *reg_ctx = reg_ctx_sp.get(); 581 582 data.PutHex32 (GPRRegSet); // Flavor 583 data.PutHex32 (GPRWordCount); 584 WriteRegister (reg_ctx, "r0", NULL, 4, data); 585 WriteRegister (reg_ctx, "r1", NULL, 4, data); 586 WriteRegister (reg_ctx, "r2", NULL, 4, data); 587 WriteRegister (reg_ctx, "r3", NULL, 4, data); 588 WriteRegister (reg_ctx, "r4", NULL, 4, data); 589 WriteRegister (reg_ctx, "r5", NULL, 4, data); 590 WriteRegister (reg_ctx, "r6", NULL, 4, data); 591 WriteRegister (reg_ctx, "r7", NULL, 4, data); 592 WriteRegister (reg_ctx, "r8", NULL, 4, data); 593 WriteRegister (reg_ctx, "r9", NULL, 4, data); 594 WriteRegister (reg_ctx, "r10", NULL, 4, data); 595 WriteRegister (reg_ctx, "r11", NULL, 4, data); 596 WriteRegister (reg_ctx, "r12", NULL, 4, data); 597 WriteRegister (reg_ctx, "sp", NULL, 4, data); 598 WriteRegister (reg_ctx, "lr", NULL, 4, data); 599 WriteRegister (reg_ctx, "pc", NULL, 4, data); 600 WriteRegister (reg_ctx, "cpsr", NULL, 4, data); 601 602 // Write out the EXC registers 603 // data.PutHex32 (EXCRegSet); 604 // data.PutHex32 (EXCWordCount); 605 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 606 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 607 // WriteRegister (reg_ctx, "far", NULL, 4, data); 608 return true; 609 } 610 return false; 611 } 612 613 protected: 614 virtual int 615 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 616 { 617 return -1; 618 } 619 620 virtual int 621 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 622 { 623 return -1; 624 } 625 626 virtual int 627 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 628 { 629 return -1; 630 } 631 632 virtual int 633 DoReadDBG (lldb::tid_t tid, int flavor, DBG &dbg) 634 { 635 return -1; 636 } 637 638 virtual int 639 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 640 { 641 return 0; 642 } 643 644 virtual int 645 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 646 { 647 return 0; 648 } 649 650 virtual int 651 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 652 { 653 return 0; 654 } 655 656 virtual int 657 DoWriteDBG (lldb::tid_t tid, int flavor, const DBG &dbg) 658 { 659 return -1; 660 } 661 }; 662 663 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 664 { 665 public: 666 RegisterContextDarwin_arm64_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 667 RegisterContextDarwin_arm64 (thread, 0) 668 { 669 SetRegisterDataFrom_LC_THREAD (data); 670 } 671 672 virtual void 673 InvalidateAllRegisters () 674 { 675 // Do nothing... registers are always valid... 676 } 677 678 void 679 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 680 { 681 lldb::offset_t offset = 0; 682 SetError (GPRRegSet, Read, -1); 683 SetError (FPURegSet, Read, -1); 684 SetError (EXCRegSet, Read, -1); 685 bool done = false; 686 while (!done) 687 { 688 int flavor = data.GetU32 (&offset); 689 uint32_t count = data.GetU32 (&offset); 690 lldb::offset_t next_thread_state = offset + (count * 4); 691 switch (flavor) 692 { 693 case GPRRegSet: 694 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 32-bit register) 695 if (count >= (33 * 2) + 1) 696 { 697 for (uint32_t i=0; i<33; ++i) 698 gpr.x[i] = data.GetU64(&offset); 699 gpr.cpsr = data.GetU32(&offset); 700 SetError (GPRRegSet, Read, 0); 701 } 702 offset = next_thread_state; 703 break; 704 case FPURegSet: 705 { 706 uint8_t *fpu_reg_buf = (uint8_t*) &fpu.v[0]; 707 const int fpu_reg_buf_size = sizeof (fpu); 708 if (fpu_reg_buf_size == count 709 && data.ExtractBytes (offset, fpu_reg_buf_size, eByteOrderLittle, fpu_reg_buf) == fpu_reg_buf_size) 710 { 711 SetError (FPURegSet, Read, 0); 712 } 713 else 714 { 715 done = true; 716 } 717 } 718 offset = next_thread_state; 719 break; 720 case EXCRegSet: 721 if (count == 4) 722 { 723 exc.far = data.GetU64(&offset); 724 exc.esr = data.GetU32(&offset); 725 exc.exception = data.GetU32(&offset); 726 SetError (EXCRegSet, Read, 0); 727 } 728 offset = next_thread_state; 729 break; 730 default: 731 done = true; 732 break; 733 } 734 } 735 } 736 737 static size_t 738 WriteRegister (RegisterContext *reg_ctx, const char *name, const char *alt_name, size_t reg_byte_size, Stream &data) 739 { 740 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 741 if (reg_info == NULL) 742 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 743 if (reg_info) 744 { 745 lldb_private::RegisterValue reg_value; 746 if (reg_ctx->ReadRegister(reg_info, reg_value)) 747 { 748 if (reg_info->byte_size >= reg_byte_size) 749 data.Write(reg_value.GetBytes(), reg_byte_size); 750 else 751 { 752 data.Write(reg_value.GetBytes(), reg_info->byte_size); 753 for (size_t i=0, n = reg_byte_size - reg_info->byte_size; i<n; ++ i) 754 data.PutChar(0); 755 } 756 return reg_byte_size; 757 } 758 } 759 // Just write zeros if all else fails 760 for (size_t i=0; i<reg_byte_size; ++ i) 761 data.PutChar(0); 762 return reg_byte_size; 763 } 764 765 static bool 766 Create_LC_THREAD (Thread *thread, Stream &data) 767 { 768 RegisterContextSP reg_ctx_sp (thread->GetRegisterContext()); 769 if (reg_ctx_sp) 770 { 771 RegisterContext *reg_ctx = reg_ctx_sp.get(); 772 773 data.PutHex32 (GPRRegSet); // Flavor 774 data.PutHex32 (GPRWordCount); 775 WriteRegister (reg_ctx, "x0", NULL, 8, data); 776 WriteRegister (reg_ctx, "x1", NULL, 8, data); 777 WriteRegister (reg_ctx, "x2", NULL, 8, data); 778 WriteRegister (reg_ctx, "x3", NULL, 8, data); 779 WriteRegister (reg_ctx, "x4", NULL, 8, data); 780 WriteRegister (reg_ctx, "x5", NULL, 8, data); 781 WriteRegister (reg_ctx, "x6", NULL, 8, data); 782 WriteRegister (reg_ctx, "x7", NULL, 8, data); 783 WriteRegister (reg_ctx, "x8", NULL, 8, data); 784 WriteRegister (reg_ctx, "x9", NULL, 8, data); 785 WriteRegister (reg_ctx, "x10", NULL, 8, data); 786 WriteRegister (reg_ctx, "x11", NULL, 8, data); 787 WriteRegister (reg_ctx, "x12", NULL, 8, data); 788 WriteRegister (reg_ctx, "x13", NULL, 8, data); 789 WriteRegister (reg_ctx, "x14", NULL, 8, data); 790 WriteRegister (reg_ctx, "x15", NULL, 8, data); 791 WriteRegister (reg_ctx, "x16", NULL, 8, data); 792 WriteRegister (reg_ctx, "x17", NULL, 8, data); 793 WriteRegister (reg_ctx, "x18", NULL, 8, data); 794 WriteRegister (reg_ctx, "x19", NULL, 8, data); 795 WriteRegister (reg_ctx, "x20", NULL, 8, data); 796 WriteRegister (reg_ctx, "x21", NULL, 8, data); 797 WriteRegister (reg_ctx, "x22", NULL, 8, data); 798 WriteRegister (reg_ctx, "x23", NULL, 8, data); 799 WriteRegister (reg_ctx, "x24", NULL, 8, data); 800 WriteRegister (reg_ctx, "x25", NULL, 8, data); 801 WriteRegister (reg_ctx, "x26", NULL, 8, data); 802 WriteRegister (reg_ctx, "x27", NULL, 8, data); 803 WriteRegister (reg_ctx, "x28", NULL, 8, data); 804 WriteRegister (reg_ctx, "fp", NULL, 8, data); 805 WriteRegister (reg_ctx, "lr", NULL, 8, data); 806 WriteRegister (reg_ctx, "sp", NULL, 8, data); 807 WriteRegister (reg_ctx, "pc", NULL, 8, data); 808 WriteRegister (reg_ctx, "cpsr", NULL, 4, data); 809 810 // Write out the EXC registers 811 // data.PutHex32 (EXCRegSet); 812 // data.PutHex32 (EXCWordCount); 813 // WriteRegister (reg_ctx, "far", NULL, 8, data); 814 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 815 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 816 return true; 817 } 818 return false; 819 } 820 821 protected: 822 virtual int 823 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 824 { 825 return -1; 826 } 827 828 virtual int 829 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 830 { 831 return -1; 832 } 833 834 virtual int 835 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 836 { 837 return -1; 838 } 839 840 virtual int 841 DoReadDBG (lldb::tid_t tid, int flavor, DBG &dbg) 842 { 843 return -1; 844 } 845 846 virtual int 847 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 848 { 849 return 0; 850 } 851 852 virtual int 853 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 854 { 855 return 0; 856 } 857 858 virtual int 859 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 860 { 861 return 0; 862 } 863 864 virtual int 865 DoWriteDBG (lldb::tid_t tid, int flavor, const DBG &dbg) 866 { 867 return -1; 868 } 869 }; 870 871 static uint32_t 872 MachHeaderSizeFromMagic(uint32_t magic) 873 { 874 switch (magic) 875 { 876 case MH_MAGIC: 877 case MH_CIGAM: 878 return sizeof(struct mach_header); 879 880 case MH_MAGIC_64: 881 case MH_CIGAM_64: 882 return sizeof(struct mach_header_64); 883 break; 884 885 default: 886 break; 887 } 888 return 0; 889 } 890 891 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 892 893 void 894 ObjectFileMachO::Initialize() 895 { 896 PluginManager::RegisterPlugin (GetPluginNameStatic(), 897 GetPluginDescriptionStatic(), 898 CreateInstance, 899 CreateMemoryInstance, 900 GetModuleSpecifications, 901 SaveCore); 902 } 903 904 void 905 ObjectFileMachO::Terminate() 906 { 907 PluginManager::UnregisterPlugin (CreateInstance); 908 } 909 910 911 lldb_private::ConstString 912 ObjectFileMachO::GetPluginNameStatic() 913 { 914 static ConstString g_name("mach-o"); 915 return g_name; 916 } 917 918 const char * 919 ObjectFileMachO::GetPluginDescriptionStatic() 920 { 921 return "Mach-o object file reader (32 and 64 bit)"; 922 } 923 924 ObjectFile * 925 ObjectFileMachO::CreateInstance (const lldb::ModuleSP &module_sp, 926 DataBufferSP& data_sp, 927 lldb::offset_t data_offset, 928 const FileSpec* file, 929 lldb::offset_t file_offset, 930 lldb::offset_t length) 931 { 932 if (!data_sp) 933 { 934 data_sp = file->MemoryMapFileContentsIfLocal(file_offset, length); 935 data_offset = 0; 936 } 937 938 if (ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 939 { 940 // Update the data to contain the entire file if it doesn't already 941 if (data_sp->GetByteSize() < length) 942 { 943 data_sp = file->MemoryMapFileContentsIfLocal(file_offset, length); 944 data_offset = 0; 945 } 946 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, data_offset, file, file_offset, length)); 947 if (objfile_ap.get() && objfile_ap->ParseHeader()) 948 return objfile_ap.release(); 949 } 950 return NULL; 951 } 952 953 ObjectFile * 954 ObjectFileMachO::CreateMemoryInstance (const lldb::ModuleSP &module_sp, 955 DataBufferSP& data_sp, 956 const ProcessSP &process_sp, 957 lldb::addr_t header_addr) 958 { 959 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) 960 { 961 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, process_sp, header_addr)); 962 if (objfile_ap.get() && objfile_ap->ParseHeader()) 963 return objfile_ap.release(); 964 } 965 return NULL; 966 } 967 968 size_t 969 ObjectFileMachO::GetModuleSpecifications (const lldb_private::FileSpec& file, 970 lldb::DataBufferSP& data_sp, 971 lldb::offset_t data_offset, 972 lldb::offset_t file_offset, 973 lldb::offset_t length, 974 lldb_private::ModuleSpecList &specs) 975 { 976 const size_t initial_count = specs.GetSize(); 977 978 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) 979 { 980 DataExtractor data; 981 data.SetData(data_sp); 982 llvm::MachO::mach_header header; 983 if (ParseHeader (data, &data_offset, header)) 984 { 985 size_t header_and_load_cmds = header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 986 if (header_and_load_cmds >= data_sp->GetByteSize()) 987 { 988 data_sp = file.ReadFileContents(file_offset, header_and_load_cmds); 989 data.SetData(data_sp); 990 data_offset = MachHeaderSizeFromMagic(header.magic); 991 } 992 if (data_sp) 993 { 994 ModuleSpec spec; 995 spec.GetFileSpec() = file; 996 spec.SetObjectOffset(file_offset); 997 spec.SetObjectSize(length); 998 999 if (GetArchitecture (header, data, data_offset, spec.GetArchitecture())) 1000 { 1001 if (spec.GetArchitecture().IsValid()) 1002 { 1003 GetUUID (header, data, data_offset, spec.GetUUID()); 1004 specs.Append(spec); 1005 } 1006 } 1007 } 1008 } 1009 } 1010 return specs.GetSize() - initial_count; 1011 } 1012 1013 1014 1015 const ConstString & 1016 ObjectFileMachO::GetSegmentNameTEXT() 1017 { 1018 static ConstString g_segment_name_TEXT ("__TEXT"); 1019 return g_segment_name_TEXT; 1020 } 1021 1022 const ConstString & 1023 ObjectFileMachO::GetSegmentNameDATA() 1024 { 1025 static ConstString g_segment_name_DATA ("__DATA"); 1026 return g_segment_name_DATA; 1027 } 1028 1029 const ConstString & 1030 ObjectFileMachO::GetSegmentNameDATA_DIRTY() 1031 { 1032 static ConstString g_segment_name ("__DATA_DIRTY"); 1033 return g_segment_name; 1034 } 1035 1036 const ConstString & 1037 ObjectFileMachO::GetSegmentNameDATA_CONST() 1038 { 1039 static ConstString g_segment_name ("__DATA_CONST"); 1040 return g_segment_name; 1041 } 1042 1043 const ConstString & 1044 ObjectFileMachO::GetSegmentNameOBJC() 1045 { 1046 static ConstString g_segment_name_OBJC ("__OBJC"); 1047 return g_segment_name_OBJC; 1048 } 1049 1050 const ConstString & 1051 ObjectFileMachO::GetSegmentNameLINKEDIT() 1052 { 1053 static ConstString g_section_name_LINKEDIT ("__LINKEDIT"); 1054 return g_section_name_LINKEDIT; 1055 } 1056 1057 const ConstString & 1058 ObjectFileMachO::GetSectionNameEHFrame() 1059 { 1060 static ConstString g_section_name_eh_frame ("__eh_frame"); 1061 return g_section_name_eh_frame; 1062 } 1063 1064 bool 1065 ObjectFileMachO::MagicBytesMatch (DataBufferSP& data_sp, 1066 lldb::addr_t data_offset, 1067 lldb::addr_t data_length) 1068 { 1069 DataExtractor data; 1070 data.SetData (data_sp, data_offset, data_length); 1071 lldb::offset_t offset = 0; 1072 uint32_t magic = data.GetU32(&offset); 1073 return MachHeaderSizeFromMagic(magic) != 0; 1074 } 1075 1076 1077 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 1078 DataBufferSP& data_sp, 1079 lldb::offset_t data_offset, 1080 const FileSpec* file, 1081 lldb::offset_t file_offset, 1082 lldb::offset_t length) : 1083 ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 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 (const lldb::ModuleSP &module_sp, 1095 lldb::DataBufferSP& header_data_sp, 1096 const lldb::ProcessSP &process_sp, 1097 lldb::addr_t header_addr) : 1098 ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 1099 m_mach_segments(), 1100 m_mach_sections(), 1101 m_entry_point_address(), 1102 m_thread_context_offsets(), 1103 m_thread_context_offsets_valid(false) 1104 { 1105 ::memset (&m_header, 0, sizeof(m_header)); 1106 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 1107 } 1108 1109 ObjectFileMachO::~ObjectFileMachO() 1110 { 1111 } 1112 1113 bool 1114 ObjectFileMachO::ParseHeader (DataExtractor &data, 1115 lldb::offset_t *data_offset_ptr, 1116 llvm::MachO::mach_header &header) 1117 { 1118 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1119 // Leave magic in the original byte order 1120 header.magic = data.GetU32(data_offset_ptr); 1121 bool can_parse = false; 1122 bool is_64_bit = false; 1123 switch (header.magic) 1124 { 1125 case MH_MAGIC: 1126 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1127 data.SetAddressByteSize(4); 1128 can_parse = true; 1129 break; 1130 1131 case MH_MAGIC_64: 1132 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1133 data.SetAddressByteSize(8); 1134 can_parse = true; 1135 is_64_bit = true; 1136 break; 1137 1138 case MH_CIGAM: 1139 data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 1140 data.SetAddressByteSize(4); 1141 can_parse = true; 1142 break; 1143 1144 case MH_CIGAM_64: 1145 data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 1146 data.SetAddressByteSize(8); 1147 is_64_bit = true; 1148 can_parse = true; 1149 break; 1150 1151 default: 1152 break; 1153 } 1154 1155 if (can_parse) 1156 { 1157 data.GetU32(data_offset_ptr, &header.cputype, 6); 1158 if (is_64_bit) 1159 *data_offset_ptr += 4; 1160 return true; 1161 } 1162 else 1163 { 1164 memset(&header, 0, sizeof(header)); 1165 } 1166 return false; 1167 } 1168 1169 bool 1170 ObjectFileMachO::ParseHeader () 1171 { 1172 ModuleSP module_sp(GetModule()); 1173 if (module_sp) 1174 { 1175 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 1176 bool can_parse = false; 1177 lldb::offset_t offset = 0; 1178 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1179 // Leave magic in the original byte order 1180 m_header.magic = m_data.GetU32(&offset); 1181 switch (m_header.magic) 1182 { 1183 case MH_MAGIC: 1184 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1185 m_data.SetAddressByteSize(4); 1186 can_parse = true; 1187 break; 1188 1189 case MH_MAGIC_64: 1190 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 1191 m_data.SetAddressByteSize(8); 1192 can_parse = true; 1193 break; 1194 1195 case MH_CIGAM: 1196 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 1197 m_data.SetAddressByteSize(4); 1198 can_parse = true; 1199 break; 1200 1201 case MH_CIGAM_64: 1202 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 1203 m_data.SetAddressByteSize(8); 1204 can_parse = true; 1205 break; 1206 1207 default: 1208 break; 1209 } 1210 1211 if (can_parse) 1212 { 1213 m_data.GetU32(&offset, &m_header.cputype, 6); 1214 1215 1216 ArchSpec mach_arch; 1217 1218 if (GetArchitecture (mach_arch)) 1219 { 1220 // Check if the module has a required architecture 1221 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1222 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1223 return false; 1224 1225 if (SetModulesArchitecture (mach_arch)) 1226 { 1227 const size_t header_and_lc_size = m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1228 if (m_data.GetByteSize() < header_and_lc_size) 1229 { 1230 DataBufferSP data_sp; 1231 ProcessSP process_sp (m_process_wp.lock()); 1232 if (process_sp) 1233 { 1234 data_sp = ReadMemory (process_sp, m_memory_addr, header_and_lc_size); 1235 } 1236 else 1237 { 1238 // Read in all only the load command data from the file on disk 1239 data_sp = m_file.ReadFileContents(m_file_offset, header_and_lc_size); 1240 if (data_sp->GetByteSize() != header_and_lc_size) 1241 return false; 1242 } 1243 if (data_sp) 1244 m_data.SetData (data_sp); 1245 } 1246 } 1247 return true; 1248 } 1249 } 1250 else 1251 { 1252 memset(&m_header, 0, sizeof(struct mach_header)); 1253 } 1254 } 1255 return false; 1256 } 1257 1258 1259 ByteOrder 1260 ObjectFileMachO::GetByteOrder () const 1261 { 1262 return m_data.GetByteOrder (); 1263 } 1264 1265 bool 1266 ObjectFileMachO::IsExecutable() const 1267 { 1268 return m_header.filetype == MH_EXECUTE; 1269 } 1270 1271 uint32_t 1272 ObjectFileMachO::GetAddressByteSize () const 1273 { 1274 return m_data.GetAddressByteSize (); 1275 } 1276 1277 AddressClass 1278 ObjectFileMachO::GetAddressClass (lldb::addr_t file_addr) 1279 { 1280 Symtab *symtab = GetSymtab(); 1281 if (symtab) 1282 { 1283 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1284 if (symbol) 1285 { 1286 if (symbol->ValueIsAddress()) 1287 { 1288 SectionSP section_sp (symbol->GetAddressRef().GetSection()); 1289 if (section_sp) 1290 { 1291 const lldb::SectionType section_type = section_sp->GetType(); 1292 switch (section_type) 1293 { 1294 case eSectionTypeInvalid: 1295 return eAddressClassUnknown; 1296 1297 case eSectionTypeCode: 1298 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) 1299 { 1300 // For ARM we have a bit in the n_desc field of the symbol 1301 // that tells us ARM/Thumb which is bit 0x0008. 1302 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1303 return eAddressClassCodeAlternateISA; 1304 } 1305 return eAddressClassCode; 1306 1307 case eSectionTypeContainer: 1308 return eAddressClassUnknown; 1309 1310 case eSectionTypeData: 1311 case eSectionTypeDataCString: 1312 case eSectionTypeDataCStringPointers: 1313 case eSectionTypeDataSymbolAddress: 1314 case eSectionTypeData4: 1315 case eSectionTypeData8: 1316 case eSectionTypeData16: 1317 case eSectionTypeDataPointers: 1318 case eSectionTypeZeroFill: 1319 case eSectionTypeDataObjCMessageRefs: 1320 case eSectionTypeDataObjCCFStrings: 1321 case eSectionTypeGoSymtab: 1322 return eAddressClassData; 1323 1324 case eSectionTypeDebug: 1325 case eSectionTypeDWARFDebugAbbrev: 1326 case eSectionTypeDWARFDebugAddr: 1327 case eSectionTypeDWARFDebugAranges: 1328 case eSectionTypeDWARFDebugFrame: 1329 case eSectionTypeDWARFDebugInfo: 1330 case eSectionTypeDWARFDebugLine: 1331 case eSectionTypeDWARFDebugLoc: 1332 case eSectionTypeDWARFDebugMacInfo: 1333 case eSectionTypeDWARFDebugPubNames: 1334 case eSectionTypeDWARFDebugPubTypes: 1335 case eSectionTypeDWARFDebugRanges: 1336 case eSectionTypeDWARFDebugStr: 1337 case eSectionTypeDWARFDebugStrOffsets: 1338 case eSectionTypeDWARFAppleNames: 1339 case eSectionTypeDWARFAppleTypes: 1340 case eSectionTypeDWARFAppleNamespaces: 1341 case eSectionTypeDWARFAppleObjC: 1342 return eAddressClassDebug; 1343 1344 case eSectionTypeEHFrame: 1345 case eSectionTypeARMexidx: 1346 case eSectionTypeARMextab: 1347 case eSectionTypeCompactUnwind: 1348 return eAddressClassRuntime; 1349 1350 case eSectionTypeELFSymbolTable: 1351 case eSectionTypeELFDynamicSymbols: 1352 case eSectionTypeELFRelocationEntries: 1353 case eSectionTypeELFDynamicLinkInfo: 1354 case eSectionTypeOther: 1355 return eAddressClassUnknown; 1356 } 1357 } 1358 } 1359 1360 const SymbolType symbol_type = symbol->GetType(); 1361 switch (symbol_type) 1362 { 1363 case eSymbolTypeAny: return eAddressClassUnknown; 1364 case eSymbolTypeAbsolute: return eAddressClassUnknown; 1365 1366 case eSymbolTypeCode: 1367 case eSymbolTypeTrampoline: 1368 case eSymbolTypeResolver: 1369 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) 1370 { 1371 // For ARM we have a bit in the n_desc field of the symbol 1372 // that tells us ARM/Thumb which is bit 0x0008. 1373 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1374 return eAddressClassCodeAlternateISA; 1375 } 1376 return eAddressClassCode; 1377 1378 case eSymbolTypeData: return eAddressClassData; 1379 case eSymbolTypeRuntime: return eAddressClassRuntime; 1380 case eSymbolTypeException: return eAddressClassRuntime; 1381 case eSymbolTypeSourceFile: return eAddressClassDebug; 1382 case eSymbolTypeHeaderFile: return eAddressClassDebug; 1383 case eSymbolTypeObjectFile: return eAddressClassDebug; 1384 case eSymbolTypeCommonBlock: return eAddressClassDebug; 1385 case eSymbolTypeBlock: return eAddressClassDebug; 1386 case eSymbolTypeLocal: return eAddressClassData; 1387 case eSymbolTypeParam: return eAddressClassData; 1388 case eSymbolTypeVariable: return eAddressClassData; 1389 case eSymbolTypeVariableType: return eAddressClassDebug; 1390 case eSymbolTypeLineEntry: return eAddressClassDebug; 1391 case eSymbolTypeLineHeader: return eAddressClassDebug; 1392 case eSymbolTypeScopeBegin: return eAddressClassDebug; 1393 case eSymbolTypeScopeEnd: return eAddressClassDebug; 1394 case eSymbolTypeAdditional: return eAddressClassUnknown; 1395 case eSymbolTypeCompiler: return eAddressClassDebug; 1396 case eSymbolTypeInstrumentation:return eAddressClassDebug; 1397 case eSymbolTypeUndefined: return eAddressClassUnknown; 1398 case eSymbolTypeObjCClass: return eAddressClassRuntime; 1399 case eSymbolTypeObjCMetaClass: return eAddressClassRuntime; 1400 case eSymbolTypeObjCIVar: return eAddressClassRuntime; 1401 case eSymbolTypeReExported: return eAddressClassRuntime; 1402 } 1403 } 1404 } 1405 return eAddressClassUnknown; 1406 } 1407 1408 Symtab * 1409 ObjectFileMachO::GetSymtab() 1410 { 1411 ModuleSP module_sp(GetModule()); 1412 if (module_sp) 1413 { 1414 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 1415 if (m_symtab_ap.get() == NULL) 1416 { 1417 m_symtab_ap.reset(new Symtab(this)); 1418 Mutex::Locker symtab_locker (m_symtab_ap->GetMutex()); 1419 ParseSymtab (); 1420 m_symtab_ap->Finalize (); 1421 } 1422 } 1423 return m_symtab_ap.get(); 1424 } 1425 1426 bool 1427 ObjectFileMachO::IsStripped () 1428 { 1429 if (m_dysymtab.cmd == 0) 1430 { 1431 ModuleSP module_sp(GetModule()); 1432 if (module_sp) 1433 { 1434 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1435 for (uint32_t i=0; i<m_header.ncmds; ++i) 1436 { 1437 const lldb::offset_t load_cmd_offset = offset; 1438 1439 load_command lc; 1440 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1441 break; 1442 if (lc.cmd == LC_DYSYMTAB) 1443 { 1444 m_dysymtab.cmd = lc.cmd; 1445 m_dysymtab.cmdsize = lc.cmdsize; 1446 if (m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == NULL) 1447 { 1448 // Clear m_dysymtab if we were unable to read all items from the load command 1449 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 1450 } 1451 } 1452 offset = load_cmd_offset + lc.cmdsize; 1453 } 1454 } 1455 } 1456 if (m_dysymtab.cmd) 1457 return m_dysymtab.nlocalsym <= 1; 1458 return false; 1459 } 1460 1461 void 1462 ObjectFileMachO::CreateSections (SectionList &unified_section_list) 1463 { 1464 if (!m_sections_ap.get()) 1465 { 1466 m_sections_ap.reset(new SectionList()); 1467 1468 const bool is_dsym = (m_header.filetype == MH_DSYM); 1469 lldb::user_id_t segID = 0; 1470 lldb::user_id_t sectID = 0; 1471 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1472 uint32_t i; 1473 const bool is_core = GetType() == eTypeCoreFile; 1474 //bool dump_sections = false; 1475 ModuleSP module_sp (GetModule()); 1476 // First look up any LC_ENCRYPTION_INFO load commands 1477 typedef RangeArray<uint32_t, uint32_t, 8> EncryptedFileRanges; 1478 EncryptedFileRanges encrypted_file_ranges; 1479 encryption_info_command encryption_cmd; 1480 for (i=0; i<m_header.ncmds; ++i) 1481 { 1482 const lldb::offset_t load_cmd_offset = offset; 1483 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1484 break; 1485 1486 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for 1487 // the 3 fields we care about, so treat them the same. 1488 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) 1489 { 1490 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) 1491 { 1492 if (encryption_cmd.cryptid != 0) 1493 { 1494 EncryptedFileRanges::Entry entry; 1495 entry.SetRangeBase(encryption_cmd.cryptoff); 1496 entry.SetByteSize(encryption_cmd.cryptsize); 1497 encrypted_file_ranges.Append(entry); 1498 } 1499 } 1500 } 1501 offset = load_cmd_offset + encryption_cmd.cmdsize; 1502 } 1503 1504 bool section_file_addresses_changed = false; 1505 1506 offset = MachHeaderSizeFromMagic(m_header.magic); 1507 1508 struct segment_command_64 load_cmd; 1509 for (i=0; i<m_header.ncmds; ++i) 1510 { 1511 const lldb::offset_t load_cmd_offset = offset; 1512 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1513 break; 1514 1515 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1516 { 1517 if (m_data.GetU8(&offset, (uint8_t*)load_cmd.segname, 16)) 1518 { 1519 bool add_section = true; 1520 bool add_to_unified = true; 1521 ConstString const_segname (load_cmd.segname, std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 1522 1523 SectionSP unified_section_sp(unified_section_list.FindSectionByName(const_segname)); 1524 if (is_dsym && unified_section_sp) 1525 { 1526 if (const_segname == GetSegmentNameLINKEDIT()) 1527 { 1528 // We need to keep the __LINKEDIT segment private to this object file only 1529 add_to_unified = false; 1530 } 1531 else 1532 { 1533 // This is the dSYM file and this section has already been created by 1534 // the object file, no need to create it. 1535 add_section = false; 1536 } 1537 } 1538 load_cmd.vmaddr = m_data.GetAddress(&offset); 1539 load_cmd.vmsize = m_data.GetAddress(&offset); 1540 load_cmd.fileoff = m_data.GetAddress(&offset); 1541 load_cmd.filesize = m_data.GetAddress(&offset); 1542 if (m_length != 0 && load_cmd.filesize != 0) 1543 { 1544 if (load_cmd.fileoff > m_length) 1545 { 1546 // We have a load command that says it extends past the end of the file. This is likely 1547 // a corrupt file. We don't have any way to return an error condition here (this method 1548 // was likely invoked from something like ObjectFile::GetSectionList()) -- all we can do 1549 // is null out the SectionList vector and if a process has been set up, dump a message 1550 // to stdout. The most common case here is core file debugging with a truncated file. 1551 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1552 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", 1553 i, 1554 lc_segment_name, 1555 load_cmd.fileoff, 1556 m_length); 1557 1558 load_cmd.fileoff = 0; 1559 load_cmd.filesize = 0; 1560 } 1561 1562 if (load_cmd.fileoff + load_cmd.filesize > m_length) 1563 { 1564 // We have a load command that says it extends past the end of the file. This is likely 1565 // a corrupt file. We don't have any way to return an error condition here (this method 1566 // was likely invoked from something like ObjectFile::GetSectionList()) -- all we can do 1567 // is null out the SectionList vector and if a process has been set up, dump a message 1568 // to stdout. The most common case here is core file debugging with a truncated file. 1569 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1570 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", 1571 i, 1572 lc_segment_name, 1573 load_cmd.fileoff + load_cmd.filesize, 1574 m_length); 1575 1576 // Tuncase the length 1577 load_cmd.filesize = m_length - load_cmd.fileoff; 1578 } 1579 } 1580 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1581 { 1582 1583 const bool segment_is_encrypted = (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1584 1585 // Keep a list of mach segments around in case we need to 1586 // get at data that isn't stored in the abstracted Sections. 1587 m_mach_segments.push_back (load_cmd); 1588 1589 // Use a segment ID of the segment index shifted left by 8 so they 1590 // never conflict with any of the sections. 1591 SectionSP segment_sp; 1592 if (add_section && (const_segname || is_core)) 1593 { 1594 segment_sp.reset(new Section (module_sp, // Module to which this section belongs 1595 this, // Object file to which this sections belongs 1596 ++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 1597 const_segname, // Name of this section 1598 eSectionTypeContainer, // This section is a container of other sections. 1599 load_cmd.vmaddr, // File VM address == addresses as they are found in the object file 1600 load_cmd.vmsize, // VM size in bytes of this section 1601 load_cmd.fileoff, // Offset to the data for this section in the file 1602 load_cmd.filesize, // Size in bytes of this section as found in the file 1603 0, // Segments have no alignment information 1604 load_cmd.flags)); // Flags for this section 1605 1606 segment_sp->SetIsEncrypted (segment_is_encrypted); 1607 m_sections_ap->AddSection(segment_sp); 1608 if (add_to_unified) 1609 unified_section_list.AddSection(segment_sp); 1610 } 1611 else if (unified_section_sp) 1612 { 1613 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) 1614 { 1615 // Check to see if the module was read from memory? 1616 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) 1617 { 1618 // We have a module that is in memory and needs to have its 1619 // file address adjusted. We need to do this because when we 1620 // load a file from memory, its addresses will be slid already, 1621 // yet the addresses in the new symbol file will still be unslid. 1622 // Since everything is stored as section offset, this shouldn't 1623 // cause any problems. 1624 1625 // Make sure we've parsed the symbol table from the 1626 // ObjectFile before we go around changing its Sections. 1627 module_sp->GetObjectFile()->GetSymtab(); 1628 // eh_frame would present the same problems but we parse that on 1629 // a per-function basis as-needed so it's more difficult to 1630 // remove its use of the Sections. Realistically, the environments 1631 // where this code path will be taken will not have eh_frame sections. 1632 1633 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1634 1635 // Notify the module that the section addresses have been changed once 1636 // we're done so any file-address caches can be updated. 1637 section_file_addresses_changed = true; 1638 } 1639 } 1640 m_sections_ap->AddSection(unified_section_sp); 1641 } 1642 1643 struct section_64 sect64; 1644 ::memset (§64, 0, sizeof(sect64)); 1645 // Push a section into our mach sections for the section at 1646 // index zero (NO_SECT) if we don't have any mach sections yet... 1647 if (m_mach_sections.empty()) 1648 m_mach_sections.push_back(sect64); 1649 uint32_t segment_sect_idx; 1650 const lldb::user_id_t first_segment_sectID = sectID + 1; 1651 1652 1653 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1654 for (segment_sect_idx=0; segment_sect_idx<load_cmd.nsects; ++segment_sect_idx) 1655 { 1656 if (m_data.GetU8(&offset, (uint8_t*)sect64.sectname, sizeof(sect64.sectname)) == NULL) 1657 break; 1658 if (m_data.GetU8(&offset, (uint8_t*)sect64.segname, sizeof(sect64.segname)) == NULL) 1659 break; 1660 sect64.addr = m_data.GetAddress(&offset); 1661 sect64.size = m_data.GetAddress(&offset); 1662 1663 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1664 break; 1665 1666 // Keep a list of mach sections around in case we need to 1667 // get at data that isn't stored in the abstracted Sections. 1668 m_mach_sections.push_back (sect64); 1669 1670 if (add_section) 1671 { 1672 ConstString section_name (sect64.sectname, std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1673 if (!const_segname) 1674 { 1675 // We have a segment with no name so we need to conjure up 1676 // segments that correspond to the section's segname if there 1677 // isn't already such a section. If there is such a section, 1678 // we resize the section so that it spans all sections. 1679 // We also mark these sections as fake so address matches don't 1680 // hit if they land in the gaps between the child sections. 1681 const_segname.SetTrimmedCStringWithLength(sect64.segname, sizeof(sect64.segname)); 1682 segment_sp = unified_section_list.FindSectionByName (const_segname); 1683 if (segment_sp.get()) 1684 { 1685 Section *segment = segment_sp.get(); 1686 // Grow the section size as needed. 1687 const lldb::addr_t sect64_min_addr = sect64.addr; 1688 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1689 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1690 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1691 const lldb::addr_t curr_seg_max_addr = curr_seg_min_addr + curr_seg_byte_size; 1692 if (sect64_min_addr >= curr_seg_min_addr) 1693 { 1694 const lldb::addr_t new_seg_byte_size = sect64_max_addr - curr_seg_min_addr; 1695 // Only grow the section size if needed 1696 if (new_seg_byte_size > curr_seg_byte_size) 1697 segment->SetByteSize (new_seg_byte_size); 1698 } 1699 else 1700 { 1701 // We need to change the base address of the segment and 1702 // adjust the child section offsets for all existing children. 1703 const lldb::addr_t slide_amount = sect64_min_addr - curr_seg_min_addr; 1704 segment->Slide(slide_amount, false); 1705 segment->GetChildren().Slide(-slide_amount, false); 1706 segment->SetByteSize (curr_seg_max_addr - sect64_min_addr); 1707 } 1708 1709 // Grow the section size as needed. 1710 if (sect64.offset) 1711 { 1712 const lldb::addr_t segment_min_file_offset = segment->GetFileOffset(); 1713 const lldb::addr_t segment_max_file_offset = segment_min_file_offset + segment->GetFileSize(); 1714 1715 const lldb::addr_t section_min_file_offset = sect64.offset; 1716 const lldb::addr_t section_max_file_offset = section_min_file_offset + sect64.size; 1717 const lldb::addr_t new_file_offset = std::min (section_min_file_offset, segment_min_file_offset); 1718 const lldb::addr_t new_file_size = std::max (section_max_file_offset, segment_max_file_offset) - new_file_offset; 1719 segment->SetFileOffset (new_file_offset); 1720 segment->SetFileSize (new_file_size); 1721 } 1722 } 1723 else 1724 { 1725 // Create a fake section for the section's named segment 1726 segment_sp.reset(new Section (segment_sp, // Parent section 1727 module_sp, // Module to which this section belongs 1728 this, // Object file to which this section belongs 1729 ++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 1730 const_segname, // Name of this section 1731 eSectionTypeContainer, // This section is a container of other sections. 1732 sect64.addr, // File VM address == addresses as they are found in the object file 1733 sect64.size, // VM size in bytes of this section 1734 sect64.offset, // Offset to the data for this section in the file 1735 sect64.offset ? sect64.size : 0, // Size in bytes of this section as found in the file 1736 sect64.align, 1737 load_cmd.flags)); // Flags for this section 1738 segment_sp->SetIsFake(true); 1739 1740 m_sections_ap->AddSection(segment_sp); 1741 if (add_to_unified) 1742 unified_section_list.AddSection(segment_sp); 1743 segment_sp->SetIsEncrypted (segment_is_encrypted); 1744 } 1745 } 1746 assert (segment_sp.get()); 1747 1748 lldb::SectionType sect_type = eSectionTypeOther; 1749 1750 if (sect64.flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1751 sect_type = eSectionTypeCode; 1752 else 1753 { 1754 uint32_t mach_sect_type = sect64.flags & SECTION_TYPE; 1755 static ConstString g_sect_name_objc_data ("__objc_data"); 1756 static ConstString g_sect_name_objc_msgrefs ("__objc_msgrefs"); 1757 static ConstString g_sect_name_objc_selrefs ("__objc_selrefs"); 1758 static ConstString g_sect_name_objc_classrefs ("__objc_classrefs"); 1759 static ConstString g_sect_name_objc_superrefs ("__objc_superrefs"); 1760 static ConstString g_sect_name_objc_const ("__objc_const"); 1761 static ConstString g_sect_name_objc_classlist ("__objc_classlist"); 1762 static ConstString g_sect_name_cfstring ("__cfstring"); 1763 1764 static ConstString g_sect_name_dwarf_debug_abbrev ("__debug_abbrev"); 1765 static ConstString g_sect_name_dwarf_debug_aranges ("__debug_aranges"); 1766 static ConstString g_sect_name_dwarf_debug_frame ("__debug_frame"); 1767 static ConstString g_sect_name_dwarf_debug_info ("__debug_info"); 1768 static ConstString g_sect_name_dwarf_debug_line ("__debug_line"); 1769 static ConstString g_sect_name_dwarf_debug_loc ("__debug_loc"); 1770 static ConstString g_sect_name_dwarf_debug_macinfo ("__debug_macinfo"); 1771 static ConstString g_sect_name_dwarf_debug_pubnames ("__debug_pubnames"); 1772 static ConstString g_sect_name_dwarf_debug_pubtypes ("__debug_pubtypes"); 1773 static ConstString g_sect_name_dwarf_debug_ranges ("__debug_ranges"); 1774 static ConstString g_sect_name_dwarf_debug_str ("__debug_str"); 1775 static ConstString g_sect_name_dwarf_apple_names ("__apple_names"); 1776 static ConstString g_sect_name_dwarf_apple_types ("__apple_types"); 1777 static ConstString g_sect_name_dwarf_apple_namespaces ("__apple_namespac"); 1778 static ConstString g_sect_name_dwarf_apple_objc ("__apple_objc"); 1779 static ConstString g_sect_name_eh_frame ("__eh_frame"); 1780 static ConstString g_sect_name_compact_unwind ("__unwind_info"); 1781 static ConstString g_sect_name_text ("__text"); 1782 static ConstString g_sect_name_data ("__data"); 1783 static ConstString g_sect_name_go_symtab ("__gosymtab"); 1784 1785 1786 if (section_name == g_sect_name_dwarf_debug_abbrev) 1787 sect_type = eSectionTypeDWARFDebugAbbrev; 1788 else if (section_name == g_sect_name_dwarf_debug_aranges) 1789 sect_type = eSectionTypeDWARFDebugAranges; 1790 else if (section_name == g_sect_name_dwarf_debug_frame) 1791 sect_type = eSectionTypeDWARFDebugFrame; 1792 else if (section_name == g_sect_name_dwarf_debug_info) 1793 sect_type = eSectionTypeDWARFDebugInfo; 1794 else if (section_name == g_sect_name_dwarf_debug_line) 1795 sect_type = eSectionTypeDWARFDebugLine; 1796 else if (section_name == g_sect_name_dwarf_debug_loc) 1797 sect_type = eSectionTypeDWARFDebugLoc; 1798 else if (section_name == g_sect_name_dwarf_debug_macinfo) 1799 sect_type = eSectionTypeDWARFDebugMacInfo; 1800 else if (section_name == g_sect_name_dwarf_debug_pubnames) 1801 sect_type = eSectionTypeDWARFDebugPubNames; 1802 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 1803 sect_type = eSectionTypeDWARFDebugPubTypes; 1804 else if (section_name == g_sect_name_dwarf_debug_ranges) 1805 sect_type = eSectionTypeDWARFDebugRanges; 1806 else if (section_name == g_sect_name_dwarf_debug_str) 1807 sect_type = eSectionTypeDWARFDebugStr; 1808 else if (section_name == g_sect_name_dwarf_apple_names) 1809 sect_type = eSectionTypeDWARFAppleNames; 1810 else if (section_name == g_sect_name_dwarf_apple_types) 1811 sect_type = eSectionTypeDWARFAppleTypes; 1812 else if (section_name == g_sect_name_dwarf_apple_namespaces) 1813 sect_type = eSectionTypeDWARFAppleNamespaces; 1814 else if (section_name == g_sect_name_dwarf_apple_objc) 1815 sect_type = eSectionTypeDWARFAppleObjC; 1816 else if (section_name == g_sect_name_objc_selrefs) 1817 sect_type = eSectionTypeDataCStringPointers; 1818 else if (section_name == g_sect_name_objc_msgrefs) 1819 sect_type = eSectionTypeDataObjCMessageRefs; 1820 else if (section_name == g_sect_name_eh_frame) 1821 sect_type = eSectionTypeEHFrame; 1822 else if (section_name == g_sect_name_compact_unwind) 1823 sect_type = eSectionTypeCompactUnwind; 1824 else if (section_name == g_sect_name_cfstring) 1825 sect_type = eSectionTypeDataObjCCFStrings; 1826 else if (section_name == g_sect_name_go_symtab) 1827 sect_type = eSectionTypeGoSymtab; 1828 else if (section_name == g_sect_name_objc_data || 1829 section_name == g_sect_name_objc_classrefs || 1830 section_name == g_sect_name_objc_superrefs || 1831 section_name == g_sect_name_objc_const || 1832 section_name == g_sect_name_objc_classlist) 1833 { 1834 sect_type = eSectionTypeDataPointers; 1835 } 1836 1837 if (sect_type == eSectionTypeOther) 1838 { 1839 switch (mach_sect_type) 1840 { 1841 // TODO: categorize sections by other flags for regular sections 1842 case S_REGULAR: 1843 if (section_name == g_sect_name_text) 1844 sect_type = eSectionTypeCode; 1845 else if (section_name == g_sect_name_data) 1846 sect_type = eSectionTypeData; 1847 else 1848 sect_type = eSectionTypeOther; 1849 break; 1850 case S_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1851 case S_CSTRING_LITERALS: sect_type = eSectionTypeDataCString; break; // section with only literal C strings 1852 case S_4BYTE_LITERALS: sect_type = eSectionTypeData4; break; // section with only 4 byte literals 1853 case S_8BYTE_LITERALS: sect_type = eSectionTypeData8; break; // section with only 8 byte literals 1854 case S_LITERAL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only pointers to literals 1855 case S_NON_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only non-lazy symbol pointers 1856 case S_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only lazy symbol pointers 1857 case S_SYMBOL_STUBS: sect_type = eSectionTypeCode; break; // section with only symbol stubs, byte size of stub in the reserved2 field 1858 case S_MOD_INIT_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for initialization 1859 case S_MOD_TERM_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for termination 1860 case S_COALESCED: sect_type = eSectionTypeOther; break; 1861 case S_GB_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1862 case S_INTERPOSING: sect_type = eSectionTypeCode; break; // section with only pairs of function pointers for interposing 1863 case S_16BYTE_LITERALS: sect_type = eSectionTypeData16; break; // section with only 16 byte literals 1864 case S_DTRACE_DOF: sect_type = eSectionTypeDebug; break; 1865 case S_LAZY_DYLIB_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; 1866 default: break; 1867 } 1868 } 1869 } 1870 1871 SectionSP section_sp(new Section (segment_sp, 1872 module_sp, 1873 this, 1874 ++sectID, 1875 section_name, 1876 sect_type, 1877 sect64.addr - segment_sp->GetFileAddress(), 1878 sect64.size, 1879 sect64.offset, 1880 sect64.offset == 0 ? 0 : sect64.size, 1881 sect64.align, 1882 sect64.flags)); 1883 // Set the section to be encrypted to match the segment 1884 1885 bool section_is_encrypted = false; 1886 if (!segment_is_encrypted && load_cmd.filesize != 0) 1887 section_is_encrypted = encrypted_file_ranges.FindEntryThatContains(sect64.offset) != NULL; 1888 1889 section_sp->SetIsEncrypted (segment_is_encrypted || section_is_encrypted); 1890 segment_sp->GetChildren().AddSection(section_sp); 1891 1892 if (segment_sp->IsFake()) 1893 { 1894 segment_sp.reset(); 1895 const_segname.Clear(); 1896 } 1897 } 1898 } 1899 if (segment_sp && is_dsym) 1900 { 1901 if (first_segment_sectID <= sectID) 1902 { 1903 lldb::user_id_t sect_uid; 1904 for (sect_uid = first_segment_sectID; sect_uid <= sectID; ++sect_uid) 1905 { 1906 SectionSP curr_section_sp(segment_sp->GetChildren().FindSectionByID (sect_uid)); 1907 SectionSP next_section_sp; 1908 if (sect_uid + 1 <= sectID) 1909 next_section_sp = segment_sp->GetChildren().FindSectionByID (sect_uid+1); 1910 1911 if (curr_section_sp.get()) 1912 { 1913 if (curr_section_sp->GetByteSize() == 0) 1914 { 1915 if (next_section_sp.get() != NULL) 1916 curr_section_sp->SetByteSize ( next_section_sp->GetFileAddress() - curr_section_sp->GetFileAddress() ); 1917 else 1918 curr_section_sp->SetByteSize ( load_cmd.vmsize ); 1919 } 1920 } 1921 } 1922 } 1923 } 1924 } 1925 } 1926 } 1927 else if (load_cmd.cmd == LC_DYSYMTAB) 1928 { 1929 m_dysymtab.cmd = load_cmd.cmd; 1930 m_dysymtab.cmdsize = load_cmd.cmdsize; 1931 m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1932 } 1933 1934 offset = load_cmd_offset + load_cmd.cmdsize; 1935 } 1936 1937 1938 if (section_file_addresses_changed && module_sp.get()) 1939 { 1940 module_sp->SectionFileAddressesChanged(); 1941 } 1942 } 1943 } 1944 1945 class MachSymtabSectionInfo 1946 { 1947 public: 1948 1949 MachSymtabSectionInfo (SectionList *section_list) : 1950 m_section_list (section_list), 1951 m_section_infos() 1952 { 1953 // Get the number of sections down to a depth of 1 to include 1954 // all segments and their sections, but no other sections that 1955 // may be added for debug map or 1956 m_section_infos.resize(section_list->GetNumSections(1)); 1957 } 1958 1959 1960 SectionSP 1961 GetSection (uint8_t n_sect, addr_t file_addr) 1962 { 1963 if (n_sect == 0) 1964 return SectionSP(); 1965 if (n_sect < m_section_infos.size()) 1966 { 1967 if (!m_section_infos[n_sect].section_sp) 1968 { 1969 SectionSP section_sp (m_section_list->FindSectionByID (n_sect)); 1970 m_section_infos[n_sect].section_sp = section_sp; 1971 if (section_sp) 1972 { 1973 m_section_infos[n_sect].vm_range.SetBaseAddress (section_sp->GetFileAddress()); 1974 m_section_infos[n_sect].vm_range.SetByteSize (section_sp->GetByteSize()); 1975 } 1976 else 1977 { 1978 Host::SystemLog (Host::eSystemLogError, "error: unable to find section for section %u\n", n_sect); 1979 } 1980 } 1981 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) 1982 { 1983 // Symbol is in section. 1984 return m_section_infos[n_sect].section_sp; 1985 } 1986 else if (m_section_infos[n_sect].vm_range.GetByteSize () == 0 && 1987 m_section_infos[n_sect].vm_range.GetBaseAddress() == file_addr) 1988 { 1989 // Symbol is in section with zero size, but has the same start 1990 // address as the section. This can happen with linker symbols 1991 // (symbols that start with the letter 'l' or 'L'. 1992 return m_section_infos[n_sect].section_sp; 1993 } 1994 } 1995 return m_section_list->FindSectionContainingFileAddress(file_addr); 1996 } 1997 1998 protected: 1999 struct SectionInfo 2000 { 2001 SectionInfo () : 2002 vm_range(), 2003 section_sp () 2004 { 2005 } 2006 2007 VMRange vm_range; 2008 SectionSP section_sp; 2009 }; 2010 SectionList *m_section_list; 2011 std::vector<SectionInfo> m_section_infos; 2012 }; 2013 2014 struct TrieEntry 2015 { 2016 TrieEntry () : 2017 name(), 2018 address(LLDB_INVALID_ADDRESS), 2019 flags (0), 2020 other(0), 2021 import_name() 2022 { 2023 } 2024 2025 void 2026 Clear () 2027 { 2028 name.Clear(); 2029 address = LLDB_INVALID_ADDRESS; 2030 flags = 0; 2031 other = 0; 2032 import_name.Clear(); 2033 } 2034 2035 void 2036 Dump () const 2037 { 2038 printf ("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 2039 static_cast<unsigned long long>(address), 2040 static_cast<unsigned long long>(flags), 2041 static_cast<unsigned long long>(other), name.GetCString()); 2042 if (import_name) 2043 printf (" -> \"%s\"\n", import_name.GetCString()); 2044 else 2045 printf ("\n"); 2046 } 2047 ConstString name; 2048 uint64_t address; 2049 uint64_t flags; 2050 uint64_t other; 2051 ConstString import_name; 2052 }; 2053 2054 struct TrieEntryWithOffset 2055 { 2056 lldb::offset_t nodeOffset; 2057 TrieEntry entry; 2058 2059 TrieEntryWithOffset (lldb::offset_t offset) : 2060 nodeOffset (offset), 2061 entry() 2062 { 2063 } 2064 2065 void 2066 Dump (uint32_t idx) const 2067 { 2068 printf ("[%3u] 0x%16.16llx: ", idx, 2069 static_cast<unsigned long long>(nodeOffset)); 2070 entry.Dump(); 2071 } 2072 2073 bool 2074 operator<(const TrieEntryWithOffset& other) const 2075 { 2076 return ( nodeOffset < other.nodeOffset ); 2077 } 2078 }; 2079 2080 static void 2081 ParseTrieEntries (DataExtractor &data, 2082 lldb::offset_t offset, 2083 const bool is_arm, 2084 std::vector<llvm::StringRef> &nameSlices, 2085 std::set<lldb::addr_t> &resolver_addresses, 2086 std::vector<TrieEntryWithOffset>& output) 2087 { 2088 if (!data.ValidOffset(offset)) 2089 return; 2090 2091 const uint64_t terminalSize = data.GetULEB128(&offset); 2092 lldb::offset_t children_offset = offset + terminalSize; 2093 if ( terminalSize != 0 ) { 2094 TrieEntryWithOffset e (offset); 2095 e.entry.flags = data.GetULEB128(&offset); 2096 const char *import_name = NULL; 2097 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT ) { 2098 e.entry.address = 0; 2099 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2100 import_name = data.GetCStr(&offset); 2101 } 2102 else { 2103 e.entry.address = data.GetULEB128(&offset); 2104 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER ) 2105 { 2106 e.entry.other = data.GetULEB128(&offset); 2107 uint64_t resolver_addr = e.entry.other; 2108 if (is_arm) 2109 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2110 resolver_addresses.insert(resolver_addr); 2111 } 2112 else 2113 e.entry.other = 0; 2114 } 2115 // Only add symbols that are reexport symbols with a valid import name 2116 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && import_name[0]) 2117 { 2118 std::string name; 2119 if (!nameSlices.empty()) 2120 { 2121 for (auto name_slice: nameSlices) 2122 name.append(name_slice.data(), name_slice.size()); 2123 } 2124 if (name.size() > 1) 2125 { 2126 // Skip the leading '_' 2127 e.entry.name.SetCStringWithLength(name.c_str() + 1,name.size() - 1); 2128 } 2129 if (import_name) 2130 { 2131 // Skip the leading '_' 2132 e.entry.import_name.SetCString(import_name+1); 2133 } 2134 output.push_back(e); 2135 } 2136 } 2137 2138 const uint8_t childrenCount = data.GetU8(&children_offset); 2139 for (uint8_t i=0; i < childrenCount; ++i) { 2140 nameSlices.push_back(data.GetCStr(&children_offset)); 2141 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2142 if (childNodeOffset) 2143 { 2144 ParseTrieEntries(data, 2145 childNodeOffset, 2146 is_arm, 2147 nameSlices, 2148 resolver_addresses, 2149 output); 2150 } 2151 nameSlices.pop_back(); 2152 } 2153 } 2154 2155 size_t 2156 ObjectFileMachO::ParseSymtab () 2157 { 2158 Timer scoped_timer(__PRETTY_FUNCTION__, 2159 "ObjectFileMachO::ParseSymtab () module = %s", 2160 m_file.GetFilename().AsCString("")); 2161 ModuleSP module_sp (GetModule()); 2162 if (!module_sp) 2163 return 0; 2164 2165 struct symtab_command symtab_load_command = { 0, 0, 0, 0, 0, 0 }; 2166 struct linkedit_data_command function_starts_load_command = { 0, 0, 0, 0 }; 2167 struct dyld_info_command dyld_info = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; 2168 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2169 FunctionStarts function_starts; 2170 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2171 uint32_t i; 2172 FileSpecList dylib_files; 2173 Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_SYMBOLS)); 2174 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 2175 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 2176 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 2177 2178 for (i=0; i<m_header.ncmds; ++i) 2179 { 2180 const lldb::offset_t cmd_offset = offset; 2181 // Read in the load command and load command size 2182 struct load_command lc; 2183 if (m_data.GetU32(&offset, &lc, 2) == NULL) 2184 break; 2185 // Watch for the symbol table load command 2186 switch (lc.cmd) 2187 { 2188 case LC_SYMTAB: 2189 symtab_load_command.cmd = lc.cmd; 2190 symtab_load_command.cmdsize = lc.cmdsize; 2191 // Read in the rest of the symtab load command 2192 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 0) // fill in symoff, nsyms, stroff, strsize fields 2193 return 0; 2194 if (symtab_load_command.symoff == 0) 2195 { 2196 if (log) 2197 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2198 return 0; 2199 } 2200 2201 if (symtab_load_command.stroff == 0) 2202 { 2203 if (log) 2204 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2205 return 0; 2206 } 2207 2208 if (symtab_load_command.nsyms == 0) 2209 { 2210 if (log) 2211 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2212 return 0; 2213 } 2214 2215 if (symtab_load_command.strsize == 0) 2216 { 2217 if (log) 2218 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2219 return 0; 2220 } 2221 break; 2222 2223 case LC_DYLD_INFO: 2224 case LC_DYLD_INFO_ONLY: 2225 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) 2226 { 2227 dyld_info.cmd = lc.cmd; 2228 dyld_info.cmdsize = lc.cmdsize; 2229 } 2230 else 2231 { 2232 memset (&dyld_info, 0, sizeof(dyld_info)); 2233 } 2234 break; 2235 2236 case LC_LOAD_DYLIB: 2237 case LC_LOAD_WEAK_DYLIB: 2238 case LC_REEXPORT_DYLIB: 2239 case LC_LOADFVMLIB: 2240 case LC_LOAD_UPWARD_DYLIB: 2241 { 2242 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2243 const char *path = m_data.PeekCStr(name_offset); 2244 if (path) 2245 { 2246 FileSpec file_spec(path, false); 2247 // Strip the path if there is @rpath, @executable, etc so we just use the basename 2248 if (path[0] == '@') 2249 file_spec.GetDirectory().Clear(); 2250 2251 if (lc.cmd == LC_REEXPORT_DYLIB) 2252 { 2253 m_reexported_dylibs.AppendIfUnique(file_spec); 2254 } 2255 2256 dylib_files.Append(file_spec); 2257 } 2258 } 2259 break; 2260 2261 case LC_FUNCTION_STARTS: 2262 function_starts_load_command.cmd = lc.cmd; 2263 function_starts_load_command.cmdsize = lc.cmdsize; 2264 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == NULL) // fill in symoff, nsyms, stroff, strsize fields 2265 memset (&function_starts_load_command, 0, sizeof(function_starts_load_command)); 2266 break; 2267 2268 default: 2269 break; 2270 } 2271 offset = cmd_offset + lc.cmdsize; 2272 } 2273 2274 if (symtab_load_command.cmd) 2275 { 2276 Symtab *symtab = m_symtab_ap.get(); 2277 SectionList *section_list = GetSectionList(); 2278 if (section_list == NULL) 2279 return 0; 2280 2281 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2282 const ByteOrder byte_order = m_data.GetByteOrder(); 2283 bool bit_width_32 = addr_byte_size == 4; 2284 const size_t nlist_byte_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2285 2286 DataExtractor nlist_data (NULL, 0, byte_order, addr_byte_size); 2287 DataExtractor strtab_data (NULL, 0, byte_order, addr_byte_size); 2288 DataExtractor function_starts_data (NULL, 0, byte_order, addr_byte_size); 2289 DataExtractor indirect_symbol_index_data (NULL, 0, byte_order, addr_byte_size); 2290 DataExtractor dyld_trie_data (NULL, 0, byte_order, addr_byte_size); 2291 2292 const addr_t nlist_data_byte_size = symtab_load_command.nsyms * nlist_byte_size; 2293 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2294 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2295 2296 ProcessSP process_sp (m_process_wp.lock()); 2297 Process *process = process_sp.get(); 2298 2299 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2300 2301 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) 2302 { 2303 Target &target = process->GetTarget(); 2304 2305 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2306 2307 SectionSP linkedit_section_sp(section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2308 // Reading mach file from memory in a process or core file... 2309 2310 if (linkedit_section_sp) 2311 { 2312 addr_t linkedit_load_addr = linkedit_section_sp->GetLoadBaseAddress(&target); 2313 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) 2314 { 2315 // We might be trying to access the symbol table before the __LINKEDIT's load 2316 // address has been set in the target. We can't fail to read the symbol table, 2317 // so calculate the right address manually 2318 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage(m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2319 } 2320 2321 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2322 const addr_t symoff_addr = linkedit_load_addr + symtab_load_command.symoff - linkedit_file_offset; 2323 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - linkedit_file_offset; 2324 2325 bool data_was_read = false; 2326 2327 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 2328 if (m_header.flags & 0x80000000u && process->GetAddressByteSize() == sizeof (void*)) 2329 { 2330 // This mach-o memory file is in the dyld shared cache. If this 2331 // program is not remote and this is iOS, then this process will 2332 // share the same shared cache as the process we are debugging and 2333 // we can read the entire __LINKEDIT from the address space in this 2334 // process. This is a needed optimization that is used for local iOS 2335 // debugging only since all shared libraries in the shared cache do 2336 // not have corresponding files that exist in the file system of the 2337 // device. They have been combined into a single file. This means we 2338 // always have to load these files from memory. All of the symbol and 2339 // string tables from all of the __LINKEDIT sections from the shared 2340 // libraries in the shared cache have been merged into a single large 2341 // symbol and string table. Reading all of this symbol and string table 2342 // data across can slow down debug launch times, so we optimize this by 2343 // reading the memory for the __LINKEDIT section from this process. 2344 2345 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 2346 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 2347 bool use_lldb_cache = true; 2348 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && lldb_shared_cache != process_shared_cache) 2349 { 2350 use_lldb_cache = false; 2351 ModuleSP module_sp (GetModule()); 2352 if (module_sp) 2353 module_sp->ReportWarning ("shared cache in process does not match lldb's own shared cache, startup will be slow."); 2354 2355 } 2356 2357 PlatformSP platform_sp (target.GetPlatform()); 2358 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) 2359 { 2360 data_was_read = true; 2361 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, eByteOrderLittle); 2362 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, eByteOrderLittle); 2363 if (function_starts_load_command.cmd) 2364 { 2365 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 2366 function_starts_data.SetData ((void *)func_start_addr, function_starts_load_command.datasize, eByteOrderLittle); 2367 } 2368 } 2369 } 2370 #endif 2371 2372 if (!data_was_read) 2373 { 2374 if (memory_module_load_level == eMemoryModuleLoadLevelComplete) 2375 { 2376 DataBufferSP nlist_data_sp (ReadMemory (process_sp, symoff_addr, nlist_data_byte_size)); 2377 if (nlist_data_sp) 2378 nlist_data.SetData (nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2379 // Load strings individually from memory when loading from memory since shared cache 2380 // string tables contain strings for all symbols from all shared cached libraries 2381 //DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, strtab_data_byte_size)); 2382 //if (strtab_data_sp) 2383 // strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 2384 if (m_dysymtab.nindirectsyms != 0) 2385 { 2386 const addr_t indirect_syms_addr = linkedit_load_addr + m_dysymtab.indirectsymoff - linkedit_file_offset; 2387 DataBufferSP indirect_syms_data_sp (ReadMemory (process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 2388 if (indirect_syms_data_sp) 2389 indirect_symbol_index_data.SetData (indirect_syms_data_sp, 0, indirect_syms_data_sp->GetByteSize()); 2390 } 2391 } 2392 2393 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) 2394 { 2395 if (function_starts_load_command.cmd) 2396 { 2397 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 2398 DataBufferSP func_start_data_sp (ReadMemory (process_sp, func_start_addr, function_starts_load_command.datasize)); 2399 if (func_start_data_sp) 2400 function_starts_data.SetData (func_start_data_sp, 0, func_start_data_sp->GetByteSize()); 2401 } 2402 } 2403 } 2404 } 2405 } 2406 else 2407 { 2408 nlist_data.SetData (m_data, 2409 symtab_load_command.symoff, 2410 nlist_data_byte_size); 2411 strtab_data.SetData (m_data, 2412 symtab_load_command.stroff, 2413 strtab_data_byte_size); 2414 2415 if (dyld_info.export_size > 0) 2416 { 2417 dyld_trie_data.SetData (m_data, 2418 dyld_info.export_off, 2419 dyld_info.export_size); 2420 } 2421 2422 if (m_dysymtab.nindirectsyms != 0) 2423 { 2424 indirect_symbol_index_data.SetData (m_data, 2425 m_dysymtab.indirectsymoff, 2426 m_dysymtab.nindirectsyms * 4); 2427 } 2428 if (function_starts_load_command.cmd) 2429 { 2430 function_starts_data.SetData (m_data, 2431 function_starts_load_command.dataoff, 2432 function_starts_load_command.datasize); 2433 } 2434 } 2435 2436 if (nlist_data.GetByteSize() == 0 && memory_module_load_level == eMemoryModuleLoadLevelComplete) 2437 { 2438 if (log) 2439 module_sp->LogMessage(log, "failed to read nlist data"); 2440 return 0; 2441 } 2442 2443 2444 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2445 if (!have_strtab_data) 2446 { 2447 if (process) 2448 { 2449 if (strtab_addr == LLDB_INVALID_ADDRESS) 2450 { 2451 if (log) 2452 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2453 return 0; 2454 } 2455 } 2456 else 2457 { 2458 if (log) 2459 module_sp->LogMessage(log, "failed to read strtab data"); 2460 return 0; 2461 } 2462 } 2463 2464 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2465 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2466 const ConstString &g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2467 const ConstString &g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2468 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2469 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2470 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 2471 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 2472 SectionSP data_dirty_section_sp(section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2473 SectionSP data_const_section_sp(section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2474 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 2475 SectionSP eh_frame_section_sp; 2476 if (text_section_sp.get()) 2477 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 2478 else 2479 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 2480 2481 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2482 2483 // lldb works best if it knows the start address of all functions in a module. 2484 // Linker symbols or debug info are normally the best source of information for start addr / size but 2485 // they may be stripped in a released binary. 2486 // Two additional sources of information exist in Mach-O binaries: 2487 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each function's start address in the 2488 // binary, relative to the text section. 2489 // eh_frame - the eh_frame FDEs have the start addr & size of each function 2490 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on all modern binaries. 2491 // Binaries built to run on older releases may need to use eh_frame information. 2492 2493 if (text_section_sp && function_starts_data.GetByteSize()) 2494 { 2495 FunctionStarts::Entry function_start_entry; 2496 function_start_entry.data = false; 2497 lldb::offset_t function_start_offset = 0; 2498 function_start_entry.addr = text_section_sp->GetFileAddress(); 2499 uint64_t delta; 2500 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 0) 2501 { 2502 // Now append the current entry 2503 function_start_entry.addr += delta; 2504 function_starts.Append(function_start_entry); 2505 } 2506 } 2507 else 2508 { 2509 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the load command claiming an eh_frame 2510 // but it doesn't actually have the eh_frame content. And if we have a dSYM, we don't need to do any 2511 // of this fill-in-the-missing-symbols works anyway - the debug info should give us all the functions in 2512 // the module. 2513 if (text_section_sp.get() && eh_frame_section_sp.get() && m_type != eTypeDebugInfo) 2514 { 2515 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, eRegisterKindEHFrame, true); 2516 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2517 eh_frame.GetFunctionAddressAndSizeVector (functions); 2518 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2519 size_t count = functions.GetSize(); 2520 for (size_t i = 0; i < count; ++i) 2521 { 2522 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = functions.GetEntryAtIndex (i); 2523 if (func) 2524 { 2525 FunctionStarts::Entry function_start_entry; 2526 function_start_entry.addr = func->base - text_base_addr; 2527 function_starts.Append(function_start_entry); 2528 } 2529 } 2530 } 2531 } 2532 2533 const size_t function_starts_count = function_starts.GetSize(); 2534 2535 const user_id_t TEXT_eh_frame_sectID = 2536 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2537 : static_cast<user_id_t>(NO_SECT); 2538 2539 lldb::offset_t nlist_data_offset = 0; 2540 2541 uint32_t N_SO_index = UINT32_MAX; 2542 2543 MachSymtabSectionInfo section_info (section_list); 2544 std::vector<uint32_t> N_FUN_indexes; 2545 std::vector<uint32_t> N_NSYM_indexes; 2546 std::vector<uint32_t> N_INCL_indexes; 2547 std::vector<uint32_t> N_BRAC_indexes; 2548 std::vector<uint32_t> N_COMM_indexes; 2549 typedef std::multimap <uint64_t, uint32_t> ValueToSymbolIndexMap; 2550 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2551 typedef std::map <const char *, uint32_t> ConstNameToSymbolIndexMap; 2552 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2553 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2554 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2555 // Any symbols that get merged into another will get an entry 2556 // in this map so we know 2557 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2558 uint32_t nlist_idx = 0; 2559 Symbol *symbol_ptr = NULL; 2560 2561 uint32_t sym_idx = 0; 2562 Symbol *sym = NULL; 2563 size_t num_syms = 0; 2564 std::string memory_symbol_name; 2565 uint32_t unmapped_local_symbols_found = 0; 2566 2567 std::vector<TrieEntryWithOffset> trie_entries; 2568 std::set<lldb::addr_t> resolver_addresses; 2569 2570 if (dyld_trie_data.GetByteSize() > 0) 2571 { 2572 std::vector<llvm::StringRef> nameSlices; 2573 ParseTrieEntries (dyld_trie_data, 2574 0, 2575 is_arm, 2576 nameSlices, 2577 resolver_addresses, 2578 trie_entries); 2579 2580 ConstString text_segment_name ("__TEXT"); 2581 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 2582 if (text_segment_sp) 2583 { 2584 const lldb::addr_t text_segment_file_addr = text_segment_sp->GetFileAddress(); 2585 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) 2586 { 2587 for (auto &e : trie_entries) 2588 e.entry.address += text_segment_file_addr; 2589 } 2590 } 2591 } 2592 2593 typedef std::set<ConstString> IndirectSymbols; 2594 IndirectSymbols indirect_symbol_names; 2595 2596 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 2597 2598 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been optimized by moving LOCAL 2599 // symbols out of the memory mapped portion of the DSC. The symbol information has all been retained, 2600 // but it isn't available in the normal nlist data. However, there *are* duplicate entries of *some* 2601 // LOCAL symbols in the normal nlist data. To handle this situation correctly, we must first attempt 2602 // to parse any DSC unmapped symbol information. If we find any, we set a flag that tells the normal 2603 // nlist parser to ignore all LOCAL symbols. 2604 2605 if (m_header.flags & 0x80000000u) 2606 { 2607 // Before we can start mapping the DSC, we need to make certain the target process is actually 2608 // using the cache we can find. 2609 2610 // Next we need to determine the correct path for the dyld shared cache. 2611 2612 ArchSpec header_arch; 2613 GetArchitecture(header_arch); 2614 char dsc_path[PATH_MAX]; 2615 2616 snprintf(dsc_path, sizeof(dsc_path), "%s%s%s", 2617 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR */ 2618 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2619 header_arch.GetArchitectureName()); 2620 2621 FileSpec dsc_filespec(dsc_path, false); 2622 2623 // We need definitions of two structures in the on-disk DSC, copy them here manually 2624 struct lldb_copy_dyld_cache_header_v0 2625 { 2626 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2627 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2628 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2629 uint32_t imagesOffset; 2630 uint32_t imagesCount; 2631 uint64_t dyldBaseAddress; 2632 uint64_t codeSignatureOffset; 2633 uint64_t codeSignatureSize; 2634 uint64_t slideInfoOffset; 2635 uint64_t slideInfoSize; 2636 uint64_t localSymbolsOffset; // file offset of where local symbols are stored 2637 uint64_t localSymbolsSize; // size of local symbols information 2638 }; 2639 struct lldb_copy_dyld_cache_header_v1 2640 { 2641 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2642 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2643 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2644 uint32_t imagesOffset; 2645 uint32_t imagesCount; 2646 uint64_t dyldBaseAddress; 2647 uint64_t codeSignatureOffset; 2648 uint64_t codeSignatureSize; 2649 uint64_t slideInfoOffset; 2650 uint64_t slideInfoSize; 2651 uint64_t localSymbolsOffset; 2652 uint64_t localSymbolsSize; 2653 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 and later 2654 }; 2655 2656 struct lldb_copy_dyld_cache_mapping_info 2657 { 2658 uint64_t address; 2659 uint64_t size; 2660 uint64_t fileOffset; 2661 uint32_t maxProt; 2662 uint32_t initProt; 2663 }; 2664 2665 struct lldb_copy_dyld_cache_local_symbols_info 2666 { 2667 uint32_t nlistOffset; 2668 uint32_t nlistCount; 2669 uint32_t stringsOffset; 2670 uint32_t stringsSize; 2671 uint32_t entriesOffset; 2672 uint32_t entriesCount; 2673 }; 2674 struct lldb_copy_dyld_cache_local_symbols_entry 2675 { 2676 uint32_t dylibOffset; 2677 uint32_t nlistStartIndex; 2678 uint32_t nlistCount; 2679 }; 2680 2681 /* The dyld_cache_header has a pointer to the dyld_cache_local_symbols_info structure (localSymbolsOffset). 2682 The dyld_cache_local_symbols_info structure gives us three things: 2683 1. The start and count of the nlist records in the dyld_shared_cache file 2684 2. The start and size of the strings for these nlist records 2685 3. The start and count of dyld_cache_local_symbols_entry entries 2686 2687 There is one dyld_cache_local_symbols_entry per dylib/framework in the dyld shared cache. 2688 The "dylibOffset" field is the Mach-O header of this dylib/framework in the dyld shared cache. 2689 The dyld_cache_local_symbols_entry also lists the start of this dylib/framework's nlist records 2690 and the count of how many nlist records there are for this dylib/framework. 2691 */ 2692 2693 // Process the dsc header to find the unmapped symbols 2694 // 2695 // Save some VM space, do not map the entire cache in one shot. 2696 2697 DataBufferSP dsc_data_sp; 2698 dsc_data_sp = dsc_filespec.MemoryMapFileContentsIfLocal(0, sizeof(struct lldb_copy_dyld_cache_header_v1)); 2699 2700 if (dsc_data_sp) 2701 { 2702 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2703 2704 char version_str[17]; 2705 int version = -1; 2706 lldb::offset_t offset = 0; 2707 memcpy (version_str, dsc_header_data.GetData (&offset, 16), 16); 2708 version_str[16] = '\0'; 2709 if (strncmp (version_str, "dyld_v", 6) == 0 && isdigit (version_str[6])) 2710 { 2711 int v; 2712 if (::sscanf (version_str + 6, "%d", &v) == 1) 2713 { 2714 version = v; 2715 } 2716 } 2717 2718 UUID dsc_uuid; 2719 if (version >= 1) 2720 { 2721 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, uuid); 2722 uint8_t uuid_bytes[sizeof (uuid_t)]; 2723 memcpy (uuid_bytes, dsc_header_data.GetData (&offset, sizeof (uuid_t)), sizeof (uuid_t)); 2724 dsc_uuid.SetBytes (uuid_bytes); 2725 } 2726 2727 bool uuid_match = true; 2728 if (dsc_uuid.IsValid() && process) 2729 { 2730 UUID shared_cache_uuid(GetProcessSharedCacheUUID(process)); 2731 2732 if (shared_cache_uuid.IsValid() && dsc_uuid != shared_cache_uuid) 2733 { 2734 // The on-disk dyld_shared_cache file is not the same as the one in this 2735 // process' memory, don't use it. 2736 uuid_match = false; 2737 ModuleSP module_sp (GetModule()); 2738 if (module_sp) 2739 module_sp->ReportWarning ("process shared cache does not match on-disk dyld_shared_cache file, some symbol names will be missing."); 2740 } 2741 } 2742 2743 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2744 2745 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2746 2747 // If the mappingOffset points to a location inside the header, we've 2748 // opened an old dyld shared cache, and should not proceed further. 2749 if (uuid_match && mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v0)) 2750 { 2751 2752 DataBufferSP dsc_mapping_info_data_sp = dsc_filespec.MemoryMapFileContentsIfLocal(mappingOffset, sizeof (struct lldb_copy_dyld_cache_mapping_info)); 2753 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, byte_order, addr_byte_size); 2754 offset = 0; 2755 2756 // The File addresses (from the in-memory Mach-O load commands) for the shared libraries 2757 // in the shared library cache need to be adjusted by an offset to match up with the 2758 // dylibOffset identifying field in the dyld_cache_local_symbol_entry's. This offset is 2759 // recorded in mapping_offset_value. 2760 const uint64_t mapping_offset_value = dsc_mapping_info_data.GetU64(&offset); 2761 2762 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2763 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2764 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2765 2766 if (localSymbolsOffset && localSymbolsSize) 2767 { 2768 // Map the local symbols 2769 if (DataBufferSP dsc_local_symbols_data_sp = dsc_filespec.MemoryMapFileContentsIfLocal(localSymbolsOffset, localSymbolsSize)) 2770 { 2771 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, byte_order, addr_byte_size); 2772 2773 offset = 0; 2774 2775 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2776 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2777 UndefinedNameToDescMap undefined_name_to_desc; 2778 SymbolIndexToName reexport_shlib_needs_fixup; 2779 2780 2781 // Read the local_symbols_infos struct in one shot 2782 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2783 dsc_local_symbols_data.GetU32(&offset, &local_symbols_info.nlistOffset, 6); 2784 2785 SectionSP text_section_sp(section_list->FindSectionByName(GetSegmentNameTEXT())); 2786 2787 uint32_t header_file_offset = (text_section_sp->GetFileAddress() - mapping_offset_value); 2788 2789 offset = local_symbols_info.entriesOffset; 2790 for (uint32_t entry_index = 0; entry_index < local_symbols_info.entriesCount; entry_index++) 2791 { 2792 struct lldb_copy_dyld_cache_local_symbols_entry local_symbols_entry; 2793 local_symbols_entry.dylibOffset = dsc_local_symbols_data.GetU32(&offset); 2794 local_symbols_entry.nlistStartIndex = dsc_local_symbols_data.GetU32(&offset); 2795 local_symbols_entry.nlistCount = dsc_local_symbols_data.GetU32(&offset); 2796 2797 if (header_file_offset == local_symbols_entry.dylibOffset) 2798 { 2799 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2800 2801 // The normal nlist code cannot correctly size the Symbols array, we need to allocate it here. 2802 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms + unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2803 num_syms = symtab->GetNumSymbols(); 2804 2805 nlist_data_offset = local_symbols_info.nlistOffset + (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2806 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2807 2808 for (uint32_t nlist_index = 0; nlist_index < local_symbols_entry.nlistCount; nlist_index++) 2809 { 2810 ///////////////////////////// 2811 { 2812 struct nlist_64 nlist; 2813 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2814 break; 2815 2816 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked(&nlist_data_offset); 2817 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2818 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2819 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked (&nlist_data_offset); 2820 nlist.n_value = dsc_local_symbols_data.GetAddress_unchecked (&nlist_data_offset); 2821 2822 SymbolType type = eSymbolTypeInvalid; 2823 const char *symbol_name = dsc_local_symbols_data.PeekCStr(string_table_offset + nlist.n_strx); 2824 2825 if (symbol_name == NULL) 2826 { 2827 // No symbol should be NULL, even the symbols with no 2828 // string values should have an offset zero which points 2829 // to an empty C-string 2830 Host::SystemLog (Host::eSystemLogError, 2831 "error: DSC unmapped local symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 2832 entry_index, 2833 nlist.n_strx, 2834 module_sp->GetFileSpec().GetPath().c_str()); 2835 continue; 2836 } 2837 if (symbol_name[0] == '\0') 2838 symbol_name = NULL; 2839 2840 const char *symbol_name_non_abi_mangled = NULL; 2841 2842 SectionSP symbol_section; 2843 uint32_t symbol_byte_size = 0; 2844 bool add_nlist = true; 2845 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2846 bool demangled_is_synthesized = false; 2847 bool is_gsym = false; 2848 bool set_value = true; 2849 2850 assert (sym_idx < num_syms); 2851 2852 sym[sym_idx].SetDebug (is_debug); 2853 2854 if (is_debug) 2855 { 2856 switch (nlist.n_type) 2857 { 2858 case N_GSYM: 2859 // global symbol: name,,NO_SECT,type,0 2860 // Sometimes the N_GSYM value contains the address. 2861 2862 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 2863 // have the same address, but we want to ensure that we always find only the real symbol, 2864 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 2865 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 2866 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 2867 // same address. 2868 2869 is_gsym = true; 2870 sym[sym_idx].SetExternal(true); 2871 2872 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') 2873 { 2874 llvm::StringRef symbol_name_ref(symbol_name); 2875 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 2876 { 2877 symbol_name_non_abi_mangled = symbol_name + 1; 2878 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2879 type = eSymbolTypeObjCClass; 2880 demangled_is_synthesized = true; 2881 2882 } 2883 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 2884 { 2885 symbol_name_non_abi_mangled = symbol_name + 1; 2886 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2887 type = eSymbolTypeObjCMetaClass; 2888 demangled_is_synthesized = true; 2889 } 2890 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 2891 { 2892 symbol_name_non_abi_mangled = symbol_name + 1; 2893 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2894 type = eSymbolTypeObjCIVar; 2895 demangled_is_synthesized = true; 2896 } 2897 } 2898 else 2899 { 2900 if (nlist.n_value != 0) 2901 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2902 type = eSymbolTypeData; 2903 } 2904 break; 2905 2906 case N_FNAME: 2907 // procedure name (f77 kludge): name,,NO_SECT,0,0 2908 type = eSymbolTypeCompiler; 2909 break; 2910 2911 case N_FUN: 2912 // procedure: name,,n_sect,linenumber,address 2913 if (symbol_name) 2914 { 2915 type = eSymbolTypeCode; 2916 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2917 2918 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 2919 // We use the current number of symbols in the symbol table in lieu of 2920 // using nlist_idx in case we ever start trimming entries out 2921 N_FUN_indexes.push_back(sym_idx); 2922 } 2923 else 2924 { 2925 type = eSymbolTypeCompiler; 2926 2927 if ( !N_FUN_indexes.empty() ) 2928 { 2929 // Copy the size of the function into the original STAB entry so we don't have 2930 // to hunt for it later 2931 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 2932 N_FUN_indexes.pop_back(); 2933 // We don't really need the end function STAB as it contains the size which 2934 // we already placed with the original symbol, so don't add it if we want a 2935 // minimal symbol table 2936 add_nlist = false; 2937 } 2938 } 2939 break; 2940 2941 case N_STSYM: 2942 // static symbol: name,,n_sect,type,address 2943 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 2944 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2945 type = eSymbolTypeData; 2946 break; 2947 2948 case N_LCSYM: 2949 // .lcomm symbol: name,,n_sect,type,address 2950 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2951 type = eSymbolTypeCommonBlock; 2952 break; 2953 2954 case N_BNSYM: 2955 // We use the current number of symbols in the symbol table in lieu of 2956 // using nlist_idx in case we ever start trimming entries out 2957 // Skip these if we want minimal symbol tables 2958 add_nlist = false; 2959 break; 2960 2961 case N_ENSYM: 2962 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 2963 // so that we can always skip the entire symbol if we need to navigate 2964 // more quickly at the source level when parsing STABS 2965 // Skip these if we want minimal symbol tables 2966 add_nlist = false; 2967 break; 2968 2969 2970 case N_OPT: 2971 // emitted with gcc2_compiled and in gcc source 2972 type = eSymbolTypeCompiler; 2973 break; 2974 2975 case N_RSYM: 2976 // register sym: name,,NO_SECT,type,register 2977 type = eSymbolTypeVariable; 2978 break; 2979 2980 case N_SLINE: 2981 // src line: 0,,n_sect,linenumber,address 2982 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2983 type = eSymbolTypeLineEntry; 2984 break; 2985 2986 case N_SSYM: 2987 // structure elt: name,,NO_SECT,type,struct_offset 2988 type = eSymbolTypeVariableType; 2989 break; 2990 2991 case N_SO: 2992 // source file name 2993 type = eSymbolTypeSourceFile; 2994 if (symbol_name == NULL) 2995 { 2996 add_nlist = false; 2997 if (N_SO_index != UINT32_MAX) 2998 { 2999 // Set the size of the N_SO to the terminating index of this N_SO 3000 // so that we can always skip the entire N_SO if we need to navigate 3001 // more quickly at the source level when parsing STABS 3002 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3003 symbol_ptr->SetByteSize(sym_idx); 3004 symbol_ptr->SetSizeIsSibling(true); 3005 } 3006 N_NSYM_indexes.clear(); 3007 N_INCL_indexes.clear(); 3008 N_BRAC_indexes.clear(); 3009 N_COMM_indexes.clear(); 3010 N_FUN_indexes.clear(); 3011 N_SO_index = UINT32_MAX; 3012 } 3013 else 3014 { 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 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3018 if (N_SO_has_full_path) 3019 { 3020 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3021 { 3022 // We have two consecutive N_SO entries where the first contains a directory 3023 // and the second contains a full path. 3024 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 3025 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3026 add_nlist = false; 3027 } 3028 else 3029 { 3030 // This is the first entry in a N_SO that contains a directory or 3031 // a full path to the source file 3032 N_SO_index = sym_idx; 3033 } 3034 } 3035 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3036 { 3037 // This is usually the second N_SO entry that contains just the filename, 3038 // so here we combine it with the first one if we are minimizing the symbol table 3039 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3040 if (so_path && so_path[0]) 3041 { 3042 std::string full_so_path (so_path); 3043 const size_t double_slash_pos = full_so_path.find("//"); 3044 if (double_slash_pos != std::string::npos) 3045 { 3046 // The linker has been generating bad N_SO entries with doubled up paths 3047 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 3048 // and the second is the directory for the source file so you end up with 3049 // a path that looks like "/tmp/src//tmp/src/" 3050 FileSpec so_dir(so_path, false); 3051 if (!so_dir.Exists()) 3052 { 3053 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 3054 if (so_dir.Exists()) 3055 { 3056 // Trim off the incorrect path 3057 full_so_path.erase(0, double_slash_pos + 1); 3058 } 3059 } 3060 } 3061 if (*full_so_path.rbegin() != '/') 3062 full_so_path += '/'; 3063 full_so_path += symbol_name; 3064 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 3065 add_nlist = false; 3066 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3067 } 3068 } 3069 else 3070 { 3071 // This could be a relative path to a N_SO 3072 N_SO_index = sym_idx; 3073 } 3074 } 3075 break; 3076 3077 case N_OSO: 3078 // object file name: name,,0,0,st_mtime 3079 type = eSymbolTypeObjectFile; 3080 break; 3081 3082 case N_LSYM: 3083 // local sym: name,,NO_SECT,type,offset 3084 type = eSymbolTypeLocal; 3085 break; 3086 3087 //---------------------------------------------------------------------- 3088 // INCL scopes 3089 //---------------------------------------------------------------------- 3090 case N_BINCL: 3091 // include file beginning: name,,NO_SECT,0,sum 3092 // We use the current number of symbols in the symbol table in lieu of 3093 // using nlist_idx in case we ever start trimming entries out 3094 N_INCL_indexes.push_back(sym_idx); 3095 type = eSymbolTypeScopeBegin; 3096 break; 3097 3098 case N_EINCL: 3099 // include file end: name,,NO_SECT,0,0 3100 // Set the size of the N_BINCL to the terminating index of this N_EINCL 3101 // so that we can always skip the entire symbol if we need to navigate 3102 // more quickly at the source level when parsing STABS 3103 if ( !N_INCL_indexes.empty() ) 3104 { 3105 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3106 symbol_ptr->SetByteSize(sym_idx + 1); 3107 symbol_ptr->SetSizeIsSibling(true); 3108 N_INCL_indexes.pop_back(); 3109 } 3110 type = eSymbolTypeScopeEnd; 3111 break; 3112 3113 case N_SOL: 3114 // #included file name: name,,n_sect,0,address 3115 type = eSymbolTypeHeaderFile; 3116 3117 // We currently don't use the header files on darwin 3118 add_nlist = false; 3119 break; 3120 3121 case N_PARAMS: 3122 // compiler parameters: name,,NO_SECT,0,0 3123 type = eSymbolTypeCompiler; 3124 break; 3125 3126 case N_VERSION: 3127 // compiler version: name,,NO_SECT,0,0 3128 type = eSymbolTypeCompiler; 3129 break; 3130 3131 case N_OLEVEL: 3132 // compiler -O level: name,,NO_SECT,0,0 3133 type = eSymbolTypeCompiler; 3134 break; 3135 3136 case N_PSYM: 3137 // parameter: name,,NO_SECT,type,offset 3138 type = eSymbolTypeVariable; 3139 break; 3140 3141 case N_ENTRY: 3142 // alternate entry: name,,n_sect,linenumber,address 3143 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3144 type = eSymbolTypeLineEntry; 3145 break; 3146 3147 //---------------------------------------------------------------------- 3148 // Left and Right Braces 3149 //---------------------------------------------------------------------- 3150 case N_LBRAC: 3151 // left bracket: 0,,NO_SECT,nesting level,address 3152 // We use the current number of symbols in the symbol table in lieu of 3153 // using nlist_idx in case we ever start trimming entries out 3154 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3155 N_BRAC_indexes.push_back(sym_idx); 3156 type = eSymbolTypeScopeBegin; 3157 break; 3158 3159 case N_RBRAC: 3160 // right bracket: 0,,NO_SECT,nesting level,address 3161 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 3162 // so that we can always skip the entire symbol if we need to navigate 3163 // more quickly at the source level when parsing STABS 3164 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3165 if ( !N_BRAC_indexes.empty() ) 3166 { 3167 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3168 symbol_ptr->SetByteSize(sym_idx + 1); 3169 symbol_ptr->SetSizeIsSibling(true); 3170 N_BRAC_indexes.pop_back(); 3171 } 3172 type = eSymbolTypeScopeEnd; 3173 break; 3174 3175 case N_EXCL: 3176 // deleted include file: name,,NO_SECT,0,sum 3177 type = eSymbolTypeHeaderFile; 3178 break; 3179 3180 //---------------------------------------------------------------------- 3181 // COMM scopes 3182 //---------------------------------------------------------------------- 3183 case N_BCOMM: 3184 // begin common: name,,NO_SECT,0,0 3185 // We use the current number of symbols in the symbol table in lieu of 3186 // using nlist_idx in case we ever start trimming entries out 3187 type = eSymbolTypeScopeBegin; 3188 N_COMM_indexes.push_back(sym_idx); 3189 break; 3190 3191 case N_ECOML: 3192 // end common (local name): 0,,n_sect,0,address 3193 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3194 // Fall through 3195 3196 case N_ECOMM: 3197 // end common: name,,n_sect,0,0 3198 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 3199 // so that we can always skip the entire symbol if we need to navigate 3200 // more quickly at the source level when parsing STABS 3201 if ( !N_COMM_indexes.empty() ) 3202 { 3203 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 3204 symbol_ptr->SetByteSize(sym_idx + 1); 3205 symbol_ptr->SetSizeIsSibling(true); 3206 N_COMM_indexes.pop_back(); 3207 } 3208 type = eSymbolTypeScopeEnd; 3209 break; 3210 3211 case N_LENG: 3212 // second stab entry with length information 3213 type = eSymbolTypeAdditional; 3214 break; 3215 3216 default: break; 3217 } 3218 } 3219 else 3220 { 3221 //uint8_t n_pext = N_PEXT & nlist.n_type; 3222 uint8_t n_type = N_TYPE & nlist.n_type; 3223 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3224 3225 switch (n_type) 3226 { 3227 case N_INDR: 3228 { 3229 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 3230 if (reexport_name_cstr && reexport_name_cstr[0]) 3231 { 3232 type = eSymbolTypeReExported; 3233 ConstString reexport_name(reexport_name_cstr + ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3234 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3235 set_value = false; 3236 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3237 indirect_symbol_names.insert(ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3238 } 3239 else 3240 type = eSymbolTypeUndefined; 3241 } 3242 break; 3243 3244 case N_UNDF: 3245 if (symbol_name && symbol_name[0]) 3246 { 3247 ConstString undefined_name(symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3248 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3249 } 3250 // Fall through 3251 case N_PBUD: 3252 type = eSymbolTypeUndefined; 3253 break; 3254 3255 case N_ABS: 3256 type = eSymbolTypeAbsolute; 3257 break; 3258 3259 case N_SECT: 3260 { 3261 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3262 3263 if (symbol_section == NULL) 3264 { 3265 // TODO: warn about this? 3266 add_nlist = false; 3267 break; 3268 } 3269 3270 if (TEXT_eh_frame_sectID == nlist.n_sect) 3271 { 3272 type = eSymbolTypeException; 3273 } 3274 else 3275 { 3276 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 3277 3278 switch (section_type) 3279 { 3280 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 3281 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 3282 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 3283 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 3284 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 3285 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 3286 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 3287 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 3288 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 3289 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 3290 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 3291 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 3292 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 3293 default: 3294 switch (symbol_section->GetType()) 3295 { 3296 case lldb::eSectionTypeCode: 3297 type = eSymbolTypeCode; 3298 break; 3299 case eSectionTypeData: 3300 case eSectionTypeDataCString: // Inlined C string data 3301 case eSectionTypeDataCStringPointers: // Pointers to C string data 3302 case eSectionTypeDataSymbolAddress: // Address of a symbol in the symbol table 3303 case eSectionTypeData4: 3304 case eSectionTypeData8: 3305 case eSectionTypeData16: 3306 type = eSymbolTypeData; 3307 break; 3308 default: 3309 break; 3310 } 3311 break; 3312 } 3313 3314 if (type == eSymbolTypeInvalid) 3315 { 3316 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 3317 if (symbol_section->IsDescendant (text_section_sp.get())) 3318 { 3319 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 3320 S_ATTR_SELF_MODIFYING_CODE | 3321 S_ATTR_SOME_INSTRUCTIONS)) 3322 type = eSymbolTypeData; 3323 else 3324 type = eSymbolTypeCode; 3325 } 3326 else if (symbol_section->IsDescendant(data_section_sp.get()) || 3327 symbol_section->IsDescendant(data_dirty_section_sp.get()) || 3328 symbol_section->IsDescendant(data_const_section_sp.get())) 3329 { 3330 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 3331 { 3332 type = eSymbolTypeRuntime; 3333 3334 if (symbol_name && 3335 symbol_name[0] == '_' && 3336 symbol_name[1] == 'O' && 3337 symbol_name[2] == 'B') 3338 { 3339 llvm::StringRef symbol_name_ref(symbol_name); 3340 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 3341 { 3342 symbol_name_non_abi_mangled = symbol_name + 1; 3343 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3344 type = eSymbolTypeObjCClass; 3345 demangled_is_synthesized = true; 3346 } 3347 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 3348 { 3349 symbol_name_non_abi_mangled = symbol_name + 1; 3350 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3351 type = eSymbolTypeObjCMetaClass; 3352 demangled_is_synthesized = true; 3353 } 3354 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 3355 { 3356 symbol_name_non_abi_mangled = symbol_name + 1; 3357 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3358 type = eSymbolTypeObjCIVar; 3359 demangled_is_synthesized = true; 3360 } 3361 } 3362 } 3363 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 3364 { 3365 type = eSymbolTypeException; 3366 } 3367 else 3368 { 3369 type = eSymbolTypeData; 3370 } 3371 } 3372 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 3373 { 3374 type = eSymbolTypeTrampoline; 3375 } 3376 else if (symbol_section->IsDescendant(objc_section_sp.get())) 3377 { 3378 type = eSymbolTypeRuntime; 3379 if (symbol_name && symbol_name[0] == '.') 3380 { 3381 llvm::StringRef symbol_name_ref(symbol_name); 3382 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 3383 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 3384 { 3385 symbol_name_non_abi_mangled = symbol_name; 3386 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 3387 type = eSymbolTypeObjCClass; 3388 demangled_is_synthesized = true; 3389 } 3390 } 3391 } 3392 } 3393 } 3394 } 3395 break; 3396 } 3397 } 3398 3399 if (add_nlist) 3400 { 3401 uint64_t symbol_value = nlist.n_value; 3402 if (symbol_name_non_abi_mangled) 3403 { 3404 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 3405 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 3406 } 3407 else 3408 { 3409 bool symbol_name_is_mangled = false; 3410 3411 if (symbol_name && symbol_name[0] == '_') 3412 { 3413 symbol_name_is_mangled = symbol_name[1] == '_'; 3414 symbol_name++; // Skip the leading underscore 3415 } 3416 3417 if (symbol_name) 3418 { 3419 ConstString const_symbol_name(symbol_name); 3420 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 3421 if (is_gsym && is_debug) 3422 { 3423 const char *gsym_name = sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString(); 3424 if (gsym_name) 3425 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3426 } 3427 } 3428 } 3429 if (symbol_section) 3430 { 3431 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3432 if (symbol_byte_size == 0 && function_starts_count > 0) 3433 { 3434 addr_t symbol_lookup_file_addr = nlist.n_value; 3435 // Do an exact address match for non-ARM addresses, else get the closest since 3436 // the symbol might be a thumb symbol which has an address with bit zero set 3437 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 3438 if (is_arm && func_start_entry) 3439 { 3440 // Verify that the function start address is the symbol address (ARM) 3441 // or the symbol address + 1 (thumb) 3442 if (func_start_entry->addr != symbol_lookup_file_addr && 3443 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 3444 { 3445 // Not the right entry, NULL it out... 3446 func_start_entry = NULL; 3447 } 3448 } 3449 if (func_start_entry) 3450 { 3451 func_start_entry->data = true; 3452 3453 addr_t symbol_file_addr = func_start_entry->addr; 3454 uint32_t symbol_flags = 0; 3455 if (is_arm) 3456 { 3457 if (symbol_file_addr & 1) 3458 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3459 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3460 } 3461 3462 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3463 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3464 if (next_func_start_entry) 3465 { 3466 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3467 // Be sure the clear the Thumb address bit when we calculate the size 3468 // from the current and next address 3469 if (is_arm) 3470 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3471 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3472 } 3473 else 3474 { 3475 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3476 } 3477 } 3478 } 3479 symbol_value -= section_file_addr; 3480 } 3481 3482 if (is_debug == false) 3483 { 3484 if (type == eSymbolTypeCode) 3485 { 3486 // See if we can find a N_FUN entry for any code symbols. 3487 // If we do find a match, and the name matches, then we 3488 // can merge the two into just the function symbol to avoid 3489 // duplicate entries in the symbol table 3490 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3491 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3492 if (range.first != range.second) 3493 { 3494 bool found_it = false; 3495 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3496 { 3497 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3498 { 3499 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3500 // We just need the flags from the linker symbol, so put these flags 3501 // into the N_FUN flags to avoid duplicate symbols in the symbol table 3502 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3503 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3504 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3505 sym[pos->second].SetType (eSymbolTypeResolver); 3506 sym[sym_idx].Clear(); 3507 found_it = true; 3508 break; 3509 } 3510 } 3511 if (found_it) 3512 continue; 3513 } 3514 else 3515 { 3516 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3517 type = eSymbolTypeResolver; 3518 } 3519 } 3520 else if (type == eSymbolTypeData || 3521 type == eSymbolTypeObjCClass || 3522 type == eSymbolTypeObjCMetaClass || 3523 type == eSymbolTypeObjCIVar ) 3524 { 3525 // See if we can find a N_STSYM entry for any data symbols. 3526 // If we do find a match, and the name matches, then we 3527 // can merge the two into just the Static symbol to avoid 3528 // duplicate entries in the symbol table 3529 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3530 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 3531 if (range.first != range.second) 3532 { 3533 bool found_it = false; 3534 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3535 { 3536 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3537 { 3538 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3539 // We just need the flags from the linker symbol, so put these flags 3540 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3541 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3542 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3543 sym[sym_idx].Clear(); 3544 found_it = true; 3545 break; 3546 } 3547 } 3548 if (found_it) 3549 continue; 3550 } 3551 else 3552 { 3553 const char *gsym_name = sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString(); 3554 if (gsym_name) 3555 { 3556 // Combine N_GSYM stab entries with the non stab symbol 3557 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(gsym_name); 3558 if (pos != N_GSYM_name_to_sym_idx.end()) 3559 { 3560 const uint32_t GSYM_sym_idx = pos->second; 3561 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 3562 // Copy the address, because often the N_GSYM address has an invalid address of zero 3563 // when the global is a common symbol 3564 sym[GSYM_sym_idx].GetAddressRef().SetSection (symbol_section); 3565 sym[GSYM_sym_idx].GetAddressRef().SetOffset (symbol_value); 3566 // We just need the flags from the linker symbol, so put these flags 3567 // into the N_GSYM flags to avoid duplicate symbols in the symbol table 3568 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3569 sym[sym_idx].Clear(); 3570 continue; 3571 } 3572 } 3573 } 3574 } 3575 } 3576 3577 sym[sym_idx].SetID (nlist_idx); 3578 sym[sym_idx].SetType (type); 3579 if (set_value) 3580 { 3581 sym[sym_idx].GetAddressRef().SetSection (symbol_section); 3582 sym[sym_idx].GetAddressRef().SetOffset (symbol_value); 3583 } 3584 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3585 3586 if (symbol_byte_size > 0) 3587 sym[sym_idx].SetByteSize(symbol_byte_size); 3588 3589 if (demangled_is_synthesized) 3590 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3591 ++sym_idx; 3592 } 3593 else 3594 { 3595 sym[sym_idx].Clear(); 3596 } 3597 3598 } 3599 ///////////////////////////// 3600 } 3601 break; // No more entries to consider 3602 } 3603 } 3604 3605 for (const auto &pos :reexport_shlib_needs_fixup) 3606 { 3607 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3608 if (undef_pos != undefined_name_to_desc.end()) 3609 { 3610 const uint8_t dylib_ordinal = llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3611 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3612 sym[pos.first].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(dylib_ordinal-1)); 3613 } 3614 } 3615 } 3616 } 3617 } 3618 } 3619 } 3620 3621 // Must reset this in case it was mutated above! 3622 nlist_data_offset = 0; 3623 #endif 3624 3625 if (nlist_data.GetByteSize() > 0) 3626 { 3627 3628 // If the sym array was not created while parsing the DSC unmapped 3629 // symbols, create it now. 3630 if (sym == NULL) 3631 { 3632 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3633 num_syms = symtab->GetNumSymbols(); 3634 } 3635 3636 if (unmapped_local_symbols_found) 3637 { 3638 assert(m_dysymtab.ilocalsym == 0); 3639 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3640 nlist_idx = m_dysymtab.nlocalsym; 3641 } 3642 else 3643 { 3644 nlist_idx = 0; 3645 } 3646 3647 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3648 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3649 UndefinedNameToDescMap undefined_name_to_desc; 3650 SymbolIndexToName reexport_shlib_needs_fixup; 3651 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 3652 { 3653 struct nlist_64 nlist; 3654 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 3655 break; 3656 3657 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3658 nlist.n_type = nlist_data.GetU8_unchecked (&nlist_data_offset); 3659 nlist.n_sect = nlist_data.GetU8_unchecked (&nlist_data_offset); 3660 nlist.n_desc = nlist_data.GetU16_unchecked (&nlist_data_offset); 3661 nlist.n_value = nlist_data.GetAddress_unchecked (&nlist_data_offset); 3662 3663 SymbolType type = eSymbolTypeInvalid; 3664 const char *symbol_name = NULL; 3665 3666 if (have_strtab_data) 3667 { 3668 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3669 3670 if (symbol_name == NULL) 3671 { 3672 // No symbol should be NULL, even the symbols with no 3673 // string values should have an offset zero which points 3674 // to an empty C-string 3675 Host::SystemLog (Host::eSystemLogError, 3676 "error: symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 3677 nlist_idx, 3678 nlist.n_strx, 3679 module_sp->GetFileSpec().GetPath().c_str()); 3680 continue; 3681 } 3682 if (symbol_name[0] == '\0') 3683 symbol_name = NULL; 3684 } 3685 else 3686 { 3687 const addr_t str_addr = strtab_addr + nlist.n_strx; 3688 Error str_error; 3689 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, str_error)) 3690 symbol_name = memory_symbol_name.c_str(); 3691 } 3692 const char *symbol_name_non_abi_mangled = NULL; 3693 3694 SectionSP symbol_section; 3695 lldb::addr_t symbol_byte_size = 0; 3696 bool add_nlist = true; 3697 bool is_gsym = false; 3698 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3699 bool demangled_is_synthesized = false; 3700 bool set_value = true; 3701 assert (sym_idx < num_syms); 3702 3703 sym[sym_idx].SetDebug (is_debug); 3704 3705 if (is_debug) 3706 { 3707 switch (nlist.n_type) 3708 { 3709 case N_GSYM: 3710 // global symbol: name,,NO_SECT,type,0 3711 // Sometimes the N_GSYM value contains the address. 3712 3713 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 3714 // have the same address, but we want to ensure that we always find only the real symbol, 3715 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 3716 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 3717 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 3718 // same address. 3719 is_gsym = true; 3720 sym[sym_idx].SetExternal(true); 3721 3722 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') 3723 { 3724 llvm::StringRef symbol_name_ref(symbol_name); 3725 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 3726 { 3727 symbol_name_non_abi_mangled = symbol_name + 1; 3728 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3729 type = eSymbolTypeObjCClass; 3730 demangled_is_synthesized = true; 3731 3732 } 3733 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 3734 { 3735 symbol_name_non_abi_mangled = symbol_name + 1; 3736 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3737 type = eSymbolTypeObjCMetaClass; 3738 demangled_is_synthesized = true; 3739 } 3740 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 3741 { 3742 symbol_name_non_abi_mangled = symbol_name + 1; 3743 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3744 type = eSymbolTypeObjCIVar; 3745 demangled_is_synthesized = true; 3746 } 3747 } 3748 else 3749 { 3750 if (nlist.n_value != 0) 3751 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3752 type = eSymbolTypeData; 3753 } 3754 break; 3755 3756 case N_FNAME: 3757 // procedure name (f77 kludge): name,,NO_SECT,0,0 3758 type = eSymbolTypeCompiler; 3759 break; 3760 3761 case N_FUN: 3762 // procedure: name,,n_sect,linenumber,address 3763 if (symbol_name) 3764 { 3765 type = eSymbolTypeCode; 3766 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3767 3768 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3769 // We use the current number of symbols in the symbol table in lieu of 3770 // using nlist_idx in case we ever start trimming entries out 3771 N_FUN_indexes.push_back(sym_idx); 3772 } 3773 else 3774 { 3775 type = eSymbolTypeCompiler; 3776 3777 if ( !N_FUN_indexes.empty() ) 3778 { 3779 // Copy the size of the function into the original STAB entry so we don't have 3780 // to hunt for it later 3781 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 3782 N_FUN_indexes.pop_back(); 3783 // We don't really need the end function STAB as it contains the size which 3784 // we already placed with the original symbol, so don't add it if we want a 3785 // minimal symbol table 3786 add_nlist = false; 3787 } 3788 } 3789 break; 3790 3791 case N_STSYM: 3792 // static symbol: name,,n_sect,type,address 3793 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3794 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3795 type = eSymbolTypeData; 3796 break; 3797 3798 case N_LCSYM: 3799 // .lcomm symbol: name,,n_sect,type,address 3800 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3801 type = eSymbolTypeCommonBlock; 3802 break; 3803 3804 case N_BNSYM: 3805 // We use the current number of symbols in the symbol table in lieu of 3806 // using nlist_idx in case we ever start trimming entries out 3807 // Skip these if we want minimal symbol tables 3808 add_nlist = false; 3809 break; 3810 3811 case N_ENSYM: 3812 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 3813 // so that we can always skip the entire symbol if we need to navigate 3814 // more quickly at the source level when parsing STABS 3815 // Skip these if we want minimal symbol tables 3816 add_nlist = false; 3817 break; 3818 3819 3820 case N_OPT: 3821 // emitted with gcc2_compiled and in gcc source 3822 type = eSymbolTypeCompiler; 3823 break; 3824 3825 case N_RSYM: 3826 // register sym: name,,NO_SECT,type,register 3827 type = eSymbolTypeVariable; 3828 break; 3829 3830 case N_SLINE: 3831 // src line: 0,,n_sect,linenumber,address 3832 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3833 type = eSymbolTypeLineEntry; 3834 break; 3835 3836 case N_SSYM: 3837 // structure elt: name,,NO_SECT,type,struct_offset 3838 type = eSymbolTypeVariableType; 3839 break; 3840 3841 case N_SO: 3842 // source file name 3843 type = eSymbolTypeSourceFile; 3844 if (symbol_name == NULL) 3845 { 3846 add_nlist = false; 3847 if (N_SO_index != UINT32_MAX) 3848 { 3849 // Set the size of the N_SO to the terminating index of this N_SO 3850 // so that we can always skip the entire N_SO if we need to navigate 3851 // more quickly at the source level when parsing STABS 3852 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3853 symbol_ptr->SetByteSize(sym_idx); 3854 symbol_ptr->SetSizeIsSibling(true); 3855 } 3856 N_NSYM_indexes.clear(); 3857 N_INCL_indexes.clear(); 3858 N_BRAC_indexes.clear(); 3859 N_COMM_indexes.clear(); 3860 N_FUN_indexes.clear(); 3861 N_SO_index = UINT32_MAX; 3862 } 3863 else 3864 { 3865 // We use the current number of symbols in the symbol table in lieu of 3866 // using nlist_idx in case we ever start trimming entries out 3867 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3868 if (N_SO_has_full_path) 3869 { 3870 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3871 { 3872 // We have two consecutive N_SO entries where the first contains a directory 3873 // and the second contains a full path. 3874 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 3875 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3876 add_nlist = false; 3877 } 3878 else 3879 { 3880 // This is the first entry in a N_SO that contains a directory or 3881 // a full path to the source file 3882 N_SO_index = sym_idx; 3883 } 3884 } 3885 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3886 { 3887 // This is usually the second N_SO entry that contains just the filename, 3888 // so here we combine it with the first one if we are minimizing the symbol table 3889 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName(lldb::eLanguageTypeUnknown).AsCString(); 3890 if (so_path && so_path[0]) 3891 { 3892 std::string full_so_path (so_path); 3893 const size_t double_slash_pos = full_so_path.find("//"); 3894 if (double_slash_pos != std::string::npos) 3895 { 3896 // The linker has been generating bad N_SO entries with doubled up paths 3897 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 3898 // and the second is the directory for the source file so you end up with 3899 // a path that looks like "/tmp/src//tmp/src/" 3900 FileSpec so_dir(so_path, false); 3901 if (!so_dir.Exists()) 3902 { 3903 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 3904 if (so_dir.Exists()) 3905 { 3906 // Trim off the incorrect path 3907 full_so_path.erase(0, double_slash_pos + 1); 3908 } 3909 } 3910 } 3911 if (*full_so_path.rbegin() != '/') 3912 full_so_path += '/'; 3913 full_so_path += symbol_name; 3914 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 3915 add_nlist = false; 3916 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3917 } 3918 } 3919 else 3920 { 3921 // This could be a relative path to a N_SO 3922 N_SO_index = sym_idx; 3923 } 3924 } 3925 3926 break; 3927 3928 case N_OSO: 3929 // object file name: name,,0,0,st_mtime 3930 type = eSymbolTypeObjectFile; 3931 break; 3932 3933 case N_LSYM: 3934 // local sym: name,,NO_SECT,type,offset 3935 type = eSymbolTypeLocal; 3936 break; 3937 3938 //---------------------------------------------------------------------- 3939 // INCL scopes 3940 //---------------------------------------------------------------------- 3941 case N_BINCL: 3942 // include file beginning: name,,NO_SECT,0,sum 3943 // We use the current number of symbols in the symbol table in lieu of 3944 // using nlist_idx in case we ever start trimming entries out 3945 N_INCL_indexes.push_back(sym_idx); 3946 type = eSymbolTypeScopeBegin; 3947 break; 3948 3949 case N_EINCL: 3950 // include file end: name,,NO_SECT,0,0 3951 // Set the size of the N_BINCL to the terminating index of this N_EINCL 3952 // so that we can always skip the entire symbol if we need to navigate 3953 // more quickly at the source level when parsing STABS 3954 if ( !N_INCL_indexes.empty() ) 3955 { 3956 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3957 symbol_ptr->SetByteSize(sym_idx + 1); 3958 symbol_ptr->SetSizeIsSibling(true); 3959 N_INCL_indexes.pop_back(); 3960 } 3961 type = eSymbolTypeScopeEnd; 3962 break; 3963 3964 case N_SOL: 3965 // #included file name: name,,n_sect,0,address 3966 type = eSymbolTypeHeaderFile; 3967 3968 // We currently don't use the header files on darwin 3969 add_nlist = false; 3970 break; 3971 3972 case N_PARAMS: 3973 // compiler parameters: name,,NO_SECT,0,0 3974 type = eSymbolTypeCompiler; 3975 break; 3976 3977 case N_VERSION: 3978 // compiler version: name,,NO_SECT,0,0 3979 type = eSymbolTypeCompiler; 3980 break; 3981 3982 case N_OLEVEL: 3983 // compiler -O level: name,,NO_SECT,0,0 3984 type = eSymbolTypeCompiler; 3985 break; 3986 3987 case N_PSYM: 3988 // parameter: name,,NO_SECT,type,offset 3989 type = eSymbolTypeVariable; 3990 break; 3991 3992 case N_ENTRY: 3993 // alternate entry: name,,n_sect,linenumber,address 3994 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3995 type = eSymbolTypeLineEntry; 3996 break; 3997 3998 //---------------------------------------------------------------------- 3999 // Left and Right Braces 4000 //---------------------------------------------------------------------- 4001 case N_LBRAC: 4002 // left bracket: 0,,NO_SECT,nesting level,address 4003 // We use the current number of symbols in the symbol table in lieu of 4004 // using nlist_idx in case we ever start trimming entries out 4005 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 4006 N_BRAC_indexes.push_back(sym_idx); 4007 type = eSymbolTypeScopeBegin; 4008 break; 4009 4010 case N_RBRAC: 4011 // right bracket: 0,,NO_SECT,nesting level,address 4012 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 4013 // so that we can always skip the entire symbol if we need to navigate 4014 // more quickly at the source level when parsing STABS 4015 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 4016 if ( !N_BRAC_indexes.empty() ) 4017 { 4018 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4019 symbol_ptr->SetByteSize(sym_idx + 1); 4020 symbol_ptr->SetSizeIsSibling(true); 4021 N_BRAC_indexes.pop_back(); 4022 } 4023 type = eSymbolTypeScopeEnd; 4024 break; 4025 4026 case N_EXCL: 4027 // deleted include file: name,,NO_SECT,0,sum 4028 type = eSymbolTypeHeaderFile; 4029 break; 4030 4031 //---------------------------------------------------------------------- 4032 // COMM scopes 4033 //---------------------------------------------------------------------- 4034 case N_BCOMM: 4035 // begin common: name,,NO_SECT,0,0 4036 // We use the current number of symbols in the symbol table in lieu of 4037 // using nlist_idx in case we ever start trimming entries out 4038 type = eSymbolTypeScopeBegin; 4039 N_COMM_indexes.push_back(sym_idx); 4040 break; 4041 4042 case N_ECOML: 4043 // end common (local name): 0,,n_sect,0,address 4044 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 4045 // Fall through 4046 4047 case N_ECOMM: 4048 // end common: name,,n_sect,0,0 4049 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 4050 // so that we can always skip the entire symbol if we need to navigate 4051 // more quickly at the source level when parsing STABS 4052 if ( !N_COMM_indexes.empty() ) 4053 { 4054 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4055 symbol_ptr->SetByteSize(sym_idx + 1); 4056 symbol_ptr->SetSizeIsSibling(true); 4057 N_COMM_indexes.pop_back(); 4058 } 4059 type = eSymbolTypeScopeEnd; 4060 break; 4061 4062 case N_LENG: 4063 // second stab entry with length information 4064 type = eSymbolTypeAdditional; 4065 break; 4066 4067 default: break; 4068 } 4069 } 4070 else 4071 { 4072 //uint8_t n_pext = N_PEXT & nlist.n_type; 4073 uint8_t n_type = N_TYPE & nlist.n_type; 4074 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4075 4076 switch (n_type) 4077 { 4078 case N_INDR: 4079 { 4080 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4081 if (reexport_name_cstr && reexport_name_cstr[0]) 4082 { 4083 type = eSymbolTypeReExported; 4084 ConstString reexport_name(reexport_name_cstr + ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4085 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4086 set_value = false; 4087 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4088 indirect_symbol_names.insert(ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4089 } 4090 else 4091 type = eSymbolTypeUndefined; 4092 } 4093 break; 4094 4095 case N_UNDF: 4096 if (symbol_name && symbol_name[0]) 4097 { 4098 ConstString undefined_name(symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 4099 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4100 } 4101 // Fall through 4102 case N_PBUD: 4103 type = eSymbolTypeUndefined; 4104 break; 4105 4106 case N_ABS: 4107 type = eSymbolTypeAbsolute; 4108 break; 4109 4110 case N_SECT: 4111 { 4112 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 4113 4114 if (!symbol_section) 4115 { 4116 // TODO: warn about this? 4117 add_nlist = false; 4118 break; 4119 } 4120 4121 if (TEXT_eh_frame_sectID == nlist.n_sect) 4122 { 4123 type = eSymbolTypeException; 4124 } 4125 else 4126 { 4127 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4128 4129 switch (section_type) 4130 { 4131 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 4132 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 4133 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 4134 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 4135 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 4136 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 4137 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 4138 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 4139 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 4140 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 4141 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 4142 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 4143 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 4144 default: 4145 switch (symbol_section->GetType()) 4146 { 4147 case lldb::eSectionTypeCode: 4148 type = eSymbolTypeCode; 4149 break; 4150 case eSectionTypeData: 4151 case eSectionTypeDataCString: // Inlined C string data 4152 case eSectionTypeDataCStringPointers: // Pointers to C string data 4153 case eSectionTypeDataSymbolAddress: // Address of a symbol in the symbol table 4154 case eSectionTypeData4: 4155 case eSectionTypeData8: 4156 case eSectionTypeData16: 4157 type = eSymbolTypeData; 4158 break; 4159 default: 4160 break; 4161 } 4162 break; 4163 } 4164 4165 if (type == eSymbolTypeInvalid) 4166 { 4167 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 4168 if (symbol_section->IsDescendant (text_section_sp.get())) 4169 { 4170 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4171 S_ATTR_SELF_MODIFYING_CODE | 4172 S_ATTR_SOME_INSTRUCTIONS)) 4173 type = eSymbolTypeData; 4174 else 4175 type = eSymbolTypeCode; 4176 } 4177 else 4178 if (symbol_section->IsDescendant(data_section_sp.get()) || 4179 symbol_section->IsDescendant(data_dirty_section_sp.get()) || 4180 symbol_section->IsDescendant(data_const_section_sp.get())) 4181 { 4182 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 4183 { 4184 type = eSymbolTypeRuntime; 4185 4186 if (symbol_name && 4187 symbol_name[0] == '_' && 4188 symbol_name[1] == 'O' && 4189 symbol_name[2] == 'B') 4190 { 4191 llvm::StringRef symbol_name_ref(symbol_name); 4192 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 4193 { 4194 symbol_name_non_abi_mangled = symbol_name + 1; 4195 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 4196 type = eSymbolTypeObjCClass; 4197 demangled_is_synthesized = true; 4198 } 4199 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 4200 { 4201 symbol_name_non_abi_mangled = symbol_name + 1; 4202 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 4203 type = eSymbolTypeObjCMetaClass; 4204 demangled_is_synthesized = true; 4205 } 4206 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 4207 { 4208 symbol_name_non_abi_mangled = symbol_name + 1; 4209 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 4210 type = eSymbolTypeObjCIVar; 4211 demangled_is_synthesized = true; 4212 } 4213 } 4214 } 4215 else 4216 if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 4217 { 4218 type = eSymbolTypeException; 4219 } 4220 else 4221 { 4222 type = eSymbolTypeData; 4223 } 4224 } 4225 else 4226 if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 4227 { 4228 type = eSymbolTypeTrampoline; 4229 } 4230 else 4231 if (symbol_section->IsDescendant(objc_section_sp.get())) 4232 { 4233 type = eSymbolTypeRuntime; 4234 if (symbol_name && symbol_name[0] == '.') 4235 { 4236 llvm::StringRef symbol_name_ref(symbol_name); 4237 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 4238 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 4239 { 4240 symbol_name_non_abi_mangled = symbol_name; 4241 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4242 type = eSymbolTypeObjCClass; 4243 demangled_is_synthesized = true; 4244 } 4245 } 4246 } 4247 } 4248 } 4249 } 4250 break; 4251 } 4252 } 4253 4254 if (add_nlist) 4255 { 4256 uint64_t symbol_value = nlist.n_value; 4257 4258 if (symbol_name_non_abi_mangled) 4259 { 4260 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 4261 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 4262 } 4263 else 4264 { 4265 bool symbol_name_is_mangled = false; 4266 4267 if (symbol_name && symbol_name[0] == '_') 4268 { 4269 symbol_name_is_mangled = symbol_name[1] == '_'; 4270 symbol_name++; // Skip the leading underscore 4271 } 4272 4273 if (symbol_name) 4274 { 4275 ConstString const_symbol_name(symbol_name); 4276 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 4277 } 4278 } 4279 4280 if (is_gsym) 4281 { 4282 const char *gsym_name = sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled).GetCString(); 4283 if (gsym_name) 4284 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4285 } 4286 4287 if (symbol_section) 4288 { 4289 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4290 if (symbol_byte_size == 0 && function_starts_count > 0) 4291 { 4292 addr_t symbol_lookup_file_addr = nlist.n_value; 4293 // Do an exact address match for non-ARM addresses, else get the closest since 4294 // the symbol might be a thumb symbol which has an address with bit zero set 4295 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 4296 if (is_arm && func_start_entry) 4297 { 4298 // Verify that the function start address is the symbol address (ARM) 4299 // or the symbol address + 1 (thumb) 4300 if (func_start_entry->addr != symbol_lookup_file_addr && 4301 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 4302 { 4303 // Not the right entry, NULL it out... 4304 func_start_entry = NULL; 4305 } 4306 } 4307 if (func_start_entry) 4308 { 4309 func_start_entry->data = true; 4310 4311 addr_t symbol_file_addr = func_start_entry->addr; 4312 if (is_arm) 4313 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4314 4315 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 4316 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 4317 if (next_func_start_entry) 4318 { 4319 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4320 // Be sure the clear the Thumb address bit when we calculate the size 4321 // from the current and next address 4322 if (is_arm) 4323 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4324 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 4325 } 4326 else 4327 { 4328 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4329 } 4330 } 4331 } 4332 symbol_value -= section_file_addr; 4333 } 4334 4335 if (is_debug == false) 4336 { 4337 if (type == eSymbolTypeCode) 4338 { 4339 // See if we can find a N_FUN entry for any code symbols. 4340 // If we do find a match, and the name matches, then we 4341 // can merge the two into just the function symbol to avoid 4342 // duplicate entries in the symbol table 4343 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 4344 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4345 if (range.first != range.second) 4346 { 4347 bool found_it = false; 4348 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 4349 { 4350 if (sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled)) 4351 { 4352 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4353 // We just need the flags from the linker symbol, so put these flags 4354 // into the N_FUN flags to avoid duplicate symbols in the symbol table 4355 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4356 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4357 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 4358 sym[pos->second].SetType (eSymbolTypeResolver); 4359 sym[sym_idx].Clear(); 4360 found_it = true; 4361 break; 4362 } 4363 } 4364 if (found_it) 4365 continue; 4366 } 4367 else 4368 { 4369 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 4370 type = eSymbolTypeResolver; 4371 } 4372 } 4373 else if (type == eSymbolTypeData || 4374 type == eSymbolTypeObjCClass || 4375 type == eSymbolTypeObjCMetaClass || 4376 type == eSymbolTypeObjCIVar ) 4377 { 4378 // See if we can find a N_STSYM entry for any data symbols. 4379 // If we do find a match, and the name matches, then we 4380 // can merge the two into just the Static symbol to avoid 4381 // duplicate entries in the symbol table 4382 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 4383 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4384 if (range.first != range.second) 4385 { 4386 bool found_it = false; 4387 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 4388 { 4389 if (sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled)) 4390 { 4391 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4392 // We just need the flags from the linker symbol, so put these flags 4393 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 4394 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4395 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4396 sym[sym_idx].Clear(); 4397 found_it = true; 4398 break; 4399 } 4400 } 4401 if (found_it) 4402 continue; 4403 } 4404 else 4405 { 4406 // Combine N_GSYM stab entries with the non stab symbol 4407 const char *gsym_name = sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled).GetCString(); 4408 if (gsym_name) 4409 { 4410 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(gsym_name); 4411 if (pos != N_GSYM_name_to_sym_idx.end()) 4412 { 4413 const uint32_t GSYM_sym_idx = pos->second; 4414 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4415 // Copy the address, because often the N_GSYM address has an invalid address of zero 4416 // when the global is a common symbol 4417 sym[GSYM_sym_idx].GetAddressRef().SetSection (symbol_section); 4418 sym[GSYM_sym_idx].GetAddressRef().SetOffset (symbol_value); 4419 // We just need the flags from the linker symbol, so put these flags 4420 // into the N_GSYM flags to avoid duplicate symbols in the symbol table 4421 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4422 sym[sym_idx].Clear(); 4423 continue; 4424 } 4425 } 4426 } 4427 } 4428 } 4429 4430 sym[sym_idx].SetID (nlist_idx); 4431 sym[sym_idx].SetType (type); 4432 if (set_value) 4433 { 4434 sym[sym_idx].GetAddressRef().SetSection (symbol_section); 4435 sym[sym_idx].GetAddressRef().SetOffset (symbol_value); 4436 } 4437 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4438 4439 if (symbol_byte_size > 0) 4440 sym[sym_idx].SetByteSize(symbol_byte_size); 4441 4442 if (demangled_is_synthesized) 4443 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4444 4445 ++sym_idx; 4446 } 4447 else 4448 { 4449 sym[sym_idx].Clear(); 4450 } 4451 } 4452 4453 for (const auto &pos :reexport_shlib_needs_fixup) 4454 { 4455 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4456 if (undef_pos != undefined_name_to_desc.end()) 4457 { 4458 const uint8_t dylib_ordinal = llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4459 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4460 sym[pos.first].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(dylib_ordinal-1)); 4461 } 4462 } 4463 4464 } 4465 4466 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4467 4468 if (function_starts_count > 0) 4469 { 4470 char synthetic_function_symbol[PATH_MAX]; 4471 uint32_t num_synthetic_function_symbols = 0; 4472 for (i=0; i<function_starts_count; ++i) 4473 { 4474 if (function_starts.GetEntryRef (i).data == false) 4475 ++num_synthetic_function_symbols; 4476 } 4477 4478 if (num_synthetic_function_symbols > 0) 4479 { 4480 if (num_syms < sym_idx + num_synthetic_function_symbols) 4481 { 4482 num_syms = sym_idx + num_synthetic_function_symbols; 4483 sym = symtab->Resize (num_syms); 4484 } 4485 uint32_t synthetic_function_symbol_idx = 0; 4486 for (i=0; i<function_starts_count; ++i) 4487 { 4488 const FunctionStarts::Entry *func_start_entry = function_starts.GetEntryAtIndex (i); 4489 if (func_start_entry->data == false) 4490 { 4491 addr_t symbol_file_addr = func_start_entry->addr; 4492 uint32_t symbol_flags = 0; 4493 if (is_arm) 4494 { 4495 if (symbol_file_addr & 1) 4496 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4497 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4498 } 4499 Address symbol_addr; 4500 if (module_sp->ResolveFileAddress (symbol_file_addr, symbol_addr)) 4501 { 4502 SectionSP symbol_section (symbol_addr.GetSection()); 4503 uint32_t symbol_byte_size = 0; 4504 if (symbol_section) 4505 { 4506 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4507 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 4508 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 4509 if (next_func_start_entry) 4510 { 4511 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4512 if (is_arm) 4513 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4514 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 4515 } 4516 else 4517 { 4518 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4519 } 4520 snprintf (synthetic_function_symbol, 4521 sizeof(synthetic_function_symbol), 4522 "___lldb_unnamed_function%u$$%s", 4523 ++synthetic_function_symbol_idx, 4524 module_sp->GetFileSpec().GetFilename().GetCString()); 4525 sym[sym_idx].SetID (synthetic_sym_id++); 4526 sym[sym_idx].GetMangled().SetDemangledName(ConstString(synthetic_function_symbol)); 4527 sym[sym_idx].SetType (eSymbolTypeCode); 4528 sym[sym_idx].SetIsSynthetic (true); 4529 sym[sym_idx].GetAddressRef() = symbol_addr; 4530 if (symbol_flags) 4531 sym[sym_idx].SetFlags (symbol_flags); 4532 if (symbol_byte_size) 4533 sym[sym_idx].SetByteSize (symbol_byte_size); 4534 ++sym_idx; 4535 } 4536 } 4537 } 4538 } 4539 } 4540 } 4541 4542 // Trim our symbols down to just what we ended up with after 4543 // removing any symbols. 4544 if (sym_idx < num_syms) 4545 { 4546 num_syms = sym_idx; 4547 sym = symtab->Resize (num_syms); 4548 } 4549 4550 // Now synthesize indirect symbols 4551 if (m_dysymtab.nindirectsyms != 0) 4552 { 4553 if (indirect_symbol_index_data.GetByteSize()) 4554 { 4555 NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end(); 4556 4557 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx) 4558 { 4559 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == S_SYMBOL_STUBS) 4560 { 4561 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4562 if (symbol_stub_byte_size == 0) 4563 continue; 4564 4565 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4566 4567 if (num_symbol_stubs == 0) 4568 continue; 4569 4570 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1; 4571 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) 4572 { 4573 const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx; 4574 const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size); 4575 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4576 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4)) 4577 { 4578 const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset); 4579 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4580 continue; 4581 4582 NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id); 4583 Symbol *stub_symbol = NULL; 4584 if (index_pos != end_index_pos) 4585 { 4586 // We have a remapping from the original nlist index to 4587 // a current symbol index, so just look this up by index 4588 stub_symbol = symtab->SymbolAtIndex (index_pos->second); 4589 } 4590 else 4591 { 4592 // We need to lookup a symbol using the original nlist 4593 // symbol index since this index is coming from the 4594 // S_SYMBOL_STUBS 4595 stub_symbol = symtab->FindSymbolByID (stub_sym_id); 4596 } 4597 4598 if (stub_symbol) 4599 { 4600 Address so_addr(symbol_stub_addr, section_list); 4601 4602 if (stub_symbol->GetType() == eSymbolTypeUndefined) 4603 { 4604 // Change the external symbol into a trampoline that makes sense 4605 // These symbols were N_UNDF N_EXT, and are useless to us, so we 4606 // can re-use them so we don't have to make up a synthetic symbol 4607 // for no good reason. 4608 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 4609 stub_symbol->SetType (eSymbolTypeTrampoline); 4610 else 4611 stub_symbol->SetType (eSymbolTypeResolver); 4612 stub_symbol->SetExternal (false); 4613 stub_symbol->GetAddressRef() = so_addr; 4614 stub_symbol->SetByteSize (symbol_stub_byte_size); 4615 } 4616 else 4617 { 4618 // Make a synthetic symbol to describe the trampoline stub 4619 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4620 if (sym_idx >= num_syms) 4621 { 4622 sym = symtab->Resize (++num_syms); 4623 stub_symbol = NULL; // this pointer no longer valid 4624 } 4625 sym[sym_idx].SetID (synthetic_sym_id++); 4626 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4627 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 4628 sym[sym_idx].SetType (eSymbolTypeTrampoline); 4629 else 4630 sym[sym_idx].SetType (eSymbolTypeResolver); 4631 sym[sym_idx].SetIsSynthetic (true); 4632 sym[sym_idx].GetAddressRef() = so_addr; 4633 sym[sym_idx].SetByteSize (symbol_stub_byte_size); 4634 ++sym_idx; 4635 } 4636 } 4637 else 4638 { 4639 if (log) 4640 log->Warning ("symbol stub referencing symbol table symbol %u that isn't in our minimal symbol table, fix this!!!", stub_sym_id); 4641 } 4642 } 4643 } 4644 } 4645 } 4646 } 4647 } 4648 4649 4650 if (!trie_entries.empty()) 4651 { 4652 for (const auto &e : trie_entries) 4653 { 4654 if (e.entry.import_name) 4655 { 4656 // Only add indirect symbols from the Trie entries if we 4657 // didn't have a N_INDR nlist entry for this already 4658 if (indirect_symbol_names.find(e.entry.name) == indirect_symbol_names.end()) 4659 { 4660 // Make a synthetic symbol to describe re-exported symbol. 4661 if (sym_idx >= num_syms) 4662 sym = symtab->Resize (++num_syms); 4663 sym[sym_idx].SetID (synthetic_sym_id++); 4664 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4665 sym[sym_idx].SetType (eSymbolTypeReExported); 4666 sym[sym_idx].SetIsSynthetic (true); 4667 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4668 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) 4669 { 4670 sym[sym_idx].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(e.entry.other-1)); 4671 } 4672 ++sym_idx; 4673 } 4674 } 4675 } 4676 } 4677 4678 4679 4680 // StreamFile s(stdout, false); 4681 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4682 // symtab->Dump(&s, NULL, eSortOrderNone); 4683 // Set symbol byte sizes correctly since mach-o nlist entries don't have sizes 4684 symtab->CalculateSymbolSizes(); 4685 4686 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4687 // symtab->Dump(&s, NULL, eSortOrderNone); 4688 4689 return symtab->GetNumSymbols(); 4690 } 4691 return 0; 4692 } 4693 4694 4695 void 4696 ObjectFileMachO::Dump (Stream *s) 4697 { 4698 ModuleSP module_sp(GetModule()); 4699 if (module_sp) 4700 { 4701 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4702 s->Printf("%p: ", static_cast<void*>(this)); 4703 s->Indent(); 4704 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4705 s->PutCString("ObjectFileMachO64"); 4706 else 4707 s->PutCString("ObjectFileMachO32"); 4708 4709 ArchSpec header_arch; 4710 GetArchitecture(header_arch); 4711 4712 *s << ", file = '" << m_file << "', arch = " << header_arch.GetArchitectureName() << "\n"; 4713 4714 SectionList *sections = GetSectionList(); 4715 if (sections) 4716 sections->Dump(s, NULL, true, UINT32_MAX); 4717 4718 if (m_symtab_ap.get()) 4719 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4720 } 4721 } 4722 4723 bool 4724 ObjectFileMachO::GetUUID (const llvm::MachO::mach_header &header, 4725 const lldb_private::DataExtractor &data, 4726 lldb::offset_t lc_offset, 4727 lldb_private::UUID& uuid) 4728 { 4729 uint32_t i; 4730 struct uuid_command load_cmd; 4731 4732 lldb::offset_t offset = lc_offset; 4733 for (i=0; i<header.ncmds; ++i) 4734 { 4735 const lldb::offset_t cmd_offset = offset; 4736 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4737 break; 4738 4739 if (load_cmd.cmd == LC_UUID) 4740 { 4741 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4742 4743 if (uuid_bytes) 4744 { 4745 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4746 // We pretend these object files have no UUID to prevent crashing. 4747 4748 const uint8_t opencl_uuid[] = { 0x8c, 0x8e, 0xb3, 0x9b, 4749 0x3b, 0xa8, 4750 0x4b, 0x16, 4751 0xb6, 0xa4, 4752 0x27, 0x63, 0xbb, 0x14, 0xf0, 0x0d }; 4753 4754 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4755 return false; 4756 4757 uuid.SetBytes (uuid_bytes); 4758 return true; 4759 } 4760 return false; 4761 } 4762 offset = cmd_offset + load_cmd.cmdsize; 4763 } 4764 return false; 4765 } 4766 4767 4768 bool 4769 ObjectFileMachO::GetArchitecture (const llvm::MachO::mach_header &header, 4770 const lldb_private::DataExtractor &data, 4771 lldb::offset_t lc_offset, 4772 ArchSpec &arch) 4773 { 4774 arch.SetArchitecture (eArchTypeMachO, header.cputype, header.cpusubtype); 4775 4776 if (arch.IsValid()) 4777 { 4778 llvm::Triple &triple = arch.GetTriple(); 4779 if (header.filetype == MH_PRELOAD) 4780 { 4781 // Set OS to "unknown" - this is a standalone binary with no dyld et al 4782 triple.setOS(llvm::Triple::UnknownOS); 4783 return true; 4784 } 4785 else 4786 { 4787 struct load_command load_cmd; 4788 4789 lldb::offset_t offset = lc_offset; 4790 for (uint32_t i=0; i<header.ncmds; ++i) 4791 { 4792 const lldb::offset_t cmd_offset = offset; 4793 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4794 break; 4795 4796 switch (load_cmd.cmd) 4797 { 4798 case LC_VERSION_MIN_IPHONEOS: 4799 triple.setOS (llvm::Triple::IOS); 4800 return true; 4801 4802 case LC_VERSION_MIN_MACOSX: 4803 triple.setOS (llvm::Triple::MacOSX); 4804 return true; 4805 4806 default: 4807 break; 4808 } 4809 4810 offset = cmd_offset + load_cmd.cmdsize; 4811 } 4812 4813 // Only set the OS to iOS for ARM, we don't want to set it for x86 and x86_64. 4814 // We do this because we now have MacOSX or iOS as the OS value for x86 and 4815 // x86_64 for normal desktop (MacOSX) and simulator (iOS) binaries. And if 4816 // we compare a "x86_64-apple-ios" to a "x86_64-apple-" triple, it will say 4817 // it is compatible (because the OS is unspecified in the second one and will 4818 // match anything in the first 4819 if (header.cputype == CPU_TYPE_ARM || header.cputype == CPU_TYPE_ARM64) 4820 triple.setOS (llvm::Triple::IOS); 4821 } 4822 } 4823 return arch.IsValid(); 4824 } 4825 4826 bool 4827 ObjectFileMachO::GetUUID (lldb_private::UUID* uuid) 4828 { 4829 ModuleSP module_sp(GetModule()); 4830 if (module_sp) 4831 { 4832 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4833 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4834 return GetUUID (m_header, m_data, offset, *uuid); 4835 } 4836 return false; 4837 } 4838 4839 4840 uint32_t 4841 ObjectFileMachO::GetDependentModules (FileSpecList& files) 4842 { 4843 uint32_t count = 0; 4844 ModuleSP module_sp(GetModule()); 4845 if (module_sp) 4846 { 4847 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4848 struct load_command load_cmd; 4849 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4850 std::vector<std::string> rpath_paths; 4851 std::vector<std::string> rpath_relative_paths; 4852 const bool resolve_path = false; // Don't resolve the dependent file paths since they may not reside on this system 4853 uint32_t i; 4854 for (i=0; i<m_header.ncmds; ++i) 4855 { 4856 const uint32_t cmd_offset = offset; 4857 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4858 break; 4859 4860 switch (load_cmd.cmd) 4861 { 4862 case LC_RPATH: 4863 case LC_LOAD_DYLIB: 4864 case LC_LOAD_WEAK_DYLIB: 4865 case LC_REEXPORT_DYLIB: 4866 case LC_LOAD_DYLINKER: 4867 case LC_LOADFVMLIB: 4868 case LC_LOAD_UPWARD_DYLIB: 4869 { 4870 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 4871 const char *path = m_data.PeekCStr(name_offset); 4872 if (path) 4873 { 4874 if (load_cmd.cmd == LC_RPATH) 4875 rpath_paths.push_back(path); 4876 else 4877 { 4878 if (path[0] == '@') 4879 { 4880 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 4881 rpath_relative_paths.push_back(path + strlen("@rpath")); 4882 } 4883 else 4884 { 4885 FileSpec file_spec(path, resolve_path); 4886 if (files.AppendIfUnique(file_spec)) 4887 count++; 4888 } 4889 } 4890 } 4891 } 4892 break; 4893 4894 default: 4895 break; 4896 } 4897 offset = cmd_offset + load_cmd.cmdsize; 4898 } 4899 4900 if (!rpath_paths.empty()) 4901 { 4902 // Fixup all LC_RPATH values to be absolute paths 4903 FileSpec this_file_spec(m_file); 4904 this_file_spec.ResolvePath(); 4905 std::string loader_path("@loader_path"); 4906 std::string executable_path("@executable_path"); 4907 for (auto &rpath : rpath_paths) 4908 { 4909 if (rpath.find(loader_path) == 0) 4910 { 4911 rpath.erase(0, loader_path.size()); 4912 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 4913 } 4914 else if (rpath.find(executable_path) == 0) 4915 { 4916 rpath.erase(0, executable_path.size()); 4917 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 4918 } 4919 } 4920 4921 for (const auto &rpath_relative_path : rpath_relative_paths) 4922 { 4923 for (const auto &rpath : rpath_paths) 4924 { 4925 std::string path = rpath; 4926 path += rpath_relative_path; 4927 // It is OK to resolve this path because we must find a file on 4928 // disk for us to accept it anyway if it is rpath relative. 4929 FileSpec file_spec(path, true); 4930 // Remove any redundant parts of the path (like "../foo") since 4931 // LC_RPATH values often contain "..". 4932 file_spec.NormalizePath (); 4933 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) 4934 { 4935 count++; 4936 break; 4937 } 4938 } 4939 } 4940 } 4941 } 4942 return count; 4943 } 4944 4945 lldb_private::Address 4946 ObjectFileMachO::GetEntryPointAddress () 4947 { 4948 // If the object file is not an executable it can't hold the entry point. m_entry_point_address 4949 // is initialized to an invalid address, so we can just return that. 4950 // If m_entry_point_address is valid it means we've found it already, so return the cached value. 4951 4952 if (!IsExecutable() || m_entry_point_address.IsValid()) 4953 return m_entry_point_address; 4954 4955 // Otherwise, look for the UnixThread or Thread command. The data for the Thread command is given in 4956 // /usr/include/mach-o.h, but it is basically: 4957 // 4958 // uint32_t flavor - this is the flavor argument you would pass to thread_get_state 4959 // uint32_t count - this is the count of longs in the thread state data 4960 // struct XXX_thread_state state - this is the structure from <machine/thread_status.h> corresponding to the flavor. 4961 // <repeat this trio> 4962 // 4963 // So we just keep reading the various register flavors till we find the GPR one, then read the PC out of there. 4964 // FIXME: We will need to have a "RegisterContext data provider" class at some point that can get all the registers 4965 // out of data in this form & attach them to a given thread. That should underlie the MacOS X User process plugin, 4966 // and we'll also need it for the MacOS X Core File process plugin. When we have that we can also use it here. 4967 // 4968 // For now we hard-code the offsets and flavors we need: 4969 // 4970 // 4971 4972 ModuleSP module_sp(GetModule()); 4973 if (module_sp) 4974 { 4975 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4976 struct load_command load_cmd; 4977 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4978 uint32_t i; 4979 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 4980 bool done = false; 4981 4982 for (i=0; i<m_header.ncmds; ++i) 4983 { 4984 const lldb::offset_t cmd_offset = offset; 4985 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4986 break; 4987 4988 switch (load_cmd.cmd) 4989 { 4990 case LC_UNIXTHREAD: 4991 case LC_THREAD: 4992 { 4993 while (offset < cmd_offset + load_cmd.cmdsize) 4994 { 4995 uint32_t flavor = m_data.GetU32(&offset); 4996 uint32_t count = m_data.GetU32(&offset); 4997 if (count == 0) 4998 { 4999 // We've gotten off somehow, log and exit; 5000 return m_entry_point_address; 5001 } 5002 5003 switch (m_header.cputype) 5004 { 5005 case llvm::MachO::CPU_TYPE_ARM: 5006 if (flavor == 1) // ARM_THREAD_STATE from mach/arm/thread_status.h 5007 { 5008 offset += 60; // This is the offset of pc in the GPR thread state data structure. 5009 start_address = m_data.GetU32(&offset); 5010 done = true; 5011 } 5012 break; 5013 case llvm::MachO::CPU_TYPE_ARM64: 5014 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5015 { 5016 offset += 256; // This is the offset of pc in the GPR thread state data structure. 5017 start_address = m_data.GetU64(&offset); 5018 done = true; 5019 } 5020 break; 5021 case llvm::MachO::CPU_TYPE_I386: 5022 if (flavor == 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5023 { 5024 offset += 40; // This is the offset of eip in the GPR thread state data structure. 5025 start_address = m_data.GetU32(&offset); 5026 done = true; 5027 } 5028 break; 5029 case llvm::MachO::CPU_TYPE_X86_64: 5030 if (flavor == 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5031 { 5032 offset += 16 * 8; // This is the offset of rip in the GPR thread state data structure. 5033 start_address = m_data.GetU64(&offset); 5034 done = true; 5035 } 5036 break; 5037 default: 5038 return m_entry_point_address; 5039 } 5040 // Haven't found the GPR flavor yet, skip over the data for this flavor: 5041 if (done) 5042 break; 5043 offset += count * 4; 5044 } 5045 } 5046 break; 5047 case LC_MAIN: 5048 { 5049 ConstString text_segment_name ("__TEXT"); 5050 uint64_t entryoffset = m_data.GetU64(&offset); 5051 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 5052 if (text_segment_sp) 5053 { 5054 done = true; 5055 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5056 } 5057 } 5058 5059 default: 5060 break; 5061 } 5062 if (done) 5063 break; 5064 5065 // Go to the next load command: 5066 offset = cmd_offset + load_cmd.cmdsize; 5067 } 5068 5069 if (start_address != LLDB_INVALID_ADDRESS) 5070 { 5071 // We got the start address from the load commands, so now resolve that address in the sections 5072 // of this ObjectFile: 5073 if (!m_entry_point_address.ResolveAddressUsingFileSections (start_address, GetSectionList())) 5074 { 5075 m_entry_point_address.Clear(); 5076 } 5077 } 5078 else 5079 { 5080 // We couldn't read the UnixThread load command - maybe it wasn't there. As a fallback look for the 5081 // "start" symbol in the main executable. 5082 5083 ModuleSP module_sp (GetModule()); 5084 5085 if (module_sp) 5086 { 5087 SymbolContextList contexts; 5088 SymbolContext context; 5089 if (module_sp->FindSymbolsWithNameAndType(ConstString ("start"), eSymbolTypeCode, contexts)) 5090 { 5091 if (contexts.GetContextAtIndex(0, context)) 5092 m_entry_point_address = context.symbol->GetAddress(); 5093 } 5094 } 5095 } 5096 } 5097 5098 return m_entry_point_address; 5099 5100 } 5101 5102 lldb_private::Address 5103 ObjectFileMachO::GetHeaderAddress () 5104 { 5105 lldb_private::Address header_addr; 5106 SectionList *section_list = GetSectionList(); 5107 if (section_list) 5108 { 5109 SectionSP text_segment_sp (section_list->FindSectionByName (GetSegmentNameTEXT())); 5110 if (text_segment_sp) 5111 { 5112 header_addr.SetSection (text_segment_sp); 5113 header_addr.SetOffset (0); 5114 } 5115 } 5116 return header_addr; 5117 } 5118 5119 uint32_t 5120 ObjectFileMachO::GetNumThreadContexts () 5121 { 5122 ModuleSP module_sp(GetModule()); 5123 if (module_sp) 5124 { 5125 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 5126 if (!m_thread_context_offsets_valid) 5127 { 5128 m_thread_context_offsets_valid = true; 5129 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5130 FileRangeArray::Entry file_range; 5131 thread_command thread_cmd; 5132 for (uint32_t i=0; i<m_header.ncmds; ++i) 5133 { 5134 const uint32_t cmd_offset = offset; 5135 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 5136 break; 5137 5138 if (thread_cmd.cmd == LC_THREAD) 5139 { 5140 file_range.SetRangeBase (offset); 5141 file_range.SetByteSize (thread_cmd.cmdsize - 8); 5142 m_thread_context_offsets.Append (file_range); 5143 } 5144 offset = cmd_offset + thread_cmd.cmdsize; 5145 } 5146 } 5147 } 5148 return m_thread_context_offsets.GetSize(); 5149 } 5150 5151 lldb::RegisterContextSP 5152 ObjectFileMachO::GetThreadContextAtIndex (uint32_t idx, lldb_private::Thread &thread) 5153 { 5154 lldb::RegisterContextSP reg_ctx_sp; 5155 5156 ModuleSP module_sp(GetModule()); 5157 if (module_sp) 5158 { 5159 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 5160 if (!m_thread_context_offsets_valid) 5161 GetNumThreadContexts (); 5162 5163 const FileRangeArray::Entry *thread_context_file_range = m_thread_context_offsets.GetEntryAtIndex (idx); 5164 if (thread_context_file_range) 5165 { 5166 5167 DataExtractor data (m_data, 5168 thread_context_file_range->GetRangeBase(), 5169 thread_context_file_range->GetByteSize()); 5170 5171 switch (m_header.cputype) 5172 { 5173 case llvm::MachO::CPU_TYPE_ARM64: 5174 reg_ctx_sp.reset (new RegisterContextDarwin_arm64_Mach (thread, data)); 5175 break; 5176 5177 case llvm::MachO::CPU_TYPE_ARM: 5178 reg_ctx_sp.reset (new RegisterContextDarwin_arm_Mach (thread, data)); 5179 break; 5180 5181 case llvm::MachO::CPU_TYPE_I386: 5182 reg_ctx_sp.reset (new RegisterContextDarwin_i386_Mach (thread, data)); 5183 break; 5184 5185 case llvm::MachO::CPU_TYPE_X86_64: 5186 reg_ctx_sp.reset (new RegisterContextDarwin_x86_64_Mach (thread, data)); 5187 break; 5188 } 5189 } 5190 } 5191 return reg_ctx_sp; 5192 } 5193 5194 5195 ObjectFile::Type 5196 ObjectFileMachO::CalculateType() 5197 { 5198 switch (m_header.filetype) 5199 { 5200 case MH_OBJECT: // 0x1u 5201 if (GetAddressByteSize () == 4) 5202 { 5203 // 32 bit kexts are just object files, but they do have a valid 5204 // UUID load command. 5205 UUID uuid; 5206 if (GetUUID(&uuid)) 5207 { 5208 // this checking for the UUID load command is not enough 5209 // we could eventually look for the symbol named 5210 // "OSKextGetCurrentIdentifier" as this is required of kexts 5211 if (m_strata == eStrataInvalid) 5212 m_strata = eStrataKernel; 5213 return eTypeSharedLibrary; 5214 } 5215 } 5216 return eTypeObjectFile; 5217 5218 case MH_EXECUTE: return eTypeExecutable; // 0x2u 5219 case MH_FVMLIB: return eTypeSharedLibrary; // 0x3u 5220 case MH_CORE: return eTypeCoreFile; // 0x4u 5221 case MH_PRELOAD: return eTypeSharedLibrary; // 0x5u 5222 case MH_DYLIB: return eTypeSharedLibrary; // 0x6u 5223 case MH_DYLINKER: return eTypeDynamicLinker; // 0x7u 5224 case MH_BUNDLE: return eTypeSharedLibrary; // 0x8u 5225 case MH_DYLIB_STUB: return eTypeStubLibrary; // 0x9u 5226 case MH_DSYM: return eTypeDebugInfo; // 0xAu 5227 case MH_KEXT_BUNDLE: return eTypeSharedLibrary; // 0xBu 5228 default: 5229 break; 5230 } 5231 return eTypeUnknown; 5232 } 5233 5234 ObjectFile::Strata 5235 ObjectFileMachO::CalculateStrata() 5236 { 5237 switch (m_header.filetype) 5238 { 5239 case MH_OBJECT: // 0x1u 5240 { 5241 // 32 bit kexts are just object files, but they do have a valid 5242 // UUID load command. 5243 UUID uuid; 5244 if (GetUUID(&uuid)) 5245 { 5246 // this checking for the UUID load command is not enough 5247 // we could eventually look for the symbol named 5248 // "OSKextGetCurrentIdentifier" as this is required of kexts 5249 if (m_type == eTypeInvalid) 5250 m_type = eTypeSharedLibrary; 5251 5252 return eStrataKernel; 5253 } 5254 } 5255 return eStrataUnknown; 5256 5257 case MH_EXECUTE: // 0x2u 5258 // Check for the MH_DYLDLINK bit in the flags 5259 if (m_header.flags & MH_DYLDLINK) 5260 { 5261 return eStrataUser; 5262 } 5263 else 5264 { 5265 SectionList *section_list = GetSectionList(); 5266 if (section_list) 5267 { 5268 static ConstString g_kld_section_name ("__KLD"); 5269 if (section_list->FindSectionByName(g_kld_section_name)) 5270 return eStrataKernel; 5271 } 5272 } 5273 return eStrataRawImage; 5274 5275 case MH_FVMLIB: return eStrataUser; // 0x3u 5276 case MH_CORE: return eStrataUnknown; // 0x4u 5277 case MH_PRELOAD: return eStrataRawImage; // 0x5u 5278 case MH_DYLIB: return eStrataUser; // 0x6u 5279 case MH_DYLINKER: return eStrataUser; // 0x7u 5280 case MH_BUNDLE: return eStrataUser; // 0x8u 5281 case MH_DYLIB_STUB: return eStrataUser; // 0x9u 5282 case MH_DSYM: return eStrataUnknown; // 0xAu 5283 case MH_KEXT_BUNDLE: return eStrataKernel; // 0xBu 5284 default: 5285 break; 5286 } 5287 return eStrataUnknown; 5288 } 5289 5290 5291 uint32_t 5292 ObjectFileMachO::GetVersion (uint32_t *versions, uint32_t num_versions) 5293 { 5294 ModuleSP module_sp(GetModule()); 5295 if (module_sp) 5296 { 5297 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 5298 struct dylib_command load_cmd; 5299 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5300 uint32_t version_cmd = 0; 5301 uint64_t version = 0; 5302 uint32_t i; 5303 for (i=0; i<m_header.ncmds; ++i) 5304 { 5305 const lldb::offset_t cmd_offset = offset; 5306 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5307 break; 5308 5309 if (load_cmd.cmd == LC_ID_DYLIB) 5310 { 5311 if (version_cmd == 0) 5312 { 5313 version_cmd = load_cmd.cmd; 5314 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5315 break; 5316 version = load_cmd.dylib.current_version; 5317 } 5318 break; // Break for now unless there is another more complete version 5319 // number load command in the future. 5320 } 5321 offset = cmd_offset + load_cmd.cmdsize; 5322 } 5323 5324 if (version_cmd == LC_ID_DYLIB) 5325 { 5326 if (versions != NULL && num_versions > 0) 5327 { 5328 if (num_versions > 0) 5329 versions[0] = (version & 0xFFFF0000ull) >> 16; 5330 if (num_versions > 1) 5331 versions[1] = (version & 0x0000FF00ull) >> 8; 5332 if (num_versions > 2) 5333 versions[2] = (version & 0x000000FFull); 5334 // Fill in an remaining version numbers with invalid values 5335 for (i=3; i<num_versions; ++i) 5336 versions[i] = UINT32_MAX; 5337 } 5338 // The LC_ID_DYLIB load command has a version with 3 version numbers 5339 // in it, so always return 3 5340 return 3; 5341 } 5342 } 5343 return false; 5344 } 5345 5346 bool 5347 ObjectFileMachO::GetArchitecture (ArchSpec &arch) 5348 { 5349 ModuleSP module_sp(GetModule()); 5350 if (module_sp) 5351 { 5352 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 5353 return GetArchitecture (m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), arch); 5354 } 5355 return false; 5356 } 5357 5358 5359 UUID 5360 ObjectFileMachO::GetProcessSharedCacheUUID (Process *process) 5361 { 5362 UUID uuid; 5363 if (process) 5364 { 5365 addr_t all_image_infos = process->GetImageInfoAddress(); 5366 5367 // The address returned by GetImageInfoAddress may be the address of dyld (don't want) 5368 // or it may be the address of the dyld_all_image_infos structure (want). The first four 5369 // bytes will be either the version field (all_image_infos) or a Mach-O file magic constant. 5370 // Version 13 and higher of dyld_all_image_infos is required to get the sharedCacheUUID field. 5371 5372 Error err; 5373 uint32_t version_or_magic = process->ReadUnsignedIntegerFromMemory (all_image_infos, 4, -1, err); 5374 if (version_or_magic != static_cast<uint32_t>(-1) 5375 && version_or_magic != MH_MAGIC 5376 && version_or_magic != MH_CIGAM 5377 && version_or_magic != MH_MAGIC_64 5378 && version_or_magic != MH_CIGAM_64 5379 && version_or_magic >= 13) 5380 { 5381 addr_t sharedCacheUUID_address = LLDB_INVALID_ADDRESS; 5382 int wordsize = process->GetAddressByteSize(); 5383 if (wordsize == 8) 5384 { 5385 sharedCacheUUID_address = all_image_infos + 160; // sharedCacheUUID <mach-o/dyld_images.h> 5386 } 5387 if (wordsize == 4) 5388 { 5389 sharedCacheUUID_address = all_image_infos + 84; // sharedCacheUUID <mach-o/dyld_images.h> 5390 } 5391 if (sharedCacheUUID_address != LLDB_INVALID_ADDRESS) 5392 { 5393 uuid_t shared_cache_uuid; 5394 if (process->ReadMemory (sharedCacheUUID_address, shared_cache_uuid, sizeof (uuid_t), err) == sizeof (uuid_t)) 5395 { 5396 uuid.SetBytes (shared_cache_uuid); 5397 } 5398 } 5399 } 5400 } 5401 return uuid; 5402 } 5403 5404 UUID 5405 ObjectFileMachO::GetLLDBSharedCacheUUID () 5406 { 5407 UUID uuid; 5408 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 5409 uint8_t *(*dyld_get_all_image_infos)(void); 5410 dyld_get_all_image_infos = (uint8_t*(*)()) dlsym (RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5411 if (dyld_get_all_image_infos) 5412 { 5413 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5414 if (dyld_all_image_infos_address) 5415 { 5416 uint32_t *version = (uint32_t*) dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5417 if (*version >= 13) 5418 { 5419 uuid_t *sharedCacheUUID_address = 0; 5420 int wordsize = sizeof (uint8_t *); 5421 if (wordsize == 8) 5422 { 5423 sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 160); // sharedCacheUUID <mach-o/dyld_images.h> 5424 } 5425 else 5426 { 5427 sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 84); // sharedCacheUUID <mach-o/dyld_images.h> 5428 } 5429 uuid.SetBytes (sharedCacheUUID_address); 5430 } 5431 } 5432 } 5433 #endif 5434 return uuid; 5435 } 5436 5437 uint32_t 5438 ObjectFileMachO::GetMinimumOSVersion (uint32_t *versions, uint32_t num_versions) 5439 { 5440 if (m_min_os_versions.empty()) 5441 { 5442 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5443 bool success = false; 5444 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 5445 { 5446 const lldb::offset_t load_cmd_offset = offset; 5447 5448 version_min_command lc; 5449 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5450 break; 5451 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 5452 { 5453 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 5454 { 5455 const uint32_t xxxx = lc.version >> 16; 5456 const uint32_t yy = (lc.version >> 8) & 0xffu; 5457 const uint32_t zz = lc.version & 0xffu; 5458 if (xxxx) 5459 { 5460 m_min_os_versions.push_back(xxxx); 5461 m_min_os_versions.push_back(yy); 5462 m_min_os_versions.push_back(zz); 5463 } 5464 success = true; 5465 } 5466 } 5467 offset = load_cmd_offset + lc.cmdsize; 5468 } 5469 5470 if (success == false) 5471 { 5472 // Push an invalid value so we don't keep trying to 5473 m_min_os_versions.push_back(UINT32_MAX); 5474 } 5475 } 5476 5477 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) 5478 { 5479 if (versions != NULL && num_versions > 0) 5480 { 5481 for (size_t i=0; i<num_versions; ++i) 5482 { 5483 if (i < m_min_os_versions.size()) 5484 versions[i] = m_min_os_versions[i]; 5485 else 5486 versions[i] = 0; 5487 } 5488 } 5489 return m_min_os_versions.size(); 5490 } 5491 // Call the superclasses version that will empty out the data 5492 return ObjectFile::GetMinimumOSVersion (versions, num_versions); 5493 } 5494 5495 uint32_t 5496 ObjectFileMachO::GetSDKVersion(uint32_t *versions, uint32_t num_versions) 5497 { 5498 if (m_sdk_versions.empty()) 5499 { 5500 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5501 bool success = false; 5502 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 5503 { 5504 const lldb::offset_t load_cmd_offset = offset; 5505 5506 version_min_command lc; 5507 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5508 break; 5509 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 5510 { 5511 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 5512 { 5513 const uint32_t xxxx = lc.sdk >> 16; 5514 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5515 const uint32_t zz = lc.sdk & 0xffu; 5516 if (xxxx) 5517 { 5518 m_sdk_versions.push_back(xxxx); 5519 m_sdk_versions.push_back(yy); 5520 m_sdk_versions.push_back(zz); 5521 } 5522 success = true; 5523 } 5524 } 5525 offset = load_cmd_offset + lc.cmdsize; 5526 } 5527 5528 if (success == false) 5529 { 5530 // Push an invalid value so we don't keep trying to 5531 m_sdk_versions.push_back(UINT32_MAX); 5532 } 5533 } 5534 5535 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) 5536 { 5537 if (versions != NULL && num_versions > 0) 5538 { 5539 for (size_t i=0; i<num_versions; ++i) 5540 { 5541 if (i < m_sdk_versions.size()) 5542 versions[i] = m_sdk_versions[i]; 5543 else 5544 versions[i] = 0; 5545 } 5546 } 5547 return m_sdk_versions.size(); 5548 } 5549 // Call the superclasses version that will empty out the data 5550 return ObjectFile::GetSDKVersion (versions, num_versions); 5551 } 5552 5553 5554 bool 5555 ObjectFileMachO::GetIsDynamicLinkEditor() 5556 { 5557 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5558 } 5559 5560 //------------------------------------------------------------------ 5561 // PluginInterface protocol 5562 //------------------------------------------------------------------ 5563 lldb_private::ConstString 5564 ObjectFileMachO::GetPluginName() 5565 { 5566 return GetPluginNameStatic(); 5567 } 5568 5569 uint32_t 5570 ObjectFileMachO::GetPluginVersion() 5571 { 5572 return 1; 5573 } 5574 5575 5576 Section * 5577 ObjectFileMachO::GetMachHeaderSection() 5578 { 5579 // Find the first address of the mach header which is the first non-zero 5580 // file sized section whose file offset is zero. This is the base file address 5581 // of the mach-o file which can be subtracted from the vmaddr of the other 5582 // segments found in memory and added to the load address 5583 ModuleSP module_sp = GetModule(); 5584 if (module_sp) 5585 { 5586 SectionList *section_list = GetSectionList (); 5587 if (section_list) 5588 { 5589 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5590 const size_t num_sections = section_list->GetSize(); 5591 5592 for (size_t sect_idx = 0; 5593 sect_idx < num_sections && mach_base_file_addr == LLDB_INVALID_ADDRESS; 5594 ++sect_idx) 5595 { 5596 Section *section = section_list->GetSectionAtIndex (sect_idx).get(); 5597 if (section && 5598 section->GetFileSize() > 0 && 5599 section->GetFileOffset() == 0 && 5600 section->IsThreadSpecific() == false && 5601 module_sp.get() == section->GetModule().get()) 5602 { 5603 return section; 5604 } 5605 } 5606 } 5607 } 5608 return nullptr; 5609 } 5610 5611 lldb::addr_t 5612 ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage(lldb::addr_t mach_header_load_address, const Section *mach_header_section, const Section *section) 5613 { 5614 ModuleSP module_sp = GetModule(); 5615 if (module_sp && mach_header_section && section && mach_header_load_address != LLDB_INVALID_ADDRESS) 5616 { 5617 lldb::addr_t mach_header_file_addr = mach_header_section->GetFileAddress(); 5618 if (mach_header_file_addr != LLDB_INVALID_ADDRESS) 5619 { 5620 if (section && 5621 section->GetFileSize() > 0 && 5622 section->IsThreadSpecific() == false && 5623 module_sp.get() == section->GetModule().get()) 5624 { 5625 // Ignore __LINKEDIT and __DWARF segments 5626 if (section->GetName() == GetSegmentNameLINKEDIT()) 5627 { 5628 // Only map __LINKEDIT if we have an in memory image and this isn't 5629 // a kernel binary like a kext or mach_kernel. 5630 const bool is_memory_image = (bool)m_process_wp.lock(); 5631 const Strata strata = GetStrata(); 5632 if (is_memory_image == false || strata == eStrataKernel) 5633 return LLDB_INVALID_ADDRESS; 5634 } 5635 return section->GetFileAddress() - mach_header_file_addr + mach_header_load_address; 5636 } 5637 } 5638 } 5639 return LLDB_INVALID_ADDRESS; 5640 } 5641 5642 bool 5643 ObjectFileMachO::SetLoadAddress (Target &target, 5644 lldb::addr_t value, 5645 bool value_is_offset) 5646 { 5647 ModuleSP module_sp = GetModule(); 5648 if (module_sp) 5649 { 5650 size_t num_loaded_sections = 0; 5651 SectionList *section_list = GetSectionList (); 5652 if (section_list) 5653 { 5654 const size_t num_sections = section_list->GetSize(); 5655 5656 if (value_is_offset) 5657 { 5658 // "value" is an offset to apply to each top level segment 5659 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) 5660 { 5661 // Iterate through the object file sections to find all 5662 // of the sections that size on disk (to avoid __PAGEZERO) 5663 // and load them 5664 SectionSP section_sp (section_list->GetSectionAtIndex (sect_idx)); 5665 if (section_sp && 5666 section_sp->GetFileSize() > 0 && 5667 section_sp->IsThreadSpecific() == false && 5668 module_sp.get() == section_sp->GetModule().get()) 5669 { 5670 // Ignore __LINKEDIT and __DWARF segments 5671 if (section_sp->GetName() == GetSegmentNameLINKEDIT()) 5672 { 5673 // Only map __LINKEDIT if we have an in memory image and this isn't 5674 // a kernel binary like a kext or mach_kernel. 5675 const bool is_memory_image = (bool)m_process_wp.lock(); 5676 const Strata strata = GetStrata(); 5677 if (is_memory_image == false || strata == eStrataKernel) 5678 continue; 5679 } 5680 if (target.GetSectionLoadList().SetSectionLoadAddress (section_sp, section_sp->GetFileAddress() + value)) 5681 ++num_loaded_sections; 5682 } 5683 } 5684 } 5685 else 5686 { 5687 // "value" is the new base address of the mach_header, adjust each 5688 // section accordingly 5689 5690 Section *mach_header_section = GetMachHeaderSection(); 5691 if (mach_header_section) 5692 { 5693 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) 5694 { 5695 SectionSP section_sp (section_list->GetSectionAtIndex (sect_idx)); 5696 5697 lldb::addr_t section_load_addr = CalculateSectionLoadAddressForMemoryImage(value, mach_header_section, section_sp.get()); 5698 if (section_load_addr != LLDB_INVALID_ADDRESS) 5699 { 5700 if (target.GetSectionLoadList().SetSectionLoadAddress (section_sp, section_load_addr)) 5701 ++num_loaded_sections; 5702 } 5703 } 5704 } 5705 } 5706 } 5707 return num_loaded_sections > 0; 5708 } 5709 return false; 5710 } 5711 5712 bool 5713 ObjectFileMachO::SaveCore (const lldb::ProcessSP &process_sp, 5714 const FileSpec &outfile, 5715 Error &error) 5716 { 5717 if (process_sp) 5718 { 5719 Target &target = process_sp->GetTarget(); 5720 const ArchSpec target_arch = target.GetArchitecture(); 5721 const llvm::Triple &target_triple = target_arch.GetTriple(); 5722 if (target_triple.getVendor() == llvm::Triple::Apple && 5723 (target_triple.getOS() == llvm::Triple::MacOSX || 5724 target_triple.getOS() == llvm::Triple::IOS)) 5725 { 5726 bool make_core = false; 5727 switch (target_arch.GetMachine()) 5728 { 5729 case llvm::Triple::aarch64: 5730 case llvm::Triple::arm: 5731 case llvm::Triple::thumb: 5732 case llvm::Triple::x86: 5733 case llvm::Triple::x86_64: 5734 make_core = true; 5735 break; 5736 default: 5737 error.SetErrorStringWithFormat ("unsupported core architecture: %s", target_triple.str().c_str()); 5738 break; 5739 } 5740 5741 if (make_core) 5742 { 5743 std::vector<segment_command_64> segment_load_commands; 5744 // uint32_t range_info_idx = 0; 5745 MemoryRegionInfo range_info; 5746 Error range_error = process_sp->GetMemoryRegionInfo(0, range_info); 5747 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 5748 const ByteOrder byte_order = target_arch.GetByteOrder(); 5749 if (range_error.Success()) 5750 { 5751 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) 5752 { 5753 const addr_t addr = range_info.GetRange().GetRangeBase(); 5754 const addr_t size = range_info.GetRange().GetByteSize(); 5755 5756 if (size == 0) 5757 break; 5758 5759 // Calculate correct protections 5760 uint32_t prot = 0; 5761 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 5762 prot |= VM_PROT_READ; 5763 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 5764 prot |= VM_PROT_WRITE; 5765 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 5766 prot |= VM_PROT_EXECUTE; 5767 5768 // printf ("[%3u] [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 ") %c%c%c\n", 5769 // range_info_idx, 5770 // addr, 5771 // size, 5772 // (prot & VM_PROT_READ ) ? 'r' : '-', 5773 // (prot & VM_PROT_WRITE ) ? 'w' : '-', 5774 // (prot & VM_PROT_EXECUTE) ? 'x' : '-'); 5775 5776 if (prot != 0) 5777 { 5778 uint32_t cmd_type = LC_SEGMENT_64; 5779 uint32_t segment_size = sizeof (segment_command_64); 5780 if (addr_byte_size == 4) 5781 { 5782 cmd_type = LC_SEGMENT; 5783 segment_size = sizeof (segment_command); 5784 } 5785 segment_command_64 segment = { 5786 cmd_type, // uint32_t cmd; 5787 segment_size, // uint32_t cmdsize; 5788 {0}, // char segname[16]; 5789 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 5790 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 5791 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 5792 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 5793 prot, // uint32_t maxprot; 5794 prot, // uint32_t initprot; 5795 0, // uint32_t nsects; 5796 0 }; // uint32_t flags; 5797 segment_load_commands.push_back(segment); 5798 } 5799 else 5800 { 5801 // No protections and a size of 1 used to be returned from old 5802 // debugservers when we asked about a region that was past the 5803 // last memory region and it indicates the end... 5804 if (size == 1) 5805 break; 5806 } 5807 5808 range_error = process_sp->GetMemoryRegionInfo(range_info.GetRange().GetRangeEnd(), range_info); 5809 if (range_error.Fail()) 5810 break; 5811 } 5812 5813 StreamString buffer (Stream::eBinary, 5814 addr_byte_size, 5815 byte_order); 5816 5817 mach_header_64 mach_header; 5818 if (addr_byte_size == 8) 5819 { 5820 mach_header.magic = MH_MAGIC_64; 5821 } 5822 else 5823 { 5824 mach_header.magic = MH_MAGIC; 5825 } 5826 mach_header.cputype = target_arch.GetMachOCPUType(); 5827 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 5828 mach_header.filetype = MH_CORE; 5829 mach_header.ncmds = segment_load_commands.size(); 5830 mach_header.flags = 0; 5831 mach_header.reserved = 0; 5832 ThreadList &thread_list = process_sp->GetThreadList(); 5833 const uint32_t num_threads = thread_list.GetSize(); 5834 5835 // Make an array of LC_THREAD data items. Each one contains 5836 // the contents of the LC_THREAD load command. The data doesn't 5837 // contain the load command + load command size, we will 5838 // add the load command and load command size as we emit the data. 5839 std::vector<StreamString> LC_THREAD_datas(num_threads); 5840 for (auto &LC_THREAD_data : LC_THREAD_datas) 5841 { 5842 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 5843 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 5844 LC_THREAD_data.SetByteOrder(byte_order); 5845 } 5846 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) 5847 { 5848 ThreadSP thread_sp (thread_list.GetThreadAtIndex(thread_idx)); 5849 if (thread_sp) 5850 { 5851 switch (mach_header.cputype) 5852 { 5853 case llvm::MachO::CPU_TYPE_ARM64: 5854 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5855 break; 5856 5857 case llvm::MachO::CPU_TYPE_ARM: 5858 RegisterContextDarwin_arm_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5859 break; 5860 5861 case llvm::MachO::CPU_TYPE_I386: 5862 RegisterContextDarwin_i386_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5863 break; 5864 5865 case llvm::MachO::CPU_TYPE_X86_64: 5866 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5867 break; 5868 } 5869 5870 } 5871 } 5872 5873 // The size of the load command is the size of the segments... 5874 if (addr_byte_size == 8) 5875 { 5876 mach_header.sizeofcmds = segment_load_commands.size() * sizeof (struct segment_command_64); 5877 } 5878 else 5879 { 5880 mach_header.sizeofcmds = segment_load_commands.size() * sizeof (struct segment_command); 5881 } 5882 5883 // and the size of all LC_THREAD load command 5884 for (const auto &LC_THREAD_data : LC_THREAD_datas) 5885 { 5886 ++mach_header.ncmds; 5887 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 5888 } 5889 5890 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", 5891 mach_header.magic, 5892 mach_header.cputype, 5893 mach_header.cpusubtype, 5894 mach_header.filetype, 5895 mach_header.ncmds, 5896 mach_header.sizeofcmds, 5897 mach_header.flags, 5898 mach_header.reserved); 5899 5900 // Write the mach header 5901 buffer.PutHex32(mach_header.magic); 5902 buffer.PutHex32(mach_header.cputype); 5903 buffer.PutHex32(mach_header.cpusubtype); 5904 buffer.PutHex32(mach_header.filetype); 5905 buffer.PutHex32(mach_header.ncmds); 5906 buffer.PutHex32(mach_header.sizeofcmds); 5907 buffer.PutHex32(mach_header.flags); 5908 if (addr_byte_size == 8) 5909 { 5910 buffer.PutHex32(mach_header.reserved); 5911 } 5912 5913 // Skip the mach header and all load commands and align to the next 5914 // 0x1000 byte boundary 5915 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 5916 if (file_offset & 0x00000fff) 5917 { 5918 file_offset += 0x00001000ull; 5919 file_offset &= (~0x00001000ull + 1); 5920 } 5921 5922 for (auto &segment : segment_load_commands) 5923 { 5924 segment.fileoff = file_offset; 5925 file_offset += segment.filesize; 5926 } 5927 5928 // Write out all of the LC_THREAD load commands 5929 for (const auto &LC_THREAD_data : LC_THREAD_datas) 5930 { 5931 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 5932 buffer.PutHex32(LC_THREAD); 5933 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 5934 buffer.Write(LC_THREAD_data.GetData(), LC_THREAD_data_size); 5935 } 5936 5937 // Write out all of the segment load commands 5938 for (const auto &segment : segment_load_commands) 5939 { 5940 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", 5941 segment.cmd, 5942 segment.cmdsize, 5943 segment.vmaddr, 5944 segment.vmaddr + segment.vmsize, 5945 segment.fileoff, 5946 segment.filesize, 5947 segment.maxprot, 5948 segment.initprot, 5949 segment.nsects, 5950 segment.flags); 5951 5952 buffer.PutHex32(segment.cmd); 5953 buffer.PutHex32(segment.cmdsize); 5954 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 5955 if (addr_byte_size == 8) 5956 { 5957 buffer.PutHex64(segment.vmaddr); 5958 buffer.PutHex64(segment.vmsize); 5959 buffer.PutHex64(segment.fileoff); 5960 buffer.PutHex64(segment.filesize); 5961 } 5962 else 5963 { 5964 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 5965 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 5966 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 5967 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 5968 } 5969 buffer.PutHex32(segment.maxprot); 5970 buffer.PutHex32(segment.initprot); 5971 buffer.PutHex32(segment.nsects); 5972 buffer.PutHex32(segment.flags); 5973 } 5974 5975 File core_file; 5976 std::string core_file_path(outfile.GetPath()); 5977 error = core_file.Open(core_file_path.c_str(), 5978 File::eOpenOptionWrite | 5979 File::eOpenOptionTruncate | 5980 File::eOpenOptionCanCreate); 5981 if (error.Success()) 5982 { 5983 // Read 1 page at a time 5984 uint8_t bytes[0x1000]; 5985 // Write the mach header and load commands out to the core file 5986 size_t bytes_written = buffer.GetString().size(); 5987 error = core_file.Write(buffer.GetString().data(), bytes_written); 5988 if (error.Success()) 5989 { 5990 // Now write the file data for all memory segments in the process 5991 for (const auto &segment : segment_load_commands) 5992 { 5993 if (core_file.SeekFromStart(segment.fileoff) == -1) 5994 { 5995 error.SetErrorStringWithFormat("unable to seek to offset 0x%" PRIx64 " in '%s'", segment.fileoff, core_file_path.c_str()); 5996 break; 5997 } 5998 5999 printf ("Saving %" PRId64 " bytes of data for memory region at 0x%" PRIx64 "\n", segment.vmsize, segment.vmaddr); 6000 addr_t bytes_left = segment.vmsize; 6001 addr_t addr = segment.vmaddr; 6002 Error memory_read_error; 6003 while (bytes_left > 0 && error.Success()) 6004 { 6005 const size_t bytes_to_read = bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6006 const size_t bytes_read = process_sp->ReadMemory(addr, bytes, bytes_to_read, memory_read_error); 6007 if (bytes_read == bytes_to_read) 6008 { 6009 size_t bytes_written = bytes_read; 6010 error = core_file.Write(bytes, bytes_written); 6011 bytes_left -= bytes_read; 6012 addr += bytes_read; 6013 } 6014 else 6015 { 6016 // Some pages within regions are not readable, those 6017 // should be zero filled 6018 memset (bytes, 0, bytes_to_read); 6019 size_t bytes_written = bytes_to_read; 6020 error = core_file.Write(bytes, bytes_written); 6021 bytes_left -= bytes_to_read; 6022 addr += bytes_to_read; 6023 } 6024 } 6025 } 6026 } 6027 } 6028 } 6029 else 6030 { 6031 error.SetErrorString("process doesn't support getting memory region info"); 6032 } 6033 } 6034 return true; // This is the right plug to handle saving core files for this process 6035 } 6036 } 6037 return false; 6038 } 6039 6040