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 log->Printf("Shared cache %s has UUID %s", dyld_shared_cache.GetPath().c_str(), 2107 dsc_uuid.GetAsString().c_str()); 2108 } 2109 return dsc_uuid; 2110 } 2111 2112 size_t ObjectFileMachO::ParseSymtab() { 2113 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2114 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2115 m_file.GetFilename().AsCString("")); 2116 ModuleSP module_sp(GetModule()); 2117 if (!module_sp) 2118 return 0; 2119 2120 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2121 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2122 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2123 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2124 FunctionStarts function_starts; 2125 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2126 uint32_t i; 2127 FileSpecList dylib_files; 2128 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2129 static const llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2130 static const llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2131 static const llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2132 2133 for (i = 0; i < m_header.ncmds; ++i) { 2134 const lldb::offset_t cmd_offset = offset; 2135 // Read in the load command and load command size 2136 struct load_command lc; 2137 if (m_data.GetU32(&offset, &lc, 2) == nullptr) 2138 break; 2139 // Watch for the symbol table load command 2140 switch (lc.cmd) { 2141 case LC_SYMTAB: 2142 symtab_load_command.cmd = lc.cmd; 2143 symtab_load_command.cmdsize = lc.cmdsize; 2144 // Read in the rest of the symtab load command 2145 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2146 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2147 return 0; 2148 if (symtab_load_command.symoff == 0) { 2149 if (log) 2150 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2151 return 0; 2152 } 2153 2154 if (symtab_load_command.stroff == 0) { 2155 if (log) 2156 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2157 return 0; 2158 } 2159 2160 if (symtab_load_command.nsyms == 0) { 2161 if (log) 2162 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2163 return 0; 2164 } 2165 2166 if (symtab_load_command.strsize == 0) { 2167 if (log) 2168 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2169 return 0; 2170 } 2171 break; 2172 2173 case LC_DYLD_INFO: 2174 case LC_DYLD_INFO_ONLY: 2175 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2176 dyld_info.cmd = lc.cmd; 2177 dyld_info.cmdsize = lc.cmdsize; 2178 } else { 2179 memset(&dyld_info, 0, sizeof(dyld_info)); 2180 } 2181 break; 2182 2183 case LC_LOAD_DYLIB: 2184 case LC_LOAD_WEAK_DYLIB: 2185 case LC_REEXPORT_DYLIB: 2186 case LC_LOADFVMLIB: 2187 case LC_LOAD_UPWARD_DYLIB: { 2188 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2189 const char *path = m_data.PeekCStr(name_offset); 2190 if (path) { 2191 FileSpec file_spec(path); 2192 // Strip the path if there is @rpath, @executable, etc so we just use 2193 // the basename 2194 if (path[0] == '@') 2195 file_spec.GetDirectory().Clear(); 2196 2197 if (lc.cmd == LC_REEXPORT_DYLIB) { 2198 m_reexported_dylibs.AppendIfUnique(file_spec); 2199 } 2200 2201 dylib_files.Append(file_spec); 2202 } 2203 } break; 2204 2205 case LC_FUNCTION_STARTS: 2206 function_starts_load_command.cmd = lc.cmd; 2207 function_starts_load_command.cmdsize = lc.cmdsize; 2208 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2209 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2210 memset(&function_starts_load_command, 0, 2211 sizeof(function_starts_load_command)); 2212 break; 2213 2214 default: 2215 break; 2216 } 2217 offset = cmd_offset + lc.cmdsize; 2218 } 2219 2220 if (symtab_load_command.cmd) { 2221 Symtab *symtab = m_symtab_up.get(); 2222 SectionList *section_list = GetSectionList(); 2223 if (section_list == nullptr) 2224 return 0; 2225 2226 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2227 const ByteOrder byte_order = m_data.GetByteOrder(); 2228 bool bit_width_32 = addr_byte_size == 4; 2229 const size_t nlist_byte_size = 2230 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2231 2232 DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size); 2233 DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size); 2234 DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size); 2235 DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order, 2236 addr_byte_size); 2237 DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size); 2238 2239 const addr_t nlist_data_byte_size = 2240 symtab_load_command.nsyms * nlist_byte_size; 2241 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2242 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2243 2244 ProcessSP process_sp(m_process_wp.lock()); 2245 Process *process = process_sp.get(); 2246 2247 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2248 2249 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2250 Target &target = process->GetTarget(); 2251 2252 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2253 2254 SectionSP linkedit_section_sp( 2255 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2256 // Reading mach file from memory in a process or core file... 2257 2258 if (linkedit_section_sp) { 2259 addr_t linkedit_load_addr = 2260 linkedit_section_sp->GetLoadBaseAddress(&target); 2261 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2262 // We might be trying to access the symbol table before the 2263 // __LINKEDIT's load address has been set in the target. We can't 2264 // fail to read the symbol table, so calculate the right address 2265 // manually 2266 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2267 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2268 } 2269 2270 const addr_t linkedit_file_offset = 2271 linkedit_section_sp->GetFileOffset(); 2272 const addr_t symoff_addr = linkedit_load_addr + 2273 symtab_load_command.symoff - 2274 linkedit_file_offset; 2275 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2276 linkedit_file_offset; 2277 2278 bool data_was_read = false; 2279 2280 #if defined(__APPLE__) && \ 2281 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2282 if (m_header.flags & 0x80000000u && 2283 process->GetAddressByteSize() == sizeof(void *)) { 2284 // This mach-o memory file is in the dyld shared cache. If this 2285 // program is not remote and this is iOS, then this process will 2286 // share the same shared cache as the process we are debugging and we 2287 // can read the entire __LINKEDIT from the address space in this 2288 // process. This is a needed optimization that is used for local iOS 2289 // debugging only since all shared libraries in the shared cache do 2290 // not have corresponding files that exist in the file system of the 2291 // device. They have been combined into a single file. This means we 2292 // always have to load these files from memory. All of the symbol and 2293 // string tables from all of the __LINKEDIT sections from the shared 2294 // libraries in the shared cache have been merged into a single large 2295 // symbol and string table. Reading all of this symbol and string 2296 // table data across can slow down debug launch times, so we optimize 2297 // this by reading the memory for the __LINKEDIT section from this 2298 // process. 2299 2300 UUID lldb_shared_cache; 2301 addr_t lldb_shared_cache_addr; 2302 GetLLDBSharedCacheUUID (lldb_shared_cache_addr, lldb_shared_cache); 2303 UUID process_shared_cache; 2304 addr_t process_shared_cache_addr; 2305 GetProcessSharedCacheUUID(process, process_shared_cache_addr, process_shared_cache); 2306 bool use_lldb_cache = true; 2307 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2308 (lldb_shared_cache != process_shared_cache 2309 || process_shared_cache_addr != lldb_shared_cache_addr)) { 2310 use_lldb_cache = false; 2311 } 2312 2313 PlatformSP platform_sp(target.GetPlatform()); 2314 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2315 data_was_read = true; 2316 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2317 eByteOrderLittle); 2318 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2319 eByteOrderLittle); 2320 if (function_starts_load_command.cmd) { 2321 const addr_t func_start_addr = 2322 linkedit_load_addr + function_starts_load_command.dataoff - 2323 linkedit_file_offset; 2324 function_starts_data.SetData( 2325 (void *)func_start_addr, 2326 function_starts_load_command.datasize, eByteOrderLittle); 2327 } 2328 } 2329 } 2330 #endif 2331 2332 if (!data_was_read) { 2333 // Always load dyld - the dynamic linker - from memory if we didn't 2334 // find a binary anywhere else. lldb will not register 2335 // dylib/framework/bundle loads/unloads if we don't have the dyld 2336 // symbols, we force dyld to load from memory despite the user's 2337 // target.memory-module-load-level setting. 2338 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2339 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2340 DataBufferSP nlist_data_sp( 2341 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2342 if (nlist_data_sp) 2343 nlist_data.SetData(nlist_data_sp, 0, 2344 nlist_data_sp->GetByteSize()); 2345 if (m_dysymtab.nindirectsyms != 0) { 2346 const addr_t indirect_syms_addr = linkedit_load_addr + 2347 m_dysymtab.indirectsymoff - 2348 linkedit_file_offset; 2349 DataBufferSP indirect_syms_data_sp( 2350 ReadMemory(process_sp, indirect_syms_addr, 2351 m_dysymtab.nindirectsyms * 4)); 2352 if (indirect_syms_data_sp) 2353 indirect_symbol_index_data.SetData( 2354 indirect_syms_data_sp, 0, 2355 indirect_syms_data_sp->GetByteSize()); 2356 // If this binary is outside the shared cache, 2357 // cache the string table. 2358 // Binaries in the shared cache all share a giant string table, and 2359 // we can't share the string tables across multiple ObjectFileMachO's, 2360 // so we'd end up re-reading this mega-strtab for every binary 2361 // in the shared cache - it would be a big perf problem. 2362 // For binaries outside the shared cache, it's faster to read the 2363 // entire strtab at once instead of piece-by-piece as we process 2364 // the nlist records. 2365 if ((m_header.flags & 0x80000000u) == 0) { 2366 DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, 2367 strtab_data_byte_size)); 2368 if (strtab_data_sp) { 2369 strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 2370 } 2371 } 2372 } 2373 } 2374 if (memory_module_load_level >= 2375 eMemoryModuleLoadLevelPartial) { 2376 if (function_starts_load_command.cmd) { 2377 const addr_t func_start_addr = 2378 linkedit_load_addr + function_starts_load_command.dataoff - 2379 linkedit_file_offset; 2380 DataBufferSP func_start_data_sp( 2381 ReadMemory(process_sp, func_start_addr, 2382 function_starts_load_command.datasize)); 2383 if (func_start_data_sp) 2384 function_starts_data.SetData(func_start_data_sp, 0, 2385 func_start_data_sp->GetByteSize()); 2386 } 2387 } 2388 } 2389 } 2390 } else { 2391 nlist_data.SetData(m_data, symtab_load_command.symoff, 2392 nlist_data_byte_size); 2393 strtab_data.SetData(m_data, symtab_load_command.stroff, 2394 strtab_data_byte_size); 2395 2396 if (dyld_info.export_size > 0) { 2397 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2398 dyld_info.export_size); 2399 } 2400 2401 if (m_dysymtab.nindirectsyms != 0) { 2402 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2403 m_dysymtab.nindirectsyms * 4); 2404 } 2405 if (function_starts_load_command.cmd) { 2406 function_starts_data.SetData(m_data, 2407 function_starts_load_command.dataoff, 2408 function_starts_load_command.datasize); 2409 } 2410 } 2411 2412 if (nlist_data.GetByteSize() == 0 && 2413 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2414 if (log) 2415 module_sp->LogMessage(log, "failed to read nlist data"); 2416 return 0; 2417 } 2418 2419 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2420 if (!have_strtab_data) { 2421 if (process) { 2422 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2423 if (log) 2424 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2425 return 0; 2426 } 2427 } else { 2428 if (log) 2429 module_sp->LogMessage(log, "failed to read strtab data"); 2430 return 0; 2431 } 2432 } 2433 2434 ConstString g_segment_name_TEXT = GetSegmentNameTEXT(); 2435 ConstString g_segment_name_DATA = GetSegmentNameDATA(); 2436 ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2437 ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2438 ConstString g_segment_name_OBJC = GetSegmentNameOBJC(); 2439 ConstString g_section_name_eh_frame = GetSectionNameEHFrame(); 2440 SectionSP text_section_sp( 2441 section_list->FindSectionByName(g_segment_name_TEXT)); 2442 SectionSP data_section_sp( 2443 section_list->FindSectionByName(g_segment_name_DATA)); 2444 SectionSP data_dirty_section_sp( 2445 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2446 SectionSP data_const_section_sp( 2447 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2448 SectionSP objc_section_sp( 2449 section_list->FindSectionByName(g_segment_name_OBJC)); 2450 SectionSP eh_frame_section_sp; 2451 if (text_section_sp.get()) 2452 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2453 g_section_name_eh_frame); 2454 else 2455 eh_frame_section_sp = 2456 section_list->FindSectionByName(g_section_name_eh_frame); 2457 2458 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2459 2460 // lldb works best if it knows the start address of all functions in a 2461 // module. Linker symbols or debug info are normally the best source of 2462 // information for start addr / size but they may be stripped in a released 2463 // binary. Two additional sources of information exist in Mach-O binaries: 2464 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2465 // function's start address in the 2466 // binary, relative to the text section. 2467 // eh_frame - the eh_frame FDEs have the start addr & size of 2468 // each function 2469 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2470 // all modern binaries. 2471 // Binaries built to run on older releases may need to use eh_frame 2472 // information. 2473 2474 if (text_section_sp && function_starts_data.GetByteSize()) { 2475 FunctionStarts::Entry function_start_entry; 2476 function_start_entry.data = false; 2477 lldb::offset_t function_start_offset = 0; 2478 function_start_entry.addr = text_section_sp->GetFileAddress(); 2479 uint64_t delta; 2480 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2481 0) { 2482 // Now append the current entry 2483 function_start_entry.addr += delta; 2484 function_starts.Append(function_start_entry); 2485 } 2486 } else { 2487 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2488 // load command claiming an eh_frame but it doesn't actually have the 2489 // eh_frame content. And if we have a dSYM, we don't need to do any of 2490 // this fill-in-the-missing-symbols works anyway - the debug info should 2491 // give us all the functions in the module. 2492 if (text_section_sp.get() && eh_frame_section_sp.get() && 2493 m_type != eTypeDebugInfo) { 2494 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2495 DWARFCallFrameInfo::EH); 2496 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2497 eh_frame.GetFunctionAddressAndSizeVector(functions); 2498 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2499 size_t count = functions.GetSize(); 2500 for (size_t i = 0; i < count; ++i) { 2501 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2502 functions.GetEntryAtIndex(i); 2503 if (func) { 2504 FunctionStarts::Entry function_start_entry; 2505 function_start_entry.addr = func->base - text_base_addr; 2506 function_starts.Append(function_start_entry); 2507 } 2508 } 2509 } 2510 } 2511 2512 const size_t function_starts_count = function_starts.GetSize(); 2513 2514 // For user process binaries (executables, dylibs, frameworks, bundles), if 2515 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2516 // going to assume the binary has been stripped. Don't allow assembly 2517 // language instruction emulation because we don't know proper function 2518 // start boundaries. 2519 // 2520 // For all other types of binaries (kernels, stand-alone bare board 2521 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2522 // sections - we should not make any assumptions about them based on that. 2523 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2524 m_allow_assembly_emulation_unwind_plans = false; 2525 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2526 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2527 2528 if (unwind_or_symbol_log) 2529 module_sp->LogMessage( 2530 unwind_or_symbol_log, 2531 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2532 } 2533 2534 const user_id_t TEXT_eh_frame_sectID = 2535 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2536 : static_cast<user_id_t>(NO_SECT); 2537 2538 lldb::offset_t nlist_data_offset = 0; 2539 2540 uint32_t N_SO_index = UINT32_MAX; 2541 2542 MachSymtabSectionInfo section_info(section_list); 2543 std::vector<uint32_t> N_FUN_indexes; 2544 std::vector<uint32_t> N_NSYM_indexes; 2545 std::vector<uint32_t> N_INCL_indexes; 2546 std::vector<uint32_t> N_BRAC_indexes; 2547 std::vector<uint32_t> N_COMM_indexes; 2548 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2549 typedef std::map<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2550 typedef std::map<const char *, uint32_t> ConstNameToSymbolIndexMap; 2551 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2552 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2553 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2554 // Any symbols that get merged into another will get an entry in this map 2555 // so we know 2556 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2557 uint32_t nlist_idx = 0; 2558 Symbol *symbol_ptr = nullptr; 2559 2560 uint32_t sym_idx = 0; 2561 Symbol *sym = nullptr; 2562 size_t num_syms = 0; 2563 std::string memory_symbol_name; 2564 uint32_t unmapped_local_symbols_found = 0; 2565 2566 std::vector<TrieEntryWithOffset> trie_entries; 2567 std::set<lldb::addr_t> resolver_addresses; 2568 2569 if (dyld_trie_data.GetByteSize() > 0) { 2570 std::vector<llvm::StringRef> nameSlices; 2571 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, 2572 resolver_addresses, trie_entries); 2573 2574 ConstString text_segment_name("__TEXT"); 2575 SectionSP text_segment_sp = 2576 GetSectionList()->FindSectionByName(text_segment_name); 2577 if (text_segment_sp) { 2578 const lldb::addr_t text_segment_file_addr = 2579 text_segment_sp->GetFileAddress(); 2580 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2581 for (auto &e : trie_entries) 2582 e.entry.address += text_segment_file_addr; 2583 } 2584 } 2585 } 2586 2587 typedef std::set<ConstString> IndirectSymbols; 2588 IndirectSymbols indirect_symbol_names; 2589 2590 #if defined(__APPLE__) && \ 2591 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2592 2593 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2594 // optimized by moving LOCAL symbols out of the memory mapped portion of 2595 // the DSC. The symbol information has all been retained, but it isn't 2596 // available in the normal nlist data. However, there *are* duplicate 2597 // entries of *some* 2598 // LOCAL symbols in the normal nlist data. To handle this situation 2599 // correctly, we must first attempt 2600 // to parse any DSC unmapped symbol information. If we find any, we set a 2601 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2602 2603 if (m_header.flags & 0x80000000u) { 2604 // Before we can start mapping the DSC, we need to make certain the 2605 // target process is actually using the cache we can find. 2606 2607 // Next we need to determine the correct path for the dyld shared cache. 2608 2609 ArchSpec header_arch; 2610 GetArchitecture(header_arch); 2611 char dsc_path[PATH_MAX]; 2612 char dsc_path_development[PATH_MAX]; 2613 2614 snprintf( 2615 dsc_path, sizeof(dsc_path), "%s%s%s", 2616 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2617 */ 2618 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2619 header_arch.GetArchitectureName()); 2620 2621 snprintf( 2622 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2623 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2624 */ 2625 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2626 header_arch.GetArchitectureName(), ".development"); 2627 2628 FileSpec dsc_nondevelopment_filespec(dsc_path, false); 2629 FileSpec dsc_development_filespec(dsc_path_development, false); 2630 FileSpec dsc_filespec; 2631 2632 UUID dsc_uuid; 2633 UUID process_shared_cache_uuid; 2634 addr_t process_shared_cache_base_addr; 2635 2636 if (process) { 2637 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, process_shared_cache_uuid); 2638 } 2639 2640 // First see if we can find an exact match for the inferior process 2641 // shared cache UUID in the development or non-development shared caches 2642 // on disk. 2643 if (process_shared_cache_uuid.IsValid()) { 2644 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2645 UUID dsc_development_uuid = GetSharedCacheUUID( 2646 dsc_development_filespec, byte_order, addr_byte_size); 2647 if (dsc_development_uuid.IsValid() && 2648 dsc_development_uuid == process_shared_cache_uuid) { 2649 dsc_filespec = dsc_development_filespec; 2650 dsc_uuid = dsc_development_uuid; 2651 } 2652 } 2653 if (!dsc_uuid.IsValid() && 2654 FileSystem::Instance().Exists(dsc_nondevelopment_filespec)) { 2655 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2656 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2657 if (dsc_nondevelopment_uuid.IsValid() && 2658 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2659 dsc_filespec = dsc_nondevelopment_filespec; 2660 dsc_uuid = dsc_nondevelopment_uuid; 2661 } 2662 } 2663 } 2664 2665 // Failing a UUID match, prefer the development dyld_shared cache if both 2666 // are present. 2667 if (!FileSystem::Instance().Exists(dsc_filespec)) { 2668 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2669 dsc_filespec = dsc_development_filespec; 2670 } else { 2671 dsc_filespec = dsc_nondevelopment_filespec; 2672 } 2673 } 2674 2675 /* The dyld_cache_header has a pointer to the 2676 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2677 The dyld_cache_local_symbols_info structure gives us three things: 2678 1. The start and count of the nlist records in the dyld_shared_cache 2679 file 2680 2. The start and size of the strings for these nlist records 2681 3. The start and count of dyld_cache_local_symbols_entry entries 2682 2683 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2684 dyld shared cache. 2685 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2686 the dyld shared cache. 2687 The dyld_cache_local_symbols_entry also lists the start of this 2688 dylib/framework's nlist records 2689 and the count of how many nlist records there are for this 2690 dylib/framework. 2691 */ 2692 2693 // Process the dyld shared cache header to find the unmapped symbols 2694 2695 DataBufferSP dsc_data_sp = MapFileData( 2696 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2697 if (!dsc_uuid.IsValid()) { 2698 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2699 } 2700 if (dsc_data_sp) { 2701 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2702 2703 bool uuid_match = true; 2704 if (dsc_uuid.IsValid() && process) { 2705 if (process_shared_cache_uuid.IsValid() && 2706 dsc_uuid != process_shared_cache_uuid) { 2707 // The on-disk dyld_shared_cache file is not the same as the one in 2708 // this process' memory, don't use it. 2709 uuid_match = false; 2710 ModuleSP module_sp(GetModule()); 2711 if (module_sp) 2712 module_sp->ReportWarning("process shared cache does not match " 2713 "on-disk dyld_shared_cache file, some " 2714 "symbol names will be missing."); 2715 } 2716 } 2717 2718 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2719 2720 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2721 2722 // If the mappingOffset points to a location inside the header, we've 2723 // opened an old dyld shared cache, and should not proceed further. 2724 if (uuid_match && 2725 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2726 2727 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2728 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2729 mappingOffset); 2730 2731 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2732 byte_order, addr_byte_size); 2733 offset = 0; 2734 2735 // The File addresses (from the in-memory Mach-O load commands) for 2736 // the shared libraries in the shared library cache need to be 2737 // adjusted by an offset to match up with the dylibOffset identifying 2738 // field in the dyld_cache_local_symbol_entry's. This offset is 2739 // recorded in mapping_offset_value. 2740 const uint64_t mapping_offset_value = 2741 dsc_mapping_info_data.GetU64(&offset); 2742 2743 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, 2744 localSymbolsOffset); 2745 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2746 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2747 2748 if (localSymbolsOffset && localSymbolsSize) { 2749 // Map the local symbols 2750 DataBufferSP dsc_local_symbols_data_sp = 2751 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2752 2753 if (dsc_local_symbols_data_sp) { 2754 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2755 byte_order, addr_byte_size); 2756 2757 offset = 0; 2758 2759 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2760 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2761 UndefinedNameToDescMap undefined_name_to_desc; 2762 SymbolIndexToName reexport_shlib_needs_fixup; 2763 2764 // Read the local_symbols_infos struct in one shot 2765 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2766 dsc_local_symbols_data.GetU32(&offset, 2767 &local_symbols_info.nlistOffset, 6); 2768 2769 SectionSP text_section_sp( 2770 section_list->FindSectionByName(GetSegmentNameTEXT())); 2771 2772 uint32_t header_file_offset = 2773 (text_section_sp->GetFileAddress() - mapping_offset_value); 2774 2775 offset = local_symbols_info.entriesOffset; 2776 for (uint32_t entry_index = 0; 2777 entry_index < local_symbols_info.entriesCount; 2778 entry_index++) { 2779 struct lldb_copy_dyld_cache_local_symbols_entry 2780 local_symbols_entry; 2781 local_symbols_entry.dylibOffset = 2782 dsc_local_symbols_data.GetU32(&offset); 2783 local_symbols_entry.nlistStartIndex = 2784 dsc_local_symbols_data.GetU32(&offset); 2785 local_symbols_entry.nlistCount = 2786 dsc_local_symbols_data.GetU32(&offset); 2787 2788 if (header_file_offset == local_symbols_entry.dylibOffset) { 2789 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2790 2791 // The normal nlist code cannot correctly size the Symbols 2792 // array, we need to allocate it here. 2793 sym = symtab->Resize( 2794 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2795 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2796 num_syms = symtab->GetNumSymbols(); 2797 2798 nlist_data_offset = 2799 local_symbols_info.nlistOffset + 2800 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2801 uint32_t string_table_offset = 2802 local_symbols_info.stringsOffset; 2803 2804 for (uint32_t nlist_index = 0; 2805 nlist_index < local_symbols_entry.nlistCount; 2806 nlist_index++) { 2807 ///////////////////////////// 2808 { 2809 struct nlist_64 nlist; 2810 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize( 2811 nlist_data_offset, nlist_byte_size)) 2812 break; 2813 2814 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked( 2815 &nlist_data_offset); 2816 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked( 2817 &nlist_data_offset); 2818 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked( 2819 &nlist_data_offset); 2820 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked( 2821 &nlist_data_offset); 2822 nlist.n_value = 2823 dsc_local_symbols_data.GetAddress_unchecked( 2824 &nlist_data_offset); 2825 2826 SymbolType type = eSymbolTypeInvalid; 2827 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2828 string_table_offset + nlist.n_strx); 2829 2830 if (symbol_name == NULL) { 2831 // No symbol should be NULL, even the symbols with no 2832 // string values should have an offset zero which 2833 // points to an empty C-string 2834 Host::SystemLog( 2835 Host::eSystemLogError, 2836 "error: DSC unmapped local symbol[%u] has invalid " 2837 "string table offset 0x%x in %s, ignoring symbol\n", 2838 entry_index, nlist.n_strx, 2839 module_sp->GetFileSpec().GetPath().c_str()); 2840 continue; 2841 } 2842 if (symbol_name[0] == '\0') 2843 symbol_name = NULL; 2844 2845 const char *symbol_name_non_abi_mangled = NULL; 2846 2847 SectionSP symbol_section; 2848 uint32_t symbol_byte_size = 0; 2849 bool add_nlist = true; 2850 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2851 bool demangled_is_synthesized = false; 2852 bool is_gsym = false; 2853 bool set_value = true; 2854 2855 assert(sym_idx < num_syms); 2856 2857 sym[sym_idx].SetDebug(is_debug); 2858 2859 if (is_debug) { 2860 switch (nlist.n_type) { 2861 case N_GSYM: 2862 // global symbol: name,,NO_SECT,type,0 2863 // Sometimes the N_GSYM value contains the address. 2864 2865 // FIXME: In the .o files, we have a GSYM and a debug 2866 // symbol for all the ObjC data. They 2867 // have the same address, but we want to ensure that 2868 // we always find only the real symbol, 'cause we 2869 // don't currently correctly attribute the 2870 // GSYM one to the ObjCClass/Ivar/MetaClass 2871 // symbol type. This is a temporary hack to make 2872 // sure the ObjectiveC symbols get treated correctly. 2873 // To do this right, we should coalesce all the GSYM 2874 // & global symbols that have the same address. 2875 2876 is_gsym = true; 2877 sym[sym_idx].SetExternal(true); 2878 2879 if (symbol_name && symbol_name[0] == '_' && 2880 symbol_name[1] == 'O') { 2881 llvm::StringRef symbol_name_ref(symbol_name); 2882 if (symbol_name_ref.startswith( 2883 g_objc_v2_prefix_class)) { 2884 symbol_name_non_abi_mangled = symbol_name + 1; 2885 symbol_name = 2886 symbol_name + g_objc_v2_prefix_class.size(); 2887 type = eSymbolTypeObjCClass; 2888 demangled_is_synthesized = true; 2889 2890 } else if (symbol_name_ref.startswith( 2891 g_objc_v2_prefix_metaclass)) { 2892 symbol_name_non_abi_mangled = symbol_name + 1; 2893 symbol_name = symbol_name + 2894 g_objc_v2_prefix_metaclass.size(); 2895 type = eSymbolTypeObjCMetaClass; 2896 demangled_is_synthesized = true; 2897 } else if (symbol_name_ref.startswith( 2898 g_objc_v2_prefix_ivar)) { 2899 symbol_name_non_abi_mangled = symbol_name + 1; 2900 symbol_name = 2901 symbol_name + g_objc_v2_prefix_ivar.size(); 2902 type = eSymbolTypeObjCIVar; 2903 demangled_is_synthesized = true; 2904 } 2905 } else { 2906 if (nlist.n_value != 0) 2907 symbol_section = section_info.GetSection( 2908 nlist.n_sect, nlist.n_value); 2909 type = eSymbolTypeData; 2910 } 2911 break; 2912 2913 case N_FNAME: 2914 // procedure name (f77 kludge): name,,NO_SECT,0,0 2915 type = eSymbolTypeCompiler; 2916 break; 2917 2918 case N_FUN: 2919 // procedure: name,,n_sect,linenumber,address 2920 if (symbol_name) { 2921 type = eSymbolTypeCode; 2922 symbol_section = section_info.GetSection( 2923 nlist.n_sect, nlist.n_value); 2924 2925 N_FUN_addr_to_sym_idx.insert( 2926 std::make_pair(nlist.n_value, sym_idx)); 2927 // We use the current number of symbols in the 2928 // symbol table in lieu of using nlist_idx in case 2929 // we ever start trimming entries out 2930 N_FUN_indexes.push_back(sym_idx); 2931 } else { 2932 type = eSymbolTypeCompiler; 2933 2934 if (!N_FUN_indexes.empty()) { 2935 // Copy the size of the function into the 2936 // original 2937 // STAB entry so we don't have 2938 // to hunt for it later 2939 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2940 ->SetByteSize(nlist.n_value); 2941 N_FUN_indexes.pop_back(); 2942 // We don't really need the end function STAB as 2943 // it contains the size which we already placed 2944 // with the original symbol, so don't add it if 2945 // we want a minimal symbol table 2946 add_nlist = false; 2947 } 2948 } 2949 break; 2950 2951 case N_STSYM: 2952 // static symbol: name,,n_sect,type,address 2953 N_STSYM_addr_to_sym_idx.insert( 2954 std::make_pair(nlist.n_value, sym_idx)); 2955 symbol_section = section_info.GetSection( 2956 nlist.n_sect, nlist.n_value); 2957 if (symbol_name && symbol_name[0]) { 2958 type = ObjectFile::GetSymbolTypeFromName( 2959 symbol_name + 1, eSymbolTypeData); 2960 } 2961 break; 2962 2963 case N_LCSYM: 2964 // .lcomm symbol: name,,n_sect,type,address 2965 symbol_section = section_info.GetSection( 2966 nlist.n_sect, nlist.n_value); 2967 type = eSymbolTypeCommonBlock; 2968 break; 2969 2970 case N_BNSYM: 2971 // We use the current number of symbols in the symbol 2972 // table in lieu of using nlist_idx in case we ever 2973 // start trimming entries out Skip these if we want 2974 // minimal symbol tables 2975 add_nlist = false; 2976 break; 2977 2978 case N_ENSYM: 2979 // Set the size of the N_BNSYM to the terminating 2980 // index of this N_ENSYM so that we can always skip 2981 // the entire symbol if we need to navigate more 2982 // quickly at the source level when parsing STABS 2983 // Skip these if we want minimal symbol tables 2984 add_nlist = false; 2985 break; 2986 2987 case N_OPT: 2988 // emitted with gcc2_compiled and in gcc source 2989 type = eSymbolTypeCompiler; 2990 break; 2991 2992 case N_RSYM: 2993 // register sym: name,,NO_SECT,type,register 2994 type = eSymbolTypeVariable; 2995 break; 2996 2997 case N_SLINE: 2998 // src line: 0,,n_sect,linenumber,address 2999 symbol_section = section_info.GetSection( 3000 nlist.n_sect, nlist.n_value); 3001 type = eSymbolTypeLineEntry; 3002 break; 3003 3004 case N_SSYM: 3005 // structure elt: name,,NO_SECT,type,struct_offset 3006 type = eSymbolTypeVariableType; 3007 break; 3008 3009 case N_SO: 3010 // source file name 3011 type = eSymbolTypeSourceFile; 3012 if (symbol_name == NULL) { 3013 add_nlist = false; 3014 if (N_SO_index != UINT32_MAX) { 3015 // Set the size of the N_SO to the terminating 3016 // index of this N_SO so that we can always skip 3017 // the entire N_SO if we need to navigate more 3018 // quickly at the source level when parsing STABS 3019 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3020 symbol_ptr->SetByteSize(sym_idx); 3021 symbol_ptr->SetSizeIsSibling(true); 3022 } 3023 N_NSYM_indexes.clear(); 3024 N_INCL_indexes.clear(); 3025 N_BRAC_indexes.clear(); 3026 N_COMM_indexes.clear(); 3027 N_FUN_indexes.clear(); 3028 N_SO_index = UINT32_MAX; 3029 } else { 3030 // We use the current number of symbols in the 3031 // symbol table in lieu of using nlist_idx in case 3032 // we ever start trimming entries out 3033 const bool N_SO_has_full_path = 3034 symbol_name[0] == '/'; 3035 if (N_SO_has_full_path) { 3036 if ((N_SO_index == sym_idx - 1) && 3037 ((sym_idx - 1) < num_syms)) { 3038 // We have two consecutive N_SO entries where 3039 // the first contains a directory and the 3040 // second contains a full path. 3041 sym[sym_idx - 1].GetMangled().SetValue( 3042 ConstString(symbol_name), false); 3043 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3044 add_nlist = false; 3045 } else { 3046 // This is the first entry in a N_SO that 3047 // contains a directory or 3048 // a full path to the source file 3049 N_SO_index = sym_idx; 3050 } 3051 } else if ((N_SO_index == sym_idx - 1) && 3052 ((sym_idx - 1) < num_syms)) { 3053 // This is usually the second N_SO entry that 3054 // contains just the filename, so here we combine 3055 // it with the first one if we are minimizing the 3056 // symbol table 3057 const char *so_path = 3058 sym[sym_idx - 1] 3059 .GetMangled() 3060 .GetDemangledName( 3061 lldb::eLanguageTypeUnknown) 3062 .AsCString(); 3063 if (so_path && so_path[0]) { 3064 std::string full_so_path(so_path); 3065 const size_t double_slash_pos = 3066 full_so_path.find("//"); 3067 if (double_slash_pos != std::string::npos) { 3068 // The linker has been generating bad N_SO 3069 // entries with doubled up paths 3070 // in the format "%s%s" where the first 3071 // string in the DW_AT_comp_dir, and the 3072 // second is the directory for the source 3073 // file so you end up with a path that looks 3074 // like "/tmp/src//tmp/src/" 3075 FileSpec so_dir(so_path, false); 3076 if (!FileSystem::Instance().Exists(so_dir)) { 3077 so_dir.SetFile( 3078 &full_so_path[double_slash_pos + 1], 3079 false); 3080 if (FileSystem::Instance().Exists(so_dir)) { 3081 // Trim off the incorrect path 3082 full_so_path.erase(0, 3083 double_slash_pos + 1); 3084 } 3085 } 3086 } 3087 if (*full_so_path.rbegin() != '/') 3088 full_so_path += '/'; 3089 full_so_path += symbol_name; 3090 sym[sym_idx - 1].GetMangled().SetValue( 3091 ConstString(full_so_path.c_str()), false); 3092 add_nlist = false; 3093 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3094 } 3095 } else { 3096 // This could be a relative path to a N_SO 3097 N_SO_index = sym_idx; 3098 } 3099 } 3100 break; 3101 3102 case N_OSO: 3103 // object file name: name,,0,0,st_mtime 3104 type = eSymbolTypeObjectFile; 3105 break; 3106 3107 case N_LSYM: 3108 // local sym: name,,NO_SECT,type,offset 3109 type = eSymbolTypeLocal; 3110 break; 3111 3112 // INCL scopes 3113 case N_BINCL: 3114 // include file beginning: name,,NO_SECT,0,sum We use 3115 // the current number of symbols in the symbol table 3116 // in lieu of using nlist_idx in case we ever start 3117 // trimming entries out 3118 N_INCL_indexes.push_back(sym_idx); 3119 type = eSymbolTypeScopeBegin; 3120 break; 3121 3122 case N_EINCL: 3123 // include file end: name,,NO_SECT,0,0 3124 // Set the size of the N_BINCL to the terminating 3125 // index of this N_EINCL so that we can always skip 3126 // the entire symbol if we need to navigate more 3127 // quickly at the source level when parsing STABS 3128 if (!N_INCL_indexes.empty()) { 3129 symbol_ptr = 3130 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3131 symbol_ptr->SetByteSize(sym_idx + 1); 3132 symbol_ptr->SetSizeIsSibling(true); 3133 N_INCL_indexes.pop_back(); 3134 } 3135 type = eSymbolTypeScopeEnd; 3136 break; 3137 3138 case N_SOL: 3139 // #included file name: name,,n_sect,0,address 3140 type = eSymbolTypeHeaderFile; 3141 3142 // We currently don't use the header files on darwin 3143 add_nlist = false; 3144 break; 3145 3146 case N_PARAMS: 3147 // compiler parameters: name,,NO_SECT,0,0 3148 type = eSymbolTypeCompiler; 3149 break; 3150 3151 case N_VERSION: 3152 // compiler version: name,,NO_SECT,0,0 3153 type = eSymbolTypeCompiler; 3154 break; 3155 3156 case N_OLEVEL: 3157 // compiler -O level: name,,NO_SECT,0,0 3158 type = eSymbolTypeCompiler; 3159 break; 3160 3161 case N_PSYM: 3162 // parameter: name,,NO_SECT,type,offset 3163 type = eSymbolTypeVariable; 3164 break; 3165 3166 case N_ENTRY: 3167 // alternate entry: name,,n_sect,linenumber,address 3168 symbol_section = section_info.GetSection( 3169 nlist.n_sect, nlist.n_value); 3170 type = eSymbolTypeLineEntry; 3171 break; 3172 3173 // Left and Right Braces 3174 case N_LBRAC: 3175 // left bracket: 0,,NO_SECT,nesting level,address We 3176 // use the current number of symbols in the symbol 3177 // table in lieu of using nlist_idx in case we ever 3178 // start trimming entries out 3179 symbol_section = section_info.GetSection( 3180 nlist.n_sect, nlist.n_value); 3181 N_BRAC_indexes.push_back(sym_idx); 3182 type = eSymbolTypeScopeBegin; 3183 break; 3184 3185 case N_RBRAC: 3186 // right bracket: 0,,NO_SECT,nesting level,address 3187 // Set the size of the N_LBRAC to the terminating 3188 // index of this N_RBRAC so that we can always skip 3189 // the entire symbol if we need to navigate more 3190 // quickly at the source level when parsing STABS 3191 symbol_section = section_info.GetSection( 3192 nlist.n_sect, nlist.n_value); 3193 if (!N_BRAC_indexes.empty()) { 3194 symbol_ptr = 3195 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3196 symbol_ptr->SetByteSize(sym_idx + 1); 3197 symbol_ptr->SetSizeIsSibling(true); 3198 N_BRAC_indexes.pop_back(); 3199 } 3200 type = eSymbolTypeScopeEnd; 3201 break; 3202 3203 case N_EXCL: 3204 // deleted include file: name,,NO_SECT,0,sum 3205 type = eSymbolTypeHeaderFile; 3206 break; 3207 3208 // COMM scopes 3209 case N_BCOMM: 3210 // begin common: name,,NO_SECT,0,0 3211 // We use the current number of symbols in the symbol 3212 // table in lieu of using nlist_idx in case we ever 3213 // start trimming entries out 3214 type = eSymbolTypeScopeBegin; 3215 N_COMM_indexes.push_back(sym_idx); 3216 break; 3217 3218 case N_ECOML: 3219 // end common (local name): 0,,n_sect,0,address 3220 symbol_section = section_info.GetSection( 3221 nlist.n_sect, nlist.n_value); 3222 // Fall through 3223 3224 case N_ECOMM: 3225 // end common: name,,n_sect,0,0 3226 // Set the size of the N_BCOMM to the terminating 3227 // index of this N_ECOMM/N_ECOML so that we can 3228 // always skip the entire symbol if we need to 3229 // navigate more quickly at the source level when 3230 // parsing STABS 3231 if (!N_COMM_indexes.empty()) { 3232 symbol_ptr = 3233 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3234 symbol_ptr->SetByteSize(sym_idx + 1); 3235 symbol_ptr->SetSizeIsSibling(true); 3236 N_COMM_indexes.pop_back(); 3237 } 3238 type = eSymbolTypeScopeEnd; 3239 break; 3240 3241 case N_LENG: 3242 // second stab entry with length information 3243 type = eSymbolTypeAdditional; 3244 break; 3245 3246 default: 3247 break; 3248 } 3249 } else { 3250 // uint8_t n_pext = N_PEXT & nlist.n_type; 3251 uint8_t n_type = N_TYPE & nlist.n_type; 3252 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3253 3254 switch (n_type) { 3255 case N_INDR: { 3256 const char *reexport_name_cstr = 3257 strtab_data.PeekCStr(nlist.n_value); 3258 if (reexport_name_cstr && reexport_name_cstr[0]) { 3259 type = eSymbolTypeReExported; 3260 ConstString reexport_name( 3261 reexport_name_cstr + 3262 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3263 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3264 set_value = false; 3265 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3266 indirect_symbol_names.insert( 3267 ConstString(symbol_name + 3268 ((symbol_name[0] == '_') ? 1 : 0))); 3269 } else 3270 type = eSymbolTypeUndefined; 3271 } break; 3272 3273 case N_UNDF: 3274 if (symbol_name && symbol_name[0]) { 3275 ConstString undefined_name( 3276 symbol_name + 3277 ((symbol_name[0] == '_') ? 1 : 0)); 3278 undefined_name_to_desc[undefined_name] = 3279 nlist.n_desc; 3280 } 3281 // Fall through 3282 case N_PBUD: 3283 type = eSymbolTypeUndefined; 3284 break; 3285 3286 case N_ABS: 3287 type = eSymbolTypeAbsolute; 3288 break; 3289 3290 case N_SECT: { 3291 symbol_section = section_info.GetSection( 3292 nlist.n_sect, nlist.n_value); 3293 3294 if (symbol_section == NULL) { 3295 // TODO: warn about this? 3296 add_nlist = false; 3297 break; 3298 } 3299 3300 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3301 type = eSymbolTypeException; 3302 } else { 3303 uint32_t section_type = 3304 symbol_section->Get() & SECTION_TYPE; 3305 3306 switch (section_type) { 3307 case S_CSTRING_LITERALS: 3308 type = eSymbolTypeData; 3309 break; // section with only literal C strings 3310 case S_4BYTE_LITERALS: 3311 type = eSymbolTypeData; 3312 break; // section with only 4 byte literals 3313 case S_8BYTE_LITERALS: 3314 type = eSymbolTypeData; 3315 break; // section with only 8 byte literals 3316 case S_LITERAL_POINTERS: 3317 type = eSymbolTypeTrampoline; 3318 break; // section with only pointers to literals 3319 case S_NON_LAZY_SYMBOL_POINTERS: 3320 type = eSymbolTypeTrampoline; 3321 break; // section with only non-lazy symbol 3322 // pointers 3323 case S_LAZY_SYMBOL_POINTERS: 3324 type = eSymbolTypeTrampoline; 3325 break; // section with only lazy symbol pointers 3326 case S_SYMBOL_STUBS: 3327 type = eSymbolTypeTrampoline; 3328 break; // section with only symbol stubs, byte 3329 // size of stub in the reserved2 field 3330 case S_MOD_INIT_FUNC_POINTERS: 3331 type = eSymbolTypeCode; 3332 break; // section with only function pointers for 3333 // initialization 3334 case S_MOD_TERM_FUNC_POINTERS: 3335 type = eSymbolTypeCode; 3336 break; // section with only function pointers for 3337 // termination 3338 case S_INTERPOSING: 3339 type = eSymbolTypeTrampoline; 3340 break; // section with only pairs of function 3341 // pointers for interposing 3342 case S_16BYTE_LITERALS: 3343 type = eSymbolTypeData; 3344 break; // section with only 16 byte literals 3345 case S_DTRACE_DOF: 3346 type = eSymbolTypeInstrumentation; 3347 break; 3348 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3349 type = eSymbolTypeTrampoline; 3350 break; 3351 default: 3352 switch (symbol_section->GetType()) { 3353 case lldb::eSectionTypeCode: 3354 type = eSymbolTypeCode; 3355 break; 3356 case eSectionTypeData: 3357 case eSectionTypeDataCString: // Inlined C string 3358 // data 3359 case eSectionTypeDataCStringPointers: // Pointers 3360 // to C 3361 // string 3362 // data 3363 case eSectionTypeDataSymbolAddress: // Address of 3364 // a symbol in 3365 // the symbol 3366 // table 3367 case eSectionTypeData4: 3368 case eSectionTypeData8: 3369 case eSectionTypeData16: 3370 type = eSymbolTypeData; 3371 break; 3372 default: 3373 break; 3374 } 3375 break; 3376 } 3377 3378 if (type == eSymbolTypeInvalid) { 3379 const char *symbol_sect_name = 3380 symbol_section->GetName().AsCString(); 3381 if (symbol_section->IsDescendant( 3382 text_section_sp.get())) { 3383 if (symbol_section->IsClear( 3384 S_ATTR_PURE_INSTRUCTIONS | 3385 S_ATTR_SELF_MODIFYING_CODE | 3386 S_ATTR_SOME_INSTRUCTIONS)) 3387 type = eSymbolTypeData; 3388 else 3389 type = eSymbolTypeCode; 3390 } else if (symbol_section->IsDescendant( 3391 data_section_sp.get()) || 3392 symbol_section->IsDescendant( 3393 data_dirty_section_sp.get()) || 3394 symbol_section->IsDescendant( 3395 data_const_section_sp.get())) { 3396 if (symbol_sect_name && 3397 ::strstr(symbol_sect_name, "__objc") == 3398 symbol_sect_name) { 3399 type = eSymbolTypeRuntime; 3400 3401 if (symbol_name) { 3402 llvm::StringRef symbol_name_ref( 3403 symbol_name); 3404 if (symbol_name_ref.startswith("_OBJC_")) { 3405 static const llvm::StringRef 3406 g_objc_v2_prefix_class( 3407 "_OBJC_CLASS_$_"); 3408 static const llvm::StringRef 3409 g_objc_v2_prefix_metaclass( 3410 "_OBJC_METACLASS_$_"); 3411 static const llvm::StringRef 3412 g_objc_v2_prefix_ivar( 3413 "_OBJC_IVAR_$_"); 3414 if (symbol_name_ref.startswith( 3415 g_objc_v2_prefix_class)) { 3416 symbol_name_non_abi_mangled = 3417 symbol_name + 1; 3418 symbol_name = 3419 symbol_name + 3420 g_objc_v2_prefix_class.size(); 3421 type = eSymbolTypeObjCClass; 3422 demangled_is_synthesized = true; 3423 } else if ( 3424 symbol_name_ref.startswith( 3425 g_objc_v2_prefix_metaclass)) { 3426 symbol_name_non_abi_mangled = 3427 symbol_name + 1; 3428 symbol_name = 3429 symbol_name + 3430 g_objc_v2_prefix_metaclass.size(); 3431 type = eSymbolTypeObjCMetaClass; 3432 demangled_is_synthesized = true; 3433 } else if (symbol_name_ref.startswith( 3434 g_objc_v2_prefix_ivar)) { 3435 symbol_name_non_abi_mangled = 3436 symbol_name + 1; 3437 symbol_name = 3438 symbol_name + 3439 g_objc_v2_prefix_ivar.size(); 3440 type = eSymbolTypeObjCIVar; 3441 demangled_is_synthesized = true; 3442 } 3443 } 3444 } 3445 } else if (symbol_sect_name && 3446 ::strstr(symbol_sect_name, 3447 "__gcc_except_tab") == 3448 symbol_sect_name) { 3449 type = eSymbolTypeException; 3450 } else { 3451 type = eSymbolTypeData; 3452 } 3453 } else if (symbol_sect_name && 3454 ::strstr(symbol_sect_name, 3455 "__IMPORT") == 3456 symbol_sect_name) { 3457 type = eSymbolTypeTrampoline; 3458 } else if (symbol_section->IsDescendant( 3459 objc_section_sp.get())) { 3460 type = eSymbolTypeRuntime; 3461 if (symbol_name && symbol_name[0] == '.') { 3462 llvm::StringRef symbol_name_ref(symbol_name); 3463 static const llvm::StringRef 3464 g_objc_v1_prefix_class( 3465 ".objc_class_name_"); 3466 if (symbol_name_ref.startswith( 3467 g_objc_v1_prefix_class)) { 3468 symbol_name_non_abi_mangled = symbol_name; 3469 symbol_name = symbol_name + 3470 g_objc_v1_prefix_class.size(); 3471 type = eSymbolTypeObjCClass; 3472 demangled_is_synthesized = true; 3473 } 3474 } 3475 } 3476 } 3477 } 3478 } break; 3479 } 3480 } 3481 3482 if (add_nlist) { 3483 uint64_t symbol_value = nlist.n_value; 3484 if (symbol_name_non_abi_mangled) { 3485 sym[sym_idx].GetMangled().SetMangledName( 3486 ConstString(symbol_name_non_abi_mangled)); 3487 sym[sym_idx].GetMangled().SetDemangledName( 3488 ConstString(symbol_name)); 3489 } else { 3490 bool symbol_name_is_mangled = false; 3491 3492 if (symbol_name && symbol_name[0] == '_') { 3493 symbol_name_is_mangled = symbol_name[1] == '_'; 3494 symbol_name++; // Skip the leading underscore 3495 } 3496 3497 if (symbol_name) { 3498 ConstString const_symbol_name(symbol_name); 3499 sym[sym_idx].GetMangled().SetValue( 3500 const_symbol_name, symbol_name_is_mangled); 3501 if (is_gsym && is_debug) { 3502 const char *gsym_name = 3503 sym[sym_idx] 3504 .GetMangled() 3505 .GetName(lldb::eLanguageTypeUnknown, 3506 Mangled::ePreferMangled) 3507 .GetCString(); 3508 if (gsym_name) 3509 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3510 } 3511 } 3512 } 3513 if (symbol_section) { 3514 const addr_t section_file_addr = 3515 symbol_section->GetFileAddress(); 3516 if (symbol_byte_size == 0 && 3517 function_starts_count > 0) { 3518 addr_t symbol_lookup_file_addr = nlist.n_value; 3519 // Do an exact address match for non-ARM addresses, 3520 // else get the closest since the symbol might be a 3521 // thumb symbol which has an address with bit zero 3522 // set 3523 FunctionStarts::Entry *func_start_entry = 3524 function_starts.FindEntry( 3525 symbol_lookup_file_addr, !is_arm); 3526 if (is_arm && func_start_entry) { 3527 // Verify that the function start address is the 3528 // symbol address (ARM) or the symbol address + 1 3529 // (thumb) 3530 if (func_start_entry->addr != 3531 symbol_lookup_file_addr && 3532 func_start_entry->addr != 3533 (symbol_lookup_file_addr + 1)) { 3534 // Not the right entry, NULL it out... 3535 func_start_entry = NULL; 3536 } 3537 } 3538 if (func_start_entry) { 3539 func_start_entry->data = true; 3540 3541 addr_t symbol_file_addr = func_start_entry->addr; 3542 uint32_t symbol_flags = 0; 3543 if (is_arm) { 3544 if (symbol_file_addr & 1) 3545 symbol_flags = 3546 MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3547 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3548 } 3549 3550 const FunctionStarts::Entry 3551 *next_func_start_entry = 3552 function_starts.FindNextEntry( 3553 func_start_entry); 3554 const addr_t section_end_file_addr = 3555 section_file_addr + 3556 symbol_section->GetByteSize(); 3557 if (next_func_start_entry) { 3558 addr_t next_symbol_file_addr = 3559 next_func_start_entry->addr; 3560 // Be sure the clear the Thumb address bit when 3561 // we calculate the size from the current and 3562 // next address 3563 if (is_arm) 3564 next_symbol_file_addr &= 3565 THUMB_ADDRESS_BIT_MASK; 3566 symbol_byte_size = std::min<lldb::addr_t>( 3567 next_symbol_file_addr - symbol_file_addr, 3568 section_end_file_addr - symbol_file_addr); 3569 } else { 3570 symbol_byte_size = 3571 section_end_file_addr - symbol_file_addr; 3572 } 3573 } 3574 } 3575 symbol_value -= section_file_addr; 3576 } 3577 3578 if (is_debug == false) { 3579 if (type == eSymbolTypeCode) { 3580 // See if we can find a N_FUN entry for any code 3581 // symbols. If we do find a match, and the name 3582 // matches, then we can merge the two into just the 3583 // function symbol to avoid duplicate entries in 3584 // the symbol table 3585 std::pair<ValueToSymbolIndexMap::const_iterator, 3586 ValueToSymbolIndexMap::const_iterator> 3587 range; 3588 range = N_FUN_addr_to_sym_idx.equal_range( 3589 nlist.n_value); 3590 if (range.first != range.second) { 3591 bool found_it = false; 3592 for (ValueToSymbolIndexMap::const_iterator pos = 3593 range.first; 3594 pos != range.second; ++pos) { 3595 if (sym[sym_idx].GetMangled().GetName( 3596 lldb::eLanguageTypeUnknown, 3597 Mangled::ePreferMangled) == 3598 sym[pos->second].GetMangled().GetName( 3599 lldb::eLanguageTypeUnknown, 3600 Mangled::ePreferMangled)) { 3601 m_nlist_idx_to_sym_idx[nlist_idx] = 3602 pos->second; 3603 // We just need the flags from the linker 3604 // symbol, so put these flags 3605 // into the N_FUN flags to avoid duplicate 3606 // symbols in the symbol table 3607 sym[pos->second].SetExternal( 3608 sym[sym_idx].IsExternal()); 3609 sym[pos->second].SetFlags(nlist.n_type << 16 | 3610 nlist.n_desc); 3611 if (resolver_addresses.find(nlist.n_value) != 3612 resolver_addresses.end()) 3613 sym[pos->second].SetType( 3614 eSymbolTypeResolver); 3615 sym[sym_idx].Clear(); 3616 found_it = true; 3617 break; 3618 } 3619 } 3620 if (found_it) 3621 continue; 3622 } else { 3623 if (resolver_addresses.find(nlist.n_value) != 3624 resolver_addresses.end()) 3625 type = eSymbolTypeResolver; 3626 } 3627 } else if (type == eSymbolTypeData || 3628 type == eSymbolTypeObjCClass || 3629 type == eSymbolTypeObjCMetaClass || 3630 type == eSymbolTypeObjCIVar) { 3631 // See if we can find a N_STSYM entry for any data 3632 // symbols. If we do find a match, and the name 3633 // matches, then we can merge the two into just the 3634 // Static symbol to avoid duplicate entries in the 3635 // symbol table 3636 std::pair<ValueToSymbolIndexMap::const_iterator, 3637 ValueToSymbolIndexMap::const_iterator> 3638 range; 3639 range = N_STSYM_addr_to_sym_idx.equal_range( 3640 nlist.n_value); 3641 if (range.first != range.second) { 3642 bool found_it = false; 3643 for (ValueToSymbolIndexMap::const_iterator pos = 3644 range.first; 3645 pos != range.second; ++pos) { 3646 if (sym[sym_idx].GetMangled().GetName( 3647 lldb::eLanguageTypeUnknown, 3648 Mangled::ePreferMangled) == 3649 sym[pos->second].GetMangled().GetName( 3650 lldb::eLanguageTypeUnknown, 3651 Mangled::ePreferMangled)) { 3652 m_nlist_idx_to_sym_idx[nlist_idx] = 3653 pos->second; 3654 // We just need the flags from the linker 3655 // symbol, so put these flags 3656 // into the N_STSYM flags to avoid duplicate 3657 // symbols in the symbol table 3658 sym[pos->second].SetExternal( 3659 sym[sym_idx].IsExternal()); 3660 sym[pos->second].SetFlags(nlist.n_type << 16 | 3661 nlist.n_desc); 3662 sym[sym_idx].Clear(); 3663 found_it = true; 3664 break; 3665 } 3666 } 3667 if (found_it) 3668 continue; 3669 } else { 3670 const char *gsym_name = 3671 sym[sym_idx] 3672 .GetMangled() 3673 .GetName(lldb::eLanguageTypeUnknown, 3674 Mangled::ePreferMangled) 3675 .GetCString(); 3676 if (gsym_name) { 3677 // Combine N_GSYM stab entries with the non 3678 // stab symbol 3679 ConstNameToSymbolIndexMap::const_iterator pos = 3680 N_GSYM_name_to_sym_idx.find(gsym_name); 3681 if (pos != N_GSYM_name_to_sym_idx.end()) { 3682 const uint32_t GSYM_sym_idx = pos->second; 3683 m_nlist_idx_to_sym_idx[nlist_idx] = 3684 GSYM_sym_idx; 3685 // Copy the address, because often the N_GSYM 3686 // address has an invalid address of zero 3687 // when the global is a common symbol 3688 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3689 symbol_section); 3690 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3691 symbol_value); 3692 // We just need the flags from the linker 3693 // symbol, so put these flags 3694 // into the N_GSYM flags to avoid duplicate 3695 // symbols in the symbol table 3696 sym[GSYM_sym_idx].SetFlags( 3697 nlist.n_type << 16 | nlist.n_desc); 3698 sym[sym_idx].Clear(); 3699 continue; 3700 } 3701 } 3702 } 3703 } 3704 } 3705 3706 sym[sym_idx].SetID(nlist_idx); 3707 sym[sym_idx].SetType(type); 3708 if (set_value) { 3709 sym[sym_idx].GetAddressRef().SetSection( 3710 symbol_section); 3711 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3712 } 3713 sym[sym_idx].SetFlags(nlist.n_type << 16 | 3714 nlist.n_desc); 3715 3716 if (symbol_byte_size > 0) 3717 sym[sym_idx].SetByteSize(symbol_byte_size); 3718 3719 if (demangled_is_synthesized) 3720 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3721 ++sym_idx; 3722 } else { 3723 sym[sym_idx].Clear(); 3724 } 3725 } 3726 ///////////////////////////// 3727 } 3728 break; // No more entries to consider 3729 } 3730 } 3731 3732 for (const auto &pos : reexport_shlib_needs_fixup) { 3733 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3734 if (undef_pos != undefined_name_to_desc.end()) { 3735 const uint8_t dylib_ordinal = 3736 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3737 if (dylib_ordinal > 0 && 3738 dylib_ordinal < dylib_files.GetSize()) 3739 sym[pos.first].SetReExportedSymbolSharedLibrary( 3740 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3741 } 3742 } 3743 } 3744 } 3745 } 3746 } 3747 } 3748 3749 // Must reset this in case it was mutated above! 3750 nlist_data_offset = 0; 3751 #endif 3752 3753 if (nlist_data.GetByteSize() > 0) { 3754 3755 // If the sym array was not created while parsing the DSC unmapped 3756 // symbols, create it now. 3757 if (sym == nullptr) { 3758 sym = symtab->Resize(symtab_load_command.nsyms + 3759 m_dysymtab.nindirectsyms); 3760 num_syms = symtab->GetNumSymbols(); 3761 } 3762 3763 if (unmapped_local_symbols_found) { 3764 assert(m_dysymtab.ilocalsym == 0); 3765 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3766 nlist_idx = m_dysymtab.nlocalsym; 3767 } else { 3768 nlist_idx = 0; 3769 } 3770 3771 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3772 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3773 UndefinedNameToDescMap undefined_name_to_desc; 3774 SymbolIndexToName reexport_shlib_needs_fixup; 3775 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 3776 struct nlist_64 nlist; 3777 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, 3778 nlist_byte_size)) 3779 break; 3780 3781 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3782 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 3783 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 3784 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 3785 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 3786 3787 SymbolType type = eSymbolTypeInvalid; 3788 const char *symbol_name = nullptr; 3789 3790 if (have_strtab_data) { 3791 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3792 3793 if (symbol_name == nullptr) { 3794 // No symbol should be NULL, even the symbols with no string values 3795 // should have an offset zero which points to an empty C-string 3796 Host::SystemLog(Host::eSystemLogError, 3797 "error: symbol[%u] has invalid string table offset " 3798 "0x%x in %s, ignoring symbol\n", 3799 nlist_idx, nlist.n_strx, 3800 module_sp->GetFileSpec().GetPath().c_str()); 3801 continue; 3802 } 3803 if (symbol_name[0] == '\0') 3804 symbol_name = nullptr; 3805 } else { 3806 const addr_t str_addr = strtab_addr + nlist.n_strx; 3807 Status str_error; 3808 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3809 str_error)) 3810 symbol_name = memory_symbol_name.c_str(); 3811 } 3812 const char *symbol_name_non_abi_mangled = nullptr; 3813 3814 SectionSP symbol_section; 3815 lldb::addr_t symbol_byte_size = 0; 3816 bool add_nlist = true; 3817 bool is_gsym = false; 3818 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3819 bool demangled_is_synthesized = false; 3820 bool set_value = true; 3821 assert(sym_idx < num_syms); 3822 3823 sym[sym_idx].SetDebug(is_debug); 3824 3825 if (is_debug) { 3826 switch (nlist.n_type) { 3827 case N_GSYM: 3828 // global symbol: name,,NO_SECT,type,0 3829 // Sometimes the N_GSYM value contains the address. 3830 3831 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3832 // the ObjC data. They 3833 // have the same address, but we want to ensure that we always find 3834 // only the real symbol, 'cause we don't currently correctly 3835 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3836 // type. This is a temporary hack to make sure the ObjectiveC 3837 // symbols get treated correctly. To do this right, we should 3838 // coalesce all the GSYM & global symbols that have the same 3839 // address. 3840 is_gsym = true; 3841 sym[sym_idx].SetExternal(true); 3842 3843 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3844 llvm::StringRef symbol_name_ref(symbol_name); 3845 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3846 symbol_name_non_abi_mangled = symbol_name + 1; 3847 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3848 type = eSymbolTypeObjCClass; 3849 demangled_is_synthesized = true; 3850 3851 } else if (symbol_name_ref.startswith( 3852 g_objc_v2_prefix_metaclass)) { 3853 symbol_name_non_abi_mangled = symbol_name + 1; 3854 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3855 type = eSymbolTypeObjCMetaClass; 3856 demangled_is_synthesized = true; 3857 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3858 symbol_name_non_abi_mangled = symbol_name + 1; 3859 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3860 type = eSymbolTypeObjCIVar; 3861 demangled_is_synthesized = true; 3862 } 3863 } else { 3864 if (nlist.n_value != 0) 3865 symbol_section = 3866 section_info.GetSection(nlist.n_sect, nlist.n_value); 3867 type = eSymbolTypeData; 3868 } 3869 break; 3870 3871 case N_FNAME: 3872 // procedure name (f77 kludge): name,,NO_SECT,0,0 3873 type = eSymbolTypeCompiler; 3874 break; 3875 3876 case N_FUN: 3877 // procedure: name,,n_sect,linenumber,address 3878 if (symbol_name) { 3879 type = eSymbolTypeCode; 3880 symbol_section = 3881 section_info.GetSection(nlist.n_sect, nlist.n_value); 3882 3883 N_FUN_addr_to_sym_idx.insert( 3884 std::make_pair(nlist.n_value, sym_idx)); 3885 // We use the current number of symbols in the symbol table in 3886 // lieu of using nlist_idx in case we ever start trimming entries 3887 // out 3888 N_FUN_indexes.push_back(sym_idx); 3889 } else { 3890 type = eSymbolTypeCompiler; 3891 3892 if (!N_FUN_indexes.empty()) { 3893 // Copy the size of the function into the original STAB entry 3894 // so we don't have to hunt for it later 3895 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3896 ->SetByteSize(nlist.n_value); 3897 N_FUN_indexes.pop_back(); 3898 // We don't really need the end function STAB as it contains 3899 // the size which we already placed with the original symbol, 3900 // so don't add it if we want a minimal symbol table 3901 add_nlist = false; 3902 } 3903 } 3904 break; 3905 3906 case N_STSYM: 3907 // static symbol: name,,n_sect,type,address 3908 N_STSYM_addr_to_sym_idx.insert( 3909 std::make_pair(nlist.n_value, sym_idx)); 3910 symbol_section = 3911 section_info.GetSection(nlist.n_sect, nlist.n_value); 3912 if (symbol_name && symbol_name[0]) { 3913 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3914 eSymbolTypeData); 3915 } 3916 break; 3917 3918 case N_LCSYM: 3919 // .lcomm symbol: name,,n_sect,type,address 3920 symbol_section = 3921 section_info.GetSection(nlist.n_sect, nlist.n_value); 3922 type = eSymbolTypeCommonBlock; 3923 break; 3924 3925 case N_BNSYM: 3926 // We use the current number of symbols in the symbol table in lieu 3927 // of using nlist_idx in case we ever start trimming entries out 3928 // Skip these if we want minimal symbol tables 3929 add_nlist = false; 3930 break; 3931 3932 case N_ENSYM: 3933 // Set the size of the N_BNSYM to the terminating index of this 3934 // N_ENSYM so that we can always skip the entire symbol if we need 3935 // to navigate more quickly at the source level when parsing STABS 3936 // Skip these if we want minimal symbol tables 3937 add_nlist = false; 3938 break; 3939 3940 case N_OPT: 3941 // emitted with gcc2_compiled and in gcc source 3942 type = eSymbolTypeCompiler; 3943 break; 3944 3945 case N_RSYM: 3946 // register sym: name,,NO_SECT,type,register 3947 type = eSymbolTypeVariable; 3948 break; 3949 3950 case N_SLINE: 3951 // src line: 0,,n_sect,linenumber,address 3952 symbol_section = 3953 section_info.GetSection(nlist.n_sect, nlist.n_value); 3954 type = eSymbolTypeLineEntry; 3955 break; 3956 3957 case N_SSYM: 3958 // structure elt: name,,NO_SECT,type,struct_offset 3959 type = eSymbolTypeVariableType; 3960 break; 3961 3962 case N_SO: 3963 // source file name 3964 type = eSymbolTypeSourceFile; 3965 if (symbol_name == nullptr) { 3966 add_nlist = false; 3967 if (N_SO_index != UINT32_MAX) { 3968 // Set the size of the N_SO to the terminating index of this 3969 // N_SO so that we can always skip the entire N_SO if we need 3970 // to navigate more quickly at the source level when parsing 3971 // STABS 3972 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3973 symbol_ptr->SetByteSize(sym_idx); 3974 symbol_ptr->SetSizeIsSibling(true); 3975 } 3976 N_NSYM_indexes.clear(); 3977 N_INCL_indexes.clear(); 3978 N_BRAC_indexes.clear(); 3979 N_COMM_indexes.clear(); 3980 N_FUN_indexes.clear(); 3981 N_SO_index = UINT32_MAX; 3982 } else { 3983 // We use the current number of symbols in the symbol table in 3984 // lieu of using nlist_idx in case we ever start trimming entries 3985 // out 3986 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3987 if (N_SO_has_full_path) { 3988 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3989 // We have two consecutive N_SO entries where the first 3990 // contains a directory and the second contains a full path. 3991 sym[sym_idx - 1].GetMangled().SetValue( 3992 ConstString(symbol_name), false); 3993 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3994 add_nlist = false; 3995 } else { 3996 // This is the first entry in a N_SO that contains a 3997 // directory or a full path to the source file 3998 N_SO_index = sym_idx; 3999 } 4000 } else if ((N_SO_index == sym_idx - 1) && 4001 ((sym_idx - 1) < num_syms)) { 4002 // This is usually the second N_SO entry that contains just the 4003 // filename, so here we combine it with the first one if we are 4004 // minimizing the symbol table 4005 const char *so_path = 4006 sym[sym_idx - 1] 4007 .GetMangled() 4008 .GetDemangledName(lldb::eLanguageTypeUnknown) 4009 .AsCString(); 4010 if (so_path && so_path[0]) { 4011 std::string full_so_path(so_path); 4012 const size_t double_slash_pos = full_so_path.find("//"); 4013 if (double_slash_pos != std::string::npos) { 4014 // The linker has been generating bad N_SO entries with 4015 // doubled up paths in the format "%s%s" where the first 4016 // string in the DW_AT_comp_dir, and the second is the 4017 // directory for the source file so you end up with a path 4018 // that looks like "/tmp/src//tmp/src/" 4019 FileSpec so_dir(so_path); 4020 if (!FileSystem::Instance().Exists(so_dir)) { 4021 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 4022 FileSpec::Style::native); 4023 if (FileSystem::Instance().Exists(so_dir)) { 4024 // Trim off the incorrect path 4025 full_so_path.erase(0, double_slash_pos + 1); 4026 } 4027 } 4028 } 4029 if (*full_so_path.rbegin() != '/') 4030 full_so_path += '/'; 4031 full_so_path += symbol_name; 4032 sym[sym_idx - 1].GetMangled().SetValue( 4033 ConstString(full_so_path.c_str()), false); 4034 add_nlist = false; 4035 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4036 } 4037 } else { 4038 // This could be a relative path to a N_SO 4039 N_SO_index = sym_idx; 4040 } 4041 } 4042 break; 4043 4044 case N_OSO: 4045 // object file name: name,,0,0,st_mtime 4046 type = eSymbolTypeObjectFile; 4047 break; 4048 4049 case N_LSYM: 4050 // local sym: name,,NO_SECT,type,offset 4051 type = eSymbolTypeLocal; 4052 break; 4053 4054 // INCL scopes 4055 case N_BINCL: 4056 // include file beginning: name,,NO_SECT,0,sum We use the current 4057 // number of symbols in the symbol table in lieu of using nlist_idx 4058 // in case we ever start trimming entries out 4059 N_INCL_indexes.push_back(sym_idx); 4060 type = eSymbolTypeScopeBegin; 4061 break; 4062 4063 case N_EINCL: 4064 // include file end: name,,NO_SECT,0,0 4065 // Set the size of the N_BINCL to the terminating index of this 4066 // N_EINCL so that we can always skip the entire symbol if we need 4067 // to navigate more quickly at the source level when parsing STABS 4068 if (!N_INCL_indexes.empty()) { 4069 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4070 symbol_ptr->SetByteSize(sym_idx + 1); 4071 symbol_ptr->SetSizeIsSibling(true); 4072 N_INCL_indexes.pop_back(); 4073 } 4074 type = eSymbolTypeScopeEnd; 4075 break; 4076 4077 case N_SOL: 4078 // #included file name: name,,n_sect,0,address 4079 type = eSymbolTypeHeaderFile; 4080 4081 // We currently don't use the header files on darwin 4082 add_nlist = false; 4083 break; 4084 4085 case N_PARAMS: 4086 // compiler parameters: name,,NO_SECT,0,0 4087 type = eSymbolTypeCompiler; 4088 break; 4089 4090 case N_VERSION: 4091 // compiler version: name,,NO_SECT,0,0 4092 type = eSymbolTypeCompiler; 4093 break; 4094 4095 case N_OLEVEL: 4096 // compiler -O level: name,,NO_SECT,0,0 4097 type = eSymbolTypeCompiler; 4098 break; 4099 4100 case N_PSYM: 4101 // parameter: name,,NO_SECT,type,offset 4102 type = eSymbolTypeVariable; 4103 break; 4104 4105 case N_ENTRY: 4106 // alternate entry: name,,n_sect,linenumber,address 4107 symbol_section = 4108 section_info.GetSection(nlist.n_sect, nlist.n_value); 4109 type = eSymbolTypeLineEntry; 4110 break; 4111 4112 // Left and Right Braces 4113 case N_LBRAC: 4114 // left bracket: 0,,NO_SECT,nesting level,address We use the 4115 // current number of symbols in the symbol table in lieu of using 4116 // nlist_idx in case we ever start trimming entries out 4117 symbol_section = 4118 section_info.GetSection(nlist.n_sect, nlist.n_value); 4119 N_BRAC_indexes.push_back(sym_idx); 4120 type = eSymbolTypeScopeBegin; 4121 break; 4122 4123 case N_RBRAC: 4124 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4125 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4126 // can always skip the entire symbol if we need to navigate more 4127 // quickly at the source level when parsing STABS 4128 symbol_section = 4129 section_info.GetSection(nlist.n_sect, nlist.n_value); 4130 if (!N_BRAC_indexes.empty()) { 4131 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4132 symbol_ptr->SetByteSize(sym_idx + 1); 4133 symbol_ptr->SetSizeIsSibling(true); 4134 N_BRAC_indexes.pop_back(); 4135 } 4136 type = eSymbolTypeScopeEnd; 4137 break; 4138 4139 case N_EXCL: 4140 // deleted include file: name,,NO_SECT,0,sum 4141 type = eSymbolTypeHeaderFile; 4142 break; 4143 4144 // COMM scopes 4145 case N_BCOMM: 4146 // begin common: name,,NO_SECT,0,0 4147 // We use the current number of symbols in the symbol table in lieu 4148 // of using nlist_idx in case we ever start trimming entries out 4149 type = eSymbolTypeScopeBegin; 4150 N_COMM_indexes.push_back(sym_idx); 4151 break; 4152 4153 case N_ECOML: 4154 // end common (local name): 0,,n_sect,0,address 4155 symbol_section = 4156 section_info.GetSection(nlist.n_sect, nlist.n_value); 4157 LLVM_FALLTHROUGH; 4158 4159 case N_ECOMM: 4160 // end common: name,,n_sect,0,0 4161 // Set the size of the N_BCOMM to the terminating index of this 4162 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4163 // we need to navigate more quickly at the source level when 4164 // parsing STABS 4165 if (!N_COMM_indexes.empty()) { 4166 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4167 symbol_ptr->SetByteSize(sym_idx + 1); 4168 symbol_ptr->SetSizeIsSibling(true); 4169 N_COMM_indexes.pop_back(); 4170 } 4171 type = eSymbolTypeScopeEnd; 4172 break; 4173 4174 case N_LENG: 4175 // second stab entry with length information 4176 type = eSymbolTypeAdditional; 4177 break; 4178 4179 default: 4180 break; 4181 } 4182 } else { 4183 // uint8_t n_pext = N_PEXT & nlist.n_type; 4184 uint8_t n_type = N_TYPE & nlist.n_type; 4185 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4186 4187 switch (n_type) { 4188 case N_INDR: { 4189 const char *reexport_name_cstr = 4190 strtab_data.PeekCStr(nlist.n_value); 4191 if (reexport_name_cstr && reexport_name_cstr[0]) { 4192 type = eSymbolTypeReExported; 4193 ConstString reexport_name( 4194 reexport_name_cstr + 4195 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4196 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4197 set_value = false; 4198 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4199 indirect_symbol_names.insert( 4200 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4201 } else 4202 type = eSymbolTypeUndefined; 4203 } break; 4204 4205 case N_UNDF: 4206 if (symbol_name && symbol_name[0]) { 4207 ConstString undefined_name(symbol_name + 4208 ((symbol_name[0] == '_') ? 1 : 0)); 4209 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4210 } 4211 LLVM_FALLTHROUGH; 4212 4213 case N_PBUD: 4214 type = eSymbolTypeUndefined; 4215 break; 4216 4217 case N_ABS: 4218 type = eSymbolTypeAbsolute; 4219 break; 4220 4221 case N_SECT: { 4222 symbol_section = 4223 section_info.GetSection(nlist.n_sect, nlist.n_value); 4224 4225 if (!symbol_section) { 4226 // TODO: warn about this? 4227 add_nlist = false; 4228 break; 4229 } 4230 4231 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4232 type = eSymbolTypeException; 4233 } else { 4234 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4235 4236 switch (section_type) { 4237 case S_CSTRING_LITERALS: 4238 type = eSymbolTypeData; 4239 break; // section with only literal C strings 4240 case S_4BYTE_LITERALS: 4241 type = eSymbolTypeData; 4242 break; // section with only 4 byte literals 4243 case S_8BYTE_LITERALS: 4244 type = eSymbolTypeData; 4245 break; // section with only 8 byte literals 4246 case S_LITERAL_POINTERS: 4247 type = eSymbolTypeTrampoline; 4248 break; // section with only pointers to literals 4249 case S_NON_LAZY_SYMBOL_POINTERS: 4250 type = eSymbolTypeTrampoline; 4251 break; // section with only non-lazy symbol pointers 4252 case S_LAZY_SYMBOL_POINTERS: 4253 type = eSymbolTypeTrampoline; 4254 break; // section with only lazy symbol pointers 4255 case S_SYMBOL_STUBS: 4256 type = eSymbolTypeTrampoline; 4257 break; // section with only symbol stubs, byte size of stub in 4258 // the reserved2 field 4259 case S_MOD_INIT_FUNC_POINTERS: 4260 type = eSymbolTypeCode; 4261 break; // section with only function pointers for initialization 4262 case S_MOD_TERM_FUNC_POINTERS: 4263 type = eSymbolTypeCode; 4264 break; // section with only function pointers for termination 4265 case S_INTERPOSING: 4266 type = eSymbolTypeTrampoline; 4267 break; // section with only pairs of function pointers for 4268 // interposing 4269 case S_16BYTE_LITERALS: 4270 type = eSymbolTypeData; 4271 break; // section with only 16 byte literals 4272 case S_DTRACE_DOF: 4273 type = eSymbolTypeInstrumentation; 4274 break; 4275 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4276 type = eSymbolTypeTrampoline; 4277 break; 4278 default: 4279 switch (symbol_section->GetType()) { 4280 case lldb::eSectionTypeCode: 4281 type = eSymbolTypeCode; 4282 break; 4283 case eSectionTypeData: 4284 case eSectionTypeDataCString: // Inlined C string data 4285 case eSectionTypeDataCStringPointers: // Pointers to C string 4286 // data 4287 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4288 // the symbol table 4289 case eSectionTypeData4: 4290 case eSectionTypeData8: 4291 case eSectionTypeData16: 4292 type = eSymbolTypeData; 4293 break; 4294 default: 4295 break; 4296 } 4297 break; 4298 } 4299 4300 if (type == eSymbolTypeInvalid) { 4301 const char *symbol_sect_name = 4302 symbol_section->GetName().AsCString(); 4303 if (symbol_section->IsDescendant(text_section_sp.get())) { 4304 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4305 S_ATTR_SELF_MODIFYING_CODE | 4306 S_ATTR_SOME_INSTRUCTIONS)) 4307 type = eSymbolTypeData; 4308 else 4309 type = eSymbolTypeCode; 4310 } else if (symbol_section->IsDescendant( 4311 data_section_sp.get()) || 4312 symbol_section->IsDescendant( 4313 data_dirty_section_sp.get()) || 4314 symbol_section->IsDescendant( 4315 data_const_section_sp.get())) { 4316 if (symbol_sect_name && 4317 ::strstr(symbol_sect_name, "__objc") == 4318 symbol_sect_name) { 4319 type = eSymbolTypeRuntime; 4320 4321 if (symbol_name) { 4322 llvm::StringRef symbol_name_ref(symbol_name); 4323 if (symbol_name_ref.startswith("_OBJC_")) { 4324 static const llvm::StringRef g_objc_v2_prefix_class( 4325 "_OBJC_CLASS_$_"); 4326 static const llvm::StringRef g_objc_v2_prefix_metaclass( 4327 "_OBJC_METACLASS_$_"); 4328 static const llvm::StringRef g_objc_v2_prefix_ivar( 4329 "_OBJC_IVAR_$_"); 4330 if (symbol_name_ref.startswith( 4331 g_objc_v2_prefix_class)) { 4332 symbol_name_non_abi_mangled = symbol_name + 1; 4333 symbol_name = 4334 symbol_name + g_objc_v2_prefix_class.size(); 4335 type = eSymbolTypeObjCClass; 4336 demangled_is_synthesized = true; 4337 } else if (symbol_name_ref.startswith( 4338 g_objc_v2_prefix_metaclass)) { 4339 symbol_name_non_abi_mangled = symbol_name + 1; 4340 symbol_name = 4341 symbol_name + g_objc_v2_prefix_metaclass.size(); 4342 type = eSymbolTypeObjCMetaClass; 4343 demangled_is_synthesized = true; 4344 } else if (symbol_name_ref.startswith( 4345 g_objc_v2_prefix_ivar)) { 4346 symbol_name_non_abi_mangled = symbol_name + 1; 4347 symbol_name = 4348 symbol_name + g_objc_v2_prefix_ivar.size(); 4349 type = eSymbolTypeObjCIVar; 4350 demangled_is_synthesized = true; 4351 } 4352 } 4353 } 4354 } else if (symbol_sect_name && 4355 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4356 symbol_sect_name) { 4357 type = eSymbolTypeException; 4358 } else { 4359 type = eSymbolTypeData; 4360 } 4361 } else if (symbol_sect_name && 4362 ::strstr(symbol_sect_name, "__IMPORT") == 4363 symbol_sect_name) { 4364 type = eSymbolTypeTrampoline; 4365 } else if (symbol_section->IsDescendant( 4366 objc_section_sp.get())) { 4367 type = eSymbolTypeRuntime; 4368 if (symbol_name && symbol_name[0] == '.') { 4369 llvm::StringRef symbol_name_ref(symbol_name); 4370 static const llvm::StringRef g_objc_v1_prefix_class( 4371 ".objc_class_name_"); 4372 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4373 symbol_name_non_abi_mangled = symbol_name; 4374 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4375 type = eSymbolTypeObjCClass; 4376 demangled_is_synthesized = true; 4377 } 4378 } 4379 } 4380 } 4381 } 4382 } break; 4383 } 4384 } 4385 4386 if (add_nlist) { 4387 uint64_t symbol_value = nlist.n_value; 4388 4389 if (symbol_name_non_abi_mangled) { 4390 sym[sym_idx].GetMangled().SetMangledName( 4391 ConstString(symbol_name_non_abi_mangled)); 4392 sym[sym_idx].GetMangled().SetDemangledName( 4393 ConstString(symbol_name)); 4394 } else { 4395 bool symbol_name_is_mangled = false; 4396 4397 if (symbol_name && symbol_name[0] == '_') { 4398 symbol_name_is_mangled = symbol_name[1] == '_'; 4399 symbol_name++; // Skip the leading underscore 4400 } 4401 4402 if (symbol_name) { 4403 ConstString const_symbol_name(symbol_name); 4404 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4405 symbol_name_is_mangled); 4406 } 4407 } 4408 4409 if (is_gsym) { 4410 const char *gsym_name = sym[sym_idx] 4411 .GetMangled() 4412 .GetName(lldb::eLanguageTypeUnknown, 4413 Mangled::ePreferMangled) 4414 .GetCString(); 4415 if (gsym_name) 4416 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4417 } 4418 4419 if (symbol_section) { 4420 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4421 if (symbol_byte_size == 0 && function_starts_count > 0) { 4422 addr_t symbol_lookup_file_addr = nlist.n_value; 4423 // Do an exact address match for non-ARM addresses, else get the 4424 // closest since the symbol might be a thumb symbol which has an 4425 // address with bit zero set 4426 FunctionStarts::Entry *func_start_entry = 4427 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4428 if (is_arm && func_start_entry) { 4429 // Verify that the function start address is the symbol address 4430 // (ARM) or the symbol address + 1 (thumb) 4431 if (func_start_entry->addr != symbol_lookup_file_addr && 4432 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4433 // Not the right entry, NULL it out... 4434 func_start_entry = nullptr; 4435 } 4436 } 4437 if (func_start_entry) { 4438 func_start_entry->data = true; 4439 4440 addr_t symbol_file_addr = func_start_entry->addr; 4441 if (is_arm) 4442 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4443 4444 const FunctionStarts::Entry *next_func_start_entry = 4445 function_starts.FindNextEntry(func_start_entry); 4446 const addr_t section_end_file_addr = 4447 section_file_addr + symbol_section->GetByteSize(); 4448 if (next_func_start_entry) { 4449 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4450 // Be sure the clear the Thumb address bit when we calculate 4451 // the size from the current and next address 4452 if (is_arm) 4453 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4454 symbol_byte_size = std::min<lldb::addr_t>( 4455 next_symbol_file_addr - symbol_file_addr, 4456 section_end_file_addr - symbol_file_addr); 4457 } else { 4458 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4459 } 4460 } 4461 } 4462 symbol_value -= section_file_addr; 4463 } 4464 4465 if (!is_debug) { 4466 if (type == eSymbolTypeCode) { 4467 // See if we can find a N_FUN entry for any code symbols. If we 4468 // do find a match, and the name matches, then we can merge the 4469 // two into just the function symbol to avoid duplicate entries 4470 // in the symbol table 4471 std::pair<ValueToSymbolIndexMap::const_iterator, 4472 ValueToSymbolIndexMap::const_iterator> 4473 range; 4474 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4475 if (range.first != range.second) { 4476 bool found_it = false; 4477 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4478 pos != range.second; ++pos) { 4479 if (sym[sym_idx].GetMangled().GetName( 4480 lldb::eLanguageTypeUnknown, 4481 Mangled::ePreferMangled) == 4482 sym[pos->second].GetMangled().GetName( 4483 lldb::eLanguageTypeUnknown, 4484 Mangled::ePreferMangled)) { 4485 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4486 // We just need the flags from the linker symbol, so put 4487 // these flags into the N_FUN flags to avoid duplicate 4488 // symbols in the symbol table 4489 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4490 sym[pos->second].SetFlags(nlist.n_type << 16 | 4491 nlist.n_desc); 4492 if (resolver_addresses.find(nlist.n_value) != 4493 resolver_addresses.end()) 4494 sym[pos->second].SetType(eSymbolTypeResolver); 4495 sym[sym_idx].Clear(); 4496 found_it = true; 4497 break; 4498 } 4499 } 4500 if (found_it) 4501 continue; 4502 } else { 4503 if (resolver_addresses.find(nlist.n_value) != 4504 resolver_addresses.end()) 4505 type = eSymbolTypeResolver; 4506 } 4507 } else if (type == eSymbolTypeData || 4508 type == eSymbolTypeObjCClass || 4509 type == eSymbolTypeObjCMetaClass || 4510 type == eSymbolTypeObjCIVar) { 4511 // See if we can find a N_STSYM entry for any data symbols. If we 4512 // do find a match, and the name matches, then we can merge the 4513 // two into just the Static symbol to avoid duplicate entries in 4514 // the symbol table 4515 std::pair<ValueToSymbolIndexMap::const_iterator, 4516 ValueToSymbolIndexMap::const_iterator> 4517 range; 4518 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4519 if (range.first != range.second) { 4520 bool found_it = false; 4521 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4522 pos != range.second; ++pos) { 4523 if (sym[sym_idx].GetMangled().GetName( 4524 lldb::eLanguageTypeUnknown, 4525 Mangled::ePreferMangled) == 4526 sym[pos->second].GetMangled().GetName( 4527 lldb::eLanguageTypeUnknown, 4528 Mangled::ePreferMangled)) { 4529 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4530 // We just need the flags from the linker symbol, so put 4531 // these flags into the N_STSYM flags to avoid duplicate 4532 // symbols in the symbol table 4533 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4534 sym[pos->second].SetFlags(nlist.n_type << 16 | 4535 nlist.n_desc); 4536 sym[sym_idx].Clear(); 4537 found_it = true; 4538 break; 4539 } 4540 } 4541 if (found_it) 4542 continue; 4543 } else { 4544 // Combine N_GSYM stab entries with the non stab symbol 4545 const char *gsym_name = sym[sym_idx] 4546 .GetMangled() 4547 .GetName(lldb::eLanguageTypeUnknown, 4548 Mangled::ePreferMangled) 4549 .GetCString(); 4550 if (gsym_name) { 4551 ConstNameToSymbolIndexMap::const_iterator pos = 4552 N_GSYM_name_to_sym_idx.find(gsym_name); 4553 if (pos != N_GSYM_name_to_sym_idx.end()) { 4554 const uint32_t GSYM_sym_idx = pos->second; 4555 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4556 // Copy the address, because often the N_GSYM address has 4557 // an invalid address of zero when the global is a common 4558 // symbol 4559 sym[GSYM_sym_idx].GetAddressRef().SetSection( 4560 symbol_section); 4561 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4562 // We just need the flags from the linker symbol, so put 4563 // these flags into the N_GSYM flags to avoid duplicate 4564 // symbols in the symbol table 4565 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 4566 nlist.n_desc); 4567 sym[sym_idx].Clear(); 4568 continue; 4569 } 4570 } 4571 } 4572 } 4573 } 4574 4575 sym[sym_idx].SetID(nlist_idx); 4576 sym[sym_idx].SetType(type); 4577 if (set_value) { 4578 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4579 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4580 } 4581 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4582 if (nlist.n_desc & N_WEAK_REF) 4583 sym[sym_idx].SetIsWeak(true); 4584 4585 if (symbol_byte_size > 0) 4586 sym[sym_idx].SetByteSize(symbol_byte_size); 4587 4588 if (demangled_is_synthesized) 4589 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4590 4591 ++sym_idx; 4592 } else { 4593 sym[sym_idx].Clear(); 4594 } 4595 } 4596 4597 for (const auto &pos : reexport_shlib_needs_fixup) { 4598 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4599 if (undef_pos != undefined_name_to_desc.end()) { 4600 const uint8_t dylib_ordinal = 4601 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4602 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4603 sym[pos.first].SetReExportedSymbolSharedLibrary( 4604 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4605 } 4606 } 4607 } 4608 4609 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4610 4611 if (function_starts_count > 0) { 4612 uint32_t num_synthetic_function_symbols = 0; 4613 for (i = 0; i < function_starts_count; ++i) { 4614 if (!function_starts.GetEntryRef(i).data) 4615 ++num_synthetic_function_symbols; 4616 } 4617 4618 if (num_synthetic_function_symbols > 0) { 4619 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4620 num_syms = sym_idx + num_synthetic_function_symbols; 4621 sym = symtab->Resize(num_syms); 4622 } 4623 for (i = 0; i < function_starts_count; ++i) { 4624 const FunctionStarts::Entry *func_start_entry = 4625 function_starts.GetEntryAtIndex(i); 4626 if (!func_start_entry->data) { 4627 addr_t symbol_file_addr = func_start_entry->addr; 4628 uint32_t symbol_flags = 0; 4629 if (is_arm) { 4630 if (symbol_file_addr & 1) 4631 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4632 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4633 } 4634 Address symbol_addr; 4635 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4636 SectionSP symbol_section(symbol_addr.GetSection()); 4637 uint32_t symbol_byte_size = 0; 4638 if (symbol_section) { 4639 const addr_t section_file_addr = 4640 symbol_section->GetFileAddress(); 4641 const FunctionStarts::Entry *next_func_start_entry = 4642 function_starts.FindNextEntry(func_start_entry); 4643 const addr_t section_end_file_addr = 4644 section_file_addr + symbol_section->GetByteSize(); 4645 if (next_func_start_entry) { 4646 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4647 if (is_arm) 4648 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4649 symbol_byte_size = std::min<lldb::addr_t>( 4650 next_symbol_file_addr - symbol_file_addr, 4651 section_end_file_addr - symbol_file_addr); 4652 } else { 4653 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4654 } 4655 sym[sym_idx].SetID(synthetic_sym_id++); 4656 sym[sym_idx].GetMangled().SetDemangledName( 4657 GetNextSyntheticSymbolName()); 4658 sym[sym_idx].SetType(eSymbolTypeCode); 4659 sym[sym_idx].SetIsSynthetic(true); 4660 sym[sym_idx].GetAddressRef() = symbol_addr; 4661 if (symbol_flags) 4662 sym[sym_idx].SetFlags(symbol_flags); 4663 if (symbol_byte_size) 4664 sym[sym_idx].SetByteSize(symbol_byte_size); 4665 ++sym_idx; 4666 } 4667 } 4668 } 4669 } 4670 } 4671 } 4672 4673 // Trim our symbols down to just what we ended up with after removing any 4674 // symbols. 4675 if (sym_idx < num_syms) { 4676 num_syms = sym_idx; 4677 sym = symtab->Resize(num_syms); 4678 } 4679 4680 // Now synthesize indirect symbols 4681 if (m_dysymtab.nindirectsyms != 0) { 4682 if (indirect_symbol_index_data.GetByteSize()) { 4683 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4684 m_nlist_idx_to_sym_idx.end(); 4685 4686 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4687 ++sect_idx) { 4688 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4689 S_SYMBOL_STUBS) { 4690 uint32_t symbol_stub_byte_size = 4691 m_mach_sections[sect_idx].reserved2; 4692 if (symbol_stub_byte_size == 0) 4693 continue; 4694 4695 const uint32_t num_symbol_stubs = 4696 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4697 4698 if (num_symbol_stubs == 0) 4699 continue; 4700 4701 const uint32_t symbol_stub_index_offset = 4702 m_mach_sections[sect_idx].reserved1; 4703 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; 4704 ++stub_idx) { 4705 const uint32_t symbol_stub_index = 4706 symbol_stub_index_offset + stub_idx; 4707 const lldb::addr_t symbol_stub_addr = 4708 m_mach_sections[sect_idx].addr + 4709 (stub_idx * symbol_stub_byte_size); 4710 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4711 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4712 symbol_stub_offset, 4)) { 4713 const uint32_t stub_sym_id = 4714 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4715 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4716 continue; 4717 4718 NListIndexToSymbolIndexMap::const_iterator index_pos = 4719 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4720 Symbol *stub_symbol = nullptr; 4721 if (index_pos != end_index_pos) { 4722 // We have a remapping from the original nlist index to a 4723 // current symbol index, so just look this up by index 4724 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4725 } else { 4726 // We need to lookup a symbol using the original nlist symbol 4727 // index since this index is coming from the S_SYMBOL_STUBS 4728 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4729 } 4730 4731 if (stub_symbol) { 4732 Address so_addr(symbol_stub_addr, section_list); 4733 4734 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4735 // Change the external symbol into a trampoline that makes 4736 // sense These symbols were N_UNDF N_EXT, and are useless 4737 // to us, so we can re-use them so we don't have to make up 4738 // a synthetic symbol for no good reason. 4739 if (resolver_addresses.find(symbol_stub_addr) == 4740 resolver_addresses.end()) 4741 stub_symbol->SetType(eSymbolTypeTrampoline); 4742 else 4743 stub_symbol->SetType(eSymbolTypeResolver); 4744 stub_symbol->SetExternal(false); 4745 stub_symbol->GetAddressRef() = so_addr; 4746 stub_symbol->SetByteSize(symbol_stub_byte_size); 4747 } else { 4748 // Make a synthetic symbol to describe the trampoline stub 4749 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4750 if (sym_idx >= num_syms) { 4751 sym = symtab->Resize(++num_syms); 4752 stub_symbol = nullptr; // this pointer no longer valid 4753 } 4754 sym[sym_idx].SetID(synthetic_sym_id++); 4755 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4756 if (resolver_addresses.find(symbol_stub_addr) == 4757 resolver_addresses.end()) 4758 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4759 else 4760 sym[sym_idx].SetType(eSymbolTypeResolver); 4761 sym[sym_idx].SetIsSynthetic(true); 4762 sym[sym_idx].GetAddressRef() = so_addr; 4763 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4764 ++sym_idx; 4765 } 4766 } else { 4767 if (log) 4768 log->Warning("symbol stub referencing symbol table symbol " 4769 "%u that isn't in our minimal symbol table, " 4770 "fix this!!!", 4771 stub_sym_id); 4772 } 4773 } 4774 } 4775 } 4776 } 4777 } 4778 } 4779 4780 if (!trie_entries.empty()) { 4781 for (const auto &e : trie_entries) { 4782 if (e.entry.import_name) { 4783 // Only add indirect symbols from the Trie entries if we didn't have 4784 // a N_INDR nlist entry for this already 4785 if (indirect_symbol_names.find(e.entry.name) == 4786 indirect_symbol_names.end()) { 4787 // Make a synthetic symbol to describe re-exported symbol. 4788 if (sym_idx >= num_syms) 4789 sym = symtab->Resize(++num_syms); 4790 sym[sym_idx].SetID(synthetic_sym_id++); 4791 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4792 sym[sym_idx].SetType(eSymbolTypeReExported); 4793 sym[sym_idx].SetIsSynthetic(true); 4794 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4795 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4796 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4797 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4798 } 4799 ++sym_idx; 4800 } 4801 } 4802 } 4803 } 4804 4805 // StreamFile s(stdout, false); 4806 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4807 // symtab->Dump(&s, NULL, eSortOrderNone); 4808 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4809 // sizes 4810 symtab->CalculateSymbolSizes(); 4811 4812 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4813 // symtab->Dump(&s, NULL, eSortOrderNone); 4814 4815 return symtab->GetNumSymbols(); 4816 } 4817 return 0; 4818 } 4819 4820 void ObjectFileMachO::Dump(Stream *s) { 4821 ModuleSP module_sp(GetModule()); 4822 if (module_sp) { 4823 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4824 s->Printf("%p: ", static_cast<void *>(this)); 4825 s->Indent(); 4826 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4827 s->PutCString("ObjectFileMachO64"); 4828 else 4829 s->PutCString("ObjectFileMachO32"); 4830 4831 ArchSpec header_arch = GetArchitecture(); 4832 4833 *s << ", file = '" << m_file 4834 << "', triple = " << header_arch.GetTriple().getTriple() << "\n"; 4835 4836 SectionList *sections = GetSectionList(); 4837 if (sections) 4838 sections->Dump(s, nullptr, true, UINT32_MAX); 4839 4840 if (m_symtab_up) 4841 m_symtab_up->Dump(s, nullptr, eSortOrderNone); 4842 } 4843 } 4844 4845 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4846 const lldb_private::DataExtractor &data, 4847 lldb::offset_t lc_offset) { 4848 uint32_t i; 4849 struct uuid_command load_cmd; 4850 4851 lldb::offset_t offset = lc_offset; 4852 for (i = 0; i < header.ncmds; ++i) { 4853 const lldb::offset_t cmd_offset = offset; 4854 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4855 break; 4856 4857 if (load_cmd.cmd == LC_UUID) { 4858 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4859 4860 if (uuid_bytes) { 4861 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4862 // We pretend these object files have no UUID to prevent crashing. 4863 4864 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4865 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4866 0xbb, 0x14, 0xf0, 0x0d}; 4867 4868 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4869 return UUID(); 4870 4871 return UUID::fromOptionalData(uuid_bytes, 16); 4872 } 4873 return UUID(); 4874 } 4875 offset = cmd_offset + load_cmd.cmdsize; 4876 } 4877 return UUID(); 4878 } 4879 4880 static llvm::StringRef GetOSName(uint32_t cmd) { 4881 switch (cmd) { 4882 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4883 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4884 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4885 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4886 case llvm::MachO::LC_VERSION_MIN_TVOS: 4887 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4888 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4889 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4890 default: 4891 llvm_unreachable("unexpected LC_VERSION load command"); 4892 } 4893 } 4894 4895 namespace { 4896 struct OSEnv { 4897 llvm::StringRef os_type; 4898 llvm::StringRef environment; 4899 OSEnv(uint32_t cmd) { 4900 switch (cmd) { 4901 case PLATFORM_MACOS: 4902 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4903 return; 4904 case PLATFORM_IOS: 4905 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4906 return; 4907 case PLATFORM_TVOS: 4908 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4909 return; 4910 case PLATFORM_WATCHOS: 4911 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4912 return; 4913 // NEED_BRIDGEOS_TRIPLE case PLATFORM_BRIDGEOS: 4914 // NEED_BRIDGEOS_TRIPLE os_type = llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4915 // NEED_BRIDGEOS_TRIPLE return; 4916 #if defined (PLATFORM_IOSSIMULATOR) && defined (PLATFORM_TVOSSIMULATOR) && defined (PLATFORM_WATCHOSSIMULATOR) 4917 case PLATFORM_IOSSIMULATOR: 4918 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4919 environment = 4920 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4921 return; 4922 case PLATFORM_TVOSSIMULATOR: 4923 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4924 environment = 4925 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4926 return; 4927 case PLATFORM_WATCHOSSIMULATOR: 4928 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4929 environment = 4930 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4931 return; 4932 #endif 4933 default: { 4934 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 4935 LIBLLDB_LOG_PROCESS)); 4936 if (log) 4937 log->Printf("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 if (log) 5722 log->Printf("inferior process shared cache has a UUID of %s, base address 0x%" PRIx64 , uuid.GetAsString().c_str(), base_addr); 5723 } 5724 5725 // From dyld SPI header dyld_process_info.h 5726 typedef void *dyld_process_info; 5727 struct lldb_copy__dyld_process_cache_info { 5728 uuid_t cacheUUID; // UUID of cache used by process 5729 uint64_t cacheBaseAddress; // load address of dyld shared cache 5730 bool noCache; // process is running without a dyld cache 5731 bool privateCache; // process is using a private copy of its dyld cache 5732 }; 5733 5734 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with llvm 5735 // enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile errors. 5736 // So we need to use the actual underlying types of task_t and kern_return_t 5737 // below. 5738 extern "C" unsigned int /*task_t*/ mach_task_self(); 5739 5740 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5741 uuid.Clear(); 5742 base_addr = LLDB_INVALID_ADDRESS; 5743 5744 #if defined(__APPLE__) && \ 5745 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5746 uint8_t *(*dyld_get_all_image_infos)(void); 5747 dyld_get_all_image_infos = 5748 (uint8_t * (*)())dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5749 if (dyld_get_all_image_infos) { 5750 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5751 if (dyld_all_image_infos_address) { 5752 uint32_t *version = (uint32_t *) 5753 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5754 if (*version >= 13) { 5755 uuid_t *sharedCacheUUID_address = 0; 5756 int wordsize = sizeof(uint8_t *); 5757 if (wordsize == 8) { 5758 sharedCacheUUID_address = 5759 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5760 160); // sharedCacheUUID <mach-o/dyld_images.h> 5761 if (*version >= 15) 5762 base_addr = *(uint64_t *) ((uint8_t *) dyld_all_image_infos_address 5763 + 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5764 } else { 5765 sharedCacheUUID_address = 5766 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5767 84); // sharedCacheUUID <mach-o/dyld_images.h> 5768 if (*version >= 15) { 5769 base_addr = 0; 5770 base_addr = *(uint32_t *) ((uint8_t *) dyld_all_image_infos_address 5771 + 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5772 } 5773 } 5774 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5775 } 5776 } 5777 } else { 5778 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5779 dyld_process_info (*dyld_process_info_create)(unsigned int /* task_t */ task, uint64_t timestamp, unsigned int /*kern_return_t*/ *kernelError); 5780 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5781 void (*dyld_process_info_release)(dyld_process_info info); 5782 5783 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, unsigned int /*kern_return_t*/ *)) 5784 dlsym (RTLD_DEFAULT, "_dyld_process_info_create"); 5785 dyld_process_info_get_cache = (void (*)(void *, void *)) 5786 dlsym (RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5787 dyld_process_info_release = (void (*)(void *)) 5788 dlsym (RTLD_DEFAULT, "_dyld_process_info_release"); 5789 5790 if (dyld_process_info_create && dyld_process_info_get_cache) { 5791 unsigned int /*kern_return_t */ kern_ret; 5792 dyld_process_info process_info = dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5793 if (process_info) { 5794 struct lldb_copy__dyld_process_cache_info sc_info; 5795 memset (&sc_info, 0, sizeof (struct lldb_copy__dyld_process_cache_info)); 5796 dyld_process_info_get_cache (process_info, &sc_info); 5797 if (sc_info.cacheBaseAddress != 0) { 5798 base_addr = sc_info.cacheBaseAddress; 5799 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 5800 } 5801 dyld_process_info_release (process_info); 5802 } 5803 } 5804 } 5805 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5806 if (log && uuid.IsValid()) 5807 log->Printf("lldb's in-memory shared cache has a UUID of %s base address of 0x%" PRIx64, uuid.GetAsString().c_str(), base_addr); 5808 #endif 5809 } 5810 5811 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5812 if (!m_min_os_version) { 5813 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5814 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5815 const lldb::offset_t load_cmd_offset = offset; 5816 5817 version_min_command lc; 5818 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5819 break; 5820 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5821 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5822 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5823 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5824 if (m_data.GetU32(&offset, &lc.version, 5825 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5826 const uint32_t xxxx = lc.version >> 16; 5827 const uint32_t yy = (lc.version >> 8) & 0xffu; 5828 const uint32_t zz = lc.version & 0xffu; 5829 if (xxxx) { 5830 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5831 break; 5832 } 5833 } 5834 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5835 // struct build_version_command { 5836 // uint32_t cmd; /* LC_BUILD_VERSION */ 5837 // uint32_t cmdsize; /* sizeof(struct build_version_command) plus */ 5838 // /* ntools * sizeof(struct build_tool_version) */ 5839 // uint32_t platform; /* platform */ 5840 // uint32_t minos; /* X.Y.Z is encoded in nibbles xxxx.yy.zz */ 5841 // uint32_t sdk; /* X.Y.Z is encoded in nibbles xxxx.yy.zz */ 5842 // uint32_t ntools; /* number of tool entries following this */ 5843 // }; 5844 5845 offset += 4; // skip platform 5846 uint32_t minos = m_data.GetU32(&offset); 5847 5848 const uint32_t xxxx = minos >> 16; 5849 const uint32_t yy = (minos >> 8) & 0xffu; 5850 const uint32_t zz = minos & 0xffu; 5851 if (xxxx) { 5852 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5853 break; 5854 } 5855 } 5856 5857 offset = load_cmd_offset + lc.cmdsize; 5858 } 5859 5860 if (!m_min_os_version) { 5861 // Set version to an empty value so we don't keep trying to 5862 m_min_os_version = llvm::VersionTuple(); 5863 } 5864 } 5865 5866 return *m_min_os_version; 5867 } 5868 5869 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { 5870 if (!m_sdk_versions.hasValue()) { 5871 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5872 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5873 const lldb::offset_t load_cmd_offset = offset; 5874 5875 version_min_command lc; 5876 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5877 break; 5878 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5879 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5880 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5881 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5882 if (m_data.GetU32(&offset, &lc.version, 5883 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5884 const uint32_t xxxx = lc.sdk >> 16; 5885 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5886 const uint32_t zz = lc.sdk & 0xffu; 5887 if (xxxx) { 5888 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 5889 break; 5890 } else { 5891 GetModule()->ReportWarning( 5892 "minimum OS version load command with invalid (0) version found."); 5893 } 5894 } 5895 } 5896 offset = load_cmd_offset + lc.cmdsize; 5897 } 5898 5899 if (!m_sdk_versions.hasValue()) { 5900 offset = MachHeaderSizeFromMagic(m_header.magic); 5901 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5902 const lldb::offset_t load_cmd_offset = offset; 5903 5904 version_min_command lc; 5905 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5906 break; 5907 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5908 // struct build_version_command { 5909 // uint32_t cmd; /* LC_BUILD_VERSION */ 5910 // uint32_t cmdsize; /* sizeof(struct 5911 // build_version_command) plus */ 5912 // /* ntools * sizeof(struct 5913 // build_tool_version) */ 5914 // uint32_t platform; /* platform */ 5915 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5916 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded 5917 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number 5918 // of tool entries following this */ 5919 // }; 5920 5921 offset += 4; // skip platform 5922 uint32_t minos = m_data.GetU32(&offset); 5923 5924 const uint32_t xxxx = minos >> 16; 5925 const uint32_t yy = (minos >> 8) & 0xffu; 5926 const uint32_t zz = minos & 0xffu; 5927 if (xxxx) { 5928 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 5929 break; 5930 } 5931 } 5932 offset = load_cmd_offset + lc.cmdsize; 5933 } 5934 } 5935 5936 if (!m_sdk_versions.hasValue()) 5937 m_sdk_versions = llvm::VersionTuple(); 5938 } 5939 5940 return m_sdk_versions.getValue(); 5941 } 5942 5943 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5944 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5945 } 5946 5947 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5948 return m_allow_assembly_emulation_unwind_plans; 5949 } 5950 5951 // PluginInterface protocol 5952 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5953 return GetPluginNameStatic(); 5954 } 5955 5956 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5957 5958 Section *ObjectFileMachO::GetMachHeaderSection() { 5959 // Find the first address of the mach header which is the first non-zero file 5960 // sized section whose file offset is zero. This is the base file address of 5961 // the mach-o file which can be subtracted from the vmaddr of the other 5962 // segments found in memory and added to the load address 5963 ModuleSP module_sp = GetModule(); 5964 if (!module_sp) 5965 return nullptr; 5966 SectionList *section_list = GetSectionList(); 5967 if (!section_list) 5968 return nullptr; 5969 const size_t num_sections = section_list->GetSize(); 5970 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5971 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5972 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 5973 return section; 5974 } 5975 return nullptr; 5976 } 5977 5978 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 5979 if (!section) 5980 return false; 5981 const bool is_dsym = (m_header.filetype == MH_DSYM); 5982 if (section->GetFileSize() == 0 && !is_dsym) 5983 return false; 5984 if (section->IsThreadSpecific()) 5985 return false; 5986 if (GetModule().get() != section->GetModule().get()) 5987 return false; 5988 // Be careful with __LINKEDIT and __DWARF segments 5989 if (section->GetName() == GetSegmentNameLINKEDIT() || 5990 section->GetName() == GetSegmentNameDWARF()) { 5991 // Only map __LINKEDIT and __DWARF if we have an in memory image and 5992 // this isn't a kernel binary like a kext or mach_kernel. 5993 const bool is_memory_image = (bool)m_process_wp.lock(); 5994 const Strata strata = GetStrata(); 5995 if (is_memory_image == false || strata == eStrataKernel) 5996 return false; 5997 } 5998 return true; 5999 } 6000 6001 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 6002 lldb::addr_t header_load_address, const Section *header_section, 6003 const Section *section) { 6004 ModuleSP module_sp = GetModule(); 6005 if (module_sp && header_section && section && 6006 header_load_address != LLDB_INVALID_ADDRESS) { 6007 lldb::addr_t file_addr = header_section->GetFileAddress(); 6008 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 6009 return section->GetFileAddress() - file_addr + header_load_address; 6010 } 6011 return LLDB_INVALID_ADDRESS; 6012 } 6013 6014 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 6015 bool value_is_offset) { 6016 ModuleSP module_sp = GetModule(); 6017 if (module_sp) { 6018 size_t num_loaded_sections = 0; 6019 SectionList *section_list = GetSectionList(); 6020 if (section_list) { 6021 const size_t num_sections = section_list->GetSize(); 6022 6023 if (value_is_offset) { 6024 // "value" is an offset to apply to each top level segment 6025 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6026 // Iterate through the object file sections to find all of the 6027 // sections that size on disk (to avoid __PAGEZERO) and load them 6028 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6029 if (SectionIsLoadable(section_sp.get())) 6030 if (target.GetSectionLoadList().SetSectionLoadAddress( 6031 section_sp, section_sp->GetFileAddress() + value)) 6032 ++num_loaded_sections; 6033 } 6034 } else { 6035 // "value" is the new base address of the mach_header, adjust each 6036 // section accordingly 6037 6038 Section *mach_header_section = GetMachHeaderSection(); 6039 if (mach_header_section) { 6040 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6041 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6042 6043 lldb::addr_t section_load_addr = 6044 CalculateSectionLoadAddressForMemoryImage( 6045 value, mach_header_section, section_sp.get()); 6046 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6047 if (target.GetSectionLoadList().SetSectionLoadAddress( 6048 section_sp, section_load_addr)) 6049 ++num_loaded_sections; 6050 } 6051 } 6052 } 6053 } 6054 } 6055 return num_loaded_sections > 0; 6056 } 6057 return false; 6058 } 6059 6060 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6061 const FileSpec &outfile, Status &error) { 6062 if (process_sp) { 6063 Target &target = process_sp->GetTarget(); 6064 const ArchSpec target_arch = target.GetArchitecture(); 6065 const llvm::Triple &target_triple = target_arch.GetTriple(); 6066 if (target_triple.getVendor() == llvm::Triple::Apple && 6067 (target_triple.getOS() == llvm::Triple::MacOSX || 6068 target_triple.getOS() == llvm::Triple::IOS || 6069 target_triple.getOS() == llvm::Triple::WatchOS || 6070 target_triple.getOS() == llvm::Triple::TvOS)) { 6071 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) { 6072 bool make_core = false; 6073 switch (target_arch.GetMachine()) { 6074 case llvm::Triple::aarch64: 6075 case llvm::Triple::arm: 6076 case llvm::Triple::thumb: 6077 case llvm::Triple::x86: 6078 case llvm::Triple::x86_64: 6079 make_core = true; 6080 break; 6081 default: 6082 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6083 target_triple.str().c_str()); 6084 break; 6085 } 6086 6087 if (make_core) { 6088 std::vector<segment_command_64> segment_load_commands; 6089 // uint32_t range_info_idx = 0; 6090 MemoryRegionInfo range_info; 6091 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6092 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6093 const ByteOrder byte_order = target_arch.GetByteOrder(); 6094 if (range_error.Success()) { 6095 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6096 const addr_t addr = range_info.GetRange().GetRangeBase(); 6097 const addr_t size = range_info.GetRange().GetByteSize(); 6098 6099 if (size == 0) 6100 break; 6101 6102 // Calculate correct protections 6103 uint32_t prot = 0; 6104 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6105 prot |= VM_PROT_READ; 6106 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6107 prot |= VM_PROT_WRITE; 6108 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6109 prot |= VM_PROT_EXECUTE; 6110 6111 if (prot != 0) { 6112 uint32_t cmd_type = LC_SEGMENT_64; 6113 uint32_t segment_size = sizeof(segment_command_64); 6114 if (addr_byte_size == 4) { 6115 cmd_type = LC_SEGMENT; 6116 segment_size = sizeof(segment_command); 6117 } 6118 segment_command_64 segment = { 6119 cmd_type, // uint32_t cmd; 6120 segment_size, // uint32_t cmdsize; 6121 {0}, // char segname[16]; 6122 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6123 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6124 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6125 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6126 prot, // uint32_t maxprot; 6127 prot, // uint32_t initprot; 6128 0, // uint32_t nsects; 6129 0}; // uint32_t flags; 6130 segment_load_commands.push_back(segment); 6131 } else { 6132 // No protections and a size of 1 used to be returned from old 6133 // debugservers when we asked about a region that was past the 6134 // last memory region and it indicates the end... 6135 if (size == 1) 6136 break; 6137 } 6138 6139 range_error = process_sp->GetMemoryRegionInfo( 6140 range_info.GetRange().GetRangeEnd(), range_info); 6141 if (range_error.Fail()) 6142 break; 6143 } 6144 6145 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6146 6147 mach_header_64 mach_header; 6148 if (addr_byte_size == 8) { 6149 mach_header.magic = MH_MAGIC_64; 6150 } else { 6151 mach_header.magic = MH_MAGIC; 6152 } 6153 mach_header.cputype = target_arch.GetMachOCPUType(); 6154 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6155 mach_header.filetype = MH_CORE; 6156 mach_header.ncmds = segment_load_commands.size(); 6157 mach_header.flags = 0; 6158 mach_header.reserved = 0; 6159 ThreadList &thread_list = process_sp->GetThreadList(); 6160 const uint32_t num_threads = thread_list.GetSize(); 6161 6162 // Make an array of LC_THREAD data items. Each one contains the 6163 // contents of the LC_THREAD load command. The data doesn't contain 6164 // the load command + load command size, we will add the load command 6165 // and load command size as we emit the data. 6166 std::vector<StreamString> LC_THREAD_datas(num_threads); 6167 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6168 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6169 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6170 LC_THREAD_data.SetByteOrder(byte_order); 6171 } 6172 for (uint32_t thread_idx = 0; thread_idx < num_threads; 6173 ++thread_idx) { 6174 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6175 if (thread_sp) { 6176 switch (mach_header.cputype) { 6177 case llvm::MachO::CPU_TYPE_ARM64: 6178 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6179 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6180 break; 6181 6182 case llvm::MachO::CPU_TYPE_ARM: 6183 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6184 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6185 break; 6186 6187 case llvm::MachO::CPU_TYPE_I386: 6188 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6189 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6190 break; 6191 6192 case llvm::MachO::CPU_TYPE_X86_64: 6193 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6194 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6195 break; 6196 } 6197 } 6198 } 6199 6200 // The size of the load command is the size of the segments... 6201 if (addr_byte_size == 8) { 6202 mach_header.sizeofcmds = segment_load_commands.size() * 6203 sizeof(struct segment_command_64); 6204 } else { 6205 mach_header.sizeofcmds = 6206 segment_load_commands.size() * sizeof(struct segment_command); 6207 } 6208 6209 // and the size of all LC_THREAD load command 6210 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6211 ++mach_header.ncmds; 6212 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6213 } 6214 6215 // Write the mach header 6216 buffer.PutHex32(mach_header.magic); 6217 buffer.PutHex32(mach_header.cputype); 6218 buffer.PutHex32(mach_header.cpusubtype); 6219 buffer.PutHex32(mach_header.filetype); 6220 buffer.PutHex32(mach_header.ncmds); 6221 buffer.PutHex32(mach_header.sizeofcmds); 6222 buffer.PutHex32(mach_header.flags); 6223 if (addr_byte_size == 8) { 6224 buffer.PutHex32(mach_header.reserved); 6225 } 6226 6227 // Skip the mach header and all load commands and align to the next 6228 // 0x1000 byte boundary 6229 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6230 if (file_offset & 0x00000fff) { 6231 file_offset += 0x00001000ull; 6232 file_offset &= (~0x00001000ull + 1); 6233 } 6234 6235 for (auto &segment : segment_load_commands) { 6236 segment.fileoff = file_offset; 6237 file_offset += segment.filesize; 6238 } 6239 6240 // Write out all of the LC_THREAD load commands 6241 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6242 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6243 buffer.PutHex32(LC_THREAD); 6244 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6245 buffer.Write(LC_THREAD_data.GetString().data(), 6246 LC_THREAD_data_size); 6247 } 6248 6249 // Write out all of the segment load commands 6250 for (const auto &segment : segment_load_commands) { 6251 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6252 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6253 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6254 segment.cmd, segment.cmdsize, segment.vmaddr, 6255 segment.vmaddr + segment.vmsize, segment.fileoff, 6256 segment.filesize, segment.maxprot, segment.initprot, 6257 segment.nsects, segment.flags); 6258 6259 buffer.PutHex32(segment.cmd); 6260 buffer.PutHex32(segment.cmdsize); 6261 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6262 if (addr_byte_size == 8) { 6263 buffer.PutHex64(segment.vmaddr); 6264 buffer.PutHex64(segment.vmsize); 6265 buffer.PutHex64(segment.fileoff); 6266 buffer.PutHex64(segment.filesize); 6267 } else { 6268 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6269 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6270 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6271 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6272 } 6273 buffer.PutHex32(segment.maxprot); 6274 buffer.PutHex32(segment.initprot); 6275 buffer.PutHex32(segment.nsects); 6276 buffer.PutHex32(segment.flags); 6277 } 6278 6279 File core_file; 6280 std::string core_file_path(outfile.GetPath()); 6281 error = FileSystem::Instance().Open(core_file, outfile, 6282 File::eOpenOptionWrite | 6283 File::eOpenOptionTruncate | 6284 File::eOpenOptionCanCreate); 6285 if (error.Success()) { 6286 // Read 1 page at a time 6287 uint8_t bytes[0x1000]; 6288 // Write the mach header and load commands out to the core file 6289 size_t bytes_written = buffer.GetString().size(); 6290 error = core_file.Write(buffer.GetString().data(), bytes_written); 6291 if (error.Success()) { 6292 // Now write the file data for all memory segments in the process 6293 for (const auto &segment : segment_load_commands) { 6294 if (core_file.SeekFromStart(segment.fileoff) == -1) { 6295 error.SetErrorStringWithFormat( 6296 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6297 segment.fileoff, core_file_path.c_str()); 6298 break; 6299 } 6300 6301 printf("Saving %" PRId64 6302 " bytes of data for memory region at 0x%" PRIx64 "\n", 6303 segment.vmsize, segment.vmaddr); 6304 addr_t bytes_left = segment.vmsize; 6305 addr_t addr = segment.vmaddr; 6306 Status memory_read_error; 6307 while (bytes_left > 0 && error.Success()) { 6308 const size_t bytes_to_read = 6309 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6310 6311 // In a savecore setting, we don't really care about caching, 6312 // as the data is dumped and very likely never read again, 6313 // so we call ReadMemoryFromInferior to bypass it. 6314 const size_t bytes_read = process_sp->ReadMemoryFromInferior( 6315 addr, bytes, bytes_to_read, memory_read_error); 6316 6317 if (bytes_read == bytes_to_read) { 6318 size_t bytes_written = bytes_read; 6319 error = core_file.Write(bytes, bytes_written); 6320 bytes_left -= bytes_read; 6321 addr += bytes_read; 6322 } else { 6323 // Some pages within regions are not readable, those should 6324 // be zero filled 6325 memset(bytes, 0, bytes_to_read); 6326 size_t bytes_written = bytes_to_read; 6327 error = core_file.Write(bytes, bytes_written); 6328 bytes_left -= bytes_to_read; 6329 addr += bytes_to_read; 6330 } 6331 } 6332 } 6333 } 6334 } 6335 } else { 6336 error.SetErrorString( 6337 "process doesn't support getting memory region info"); 6338 } 6339 } 6340 return true; // This is the right plug to handle saving core files for 6341 // this process 6342 } 6343 } 6344 return false; 6345 } 6346