1 //===-- ObjectFileMachO.cpp -------------------------------------*- C++ -*-===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 9 #include "llvm/ADT/StringRef.h" 10 11 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 12 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 13 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 14 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 15 #include "lldb/Core/Debugger.h" 16 #include "lldb/Core/FileSpecList.h" 17 #include "lldb/Core/Module.h" 18 #include "lldb/Core/ModuleSpec.h" 19 #include "lldb/Core/PluginManager.h" 20 #include "lldb/Core/Section.h" 21 #include "lldb/Core/StreamFile.h" 22 #include "lldb/Host/Host.h" 23 #include "lldb/Symbol/DWARFCallFrameInfo.h" 24 #include "lldb/Symbol/ObjectFile.h" 25 #include "lldb/Target/DynamicLoader.h" 26 #include "lldb/Target/MemoryRegionInfo.h" 27 #include "lldb/Target/Platform.h" 28 #include "lldb/Target/Process.h" 29 #include "lldb/Target/SectionLoadList.h" 30 #include "lldb/Target/Target.h" 31 #include "lldb/Target/Thread.h" 32 #include "lldb/Target/ThreadList.h" 33 #include "lldb/Utility/ArchSpec.h" 34 #include "lldb/Utility/DataBuffer.h" 35 #include "lldb/Utility/FileSpec.h" 36 #include "lldb/Utility/Log.h" 37 #include "lldb/Utility/RangeMap.h" 38 #include "lldb/Utility/RegisterValue.h" 39 #include "lldb/Utility/Status.h" 40 #include "lldb/Utility/StreamString.h" 41 #include "lldb/Utility/Timer.h" 42 #include "lldb/Utility/UUID.h" 43 44 #include "lldb/Host/SafeMachO.h" 45 46 #include "llvm/Support/MemoryBuffer.h" 47 48 #include "ObjectFileMachO.h" 49 50 #if defined(__APPLE__) && \ 51 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 52 // GetLLDBSharedCacheUUID() needs to call dlsym() 53 #include <dlfcn.h> 54 #endif 55 56 #ifndef __APPLE__ 57 #include "Utility/UuidCompatibility.h" 58 #else 59 #include <uuid/uuid.h> 60 #endif 61 62 #include <memory> 63 64 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 65 using namespace lldb; 66 using namespace lldb_private; 67 using namespace llvm::MachO; 68 69 // Some structure definitions needed for parsing the dyld shared cache files 70 // found on iOS devices. 71 72 struct lldb_copy_dyld_cache_header_v1 { 73 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 74 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 75 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 76 uint32_t imagesOffset; 77 uint32_t imagesCount; 78 uint64_t dyldBaseAddress; 79 uint64_t codeSignatureOffset; 80 uint64_t codeSignatureSize; 81 uint64_t slideInfoOffset; 82 uint64_t slideInfoSize; 83 uint64_t localSymbolsOffset; 84 uint64_t localSymbolsSize; 85 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 86 // and later 87 }; 88 89 struct lldb_copy_dyld_cache_mapping_info { 90 uint64_t address; 91 uint64_t size; 92 uint64_t fileOffset; 93 uint32_t maxProt; 94 uint32_t initProt; 95 }; 96 97 struct lldb_copy_dyld_cache_local_symbols_info { 98 uint32_t nlistOffset; 99 uint32_t nlistCount; 100 uint32_t stringsOffset; 101 uint32_t stringsSize; 102 uint32_t entriesOffset; 103 uint32_t entriesCount; 104 }; 105 struct lldb_copy_dyld_cache_local_symbols_entry { 106 uint32_t dylibOffset; 107 uint32_t nlistStartIndex; 108 uint32_t nlistCount; 109 }; 110 111 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 112 public: 113 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 114 const DataExtractor &data) 115 : RegisterContextDarwin_x86_64(thread, 0) { 116 SetRegisterDataFrom_LC_THREAD(data); 117 } 118 119 void InvalidateAllRegisters() override { 120 // Do nothing... registers are always valid... 121 } 122 123 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 124 lldb::offset_t offset = 0; 125 SetError(GPRRegSet, Read, -1); 126 SetError(FPURegSet, Read, -1); 127 SetError(EXCRegSet, Read, -1); 128 bool done = false; 129 130 while (!done) { 131 int flavor = data.GetU32(&offset); 132 if (flavor == 0) 133 done = true; 134 else { 135 uint32_t i; 136 uint32_t count = data.GetU32(&offset); 137 switch (flavor) { 138 case GPRRegSet: 139 for (i = 0; i < count; ++i) 140 (&gpr.rax)[i] = data.GetU64(&offset); 141 SetError(GPRRegSet, Read, 0); 142 done = true; 143 144 break; 145 case FPURegSet: 146 // TODO: fill in FPU regs.... 147 // SetError (FPURegSet, Read, -1); 148 done = true; 149 150 break; 151 case EXCRegSet: 152 exc.trapno = data.GetU32(&offset); 153 exc.err = data.GetU32(&offset); 154 exc.faultvaddr = data.GetU64(&offset); 155 SetError(EXCRegSet, Read, 0); 156 done = true; 157 break; 158 case 7: 159 case 8: 160 case 9: 161 // fancy flavors that encapsulate of the above flavors... 162 break; 163 164 default: 165 done = true; 166 break; 167 } 168 } 169 } 170 } 171 172 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 173 const char *alt_name, size_t reg_byte_size, 174 Stream &data) { 175 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 176 if (reg_info == nullptr) 177 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 178 if (reg_info) { 179 lldb_private::RegisterValue reg_value; 180 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 181 if (reg_info->byte_size >= reg_byte_size) 182 data.Write(reg_value.GetBytes(), reg_byte_size); 183 else { 184 data.Write(reg_value.GetBytes(), reg_info->byte_size); 185 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 186 ++i) 187 data.PutChar(0); 188 } 189 return reg_byte_size; 190 } 191 } 192 // Just write zeros if all else fails 193 for (size_t i = 0; i < reg_byte_size; ++i) 194 data.PutChar(0); 195 return reg_byte_size; 196 } 197 198 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 199 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 200 if (reg_ctx_sp) { 201 RegisterContext *reg_ctx = reg_ctx_sp.get(); 202 203 data.PutHex32(GPRRegSet); // Flavor 204 data.PutHex32(GPRWordCount); 205 WriteRegister(reg_ctx, "rax", nullptr, 8, data); 206 WriteRegister(reg_ctx, "rbx", nullptr, 8, data); 207 WriteRegister(reg_ctx, "rcx", nullptr, 8, data); 208 WriteRegister(reg_ctx, "rdx", nullptr, 8, data); 209 WriteRegister(reg_ctx, "rdi", nullptr, 8, data); 210 WriteRegister(reg_ctx, "rsi", nullptr, 8, data); 211 WriteRegister(reg_ctx, "rbp", nullptr, 8, data); 212 WriteRegister(reg_ctx, "rsp", nullptr, 8, data); 213 WriteRegister(reg_ctx, "r8", nullptr, 8, data); 214 WriteRegister(reg_ctx, "r9", nullptr, 8, data); 215 WriteRegister(reg_ctx, "r10", nullptr, 8, data); 216 WriteRegister(reg_ctx, "r11", nullptr, 8, data); 217 WriteRegister(reg_ctx, "r12", nullptr, 8, data); 218 WriteRegister(reg_ctx, "r13", nullptr, 8, data); 219 WriteRegister(reg_ctx, "r14", nullptr, 8, data); 220 WriteRegister(reg_ctx, "r15", nullptr, 8, data); 221 WriteRegister(reg_ctx, "rip", nullptr, 8, data); 222 WriteRegister(reg_ctx, "rflags", nullptr, 8, data); 223 WriteRegister(reg_ctx, "cs", nullptr, 8, data); 224 WriteRegister(reg_ctx, "fs", nullptr, 8, data); 225 WriteRegister(reg_ctx, "gs", nullptr, 8, data); 226 227 // // Write out the FPU registers 228 // const size_t fpu_byte_size = sizeof(FPU); 229 // size_t bytes_written = 0; 230 // data.PutHex32 (FPURegSet); 231 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 232 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 233 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 234 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 235 // data); // uint16_t fcw; // "fctrl" 236 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 237 // data); // uint16_t fsw; // "fstat" 238 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 239 // data); // uint8_t ftw; // "ftag" 240 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 241 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 242 // data); // uint16_t fop; // "fop" 243 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 244 // data); // uint32_t ip; // "fioff" 245 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 246 // data); // uint16_t cs; // "fiseg" 247 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 248 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 249 // data); // uint32_t dp; // "fooff" 250 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 251 // data); // uint16_t ds; // "foseg" 252 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 253 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 254 // data); // uint32_t mxcsr; 255 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 256 // 4, data);// uint32_t mxcsrmask; 257 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 258 // sizeof(MMSReg), data); 259 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 260 // sizeof(MMSReg), data); 261 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 262 // sizeof(MMSReg), data); 263 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 264 // sizeof(MMSReg), data); 265 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 266 // sizeof(MMSReg), data); 267 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 268 // sizeof(MMSReg), data); 269 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 270 // sizeof(MMSReg), data); 271 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 272 // sizeof(MMSReg), data); 273 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 274 // sizeof(XMMReg), data); 275 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 276 // sizeof(XMMReg), data); 277 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 278 // sizeof(XMMReg), data); 279 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 280 // sizeof(XMMReg), data); 281 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 282 // sizeof(XMMReg), data); 283 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 284 // sizeof(XMMReg), data); 285 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 286 // sizeof(XMMReg), data); 287 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 288 // sizeof(XMMReg), data); 289 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 290 // sizeof(XMMReg), data); 291 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 292 // sizeof(XMMReg), data); 293 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 294 // sizeof(XMMReg), data); 295 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 296 // sizeof(XMMReg), data); 297 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 298 // sizeof(XMMReg), data); 299 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 300 // sizeof(XMMReg), data); 301 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 302 // sizeof(XMMReg), data); 303 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 304 // sizeof(XMMReg), data); 305 // 306 // // Fill rest with zeros 307 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 308 // i) 309 // data.PutChar(0); 310 311 // Write out the EXC registers 312 data.PutHex32(EXCRegSet); 313 data.PutHex32(EXCWordCount); 314 WriteRegister(reg_ctx, "trapno", nullptr, 4, data); 315 WriteRegister(reg_ctx, "err", nullptr, 4, data); 316 WriteRegister(reg_ctx, "faultvaddr", nullptr, 8, data); 317 return true; 318 } 319 return false; 320 } 321 322 protected: 323 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 324 325 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 326 327 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 328 329 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 330 return 0; 331 } 332 333 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 334 return 0; 335 } 336 337 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 338 return 0; 339 } 340 }; 341 342 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 343 public: 344 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 345 const DataExtractor &data) 346 : RegisterContextDarwin_i386(thread, 0) { 347 SetRegisterDataFrom_LC_THREAD(data); 348 } 349 350 void InvalidateAllRegisters() override { 351 // Do nothing... registers are always valid... 352 } 353 354 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 355 lldb::offset_t offset = 0; 356 SetError(GPRRegSet, Read, -1); 357 SetError(FPURegSet, Read, -1); 358 SetError(EXCRegSet, Read, -1); 359 bool done = false; 360 361 while (!done) { 362 int flavor = data.GetU32(&offset); 363 if (flavor == 0) 364 done = true; 365 else { 366 uint32_t i; 367 uint32_t count = data.GetU32(&offset); 368 switch (flavor) { 369 case GPRRegSet: 370 for (i = 0; i < count; ++i) 371 (&gpr.eax)[i] = data.GetU32(&offset); 372 SetError(GPRRegSet, Read, 0); 373 done = true; 374 375 break; 376 case FPURegSet: 377 // TODO: fill in FPU regs.... 378 // SetError (FPURegSet, Read, -1); 379 done = true; 380 381 break; 382 case EXCRegSet: 383 exc.trapno = data.GetU32(&offset); 384 exc.err = data.GetU32(&offset); 385 exc.faultvaddr = data.GetU32(&offset); 386 SetError(EXCRegSet, Read, 0); 387 done = true; 388 break; 389 case 7: 390 case 8: 391 case 9: 392 // fancy flavors that encapsulate of the above flavors... 393 break; 394 395 default: 396 done = true; 397 break; 398 } 399 } 400 } 401 } 402 403 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 404 const char *alt_name, size_t reg_byte_size, 405 Stream &data) { 406 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 407 if (reg_info == nullptr) 408 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 409 if (reg_info) { 410 lldb_private::RegisterValue reg_value; 411 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 412 if (reg_info->byte_size >= reg_byte_size) 413 data.Write(reg_value.GetBytes(), reg_byte_size); 414 else { 415 data.Write(reg_value.GetBytes(), reg_info->byte_size); 416 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 417 ++i) 418 data.PutChar(0); 419 } 420 return reg_byte_size; 421 } 422 } 423 // Just write zeros if all else fails 424 for (size_t i = 0; i < reg_byte_size; ++i) 425 data.PutChar(0); 426 return reg_byte_size; 427 } 428 429 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 430 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 431 if (reg_ctx_sp) { 432 RegisterContext *reg_ctx = reg_ctx_sp.get(); 433 434 data.PutHex32(GPRRegSet); // Flavor 435 data.PutHex32(GPRWordCount); 436 WriteRegister(reg_ctx, "eax", nullptr, 4, data); 437 WriteRegister(reg_ctx, "ebx", nullptr, 4, data); 438 WriteRegister(reg_ctx, "ecx", nullptr, 4, data); 439 WriteRegister(reg_ctx, "edx", nullptr, 4, data); 440 WriteRegister(reg_ctx, "edi", nullptr, 4, data); 441 WriteRegister(reg_ctx, "esi", nullptr, 4, data); 442 WriteRegister(reg_ctx, "ebp", nullptr, 4, data); 443 WriteRegister(reg_ctx, "esp", nullptr, 4, data); 444 WriteRegister(reg_ctx, "ss", nullptr, 4, data); 445 WriteRegister(reg_ctx, "eflags", nullptr, 4, data); 446 WriteRegister(reg_ctx, "eip", nullptr, 4, data); 447 WriteRegister(reg_ctx, "cs", nullptr, 4, data); 448 WriteRegister(reg_ctx, "ds", nullptr, 4, data); 449 WriteRegister(reg_ctx, "es", nullptr, 4, data); 450 WriteRegister(reg_ctx, "fs", nullptr, 4, data); 451 WriteRegister(reg_ctx, "gs", nullptr, 4, data); 452 453 // Write out the EXC registers 454 data.PutHex32(EXCRegSet); 455 data.PutHex32(EXCWordCount); 456 WriteRegister(reg_ctx, "trapno", nullptr, 4, data); 457 WriteRegister(reg_ctx, "err", nullptr, 4, data); 458 WriteRegister(reg_ctx, "faultvaddr", nullptr, 4, data); 459 return true; 460 } 461 return false; 462 } 463 464 protected: 465 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 466 467 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 468 469 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 470 471 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 472 return 0; 473 } 474 475 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 476 return 0; 477 } 478 479 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 480 return 0; 481 } 482 }; 483 484 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 485 public: 486 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 487 const DataExtractor &data) 488 : RegisterContextDarwin_arm(thread, 0) { 489 SetRegisterDataFrom_LC_THREAD(data); 490 } 491 492 void InvalidateAllRegisters() override { 493 // Do nothing... registers are always valid... 494 } 495 496 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 497 lldb::offset_t offset = 0; 498 SetError(GPRRegSet, Read, -1); 499 SetError(FPURegSet, Read, -1); 500 SetError(EXCRegSet, Read, -1); 501 bool done = false; 502 503 while (!done) { 504 int flavor = data.GetU32(&offset); 505 uint32_t count = data.GetU32(&offset); 506 lldb::offset_t next_thread_state = offset + (count * 4); 507 switch (flavor) { 508 case GPRAltRegSet: 509 case GPRRegSet: 510 for (uint32_t i = 0; i < count; ++i) { 511 gpr.r[i] = data.GetU32(&offset); 512 } 513 514 // Note that gpr.cpsr is also copied by the above loop; this loop 515 // technically extends one element past the end of the gpr.r[] array. 516 517 SetError(GPRRegSet, Read, 0); 518 offset = next_thread_state; 519 break; 520 521 case FPURegSet: { 522 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 523 const int fpu_reg_buf_size = sizeof(fpu.floats); 524 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 525 fpu_reg_buf) == fpu_reg_buf_size) { 526 offset += fpu_reg_buf_size; 527 fpu.fpscr = data.GetU32(&offset); 528 SetError(FPURegSet, Read, 0); 529 } else { 530 done = true; 531 } 532 } 533 offset = next_thread_state; 534 break; 535 536 case EXCRegSet: 537 if (count == 3) { 538 exc.exception = data.GetU32(&offset); 539 exc.fsr = data.GetU32(&offset); 540 exc.far = data.GetU32(&offset); 541 SetError(EXCRegSet, Read, 0); 542 } 543 done = true; 544 offset = next_thread_state; 545 break; 546 547 // Unknown register set flavor, stop trying to parse. 548 default: 549 done = true; 550 } 551 } 552 } 553 554 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 555 const char *alt_name, size_t reg_byte_size, 556 Stream &data) { 557 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 558 if (reg_info == nullptr) 559 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 560 if (reg_info) { 561 lldb_private::RegisterValue reg_value; 562 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 563 if (reg_info->byte_size >= reg_byte_size) 564 data.Write(reg_value.GetBytes(), reg_byte_size); 565 else { 566 data.Write(reg_value.GetBytes(), reg_info->byte_size); 567 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 568 ++i) 569 data.PutChar(0); 570 } 571 return reg_byte_size; 572 } 573 } 574 // Just write zeros if all else fails 575 for (size_t i = 0; i < reg_byte_size; ++i) 576 data.PutChar(0); 577 return reg_byte_size; 578 } 579 580 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 581 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 582 if (reg_ctx_sp) { 583 RegisterContext *reg_ctx = reg_ctx_sp.get(); 584 585 data.PutHex32(GPRRegSet); // Flavor 586 data.PutHex32(GPRWordCount); 587 WriteRegister(reg_ctx, "r0", nullptr, 4, data); 588 WriteRegister(reg_ctx, "r1", nullptr, 4, data); 589 WriteRegister(reg_ctx, "r2", nullptr, 4, data); 590 WriteRegister(reg_ctx, "r3", nullptr, 4, data); 591 WriteRegister(reg_ctx, "r4", nullptr, 4, data); 592 WriteRegister(reg_ctx, "r5", nullptr, 4, data); 593 WriteRegister(reg_ctx, "r6", nullptr, 4, data); 594 WriteRegister(reg_ctx, "r7", nullptr, 4, data); 595 WriteRegister(reg_ctx, "r8", nullptr, 4, data); 596 WriteRegister(reg_ctx, "r9", nullptr, 4, data); 597 WriteRegister(reg_ctx, "r10", nullptr, 4, data); 598 WriteRegister(reg_ctx, "r11", nullptr, 4, data); 599 WriteRegister(reg_ctx, "r12", nullptr, 4, data); 600 WriteRegister(reg_ctx, "sp", nullptr, 4, data); 601 WriteRegister(reg_ctx, "lr", nullptr, 4, data); 602 WriteRegister(reg_ctx, "pc", nullptr, 4, data); 603 WriteRegister(reg_ctx, "cpsr", nullptr, 4, data); 604 605 // Write out the EXC registers 606 // data.PutHex32 (EXCRegSet); 607 // data.PutHex32 (EXCWordCount); 608 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 609 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 610 // WriteRegister (reg_ctx, "far", NULL, 4, data); 611 return true; 612 } 613 return false; 614 } 615 616 protected: 617 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 618 619 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 620 621 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 622 623 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 624 625 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 626 return 0; 627 } 628 629 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 630 return 0; 631 } 632 633 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 634 return 0; 635 } 636 637 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 638 return -1; 639 } 640 }; 641 642 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 643 public: 644 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 645 const DataExtractor &data) 646 : RegisterContextDarwin_arm64(thread, 0) { 647 SetRegisterDataFrom_LC_THREAD(data); 648 } 649 650 void InvalidateAllRegisters() override { 651 // Do nothing... registers are always valid... 652 } 653 654 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 655 lldb::offset_t offset = 0; 656 SetError(GPRRegSet, Read, -1); 657 SetError(FPURegSet, Read, -1); 658 SetError(EXCRegSet, Read, -1); 659 bool done = false; 660 while (!done) { 661 int flavor = data.GetU32(&offset); 662 uint32_t count = data.GetU32(&offset); 663 lldb::offset_t next_thread_state = offset + (count * 4); 664 switch (flavor) { 665 case GPRRegSet: 666 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 667 // 32-bit register) 668 if (count >= (33 * 2) + 1) { 669 for (uint32_t i = 0; i < 29; ++i) 670 gpr.x[i] = data.GetU64(&offset); 671 gpr.fp = data.GetU64(&offset); 672 gpr.lr = data.GetU64(&offset); 673 gpr.sp = data.GetU64(&offset); 674 gpr.pc = data.GetU64(&offset); 675 gpr.cpsr = data.GetU32(&offset); 676 SetError(GPRRegSet, Read, 0); 677 } 678 offset = next_thread_state; 679 break; 680 case FPURegSet: { 681 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 682 const int fpu_reg_buf_size = sizeof(fpu); 683 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 684 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 685 fpu_reg_buf) == fpu_reg_buf_size) { 686 SetError(FPURegSet, Read, 0); 687 } else { 688 done = true; 689 } 690 } 691 offset = next_thread_state; 692 break; 693 case EXCRegSet: 694 if (count == 4) { 695 exc.far = data.GetU64(&offset); 696 exc.esr = data.GetU32(&offset); 697 exc.exception = data.GetU32(&offset); 698 SetError(EXCRegSet, Read, 0); 699 } 700 offset = next_thread_state; 701 break; 702 default: 703 done = true; 704 break; 705 } 706 } 707 } 708 709 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 710 const char *alt_name, size_t reg_byte_size, 711 Stream &data) { 712 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 713 if (reg_info == nullptr) 714 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 715 if (reg_info) { 716 lldb_private::RegisterValue reg_value; 717 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 718 if (reg_info->byte_size >= reg_byte_size) 719 data.Write(reg_value.GetBytes(), reg_byte_size); 720 else { 721 data.Write(reg_value.GetBytes(), reg_info->byte_size); 722 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 723 ++i) 724 data.PutChar(0); 725 } 726 return reg_byte_size; 727 } 728 } 729 // Just write zeros if all else fails 730 for (size_t i = 0; i < reg_byte_size; ++i) 731 data.PutChar(0); 732 return reg_byte_size; 733 } 734 735 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 736 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 737 if (reg_ctx_sp) { 738 RegisterContext *reg_ctx = reg_ctx_sp.get(); 739 740 data.PutHex32(GPRRegSet); // Flavor 741 data.PutHex32(GPRWordCount); 742 WriteRegister(reg_ctx, "x0", nullptr, 8, data); 743 WriteRegister(reg_ctx, "x1", nullptr, 8, data); 744 WriteRegister(reg_ctx, "x2", nullptr, 8, data); 745 WriteRegister(reg_ctx, "x3", nullptr, 8, data); 746 WriteRegister(reg_ctx, "x4", nullptr, 8, data); 747 WriteRegister(reg_ctx, "x5", nullptr, 8, data); 748 WriteRegister(reg_ctx, "x6", nullptr, 8, data); 749 WriteRegister(reg_ctx, "x7", nullptr, 8, data); 750 WriteRegister(reg_ctx, "x8", nullptr, 8, data); 751 WriteRegister(reg_ctx, "x9", nullptr, 8, data); 752 WriteRegister(reg_ctx, "x10", nullptr, 8, data); 753 WriteRegister(reg_ctx, "x11", nullptr, 8, data); 754 WriteRegister(reg_ctx, "x12", nullptr, 8, data); 755 WriteRegister(reg_ctx, "x13", nullptr, 8, data); 756 WriteRegister(reg_ctx, "x14", nullptr, 8, data); 757 WriteRegister(reg_ctx, "x15", nullptr, 8, data); 758 WriteRegister(reg_ctx, "x16", nullptr, 8, data); 759 WriteRegister(reg_ctx, "x17", nullptr, 8, data); 760 WriteRegister(reg_ctx, "x18", nullptr, 8, data); 761 WriteRegister(reg_ctx, "x19", nullptr, 8, data); 762 WriteRegister(reg_ctx, "x20", nullptr, 8, data); 763 WriteRegister(reg_ctx, "x21", nullptr, 8, data); 764 WriteRegister(reg_ctx, "x22", nullptr, 8, data); 765 WriteRegister(reg_ctx, "x23", nullptr, 8, data); 766 WriteRegister(reg_ctx, "x24", nullptr, 8, data); 767 WriteRegister(reg_ctx, "x25", nullptr, 8, data); 768 WriteRegister(reg_ctx, "x26", nullptr, 8, data); 769 WriteRegister(reg_ctx, "x27", nullptr, 8, data); 770 WriteRegister(reg_ctx, "x28", nullptr, 8, data); 771 WriteRegister(reg_ctx, "fp", nullptr, 8, data); 772 WriteRegister(reg_ctx, "lr", nullptr, 8, data); 773 WriteRegister(reg_ctx, "sp", nullptr, 8, data); 774 WriteRegister(reg_ctx, "pc", nullptr, 8, data); 775 WriteRegister(reg_ctx, "cpsr", nullptr, 4, data); 776 777 // Write out the EXC registers 778 // data.PutHex32 (EXCRegSet); 779 // data.PutHex32 (EXCWordCount); 780 // WriteRegister (reg_ctx, "far", NULL, 8, data); 781 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 782 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 783 return true; 784 } 785 return false; 786 } 787 788 protected: 789 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 790 791 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 792 793 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 794 795 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 796 797 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 798 return 0; 799 } 800 801 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 802 return 0; 803 } 804 805 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 806 return 0; 807 } 808 809 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 810 return -1; 811 } 812 }; 813 814 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 815 switch (magic) { 816 case MH_MAGIC: 817 case MH_CIGAM: 818 return sizeof(struct mach_header); 819 820 case MH_MAGIC_64: 821 case MH_CIGAM_64: 822 return sizeof(struct mach_header_64); 823 break; 824 825 default: 826 break; 827 } 828 return 0; 829 } 830 831 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 832 833 void ObjectFileMachO::Initialize() { 834 PluginManager::RegisterPlugin( 835 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 836 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 837 } 838 839 void ObjectFileMachO::Terminate() { 840 PluginManager::UnregisterPlugin(CreateInstance); 841 } 842 843 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 844 static ConstString g_name("mach-o"); 845 return g_name; 846 } 847 848 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 849 return "Mach-o object file reader (32 and 64 bit)"; 850 } 851 852 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 853 DataBufferSP &data_sp, 854 lldb::offset_t data_offset, 855 const FileSpec *file, 856 lldb::offset_t file_offset, 857 lldb::offset_t length) { 858 if (!data_sp) { 859 data_sp = MapFileData(*file, length, file_offset); 860 if (!data_sp) 861 return nullptr; 862 data_offset = 0; 863 } 864 865 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 866 return nullptr; 867 868 // Update the data to contain the entire file if it doesn't already 869 if (data_sp->GetByteSize() < length) { 870 data_sp = MapFileData(*file, length, file_offset); 871 if (!data_sp) 872 return nullptr; 873 data_offset = 0; 874 } 875 auto objfile_up = llvm::make_unique<ObjectFileMachO>( 876 module_sp, data_sp, data_offset, file, file_offset, length); 877 if (!objfile_up || !objfile_up->ParseHeader()) 878 return nullptr; 879 880 return objfile_up.release(); 881 } 882 883 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 884 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 885 const ProcessSP &process_sp, lldb::addr_t header_addr) { 886 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 887 std::unique_ptr<ObjectFile> objfile_up( 888 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 889 if (objfile_up.get() && objfile_up->ParseHeader()) 890 return objfile_up.release(); 891 } 892 return nullptr; 893 } 894 895 size_t ObjectFileMachO::GetModuleSpecifications( 896 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 897 lldb::offset_t data_offset, lldb::offset_t file_offset, 898 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 899 const size_t initial_count = specs.GetSize(); 900 901 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 902 DataExtractor data; 903 data.SetData(data_sp); 904 llvm::MachO::mach_header header; 905 if (ParseHeader(data, &data_offset, header)) { 906 size_t header_and_load_cmds = 907 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 908 if (header_and_load_cmds >= data_sp->GetByteSize()) { 909 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 910 data.SetData(data_sp); 911 data_offset = MachHeaderSizeFromMagic(header.magic); 912 } 913 if (data_sp) { 914 ModuleSpec spec; 915 spec.GetFileSpec() = file; 916 spec.SetObjectOffset(file_offset); 917 spec.SetObjectSize(length); 918 919 spec.GetArchitecture() = GetArchitecture(header, data, data_offset); 920 if (spec.GetArchitecture().IsValid()) { 921 spec.GetUUID() = GetUUID(header, data, data_offset); 922 specs.Append(spec); 923 } 924 } 925 } 926 } 927 return specs.GetSize() - initial_count; 928 } 929 930 ConstString ObjectFileMachO::GetSegmentNameTEXT() { 931 static ConstString g_segment_name_TEXT("__TEXT"); 932 return g_segment_name_TEXT; 933 } 934 935 ConstString ObjectFileMachO::GetSegmentNameDATA() { 936 static ConstString g_segment_name_DATA("__DATA"); 937 return g_segment_name_DATA; 938 } 939 940 ConstString ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 941 static ConstString g_segment_name("__DATA_DIRTY"); 942 return g_segment_name; 943 } 944 945 ConstString ObjectFileMachO::GetSegmentNameDATA_CONST() { 946 static ConstString g_segment_name("__DATA_CONST"); 947 return g_segment_name; 948 } 949 950 ConstString ObjectFileMachO::GetSegmentNameOBJC() { 951 static ConstString g_segment_name_OBJC("__OBJC"); 952 return g_segment_name_OBJC; 953 } 954 955 ConstString ObjectFileMachO::GetSegmentNameLINKEDIT() { 956 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 957 return g_section_name_LINKEDIT; 958 } 959 960 ConstString ObjectFileMachO::GetSegmentNameDWARF() { 961 static ConstString g_section_name("__DWARF"); 962 return g_section_name; 963 } 964 965 ConstString ObjectFileMachO::GetSectionNameEHFrame() { 966 static ConstString g_section_name_eh_frame("__eh_frame"); 967 return g_section_name_eh_frame; 968 } 969 970 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 971 lldb::addr_t data_offset, 972 lldb::addr_t data_length) { 973 DataExtractor data; 974 data.SetData(data_sp, data_offset, data_length); 975 lldb::offset_t offset = 0; 976 uint32_t magic = data.GetU32(&offset); 977 return MachHeaderSizeFromMagic(magic) != 0; 978 } 979 980 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 981 DataBufferSP &data_sp, 982 lldb::offset_t data_offset, 983 const FileSpec *file, 984 lldb::offset_t file_offset, 985 lldb::offset_t length) 986 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 987 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 988 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 989 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 990 ::memset(&m_header, 0, sizeof(m_header)); 991 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 992 } 993 994 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 995 lldb::DataBufferSP &header_data_sp, 996 const lldb::ProcessSP &process_sp, 997 lldb::addr_t header_addr) 998 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 999 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 1000 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 1001 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 1002 ::memset(&m_header, 0, sizeof(m_header)); 1003 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1004 } 1005 1006 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 1007 lldb::offset_t *data_offset_ptr, 1008 llvm::MachO::mach_header &header) { 1009 data.SetByteOrder(endian::InlHostByteOrder()); 1010 // Leave magic in the original byte order 1011 header.magic = data.GetU32(data_offset_ptr); 1012 bool can_parse = false; 1013 bool is_64_bit = false; 1014 switch (header.magic) { 1015 case MH_MAGIC: 1016 data.SetByteOrder(endian::InlHostByteOrder()); 1017 data.SetAddressByteSize(4); 1018 can_parse = true; 1019 break; 1020 1021 case MH_MAGIC_64: 1022 data.SetByteOrder(endian::InlHostByteOrder()); 1023 data.SetAddressByteSize(8); 1024 can_parse = true; 1025 is_64_bit = true; 1026 break; 1027 1028 case MH_CIGAM: 1029 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1030 ? eByteOrderLittle 1031 : eByteOrderBig); 1032 data.SetAddressByteSize(4); 1033 can_parse = true; 1034 break; 1035 1036 case MH_CIGAM_64: 1037 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1038 ? eByteOrderLittle 1039 : eByteOrderBig); 1040 data.SetAddressByteSize(8); 1041 is_64_bit = true; 1042 can_parse = true; 1043 break; 1044 1045 default: 1046 break; 1047 } 1048 1049 if (can_parse) { 1050 data.GetU32(data_offset_ptr, &header.cputype, 6); 1051 if (is_64_bit) 1052 *data_offset_ptr += 4; 1053 return true; 1054 } else { 1055 memset(&header, 0, sizeof(header)); 1056 } 1057 return false; 1058 } 1059 1060 bool ObjectFileMachO::ParseHeader() { 1061 ModuleSP module_sp(GetModule()); 1062 if (module_sp) { 1063 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1064 bool can_parse = false; 1065 lldb::offset_t offset = 0; 1066 m_data.SetByteOrder(endian::InlHostByteOrder()); 1067 // Leave magic in the original byte order 1068 m_header.magic = m_data.GetU32(&offset); 1069 switch (m_header.magic) { 1070 case MH_MAGIC: 1071 m_data.SetByteOrder(endian::InlHostByteOrder()); 1072 m_data.SetAddressByteSize(4); 1073 can_parse = true; 1074 break; 1075 1076 case MH_MAGIC_64: 1077 m_data.SetByteOrder(endian::InlHostByteOrder()); 1078 m_data.SetAddressByteSize(8); 1079 can_parse = true; 1080 break; 1081 1082 case MH_CIGAM: 1083 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1084 ? eByteOrderLittle 1085 : eByteOrderBig); 1086 m_data.SetAddressByteSize(4); 1087 can_parse = true; 1088 break; 1089 1090 case MH_CIGAM_64: 1091 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1092 ? eByteOrderLittle 1093 : eByteOrderBig); 1094 m_data.SetAddressByteSize(8); 1095 can_parse = true; 1096 break; 1097 1098 default: 1099 break; 1100 } 1101 1102 if (can_parse) { 1103 m_data.GetU32(&offset, &m_header.cputype, 6); 1104 1105 if (ArchSpec mach_arch = GetArchitecture()) { 1106 // Check if the module has a required architecture 1107 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1108 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1109 return false; 1110 1111 if (SetModulesArchitecture(mach_arch)) { 1112 const size_t header_and_lc_size = 1113 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1114 if (m_data.GetByteSize() < header_and_lc_size) { 1115 DataBufferSP data_sp; 1116 ProcessSP process_sp(m_process_wp.lock()); 1117 if (process_sp) { 1118 data_sp = 1119 ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1120 } else { 1121 // Read in all only the load command data from the file on disk 1122 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset); 1123 if (data_sp->GetByteSize() != header_and_lc_size) 1124 return false; 1125 } 1126 if (data_sp) 1127 m_data.SetData(data_sp); 1128 } 1129 } 1130 return true; 1131 } 1132 } else { 1133 memset(&m_header, 0, sizeof(struct mach_header)); 1134 } 1135 } 1136 return false; 1137 } 1138 1139 ByteOrder ObjectFileMachO::GetByteOrder() const { 1140 return m_data.GetByteOrder(); 1141 } 1142 1143 bool ObjectFileMachO::IsExecutable() const { 1144 return m_header.filetype == MH_EXECUTE; 1145 } 1146 1147 bool ObjectFileMachO::IsDynamicLoader() const { 1148 return m_header.filetype == MH_DYLINKER; 1149 } 1150 1151 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1152 return m_data.GetAddressByteSize(); 1153 } 1154 1155 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1156 Symtab *symtab = GetSymtab(); 1157 if (symtab) { 1158 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1159 if (symbol) { 1160 if (symbol->ValueIsAddress()) { 1161 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1162 if (section_sp) { 1163 const lldb::SectionType section_type = section_sp->GetType(); 1164 switch (section_type) { 1165 case eSectionTypeInvalid: 1166 return AddressClass::eUnknown; 1167 1168 case eSectionTypeCode: 1169 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1170 // For ARM we have a bit in the n_desc field of the symbol that 1171 // tells us ARM/Thumb which is bit 0x0008. 1172 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1173 return AddressClass::eCodeAlternateISA; 1174 } 1175 return AddressClass::eCode; 1176 1177 case eSectionTypeContainer: 1178 return AddressClass::eUnknown; 1179 1180 case eSectionTypeData: 1181 case eSectionTypeDataCString: 1182 case eSectionTypeDataCStringPointers: 1183 case eSectionTypeDataSymbolAddress: 1184 case eSectionTypeData4: 1185 case eSectionTypeData8: 1186 case eSectionTypeData16: 1187 case eSectionTypeDataPointers: 1188 case eSectionTypeZeroFill: 1189 case eSectionTypeDataObjCMessageRefs: 1190 case eSectionTypeDataObjCCFStrings: 1191 case eSectionTypeGoSymtab: 1192 return AddressClass::eData; 1193 1194 case eSectionTypeDebug: 1195 case eSectionTypeDWARFDebugAbbrev: 1196 case eSectionTypeDWARFDebugAbbrevDwo: 1197 case eSectionTypeDWARFDebugAddr: 1198 case eSectionTypeDWARFDebugAranges: 1199 case eSectionTypeDWARFDebugCuIndex: 1200 case eSectionTypeDWARFDebugFrame: 1201 case eSectionTypeDWARFDebugInfo: 1202 case eSectionTypeDWARFDebugInfoDwo: 1203 case eSectionTypeDWARFDebugLine: 1204 case eSectionTypeDWARFDebugLineStr: 1205 case eSectionTypeDWARFDebugLoc: 1206 case eSectionTypeDWARFDebugLocLists: 1207 case eSectionTypeDWARFDebugMacInfo: 1208 case eSectionTypeDWARFDebugMacro: 1209 case eSectionTypeDWARFDebugNames: 1210 case eSectionTypeDWARFDebugPubNames: 1211 case eSectionTypeDWARFDebugPubTypes: 1212 case eSectionTypeDWARFDebugRanges: 1213 case eSectionTypeDWARFDebugRngLists: 1214 case eSectionTypeDWARFDebugStr: 1215 case eSectionTypeDWARFDebugStrDwo: 1216 case eSectionTypeDWARFDebugStrOffsets: 1217 case eSectionTypeDWARFDebugStrOffsetsDwo: 1218 case eSectionTypeDWARFDebugTypes: 1219 case eSectionTypeDWARFDebugTypesDwo: 1220 case eSectionTypeDWARFAppleNames: 1221 case eSectionTypeDWARFAppleTypes: 1222 case eSectionTypeDWARFAppleNamespaces: 1223 case eSectionTypeDWARFAppleObjC: 1224 case eSectionTypeDWARFGNUDebugAltLink: 1225 return AddressClass::eDebug; 1226 1227 case eSectionTypeEHFrame: 1228 case eSectionTypeARMexidx: 1229 case eSectionTypeARMextab: 1230 case eSectionTypeCompactUnwind: 1231 return AddressClass::eRuntime; 1232 1233 case eSectionTypeAbsoluteAddress: 1234 case eSectionTypeELFSymbolTable: 1235 case eSectionTypeELFDynamicSymbols: 1236 case eSectionTypeELFRelocationEntries: 1237 case eSectionTypeELFDynamicLinkInfo: 1238 case eSectionTypeOther: 1239 return AddressClass::eUnknown; 1240 } 1241 } 1242 } 1243 1244 const SymbolType symbol_type = symbol->GetType(); 1245 switch (symbol_type) { 1246 case eSymbolTypeAny: 1247 return AddressClass::eUnknown; 1248 case eSymbolTypeAbsolute: 1249 return AddressClass::eUnknown; 1250 1251 case eSymbolTypeCode: 1252 case eSymbolTypeTrampoline: 1253 case eSymbolTypeResolver: 1254 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1255 // For ARM we have a bit in the n_desc field of the symbol that tells 1256 // us ARM/Thumb which is bit 0x0008. 1257 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1258 return AddressClass::eCodeAlternateISA; 1259 } 1260 return AddressClass::eCode; 1261 1262 case eSymbolTypeData: 1263 return AddressClass::eData; 1264 case eSymbolTypeRuntime: 1265 return AddressClass::eRuntime; 1266 case eSymbolTypeException: 1267 return AddressClass::eRuntime; 1268 case eSymbolTypeSourceFile: 1269 return AddressClass::eDebug; 1270 case eSymbolTypeHeaderFile: 1271 return AddressClass::eDebug; 1272 case eSymbolTypeObjectFile: 1273 return AddressClass::eDebug; 1274 case eSymbolTypeCommonBlock: 1275 return AddressClass::eDebug; 1276 case eSymbolTypeBlock: 1277 return AddressClass::eDebug; 1278 case eSymbolTypeLocal: 1279 return AddressClass::eData; 1280 case eSymbolTypeParam: 1281 return AddressClass::eData; 1282 case eSymbolTypeVariable: 1283 return AddressClass::eData; 1284 case eSymbolTypeVariableType: 1285 return AddressClass::eDebug; 1286 case eSymbolTypeLineEntry: 1287 return AddressClass::eDebug; 1288 case eSymbolTypeLineHeader: 1289 return AddressClass::eDebug; 1290 case eSymbolTypeScopeBegin: 1291 return AddressClass::eDebug; 1292 case eSymbolTypeScopeEnd: 1293 return AddressClass::eDebug; 1294 case eSymbolTypeAdditional: 1295 return AddressClass::eUnknown; 1296 case eSymbolTypeCompiler: 1297 return AddressClass::eDebug; 1298 case eSymbolTypeInstrumentation: 1299 return AddressClass::eDebug; 1300 case eSymbolTypeUndefined: 1301 return AddressClass::eUnknown; 1302 case eSymbolTypeObjCClass: 1303 return AddressClass::eRuntime; 1304 case eSymbolTypeObjCMetaClass: 1305 return AddressClass::eRuntime; 1306 case eSymbolTypeObjCIVar: 1307 return AddressClass::eRuntime; 1308 case eSymbolTypeReExported: 1309 return AddressClass::eRuntime; 1310 } 1311 } 1312 } 1313 return AddressClass::eUnknown; 1314 } 1315 1316 Symtab *ObjectFileMachO::GetSymtab() { 1317 ModuleSP module_sp(GetModule()); 1318 if (module_sp) { 1319 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1320 if (m_symtab_up == nullptr) { 1321 m_symtab_up.reset(new Symtab(this)); 1322 std::lock_guard<std::recursive_mutex> symtab_guard( 1323 m_symtab_up->GetMutex()); 1324 ParseSymtab(); 1325 m_symtab_up->Finalize(); 1326 } 1327 } 1328 return m_symtab_up.get(); 1329 } 1330 1331 bool ObjectFileMachO::IsStripped() { 1332 if (m_dysymtab.cmd == 0) { 1333 ModuleSP module_sp(GetModule()); 1334 if (module_sp) { 1335 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1336 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1337 const lldb::offset_t load_cmd_offset = offset; 1338 1339 load_command lc; 1340 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 1341 break; 1342 if (lc.cmd == LC_DYSYMTAB) { 1343 m_dysymtab.cmd = lc.cmd; 1344 m_dysymtab.cmdsize = lc.cmdsize; 1345 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1346 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1347 nullptr) { 1348 // Clear m_dysymtab if we were unable to read all items from the 1349 // load command 1350 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1351 } 1352 } 1353 offset = load_cmd_offset + lc.cmdsize; 1354 } 1355 } 1356 } 1357 if (m_dysymtab.cmd) 1358 return m_dysymtab.nlocalsym <= 1; 1359 return false; 1360 } 1361 1362 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1363 EncryptedFileRanges result; 1364 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1365 1366 encryption_info_command encryption_cmd; 1367 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1368 const lldb::offset_t load_cmd_offset = offset; 1369 if (m_data.GetU32(&offset, &encryption_cmd, 2) == nullptr) 1370 break; 1371 1372 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the 1373 // 3 fields we care about, so treat them the same. 1374 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1375 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1376 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1377 if (encryption_cmd.cryptid != 0) { 1378 EncryptedFileRanges::Entry entry; 1379 entry.SetRangeBase(encryption_cmd.cryptoff); 1380 entry.SetByteSize(encryption_cmd.cryptsize); 1381 result.Append(entry); 1382 } 1383 } 1384 } 1385 offset = load_cmd_offset + encryption_cmd.cmdsize; 1386 } 1387 1388 return result; 1389 } 1390 1391 void ObjectFileMachO::SanitizeSegmentCommand(segment_command_64 &seg_cmd, 1392 uint32_t cmd_idx) { 1393 if (m_length == 0 || seg_cmd.filesize == 0) 1394 return; 1395 1396 if (seg_cmd.fileoff > m_length) { 1397 // We have a load command that says it extends past the end of the file. 1398 // This is likely a corrupt file. We don't have any way to return an error 1399 // condition here (this method was likely invoked from something like 1400 // ObjectFile::GetSectionList()), so we just null out the section contents, 1401 // and dump a message to stdout. The most common case here is core file 1402 // debugging with a truncated file. 1403 const char *lc_segment_name = 1404 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1405 GetModule()->ReportWarning( 1406 "load command %u %s has a fileoff (0x%" PRIx64 1407 ") that extends beyond the end of the file (0x%" PRIx64 1408 "), ignoring this section", 1409 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1410 1411 seg_cmd.fileoff = 0; 1412 seg_cmd.filesize = 0; 1413 } 1414 1415 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1416 // We have a load command that says it extends past the end of the file. 1417 // This is likely a corrupt file. We don't have any way to return an error 1418 // condition here (this method was likely invoked from something like 1419 // ObjectFile::GetSectionList()), so we just null out the section contents, 1420 // and dump a message to stdout. The most common case here is core file 1421 // debugging with a truncated file. 1422 const char *lc_segment_name = 1423 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1424 GetModule()->ReportWarning( 1425 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1426 ") that extends beyond the end of the file (0x%" PRIx64 1427 "), the segment will be truncated to match", 1428 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1429 1430 // Truncate the length 1431 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1432 } 1433 } 1434 1435 static uint32_t GetSegmentPermissions(const segment_command_64 &seg_cmd) { 1436 uint32_t result = 0; 1437 if (seg_cmd.initprot & VM_PROT_READ) 1438 result |= ePermissionsReadable; 1439 if (seg_cmd.initprot & VM_PROT_WRITE) 1440 result |= ePermissionsWritable; 1441 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1442 result |= ePermissionsExecutable; 1443 return result; 1444 } 1445 1446 static lldb::SectionType GetSectionType(uint32_t flags, 1447 ConstString section_name) { 1448 1449 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1450 return eSectionTypeCode; 1451 1452 uint32_t mach_sect_type = flags & SECTION_TYPE; 1453 static ConstString g_sect_name_objc_data("__objc_data"); 1454 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1455 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1456 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1457 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1458 static ConstString g_sect_name_objc_const("__objc_const"); 1459 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1460 static ConstString g_sect_name_cfstring("__cfstring"); 1461 1462 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1463 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1464 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1465 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1466 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1467 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1468 static ConstString g_sect_name_dwarf_debug_loclists("__debug_loclists"); 1469 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1470 static ConstString g_sect_name_dwarf_debug_names("__debug_names"); 1471 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1472 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1473 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1474 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1475 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1476 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1477 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1478 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1479 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1480 static ConstString g_sect_name_eh_frame("__eh_frame"); 1481 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1482 static ConstString g_sect_name_text("__text"); 1483 static ConstString g_sect_name_data("__data"); 1484 static ConstString g_sect_name_go_symtab("__gosymtab"); 1485 1486 if (section_name == g_sect_name_dwarf_debug_abbrev) 1487 return eSectionTypeDWARFDebugAbbrev; 1488 if (section_name == g_sect_name_dwarf_debug_aranges) 1489 return eSectionTypeDWARFDebugAranges; 1490 if (section_name == g_sect_name_dwarf_debug_frame) 1491 return eSectionTypeDWARFDebugFrame; 1492 if (section_name == g_sect_name_dwarf_debug_info) 1493 return eSectionTypeDWARFDebugInfo; 1494 if (section_name == g_sect_name_dwarf_debug_line) 1495 return eSectionTypeDWARFDebugLine; 1496 if (section_name == g_sect_name_dwarf_debug_loc) 1497 return eSectionTypeDWARFDebugLoc; 1498 if (section_name == g_sect_name_dwarf_debug_loclists) 1499 return eSectionTypeDWARFDebugLocLists; 1500 if (section_name == g_sect_name_dwarf_debug_macinfo) 1501 return eSectionTypeDWARFDebugMacInfo; 1502 if (section_name == g_sect_name_dwarf_debug_names) 1503 return eSectionTypeDWARFDebugNames; 1504 if (section_name == g_sect_name_dwarf_debug_pubnames) 1505 return eSectionTypeDWARFDebugPubNames; 1506 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1507 return eSectionTypeDWARFDebugPubTypes; 1508 if (section_name == g_sect_name_dwarf_debug_ranges) 1509 return eSectionTypeDWARFDebugRanges; 1510 if (section_name == g_sect_name_dwarf_debug_str) 1511 return eSectionTypeDWARFDebugStr; 1512 if (section_name == g_sect_name_dwarf_debug_types) 1513 return eSectionTypeDWARFDebugTypes; 1514 if (section_name == g_sect_name_dwarf_apple_names) 1515 return eSectionTypeDWARFAppleNames; 1516 if (section_name == g_sect_name_dwarf_apple_types) 1517 return eSectionTypeDWARFAppleTypes; 1518 if (section_name == g_sect_name_dwarf_apple_namespaces) 1519 return eSectionTypeDWARFAppleNamespaces; 1520 if (section_name == g_sect_name_dwarf_apple_objc) 1521 return eSectionTypeDWARFAppleObjC; 1522 if (section_name == g_sect_name_objc_selrefs) 1523 return eSectionTypeDataCStringPointers; 1524 if (section_name == g_sect_name_objc_msgrefs) 1525 return eSectionTypeDataObjCMessageRefs; 1526 if (section_name == g_sect_name_eh_frame) 1527 return eSectionTypeEHFrame; 1528 if (section_name == g_sect_name_compact_unwind) 1529 return eSectionTypeCompactUnwind; 1530 if (section_name == g_sect_name_cfstring) 1531 return eSectionTypeDataObjCCFStrings; 1532 if (section_name == g_sect_name_go_symtab) 1533 return eSectionTypeGoSymtab; 1534 if (section_name == g_sect_name_objc_data || 1535 section_name == g_sect_name_objc_classrefs || 1536 section_name == g_sect_name_objc_superrefs || 1537 section_name == g_sect_name_objc_const || 1538 section_name == g_sect_name_objc_classlist) { 1539 return eSectionTypeDataPointers; 1540 } 1541 1542 switch (mach_sect_type) { 1543 // TODO: categorize sections by other flags for regular sections 1544 case S_REGULAR: 1545 if (section_name == g_sect_name_text) 1546 return eSectionTypeCode; 1547 if (section_name == g_sect_name_data) 1548 return eSectionTypeData; 1549 return eSectionTypeOther; 1550 case S_ZEROFILL: 1551 return eSectionTypeZeroFill; 1552 case S_CSTRING_LITERALS: // section with only literal C strings 1553 return eSectionTypeDataCString; 1554 case S_4BYTE_LITERALS: // section with only 4 byte literals 1555 return eSectionTypeData4; 1556 case S_8BYTE_LITERALS: // section with only 8 byte literals 1557 return eSectionTypeData8; 1558 case S_LITERAL_POINTERS: // section with only pointers to literals 1559 return eSectionTypeDataPointers; 1560 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1561 return eSectionTypeDataPointers; 1562 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1563 return eSectionTypeDataPointers; 1564 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1565 // the reserved2 field 1566 return eSectionTypeCode; 1567 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1568 // initialization 1569 return eSectionTypeDataPointers; 1570 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1571 // termination 1572 return eSectionTypeDataPointers; 1573 case S_COALESCED: 1574 return eSectionTypeOther; 1575 case S_GB_ZEROFILL: 1576 return eSectionTypeZeroFill; 1577 case S_INTERPOSING: // section with only pairs of function pointers for 1578 // interposing 1579 return eSectionTypeCode; 1580 case S_16BYTE_LITERALS: // section with only 16 byte literals 1581 return eSectionTypeData16; 1582 case S_DTRACE_DOF: 1583 return eSectionTypeDebug; 1584 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1585 return eSectionTypeDataPointers; 1586 default: 1587 return eSectionTypeOther; 1588 } 1589 } 1590 1591 struct ObjectFileMachO::SegmentParsingContext { 1592 const EncryptedFileRanges EncryptedRanges; 1593 lldb_private::SectionList &UnifiedList; 1594 uint32_t NextSegmentIdx = 0; 1595 uint32_t NextSectionIdx = 0; 1596 bool FileAddressesChanged = false; 1597 1598 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1599 lldb_private::SectionList &UnifiedList) 1600 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1601 }; 1602 1603 void ObjectFileMachO::ProcessSegmentCommand(const load_command &load_cmd_, 1604 lldb::offset_t offset, 1605 uint32_t cmd_idx, 1606 SegmentParsingContext &context) { 1607 segment_command_64 load_cmd; 1608 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1609 1610 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1611 return; 1612 1613 ModuleSP module_sp = GetModule(); 1614 const bool is_core = GetType() == eTypeCoreFile; 1615 const bool is_dsym = (m_header.filetype == MH_DSYM); 1616 bool add_section = true; 1617 bool add_to_unified = true; 1618 ConstString const_segname( 1619 load_cmd.segname, strnlen(load_cmd.segname, sizeof(load_cmd.segname))); 1620 1621 SectionSP unified_section_sp( 1622 context.UnifiedList.FindSectionByName(const_segname)); 1623 if (is_dsym && unified_section_sp) { 1624 if (const_segname == GetSegmentNameLINKEDIT()) { 1625 // We need to keep the __LINKEDIT segment private to this object file 1626 // only 1627 add_to_unified = false; 1628 } else { 1629 // This is the dSYM file and this section has already been created by the 1630 // object file, no need to create it. 1631 add_section = false; 1632 } 1633 } 1634 load_cmd.vmaddr = m_data.GetAddress(&offset); 1635 load_cmd.vmsize = m_data.GetAddress(&offset); 1636 load_cmd.fileoff = m_data.GetAddress(&offset); 1637 load_cmd.filesize = m_data.GetAddress(&offset); 1638 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1639 return; 1640 1641 SanitizeSegmentCommand(load_cmd, cmd_idx); 1642 1643 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1644 const bool segment_is_encrypted = 1645 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1646 1647 // Keep a list of mach segments around in case we need to get at data that 1648 // isn't stored in the abstracted Sections. 1649 m_mach_segments.push_back(load_cmd); 1650 1651 // Use a segment ID of the segment index shifted left by 8 so they never 1652 // conflict with any of the sections. 1653 SectionSP segment_sp; 1654 if (add_section && (const_segname || is_core)) { 1655 segment_sp = std::make_shared<Section>( 1656 module_sp, // Module to which this section belongs 1657 this, // Object file to which this sections belongs 1658 ++context.NextSegmentIdx 1659 << 8, // Section ID is the 1 based segment index 1660 // shifted right by 8 bits as not to collide with any of the 256 1661 // section IDs that are possible 1662 const_segname, // Name of this section 1663 eSectionTypeContainer, // This section is a container of other 1664 // sections. 1665 load_cmd.vmaddr, // File VM address == addresses as they are 1666 // found in the object file 1667 load_cmd.vmsize, // VM size in bytes of this section 1668 load_cmd.fileoff, // Offset to the data for this section in 1669 // the file 1670 load_cmd.filesize, // Size in bytes of this section as found 1671 // in the file 1672 0, // Segments have no alignment information 1673 load_cmd.flags); // Flags for this section 1674 1675 segment_sp->SetIsEncrypted(segment_is_encrypted); 1676 m_sections_up->AddSection(segment_sp); 1677 segment_sp->SetPermissions(segment_permissions); 1678 if (add_to_unified) 1679 context.UnifiedList.AddSection(segment_sp); 1680 } else if (unified_section_sp) { 1681 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1682 // Check to see if the module was read from memory? 1683 if (module_sp->GetObjectFile()->GetBaseAddress().IsValid()) { 1684 // We have a module that is in memory and needs to have its file 1685 // address adjusted. We need to do this because when we load a file 1686 // from memory, its addresses will be slid already, yet the addresses 1687 // in the new symbol file will still be unslid. Since everything is 1688 // stored as section offset, this shouldn't cause any problems. 1689 1690 // Make sure we've parsed the symbol table from the ObjectFile before 1691 // we go around changing its Sections. 1692 module_sp->GetObjectFile()->GetSymtab(); 1693 // eh_frame would present the same problems but we parse that on a per- 1694 // function basis as-needed so it's more difficult to remove its use of 1695 // the Sections. Realistically, the environments where this code path 1696 // will be taken will not have eh_frame sections. 1697 1698 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1699 1700 // Notify the module that the section addresses have been changed once 1701 // we're done so any file-address caches can be updated. 1702 context.FileAddressesChanged = true; 1703 } 1704 } 1705 m_sections_up->AddSection(unified_section_sp); 1706 } 1707 1708 struct section_64 sect64; 1709 ::memset(§64, 0, sizeof(sect64)); 1710 // Push a section into our mach sections for the section at index zero 1711 // (NO_SECT) if we don't have any mach sections yet... 1712 if (m_mach_sections.empty()) 1713 m_mach_sections.push_back(sect64); 1714 uint32_t segment_sect_idx; 1715 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1716 1717 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1718 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1719 ++segment_sect_idx) { 1720 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1721 sizeof(sect64.sectname)) == nullptr) 1722 break; 1723 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1724 sizeof(sect64.segname)) == nullptr) 1725 break; 1726 sect64.addr = m_data.GetAddress(&offset); 1727 sect64.size = m_data.GetAddress(&offset); 1728 1729 if (m_data.GetU32(&offset, §64.offset, num_u32s) == nullptr) 1730 break; 1731 1732 // Keep a list of mach sections around in case we need to get at data that 1733 // isn't stored in the abstracted Sections. 1734 m_mach_sections.push_back(sect64); 1735 1736 if (add_section) { 1737 ConstString section_name( 1738 sect64.sectname, strnlen(sect64.sectname, sizeof(sect64.sectname))); 1739 if (!const_segname) { 1740 // We have a segment with no name so we need to conjure up segments 1741 // that correspond to the section's segname if there isn't already such 1742 // a section. If there is such a section, we resize the section so that 1743 // it spans all sections. We also mark these sections as fake so 1744 // address matches don't hit if they land in the gaps between the child 1745 // sections. 1746 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1747 sizeof(sect64.segname)); 1748 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1749 if (segment_sp.get()) { 1750 Section *segment = segment_sp.get(); 1751 // Grow the section size as needed. 1752 const lldb::addr_t sect64_min_addr = sect64.addr; 1753 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1754 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1755 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1756 const lldb::addr_t curr_seg_max_addr = 1757 curr_seg_min_addr + curr_seg_byte_size; 1758 if (sect64_min_addr >= curr_seg_min_addr) { 1759 const lldb::addr_t new_seg_byte_size = 1760 sect64_max_addr - curr_seg_min_addr; 1761 // Only grow the section size if needed 1762 if (new_seg_byte_size > curr_seg_byte_size) 1763 segment->SetByteSize(new_seg_byte_size); 1764 } else { 1765 // We need to change the base address of the segment and adjust the 1766 // child section offsets for all existing children. 1767 const lldb::addr_t slide_amount = 1768 sect64_min_addr - curr_seg_min_addr; 1769 segment->Slide(slide_amount, false); 1770 segment->GetChildren().Slide(-slide_amount, false); 1771 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1772 } 1773 1774 // Grow the section size as needed. 1775 if (sect64.offset) { 1776 const lldb::addr_t segment_min_file_offset = 1777 segment->GetFileOffset(); 1778 const lldb::addr_t segment_max_file_offset = 1779 segment_min_file_offset + segment->GetFileSize(); 1780 1781 const lldb::addr_t section_min_file_offset = sect64.offset; 1782 const lldb::addr_t section_max_file_offset = 1783 section_min_file_offset + sect64.size; 1784 const lldb::addr_t new_file_offset = 1785 std::min(section_min_file_offset, segment_min_file_offset); 1786 const lldb::addr_t new_file_size = 1787 std::max(section_max_file_offset, segment_max_file_offset) - 1788 new_file_offset; 1789 segment->SetFileOffset(new_file_offset); 1790 segment->SetFileSize(new_file_size); 1791 } 1792 } else { 1793 // Create a fake section for the section's named segment 1794 segment_sp = std::make_shared<Section>( 1795 segment_sp, // Parent section 1796 module_sp, // Module to which this section belongs 1797 this, // Object file to which this section belongs 1798 ++context.NextSegmentIdx 1799 << 8, // Section ID is the 1 based segment index 1800 // shifted right by 8 bits as not to 1801 // collide with any of the 256 section IDs 1802 // that are possible 1803 const_segname, // Name of this section 1804 eSectionTypeContainer, // This section is a container of 1805 // other sections. 1806 sect64.addr, // File VM address == addresses as they are 1807 // found in the object file 1808 sect64.size, // VM size in bytes of this section 1809 sect64.offset, // Offset to the data for this section in 1810 // the file 1811 sect64.offset ? sect64.size : 0, // Size in bytes of 1812 // this section as 1813 // found in the file 1814 sect64.align, 1815 load_cmd.flags); // Flags for this section 1816 segment_sp->SetIsFake(true); 1817 segment_sp->SetPermissions(segment_permissions); 1818 m_sections_up->AddSection(segment_sp); 1819 if (add_to_unified) 1820 context.UnifiedList.AddSection(segment_sp); 1821 segment_sp->SetIsEncrypted(segment_is_encrypted); 1822 } 1823 } 1824 assert(segment_sp.get()); 1825 1826 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1827 1828 SectionSP section_sp(new Section( 1829 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1830 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1831 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1832 sect64.flags)); 1833 // Set the section to be encrypted to match the segment 1834 1835 bool section_is_encrypted = false; 1836 if (!segment_is_encrypted && load_cmd.filesize != 0) 1837 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1838 sect64.offset) != nullptr; 1839 1840 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1841 section_sp->SetPermissions(segment_permissions); 1842 segment_sp->GetChildren().AddSection(section_sp); 1843 1844 if (segment_sp->IsFake()) { 1845 segment_sp.reset(); 1846 const_segname.Clear(); 1847 } 1848 } 1849 } 1850 if (segment_sp && is_dsym) { 1851 if (first_segment_sectID <= context.NextSectionIdx) { 1852 lldb::user_id_t sect_uid; 1853 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1854 ++sect_uid) { 1855 SectionSP curr_section_sp( 1856 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1857 SectionSP next_section_sp; 1858 if (sect_uid + 1 <= context.NextSectionIdx) 1859 next_section_sp = 1860 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1861 1862 if (curr_section_sp.get()) { 1863 if (curr_section_sp->GetByteSize() == 0) { 1864 if (next_section_sp.get() != nullptr) 1865 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1866 curr_section_sp->GetFileAddress()); 1867 else 1868 curr_section_sp->SetByteSize(load_cmd.vmsize); 1869 } 1870 } 1871 } 1872 } 1873 } 1874 } 1875 1876 void ObjectFileMachO::ProcessDysymtabCommand(const load_command &load_cmd, 1877 lldb::offset_t offset) { 1878 m_dysymtab.cmd = load_cmd.cmd; 1879 m_dysymtab.cmdsize = load_cmd.cmdsize; 1880 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1881 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1882 } 1883 1884 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1885 if (m_sections_up) 1886 return; 1887 1888 m_sections_up.reset(new SectionList()); 1889 1890 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1891 // bool dump_sections = false; 1892 ModuleSP module_sp(GetModule()); 1893 1894 offset = MachHeaderSizeFromMagic(m_header.magic); 1895 1896 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1897 struct load_command load_cmd; 1898 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1899 const lldb::offset_t load_cmd_offset = offset; 1900 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 1901 break; 1902 1903 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1904 ProcessSegmentCommand(load_cmd, offset, i, context); 1905 else if (load_cmd.cmd == LC_DYSYMTAB) 1906 ProcessDysymtabCommand(load_cmd, offset); 1907 1908 offset = load_cmd_offset + load_cmd.cmdsize; 1909 } 1910 1911 if (context.FileAddressesChanged && module_sp) 1912 module_sp->SectionFileAddressesChanged(); 1913 } 1914 1915 class MachSymtabSectionInfo { 1916 public: 1917 MachSymtabSectionInfo(SectionList *section_list) 1918 : m_section_list(section_list), m_section_infos() { 1919 // Get the number of sections down to a depth of 1 to include all segments 1920 // and their sections, but no other sections that may be added for debug 1921 // map or 1922 m_section_infos.resize(section_list->GetNumSections(1)); 1923 } 1924 1925 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1926 if (n_sect == 0) 1927 return SectionSP(); 1928 if (n_sect < m_section_infos.size()) { 1929 if (!m_section_infos[n_sect].section_sp) { 1930 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1931 m_section_infos[n_sect].section_sp = section_sp; 1932 if (section_sp) { 1933 m_section_infos[n_sect].vm_range.SetBaseAddress( 1934 section_sp->GetFileAddress()); 1935 m_section_infos[n_sect].vm_range.SetByteSize( 1936 section_sp->GetByteSize()); 1937 } else { 1938 Host::SystemLog(Host::eSystemLogError, 1939 "error: unable to find section for section %u\n", 1940 n_sect); 1941 } 1942 } 1943 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1944 // Symbol is in section. 1945 return m_section_infos[n_sect].section_sp; 1946 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1947 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1948 file_addr) { 1949 // Symbol is in section with zero size, but has the same start address 1950 // as the section. This can happen with linker symbols (symbols that 1951 // start with the letter 'l' or 'L'. 1952 return m_section_infos[n_sect].section_sp; 1953 } 1954 } 1955 return m_section_list->FindSectionContainingFileAddress(file_addr); 1956 } 1957 1958 protected: 1959 struct SectionInfo { 1960 SectionInfo() : vm_range(), section_sp() {} 1961 1962 VMRange vm_range; 1963 SectionSP section_sp; 1964 }; 1965 SectionList *m_section_list; 1966 std::vector<SectionInfo> m_section_infos; 1967 }; 1968 1969 struct TrieEntry { 1970 TrieEntry() 1971 : name(), address(LLDB_INVALID_ADDRESS), flags(0), other(0), 1972 import_name() {} 1973 1974 void Clear() { 1975 name.Clear(); 1976 address = LLDB_INVALID_ADDRESS; 1977 flags = 0; 1978 other = 0; 1979 import_name.Clear(); 1980 } 1981 1982 void Dump() const { 1983 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1984 static_cast<unsigned long long>(address), 1985 static_cast<unsigned long long>(flags), 1986 static_cast<unsigned long long>(other), name.GetCString()); 1987 if (import_name) 1988 printf(" -> \"%s\"\n", import_name.GetCString()); 1989 else 1990 printf("\n"); 1991 } 1992 ConstString name; 1993 uint64_t address; 1994 uint64_t flags; 1995 uint64_t other; 1996 ConstString import_name; 1997 }; 1998 1999 struct TrieEntryWithOffset { 2000 lldb::offset_t nodeOffset; 2001 TrieEntry entry; 2002 2003 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 2004 2005 void Dump(uint32_t idx) const { 2006 printf("[%3u] 0x%16.16llx: ", idx, 2007 static_cast<unsigned long long>(nodeOffset)); 2008 entry.Dump(); 2009 } 2010 2011 bool operator<(const TrieEntryWithOffset &other) const { 2012 return (nodeOffset < other.nodeOffset); 2013 } 2014 }; 2015 2016 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2017 const bool is_arm, 2018 std::vector<llvm::StringRef> &nameSlices, 2019 std::set<lldb::addr_t> &resolver_addresses, 2020 std::vector<TrieEntryWithOffset> &output) { 2021 if (!data.ValidOffset(offset)) 2022 return true; 2023 2024 const uint64_t terminalSize = data.GetULEB128(&offset); 2025 lldb::offset_t children_offset = offset + terminalSize; 2026 if (terminalSize != 0) { 2027 TrieEntryWithOffset e(offset); 2028 e.entry.flags = data.GetULEB128(&offset); 2029 const char *import_name = nullptr; 2030 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2031 e.entry.address = 0; 2032 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2033 import_name = data.GetCStr(&offset); 2034 } else { 2035 e.entry.address = data.GetULEB128(&offset); 2036 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2037 e.entry.other = data.GetULEB128(&offset); 2038 uint64_t resolver_addr = e.entry.other; 2039 if (is_arm) 2040 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2041 resolver_addresses.insert(resolver_addr); 2042 } else 2043 e.entry.other = 0; 2044 } 2045 // Only add symbols that are reexport symbols with a valid import name 2046 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && 2047 import_name[0]) { 2048 std::string name; 2049 if (!nameSlices.empty()) { 2050 for (auto name_slice : nameSlices) 2051 name.append(name_slice.data(), name_slice.size()); 2052 } 2053 if (name.size() > 1) { 2054 // Skip the leading '_' 2055 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2056 } 2057 if (import_name) { 2058 // Skip the leading '_' 2059 e.entry.import_name.SetCString(import_name + 1); 2060 } 2061 output.push_back(e); 2062 } 2063 } 2064 2065 const uint8_t childrenCount = data.GetU8(&children_offset); 2066 for (uint8_t i = 0; i < childrenCount; ++i) { 2067 const char *cstr = data.GetCStr(&children_offset); 2068 if (cstr) 2069 nameSlices.push_back(llvm::StringRef(cstr)); 2070 else 2071 return false; // Corrupt data 2072 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2073 if (childNodeOffset) { 2074 if (!ParseTrieEntries(data, childNodeOffset, is_arm, nameSlices, 2075 resolver_addresses, output)) { 2076 return false; 2077 } 2078 } 2079 nameSlices.pop_back(); 2080 } 2081 return true; 2082 } 2083 2084 // Read the UUID out of a dyld_shared_cache file on-disk. 2085 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2086 const ByteOrder byte_order, 2087 const uint32_t addr_byte_size) { 2088 UUID dsc_uuid; 2089 DataBufferSP DscData = MapFileData( 2090 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2091 if (!DscData) 2092 return dsc_uuid; 2093 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2094 2095 char version_str[7]; 2096 lldb::offset_t offset = 0; 2097 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2098 version_str[6] = '\0'; 2099 if (strcmp(version_str, "dyld_v") == 0) { 2100 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2101 dsc_uuid = UUID::fromOptionalData( 2102 dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t)); 2103 } 2104 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2105 if (log && dsc_uuid.IsValid()) { 2106 LLDB_LOGF(log, "Shared cache %s has UUID %s", 2107 dyld_shared_cache.GetPath().c_str(), 2108 dsc_uuid.GetAsString().c_str()); 2109 } 2110 return dsc_uuid; 2111 } 2112 2113 size_t ObjectFileMachO::ParseSymtab() { 2114 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2115 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2116 m_file.GetFilename().AsCString("")); 2117 ModuleSP module_sp(GetModule()); 2118 if (!module_sp) 2119 return 0; 2120 2121 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2122 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2123 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2124 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2125 FunctionStarts function_starts; 2126 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2127 uint32_t i; 2128 FileSpecList dylib_files; 2129 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2130 static const llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2131 static const llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2132 static const llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2133 2134 for (i = 0; i < m_header.ncmds; ++i) { 2135 const lldb::offset_t cmd_offset = offset; 2136 // Read in the load command and load command size 2137 struct load_command lc; 2138 if (m_data.GetU32(&offset, &lc, 2) == nullptr) 2139 break; 2140 // Watch for the symbol table load command 2141 switch (lc.cmd) { 2142 case LC_SYMTAB: 2143 symtab_load_command.cmd = lc.cmd; 2144 symtab_load_command.cmdsize = lc.cmdsize; 2145 // Read in the rest of the symtab load command 2146 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2147 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2148 return 0; 2149 if (symtab_load_command.symoff == 0) { 2150 if (log) 2151 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2152 return 0; 2153 } 2154 2155 if (symtab_load_command.stroff == 0) { 2156 if (log) 2157 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2158 return 0; 2159 } 2160 2161 if (symtab_load_command.nsyms == 0) { 2162 if (log) 2163 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2164 return 0; 2165 } 2166 2167 if (symtab_load_command.strsize == 0) { 2168 if (log) 2169 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2170 return 0; 2171 } 2172 break; 2173 2174 case LC_DYLD_INFO: 2175 case LC_DYLD_INFO_ONLY: 2176 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2177 dyld_info.cmd = lc.cmd; 2178 dyld_info.cmdsize = lc.cmdsize; 2179 } else { 2180 memset(&dyld_info, 0, sizeof(dyld_info)); 2181 } 2182 break; 2183 2184 case LC_LOAD_DYLIB: 2185 case LC_LOAD_WEAK_DYLIB: 2186 case LC_REEXPORT_DYLIB: 2187 case LC_LOADFVMLIB: 2188 case LC_LOAD_UPWARD_DYLIB: { 2189 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2190 const char *path = m_data.PeekCStr(name_offset); 2191 if (path) { 2192 FileSpec file_spec(path); 2193 // Strip the path if there is @rpath, @executable, etc so we just use 2194 // the basename 2195 if (path[0] == '@') 2196 file_spec.GetDirectory().Clear(); 2197 2198 if (lc.cmd == LC_REEXPORT_DYLIB) { 2199 m_reexported_dylibs.AppendIfUnique(file_spec); 2200 } 2201 2202 dylib_files.Append(file_spec); 2203 } 2204 } break; 2205 2206 case LC_FUNCTION_STARTS: 2207 function_starts_load_command.cmd = lc.cmd; 2208 function_starts_load_command.cmdsize = lc.cmdsize; 2209 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2210 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2211 memset(&function_starts_load_command, 0, 2212 sizeof(function_starts_load_command)); 2213 break; 2214 2215 default: 2216 break; 2217 } 2218 offset = cmd_offset + lc.cmdsize; 2219 } 2220 2221 if (symtab_load_command.cmd) { 2222 Symtab *symtab = m_symtab_up.get(); 2223 SectionList *section_list = GetSectionList(); 2224 if (section_list == nullptr) 2225 return 0; 2226 2227 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2228 const ByteOrder byte_order = m_data.GetByteOrder(); 2229 bool bit_width_32 = addr_byte_size == 4; 2230 const size_t nlist_byte_size = 2231 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2232 2233 DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size); 2234 DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size); 2235 DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size); 2236 DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order, 2237 addr_byte_size); 2238 DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size); 2239 2240 const addr_t nlist_data_byte_size = 2241 symtab_load_command.nsyms * nlist_byte_size; 2242 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2243 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2244 2245 ProcessSP process_sp(m_process_wp.lock()); 2246 Process *process = process_sp.get(); 2247 2248 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2249 2250 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2251 Target &target = process->GetTarget(); 2252 2253 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2254 2255 SectionSP linkedit_section_sp( 2256 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2257 // Reading mach file from memory in a process or core file... 2258 2259 if (linkedit_section_sp) { 2260 addr_t linkedit_load_addr = 2261 linkedit_section_sp->GetLoadBaseAddress(&target); 2262 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2263 // We might be trying to access the symbol table before the 2264 // __LINKEDIT's load address has been set in the target. We can't 2265 // fail to read the symbol table, so calculate the right address 2266 // manually 2267 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2268 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2269 } 2270 2271 const addr_t linkedit_file_offset = 2272 linkedit_section_sp->GetFileOffset(); 2273 const addr_t symoff_addr = linkedit_load_addr + 2274 symtab_load_command.symoff - 2275 linkedit_file_offset; 2276 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2277 linkedit_file_offset; 2278 2279 bool data_was_read = false; 2280 2281 #if defined(__APPLE__) && \ 2282 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2283 if (m_header.flags & 0x80000000u && 2284 process->GetAddressByteSize() == sizeof(void *)) { 2285 // This mach-o memory file is in the dyld shared cache. If this 2286 // program is not remote and this is iOS, then this process will 2287 // share the same shared cache as the process we are debugging and we 2288 // can read the entire __LINKEDIT from the address space in this 2289 // process. This is a needed optimization that is used for local iOS 2290 // debugging only since all shared libraries in the shared cache do 2291 // not have corresponding files that exist in the file system of the 2292 // device. They have been combined into a single file. This means we 2293 // always have to load these files from memory. All of the symbol and 2294 // string tables from all of the __LINKEDIT sections from the shared 2295 // libraries in the shared cache have been merged into a single large 2296 // symbol and string table. Reading all of this symbol and string 2297 // table data across can slow down debug launch times, so we optimize 2298 // this by reading the memory for the __LINKEDIT section from this 2299 // process. 2300 2301 UUID lldb_shared_cache; 2302 addr_t lldb_shared_cache_addr; 2303 GetLLDBSharedCacheUUID (lldb_shared_cache_addr, lldb_shared_cache); 2304 UUID process_shared_cache; 2305 addr_t process_shared_cache_addr; 2306 GetProcessSharedCacheUUID(process, process_shared_cache_addr, process_shared_cache); 2307 bool use_lldb_cache = true; 2308 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2309 (lldb_shared_cache != process_shared_cache 2310 || process_shared_cache_addr != lldb_shared_cache_addr)) { 2311 use_lldb_cache = false; 2312 } 2313 2314 PlatformSP platform_sp(target.GetPlatform()); 2315 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2316 data_was_read = true; 2317 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2318 eByteOrderLittle); 2319 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2320 eByteOrderLittle); 2321 if (function_starts_load_command.cmd) { 2322 const addr_t func_start_addr = 2323 linkedit_load_addr + function_starts_load_command.dataoff - 2324 linkedit_file_offset; 2325 function_starts_data.SetData( 2326 (void *)func_start_addr, 2327 function_starts_load_command.datasize, eByteOrderLittle); 2328 } 2329 } 2330 } 2331 #endif 2332 2333 if (!data_was_read) { 2334 // Always load dyld - the dynamic linker - from memory if we didn't 2335 // find a binary anywhere else. lldb will not register 2336 // dylib/framework/bundle loads/unloads if we don't have the dyld 2337 // symbols, we force dyld to load from memory despite the user's 2338 // target.memory-module-load-level setting. 2339 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2340 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2341 DataBufferSP nlist_data_sp( 2342 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2343 if (nlist_data_sp) 2344 nlist_data.SetData(nlist_data_sp, 0, 2345 nlist_data_sp->GetByteSize()); 2346 if (m_dysymtab.nindirectsyms != 0) { 2347 const addr_t indirect_syms_addr = linkedit_load_addr + 2348 m_dysymtab.indirectsymoff - 2349 linkedit_file_offset; 2350 DataBufferSP indirect_syms_data_sp( 2351 ReadMemory(process_sp, indirect_syms_addr, 2352 m_dysymtab.nindirectsyms * 4)); 2353 if (indirect_syms_data_sp) 2354 indirect_symbol_index_data.SetData( 2355 indirect_syms_data_sp, 0, 2356 indirect_syms_data_sp->GetByteSize()); 2357 // If this binary is outside the shared cache, 2358 // cache the string table. 2359 // Binaries in the shared cache all share a giant string table, and 2360 // we can't share the string tables across multiple ObjectFileMachO's, 2361 // so we'd end up re-reading this mega-strtab for every binary 2362 // in the shared cache - it would be a big perf problem. 2363 // For binaries outside the shared cache, it's faster to read the 2364 // entire strtab at once instead of piece-by-piece as we process 2365 // the nlist records. 2366 if ((m_header.flags & 0x80000000u) == 0) { 2367 DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, 2368 strtab_data_byte_size)); 2369 if (strtab_data_sp) { 2370 strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 2371 } 2372 } 2373 } 2374 } 2375 if (memory_module_load_level >= 2376 eMemoryModuleLoadLevelPartial) { 2377 if (function_starts_load_command.cmd) { 2378 const addr_t func_start_addr = 2379 linkedit_load_addr + function_starts_load_command.dataoff - 2380 linkedit_file_offset; 2381 DataBufferSP func_start_data_sp( 2382 ReadMemory(process_sp, func_start_addr, 2383 function_starts_load_command.datasize)); 2384 if (func_start_data_sp) 2385 function_starts_data.SetData(func_start_data_sp, 0, 2386 func_start_data_sp->GetByteSize()); 2387 } 2388 } 2389 } 2390 } 2391 } else { 2392 nlist_data.SetData(m_data, symtab_load_command.symoff, 2393 nlist_data_byte_size); 2394 strtab_data.SetData(m_data, symtab_load_command.stroff, 2395 strtab_data_byte_size); 2396 2397 if (dyld_info.export_size > 0) { 2398 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2399 dyld_info.export_size); 2400 } 2401 2402 if (m_dysymtab.nindirectsyms != 0) { 2403 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2404 m_dysymtab.nindirectsyms * 4); 2405 } 2406 if (function_starts_load_command.cmd) { 2407 function_starts_data.SetData(m_data, 2408 function_starts_load_command.dataoff, 2409 function_starts_load_command.datasize); 2410 } 2411 } 2412 2413 if (nlist_data.GetByteSize() == 0 && 2414 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2415 if (log) 2416 module_sp->LogMessage(log, "failed to read nlist data"); 2417 return 0; 2418 } 2419 2420 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2421 if (!have_strtab_data) { 2422 if (process) { 2423 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2424 if (log) 2425 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2426 return 0; 2427 } 2428 } else { 2429 if (log) 2430 module_sp->LogMessage(log, "failed to read strtab data"); 2431 return 0; 2432 } 2433 } 2434 2435 ConstString g_segment_name_TEXT = GetSegmentNameTEXT(); 2436 ConstString g_segment_name_DATA = GetSegmentNameDATA(); 2437 ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2438 ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2439 ConstString g_segment_name_OBJC = GetSegmentNameOBJC(); 2440 ConstString g_section_name_eh_frame = GetSectionNameEHFrame(); 2441 SectionSP text_section_sp( 2442 section_list->FindSectionByName(g_segment_name_TEXT)); 2443 SectionSP data_section_sp( 2444 section_list->FindSectionByName(g_segment_name_DATA)); 2445 SectionSP data_dirty_section_sp( 2446 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2447 SectionSP data_const_section_sp( 2448 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2449 SectionSP objc_section_sp( 2450 section_list->FindSectionByName(g_segment_name_OBJC)); 2451 SectionSP eh_frame_section_sp; 2452 if (text_section_sp.get()) 2453 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2454 g_section_name_eh_frame); 2455 else 2456 eh_frame_section_sp = 2457 section_list->FindSectionByName(g_section_name_eh_frame); 2458 2459 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2460 2461 // lldb works best if it knows the start address of all functions in a 2462 // module. Linker symbols or debug info are normally the best source of 2463 // information for start addr / size but they may be stripped in a released 2464 // binary. Two additional sources of information exist in Mach-O binaries: 2465 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2466 // function's start address in the 2467 // binary, relative to the text section. 2468 // eh_frame - the eh_frame FDEs have the start addr & size of 2469 // each function 2470 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2471 // all modern binaries. 2472 // Binaries built to run on older releases may need to use eh_frame 2473 // information. 2474 2475 if (text_section_sp && function_starts_data.GetByteSize()) { 2476 FunctionStarts::Entry function_start_entry; 2477 function_start_entry.data = false; 2478 lldb::offset_t function_start_offset = 0; 2479 function_start_entry.addr = text_section_sp->GetFileAddress(); 2480 uint64_t delta; 2481 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2482 0) { 2483 // Now append the current entry 2484 function_start_entry.addr += delta; 2485 function_starts.Append(function_start_entry); 2486 } 2487 } else { 2488 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2489 // load command claiming an eh_frame but it doesn't actually have the 2490 // eh_frame content. And if we have a dSYM, we don't need to do any of 2491 // this fill-in-the-missing-symbols works anyway - the debug info should 2492 // give us all the functions in the module. 2493 if (text_section_sp.get() && eh_frame_section_sp.get() && 2494 m_type != eTypeDebugInfo) { 2495 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2496 DWARFCallFrameInfo::EH); 2497 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2498 eh_frame.GetFunctionAddressAndSizeVector(functions); 2499 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2500 size_t count = functions.GetSize(); 2501 for (size_t i = 0; i < count; ++i) { 2502 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2503 functions.GetEntryAtIndex(i); 2504 if (func) { 2505 FunctionStarts::Entry function_start_entry; 2506 function_start_entry.addr = func->base - text_base_addr; 2507 function_starts.Append(function_start_entry); 2508 } 2509 } 2510 } 2511 } 2512 2513 const size_t function_starts_count = function_starts.GetSize(); 2514 2515 // For user process binaries (executables, dylibs, frameworks, bundles), if 2516 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2517 // going to assume the binary has been stripped. Don't allow assembly 2518 // language instruction emulation because we don't know proper function 2519 // start boundaries. 2520 // 2521 // For all other types of binaries (kernels, stand-alone bare board 2522 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2523 // sections - we should not make any assumptions about them based on that. 2524 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2525 m_allow_assembly_emulation_unwind_plans = false; 2526 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2527 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2528 2529 if (unwind_or_symbol_log) 2530 module_sp->LogMessage( 2531 unwind_or_symbol_log, 2532 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2533 } 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 in this map 2556 // so we know 2557 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2558 uint32_t nlist_idx = 0; 2559 Symbol *symbol_ptr = nullptr; 2560 2561 uint32_t sym_idx = 0; 2562 Symbol *sym = nullptr; 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 std::vector<llvm::StringRef> nameSlices; 2572 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, 2573 resolver_addresses, trie_entries); 2574 2575 ConstString text_segment_name("__TEXT"); 2576 SectionSP text_segment_sp = 2577 GetSectionList()->FindSectionByName(text_segment_name); 2578 if (text_segment_sp) { 2579 const lldb::addr_t text_segment_file_addr = 2580 text_segment_sp->GetFileAddress(); 2581 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2582 for (auto &e : trie_entries) 2583 e.entry.address += text_segment_file_addr; 2584 } 2585 } 2586 } 2587 2588 typedef std::set<ConstString> IndirectSymbols; 2589 IndirectSymbols indirect_symbol_names; 2590 2591 #if defined(__APPLE__) && \ 2592 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2593 2594 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2595 // optimized by moving LOCAL symbols out of the memory mapped portion of 2596 // the DSC. The symbol information has all been retained, but it isn't 2597 // available in the normal nlist data. However, there *are* duplicate 2598 // entries of *some* 2599 // LOCAL symbols in the normal nlist data. To handle this situation 2600 // correctly, we must first attempt 2601 // to parse any DSC unmapped symbol information. If we find any, we set a 2602 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2603 2604 if (m_header.flags & 0x80000000u) { 2605 // Before we can start mapping the DSC, we need to make certain the 2606 // target process is actually using the cache we can find. 2607 2608 // Next we need to determine the correct path for the dyld shared cache. 2609 2610 ArchSpec header_arch; 2611 GetArchitecture(header_arch); 2612 char dsc_path[PATH_MAX]; 2613 char dsc_path_development[PATH_MAX]; 2614 2615 snprintf( 2616 dsc_path, sizeof(dsc_path), "%s%s%s", 2617 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2618 */ 2619 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2620 header_arch.GetArchitectureName()); 2621 2622 snprintf( 2623 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2624 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2625 */ 2626 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2627 header_arch.GetArchitectureName(), ".development"); 2628 2629 FileSpec dsc_nondevelopment_filespec(dsc_path, false); 2630 FileSpec dsc_development_filespec(dsc_path_development, false); 2631 FileSpec dsc_filespec; 2632 2633 UUID dsc_uuid; 2634 UUID process_shared_cache_uuid; 2635 addr_t process_shared_cache_base_addr; 2636 2637 if (process) { 2638 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, process_shared_cache_uuid); 2639 } 2640 2641 // First see if we can find an exact match for the inferior process 2642 // shared cache UUID in the development or non-development shared caches 2643 // on disk. 2644 if (process_shared_cache_uuid.IsValid()) { 2645 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2646 UUID dsc_development_uuid = GetSharedCacheUUID( 2647 dsc_development_filespec, byte_order, addr_byte_size); 2648 if (dsc_development_uuid.IsValid() && 2649 dsc_development_uuid == process_shared_cache_uuid) { 2650 dsc_filespec = dsc_development_filespec; 2651 dsc_uuid = dsc_development_uuid; 2652 } 2653 } 2654 if (!dsc_uuid.IsValid() && 2655 FileSystem::Instance().Exists(dsc_nondevelopment_filespec)) { 2656 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2657 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2658 if (dsc_nondevelopment_uuid.IsValid() && 2659 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2660 dsc_filespec = dsc_nondevelopment_filespec; 2661 dsc_uuid = dsc_nondevelopment_uuid; 2662 } 2663 } 2664 } 2665 2666 // Failing a UUID match, prefer the development dyld_shared cache if both 2667 // are present. 2668 if (!FileSystem::Instance().Exists(dsc_filespec)) { 2669 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2670 dsc_filespec = dsc_development_filespec; 2671 } else { 2672 dsc_filespec = dsc_nondevelopment_filespec; 2673 } 2674 } 2675 2676 /* The dyld_cache_header has a pointer to the 2677 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2678 The dyld_cache_local_symbols_info structure gives us three things: 2679 1. The start and count of the nlist records in the dyld_shared_cache 2680 file 2681 2. The start and size of the strings for these nlist records 2682 3. The start and count of dyld_cache_local_symbols_entry entries 2683 2684 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2685 dyld shared cache. 2686 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2687 the dyld shared cache. 2688 The dyld_cache_local_symbols_entry also lists the start of this 2689 dylib/framework's nlist records 2690 and the count of how many nlist records there are for this 2691 dylib/framework. 2692 */ 2693 2694 // Process the dyld shared cache header to find the unmapped symbols 2695 2696 DataBufferSP dsc_data_sp = MapFileData( 2697 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2698 if (!dsc_uuid.IsValid()) { 2699 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2700 } 2701 if (dsc_data_sp) { 2702 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2703 2704 bool uuid_match = true; 2705 if (dsc_uuid.IsValid() && process) { 2706 if (process_shared_cache_uuid.IsValid() && 2707 dsc_uuid != process_shared_cache_uuid) { 2708 // The on-disk dyld_shared_cache file is not the same as the one in 2709 // this process' memory, don't use it. 2710 uuid_match = false; 2711 ModuleSP module_sp(GetModule()); 2712 if (module_sp) 2713 module_sp->ReportWarning("process shared cache does not match " 2714 "on-disk dyld_shared_cache file, some " 2715 "symbol names will be missing."); 2716 } 2717 } 2718 2719 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2720 2721 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2722 2723 // If the mappingOffset points to a location inside the header, we've 2724 // opened an old dyld shared cache, and should not proceed further. 2725 if (uuid_match && 2726 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2727 2728 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2729 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2730 mappingOffset); 2731 2732 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2733 byte_order, addr_byte_size); 2734 offset = 0; 2735 2736 // The File addresses (from the in-memory Mach-O load commands) for 2737 // the shared libraries in the shared library cache need to be 2738 // adjusted by an offset to match up with the dylibOffset identifying 2739 // field in the dyld_cache_local_symbol_entry's. This offset is 2740 // recorded in mapping_offset_value. 2741 const uint64_t mapping_offset_value = 2742 dsc_mapping_info_data.GetU64(&offset); 2743 2744 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, 2745 localSymbolsOffset); 2746 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2747 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2748 2749 if (localSymbolsOffset && localSymbolsSize) { 2750 // Map the local symbols 2751 DataBufferSP dsc_local_symbols_data_sp = 2752 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2753 2754 if (dsc_local_symbols_data_sp) { 2755 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2756 byte_order, addr_byte_size); 2757 2758 offset = 0; 2759 2760 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2761 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2762 UndefinedNameToDescMap undefined_name_to_desc; 2763 SymbolIndexToName reexport_shlib_needs_fixup; 2764 2765 // Read the local_symbols_infos struct in one shot 2766 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2767 dsc_local_symbols_data.GetU32(&offset, 2768 &local_symbols_info.nlistOffset, 6); 2769 2770 SectionSP text_section_sp( 2771 section_list->FindSectionByName(GetSegmentNameTEXT())); 2772 2773 uint32_t header_file_offset = 2774 (text_section_sp->GetFileAddress() - mapping_offset_value); 2775 2776 offset = local_symbols_info.entriesOffset; 2777 for (uint32_t entry_index = 0; 2778 entry_index < local_symbols_info.entriesCount; 2779 entry_index++) { 2780 struct lldb_copy_dyld_cache_local_symbols_entry 2781 local_symbols_entry; 2782 local_symbols_entry.dylibOffset = 2783 dsc_local_symbols_data.GetU32(&offset); 2784 local_symbols_entry.nlistStartIndex = 2785 dsc_local_symbols_data.GetU32(&offset); 2786 local_symbols_entry.nlistCount = 2787 dsc_local_symbols_data.GetU32(&offset); 2788 2789 if (header_file_offset == local_symbols_entry.dylibOffset) { 2790 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2791 2792 // The normal nlist code cannot correctly size the Symbols 2793 // array, we need to allocate it here. 2794 sym = symtab->Resize( 2795 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2796 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2797 num_syms = symtab->GetNumSymbols(); 2798 2799 nlist_data_offset = 2800 local_symbols_info.nlistOffset + 2801 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2802 uint32_t string_table_offset = 2803 local_symbols_info.stringsOffset; 2804 2805 for (uint32_t nlist_index = 0; 2806 nlist_index < local_symbols_entry.nlistCount; 2807 nlist_index++) { 2808 ///////////////////////////// 2809 { 2810 struct nlist_64 nlist; 2811 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize( 2812 nlist_data_offset, nlist_byte_size)) 2813 break; 2814 2815 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked( 2816 &nlist_data_offset); 2817 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked( 2818 &nlist_data_offset); 2819 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked( 2820 &nlist_data_offset); 2821 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked( 2822 &nlist_data_offset); 2823 nlist.n_value = 2824 dsc_local_symbols_data.GetAddress_unchecked( 2825 &nlist_data_offset); 2826 2827 SymbolType type = eSymbolTypeInvalid; 2828 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2829 string_table_offset + nlist.n_strx); 2830 2831 if (symbol_name == NULL) { 2832 // No symbol should be NULL, even the symbols with no 2833 // string values should have an offset zero which 2834 // points to an empty C-string 2835 Host::SystemLog( 2836 Host::eSystemLogError, 2837 "error: DSC unmapped local symbol[%u] has invalid " 2838 "string table offset 0x%x in %s, ignoring symbol\n", 2839 entry_index, nlist.n_strx, 2840 module_sp->GetFileSpec().GetPath().c_str()); 2841 continue; 2842 } 2843 if (symbol_name[0] == '\0') 2844 symbol_name = NULL; 2845 2846 const char *symbol_name_non_abi_mangled = NULL; 2847 2848 SectionSP symbol_section; 2849 uint32_t symbol_byte_size = 0; 2850 bool add_nlist = true; 2851 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2852 bool demangled_is_synthesized = false; 2853 bool is_gsym = false; 2854 bool set_value = true; 2855 2856 assert(sym_idx < num_syms); 2857 2858 sym[sym_idx].SetDebug(is_debug); 2859 2860 if (is_debug) { 2861 switch (nlist.n_type) { 2862 case N_GSYM: 2863 // global symbol: name,,NO_SECT,type,0 2864 // Sometimes the N_GSYM value contains the address. 2865 2866 // FIXME: In the .o files, we have a GSYM and a debug 2867 // symbol for all the ObjC data. They 2868 // have the same address, but we want to ensure that 2869 // we always find only the real symbol, 'cause we 2870 // don't currently correctly attribute the 2871 // GSYM one to the ObjCClass/Ivar/MetaClass 2872 // symbol type. This is a temporary hack to make 2873 // sure the ObjectiveC symbols get treated correctly. 2874 // To do this right, we should coalesce all the GSYM 2875 // & global symbols that have the same address. 2876 2877 is_gsym = true; 2878 sym[sym_idx].SetExternal(true); 2879 2880 if (symbol_name && symbol_name[0] == '_' && 2881 symbol_name[1] == 'O') { 2882 llvm::StringRef symbol_name_ref(symbol_name); 2883 if (symbol_name_ref.startswith( 2884 g_objc_v2_prefix_class)) { 2885 symbol_name_non_abi_mangled = symbol_name + 1; 2886 symbol_name = 2887 symbol_name + g_objc_v2_prefix_class.size(); 2888 type = eSymbolTypeObjCClass; 2889 demangled_is_synthesized = true; 2890 2891 } else if (symbol_name_ref.startswith( 2892 g_objc_v2_prefix_metaclass)) { 2893 symbol_name_non_abi_mangled = symbol_name + 1; 2894 symbol_name = symbol_name + 2895 g_objc_v2_prefix_metaclass.size(); 2896 type = eSymbolTypeObjCMetaClass; 2897 demangled_is_synthesized = true; 2898 } else if (symbol_name_ref.startswith( 2899 g_objc_v2_prefix_ivar)) { 2900 symbol_name_non_abi_mangled = symbol_name + 1; 2901 symbol_name = 2902 symbol_name + g_objc_v2_prefix_ivar.size(); 2903 type = eSymbolTypeObjCIVar; 2904 demangled_is_synthesized = true; 2905 } 2906 } else { 2907 if (nlist.n_value != 0) 2908 symbol_section = section_info.GetSection( 2909 nlist.n_sect, nlist.n_value); 2910 type = eSymbolTypeData; 2911 } 2912 break; 2913 2914 case N_FNAME: 2915 // procedure name (f77 kludge): name,,NO_SECT,0,0 2916 type = eSymbolTypeCompiler; 2917 break; 2918 2919 case N_FUN: 2920 // procedure: name,,n_sect,linenumber,address 2921 if (symbol_name) { 2922 type = eSymbolTypeCode; 2923 symbol_section = section_info.GetSection( 2924 nlist.n_sect, nlist.n_value); 2925 2926 N_FUN_addr_to_sym_idx.insert( 2927 std::make_pair(nlist.n_value, sym_idx)); 2928 // We use the current number of symbols in the 2929 // symbol table in lieu of using nlist_idx in case 2930 // we ever start trimming entries out 2931 N_FUN_indexes.push_back(sym_idx); 2932 } else { 2933 type = eSymbolTypeCompiler; 2934 2935 if (!N_FUN_indexes.empty()) { 2936 // Copy the size of the function into the 2937 // original 2938 // STAB entry so we don't have 2939 // to hunt for it later 2940 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2941 ->SetByteSize(nlist.n_value); 2942 N_FUN_indexes.pop_back(); 2943 // We don't really need the end function STAB as 2944 // it contains the size which we already placed 2945 // with the original symbol, so don't add it if 2946 // we want a minimal symbol table 2947 add_nlist = false; 2948 } 2949 } 2950 break; 2951 2952 case N_STSYM: 2953 // static symbol: name,,n_sect,type,address 2954 N_STSYM_addr_to_sym_idx.insert( 2955 std::make_pair(nlist.n_value, sym_idx)); 2956 symbol_section = section_info.GetSection( 2957 nlist.n_sect, nlist.n_value); 2958 if (symbol_name && symbol_name[0]) { 2959 type = ObjectFile::GetSymbolTypeFromName( 2960 symbol_name + 1, eSymbolTypeData); 2961 } 2962 break; 2963 2964 case N_LCSYM: 2965 // .lcomm symbol: name,,n_sect,type,address 2966 symbol_section = section_info.GetSection( 2967 nlist.n_sect, nlist.n_value); 2968 type = eSymbolTypeCommonBlock; 2969 break; 2970 2971 case N_BNSYM: 2972 // We use the current number of symbols in the symbol 2973 // table in lieu of using nlist_idx in case we ever 2974 // start trimming entries out Skip these if we want 2975 // minimal symbol tables 2976 add_nlist = false; 2977 break; 2978 2979 case N_ENSYM: 2980 // Set the size of the N_BNSYM to the terminating 2981 // index of this N_ENSYM so that we can always skip 2982 // the entire symbol if we need to navigate more 2983 // quickly at the source level when parsing STABS 2984 // Skip these if we want minimal symbol tables 2985 add_nlist = false; 2986 break; 2987 2988 case N_OPT: 2989 // emitted with gcc2_compiled and in gcc source 2990 type = eSymbolTypeCompiler; 2991 break; 2992 2993 case N_RSYM: 2994 // register sym: name,,NO_SECT,type,register 2995 type = eSymbolTypeVariable; 2996 break; 2997 2998 case N_SLINE: 2999 // src line: 0,,n_sect,linenumber,address 3000 symbol_section = section_info.GetSection( 3001 nlist.n_sect, nlist.n_value); 3002 type = eSymbolTypeLineEntry; 3003 break; 3004 3005 case N_SSYM: 3006 // structure elt: name,,NO_SECT,type,struct_offset 3007 type = eSymbolTypeVariableType; 3008 break; 3009 3010 case N_SO: 3011 // source file name 3012 type = eSymbolTypeSourceFile; 3013 if (symbol_name == NULL) { 3014 add_nlist = false; 3015 if (N_SO_index != UINT32_MAX) { 3016 // Set the size of the N_SO to the terminating 3017 // index of this N_SO so that we can always skip 3018 // the entire N_SO if we need to navigate more 3019 // quickly at the source level when parsing STABS 3020 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3021 symbol_ptr->SetByteSize(sym_idx); 3022 symbol_ptr->SetSizeIsSibling(true); 3023 } 3024 N_NSYM_indexes.clear(); 3025 N_INCL_indexes.clear(); 3026 N_BRAC_indexes.clear(); 3027 N_COMM_indexes.clear(); 3028 N_FUN_indexes.clear(); 3029 N_SO_index = UINT32_MAX; 3030 } else { 3031 // We use the current number of symbols in the 3032 // symbol table in lieu of using nlist_idx in case 3033 // we ever start trimming entries out 3034 const bool N_SO_has_full_path = 3035 symbol_name[0] == '/'; 3036 if (N_SO_has_full_path) { 3037 if ((N_SO_index == sym_idx - 1) && 3038 ((sym_idx - 1) < num_syms)) { 3039 // We have two consecutive N_SO entries where 3040 // the first contains a directory and the 3041 // second contains a full path. 3042 sym[sym_idx - 1].GetMangled().SetValue( 3043 ConstString(symbol_name), false); 3044 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3045 add_nlist = false; 3046 } else { 3047 // This is the first entry in a N_SO that 3048 // contains a directory or 3049 // a full path to the source file 3050 N_SO_index = sym_idx; 3051 } 3052 } else if ((N_SO_index == sym_idx - 1) && 3053 ((sym_idx - 1) < num_syms)) { 3054 // This is usually the second N_SO entry that 3055 // contains just the filename, so here we combine 3056 // it with the first one if we are minimizing the 3057 // symbol table 3058 const char *so_path = 3059 sym[sym_idx - 1] 3060 .GetMangled() 3061 .GetDemangledName( 3062 lldb::eLanguageTypeUnknown) 3063 .AsCString(); 3064 if (so_path && so_path[0]) { 3065 std::string full_so_path(so_path); 3066 const size_t double_slash_pos = 3067 full_so_path.find("//"); 3068 if (double_slash_pos != std::string::npos) { 3069 // The linker has been generating bad N_SO 3070 // entries with doubled up paths 3071 // in the format "%s%s" where the first 3072 // string in the DW_AT_comp_dir, and the 3073 // second is the directory for the source 3074 // file so you end up with a path that looks 3075 // like "/tmp/src//tmp/src/" 3076 FileSpec so_dir(so_path, false); 3077 if (!FileSystem::Instance().Exists(so_dir)) { 3078 so_dir.SetFile( 3079 &full_so_path[double_slash_pos + 1], 3080 false); 3081 if (FileSystem::Instance().Exists(so_dir)) { 3082 // Trim off the incorrect path 3083 full_so_path.erase(0, 3084 double_slash_pos + 1); 3085 } 3086 } 3087 } 3088 if (*full_so_path.rbegin() != '/') 3089 full_so_path += '/'; 3090 full_so_path += symbol_name; 3091 sym[sym_idx - 1].GetMangled().SetValue( 3092 ConstString(full_so_path.c_str()), false); 3093 add_nlist = false; 3094 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3095 } 3096 } else { 3097 // This could be a relative path to a N_SO 3098 N_SO_index = sym_idx; 3099 } 3100 } 3101 break; 3102 3103 case N_OSO: 3104 // object file name: name,,0,0,st_mtime 3105 type = eSymbolTypeObjectFile; 3106 break; 3107 3108 case N_LSYM: 3109 // local sym: name,,NO_SECT,type,offset 3110 type = eSymbolTypeLocal; 3111 break; 3112 3113 // INCL scopes 3114 case N_BINCL: 3115 // include file beginning: name,,NO_SECT,0,sum We use 3116 // the current number of symbols in the symbol table 3117 // in lieu of using nlist_idx in case we ever start 3118 // trimming entries out 3119 N_INCL_indexes.push_back(sym_idx); 3120 type = eSymbolTypeScopeBegin; 3121 break; 3122 3123 case N_EINCL: 3124 // include file end: name,,NO_SECT,0,0 3125 // Set the size of the N_BINCL to the terminating 3126 // index of this N_EINCL so that we can always skip 3127 // the entire symbol if we need to navigate more 3128 // quickly at the source level when parsing STABS 3129 if (!N_INCL_indexes.empty()) { 3130 symbol_ptr = 3131 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3132 symbol_ptr->SetByteSize(sym_idx + 1); 3133 symbol_ptr->SetSizeIsSibling(true); 3134 N_INCL_indexes.pop_back(); 3135 } 3136 type = eSymbolTypeScopeEnd; 3137 break; 3138 3139 case N_SOL: 3140 // #included file name: name,,n_sect,0,address 3141 type = eSymbolTypeHeaderFile; 3142 3143 // We currently don't use the header files on darwin 3144 add_nlist = false; 3145 break; 3146 3147 case N_PARAMS: 3148 // compiler parameters: name,,NO_SECT,0,0 3149 type = eSymbolTypeCompiler; 3150 break; 3151 3152 case N_VERSION: 3153 // compiler version: name,,NO_SECT,0,0 3154 type = eSymbolTypeCompiler; 3155 break; 3156 3157 case N_OLEVEL: 3158 // compiler -O level: name,,NO_SECT,0,0 3159 type = eSymbolTypeCompiler; 3160 break; 3161 3162 case N_PSYM: 3163 // parameter: name,,NO_SECT,type,offset 3164 type = eSymbolTypeVariable; 3165 break; 3166 3167 case N_ENTRY: 3168 // alternate entry: name,,n_sect,linenumber,address 3169 symbol_section = section_info.GetSection( 3170 nlist.n_sect, nlist.n_value); 3171 type = eSymbolTypeLineEntry; 3172 break; 3173 3174 // Left and Right Braces 3175 case N_LBRAC: 3176 // left bracket: 0,,NO_SECT,nesting level,address We 3177 // use the current number of symbols in the symbol 3178 // table in lieu of using nlist_idx in case we ever 3179 // start trimming entries out 3180 symbol_section = section_info.GetSection( 3181 nlist.n_sect, nlist.n_value); 3182 N_BRAC_indexes.push_back(sym_idx); 3183 type = eSymbolTypeScopeBegin; 3184 break; 3185 3186 case N_RBRAC: 3187 // right bracket: 0,,NO_SECT,nesting level,address 3188 // Set the size of the N_LBRAC to the terminating 3189 // index of this N_RBRAC so that we can always skip 3190 // the entire symbol if we need to navigate more 3191 // quickly at the source level when parsing STABS 3192 symbol_section = section_info.GetSection( 3193 nlist.n_sect, nlist.n_value); 3194 if (!N_BRAC_indexes.empty()) { 3195 symbol_ptr = 3196 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3197 symbol_ptr->SetByteSize(sym_idx + 1); 3198 symbol_ptr->SetSizeIsSibling(true); 3199 N_BRAC_indexes.pop_back(); 3200 } 3201 type = eSymbolTypeScopeEnd; 3202 break; 3203 3204 case N_EXCL: 3205 // deleted include file: name,,NO_SECT,0,sum 3206 type = eSymbolTypeHeaderFile; 3207 break; 3208 3209 // COMM scopes 3210 case N_BCOMM: 3211 // begin common: name,,NO_SECT,0,0 3212 // We use the current number of symbols in the symbol 3213 // table in lieu of using nlist_idx in case we ever 3214 // start trimming entries out 3215 type = eSymbolTypeScopeBegin; 3216 N_COMM_indexes.push_back(sym_idx); 3217 break; 3218 3219 case N_ECOML: 3220 // end common (local name): 0,,n_sect,0,address 3221 symbol_section = section_info.GetSection( 3222 nlist.n_sect, nlist.n_value); 3223 // Fall through 3224 3225 case N_ECOMM: 3226 // end common: name,,n_sect,0,0 3227 // Set the size of the N_BCOMM to the terminating 3228 // index of this N_ECOMM/N_ECOML so that we can 3229 // always skip the entire symbol if we need to 3230 // navigate more quickly at the source level when 3231 // parsing STABS 3232 if (!N_COMM_indexes.empty()) { 3233 symbol_ptr = 3234 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3235 symbol_ptr->SetByteSize(sym_idx + 1); 3236 symbol_ptr->SetSizeIsSibling(true); 3237 N_COMM_indexes.pop_back(); 3238 } 3239 type = eSymbolTypeScopeEnd; 3240 break; 3241 3242 case N_LENG: 3243 // second stab entry with length information 3244 type = eSymbolTypeAdditional; 3245 break; 3246 3247 default: 3248 break; 3249 } 3250 } else { 3251 // uint8_t n_pext = N_PEXT & nlist.n_type; 3252 uint8_t n_type = N_TYPE & nlist.n_type; 3253 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3254 3255 switch (n_type) { 3256 case N_INDR: { 3257 const char *reexport_name_cstr = 3258 strtab_data.PeekCStr(nlist.n_value); 3259 if (reexport_name_cstr && reexport_name_cstr[0]) { 3260 type = eSymbolTypeReExported; 3261 ConstString reexport_name( 3262 reexport_name_cstr + 3263 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3264 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3265 set_value = false; 3266 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3267 indirect_symbol_names.insert( 3268 ConstString(symbol_name + 3269 ((symbol_name[0] == '_') ? 1 : 0))); 3270 } else 3271 type = eSymbolTypeUndefined; 3272 } break; 3273 3274 case N_UNDF: 3275 if (symbol_name && symbol_name[0]) { 3276 ConstString undefined_name( 3277 symbol_name + 3278 ((symbol_name[0] == '_') ? 1 : 0)); 3279 undefined_name_to_desc[undefined_name] = 3280 nlist.n_desc; 3281 } 3282 // Fall through 3283 case N_PBUD: 3284 type = eSymbolTypeUndefined; 3285 break; 3286 3287 case N_ABS: 3288 type = eSymbolTypeAbsolute; 3289 break; 3290 3291 case N_SECT: { 3292 symbol_section = section_info.GetSection( 3293 nlist.n_sect, nlist.n_value); 3294 3295 if (symbol_section == NULL) { 3296 // TODO: warn about this? 3297 add_nlist = false; 3298 break; 3299 } 3300 3301 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3302 type = eSymbolTypeException; 3303 } else { 3304 uint32_t section_type = 3305 symbol_section->Get() & SECTION_TYPE; 3306 3307 switch (section_type) { 3308 case S_CSTRING_LITERALS: 3309 type = eSymbolTypeData; 3310 break; // section with only literal C strings 3311 case S_4BYTE_LITERALS: 3312 type = eSymbolTypeData; 3313 break; // section with only 4 byte literals 3314 case S_8BYTE_LITERALS: 3315 type = eSymbolTypeData; 3316 break; // section with only 8 byte literals 3317 case S_LITERAL_POINTERS: 3318 type = eSymbolTypeTrampoline; 3319 break; // section with only pointers to literals 3320 case S_NON_LAZY_SYMBOL_POINTERS: 3321 type = eSymbolTypeTrampoline; 3322 break; // section with only non-lazy symbol 3323 // pointers 3324 case S_LAZY_SYMBOL_POINTERS: 3325 type = eSymbolTypeTrampoline; 3326 break; // section with only lazy symbol pointers 3327 case S_SYMBOL_STUBS: 3328 type = eSymbolTypeTrampoline; 3329 break; // section with only symbol stubs, byte 3330 // size of stub in the reserved2 field 3331 case S_MOD_INIT_FUNC_POINTERS: 3332 type = eSymbolTypeCode; 3333 break; // section with only function pointers for 3334 // initialization 3335 case S_MOD_TERM_FUNC_POINTERS: 3336 type = eSymbolTypeCode; 3337 break; // section with only function pointers for 3338 // termination 3339 case S_INTERPOSING: 3340 type = eSymbolTypeTrampoline; 3341 break; // section with only pairs of function 3342 // pointers for interposing 3343 case S_16BYTE_LITERALS: 3344 type = eSymbolTypeData; 3345 break; // section with only 16 byte literals 3346 case S_DTRACE_DOF: 3347 type = eSymbolTypeInstrumentation; 3348 break; 3349 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3350 type = eSymbolTypeTrampoline; 3351 break; 3352 default: 3353 switch (symbol_section->GetType()) { 3354 case lldb::eSectionTypeCode: 3355 type = eSymbolTypeCode; 3356 break; 3357 case eSectionTypeData: 3358 case eSectionTypeDataCString: // Inlined C string 3359 // data 3360 case eSectionTypeDataCStringPointers: // Pointers 3361 // to C 3362 // string 3363 // data 3364 case eSectionTypeDataSymbolAddress: // Address of 3365 // a symbol in 3366 // the symbol 3367 // table 3368 case eSectionTypeData4: 3369 case eSectionTypeData8: 3370 case eSectionTypeData16: 3371 type = eSymbolTypeData; 3372 break; 3373 default: 3374 break; 3375 } 3376 break; 3377 } 3378 3379 if (type == eSymbolTypeInvalid) { 3380 const char *symbol_sect_name = 3381 symbol_section->GetName().AsCString(); 3382 if (symbol_section->IsDescendant( 3383 text_section_sp.get())) { 3384 if (symbol_section->IsClear( 3385 S_ATTR_PURE_INSTRUCTIONS | 3386 S_ATTR_SELF_MODIFYING_CODE | 3387 S_ATTR_SOME_INSTRUCTIONS)) 3388 type = eSymbolTypeData; 3389 else 3390 type = eSymbolTypeCode; 3391 } else if (symbol_section->IsDescendant( 3392 data_section_sp.get()) || 3393 symbol_section->IsDescendant( 3394 data_dirty_section_sp.get()) || 3395 symbol_section->IsDescendant( 3396 data_const_section_sp.get())) { 3397 if (symbol_sect_name && 3398 ::strstr(symbol_sect_name, "__objc") == 3399 symbol_sect_name) { 3400 type = eSymbolTypeRuntime; 3401 3402 if (symbol_name) { 3403 llvm::StringRef symbol_name_ref( 3404 symbol_name); 3405 if (symbol_name_ref.startswith("_OBJC_")) { 3406 static const llvm::StringRef 3407 g_objc_v2_prefix_class( 3408 "_OBJC_CLASS_$_"); 3409 static const llvm::StringRef 3410 g_objc_v2_prefix_metaclass( 3411 "_OBJC_METACLASS_$_"); 3412 static const llvm::StringRef 3413 g_objc_v2_prefix_ivar( 3414 "_OBJC_IVAR_$_"); 3415 if (symbol_name_ref.startswith( 3416 g_objc_v2_prefix_class)) { 3417 symbol_name_non_abi_mangled = 3418 symbol_name + 1; 3419 symbol_name = 3420 symbol_name + 3421 g_objc_v2_prefix_class.size(); 3422 type = eSymbolTypeObjCClass; 3423 demangled_is_synthesized = true; 3424 } else if ( 3425 symbol_name_ref.startswith( 3426 g_objc_v2_prefix_metaclass)) { 3427 symbol_name_non_abi_mangled = 3428 symbol_name + 1; 3429 symbol_name = 3430 symbol_name + 3431 g_objc_v2_prefix_metaclass.size(); 3432 type = eSymbolTypeObjCMetaClass; 3433 demangled_is_synthesized = true; 3434 } else if (symbol_name_ref.startswith( 3435 g_objc_v2_prefix_ivar)) { 3436 symbol_name_non_abi_mangled = 3437 symbol_name + 1; 3438 symbol_name = 3439 symbol_name + 3440 g_objc_v2_prefix_ivar.size(); 3441 type = eSymbolTypeObjCIVar; 3442 demangled_is_synthesized = true; 3443 } 3444 } 3445 } 3446 } else if (symbol_sect_name && 3447 ::strstr(symbol_sect_name, 3448 "__gcc_except_tab") == 3449 symbol_sect_name) { 3450 type = eSymbolTypeException; 3451 } else { 3452 type = eSymbolTypeData; 3453 } 3454 } else if (symbol_sect_name && 3455 ::strstr(symbol_sect_name, 3456 "__IMPORT") == 3457 symbol_sect_name) { 3458 type = eSymbolTypeTrampoline; 3459 } else if (symbol_section->IsDescendant( 3460 objc_section_sp.get())) { 3461 type = eSymbolTypeRuntime; 3462 if (symbol_name && symbol_name[0] == '.') { 3463 llvm::StringRef symbol_name_ref(symbol_name); 3464 static const llvm::StringRef 3465 g_objc_v1_prefix_class( 3466 ".objc_class_name_"); 3467 if (symbol_name_ref.startswith( 3468 g_objc_v1_prefix_class)) { 3469 symbol_name_non_abi_mangled = symbol_name; 3470 symbol_name = symbol_name + 3471 g_objc_v1_prefix_class.size(); 3472 type = eSymbolTypeObjCClass; 3473 demangled_is_synthesized = true; 3474 } 3475 } 3476 } 3477 } 3478 } 3479 } break; 3480 } 3481 } 3482 3483 if (add_nlist) { 3484 uint64_t symbol_value = nlist.n_value; 3485 if (symbol_name_non_abi_mangled) { 3486 sym[sym_idx].GetMangled().SetMangledName( 3487 ConstString(symbol_name_non_abi_mangled)); 3488 sym[sym_idx].GetMangled().SetDemangledName( 3489 ConstString(symbol_name)); 3490 } else { 3491 bool symbol_name_is_mangled = false; 3492 3493 if (symbol_name && symbol_name[0] == '_') { 3494 symbol_name_is_mangled = symbol_name[1] == '_'; 3495 symbol_name++; // Skip the leading underscore 3496 } 3497 3498 if (symbol_name) { 3499 ConstString const_symbol_name(symbol_name); 3500 sym[sym_idx].GetMangled().SetValue( 3501 const_symbol_name, symbol_name_is_mangled); 3502 if (is_gsym && is_debug) { 3503 const char *gsym_name = 3504 sym[sym_idx] 3505 .GetMangled() 3506 .GetName(lldb::eLanguageTypeUnknown, 3507 Mangled::ePreferMangled) 3508 .GetCString(); 3509 if (gsym_name) 3510 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3511 } 3512 } 3513 } 3514 if (symbol_section) { 3515 const addr_t section_file_addr = 3516 symbol_section->GetFileAddress(); 3517 if (symbol_byte_size == 0 && 3518 function_starts_count > 0) { 3519 addr_t symbol_lookup_file_addr = nlist.n_value; 3520 // Do an exact address match for non-ARM addresses, 3521 // else get the closest since the symbol might be a 3522 // thumb symbol which has an address with bit zero 3523 // set 3524 FunctionStarts::Entry *func_start_entry = 3525 function_starts.FindEntry( 3526 symbol_lookup_file_addr, !is_arm); 3527 if (is_arm && func_start_entry) { 3528 // Verify that the function start address is the 3529 // symbol address (ARM) or the symbol address + 1 3530 // (thumb) 3531 if (func_start_entry->addr != 3532 symbol_lookup_file_addr && 3533 func_start_entry->addr != 3534 (symbol_lookup_file_addr + 1)) { 3535 // Not the right entry, NULL it out... 3536 func_start_entry = NULL; 3537 } 3538 } 3539 if (func_start_entry) { 3540 func_start_entry->data = true; 3541 3542 addr_t symbol_file_addr = func_start_entry->addr; 3543 uint32_t symbol_flags = 0; 3544 if (is_arm) { 3545 if (symbol_file_addr & 1) 3546 symbol_flags = 3547 MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3548 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3549 } 3550 3551 const FunctionStarts::Entry 3552 *next_func_start_entry = 3553 function_starts.FindNextEntry( 3554 func_start_entry); 3555 const addr_t section_end_file_addr = 3556 section_file_addr + 3557 symbol_section->GetByteSize(); 3558 if (next_func_start_entry) { 3559 addr_t next_symbol_file_addr = 3560 next_func_start_entry->addr; 3561 // Be sure the clear the Thumb address bit when 3562 // we calculate the size from the current and 3563 // next address 3564 if (is_arm) 3565 next_symbol_file_addr &= 3566 THUMB_ADDRESS_BIT_MASK; 3567 symbol_byte_size = std::min<lldb::addr_t>( 3568 next_symbol_file_addr - symbol_file_addr, 3569 section_end_file_addr - symbol_file_addr); 3570 } else { 3571 symbol_byte_size = 3572 section_end_file_addr - symbol_file_addr; 3573 } 3574 } 3575 } 3576 symbol_value -= section_file_addr; 3577 } 3578 3579 if (is_debug == false) { 3580 if (type == eSymbolTypeCode) { 3581 // See if we can find a N_FUN entry for any code 3582 // symbols. If we do find a match, and the name 3583 // matches, then we can merge the two into just the 3584 // function symbol to avoid duplicate entries in 3585 // the symbol table 3586 std::pair<ValueToSymbolIndexMap::const_iterator, 3587 ValueToSymbolIndexMap::const_iterator> 3588 range; 3589 range = N_FUN_addr_to_sym_idx.equal_range( 3590 nlist.n_value); 3591 if (range.first != range.second) { 3592 bool found_it = false; 3593 for (ValueToSymbolIndexMap::const_iterator pos = 3594 range.first; 3595 pos != range.second; ++pos) { 3596 if (sym[sym_idx].GetMangled().GetName( 3597 lldb::eLanguageTypeUnknown, 3598 Mangled::ePreferMangled) == 3599 sym[pos->second].GetMangled().GetName( 3600 lldb::eLanguageTypeUnknown, 3601 Mangled::ePreferMangled)) { 3602 m_nlist_idx_to_sym_idx[nlist_idx] = 3603 pos->second; 3604 // We just need the flags from the linker 3605 // symbol, so put these flags 3606 // into the N_FUN flags to avoid duplicate 3607 // symbols in the symbol table 3608 sym[pos->second].SetExternal( 3609 sym[sym_idx].IsExternal()); 3610 sym[pos->second].SetFlags(nlist.n_type << 16 | 3611 nlist.n_desc); 3612 if (resolver_addresses.find(nlist.n_value) != 3613 resolver_addresses.end()) 3614 sym[pos->second].SetType( 3615 eSymbolTypeResolver); 3616 sym[sym_idx].Clear(); 3617 found_it = true; 3618 break; 3619 } 3620 } 3621 if (found_it) 3622 continue; 3623 } else { 3624 if (resolver_addresses.find(nlist.n_value) != 3625 resolver_addresses.end()) 3626 type = eSymbolTypeResolver; 3627 } 3628 } else if (type == eSymbolTypeData || 3629 type == eSymbolTypeObjCClass || 3630 type == eSymbolTypeObjCMetaClass || 3631 type == eSymbolTypeObjCIVar) { 3632 // See if we can find a N_STSYM entry for any data 3633 // symbols. If we do find a match, and the name 3634 // matches, then we can merge the two into just the 3635 // Static symbol to avoid duplicate entries in the 3636 // symbol table 3637 std::pair<ValueToSymbolIndexMap::const_iterator, 3638 ValueToSymbolIndexMap::const_iterator> 3639 range; 3640 range = N_STSYM_addr_to_sym_idx.equal_range( 3641 nlist.n_value); 3642 if (range.first != range.second) { 3643 bool found_it = false; 3644 for (ValueToSymbolIndexMap::const_iterator pos = 3645 range.first; 3646 pos != range.second; ++pos) { 3647 if (sym[sym_idx].GetMangled().GetName( 3648 lldb::eLanguageTypeUnknown, 3649 Mangled::ePreferMangled) == 3650 sym[pos->second].GetMangled().GetName( 3651 lldb::eLanguageTypeUnknown, 3652 Mangled::ePreferMangled)) { 3653 m_nlist_idx_to_sym_idx[nlist_idx] = 3654 pos->second; 3655 // We just need the flags from the linker 3656 // symbol, so put these flags 3657 // into the N_STSYM flags to avoid duplicate 3658 // symbols in the symbol table 3659 sym[pos->second].SetExternal( 3660 sym[sym_idx].IsExternal()); 3661 sym[pos->second].SetFlags(nlist.n_type << 16 | 3662 nlist.n_desc); 3663 sym[sym_idx].Clear(); 3664 found_it = true; 3665 break; 3666 } 3667 } 3668 if (found_it) 3669 continue; 3670 } else { 3671 const char *gsym_name = 3672 sym[sym_idx] 3673 .GetMangled() 3674 .GetName(lldb::eLanguageTypeUnknown, 3675 Mangled::ePreferMangled) 3676 .GetCString(); 3677 if (gsym_name) { 3678 // Combine N_GSYM stab entries with the non 3679 // stab symbol 3680 ConstNameToSymbolIndexMap::const_iterator pos = 3681 N_GSYM_name_to_sym_idx.find(gsym_name); 3682 if (pos != N_GSYM_name_to_sym_idx.end()) { 3683 const uint32_t GSYM_sym_idx = pos->second; 3684 m_nlist_idx_to_sym_idx[nlist_idx] = 3685 GSYM_sym_idx; 3686 // Copy the address, because often the N_GSYM 3687 // address has an invalid address of zero 3688 // when the global is a common symbol 3689 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3690 symbol_section); 3691 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3692 symbol_value); 3693 // We just need the flags from the linker 3694 // symbol, so put these flags 3695 // into the N_GSYM flags to avoid duplicate 3696 // symbols in the symbol table 3697 sym[GSYM_sym_idx].SetFlags( 3698 nlist.n_type << 16 | nlist.n_desc); 3699 sym[sym_idx].Clear(); 3700 continue; 3701 } 3702 } 3703 } 3704 } 3705 } 3706 3707 sym[sym_idx].SetID(nlist_idx); 3708 sym[sym_idx].SetType(type); 3709 if (set_value) { 3710 sym[sym_idx].GetAddressRef().SetSection( 3711 symbol_section); 3712 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3713 } 3714 sym[sym_idx].SetFlags(nlist.n_type << 16 | 3715 nlist.n_desc); 3716 3717 if (symbol_byte_size > 0) 3718 sym[sym_idx].SetByteSize(symbol_byte_size); 3719 3720 if (demangled_is_synthesized) 3721 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3722 ++sym_idx; 3723 } else { 3724 sym[sym_idx].Clear(); 3725 } 3726 } 3727 ///////////////////////////// 3728 } 3729 break; // No more entries to consider 3730 } 3731 } 3732 3733 for (const auto &pos : reexport_shlib_needs_fixup) { 3734 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3735 if (undef_pos != undefined_name_to_desc.end()) { 3736 const uint8_t dylib_ordinal = 3737 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3738 if (dylib_ordinal > 0 && 3739 dylib_ordinal < dylib_files.GetSize()) 3740 sym[pos.first].SetReExportedSymbolSharedLibrary( 3741 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3742 } 3743 } 3744 } 3745 } 3746 } 3747 } 3748 } 3749 3750 // Must reset this in case it was mutated above! 3751 nlist_data_offset = 0; 3752 #endif 3753 3754 if (nlist_data.GetByteSize() > 0) { 3755 3756 // If the sym array was not created while parsing the DSC unmapped 3757 // symbols, create it now. 3758 if (sym == nullptr) { 3759 sym = symtab->Resize(symtab_load_command.nsyms + 3760 m_dysymtab.nindirectsyms); 3761 num_syms = symtab->GetNumSymbols(); 3762 } 3763 3764 if (unmapped_local_symbols_found) { 3765 assert(m_dysymtab.ilocalsym == 0); 3766 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3767 nlist_idx = m_dysymtab.nlocalsym; 3768 } else { 3769 nlist_idx = 0; 3770 } 3771 3772 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3773 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3774 UndefinedNameToDescMap undefined_name_to_desc; 3775 SymbolIndexToName reexport_shlib_needs_fixup; 3776 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 3777 struct nlist_64 nlist; 3778 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, 3779 nlist_byte_size)) 3780 break; 3781 3782 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3783 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 3784 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 3785 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 3786 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 3787 3788 SymbolType type = eSymbolTypeInvalid; 3789 const char *symbol_name = nullptr; 3790 3791 if (have_strtab_data) { 3792 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3793 3794 if (symbol_name == nullptr) { 3795 // No symbol should be NULL, even the symbols with no string values 3796 // should have an offset zero which points to an empty C-string 3797 Host::SystemLog(Host::eSystemLogError, 3798 "error: symbol[%u] has invalid string table offset " 3799 "0x%x in %s, ignoring symbol\n", 3800 nlist_idx, nlist.n_strx, 3801 module_sp->GetFileSpec().GetPath().c_str()); 3802 continue; 3803 } 3804 if (symbol_name[0] == '\0') 3805 symbol_name = nullptr; 3806 } else { 3807 const addr_t str_addr = strtab_addr + nlist.n_strx; 3808 Status str_error; 3809 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3810 str_error)) 3811 symbol_name = memory_symbol_name.c_str(); 3812 } 3813 const char *symbol_name_non_abi_mangled = nullptr; 3814 3815 SectionSP symbol_section; 3816 lldb::addr_t symbol_byte_size = 0; 3817 bool add_nlist = true; 3818 bool is_gsym = false; 3819 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3820 bool demangled_is_synthesized = false; 3821 bool set_value = true; 3822 assert(sym_idx < num_syms); 3823 3824 sym[sym_idx].SetDebug(is_debug); 3825 3826 if (is_debug) { 3827 switch (nlist.n_type) { 3828 case N_GSYM: 3829 // global symbol: name,,NO_SECT,type,0 3830 // Sometimes the N_GSYM value contains the address. 3831 3832 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3833 // the ObjC data. They 3834 // have the same address, but we want to ensure that we always find 3835 // only the real symbol, 'cause we don't currently correctly 3836 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3837 // type. This is a temporary hack to make sure the ObjectiveC 3838 // symbols get treated correctly. To do this right, we should 3839 // coalesce all the GSYM & global symbols that have the same 3840 // address. 3841 is_gsym = true; 3842 sym[sym_idx].SetExternal(true); 3843 3844 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3845 llvm::StringRef symbol_name_ref(symbol_name); 3846 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3847 symbol_name_non_abi_mangled = symbol_name + 1; 3848 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3849 type = eSymbolTypeObjCClass; 3850 demangled_is_synthesized = true; 3851 3852 } else if (symbol_name_ref.startswith( 3853 g_objc_v2_prefix_metaclass)) { 3854 symbol_name_non_abi_mangled = symbol_name + 1; 3855 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3856 type = eSymbolTypeObjCMetaClass; 3857 demangled_is_synthesized = true; 3858 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3859 symbol_name_non_abi_mangled = symbol_name + 1; 3860 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3861 type = eSymbolTypeObjCIVar; 3862 demangled_is_synthesized = true; 3863 } 3864 } else { 3865 if (nlist.n_value != 0) 3866 symbol_section = 3867 section_info.GetSection(nlist.n_sect, nlist.n_value); 3868 type = eSymbolTypeData; 3869 } 3870 break; 3871 3872 case N_FNAME: 3873 // procedure name (f77 kludge): name,,NO_SECT,0,0 3874 type = eSymbolTypeCompiler; 3875 break; 3876 3877 case N_FUN: 3878 // procedure: name,,n_sect,linenumber,address 3879 if (symbol_name) { 3880 type = eSymbolTypeCode; 3881 symbol_section = 3882 section_info.GetSection(nlist.n_sect, nlist.n_value); 3883 3884 N_FUN_addr_to_sym_idx.insert( 3885 std::make_pair(nlist.n_value, sym_idx)); 3886 // We use the current number of symbols in the symbol table in 3887 // lieu of using nlist_idx in case we ever start trimming entries 3888 // out 3889 N_FUN_indexes.push_back(sym_idx); 3890 } else { 3891 type = eSymbolTypeCompiler; 3892 3893 if (!N_FUN_indexes.empty()) { 3894 // Copy the size of the function into the original STAB entry 3895 // so we don't have to hunt for it later 3896 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3897 ->SetByteSize(nlist.n_value); 3898 N_FUN_indexes.pop_back(); 3899 // We don't really need the end function STAB as it contains 3900 // the size which we already placed with the original symbol, 3901 // so don't add it if we want a minimal symbol table 3902 add_nlist = false; 3903 } 3904 } 3905 break; 3906 3907 case N_STSYM: 3908 // static symbol: name,,n_sect,type,address 3909 N_STSYM_addr_to_sym_idx.insert( 3910 std::make_pair(nlist.n_value, sym_idx)); 3911 symbol_section = 3912 section_info.GetSection(nlist.n_sect, nlist.n_value); 3913 if (symbol_name && symbol_name[0]) { 3914 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3915 eSymbolTypeData); 3916 } 3917 break; 3918 3919 case N_LCSYM: 3920 // .lcomm symbol: name,,n_sect,type,address 3921 symbol_section = 3922 section_info.GetSection(nlist.n_sect, nlist.n_value); 3923 type = eSymbolTypeCommonBlock; 3924 break; 3925 3926 case N_BNSYM: 3927 // We use the current number of symbols in the symbol table in lieu 3928 // of using nlist_idx in case we ever start trimming entries out 3929 // Skip these if we want minimal symbol tables 3930 add_nlist = false; 3931 break; 3932 3933 case N_ENSYM: 3934 // Set the size of the N_BNSYM to the terminating index of this 3935 // N_ENSYM so that we can always skip the entire symbol if we need 3936 // to navigate more quickly at the source level when parsing STABS 3937 // Skip these if we want minimal symbol tables 3938 add_nlist = false; 3939 break; 3940 3941 case N_OPT: 3942 // emitted with gcc2_compiled and in gcc source 3943 type = eSymbolTypeCompiler; 3944 break; 3945 3946 case N_RSYM: 3947 // register sym: name,,NO_SECT,type,register 3948 type = eSymbolTypeVariable; 3949 break; 3950 3951 case N_SLINE: 3952 // src line: 0,,n_sect,linenumber,address 3953 symbol_section = 3954 section_info.GetSection(nlist.n_sect, nlist.n_value); 3955 type = eSymbolTypeLineEntry; 3956 break; 3957 3958 case N_SSYM: 3959 // structure elt: name,,NO_SECT,type,struct_offset 3960 type = eSymbolTypeVariableType; 3961 break; 3962 3963 case N_SO: 3964 // source file name 3965 type = eSymbolTypeSourceFile; 3966 if (symbol_name == nullptr) { 3967 add_nlist = false; 3968 if (N_SO_index != UINT32_MAX) { 3969 // Set the size of the N_SO to the terminating index of this 3970 // N_SO so that we can always skip the entire N_SO if we need 3971 // to navigate more quickly at the source level when parsing 3972 // STABS 3973 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3974 symbol_ptr->SetByteSize(sym_idx); 3975 symbol_ptr->SetSizeIsSibling(true); 3976 } 3977 N_NSYM_indexes.clear(); 3978 N_INCL_indexes.clear(); 3979 N_BRAC_indexes.clear(); 3980 N_COMM_indexes.clear(); 3981 N_FUN_indexes.clear(); 3982 N_SO_index = UINT32_MAX; 3983 } else { 3984 // We use the current number of symbols in the symbol table in 3985 // lieu of using nlist_idx in case we ever start trimming entries 3986 // out 3987 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3988 if (N_SO_has_full_path) { 3989 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3990 // We have two consecutive N_SO entries where the first 3991 // contains a directory and the second contains a full path. 3992 sym[sym_idx - 1].GetMangled().SetValue( 3993 ConstString(symbol_name), false); 3994 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3995 add_nlist = false; 3996 } else { 3997 // This is the first entry in a N_SO that contains a 3998 // directory or a full path to the source file 3999 N_SO_index = sym_idx; 4000 } 4001 } else if ((N_SO_index == sym_idx - 1) && 4002 ((sym_idx - 1) < num_syms)) { 4003 // This is usually the second N_SO entry that contains just the 4004 // filename, so here we combine it with the first one if we are 4005 // minimizing the symbol table 4006 const char *so_path = 4007 sym[sym_idx - 1] 4008 .GetMangled() 4009 .GetDemangledName(lldb::eLanguageTypeUnknown) 4010 .AsCString(); 4011 if (so_path && so_path[0]) { 4012 std::string full_so_path(so_path); 4013 const size_t double_slash_pos = full_so_path.find("//"); 4014 if (double_slash_pos != std::string::npos) { 4015 // The linker has been generating bad N_SO entries with 4016 // doubled up paths in the format "%s%s" where the first 4017 // string in the DW_AT_comp_dir, and the second is the 4018 // directory for the source file so you end up with a path 4019 // that looks like "/tmp/src//tmp/src/" 4020 FileSpec so_dir(so_path); 4021 if (!FileSystem::Instance().Exists(so_dir)) { 4022 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 4023 FileSpec::Style::native); 4024 if (FileSystem::Instance().Exists(so_dir)) { 4025 // Trim off the incorrect path 4026 full_so_path.erase(0, double_slash_pos + 1); 4027 } 4028 } 4029 } 4030 if (*full_so_path.rbegin() != '/') 4031 full_so_path += '/'; 4032 full_so_path += symbol_name; 4033 sym[sym_idx - 1].GetMangled().SetValue( 4034 ConstString(full_so_path.c_str()), false); 4035 add_nlist = false; 4036 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4037 } 4038 } else { 4039 // This could be a relative path to a N_SO 4040 N_SO_index = sym_idx; 4041 } 4042 } 4043 break; 4044 4045 case N_OSO: 4046 // object file name: name,,0,0,st_mtime 4047 type = eSymbolTypeObjectFile; 4048 break; 4049 4050 case N_LSYM: 4051 // local sym: name,,NO_SECT,type,offset 4052 type = eSymbolTypeLocal; 4053 break; 4054 4055 // INCL scopes 4056 case N_BINCL: 4057 // include file beginning: name,,NO_SECT,0,sum We use the current 4058 // number of symbols in the symbol table in lieu of using nlist_idx 4059 // in case we ever start trimming entries out 4060 N_INCL_indexes.push_back(sym_idx); 4061 type = eSymbolTypeScopeBegin; 4062 break; 4063 4064 case N_EINCL: 4065 // include file end: name,,NO_SECT,0,0 4066 // Set the size of the N_BINCL to the terminating index of this 4067 // N_EINCL so that we can always skip the entire symbol if we need 4068 // to navigate more quickly at the source level when parsing STABS 4069 if (!N_INCL_indexes.empty()) { 4070 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4071 symbol_ptr->SetByteSize(sym_idx + 1); 4072 symbol_ptr->SetSizeIsSibling(true); 4073 N_INCL_indexes.pop_back(); 4074 } 4075 type = eSymbolTypeScopeEnd; 4076 break; 4077 4078 case N_SOL: 4079 // #included file name: name,,n_sect,0,address 4080 type = eSymbolTypeHeaderFile; 4081 4082 // We currently don't use the header files on darwin 4083 add_nlist = false; 4084 break; 4085 4086 case N_PARAMS: 4087 // compiler parameters: name,,NO_SECT,0,0 4088 type = eSymbolTypeCompiler; 4089 break; 4090 4091 case N_VERSION: 4092 // compiler version: name,,NO_SECT,0,0 4093 type = eSymbolTypeCompiler; 4094 break; 4095 4096 case N_OLEVEL: 4097 // compiler -O level: name,,NO_SECT,0,0 4098 type = eSymbolTypeCompiler; 4099 break; 4100 4101 case N_PSYM: 4102 // parameter: name,,NO_SECT,type,offset 4103 type = eSymbolTypeVariable; 4104 break; 4105 4106 case N_ENTRY: 4107 // alternate entry: name,,n_sect,linenumber,address 4108 symbol_section = 4109 section_info.GetSection(nlist.n_sect, nlist.n_value); 4110 type = eSymbolTypeLineEntry; 4111 break; 4112 4113 // Left and Right Braces 4114 case N_LBRAC: 4115 // left bracket: 0,,NO_SECT,nesting level,address We use the 4116 // current number of symbols in the symbol table in lieu of using 4117 // nlist_idx in case we ever start trimming entries out 4118 symbol_section = 4119 section_info.GetSection(nlist.n_sect, nlist.n_value); 4120 N_BRAC_indexes.push_back(sym_idx); 4121 type = eSymbolTypeScopeBegin; 4122 break; 4123 4124 case N_RBRAC: 4125 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4126 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4127 // can always skip the entire symbol if we need to navigate more 4128 // quickly at the source level when parsing STABS 4129 symbol_section = 4130 section_info.GetSection(nlist.n_sect, nlist.n_value); 4131 if (!N_BRAC_indexes.empty()) { 4132 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4133 symbol_ptr->SetByteSize(sym_idx + 1); 4134 symbol_ptr->SetSizeIsSibling(true); 4135 N_BRAC_indexes.pop_back(); 4136 } 4137 type = eSymbolTypeScopeEnd; 4138 break; 4139 4140 case N_EXCL: 4141 // deleted include file: name,,NO_SECT,0,sum 4142 type = eSymbolTypeHeaderFile; 4143 break; 4144 4145 // COMM scopes 4146 case N_BCOMM: 4147 // begin common: name,,NO_SECT,0,0 4148 // We use the current number of symbols in the symbol table in lieu 4149 // of using nlist_idx in case we ever start trimming entries out 4150 type = eSymbolTypeScopeBegin; 4151 N_COMM_indexes.push_back(sym_idx); 4152 break; 4153 4154 case N_ECOML: 4155 // end common (local name): 0,,n_sect,0,address 4156 symbol_section = 4157 section_info.GetSection(nlist.n_sect, nlist.n_value); 4158 LLVM_FALLTHROUGH; 4159 4160 case N_ECOMM: 4161 // end common: name,,n_sect,0,0 4162 // Set the size of the N_BCOMM to the terminating index of this 4163 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4164 // we need to navigate more quickly at the source level when 4165 // parsing STABS 4166 if (!N_COMM_indexes.empty()) { 4167 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4168 symbol_ptr->SetByteSize(sym_idx + 1); 4169 symbol_ptr->SetSizeIsSibling(true); 4170 N_COMM_indexes.pop_back(); 4171 } 4172 type = eSymbolTypeScopeEnd; 4173 break; 4174 4175 case N_LENG: 4176 // second stab entry with length information 4177 type = eSymbolTypeAdditional; 4178 break; 4179 4180 default: 4181 break; 4182 } 4183 } else { 4184 // uint8_t n_pext = N_PEXT & nlist.n_type; 4185 uint8_t n_type = N_TYPE & nlist.n_type; 4186 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4187 4188 switch (n_type) { 4189 case N_INDR: { 4190 const char *reexport_name_cstr = 4191 strtab_data.PeekCStr(nlist.n_value); 4192 if (reexport_name_cstr && reexport_name_cstr[0]) { 4193 type = eSymbolTypeReExported; 4194 ConstString reexport_name( 4195 reexport_name_cstr + 4196 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4197 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4198 set_value = false; 4199 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4200 indirect_symbol_names.insert( 4201 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4202 } else 4203 type = eSymbolTypeUndefined; 4204 } break; 4205 4206 case N_UNDF: 4207 if (symbol_name && symbol_name[0]) { 4208 ConstString undefined_name(symbol_name + 4209 ((symbol_name[0] == '_') ? 1 : 0)); 4210 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4211 } 4212 LLVM_FALLTHROUGH; 4213 4214 case N_PBUD: 4215 type = eSymbolTypeUndefined; 4216 break; 4217 4218 case N_ABS: 4219 type = eSymbolTypeAbsolute; 4220 break; 4221 4222 case N_SECT: { 4223 symbol_section = 4224 section_info.GetSection(nlist.n_sect, nlist.n_value); 4225 4226 if (!symbol_section) { 4227 // TODO: warn about this? 4228 add_nlist = false; 4229 break; 4230 } 4231 4232 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4233 type = eSymbolTypeException; 4234 } else { 4235 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4236 4237 switch (section_type) { 4238 case S_CSTRING_LITERALS: 4239 type = eSymbolTypeData; 4240 break; // section with only literal C strings 4241 case S_4BYTE_LITERALS: 4242 type = eSymbolTypeData; 4243 break; // section with only 4 byte literals 4244 case S_8BYTE_LITERALS: 4245 type = eSymbolTypeData; 4246 break; // section with only 8 byte literals 4247 case S_LITERAL_POINTERS: 4248 type = eSymbolTypeTrampoline; 4249 break; // section with only pointers to literals 4250 case S_NON_LAZY_SYMBOL_POINTERS: 4251 type = eSymbolTypeTrampoline; 4252 break; // section with only non-lazy symbol pointers 4253 case S_LAZY_SYMBOL_POINTERS: 4254 type = eSymbolTypeTrampoline; 4255 break; // section with only lazy symbol pointers 4256 case S_SYMBOL_STUBS: 4257 type = eSymbolTypeTrampoline; 4258 break; // section with only symbol stubs, byte size of stub in 4259 // the reserved2 field 4260 case S_MOD_INIT_FUNC_POINTERS: 4261 type = eSymbolTypeCode; 4262 break; // section with only function pointers for initialization 4263 case S_MOD_TERM_FUNC_POINTERS: 4264 type = eSymbolTypeCode; 4265 break; // section with only function pointers for termination 4266 case S_INTERPOSING: 4267 type = eSymbolTypeTrampoline; 4268 break; // section with only pairs of function pointers for 4269 // interposing 4270 case S_16BYTE_LITERALS: 4271 type = eSymbolTypeData; 4272 break; // section with only 16 byte literals 4273 case S_DTRACE_DOF: 4274 type = eSymbolTypeInstrumentation; 4275 break; 4276 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4277 type = eSymbolTypeTrampoline; 4278 break; 4279 default: 4280 switch (symbol_section->GetType()) { 4281 case lldb::eSectionTypeCode: 4282 type = eSymbolTypeCode; 4283 break; 4284 case eSectionTypeData: 4285 case eSectionTypeDataCString: // Inlined C string data 4286 case eSectionTypeDataCStringPointers: // Pointers to C string 4287 // data 4288 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4289 // the symbol table 4290 case eSectionTypeData4: 4291 case eSectionTypeData8: 4292 case eSectionTypeData16: 4293 type = eSymbolTypeData; 4294 break; 4295 default: 4296 break; 4297 } 4298 break; 4299 } 4300 4301 if (type == eSymbolTypeInvalid) { 4302 const char *symbol_sect_name = 4303 symbol_section->GetName().AsCString(); 4304 if (symbol_section->IsDescendant(text_section_sp.get())) { 4305 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4306 S_ATTR_SELF_MODIFYING_CODE | 4307 S_ATTR_SOME_INSTRUCTIONS)) 4308 type = eSymbolTypeData; 4309 else 4310 type = eSymbolTypeCode; 4311 } else if (symbol_section->IsDescendant( 4312 data_section_sp.get()) || 4313 symbol_section->IsDescendant( 4314 data_dirty_section_sp.get()) || 4315 symbol_section->IsDescendant( 4316 data_const_section_sp.get())) { 4317 if (symbol_sect_name && 4318 ::strstr(symbol_sect_name, "__objc") == 4319 symbol_sect_name) { 4320 type = eSymbolTypeRuntime; 4321 4322 if (symbol_name) { 4323 llvm::StringRef symbol_name_ref(symbol_name); 4324 if (symbol_name_ref.startswith("_OBJC_")) { 4325 static const llvm::StringRef g_objc_v2_prefix_class( 4326 "_OBJC_CLASS_$_"); 4327 static const llvm::StringRef g_objc_v2_prefix_metaclass( 4328 "_OBJC_METACLASS_$_"); 4329 static const llvm::StringRef g_objc_v2_prefix_ivar( 4330 "_OBJC_IVAR_$_"); 4331 if (symbol_name_ref.startswith( 4332 g_objc_v2_prefix_class)) { 4333 symbol_name_non_abi_mangled = symbol_name + 1; 4334 symbol_name = 4335 symbol_name + g_objc_v2_prefix_class.size(); 4336 type = eSymbolTypeObjCClass; 4337 demangled_is_synthesized = true; 4338 } else if (symbol_name_ref.startswith( 4339 g_objc_v2_prefix_metaclass)) { 4340 symbol_name_non_abi_mangled = symbol_name + 1; 4341 symbol_name = 4342 symbol_name + g_objc_v2_prefix_metaclass.size(); 4343 type = eSymbolTypeObjCMetaClass; 4344 demangled_is_synthesized = true; 4345 } else if (symbol_name_ref.startswith( 4346 g_objc_v2_prefix_ivar)) { 4347 symbol_name_non_abi_mangled = symbol_name + 1; 4348 symbol_name = 4349 symbol_name + g_objc_v2_prefix_ivar.size(); 4350 type = eSymbolTypeObjCIVar; 4351 demangled_is_synthesized = true; 4352 } 4353 } 4354 } 4355 } else if (symbol_sect_name && 4356 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4357 symbol_sect_name) { 4358 type = eSymbolTypeException; 4359 } else { 4360 type = eSymbolTypeData; 4361 } 4362 } else if (symbol_sect_name && 4363 ::strstr(symbol_sect_name, "__IMPORT") == 4364 symbol_sect_name) { 4365 type = eSymbolTypeTrampoline; 4366 } else if (symbol_section->IsDescendant( 4367 objc_section_sp.get())) { 4368 type = eSymbolTypeRuntime; 4369 if (symbol_name && symbol_name[0] == '.') { 4370 llvm::StringRef symbol_name_ref(symbol_name); 4371 static const llvm::StringRef g_objc_v1_prefix_class( 4372 ".objc_class_name_"); 4373 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4374 symbol_name_non_abi_mangled = symbol_name; 4375 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4376 type = eSymbolTypeObjCClass; 4377 demangled_is_synthesized = true; 4378 } 4379 } 4380 } 4381 } 4382 } 4383 } break; 4384 } 4385 } 4386 4387 if (add_nlist) { 4388 uint64_t symbol_value = nlist.n_value; 4389 4390 if (symbol_name_non_abi_mangled) { 4391 sym[sym_idx].GetMangled().SetMangledName( 4392 ConstString(symbol_name_non_abi_mangled)); 4393 sym[sym_idx].GetMangled().SetDemangledName( 4394 ConstString(symbol_name)); 4395 } else { 4396 bool symbol_name_is_mangled = false; 4397 4398 if (symbol_name && symbol_name[0] == '_') { 4399 symbol_name_is_mangled = symbol_name[1] == '_'; 4400 symbol_name++; // Skip the leading underscore 4401 } 4402 4403 if (symbol_name) { 4404 ConstString const_symbol_name(symbol_name); 4405 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4406 symbol_name_is_mangled); 4407 } 4408 } 4409 4410 if (is_gsym) { 4411 const char *gsym_name = sym[sym_idx] 4412 .GetMangled() 4413 .GetName(lldb::eLanguageTypeUnknown, 4414 Mangled::ePreferMangled) 4415 .GetCString(); 4416 if (gsym_name) 4417 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4418 } 4419 4420 if (symbol_section) { 4421 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4422 if (symbol_byte_size == 0 && function_starts_count > 0) { 4423 addr_t symbol_lookup_file_addr = nlist.n_value; 4424 // Do an exact address match for non-ARM addresses, else get the 4425 // closest since the symbol might be a thumb symbol which has an 4426 // address with bit zero set 4427 FunctionStarts::Entry *func_start_entry = 4428 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4429 if (is_arm && func_start_entry) { 4430 // Verify that the function start address is the symbol address 4431 // (ARM) or the symbol address + 1 (thumb) 4432 if (func_start_entry->addr != symbol_lookup_file_addr && 4433 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4434 // Not the right entry, NULL it out... 4435 func_start_entry = nullptr; 4436 } 4437 } 4438 if (func_start_entry) { 4439 func_start_entry->data = true; 4440 4441 addr_t symbol_file_addr = func_start_entry->addr; 4442 if (is_arm) 4443 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4444 4445 const FunctionStarts::Entry *next_func_start_entry = 4446 function_starts.FindNextEntry(func_start_entry); 4447 const addr_t section_end_file_addr = 4448 section_file_addr + symbol_section->GetByteSize(); 4449 if (next_func_start_entry) { 4450 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4451 // Be sure the clear the Thumb address bit when we calculate 4452 // the size from the current and next address 4453 if (is_arm) 4454 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4455 symbol_byte_size = std::min<lldb::addr_t>( 4456 next_symbol_file_addr - symbol_file_addr, 4457 section_end_file_addr - symbol_file_addr); 4458 } else { 4459 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4460 } 4461 } 4462 } 4463 symbol_value -= section_file_addr; 4464 } 4465 4466 if (!is_debug) { 4467 if (type == eSymbolTypeCode) { 4468 // See if we can find a N_FUN entry for any code symbols. If we 4469 // do find a match, and the name matches, then we can merge the 4470 // two into just the function symbol to avoid duplicate entries 4471 // in the symbol table 4472 std::pair<ValueToSymbolIndexMap::const_iterator, 4473 ValueToSymbolIndexMap::const_iterator> 4474 range; 4475 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4476 if (range.first != range.second) { 4477 bool found_it = false; 4478 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4479 pos != range.second; ++pos) { 4480 if (sym[sym_idx].GetMangled().GetName( 4481 lldb::eLanguageTypeUnknown, 4482 Mangled::ePreferMangled) == 4483 sym[pos->second].GetMangled().GetName( 4484 lldb::eLanguageTypeUnknown, 4485 Mangled::ePreferMangled)) { 4486 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4487 // We just need the flags from the linker symbol, so put 4488 // these flags into the N_FUN flags to avoid duplicate 4489 // symbols in the symbol table 4490 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4491 sym[pos->second].SetFlags(nlist.n_type << 16 | 4492 nlist.n_desc); 4493 if (resolver_addresses.find(nlist.n_value) != 4494 resolver_addresses.end()) 4495 sym[pos->second].SetType(eSymbolTypeResolver); 4496 sym[sym_idx].Clear(); 4497 found_it = true; 4498 break; 4499 } 4500 } 4501 if (found_it) 4502 continue; 4503 } else { 4504 if (resolver_addresses.find(nlist.n_value) != 4505 resolver_addresses.end()) 4506 type = eSymbolTypeResolver; 4507 } 4508 } else if (type == eSymbolTypeData || 4509 type == eSymbolTypeObjCClass || 4510 type == eSymbolTypeObjCMetaClass || 4511 type == eSymbolTypeObjCIVar) { 4512 // See if we can find a N_STSYM entry for any data symbols. If we 4513 // do find a match, and the name matches, then we can merge the 4514 // two into just the Static symbol to avoid duplicate entries in 4515 // the symbol table 4516 std::pair<ValueToSymbolIndexMap::const_iterator, 4517 ValueToSymbolIndexMap::const_iterator> 4518 range; 4519 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4520 if (range.first != range.second) { 4521 bool found_it = false; 4522 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4523 pos != range.second; ++pos) { 4524 if (sym[sym_idx].GetMangled().GetName( 4525 lldb::eLanguageTypeUnknown, 4526 Mangled::ePreferMangled) == 4527 sym[pos->second].GetMangled().GetName( 4528 lldb::eLanguageTypeUnknown, 4529 Mangled::ePreferMangled)) { 4530 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4531 // We just need the flags from the linker symbol, so put 4532 // these flags into the N_STSYM flags to avoid duplicate 4533 // symbols in the symbol table 4534 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4535 sym[pos->second].SetFlags(nlist.n_type << 16 | 4536 nlist.n_desc); 4537 sym[sym_idx].Clear(); 4538 found_it = true; 4539 break; 4540 } 4541 } 4542 if (found_it) 4543 continue; 4544 } else { 4545 // Combine N_GSYM stab entries with the non stab symbol 4546 const char *gsym_name = sym[sym_idx] 4547 .GetMangled() 4548 .GetName(lldb::eLanguageTypeUnknown, 4549 Mangled::ePreferMangled) 4550 .GetCString(); 4551 if (gsym_name) { 4552 ConstNameToSymbolIndexMap::const_iterator pos = 4553 N_GSYM_name_to_sym_idx.find(gsym_name); 4554 if (pos != N_GSYM_name_to_sym_idx.end()) { 4555 const uint32_t GSYM_sym_idx = pos->second; 4556 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4557 // Copy the address, because often the N_GSYM address has 4558 // an invalid address of zero when the global is a common 4559 // symbol 4560 sym[GSYM_sym_idx].GetAddressRef().SetSection( 4561 symbol_section); 4562 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4563 // We just need the flags from the linker symbol, so put 4564 // these flags into the N_GSYM flags to avoid duplicate 4565 // symbols in the symbol table 4566 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 4567 nlist.n_desc); 4568 sym[sym_idx].Clear(); 4569 continue; 4570 } 4571 } 4572 } 4573 } 4574 } 4575 4576 sym[sym_idx].SetID(nlist_idx); 4577 sym[sym_idx].SetType(type); 4578 if (set_value) { 4579 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4580 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4581 } 4582 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4583 if (nlist.n_desc & N_WEAK_REF) 4584 sym[sym_idx].SetIsWeak(true); 4585 4586 if (symbol_byte_size > 0) 4587 sym[sym_idx].SetByteSize(symbol_byte_size); 4588 4589 if (demangled_is_synthesized) 4590 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4591 4592 ++sym_idx; 4593 } else { 4594 sym[sym_idx].Clear(); 4595 } 4596 } 4597 4598 for (const auto &pos : reexport_shlib_needs_fixup) { 4599 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4600 if (undef_pos != undefined_name_to_desc.end()) { 4601 const uint8_t dylib_ordinal = 4602 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4603 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4604 sym[pos.first].SetReExportedSymbolSharedLibrary( 4605 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4606 } 4607 } 4608 } 4609 4610 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4611 4612 if (function_starts_count > 0) { 4613 uint32_t num_synthetic_function_symbols = 0; 4614 for (i = 0; i < function_starts_count; ++i) { 4615 if (!function_starts.GetEntryRef(i).data) 4616 ++num_synthetic_function_symbols; 4617 } 4618 4619 if (num_synthetic_function_symbols > 0) { 4620 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4621 num_syms = sym_idx + num_synthetic_function_symbols; 4622 sym = symtab->Resize(num_syms); 4623 } 4624 for (i = 0; i < function_starts_count; ++i) { 4625 const FunctionStarts::Entry *func_start_entry = 4626 function_starts.GetEntryAtIndex(i); 4627 if (!func_start_entry->data) { 4628 addr_t symbol_file_addr = func_start_entry->addr; 4629 uint32_t symbol_flags = 0; 4630 if (is_arm) { 4631 if (symbol_file_addr & 1) 4632 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4633 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4634 } 4635 Address symbol_addr; 4636 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4637 SectionSP symbol_section(symbol_addr.GetSection()); 4638 uint32_t symbol_byte_size = 0; 4639 if (symbol_section) { 4640 const addr_t section_file_addr = 4641 symbol_section->GetFileAddress(); 4642 const FunctionStarts::Entry *next_func_start_entry = 4643 function_starts.FindNextEntry(func_start_entry); 4644 const addr_t section_end_file_addr = 4645 section_file_addr + symbol_section->GetByteSize(); 4646 if (next_func_start_entry) { 4647 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4648 if (is_arm) 4649 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4650 symbol_byte_size = std::min<lldb::addr_t>( 4651 next_symbol_file_addr - symbol_file_addr, 4652 section_end_file_addr - symbol_file_addr); 4653 } else { 4654 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4655 } 4656 sym[sym_idx].SetID(synthetic_sym_id++); 4657 sym[sym_idx].GetMangled().SetDemangledName( 4658 GetNextSyntheticSymbolName()); 4659 sym[sym_idx].SetType(eSymbolTypeCode); 4660 sym[sym_idx].SetIsSynthetic(true); 4661 sym[sym_idx].GetAddressRef() = symbol_addr; 4662 if (symbol_flags) 4663 sym[sym_idx].SetFlags(symbol_flags); 4664 if (symbol_byte_size) 4665 sym[sym_idx].SetByteSize(symbol_byte_size); 4666 ++sym_idx; 4667 } 4668 } 4669 } 4670 } 4671 } 4672 } 4673 4674 // Trim our symbols down to just what we ended up with after removing any 4675 // symbols. 4676 if (sym_idx < num_syms) { 4677 num_syms = sym_idx; 4678 sym = symtab->Resize(num_syms); 4679 } 4680 4681 // Now synthesize indirect symbols 4682 if (m_dysymtab.nindirectsyms != 0) { 4683 if (indirect_symbol_index_data.GetByteSize()) { 4684 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4685 m_nlist_idx_to_sym_idx.end(); 4686 4687 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4688 ++sect_idx) { 4689 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4690 S_SYMBOL_STUBS) { 4691 uint32_t symbol_stub_byte_size = 4692 m_mach_sections[sect_idx].reserved2; 4693 if (symbol_stub_byte_size == 0) 4694 continue; 4695 4696 const uint32_t num_symbol_stubs = 4697 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4698 4699 if (num_symbol_stubs == 0) 4700 continue; 4701 4702 const uint32_t symbol_stub_index_offset = 4703 m_mach_sections[sect_idx].reserved1; 4704 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; 4705 ++stub_idx) { 4706 const uint32_t symbol_stub_index = 4707 symbol_stub_index_offset + stub_idx; 4708 const lldb::addr_t symbol_stub_addr = 4709 m_mach_sections[sect_idx].addr + 4710 (stub_idx * symbol_stub_byte_size); 4711 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4712 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4713 symbol_stub_offset, 4)) { 4714 const uint32_t stub_sym_id = 4715 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4716 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4717 continue; 4718 4719 NListIndexToSymbolIndexMap::const_iterator index_pos = 4720 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4721 Symbol *stub_symbol = nullptr; 4722 if (index_pos != end_index_pos) { 4723 // We have a remapping from the original nlist index to a 4724 // current symbol index, so just look this up by index 4725 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4726 } else { 4727 // We need to lookup a symbol using the original nlist symbol 4728 // index since this index is coming from the S_SYMBOL_STUBS 4729 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4730 } 4731 4732 if (stub_symbol) { 4733 Address so_addr(symbol_stub_addr, section_list); 4734 4735 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4736 // Change the external symbol into a trampoline that makes 4737 // sense These symbols were N_UNDF N_EXT, and are useless 4738 // to us, so we can re-use them so we don't have to make up 4739 // a synthetic symbol for no good reason. 4740 if (resolver_addresses.find(symbol_stub_addr) == 4741 resolver_addresses.end()) 4742 stub_symbol->SetType(eSymbolTypeTrampoline); 4743 else 4744 stub_symbol->SetType(eSymbolTypeResolver); 4745 stub_symbol->SetExternal(false); 4746 stub_symbol->GetAddressRef() = so_addr; 4747 stub_symbol->SetByteSize(symbol_stub_byte_size); 4748 } else { 4749 // Make a synthetic symbol to describe the trampoline stub 4750 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4751 if (sym_idx >= num_syms) { 4752 sym = symtab->Resize(++num_syms); 4753 stub_symbol = nullptr; // this pointer no longer valid 4754 } 4755 sym[sym_idx].SetID(synthetic_sym_id++); 4756 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4757 if (resolver_addresses.find(symbol_stub_addr) == 4758 resolver_addresses.end()) 4759 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4760 else 4761 sym[sym_idx].SetType(eSymbolTypeResolver); 4762 sym[sym_idx].SetIsSynthetic(true); 4763 sym[sym_idx].GetAddressRef() = so_addr; 4764 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4765 ++sym_idx; 4766 } 4767 } else { 4768 if (log) 4769 log->Warning("symbol stub referencing symbol table symbol " 4770 "%u that isn't in our minimal symbol table, " 4771 "fix this!!!", 4772 stub_sym_id); 4773 } 4774 } 4775 } 4776 } 4777 } 4778 } 4779 } 4780 4781 if (!trie_entries.empty()) { 4782 for (const auto &e : trie_entries) { 4783 if (e.entry.import_name) { 4784 // Only add indirect symbols from the Trie entries if we didn't have 4785 // a N_INDR nlist entry for this already 4786 if (indirect_symbol_names.find(e.entry.name) == 4787 indirect_symbol_names.end()) { 4788 // Make a synthetic symbol to describe re-exported symbol. 4789 if (sym_idx >= num_syms) 4790 sym = symtab->Resize(++num_syms); 4791 sym[sym_idx].SetID(synthetic_sym_id++); 4792 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4793 sym[sym_idx].SetType(eSymbolTypeReExported); 4794 sym[sym_idx].SetIsSynthetic(true); 4795 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4796 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4797 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4798 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4799 } 4800 ++sym_idx; 4801 } 4802 } 4803 } 4804 } 4805 4806 // StreamFile s(stdout, false); 4807 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4808 // symtab->Dump(&s, NULL, eSortOrderNone); 4809 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4810 // sizes 4811 symtab->CalculateSymbolSizes(); 4812 4813 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4814 // symtab->Dump(&s, NULL, eSortOrderNone); 4815 4816 return symtab->GetNumSymbols(); 4817 } 4818 return 0; 4819 } 4820 4821 void ObjectFileMachO::Dump(Stream *s) { 4822 ModuleSP module_sp(GetModule()); 4823 if (module_sp) { 4824 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4825 s->Printf("%p: ", static_cast<void *>(this)); 4826 s->Indent(); 4827 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4828 s->PutCString("ObjectFileMachO64"); 4829 else 4830 s->PutCString("ObjectFileMachO32"); 4831 4832 ArchSpec header_arch = GetArchitecture(); 4833 4834 *s << ", file = '" << m_file 4835 << "', triple = " << header_arch.GetTriple().getTriple() << "\n"; 4836 4837 SectionList *sections = GetSectionList(); 4838 if (sections) 4839 sections->Dump(s, nullptr, true, UINT32_MAX); 4840 4841 if (m_symtab_up) 4842 m_symtab_up->Dump(s, nullptr, eSortOrderNone); 4843 } 4844 } 4845 4846 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4847 const lldb_private::DataExtractor &data, 4848 lldb::offset_t lc_offset) { 4849 uint32_t i; 4850 struct uuid_command load_cmd; 4851 4852 lldb::offset_t offset = lc_offset; 4853 for (i = 0; i < header.ncmds; ++i) { 4854 const lldb::offset_t cmd_offset = offset; 4855 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4856 break; 4857 4858 if (load_cmd.cmd == LC_UUID) { 4859 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4860 4861 if (uuid_bytes) { 4862 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4863 // We pretend these object files have no UUID to prevent crashing. 4864 4865 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4866 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4867 0xbb, 0x14, 0xf0, 0x0d}; 4868 4869 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4870 return UUID(); 4871 4872 return UUID::fromOptionalData(uuid_bytes, 16); 4873 } 4874 return UUID(); 4875 } 4876 offset = cmd_offset + load_cmd.cmdsize; 4877 } 4878 return UUID(); 4879 } 4880 4881 static llvm::StringRef GetOSName(uint32_t cmd) { 4882 switch (cmd) { 4883 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4884 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4885 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4886 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4887 case llvm::MachO::LC_VERSION_MIN_TVOS: 4888 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4889 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4890 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4891 default: 4892 llvm_unreachable("unexpected LC_VERSION load command"); 4893 } 4894 } 4895 4896 namespace { 4897 struct OSEnv { 4898 llvm::StringRef os_type; 4899 llvm::StringRef environment; 4900 OSEnv(uint32_t cmd) { 4901 switch (cmd) { 4902 case PLATFORM_MACOS: 4903 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4904 return; 4905 case PLATFORM_IOS: 4906 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4907 return; 4908 case PLATFORM_TVOS: 4909 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4910 return; 4911 case PLATFORM_WATCHOS: 4912 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4913 return; 4914 // NEED_BRIDGEOS_TRIPLE case PLATFORM_BRIDGEOS: 4915 // NEED_BRIDGEOS_TRIPLE os_type = llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4916 // NEED_BRIDGEOS_TRIPLE return; 4917 #if defined (PLATFORM_IOSSIMULATOR) && defined (PLATFORM_TVOSSIMULATOR) && defined (PLATFORM_WATCHOSSIMULATOR) 4918 case PLATFORM_IOSSIMULATOR: 4919 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4920 environment = 4921 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4922 return; 4923 case PLATFORM_TVOSSIMULATOR: 4924 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4925 environment = 4926 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4927 return; 4928 case PLATFORM_WATCHOSSIMULATOR: 4929 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4930 environment = 4931 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4932 return; 4933 #endif 4934 default: { 4935 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 4936 LIBLLDB_LOG_PROCESS)); 4937 LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION"); 4938 } 4939 } 4940 } 4941 }; 4942 4943 struct MinOS { 4944 uint32_t major_version, minor_version, patch_version; 4945 MinOS(uint32_t version) 4946 : major_version(version >> 16), 4947 minor_version((version >> 8) & 0xffu), 4948 patch_version(version & 0xffu) {} 4949 }; 4950 } // namespace 4951 4952 ArchSpec 4953 ObjectFileMachO::GetArchitecture(const llvm::MachO::mach_header &header, 4954 const lldb_private::DataExtractor &data, 4955 lldb::offset_t lc_offset) { 4956 ArchSpec arch; 4957 arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4958 4959 if (arch.IsValid()) { 4960 llvm::Triple &triple = arch.GetTriple(); 4961 4962 // Set OS to an unspecified unknown or a "*" so it can match any OS 4963 triple.setOS(llvm::Triple::UnknownOS); 4964 triple.setOSName(llvm::StringRef()); 4965 4966 if (header.filetype == MH_PRELOAD) { 4967 if (header.cputype == CPU_TYPE_ARM) { 4968 // If this is a 32-bit arm binary, and it's a standalone binary, force 4969 // the Vendor to Apple so we don't accidentally pick up the generic 4970 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4971 // frame pointer register; most other armv7 ABIs use a combination of 4972 // r7 and r11. 4973 triple.setVendor(llvm::Triple::Apple); 4974 } else { 4975 // Set vendor to an unspecified unknown or a "*" so it can match any 4976 // vendor This is required for correct behavior of EFI debugging on 4977 // x86_64 4978 triple.setVendor(llvm::Triple::UnknownVendor); 4979 triple.setVendorName(llvm::StringRef()); 4980 } 4981 return arch; 4982 } else { 4983 struct load_command load_cmd; 4984 llvm::SmallString<16> os_name; 4985 llvm::raw_svector_ostream os(os_name); 4986 4987 // See if there is an LC_VERSION_MIN_* load command that can give 4988 // us the OS type. 4989 lldb::offset_t offset = lc_offset; 4990 for (uint32_t i = 0; i < header.ncmds; ++i) { 4991 const lldb::offset_t cmd_offset = offset; 4992 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4993 break; 4994 4995 struct version_min_command version_min; 4996 switch (load_cmd.cmd) { 4997 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4998 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4999 case llvm::MachO::LC_VERSION_MIN_TVOS: 5000 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 5001 if (load_cmd.cmdsize != sizeof(version_min)) 5002 break; 5003 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 5004 data.GetByteOrder(), &version_min) == 0) 5005 break; 5006 MinOS min_os(version_min.version); 5007 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 5008 << min_os.minor_version << '.' << min_os.patch_version; 5009 triple.setOSName(os.str()); 5010 return arch; 5011 } 5012 default: 5013 break; 5014 } 5015 5016 offset = cmd_offset + load_cmd.cmdsize; 5017 } 5018 5019 // See if there is an LC_BUILD_VERSION load command that can give 5020 // us the OS type. 5021 5022 offset = lc_offset; 5023 for (uint32_t i = 0; i < header.ncmds; ++i) { 5024 const lldb::offset_t cmd_offset = offset; 5025 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 5026 break; 5027 do { 5028 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 5029 struct build_version_command build_version; 5030 if (load_cmd.cmdsize < sizeof(build_version)) { 5031 // Malformed load command. 5032 break; 5033 } 5034 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 5035 data.GetByteOrder(), &build_version) == 0) 5036 break; 5037 MinOS min_os(build_version.minos); 5038 OSEnv os_env(build_version.platform); 5039 if (os_env.os_type.empty()) 5040 break; 5041 os << os_env.os_type << min_os.major_version << '.' 5042 << min_os.minor_version << '.' << min_os.patch_version; 5043 triple.setOSName(os.str()); 5044 if (!os_env.environment.empty()) 5045 triple.setEnvironmentName(os_env.environment); 5046 return arch; 5047 } 5048 } while (false); 5049 offset = cmd_offset + load_cmd.cmdsize; 5050 } 5051 5052 if (header.filetype != MH_KEXT_BUNDLE) { 5053 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 5054 // so lets not say our Vendor is Apple, leave it as an unspecified 5055 // unknown 5056 triple.setVendor(llvm::Triple::UnknownVendor); 5057 triple.setVendorName(llvm::StringRef()); 5058 } 5059 } 5060 } 5061 return arch; 5062 } 5063 5064 UUID ObjectFileMachO::GetUUID() { 5065 ModuleSP module_sp(GetModule()); 5066 if (module_sp) { 5067 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5068 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5069 return GetUUID(m_header, m_data, offset); 5070 } 5071 return UUID(); 5072 } 5073 5074 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5075 uint32_t count = 0; 5076 ModuleSP module_sp(GetModule()); 5077 if (module_sp) { 5078 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5079 struct load_command load_cmd; 5080 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5081 std::vector<std::string> rpath_paths; 5082 std::vector<std::string> rpath_relative_paths; 5083 std::vector<std::string> at_exec_relative_paths; 5084 uint32_t i; 5085 for (i = 0; i < m_header.ncmds; ++i) { 5086 const uint32_t cmd_offset = offset; 5087 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5088 break; 5089 5090 switch (load_cmd.cmd) { 5091 case LC_RPATH: 5092 case LC_LOAD_DYLIB: 5093 case LC_LOAD_WEAK_DYLIB: 5094 case LC_REEXPORT_DYLIB: 5095 case LC_LOAD_DYLINKER: 5096 case LC_LOADFVMLIB: 5097 case LC_LOAD_UPWARD_DYLIB: { 5098 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5099 const char *path = m_data.PeekCStr(name_offset); 5100 if (path) { 5101 if (load_cmd.cmd == LC_RPATH) 5102 rpath_paths.push_back(path); 5103 else { 5104 if (path[0] == '@') { 5105 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5106 rpath_relative_paths.push_back(path + strlen("@rpath")); 5107 else if (strncmp(path, "@executable_path", 5108 strlen("@executable_path")) == 0) 5109 at_exec_relative_paths.push_back(path 5110 + strlen("@executable_path")); 5111 } else { 5112 FileSpec file_spec(path); 5113 if (files.AppendIfUnique(file_spec)) 5114 count++; 5115 } 5116 } 5117 } 5118 } break; 5119 5120 default: 5121 break; 5122 } 5123 offset = cmd_offset + load_cmd.cmdsize; 5124 } 5125 5126 FileSpec this_file_spec(m_file); 5127 FileSystem::Instance().Resolve(this_file_spec); 5128 5129 if (!rpath_paths.empty()) { 5130 // Fixup all LC_RPATH values to be absolute paths 5131 std::string loader_path("@loader_path"); 5132 std::string executable_path("@executable_path"); 5133 for (auto &rpath : rpath_paths) { 5134 if (rpath.find(loader_path) == 0) { 5135 rpath.erase(0, loader_path.size()); 5136 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5137 } else if (rpath.find(executable_path) == 0) { 5138 rpath.erase(0, executable_path.size()); 5139 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5140 } 5141 } 5142 5143 for (const auto &rpath_relative_path : rpath_relative_paths) { 5144 for (const auto &rpath : rpath_paths) { 5145 std::string path = rpath; 5146 path += rpath_relative_path; 5147 // It is OK to resolve this path because we must find a file on disk 5148 // for us to accept it anyway if it is rpath relative. 5149 FileSpec file_spec(path); 5150 FileSystem::Instance().Resolve(file_spec); 5151 if (FileSystem::Instance().Exists(file_spec) && 5152 files.AppendIfUnique(file_spec)) { 5153 count++; 5154 break; 5155 } 5156 } 5157 } 5158 } 5159 5160 // We may have @executable_paths but no RPATHS. Figure those out here. 5161 // Only do this if this object file is the executable. We have no way to 5162 // get back to the actual executable otherwise, so we won't get the right 5163 // path. 5164 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5165 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5166 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5167 FileSpec file_spec = 5168 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5169 if (FileSystem::Instance().Exists(file_spec) && 5170 files.AppendIfUnique(file_spec)) 5171 count++; 5172 } 5173 } 5174 } 5175 return count; 5176 } 5177 5178 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5179 // If the object file is not an executable it can't hold the entry point. 5180 // m_entry_point_address is initialized to an invalid address, so we can just 5181 // return that. If m_entry_point_address is valid it means we've found it 5182 // already, so return the cached value. 5183 5184 if ((!IsExecutable() && !IsDynamicLoader()) || 5185 m_entry_point_address.IsValid()) { 5186 return m_entry_point_address; 5187 } 5188 5189 // Otherwise, look for the UnixThread or Thread command. The data for the 5190 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5191 // 5192 // uint32_t flavor - this is the flavor argument you would pass to 5193 // thread_get_state 5194 // uint32_t count - this is the count of longs in the thread state data 5195 // struct XXX_thread_state state - this is the structure from 5196 // <machine/thread_status.h> corresponding to the flavor. 5197 // <repeat this trio> 5198 // 5199 // So we just keep reading the various register flavors till we find the GPR 5200 // one, then read the PC out of there. 5201 // FIXME: We will need to have a "RegisterContext data provider" class at some 5202 // point that can get all the registers 5203 // out of data in this form & attach them to a given thread. That should 5204 // underlie the MacOS X User process plugin, and we'll also need it for the 5205 // MacOS X Core File process plugin. When we have that we can also use it 5206 // here. 5207 // 5208 // For now we hard-code the offsets and flavors we need: 5209 // 5210 // 5211 5212 ModuleSP module_sp(GetModule()); 5213 if (module_sp) { 5214 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5215 struct load_command load_cmd; 5216 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5217 uint32_t i; 5218 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5219 bool done = false; 5220 5221 for (i = 0; i < m_header.ncmds; ++i) { 5222 const lldb::offset_t cmd_offset = offset; 5223 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5224 break; 5225 5226 switch (load_cmd.cmd) { 5227 case LC_UNIXTHREAD: 5228 case LC_THREAD: { 5229 while (offset < cmd_offset + load_cmd.cmdsize) { 5230 uint32_t flavor = m_data.GetU32(&offset); 5231 uint32_t count = m_data.GetU32(&offset); 5232 if (count == 0) { 5233 // We've gotten off somehow, log and exit; 5234 return m_entry_point_address; 5235 } 5236 5237 switch (m_header.cputype) { 5238 case llvm::MachO::CPU_TYPE_ARM: 5239 if (flavor == 1 || 5240 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 from 5241 // mach/arm/thread_status.h 5242 { 5243 offset += 60; // This is the offset of pc in the GPR thread state 5244 // data structure. 5245 start_address = m_data.GetU32(&offset); 5246 done = true; 5247 } 5248 break; 5249 case llvm::MachO::CPU_TYPE_ARM64: 5250 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5251 { 5252 offset += 256; // This is the offset of pc in the GPR thread state 5253 // data structure. 5254 start_address = m_data.GetU64(&offset); 5255 done = true; 5256 } 5257 break; 5258 case llvm::MachO::CPU_TYPE_I386: 5259 if (flavor == 5260 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5261 { 5262 offset += 40; // This is the offset of eip in the GPR thread state 5263 // data structure. 5264 start_address = m_data.GetU32(&offset); 5265 done = true; 5266 } 5267 break; 5268 case llvm::MachO::CPU_TYPE_X86_64: 5269 if (flavor == 5270 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5271 { 5272 offset += 16 * 8; // This is the offset of rip in the GPR thread 5273 // state data structure. 5274 start_address = m_data.GetU64(&offset); 5275 done = true; 5276 } 5277 break; 5278 default: 5279 return m_entry_point_address; 5280 } 5281 // Haven't found the GPR flavor yet, skip over the data for this 5282 // flavor: 5283 if (done) 5284 break; 5285 offset += count * 4; 5286 } 5287 } break; 5288 case LC_MAIN: { 5289 ConstString text_segment_name("__TEXT"); 5290 uint64_t entryoffset = m_data.GetU64(&offset); 5291 SectionSP text_segment_sp = 5292 GetSectionList()->FindSectionByName(text_segment_name); 5293 if (text_segment_sp) { 5294 done = true; 5295 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5296 } 5297 } break; 5298 5299 default: 5300 break; 5301 } 5302 if (done) 5303 break; 5304 5305 // Go to the next load command: 5306 offset = cmd_offset + load_cmd.cmdsize; 5307 } 5308 5309 if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) { 5310 if (GetSymtab()) { 5311 Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType( 5312 ConstString("_dyld_start"), SymbolType::eSymbolTypeCode, 5313 Symtab::eDebugAny, Symtab::eVisibilityAny); 5314 if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) { 5315 start_address = dyld_start_sym->GetAddress().GetFileAddress(); 5316 } 5317 } 5318 } 5319 5320 if (start_address != LLDB_INVALID_ADDRESS) { 5321 // We got the start address from the load commands, so now resolve that 5322 // address in the sections of this ObjectFile: 5323 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5324 start_address, GetSectionList())) { 5325 m_entry_point_address.Clear(); 5326 } 5327 } else { 5328 // We couldn't read the UnixThread load command - maybe it wasn't there. 5329 // As a fallback look for the "start" symbol in the main executable. 5330 5331 ModuleSP module_sp(GetModule()); 5332 5333 if (module_sp) { 5334 SymbolContextList contexts; 5335 SymbolContext context; 5336 if (module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5337 eSymbolTypeCode, contexts)) { 5338 if (contexts.GetContextAtIndex(0, context)) 5339 m_entry_point_address = context.symbol->GetAddress(); 5340 } 5341 } 5342 } 5343 } 5344 5345 return m_entry_point_address; 5346 } 5347 5348 lldb_private::Address ObjectFileMachO::GetBaseAddress() { 5349 lldb_private::Address header_addr; 5350 SectionList *section_list = GetSectionList(); 5351 if (section_list) { 5352 SectionSP text_segment_sp( 5353 section_list->FindSectionByName(GetSegmentNameTEXT())); 5354 if (text_segment_sp) { 5355 header_addr.SetSection(text_segment_sp); 5356 header_addr.SetOffset(0); 5357 } 5358 } 5359 return header_addr; 5360 } 5361 5362 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5363 ModuleSP module_sp(GetModule()); 5364 if (module_sp) { 5365 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5366 if (!m_thread_context_offsets_valid) { 5367 m_thread_context_offsets_valid = true; 5368 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5369 FileRangeArray::Entry file_range; 5370 thread_command thread_cmd; 5371 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5372 const uint32_t cmd_offset = offset; 5373 if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr) 5374 break; 5375 5376 if (thread_cmd.cmd == LC_THREAD) { 5377 file_range.SetRangeBase(offset); 5378 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5379 m_thread_context_offsets.Append(file_range); 5380 } 5381 offset = cmd_offset + thread_cmd.cmdsize; 5382 } 5383 } 5384 } 5385 return m_thread_context_offsets.GetSize(); 5386 } 5387 5388 std::string ObjectFileMachO::GetIdentifierString() { 5389 std::string result; 5390 ModuleSP module_sp(GetModule()); 5391 if (module_sp) { 5392 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5393 5394 // First, look over the load commands for an LC_NOTE load command with 5395 // data_owner string "kern ver str" & use that if found. 5396 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5397 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5398 const uint32_t cmd_offset = offset; 5399 load_command lc; 5400 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5401 break; 5402 if (lc.cmd == LC_NOTE) 5403 { 5404 char data_owner[17]; 5405 m_data.CopyData (offset, 16, data_owner); 5406 data_owner[16] = '\0'; 5407 offset += 16; 5408 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5409 uint64_t size = m_data.GetU64_unchecked (&offset); 5410 5411 // "kern ver str" has a uint32_t version and then a nul terminated 5412 // c-string. 5413 if (strcmp ("kern ver str", data_owner) == 0) 5414 { 5415 offset = fileoff; 5416 uint32_t version; 5417 if (m_data.GetU32 (&offset, &version, 1) != nullptr) 5418 { 5419 if (version == 1) 5420 { 5421 uint32_t strsize = size - sizeof (uint32_t); 5422 char *buf = (char*) malloc (strsize); 5423 if (buf) 5424 { 5425 m_data.CopyData (offset, strsize, buf); 5426 buf[strsize - 1] = '\0'; 5427 result = buf; 5428 if (buf) 5429 free (buf); 5430 return result; 5431 } 5432 } 5433 } 5434 } 5435 } 5436 offset = cmd_offset + lc.cmdsize; 5437 } 5438 5439 // Second, make a pass over the load commands looking for an obsolete 5440 // LC_IDENT load command. 5441 offset = MachHeaderSizeFromMagic(m_header.magic); 5442 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5443 const uint32_t cmd_offset = offset; 5444 struct ident_command ident_command; 5445 if (m_data.GetU32(&offset, &ident_command, 2) == nullptr) 5446 break; 5447 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5448 char *buf = (char *) malloc (ident_command.cmdsize); 5449 if (buf != nullptr 5450 && m_data.CopyData (offset, ident_command.cmdsize, buf) == ident_command.cmdsize) { 5451 buf[ident_command.cmdsize - 1] = '\0'; 5452 result = buf; 5453 } 5454 if (buf) 5455 free (buf); 5456 } 5457 offset = cmd_offset + ident_command.cmdsize; 5458 } 5459 5460 } 5461 return result; 5462 } 5463 5464 bool ObjectFileMachO::GetCorefileMainBinaryInfo (addr_t &address, UUID &uuid) { 5465 address = LLDB_INVALID_ADDRESS; 5466 uuid.Clear(); 5467 ModuleSP module_sp(GetModule()); 5468 if (module_sp) { 5469 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5470 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5471 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5472 const uint32_t cmd_offset = offset; 5473 load_command lc; 5474 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5475 break; 5476 if (lc.cmd == LC_NOTE) 5477 { 5478 char data_owner[17]; 5479 memset (data_owner, 0, sizeof (data_owner)); 5480 m_data.CopyData (offset, 16, data_owner); 5481 offset += 16; 5482 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5483 uint64_t size = m_data.GetU64_unchecked (&offset); 5484 5485 // "main bin spec" (main binary specification) data payload is 5486 // formatted: 5487 // uint32_t version [currently 1] 5488 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user process] 5489 // uint64_t address [ UINT64_MAX if address not specified ] 5490 // uuid_t uuid [ all zero's if uuid not specified ] 5491 // uint32_t log2_pagesize [ process page size in log base 2, e.g. 4k pages are 12. 0 for unspecified ] 5492 5493 if (strcmp ("main bin spec", data_owner) == 0 && size >= 32) 5494 { 5495 offset = fileoff; 5496 uint32_t version; 5497 if (m_data.GetU32 (&offset, &version, 1) != nullptr && version == 1) 5498 { 5499 uint32_t type = 0; 5500 uuid_t raw_uuid; 5501 memset (raw_uuid, 0, sizeof (uuid_t)); 5502 5503 if (m_data.GetU32(&offset, &type, 1) && 5504 m_data.GetU64(&offset, &address, 1) && 5505 m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5506 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5507 return true; 5508 } 5509 } 5510 } 5511 } 5512 offset = cmd_offset + lc.cmdsize; 5513 } 5514 } 5515 return false; 5516 } 5517 5518 lldb::RegisterContextSP 5519 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5520 lldb_private::Thread &thread) { 5521 lldb::RegisterContextSP reg_ctx_sp; 5522 5523 ModuleSP module_sp(GetModule()); 5524 if (module_sp) { 5525 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5526 if (!m_thread_context_offsets_valid) 5527 GetNumThreadContexts(); 5528 5529 const FileRangeArray::Entry *thread_context_file_range = 5530 m_thread_context_offsets.GetEntryAtIndex(idx); 5531 if (thread_context_file_range) { 5532 5533 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5534 thread_context_file_range->GetByteSize()); 5535 5536 switch (m_header.cputype) { 5537 case llvm::MachO::CPU_TYPE_ARM64: 5538 reg_ctx_sp = 5539 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data); 5540 break; 5541 5542 case llvm::MachO::CPU_TYPE_ARM: 5543 reg_ctx_sp = 5544 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data); 5545 break; 5546 5547 case llvm::MachO::CPU_TYPE_I386: 5548 reg_ctx_sp = 5549 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data); 5550 break; 5551 5552 case llvm::MachO::CPU_TYPE_X86_64: 5553 reg_ctx_sp = 5554 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data); 5555 break; 5556 } 5557 } 5558 } 5559 return reg_ctx_sp; 5560 } 5561 5562 ObjectFile::Type ObjectFileMachO::CalculateType() { 5563 switch (m_header.filetype) { 5564 case MH_OBJECT: // 0x1u 5565 if (GetAddressByteSize() == 4) { 5566 // 32 bit kexts are just object files, but they do have a valid 5567 // UUID load command. 5568 if (GetUUID()) { 5569 // this checking for the UUID load command is not enough we could 5570 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5571 // this is required of kexts 5572 if (m_strata == eStrataInvalid) 5573 m_strata = eStrataKernel; 5574 return eTypeSharedLibrary; 5575 } 5576 } 5577 return eTypeObjectFile; 5578 5579 case MH_EXECUTE: 5580 return eTypeExecutable; // 0x2u 5581 case MH_FVMLIB: 5582 return eTypeSharedLibrary; // 0x3u 5583 case MH_CORE: 5584 return eTypeCoreFile; // 0x4u 5585 case MH_PRELOAD: 5586 return eTypeSharedLibrary; // 0x5u 5587 case MH_DYLIB: 5588 return eTypeSharedLibrary; // 0x6u 5589 case MH_DYLINKER: 5590 return eTypeDynamicLinker; // 0x7u 5591 case MH_BUNDLE: 5592 return eTypeSharedLibrary; // 0x8u 5593 case MH_DYLIB_STUB: 5594 return eTypeStubLibrary; // 0x9u 5595 case MH_DSYM: 5596 return eTypeDebugInfo; // 0xAu 5597 case MH_KEXT_BUNDLE: 5598 return eTypeSharedLibrary; // 0xBu 5599 default: 5600 break; 5601 } 5602 return eTypeUnknown; 5603 } 5604 5605 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5606 switch (m_header.filetype) { 5607 case MH_OBJECT: // 0x1u 5608 { 5609 // 32 bit kexts are just object files, but they do have a valid 5610 // UUID load command. 5611 if (GetUUID()) { 5612 // this checking for the UUID load command is not enough we could 5613 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5614 // this is required of kexts 5615 if (m_type == eTypeInvalid) 5616 m_type = eTypeSharedLibrary; 5617 5618 return eStrataKernel; 5619 } 5620 } 5621 return eStrataUnknown; 5622 5623 case MH_EXECUTE: // 0x2u 5624 // Check for the MH_DYLDLINK bit in the flags 5625 if (m_header.flags & MH_DYLDLINK) { 5626 return eStrataUser; 5627 } else { 5628 SectionList *section_list = GetSectionList(); 5629 if (section_list) { 5630 static ConstString g_kld_section_name("__KLD"); 5631 if (section_list->FindSectionByName(g_kld_section_name)) 5632 return eStrataKernel; 5633 } 5634 } 5635 return eStrataRawImage; 5636 5637 case MH_FVMLIB: 5638 return eStrataUser; // 0x3u 5639 case MH_CORE: 5640 return eStrataUnknown; // 0x4u 5641 case MH_PRELOAD: 5642 return eStrataRawImage; // 0x5u 5643 case MH_DYLIB: 5644 return eStrataUser; // 0x6u 5645 case MH_DYLINKER: 5646 return eStrataUser; // 0x7u 5647 case MH_BUNDLE: 5648 return eStrataUser; // 0x8u 5649 case MH_DYLIB_STUB: 5650 return eStrataUser; // 0x9u 5651 case MH_DSYM: 5652 return eStrataUnknown; // 0xAu 5653 case MH_KEXT_BUNDLE: 5654 return eStrataKernel; // 0xBu 5655 default: 5656 break; 5657 } 5658 return eStrataUnknown; 5659 } 5660 5661 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5662 ModuleSP module_sp(GetModule()); 5663 if (module_sp) { 5664 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5665 struct dylib_command load_cmd; 5666 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5667 uint32_t version_cmd = 0; 5668 uint64_t version = 0; 5669 uint32_t i; 5670 for (i = 0; i < m_header.ncmds; ++i) { 5671 const lldb::offset_t cmd_offset = offset; 5672 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5673 break; 5674 5675 if (load_cmd.cmd == LC_ID_DYLIB) { 5676 if (version_cmd == 0) { 5677 version_cmd = load_cmd.cmd; 5678 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr) 5679 break; 5680 version = load_cmd.dylib.current_version; 5681 } 5682 break; // Break for now unless there is another more complete version 5683 // number load command in the future. 5684 } 5685 offset = cmd_offset + load_cmd.cmdsize; 5686 } 5687 5688 if (version_cmd == LC_ID_DYLIB) { 5689 unsigned major = (version & 0xFFFF0000ull) >> 16; 5690 unsigned minor = (version & 0x0000FF00ull) >> 8; 5691 unsigned subminor = (version & 0x000000FFull); 5692 return llvm::VersionTuple(major, minor, subminor); 5693 } 5694 } 5695 return llvm::VersionTuple(); 5696 } 5697 5698 ArchSpec ObjectFileMachO::GetArchitecture() { 5699 ModuleSP module_sp(GetModule()); 5700 ArchSpec arch; 5701 if (module_sp) { 5702 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5703 5704 return GetArchitecture(m_header, m_data, 5705 MachHeaderSizeFromMagic(m_header.magic)); 5706 } 5707 return arch; 5708 } 5709 5710 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, addr_t &base_addr, UUID &uuid) { 5711 uuid.Clear(); 5712 base_addr = LLDB_INVALID_ADDRESS; 5713 if (process && process->GetDynamicLoader()) { 5714 DynamicLoader *dl = process->GetDynamicLoader(); 5715 LazyBool using_shared_cache; 5716 LazyBool private_shared_cache; 5717 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5718 private_shared_cache); 5719 } 5720 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5721 LLDB_LOGF( 5722 log, 5723 "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64, 5724 uuid.GetAsString().c_str(), base_addr); 5725 } 5726 5727 // From dyld SPI header dyld_process_info.h 5728 typedef void *dyld_process_info; 5729 struct lldb_copy__dyld_process_cache_info { 5730 uuid_t cacheUUID; // UUID of cache used by process 5731 uint64_t cacheBaseAddress; // load address of dyld shared cache 5732 bool noCache; // process is running without a dyld cache 5733 bool privateCache; // process is using a private copy of its dyld cache 5734 }; 5735 5736 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with llvm 5737 // enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile errors. 5738 // So we need to use the actual underlying types of task_t and kern_return_t 5739 // below. 5740 extern "C" unsigned int /*task_t*/ mach_task_self(); 5741 5742 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5743 uuid.Clear(); 5744 base_addr = LLDB_INVALID_ADDRESS; 5745 5746 #if defined(__APPLE__) && \ 5747 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5748 uint8_t *(*dyld_get_all_image_infos)(void); 5749 dyld_get_all_image_infos = 5750 (uint8_t * (*)())dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5751 if (dyld_get_all_image_infos) { 5752 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5753 if (dyld_all_image_infos_address) { 5754 uint32_t *version = (uint32_t *) 5755 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5756 if (*version >= 13) { 5757 uuid_t *sharedCacheUUID_address = 0; 5758 int wordsize = sizeof(uint8_t *); 5759 if (wordsize == 8) { 5760 sharedCacheUUID_address = 5761 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5762 160); // sharedCacheUUID <mach-o/dyld_images.h> 5763 if (*version >= 15) 5764 base_addr = *(uint64_t *) ((uint8_t *) dyld_all_image_infos_address 5765 + 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5766 } else { 5767 sharedCacheUUID_address = 5768 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5769 84); // sharedCacheUUID <mach-o/dyld_images.h> 5770 if (*version >= 15) { 5771 base_addr = 0; 5772 base_addr = *(uint32_t *) ((uint8_t *) dyld_all_image_infos_address 5773 + 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5774 } 5775 } 5776 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5777 } 5778 } 5779 } else { 5780 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5781 dyld_process_info (*dyld_process_info_create)(unsigned int /* task_t */ task, uint64_t timestamp, unsigned int /*kern_return_t*/ *kernelError); 5782 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5783 void (*dyld_process_info_release)(dyld_process_info info); 5784 5785 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, unsigned int /*kern_return_t*/ *)) 5786 dlsym (RTLD_DEFAULT, "_dyld_process_info_create"); 5787 dyld_process_info_get_cache = (void (*)(void *, void *)) 5788 dlsym (RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5789 dyld_process_info_release = (void (*)(void *)) 5790 dlsym (RTLD_DEFAULT, "_dyld_process_info_release"); 5791 5792 if (dyld_process_info_create && dyld_process_info_get_cache) { 5793 unsigned int /*kern_return_t */ kern_ret; 5794 dyld_process_info process_info = dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5795 if (process_info) { 5796 struct lldb_copy__dyld_process_cache_info sc_info; 5797 memset (&sc_info, 0, sizeof (struct lldb_copy__dyld_process_cache_info)); 5798 dyld_process_info_get_cache (process_info, &sc_info); 5799 if (sc_info.cacheBaseAddress != 0) { 5800 base_addr = sc_info.cacheBaseAddress; 5801 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 5802 } 5803 dyld_process_info_release (process_info); 5804 } 5805 } 5806 } 5807 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5808 if (log && uuid.IsValid()) 5809 LLDB_LOGF(log, 5810 "lldb's in-memory shared cache has a UUID of %s base address of " 5811 "0x%" PRIx64, 5812 uuid.GetAsString().c_str(), base_addr); 5813 #endif 5814 } 5815 5816 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5817 if (!m_min_os_version) { 5818 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5819 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5820 const lldb::offset_t load_cmd_offset = offset; 5821 5822 version_min_command lc; 5823 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5824 break; 5825 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5826 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5827 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5828 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5829 if (m_data.GetU32(&offset, &lc.version, 5830 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5831 const uint32_t xxxx = lc.version >> 16; 5832 const uint32_t yy = (lc.version >> 8) & 0xffu; 5833 const uint32_t zz = lc.version & 0xffu; 5834 if (xxxx) { 5835 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5836 break; 5837 } 5838 } 5839 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5840 // struct build_version_command { 5841 // uint32_t cmd; /* LC_BUILD_VERSION */ 5842 // uint32_t cmdsize; /* sizeof(struct build_version_command) plus */ 5843 // /* ntools * sizeof(struct build_tool_version) */ 5844 // uint32_t platform; /* platform */ 5845 // uint32_t minos; /* X.Y.Z is encoded in nibbles xxxx.yy.zz */ 5846 // uint32_t sdk; /* X.Y.Z is encoded in nibbles xxxx.yy.zz */ 5847 // uint32_t ntools; /* number of tool entries following this */ 5848 // }; 5849 5850 offset += 4; // skip platform 5851 uint32_t minos = m_data.GetU32(&offset); 5852 5853 const uint32_t xxxx = minos >> 16; 5854 const uint32_t yy = (minos >> 8) & 0xffu; 5855 const uint32_t zz = minos & 0xffu; 5856 if (xxxx) { 5857 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5858 break; 5859 } 5860 } 5861 5862 offset = load_cmd_offset + lc.cmdsize; 5863 } 5864 5865 if (!m_min_os_version) { 5866 // Set version to an empty value so we don't keep trying to 5867 m_min_os_version = llvm::VersionTuple(); 5868 } 5869 } 5870 5871 return *m_min_os_version; 5872 } 5873 5874 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { 5875 if (!m_sdk_versions.hasValue()) { 5876 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5877 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5878 const lldb::offset_t load_cmd_offset = offset; 5879 5880 version_min_command lc; 5881 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5882 break; 5883 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5884 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5885 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5886 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5887 if (m_data.GetU32(&offset, &lc.version, 5888 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5889 const uint32_t xxxx = lc.sdk >> 16; 5890 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5891 const uint32_t zz = lc.sdk & 0xffu; 5892 if (xxxx) { 5893 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 5894 break; 5895 } else { 5896 GetModule()->ReportWarning( 5897 "minimum OS version load command with invalid (0) version found."); 5898 } 5899 } 5900 } 5901 offset = load_cmd_offset + lc.cmdsize; 5902 } 5903 5904 if (!m_sdk_versions.hasValue()) { 5905 offset = MachHeaderSizeFromMagic(m_header.magic); 5906 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5907 const lldb::offset_t load_cmd_offset = offset; 5908 5909 version_min_command lc; 5910 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5911 break; 5912 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5913 // struct build_version_command { 5914 // uint32_t cmd; /* LC_BUILD_VERSION */ 5915 // uint32_t cmdsize; /* sizeof(struct 5916 // build_version_command) plus */ 5917 // /* ntools * sizeof(struct 5918 // build_tool_version) */ 5919 // uint32_t platform; /* platform */ 5920 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5921 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded 5922 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number 5923 // of tool entries following this */ 5924 // }; 5925 5926 offset += 4; // skip platform 5927 uint32_t minos = m_data.GetU32(&offset); 5928 5929 const uint32_t xxxx = minos >> 16; 5930 const uint32_t yy = (minos >> 8) & 0xffu; 5931 const uint32_t zz = minos & 0xffu; 5932 if (xxxx) { 5933 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 5934 break; 5935 } 5936 } 5937 offset = load_cmd_offset + lc.cmdsize; 5938 } 5939 } 5940 5941 if (!m_sdk_versions.hasValue()) 5942 m_sdk_versions = llvm::VersionTuple(); 5943 } 5944 5945 return m_sdk_versions.getValue(); 5946 } 5947 5948 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5949 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5950 } 5951 5952 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5953 return m_allow_assembly_emulation_unwind_plans; 5954 } 5955 5956 // PluginInterface protocol 5957 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5958 return GetPluginNameStatic(); 5959 } 5960 5961 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5962 5963 Section *ObjectFileMachO::GetMachHeaderSection() { 5964 // Find the first address of the mach header which is the first non-zero file 5965 // sized section whose file offset is zero. This is the base file address of 5966 // the mach-o file which can be subtracted from the vmaddr of the other 5967 // segments found in memory and added to the load address 5968 ModuleSP module_sp = GetModule(); 5969 if (!module_sp) 5970 return nullptr; 5971 SectionList *section_list = GetSectionList(); 5972 if (!section_list) 5973 return nullptr; 5974 const size_t num_sections = section_list->GetSize(); 5975 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5976 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5977 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 5978 return section; 5979 } 5980 return nullptr; 5981 } 5982 5983 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 5984 if (!section) 5985 return false; 5986 const bool is_dsym = (m_header.filetype == MH_DSYM); 5987 if (section->GetFileSize() == 0 && !is_dsym) 5988 return false; 5989 if (section->IsThreadSpecific()) 5990 return false; 5991 if (GetModule().get() != section->GetModule().get()) 5992 return false; 5993 // Be careful with __LINKEDIT and __DWARF segments 5994 if (section->GetName() == GetSegmentNameLINKEDIT() || 5995 section->GetName() == GetSegmentNameDWARF()) { 5996 // Only map __LINKEDIT and __DWARF if we have an in memory image and 5997 // this isn't a kernel binary like a kext or mach_kernel. 5998 const bool is_memory_image = (bool)m_process_wp.lock(); 5999 const Strata strata = GetStrata(); 6000 if (is_memory_image == false || strata == eStrataKernel) 6001 return false; 6002 } 6003 return true; 6004 } 6005 6006 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 6007 lldb::addr_t header_load_address, const Section *header_section, 6008 const Section *section) { 6009 ModuleSP module_sp = GetModule(); 6010 if (module_sp && header_section && section && 6011 header_load_address != LLDB_INVALID_ADDRESS) { 6012 lldb::addr_t file_addr = header_section->GetFileAddress(); 6013 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 6014 return section->GetFileAddress() - file_addr + header_load_address; 6015 } 6016 return LLDB_INVALID_ADDRESS; 6017 } 6018 6019 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 6020 bool value_is_offset) { 6021 ModuleSP module_sp = GetModule(); 6022 if (module_sp) { 6023 size_t num_loaded_sections = 0; 6024 SectionList *section_list = GetSectionList(); 6025 if (section_list) { 6026 const size_t num_sections = section_list->GetSize(); 6027 6028 if (value_is_offset) { 6029 // "value" is an offset to apply to each top level segment 6030 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6031 // Iterate through the object file sections to find all of the 6032 // sections that size on disk (to avoid __PAGEZERO) and load them 6033 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6034 if (SectionIsLoadable(section_sp.get())) 6035 if (target.GetSectionLoadList().SetSectionLoadAddress( 6036 section_sp, section_sp->GetFileAddress() + value)) 6037 ++num_loaded_sections; 6038 } 6039 } else { 6040 // "value" is the new base address of the mach_header, adjust each 6041 // section accordingly 6042 6043 Section *mach_header_section = GetMachHeaderSection(); 6044 if (mach_header_section) { 6045 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6046 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6047 6048 lldb::addr_t section_load_addr = 6049 CalculateSectionLoadAddressForMemoryImage( 6050 value, mach_header_section, section_sp.get()); 6051 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6052 if (target.GetSectionLoadList().SetSectionLoadAddress( 6053 section_sp, section_load_addr)) 6054 ++num_loaded_sections; 6055 } 6056 } 6057 } 6058 } 6059 } 6060 return num_loaded_sections > 0; 6061 } 6062 return false; 6063 } 6064 6065 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6066 const FileSpec &outfile, Status &error) { 6067 if (process_sp) { 6068 Target &target = process_sp->GetTarget(); 6069 const ArchSpec target_arch = target.GetArchitecture(); 6070 const llvm::Triple &target_triple = target_arch.GetTriple(); 6071 if (target_triple.getVendor() == llvm::Triple::Apple && 6072 (target_triple.getOS() == llvm::Triple::MacOSX || 6073 target_triple.getOS() == llvm::Triple::IOS || 6074 target_triple.getOS() == llvm::Triple::WatchOS || 6075 target_triple.getOS() == llvm::Triple::TvOS)) { 6076 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) { 6077 bool make_core = false; 6078 switch (target_arch.GetMachine()) { 6079 case llvm::Triple::aarch64: 6080 case llvm::Triple::arm: 6081 case llvm::Triple::thumb: 6082 case llvm::Triple::x86: 6083 case llvm::Triple::x86_64: 6084 make_core = true; 6085 break; 6086 default: 6087 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6088 target_triple.str().c_str()); 6089 break; 6090 } 6091 6092 if (make_core) { 6093 std::vector<segment_command_64> segment_load_commands; 6094 // uint32_t range_info_idx = 0; 6095 MemoryRegionInfo range_info; 6096 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6097 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6098 const ByteOrder byte_order = target_arch.GetByteOrder(); 6099 if (range_error.Success()) { 6100 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6101 const addr_t addr = range_info.GetRange().GetRangeBase(); 6102 const addr_t size = range_info.GetRange().GetByteSize(); 6103 6104 if (size == 0) 6105 break; 6106 6107 // Calculate correct protections 6108 uint32_t prot = 0; 6109 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6110 prot |= VM_PROT_READ; 6111 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6112 prot |= VM_PROT_WRITE; 6113 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6114 prot |= VM_PROT_EXECUTE; 6115 6116 if (prot != 0) { 6117 uint32_t cmd_type = LC_SEGMENT_64; 6118 uint32_t segment_size = sizeof(segment_command_64); 6119 if (addr_byte_size == 4) { 6120 cmd_type = LC_SEGMENT; 6121 segment_size = sizeof(segment_command); 6122 } 6123 segment_command_64 segment = { 6124 cmd_type, // uint32_t cmd; 6125 segment_size, // uint32_t cmdsize; 6126 {0}, // char segname[16]; 6127 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6128 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6129 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6130 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6131 prot, // uint32_t maxprot; 6132 prot, // uint32_t initprot; 6133 0, // uint32_t nsects; 6134 0}; // uint32_t flags; 6135 segment_load_commands.push_back(segment); 6136 } else { 6137 // No protections and a size of 1 used to be returned from old 6138 // debugservers when we asked about a region that was past the 6139 // last memory region and it indicates the end... 6140 if (size == 1) 6141 break; 6142 } 6143 6144 range_error = process_sp->GetMemoryRegionInfo( 6145 range_info.GetRange().GetRangeEnd(), range_info); 6146 if (range_error.Fail()) 6147 break; 6148 } 6149 6150 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6151 6152 mach_header_64 mach_header; 6153 if (addr_byte_size == 8) { 6154 mach_header.magic = MH_MAGIC_64; 6155 } else { 6156 mach_header.magic = MH_MAGIC; 6157 } 6158 mach_header.cputype = target_arch.GetMachOCPUType(); 6159 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6160 mach_header.filetype = MH_CORE; 6161 mach_header.ncmds = segment_load_commands.size(); 6162 mach_header.flags = 0; 6163 mach_header.reserved = 0; 6164 ThreadList &thread_list = process_sp->GetThreadList(); 6165 const uint32_t num_threads = thread_list.GetSize(); 6166 6167 // Make an array of LC_THREAD data items. Each one contains the 6168 // contents of the LC_THREAD load command. The data doesn't contain 6169 // the load command + load command size, we will add the load command 6170 // and load command size as we emit the data. 6171 std::vector<StreamString> LC_THREAD_datas(num_threads); 6172 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6173 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6174 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6175 LC_THREAD_data.SetByteOrder(byte_order); 6176 } 6177 for (uint32_t thread_idx = 0; thread_idx < num_threads; 6178 ++thread_idx) { 6179 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6180 if (thread_sp) { 6181 switch (mach_header.cputype) { 6182 case llvm::MachO::CPU_TYPE_ARM64: 6183 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6184 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6185 break; 6186 6187 case llvm::MachO::CPU_TYPE_ARM: 6188 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6189 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6190 break; 6191 6192 case llvm::MachO::CPU_TYPE_I386: 6193 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6194 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6195 break; 6196 6197 case llvm::MachO::CPU_TYPE_X86_64: 6198 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6199 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6200 break; 6201 } 6202 } 6203 } 6204 6205 // The size of the load command is the size of the segments... 6206 if (addr_byte_size == 8) { 6207 mach_header.sizeofcmds = segment_load_commands.size() * 6208 sizeof(struct segment_command_64); 6209 } else { 6210 mach_header.sizeofcmds = 6211 segment_load_commands.size() * sizeof(struct segment_command); 6212 } 6213 6214 // and the size of all LC_THREAD load command 6215 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6216 ++mach_header.ncmds; 6217 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6218 } 6219 6220 // Write the mach header 6221 buffer.PutHex32(mach_header.magic); 6222 buffer.PutHex32(mach_header.cputype); 6223 buffer.PutHex32(mach_header.cpusubtype); 6224 buffer.PutHex32(mach_header.filetype); 6225 buffer.PutHex32(mach_header.ncmds); 6226 buffer.PutHex32(mach_header.sizeofcmds); 6227 buffer.PutHex32(mach_header.flags); 6228 if (addr_byte_size == 8) { 6229 buffer.PutHex32(mach_header.reserved); 6230 } 6231 6232 // Skip the mach header and all load commands and align to the next 6233 // 0x1000 byte boundary 6234 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6235 if (file_offset & 0x00000fff) { 6236 file_offset += 0x00001000ull; 6237 file_offset &= (~0x00001000ull + 1); 6238 } 6239 6240 for (auto &segment : segment_load_commands) { 6241 segment.fileoff = file_offset; 6242 file_offset += segment.filesize; 6243 } 6244 6245 // Write out all of the LC_THREAD load commands 6246 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6247 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6248 buffer.PutHex32(LC_THREAD); 6249 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6250 buffer.Write(LC_THREAD_data.GetString().data(), 6251 LC_THREAD_data_size); 6252 } 6253 6254 // Write out all of the segment load commands 6255 for (const auto &segment : segment_load_commands) { 6256 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6257 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6258 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6259 segment.cmd, segment.cmdsize, segment.vmaddr, 6260 segment.vmaddr + segment.vmsize, segment.fileoff, 6261 segment.filesize, segment.maxprot, segment.initprot, 6262 segment.nsects, segment.flags); 6263 6264 buffer.PutHex32(segment.cmd); 6265 buffer.PutHex32(segment.cmdsize); 6266 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6267 if (addr_byte_size == 8) { 6268 buffer.PutHex64(segment.vmaddr); 6269 buffer.PutHex64(segment.vmsize); 6270 buffer.PutHex64(segment.fileoff); 6271 buffer.PutHex64(segment.filesize); 6272 } else { 6273 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6274 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6275 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6276 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6277 } 6278 buffer.PutHex32(segment.maxprot); 6279 buffer.PutHex32(segment.initprot); 6280 buffer.PutHex32(segment.nsects); 6281 buffer.PutHex32(segment.flags); 6282 } 6283 6284 File core_file; 6285 std::string core_file_path(outfile.GetPath()); 6286 error = FileSystem::Instance().Open(core_file, outfile, 6287 File::eOpenOptionWrite | 6288 File::eOpenOptionTruncate | 6289 File::eOpenOptionCanCreate); 6290 if (error.Success()) { 6291 // Read 1 page at a time 6292 uint8_t bytes[0x1000]; 6293 // Write the mach header and load commands out to the core file 6294 size_t bytes_written = buffer.GetString().size(); 6295 error = core_file.Write(buffer.GetString().data(), bytes_written); 6296 if (error.Success()) { 6297 // Now write the file data for all memory segments in the process 6298 for (const auto &segment : segment_load_commands) { 6299 if (core_file.SeekFromStart(segment.fileoff) == -1) { 6300 error.SetErrorStringWithFormat( 6301 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6302 segment.fileoff, core_file_path.c_str()); 6303 break; 6304 } 6305 6306 printf("Saving %" PRId64 6307 " bytes of data for memory region at 0x%" PRIx64 "\n", 6308 segment.vmsize, segment.vmaddr); 6309 addr_t bytes_left = segment.vmsize; 6310 addr_t addr = segment.vmaddr; 6311 Status memory_read_error; 6312 while (bytes_left > 0 && error.Success()) { 6313 const size_t bytes_to_read = 6314 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6315 6316 // In a savecore setting, we don't really care about caching, 6317 // as the data is dumped and very likely never read again, 6318 // so we call ReadMemoryFromInferior to bypass it. 6319 const size_t bytes_read = process_sp->ReadMemoryFromInferior( 6320 addr, bytes, bytes_to_read, memory_read_error); 6321 6322 if (bytes_read == bytes_to_read) { 6323 size_t bytes_written = bytes_read; 6324 error = core_file.Write(bytes, bytes_written); 6325 bytes_left -= bytes_read; 6326 addr += bytes_read; 6327 } else { 6328 // Some pages within regions are not readable, those should 6329 // be zero filled 6330 memset(bytes, 0, bytes_to_read); 6331 size_t bytes_written = bytes_to_read; 6332 error = core_file.Write(bytes, bytes_written); 6333 bytes_left -= bytes_to_read; 6334 addr += bytes_to_read; 6335 } 6336 } 6337 } 6338 } 6339 } 6340 } else { 6341 error.SetErrorString( 6342 "process doesn't support getting memory region info"); 6343 } 6344 } 6345 return true; // This is the right plug to handle saving core files for 6346 // this process 6347 } 6348 } 6349 return false; 6350 } 6351