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 static void PrintRegisterValue(RegisterContext *reg_ctx, const char *name, 112 const char *alt_name, size_t reg_byte_size, 113 Stream &data) { 114 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 115 if (reg_info == nullptr) 116 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 117 if (reg_info) { 118 lldb_private::RegisterValue reg_value; 119 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 120 if (reg_info->byte_size >= reg_byte_size) 121 data.Write(reg_value.GetBytes(), reg_byte_size); 122 else { 123 data.Write(reg_value.GetBytes(), reg_info->byte_size); 124 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; ++i) 125 data.PutChar(0); 126 } 127 return; 128 } 129 } 130 // Just write zeros if all else fails 131 for (size_t i = 0; i < reg_byte_size; ++i) 132 data.PutChar(0); 133 } 134 135 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 136 public: 137 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 138 const DataExtractor &data) 139 : RegisterContextDarwin_x86_64(thread, 0) { 140 SetRegisterDataFrom_LC_THREAD(data); 141 } 142 143 void InvalidateAllRegisters() override { 144 // Do nothing... registers are always valid... 145 } 146 147 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 148 lldb::offset_t offset = 0; 149 SetError(GPRRegSet, Read, -1); 150 SetError(FPURegSet, Read, -1); 151 SetError(EXCRegSet, Read, -1); 152 bool done = false; 153 154 while (!done) { 155 int flavor = data.GetU32(&offset); 156 if (flavor == 0) 157 done = true; 158 else { 159 uint32_t i; 160 uint32_t count = data.GetU32(&offset); 161 switch (flavor) { 162 case GPRRegSet: 163 for (i = 0; i < count; ++i) 164 (&gpr.rax)[i] = data.GetU64(&offset); 165 SetError(GPRRegSet, Read, 0); 166 done = true; 167 168 break; 169 case FPURegSet: 170 // TODO: fill in FPU regs.... 171 // SetError (FPURegSet, Read, -1); 172 done = true; 173 174 break; 175 case EXCRegSet: 176 exc.trapno = data.GetU32(&offset); 177 exc.err = data.GetU32(&offset); 178 exc.faultvaddr = data.GetU64(&offset); 179 SetError(EXCRegSet, Read, 0); 180 done = true; 181 break; 182 case 7: 183 case 8: 184 case 9: 185 // fancy flavors that encapsulate of the above flavors... 186 break; 187 188 default: 189 done = true; 190 break; 191 } 192 } 193 } 194 } 195 196 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 197 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 198 if (reg_ctx_sp) { 199 RegisterContext *reg_ctx = reg_ctx_sp.get(); 200 201 data.PutHex32(GPRRegSet); // Flavor 202 data.PutHex32(GPRWordCount); 203 PrintRegisterValue(reg_ctx, "rax", nullptr, 8, data); 204 PrintRegisterValue(reg_ctx, "rbx", nullptr, 8, data); 205 PrintRegisterValue(reg_ctx, "rcx", nullptr, 8, data); 206 PrintRegisterValue(reg_ctx, "rdx", nullptr, 8, data); 207 PrintRegisterValue(reg_ctx, "rdi", nullptr, 8, data); 208 PrintRegisterValue(reg_ctx, "rsi", nullptr, 8, data); 209 PrintRegisterValue(reg_ctx, "rbp", nullptr, 8, data); 210 PrintRegisterValue(reg_ctx, "rsp", nullptr, 8, data); 211 PrintRegisterValue(reg_ctx, "r8", nullptr, 8, data); 212 PrintRegisterValue(reg_ctx, "r9", nullptr, 8, data); 213 PrintRegisterValue(reg_ctx, "r10", nullptr, 8, data); 214 PrintRegisterValue(reg_ctx, "r11", nullptr, 8, data); 215 PrintRegisterValue(reg_ctx, "r12", nullptr, 8, data); 216 PrintRegisterValue(reg_ctx, "r13", nullptr, 8, data); 217 PrintRegisterValue(reg_ctx, "r14", nullptr, 8, data); 218 PrintRegisterValue(reg_ctx, "r15", nullptr, 8, data); 219 PrintRegisterValue(reg_ctx, "rip", nullptr, 8, data); 220 PrintRegisterValue(reg_ctx, "rflags", nullptr, 8, data); 221 PrintRegisterValue(reg_ctx, "cs", nullptr, 8, data); 222 PrintRegisterValue(reg_ctx, "fs", nullptr, 8, data); 223 PrintRegisterValue(reg_ctx, "gs", nullptr, 8, data); 224 225 // // Write out the FPU registers 226 // const size_t fpu_byte_size = sizeof(FPU); 227 // size_t bytes_written = 0; 228 // data.PutHex32 (FPURegSet); 229 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 230 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 231 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 232 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 233 // data); // uint16_t fcw; // "fctrl" 234 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 235 // data); // uint16_t fsw; // "fstat" 236 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 237 // data); // uint8_t ftw; // "ftag" 238 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 239 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 240 // data); // uint16_t fop; // "fop" 241 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 242 // data); // uint32_t ip; // "fioff" 243 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 244 // data); // uint16_t cs; // "fiseg" 245 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 246 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 247 // data); // uint32_t dp; // "fooff" 248 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 249 // data); // uint16_t ds; // "foseg" 250 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 251 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 252 // data); // uint32_t mxcsr; 253 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 254 // 4, data);// uint32_t mxcsrmask; 255 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 256 // sizeof(MMSReg), data); 257 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 258 // sizeof(MMSReg), data); 259 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 260 // sizeof(MMSReg), data); 261 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 262 // sizeof(MMSReg), data); 263 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 264 // sizeof(MMSReg), data); 265 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 266 // sizeof(MMSReg), data); 267 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 268 // sizeof(MMSReg), data); 269 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 270 // sizeof(MMSReg), data); 271 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 272 // sizeof(XMMReg), data); 273 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 274 // sizeof(XMMReg), data); 275 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 276 // sizeof(XMMReg), data); 277 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 278 // sizeof(XMMReg), data); 279 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 280 // sizeof(XMMReg), data); 281 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 282 // sizeof(XMMReg), data); 283 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 284 // sizeof(XMMReg), data); 285 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 286 // sizeof(XMMReg), data); 287 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 288 // sizeof(XMMReg), data); 289 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 290 // sizeof(XMMReg), data); 291 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 292 // sizeof(XMMReg), data); 293 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 294 // sizeof(XMMReg), data); 295 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 296 // sizeof(XMMReg), data); 297 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 298 // sizeof(XMMReg), data); 299 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 300 // sizeof(XMMReg), data); 301 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 302 // sizeof(XMMReg), data); 303 // 304 // // Fill rest with zeros 305 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 306 // i) 307 // data.PutChar(0); 308 309 // Write out the EXC registers 310 data.PutHex32(EXCRegSet); 311 data.PutHex32(EXCWordCount); 312 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 313 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 314 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 8, data); 315 return true; 316 } 317 return false; 318 } 319 320 protected: 321 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 322 323 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 324 325 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 326 327 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 328 return 0; 329 } 330 331 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 332 return 0; 333 } 334 335 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 336 return 0; 337 } 338 }; 339 340 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 341 public: 342 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 343 const DataExtractor &data) 344 : RegisterContextDarwin_i386(thread, 0) { 345 SetRegisterDataFrom_LC_THREAD(data); 346 } 347 348 void InvalidateAllRegisters() override { 349 // Do nothing... registers are always valid... 350 } 351 352 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 353 lldb::offset_t offset = 0; 354 SetError(GPRRegSet, Read, -1); 355 SetError(FPURegSet, Read, -1); 356 SetError(EXCRegSet, Read, -1); 357 bool done = false; 358 359 while (!done) { 360 int flavor = data.GetU32(&offset); 361 if (flavor == 0) 362 done = true; 363 else { 364 uint32_t i; 365 uint32_t count = data.GetU32(&offset); 366 switch (flavor) { 367 case GPRRegSet: 368 for (i = 0; i < count; ++i) 369 (&gpr.eax)[i] = data.GetU32(&offset); 370 SetError(GPRRegSet, Read, 0); 371 done = true; 372 373 break; 374 case FPURegSet: 375 // TODO: fill in FPU regs.... 376 // SetError (FPURegSet, Read, -1); 377 done = true; 378 379 break; 380 case EXCRegSet: 381 exc.trapno = data.GetU32(&offset); 382 exc.err = data.GetU32(&offset); 383 exc.faultvaddr = data.GetU32(&offset); 384 SetError(EXCRegSet, Read, 0); 385 done = true; 386 break; 387 case 7: 388 case 8: 389 case 9: 390 // fancy flavors that encapsulate of the above flavors... 391 break; 392 393 default: 394 done = true; 395 break; 396 } 397 } 398 } 399 } 400 401 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 402 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 403 if (reg_ctx_sp) { 404 RegisterContext *reg_ctx = reg_ctx_sp.get(); 405 406 data.PutHex32(GPRRegSet); // Flavor 407 data.PutHex32(GPRWordCount); 408 PrintRegisterValue(reg_ctx, "eax", nullptr, 4, data); 409 PrintRegisterValue(reg_ctx, "ebx", nullptr, 4, data); 410 PrintRegisterValue(reg_ctx, "ecx", nullptr, 4, data); 411 PrintRegisterValue(reg_ctx, "edx", nullptr, 4, data); 412 PrintRegisterValue(reg_ctx, "edi", nullptr, 4, data); 413 PrintRegisterValue(reg_ctx, "esi", nullptr, 4, data); 414 PrintRegisterValue(reg_ctx, "ebp", nullptr, 4, data); 415 PrintRegisterValue(reg_ctx, "esp", nullptr, 4, data); 416 PrintRegisterValue(reg_ctx, "ss", nullptr, 4, data); 417 PrintRegisterValue(reg_ctx, "eflags", nullptr, 4, data); 418 PrintRegisterValue(reg_ctx, "eip", nullptr, 4, data); 419 PrintRegisterValue(reg_ctx, "cs", nullptr, 4, data); 420 PrintRegisterValue(reg_ctx, "ds", nullptr, 4, data); 421 PrintRegisterValue(reg_ctx, "es", nullptr, 4, data); 422 PrintRegisterValue(reg_ctx, "fs", nullptr, 4, data); 423 PrintRegisterValue(reg_ctx, "gs", nullptr, 4, data); 424 425 // Write out the EXC registers 426 data.PutHex32(EXCRegSet); 427 data.PutHex32(EXCWordCount); 428 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 429 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 430 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 4, data); 431 return true; 432 } 433 return false; 434 } 435 436 protected: 437 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 438 439 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 440 441 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 442 443 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 444 return 0; 445 } 446 447 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 448 return 0; 449 } 450 451 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 452 return 0; 453 } 454 }; 455 456 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 457 public: 458 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 459 const DataExtractor &data) 460 : RegisterContextDarwin_arm(thread, 0) { 461 SetRegisterDataFrom_LC_THREAD(data); 462 } 463 464 void InvalidateAllRegisters() override { 465 // Do nothing... registers are always valid... 466 } 467 468 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 469 lldb::offset_t offset = 0; 470 SetError(GPRRegSet, Read, -1); 471 SetError(FPURegSet, Read, -1); 472 SetError(EXCRegSet, Read, -1); 473 bool done = false; 474 475 while (!done) { 476 int flavor = data.GetU32(&offset); 477 uint32_t count = data.GetU32(&offset); 478 lldb::offset_t next_thread_state = offset + (count * 4); 479 switch (flavor) { 480 case GPRAltRegSet: 481 case GPRRegSet: 482 for (uint32_t i = 0; i < count; ++i) { 483 gpr.r[i] = data.GetU32(&offset); 484 } 485 486 // Note that gpr.cpsr is also copied by the above loop; this loop 487 // technically extends one element past the end of the gpr.r[] array. 488 489 SetError(GPRRegSet, Read, 0); 490 offset = next_thread_state; 491 break; 492 493 case FPURegSet: { 494 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 495 const int fpu_reg_buf_size = sizeof(fpu.floats); 496 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 497 fpu_reg_buf) == fpu_reg_buf_size) { 498 offset += fpu_reg_buf_size; 499 fpu.fpscr = data.GetU32(&offset); 500 SetError(FPURegSet, Read, 0); 501 } else { 502 done = true; 503 } 504 } 505 offset = next_thread_state; 506 break; 507 508 case EXCRegSet: 509 if (count == 3) { 510 exc.exception = data.GetU32(&offset); 511 exc.fsr = data.GetU32(&offset); 512 exc.far = data.GetU32(&offset); 513 SetError(EXCRegSet, Read, 0); 514 } 515 done = true; 516 offset = next_thread_state; 517 break; 518 519 // Unknown register set flavor, stop trying to parse. 520 default: 521 done = true; 522 } 523 } 524 } 525 526 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 527 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 528 if (reg_ctx_sp) { 529 RegisterContext *reg_ctx = reg_ctx_sp.get(); 530 531 data.PutHex32(GPRRegSet); // Flavor 532 data.PutHex32(GPRWordCount); 533 PrintRegisterValue(reg_ctx, "r0", nullptr, 4, data); 534 PrintRegisterValue(reg_ctx, "r1", nullptr, 4, data); 535 PrintRegisterValue(reg_ctx, "r2", nullptr, 4, data); 536 PrintRegisterValue(reg_ctx, "r3", nullptr, 4, data); 537 PrintRegisterValue(reg_ctx, "r4", nullptr, 4, data); 538 PrintRegisterValue(reg_ctx, "r5", nullptr, 4, data); 539 PrintRegisterValue(reg_ctx, "r6", nullptr, 4, data); 540 PrintRegisterValue(reg_ctx, "r7", nullptr, 4, data); 541 PrintRegisterValue(reg_ctx, "r8", nullptr, 4, data); 542 PrintRegisterValue(reg_ctx, "r9", nullptr, 4, data); 543 PrintRegisterValue(reg_ctx, "r10", nullptr, 4, data); 544 PrintRegisterValue(reg_ctx, "r11", nullptr, 4, data); 545 PrintRegisterValue(reg_ctx, "r12", nullptr, 4, data); 546 PrintRegisterValue(reg_ctx, "sp", nullptr, 4, data); 547 PrintRegisterValue(reg_ctx, "lr", nullptr, 4, data); 548 PrintRegisterValue(reg_ctx, "pc", nullptr, 4, data); 549 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 550 551 // Write out the EXC registers 552 // data.PutHex32 (EXCRegSet); 553 // data.PutHex32 (EXCWordCount); 554 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 555 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 556 // WriteRegister (reg_ctx, "far", NULL, 4, data); 557 return true; 558 } 559 return false; 560 } 561 562 protected: 563 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 564 565 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 566 567 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 568 569 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 570 571 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 572 return 0; 573 } 574 575 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 576 return 0; 577 } 578 579 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 580 return 0; 581 } 582 583 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 584 return -1; 585 } 586 }; 587 588 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 589 public: 590 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 591 const DataExtractor &data) 592 : RegisterContextDarwin_arm64(thread, 0) { 593 SetRegisterDataFrom_LC_THREAD(data); 594 } 595 596 void InvalidateAllRegisters() override { 597 // Do nothing... registers are always valid... 598 } 599 600 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 601 lldb::offset_t offset = 0; 602 SetError(GPRRegSet, Read, -1); 603 SetError(FPURegSet, Read, -1); 604 SetError(EXCRegSet, Read, -1); 605 bool done = false; 606 while (!done) { 607 int flavor = data.GetU32(&offset); 608 uint32_t count = data.GetU32(&offset); 609 lldb::offset_t next_thread_state = offset + (count * 4); 610 switch (flavor) { 611 case GPRRegSet: 612 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 613 // 32-bit register) 614 if (count >= (33 * 2) + 1) { 615 for (uint32_t i = 0; i < 29; ++i) 616 gpr.x[i] = data.GetU64(&offset); 617 gpr.fp = data.GetU64(&offset); 618 gpr.lr = data.GetU64(&offset); 619 gpr.sp = data.GetU64(&offset); 620 gpr.pc = data.GetU64(&offset); 621 gpr.cpsr = data.GetU32(&offset); 622 SetError(GPRRegSet, Read, 0); 623 } 624 offset = next_thread_state; 625 break; 626 case FPURegSet: { 627 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 628 const int fpu_reg_buf_size = sizeof(fpu); 629 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 630 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 631 fpu_reg_buf) == fpu_reg_buf_size) { 632 SetError(FPURegSet, Read, 0); 633 } else { 634 done = true; 635 } 636 } 637 offset = next_thread_state; 638 break; 639 case EXCRegSet: 640 if (count == 4) { 641 exc.far = data.GetU64(&offset); 642 exc.esr = data.GetU32(&offset); 643 exc.exception = data.GetU32(&offset); 644 SetError(EXCRegSet, Read, 0); 645 } 646 offset = next_thread_state; 647 break; 648 default: 649 done = true; 650 break; 651 } 652 } 653 } 654 655 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 656 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 657 if (reg_ctx_sp) { 658 RegisterContext *reg_ctx = reg_ctx_sp.get(); 659 660 data.PutHex32(GPRRegSet); // Flavor 661 data.PutHex32(GPRWordCount); 662 PrintRegisterValue(reg_ctx, "x0", nullptr, 8, data); 663 PrintRegisterValue(reg_ctx, "x1", nullptr, 8, data); 664 PrintRegisterValue(reg_ctx, "x2", nullptr, 8, data); 665 PrintRegisterValue(reg_ctx, "x3", nullptr, 8, data); 666 PrintRegisterValue(reg_ctx, "x4", nullptr, 8, data); 667 PrintRegisterValue(reg_ctx, "x5", nullptr, 8, data); 668 PrintRegisterValue(reg_ctx, "x6", nullptr, 8, data); 669 PrintRegisterValue(reg_ctx, "x7", nullptr, 8, data); 670 PrintRegisterValue(reg_ctx, "x8", nullptr, 8, data); 671 PrintRegisterValue(reg_ctx, "x9", nullptr, 8, data); 672 PrintRegisterValue(reg_ctx, "x10", nullptr, 8, data); 673 PrintRegisterValue(reg_ctx, "x11", nullptr, 8, data); 674 PrintRegisterValue(reg_ctx, "x12", nullptr, 8, data); 675 PrintRegisterValue(reg_ctx, "x13", nullptr, 8, data); 676 PrintRegisterValue(reg_ctx, "x14", nullptr, 8, data); 677 PrintRegisterValue(reg_ctx, "x15", nullptr, 8, data); 678 PrintRegisterValue(reg_ctx, "x16", nullptr, 8, data); 679 PrintRegisterValue(reg_ctx, "x17", nullptr, 8, data); 680 PrintRegisterValue(reg_ctx, "x18", nullptr, 8, data); 681 PrintRegisterValue(reg_ctx, "x19", nullptr, 8, data); 682 PrintRegisterValue(reg_ctx, "x20", nullptr, 8, data); 683 PrintRegisterValue(reg_ctx, "x21", nullptr, 8, data); 684 PrintRegisterValue(reg_ctx, "x22", nullptr, 8, data); 685 PrintRegisterValue(reg_ctx, "x23", nullptr, 8, data); 686 PrintRegisterValue(reg_ctx, "x24", nullptr, 8, data); 687 PrintRegisterValue(reg_ctx, "x25", nullptr, 8, data); 688 PrintRegisterValue(reg_ctx, "x26", nullptr, 8, data); 689 PrintRegisterValue(reg_ctx, "x27", nullptr, 8, data); 690 PrintRegisterValue(reg_ctx, "x28", nullptr, 8, data); 691 PrintRegisterValue(reg_ctx, "fp", nullptr, 8, data); 692 PrintRegisterValue(reg_ctx, "lr", nullptr, 8, data); 693 PrintRegisterValue(reg_ctx, "sp", nullptr, 8, data); 694 PrintRegisterValue(reg_ctx, "pc", nullptr, 8, data); 695 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 696 697 // Write out the EXC registers 698 // data.PutHex32 (EXCRegSet); 699 // data.PutHex32 (EXCWordCount); 700 // WriteRegister (reg_ctx, "far", NULL, 8, data); 701 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 702 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 703 return true; 704 } 705 return false; 706 } 707 708 protected: 709 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 710 711 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 712 713 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 714 715 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 716 717 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 718 return 0; 719 } 720 721 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 722 return 0; 723 } 724 725 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 726 return 0; 727 } 728 729 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 730 return -1; 731 } 732 }; 733 734 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 735 switch (magic) { 736 case MH_MAGIC: 737 case MH_CIGAM: 738 return sizeof(struct mach_header); 739 740 case MH_MAGIC_64: 741 case MH_CIGAM_64: 742 return sizeof(struct mach_header_64); 743 break; 744 745 default: 746 break; 747 } 748 return 0; 749 } 750 751 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 752 753 char ObjectFileMachO::ID; 754 755 void ObjectFileMachO::Initialize() { 756 PluginManager::RegisterPlugin( 757 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 758 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 759 } 760 761 void ObjectFileMachO::Terminate() { 762 PluginManager::UnregisterPlugin(CreateInstance); 763 } 764 765 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 766 static ConstString g_name("mach-o"); 767 return g_name; 768 } 769 770 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 771 return "Mach-o object file reader (32 and 64 bit)"; 772 } 773 774 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 775 DataBufferSP &data_sp, 776 lldb::offset_t data_offset, 777 const FileSpec *file, 778 lldb::offset_t file_offset, 779 lldb::offset_t length) { 780 if (!data_sp) { 781 data_sp = MapFileData(*file, length, file_offset); 782 if (!data_sp) 783 return nullptr; 784 data_offset = 0; 785 } 786 787 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 788 return nullptr; 789 790 // Update the data to contain the entire file if it doesn't already 791 if (data_sp->GetByteSize() < length) { 792 data_sp = MapFileData(*file, length, file_offset); 793 if (!data_sp) 794 return nullptr; 795 data_offset = 0; 796 } 797 auto objfile_up = std::make_unique<ObjectFileMachO>( 798 module_sp, data_sp, data_offset, file, file_offset, length); 799 if (!objfile_up || !objfile_up->ParseHeader()) 800 return nullptr; 801 802 return objfile_up.release(); 803 } 804 805 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 806 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 807 const ProcessSP &process_sp, lldb::addr_t header_addr) { 808 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 809 std::unique_ptr<ObjectFile> objfile_up( 810 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 811 if (objfile_up.get() && objfile_up->ParseHeader()) 812 return objfile_up.release(); 813 } 814 return nullptr; 815 } 816 817 size_t ObjectFileMachO::GetModuleSpecifications( 818 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 819 lldb::offset_t data_offset, lldb::offset_t file_offset, 820 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 821 const size_t initial_count = specs.GetSize(); 822 823 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 824 DataExtractor data; 825 data.SetData(data_sp); 826 llvm::MachO::mach_header header; 827 if (ParseHeader(data, &data_offset, header)) { 828 size_t header_and_load_cmds = 829 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 830 if (header_and_load_cmds >= data_sp->GetByteSize()) { 831 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 832 data.SetData(data_sp); 833 data_offset = MachHeaderSizeFromMagic(header.magic); 834 } 835 if (data_sp) { 836 ModuleSpec base_spec; 837 base_spec.GetFileSpec() = file; 838 base_spec.SetObjectOffset(file_offset); 839 base_spec.SetObjectSize(length); 840 GetAllArchSpecs(header, data, data_offset, base_spec, specs); 841 } 842 } 843 } 844 return specs.GetSize() - initial_count; 845 } 846 847 ConstString ObjectFileMachO::GetSegmentNameTEXT() { 848 static ConstString g_segment_name_TEXT("__TEXT"); 849 return g_segment_name_TEXT; 850 } 851 852 ConstString ObjectFileMachO::GetSegmentNameDATA() { 853 static ConstString g_segment_name_DATA("__DATA"); 854 return g_segment_name_DATA; 855 } 856 857 ConstString ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 858 static ConstString g_segment_name("__DATA_DIRTY"); 859 return g_segment_name; 860 } 861 862 ConstString ObjectFileMachO::GetSegmentNameDATA_CONST() { 863 static ConstString g_segment_name("__DATA_CONST"); 864 return g_segment_name; 865 } 866 867 ConstString ObjectFileMachO::GetSegmentNameOBJC() { 868 static ConstString g_segment_name_OBJC("__OBJC"); 869 return g_segment_name_OBJC; 870 } 871 872 ConstString ObjectFileMachO::GetSegmentNameLINKEDIT() { 873 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 874 return g_section_name_LINKEDIT; 875 } 876 877 ConstString ObjectFileMachO::GetSegmentNameDWARF() { 878 static ConstString g_section_name("__DWARF"); 879 return g_section_name; 880 } 881 882 ConstString ObjectFileMachO::GetSectionNameEHFrame() { 883 static ConstString g_section_name_eh_frame("__eh_frame"); 884 return g_section_name_eh_frame; 885 } 886 887 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 888 lldb::addr_t data_offset, 889 lldb::addr_t data_length) { 890 DataExtractor data; 891 data.SetData(data_sp, data_offset, data_length); 892 lldb::offset_t offset = 0; 893 uint32_t magic = data.GetU32(&offset); 894 return MachHeaderSizeFromMagic(magic) != 0; 895 } 896 897 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 898 DataBufferSP &data_sp, 899 lldb::offset_t data_offset, 900 const FileSpec *file, 901 lldb::offset_t file_offset, 902 lldb::offset_t length) 903 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 904 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 905 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 906 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 907 ::memset(&m_header, 0, sizeof(m_header)); 908 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 909 } 910 911 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 912 lldb::DataBufferSP &header_data_sp, 913 const lldb::ProcessSP &process_sp, 914 lldb::addr_t header_addr) 915 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 916 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 917 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 918 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 919 ::memset(&m_header, 0, sizeof(m_header)); 920 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 921 } 922 923 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 924 lldb::offset_t *data_offset_ptr, 925 llvm::MachO::mach_header &header) { 926 data.SetByteOrder(endian::InlHostByteOrder()); 927 // Leave magic in the original byte order 928 header.magic = data.GetU32(data_offset_ptr); 929 bool can_parse = false; 930 bool is_64_bit = false; 931 switch (header.magic) { 932 case MH_MAGIC: 933 data.SetByteOrder(endian::InlHostByteOrder()); 934 data.SetAddressByteSize(4); 935 can_parse = true; 936 break; 937 938 case MH_MAGIC_64: 939 data.SetByteOrder(endian::InlHostByteOrder()); 940 data.SetAddressByteSize(8); 941 can_parse = true; 942 is_64_bit = true; 943 break; 944 945 case MH_CIGAM: 946 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 947 ? eByteOrderLittle 948 : eByteOrderBig); 949 data.SetAddressByteSize(4); 950 can_parse = true; 951 break; 952 953 case MH_CIGAM_64: 954 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 955 ? eByteOrderLittle 956 : eByteOrderBig); 957 data.SetAddressByteSize(8); 958 is_64_bit = true; 959 can_parse = true; 960 break; 961 962 default: 963 break; 964 } 965 966 if (can_parse) { 967 data.GetU32(data_offset_ptr, &header.cputype, 6); 968 if (is_64_bit) 969 *data_offset_ptr += 4; 970 return true; 971 } else { 972 memset(&header, 0, sizeof(header)); 973 } 974 return false; 975 } 976 977 bool ObjectFileMachO::ParseHeader() { 978 ModuleSP module_sp(GetModule()); 979 if (!module_sp) 980 return false; 981 982 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 983 bool can_parse = false; 984 lldb::offset_t offset = 0; 985 m_data.SetByteOrder(endian::InlHostByteOrder()); 986 // Leave magic in the original byte order 987 m_header.magic = m_data.GetU32(&offset); 988 switch (m_header.magic) { 989 case MH_MAGIC: 990 m_data.SetByteOrder(endian::InlHostByteOrder()); 991 m_data.SetAddressByteSize(4); 992 can_parse = true; 993 break; 994 995 case MH_MAGIC_64: 996 m_data.SetByteOrder(endian::InlHostByteOrder()); 997 m_data.SetAddressByteSize(8); 998 can_parse = true; 999 break; 1000 1001 case MH_CIGAM: 1002 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1003 ? eByteOrderLittle 1004 : eByteOrderBig); 1005 m_data.SetAddressByteSize(4); 1006 can_parse = true; 1007 break; 1008 1009 case MH_CIGAM_64: 1010 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1011 ? eByteOrderLittle 1012 : eByteOrderBig); 1013 m_data.SetAddressByteSize(8); 1014 can_parse = true; 1015 break; 1016 1017 default: 1018 break; 1019 } 1020 1021 if (can_parse) { 1022 m_data.GetU32(&offset, &m_header.cputype, 6); 1023 1024 ModuleSpecList all_specs; 1025 ModuleSpec base_spec; 1026 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 1027 base_spec, all_specs); 1028 1029 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 1030 ArchSpec mach_arch = 1031 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 1032 1033 // Check if the module has a required architecture 1034 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1035 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1036 continue; 1037 1038 if (SetModulesArchitecture(mach_arch)) { 1039 const size_t header_and_lc_size = 1040 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1041 if (m_data.GetByteSize() < header_and_lc_size) { 1042 DataBufferSP data_sp; 1043 ProcessSP process_sp(m_process_wp.lock()); 1044 if (process_sp) { 1045 data_sp = ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1046 } else { 1047 // Read in all only the load command data from the file on disk 1048 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset); 1049 if (data_sp->GetByteSize() != header_and_lc_size) 1050 continue; 1051 } 1052 if (data_sp) 1053 m_data.SetData(data_sp); 1054 } 1055 } 1056 return true; 1057 } 1058 // None found. 1059 return false; 1060 } else { 1061 memset(&m_header, 0, sizeof(struct mach_header)); 1062 } 1063 return false; 1064 } 1065 1066 ByteOrder ObjectFileMachO::GetByteOrder() const { 1067 return m_data.GetByteOrder(); 1068 } 1069 1070 bool ObjectFileMachO::IsExecutable() const { 1071 return m_header.filetype == MH_EXECUTE; 1072 } 1073 1074 bool ObjectFileMachO::IsDynamicLoader() const { 1075 return m_header.filetype == MH_DYLINKER; 1076 } 1077 1078 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1079 return m_data.GetAddressByteSize(); 1080 } 1081 1082 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1083 Symtab *symtab = GetSymtab(); 1084 if (!symtab) 1085 return AddressClass::eUnknown; 1086 1087 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1088 if (symbol) { 1089 if (symbol->ValueIsAddress()) { 1090 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1091 if (section_sp) { 1092 const lldb::SectionType section_type = section_sp->GetType(); 1093 switch (section_type) { 1094 case eSectionTypeInvalid: 1095 return AddressClass::eUnknown; 1096 1097 case eSectionTypeCode: 1098 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1099 // For ARM we have a bit in the n_desc field of the symbol that 1100 // tells us ARM/Thumb which is bit 0x0008. 1101 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1102 return AddressClass::eCodeAlternateISA; 1103 } 1104 return AddressClass::eCode; 1105 1106 case eSectionTypeContainer: 1107 return AddressClass::eUnknown; 1108 1109 case eSectionTypeData: 1110 case eSectionTypeDataCString: 1111 case eSectionTypeDataCStringPointers: 1112 case eSectionTypeDataSymbolAddress: 1113 case eSectionTypeData4: 1114 case eSectionTypeData8: 1115 case eSectionTypeData16: 1116 case eSectionTypeDataPointers: 1117 case eSectionTypeZeroFill: 1118 case eSectionTypeDataObjCMessageRefs: 1119 case eSectionTypeDataObjCCFStrings: 1120 case eSectionTypeGoSymtab: 1121 return AddressClass::eData; 1122 1123 case eSectionTypeDebug: 1124 case eSectionTypeDWARFDebugAbbrev: 1125 case eSectionTypeDWARFDebugAbbrevDwo: 1126 case eSectionTypeDWARFDebugAddr: 1127 case eSectionTypeDWARFDebugAranges: 1128 case eSectionTypeDWARFDebugCuIndex: 1129 case eSectionTypeDWARFDebugFrame: 1130 case eSectionTypeDWARFDebugInfo: 1131 case eSectionTypeDWARFDebugInfoDwo: 1132 case eSectionTypeDWARFDebugLine: 1133 case eSectionTypeDWARFDebugLineStr: 1134 case eSectionTypeDWARFDebugLoc: 1135 case eSectionTypeDWARFDebugLocLists: 1136 case eSectionTypeDWARFDebugMacInfo: 1137 case eSectionTypeDWARFDebugMacro: 1138 case eSectionTypeDWARFDebugNames: 1139 case eSectionTypeDWARFDebugPubNames: 1140 case eSectionTypeDWARFDebugPubTypes: 1141 case eSectionTypeDWARFDebugRanges: 1142 case eSectionTypeDWARFDebugRngLists: 1143 case eSectionTypeDWARFDebugStr: 1144 case eSectionTypeDWARFDebugStrDwo: 1145 case eSectionTypeDWARFDebugStrOffsets: 1146 case eSectionTypeDWARFDebugStrOffsetsDwo: 1147 case eSectionTypeDWARFDebugTypes: 1148 case eSectionTypeDWARFDebugTypesDwo: 1149 case eSectionTypeDWARFAppleNames: 1150 case eSectionTypeDWARFAppleTypes: 1151 case eSectionTypeDWARFAppleNamespaces: 1152 case eSectionTypeDWARFAppleObjC: 1153 case eSectionTypeDWARFGNUDebugAltLink: 1154 return AddressClass::eDebug; 1155 1156 case eSectionTypeEHFrame: 1157 case eSectionTypeARMexidx: 1158 case eSectionTypeARMextab: 1159 case eSectionTypeCompactUnwind: 1160 return AddressClass::eRuntime; 1161 1162 case eSectionTypeAbsoluteAddress: 1163 case eSectionTypeELFSymbolTable: 1164 case eSectionTypeELFDynamicSymbols: 1165 case eSectionTypeELFRelocationEntries: 1166 case eSectionTypeELFDynamicLinkInfo: 1167 case eSectionTypeOther: 1168 return AddressClass::eUnknown; 1169 } 1170 } 1171 } 1172 1173 const SymbolType symbol_type = symbol->GetType(); 1174 switch (symbol_type) { 1175 case eSymbolTypeAny: 1176 return AddressClass::eUnknown; 1177 case eSymbolTypeAbsolute: 1178 return AddressClass::eUnknown; 1179 1180 case eSymbolTypeCode: 1181 case eSymbolTypeTrampoline: 1182 case eSymbolTypeResolver: 1183 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1184 // For ARM we have a bit in the n_desc field of the symbol that tells 1185 // us ARM/Thumb which is bit 0x0008. 1186 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1187 return AddressClass::eCodeAlternateISA; 1188 } 1189 return AddressClass::eCode; 1190 1191 case eSymbolTypeData: 1192 return AddressClass::eData; 1193 case eSymbolTypeRuntime: 1194 return AddressClass::eRuntime; 1195 case eSymbolTypeException: 1196 return AddressClass::eRuntime; 1197 case eSymbolTypeSourceFile: 1198 return AddressClass::eDebug; 1199 case eSymbolTypeHeaderFile: 1200 return AddressClass::eDebug; 1201 case eSymbolTypeObjectFile: 1202 return AddressClass::eDebug; 1203 case eSymbolTypeCommonBlock: 1204 return AddressClass::eDebug; 1205 case eSymbolTypeBlock: 1206 return AddressClass::eDebug; 1207 case eSymbolTypeLocal: 1208 return AddressClass::eData; 1209 case eSymbolTypeParam: 1210 return AddressClass::eData; 1211 case eSymbolTypeVariable: 1212 return AddressClass::eData; 1213 case eSymbolTypeVariableType: 1214 return AddressClass::eDebug; 1215 case eSymbolTypeLineEntry: 1216 return AddressClass::eDebug; 1217 case eSymbolTypeLineHeader: 1218 return AddressClass::eDebug; 1219 case eSymbolTypeScopeBegin: 1220 return AddressClass::eDebug; 1221 case eSymbolTypeScopeEnd: 1222 return AddressClass::eDebug; 1223 case eSymbolTypeAdditional: 1224 return AddressClass::eUnknown; 1225 case eSymbolTypeCompiler: 1226 return AddressClass::eDebug; 1227 case eSymbolTypeInstrumentation: 1228 return AddressClass::eDebug; 1229 case eSymbolTypeUndefined: 1230 return AddressClass::eUnknown; 1231 case eSymbolTypeObjCClass: 1232 return AddressClass::eRuntime; 1233 case eSymbolTypeObjCMetaClass: 1234 return AddressClass::eRuntime; 1235 case eSymbolTypeObjCIVar: 1236 return AddressClass::eRuntime; 1237 case eSymbolTypeReExported: 1238 return AddressClass::eRuntime; 1239 } 1240 } 1241 return AddressClass::eUnknown; 1242 } 1243 1244 Symtab *ObjectFileMachO::GetSymtab() { 1245 ModuleSP module_sp(GetModule()); 1246 if (module_sp) { 1247 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1248 if (m_symtab_up == nullptr) { 1249 m_symtab_up.reset(new Symtab(this)); 1250 std::lock_guard<std::recursive_mutex> symtab_guard( 1251 m_symtab_up->GetMutex()); 1252 ParseSymtab(); 1253 m_symtab_up->Finalize(); 1254 } 1255 } 1256 return m_symtab_up.get(); 1257 } 1258 1259 bool ObjectFileMachO::IsStripped() { 1260 if (m_dysymtab.cmd == 0) { 1261 ModuleSP module_sp(GetModule()); 1262 if (module_sp) { 1263 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1264 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1265 const lldb::offset_t load_cmd_offset = offset; 1266 1267 load_command lc; 1268 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 1269 break; 1270 if (lc.cmd == LC_DYSYMTAB) { 1271 m_dysymtab.cmd = lc.cmd; 1272 m_dysymtab.cmdsize = lc.cmdsize; 1273 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1274 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1275 nullptr) { 1276 // Clear m_dysymtab if we were unable to read all items from the 1277 // load command 1278 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1279 } 1280 } 1281 offset = load_cmd_offset + lc.cmdsize; 1282 } 1283 } 1284 } 1285 if (m_dysymtab.cmd) 1286 return m_dysymtab.nlocalsym <= 1; 1287 return false; 1288 } 1289 1290 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1291 EncryptedFileRanges result; 1292 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1293 1294 encryption_info_command encryption_cmd; 1295 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1296 const lldb::offset_t load_cmd_offset = offset; 1297 if (m_data.GetU32(&offset, &encryption_cmd, 2) == nullptr) 1298 break; 1299 1300 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the 1301 // 3 fields we care about, so treat them the same. 1302 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1303 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1304 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1305 if (encryption_cmd.cryptid != 0) { 1306 EncryptedFileRanges::Entry entry; 1307 entry.SetRangeBase(encryption_cmd.cryptoff); 1308 entry.SetByteSize(encryption_cmd.cryptsize); 1309 result.Append(entry); 1310 } 1311 } 1312 } 1313 offset = load_cmd_offset + encryption_cmd.cmdsize; 1314 } 1315 1316 return result; 1317 } 1318 1319 void ObjectFileMachO::SanitizeSegmentCommand(segment_command_64 &seg_cmd, 1320 uint32_t cmd_idx) { 1321 if (m_length == 0 || seg_cmd.filesize == 0) 1322 return; 1323 1324 if (seg_cmd.fileoff > m_length) { 1325 // We have a load command that says it extends past the end of the file. 1326 // This is likely a corrupt file. We don't have any way to return an error 1327 // condition here (this method was likely invoked from something like 1328 // ObjectFile::GetSectionList()), so we just null out the section contents, 1329 // and dump a message to stdout. The most common case here is core file 1330 // debugging with a truncated file. 1331 const char *lc_segment_name = 1332 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1333 GetModule()->ReportWarning( 1334 "load command %u %s has a fileoff (0x%" PRIx64 1335 ") that extends beyond the end of the file (0x%" PRIx64 1336 "), ignoring this section", 1337 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1338 1339 seg_cmd.fileoff = 0; 1340 seg_cmd.filesize = 0; 1341 } 1342 1343 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1344 // We have a load command that says it extends past the end of the file. 1345 // This is likely a corrupt file. We don't have any way to return an error 1346 // condition here (this method was likely invoked from something like 1347 // ObjectFile::GetSectionList()), so we just null out the section contents, 1348 // and dump a message to stdout. The most common case here is core file 1349 // debugging with a truncated file. 1350 const char *lc_segment_name = 1351 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1352 GetModule()->ReportWarning( 1353 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1354 ") that extends beyond the end of the file (0x%" PRIx64 1355 "), the segment will be truncated to match", 1356 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1357 1358 // Truncate the length 1359 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1360 } 1361 } 1362 1363 static uint32_t GetSegmentPermissions(const segment_command_64 &seg_cmd) { 1364 uint32_t result = 0; 1365 if (seg_cmd.initprot & VM_PROT_READ) 1366 result |= ePermissionsReadable; 1367 if (seg_cmd.initprot & VM_PROT_WRITE) 1368 result |= ePermissionsWritable; 1369 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1370 result |= ePermissionsExecutable; 1371 return result; 1372 } 1373 1374 static lldb::SectionType GetSectionType(uint32_t flags, 1375 ConstString section_name) { 1376 1377 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1378 return eSectionTypeCode; 1379 1380 uint32_t mach_sect_type = flags & SECTION_TYPE; 1381 static ConstString g_sect_name_objc_data("__objc_data"); 1382 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1383 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1384 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1385 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1386 static ConstString g_sect_name_objc_const("__objc_const"); 1387 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1388 static ConstString g_sect_name_cfstring("__cfstring"); 1389 1390 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1391 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1392 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1393 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1394 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1395 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1396 static ConstString g_sect_name_dwarf_debug_loclists("__debug_loclists"); 1397 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1398 static ConstString g_sect_name_dwarf_debug_names("__debug_names"); 1399 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1400 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1401 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1402 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1403 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1404 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1405 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1406 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1407 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1408 static ConstString g_sect_name_eh_frame("__eh_frame"); 1409 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1410 static ConstString g_sect_name_text("__text"); 1411 static ConstString g_sect_name_data("__data"); 1412 static ConstString g_sect_name_go_symtab("__gosymtab"); 1413 1414 if (section_name == g_sect_name_dwarf_debug_abbrev) 1415 return eSectionTypeDWARFDebugAbbrev; 1416 if (section_name == g_sect_name_dwarf_debug_aranges) 1417 return eSectionTypeDWARFDebugAranges; 1418 if (section_name == g_sect_name_dwarf_debug_frame) 1419 return eSectionTypeDWARFDebugFrame; 1420 if (section_name == g_sect_name_dwarf_debug_info) 1421 return eSectionTypeDWARFDebugInfo; 1422 if (section_name == g_sect_name_dwarf_debug_line) 1423 return eSectionTypeDWARFDebugLine; 1424 if (section_name == g_sect_name_dwarf_debug_loc) 1425 return eSectionTypeDWARFDebugLoc; 1426 if (section_name == g_sect_name_dwarf_debug_loclists) 1427 return eSectionTypeDWARFDebugLocLists; 1428 if (section_name == g_sect_name_dwarf_debug_macinfo) 1429 return eSectionTypeDWARFDebugMacInfo; 1430 if (section_name == g_sect_name_dwarf_debug_names) 1431 return eSectionTypeDWARFDebugNames; 1432 if (section_name == g_sect_name_dwarf_debug_pubnames) 1433 return eSectionTypeDWARFDebugPubNames; 1434 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1435 return eSectionTypeDWARFDebugPubTypes; 1436 if (section_name == g_sect_name_dwarf_debug_ranges) 1437 return eSectionTypeDWARFDebugRanges; 1438 if (section_name == g_sect_name_dwarf_debug_str) 1439 return eSectionTypeDWARFDebugStr; 1440 if (section_name == g_sect_name_dwarf_debug_types) 1441 return eSectionTypeDWARFDebugTypes; 1442 if (section_name == g_sect_name_dwarf_apple_names) 1443 return eSectionTypeDWARFAppleNames; 1444 if (section_name == g_sect_name_dwarf_apple_types) 1445 return eSectionTypeDWARFAppleTypes; 1446 if (section_name == g_sect_name_dwarf_apple_namespaces) 1447 return eSectionTypeDWARFAppleNamespaces; 1448 if (section_name == g_sect_name_dwarf_apple_objc) 1449 return eSectionTypeDWARFAppleObjC; 1450 if (section_name == g_sect_name_objc_selrefs) 1451 return eSectionTypeDataCStringPointers; 1452 if (section_name == g_sect_name_objc_msgrefs) 1453 return eSectionTypeDataObjCMessageRefs; 1454 if (section_name == g_sect_name_eh_frame) 1455 return eSectionTypeEHFrame; 1456 if (section_name == g_sect_name_compact_unwind) 1457 return eSectionTypeCompactUnwind; 1458 if (section_name == g_sect_name_cfstring) 1459 return eSectionTypeDataObjCCFStrings; 1460 if (section_name == g_sect_name_go_symtab) 1461 return eSectionTypeGoSymtab; 1462 if (section_name == g_sect_name_objc_data || 1463 section_name == g_sect_name_objc_classrefs || 1464 section_name == g_sect_name_objc_superrefs || 1465 section_name == g_sect_name_objc_const || 1466 section_name == g_sect_name_objc_classlist) { 1467 return eSectionTypeDataPointers; 1468 } 1469 1470 switch (mach_sect_type) { 1471 // TODO: categorize sections by other flags for regular sections 1472 case S_REGULAR: 1473 if (section_name == g_sect_name_text) 1474 return eSectionTypeCode; 1475 if (section_name == g_sect_name_data) 1476 return eSectionTypeData; 1477 return eSectionTypeOther; 1478 case S_ZEROFILL: 1479 return eSectionTypeZeroFill; 1480 case S_CSTRING_LITERALS: // section with only literal C strings 1481 return eSectionTypeDataCString; 1482 case S_4BYTE_LITERALS: // section with only 4 byte literals 1483 return eSectionTypeData4; 1484 case S_8BYTE_LITERALS: // section with only 8 byte literals 1485 return eSectionTypeData8; 1486 case S_LITERAL_POINTERS: // section with only pointers to literals 1487 return eSectionTypeDataPointers; 1488 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1489 return eSectionTypeDataPointers; 1490 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1491 return eSectionTypeDataPointers; 1492 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1493 // the reserved2 field 1494 return eSectionTypeCode; 1495 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1496 // initialization 1497 return eSectionTypeDataPointers; 1498 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1499 // termination 1500 return eSectionTypeDataPointers; 1501 case S_COALESCED: 1502 return eSectionTypeOther; 1503 case S_GB_ZEROFILL: 1504 return eSectionTypeZeroFill; 1505 case S_INTERPOSING: // section with only pairs of function pointers for 1506 // interposing 1507 return eSectionTypeCode; 1508 case S_16BYTE_LITERALS: // section with only 16 byte literals 1509 return eSectionTypeData16; 1510 case S_DTRACE_DOF: 1511 return eSectionTypeDebug; 1512 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1513 return eSectionTypeDataPointers; 1514 default: 1515 return eSectionTypeOther; 1516 } 1517 } 1518 1519 struct ObjectFileMachO::SegmentParsingContext { 1520 const EncryptedFileRanges EncryptedRanges; 1521 lldb_private::SectionList &UnifiedList; 1522 uint32_t NextSegmentIdx = 0; 1523 uint32_t NextSectionIdx = 0; 1524 bool FileAddressesChanged = false; 1525 1526 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1527 lldb_private::SectionList &UnifiedList) 1528 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1529 }; 1530 1531 void ObjectFileMachO::ProcessSegmentCommand(const load_command &load_cmd_, 1532 lldb::offset_t offset, 1533 uint32_t cmd_idx, 1534 SegmentParsingContext &context) { 1535 segment_command_64 load_cmd; 1536 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1537 1538 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1539 return; 1540 1541 ModuleSP module_sp = GetModule(); 1542 const bool is_core = GetType() == eTypeCoreFile; 1543 const bool is_dsym = (m_header.filetype == MH_DSYM); 1544 bool add_section = true; 1545 bool add_to_unified = true; 1546 ConstString const_segname( 1547 load_cmd.segname, strnlen(load_cmd.segname, sizeof(load_cmd.segname))); 1548 1549 SectionSP unified_section_sp( 1550 context.UnifiedList.FindSectionByName(const_segname)); 1551 if (is_dsym && unified_section_sp) { 1552 if (const_segname == GetSegmentNameLINKEDIT()) { 1553 // We need to keep the __LINKEDIT segment private to this object file 1554 // only 1555 add_to_unified = false; 1556 } else { 1557 // This is the dSYM file and this section has already been created by the 1558 // object file, no need to create it. 1559 add_section = false; 1560 } 1561 } 1562 load_cmd.vmaddr = m_data.GetAddress(&offset); 1563 load_cmd.vmsize = m_data.GetAddress(&offset); 1564 load_cmd.fileoff = m_data.GetAddress(&offset); 1565 load_cmd.filesize = m_data.GetAddress(&offset); 1566 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1567 return; 1568 1569 SanitizeSegmentCommand(load_cmd, cmd_idx); 1570 1571 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1572 const bool segment_is_encrypted = 1573 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1574 1575 // Keep a list of mach segments around in case we need to get at data that 1576 // isn't stored in the abstracted Sections. 1577 m_mach_segments.push_back(load_cmd); 1578 1579 // Use a segment ID of the segment index shifted left by 8 so they never 1580 // conflict with any of the sections. 1581 SectionSP segment_sp; 1582 if (add_section && (const_segname || is_core)) { 1583 segment_sp = std::make_shared<Section>( 1584 module_sp, // Module to which this section belongs 1585 this, // Object file to which this sections belongs 1586 ++context.NextSegmentIdx 1587 << 8, // Section ID is the 1 based segment index 1588 // shifted right by 8 bits as not to collide with any of the 256 1589 // section IDs that are possible 1590 const_segname, // Name of this section 1591 eSectionTypeContainer, // This section is a container of other 1592 // sections. 1593 load_cmd.vmaddr, // File VM address == addresses as they are 1594 // found in the object file 1595 load_cmd.vmsize, // VM size in bytes of this section 1596 load_cmd.fileoff, // Offset to the data for this section in 1597 // the file 1598 load_cmd.filesize, // Size in bytes of this section as found 1599 // in the file 1600 0, // Segments have no alignment information 1601 load_cmd.flags); // Flags for this section 1602 1603 segment_sp->SetIsEncrypted(segment_is_encrypted); 1604 m_sections_up->AddSection(segment_sp); 1605 segment_sp->SetPermissions(segment_permissions); 1606 if (add_to_unified) 1607 context.UnifiedList.AddSection(segment_sp); 1608 } else if (unified_section_sp) { 1609 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1610 // Check to see if the module was read from memory? 1611 if (module_sp->GetObjectFile()->GetBaseAddress().IsValid()) { 1612 // We have a module that is in memory and needs to have its file 1613 // address adjusted. We need to do this because when we load a file 1614 // from memory, its addresses will be slid already, yet the addresses 1615 // in the new symbol file will still be unslid. Since everything is 1616 // stored as section offset, this shouldn't cause any problems. 1617 1618 // Make sure we've parsed the symbol table from the ObjectFile before 1619 // we go around changing its Sections. 1620 module_sp->GetObjectFile()->GetSymtab(); 1621 // eh_frame would present the same problems but we parse that on a per- 1622 // function basis as-needed so it's more difficult to remove its use of 1623 // the Sections. Realistically, the environments where this code path 1624 // will be taken will not have eh_frame sections. 1625 1626 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1627 1628 // Notify the module that the section addresses have been changed once 1629 // we're done so any file-address caches can be updated. 1630 context.FileAddressesChanged = true; 1631 } 1632 } 1633 m_sections_up->AddSection(unified_section_sp); 1634 } 1635 1636 struct section_64 sect64; 1637 ::memset(§64, 0, sizeof(sect64)); 1638 // Push a section into our mach sections for the section at index zero 1639 // (NO_SECT) if we don't have any mach sections yet... 1640 if (m_mach_sections.empty()) 1641 m_mach_sections.push_back(sect64); 1642 uint32_t segment_sect_idx; 1643 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1644 1645 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1646 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1647 ++segment_sect_idx) { 1648 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1649 sizeof(sect64.sectname)) == nullptr) 1650 break; 1651 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1652 sizeof(sect64.segname)) == nullptr) 1653 break; 1654 sect64.addr = m_data.GetAddress(&offset); 1655 sect64.size = m_data.GetAddress(&offset); 1656 1657 if (m_data.GetU32(&offset, §64.offset, num_u32s) == nullptr) 1658 break; 1659 1660 // Keep a list of mach sections around in case we need to get at data that 1661 // isn't stored in the abstracted Sections. 1662 m_mach_sections.push_back(sect64); 1663 1664 if (add_section) { 1665 ConstString section_name( 1666 sect64.sectname, strnlen(sect64.sectname, sizeof(sect64.sectname))); 1667 if (!const_segname) { 1668 // We have a segment with no name so we need to conjure up segments 1669 // that correspond to the section's segname if there isn't already such 1670 // a section. If there is such a section, we resize the section so that 1671 // it spans all sections. We also mark these sections as fake so 1672 // address matches don't hit if they land in the gaps between the child 1673 // sections. 1674 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1675 sizeof(sect64.segname)); 1676 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1677 if (segment_sp.get()) { 1678 Section *segment = segment_sp.get(); 1679 // Grow the section size as needed. 1680 const lldb::addr_t sect64_min_addr = sect64.addr; 1681 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1682 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1683 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1684 const lldb::addr_t curr_seg_max_addr = 1685 curr_seg_min_addr + curr_seg_byte_size; 1686 if (sect64_min_addr >= curr_seg_min_addr) { 1687 const lldb::addr_t new_seg_byte_size = 1688 sect64_max_addr - curr_seg_min_addr; 1689 // Only grow the section size if needed 1690 if (new_seg_byte_size > curr_seg_byte_size) 1691 segment->SetByteSize(new_seg_byte_size); 1692 } else { 1693 // We need to change the base address of the segment and adjust the 1694 // child section offsets for all existing children. 1695 const lldb::addr_t slide_amount = 1696 sect64_min_addr - curr_seg_min_addr; 1697 segment->Slide(slide_amount, false); 1698 segment->GetChildren().Slide(-slide_amount, false); 1699 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1700 } 1701 1702 // Grow the section size as needed. 1703 if (sect64.offset) { 1704 const lldb::addr_t segment_min_file_offset = 1705 segment->GetFileOffset(); 1706 const lldb::addr_t segment_max_file_offset = 1707 segment_min_file_offset + segment->GetFileSize(); 1708 1709 const lldb::addr_t section_min_file_offset = sect64.offset; 1710 const lldb::addr_t section_max_file_offset = 1711 section_min_file_offset + sect64.size; 1712 const lldb::addr_t new_file_offset = 1713 std::min(section_min_file_offset, segment_min_file_offset); 1714 const lldb::addr_t new_file_size = 1715 std::max(section_max_file_offset, segment_max_file_offset) - 1716 new_file_offset; 1717 segment->SetFileOffset(new_file_offset); 1718 segment->SetFileSize(new_file_size); 1719 } 1720 } else { 1721 // Create a fake section for the section's named segment 1722 segment_sp = std::make_shared<Section>( 1723 segment_sp, // Parent section 1724 module_sp, // Module to which this section belongs 1725 this, // Object file to which this section belongs 1726 ++context.NextSegmentIdx 1727 << 8, // Section ID is the 1 based segment index 1728 // shifted right by 8 bits as not to 1729 // collide with any of the 256 section IDs 1730 // that are possible 1731 const_segname, // Name of this section 1732 eSectionTypeContainer, // This section is a container of 1733 // other sections. 1734 sect64.addr, // File VM address == addresses as they are 1735 // found in the object file 1736 sect64.size, // VM size in bytes of this section 1737 sect64.offset, // Offset to the data for this section in 1738 // the file 1739 sect64.offset ? sect64.size : 0, // Size in bytes of 1740 // this section as 1741 // found in the file 1742 sect64.align, 1743 load_cmd.flags); // Flags for this section 1744 segment_sp->SetIsFake(true); 1745 segment_sp->SetPermissions(segment_permissions); 1746 m_sections_up->AddSection(segment_sp); 1747 if (add_to_unified) 1748 context.UnifiedList.AddSection(segment_sp); 1749 segment_sp->SetIsEncrypted(segment_is_encrypted); 1750 } 1751 } 1752 assert(segment_sp.get()); 1753 1754 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1755 1756 SectionSP section_sp(new Section( 1757 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1758 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1759 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1760 sect64.flags)); 1761 // Set the section to be encrypted to match the segment 1762 1763 bool section_is_encrypted = false; 1764 if (!segment_is_encrypted && load_cmd.filesize != 0) 1765 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1766 sect64.offset) != nullptr; 1767 1768 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1769 section_sp->SetPermissions(segment_permissions); 1770 segment_sp->GetChildren().AddSection(section_sp); 1771 1772 if (segment_sp->IsFake()) { 1773 segment_sp.reset(); 1774 const_segname.Clear(); 1775 } 1776 } 1777 } 1778 if (segment_sp && is_dsym) { 1779 if (first_segment_sectID <= context.NextSectionIdx) { 1780 lldb::user_id_t sect_uid; 1781 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1782 ++sect_uid) { 1783 SectionSP curr_section_sp( 1784 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1785 SectionSP next_section_sp; 1786 if (sect_uid + 1 <= context.NextSectionIdx) 1787 next_section_sp = 1788 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1789 1790 if (curr_section_sp.get()) { 1791 if (curr_section_sp->GetByteSize() == 0) { 1792 if (next_section_sp.get() != nullptr) 1793 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1794 curr_section_sp->GetFileAddress()); 1795 else 1796 curr_section_sp->SetByteSize(load_cmd.vmsize); 1797 } 1798 } 1799 } 1800 } 1801 } 1802 } 1803 1804 void ObjectFileMachO::ProcessDysymtabCommand(const load_command &load_cmd, 1805 lldb::offset_t offset) { 1806 m_dysymtab.cmd = load_cmd.cmd; 1807 m_dysymtab.cmdsize = load_cmd.cmdsize; 1808 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1809 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1810 } 1811 1812 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1813 if (m_sections_up) 1814 return; 1815 1816 m_sections_up.reset(new SectionList()); 1817 1818 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1819 // bool dump_sections = false; 1820 ModuleSP module_sp(GetModule()); 1821 1822 offset = MachHeaderSizeFromMagic(m_header.magic); 1823 1824 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1825 struct load_command load_cmd; 1826 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1827 const lldb::offset_t load_cmd_offset = offset; 1828 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 1829 break; 1830 1831 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1832 ProcessSegmentCommand(load_cmd, offset, i, context); 1833 else if (load_cmd.cmd == LC_DYSYMTAB) 1834 ProcessDysymtabCommand(load_cmd, offset); 1835 1836 offset = load_cmd_offset + load_cmd.cmdsize; 1837 } 1838 1839 if (context.FileAddressesChanged && module_sp) 1840 module_sp->SectionFileAddressesChanged(); 1841 } 1842 1843 class MachSymtabSectionInfo { 1844 public: 1845 MachSymtabSectionInfo(SectionList *section_list) 1846 : m_section_list(section_list), m_section_infos() { 1847 // Get the number of sections down to a depth of 1 to include all segments 1848 // and their sections, but no other sections that may be added for debug 1849 // map or 1850 m_section_infos.resize(section_list->GetNumSections(1)); 1851 } 1852 1853 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1854 if (n_sect == 0) 1855 return SectionSP(); 1856 if (n_sect < m_section_infos.size()) { 1857 if (!m_section_infos[n_sect].section_sp) { 1858 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1859 m_section_infos[n_sect].section_sp = section_sp; 1860 if (section_sp) { 1861 m_section_infos[n_sect].vm_range.SetBaseAddress( 1862 section_sp->GetFileAddress()); 1863 m_section_infos[n_sect].vm_range.SetByteSize( 1864 section_sp->GetByteSize()); 1865 } else { 1866 Host::SystemLog(Host::eSystemLogError, 1867 "error: unable to find section for section %u\n", 1868 n_sect); 1869 } 1870 } 1871 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1872 // Symbol is in section. 1873 return m_section_infos[n_sect].section_sp; 1874 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1875 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1876 file_addr) { 1877 // Symbol is in section with zero size, but has the same start address 1878 // as the section. This can happen with linker symbols (symbols that 1879 // start with the letter 'l' or 'L'. 1880 return m_section_infos[n_sect].section_sp; 1881 } 1882 } 1883 return m_section_list->FindSectionContainingFileAddress(file_addr); 1884 } 1885 1886 protected: 1887 struct SectionInfo { 1888 SectionInfo() : vm_range(), section_sp() {} 1889 1890 VMRange vm_range; 1891 SectionSP section_sp; 1892 }; 1893 SectionList *m_section_list; 1894 std::vector<SectionInfo> m_section_infos; 1895 }; 1896 1897 struct TrieEntry { 1898 void Dump() const { 1899 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1900 static_cast<unsigned long long>(address), 1901 static_cast<unsigned long long>(flags), 1902 static_cast<unsigned long long>(other), name.GetCString()); 1903 if (import_name) 1904 printf(" -> \"%s\"\n", import_name.GetCString()); 1905 else 1906 printf("\n"); 1907 } 1908 ConstString name; 1909 uint64_t address = LLDB_INVALID_ADDRESS; 1910 uint64_t flags = 0; 1911 uint64_t other = 0; 1912 ConstString import_name; 1913 }; 1914 1915 struct TrieEntryWithOffset { 1916 lldb::offset_t nodeOffset; 1917 TrieEntry entry; 1918 1919 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1920 1921 void Dump(uint32_t idx) const { 1922 printf("[%3u] 0x%16.16llx: ", idx, 1923 static_cast<unsigned long long>(nodeOffset)); 1924 entry.Dump(); 1925 } 1926 1927 bool operator<(const TrieEntryWithOffset &other) const { 1928 return (nodeOffset < other.nodeOffset); 1929 } 1930 }; 1931 1932 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 1933 const bool is_arm, 1934 std::vector<llvm::StringRef> &nameSlices, 1935 std::set<lldb::addr_t> &resolver_addresses, 1936 std::vector<TrieEntryWithOffset> &output) { 1937 if (!data.ValidOffset(offset)) 1938 return true; 1939 1940 const uint64_t terminalSize = data.GetULEB128(&offset); 1941 lldb::offset_t children_offset = offset + terminalSize; 1942 if (terminalSize != 0) { 1943 TrieEntryWithOffset e(offset); 1944 e.entry.flags = data.GetULEB128(&offset); 1945 const char *import_name = nullptr; 1946 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 1947 e.entry.address = 0; 1948 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 1949 import_name = data.GetCStr(&offset); 1950 } else { 1951 e.entry.address = data.GetULEB128(&offset); 1952 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 1953 e.entry.other = data.GetULEB128(&offset); 1954 uint64_t resolver_addr = e.entry.other; 1955 if (is_arm) 1956 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 1957 resolver_addresses.insert(resolver_addr); 1958 } else 1959 e.entry.other = 0; 1960 } 1961 // Only add symbols that are reexport symbols with a valid import name 1962 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && 1963 import_name[0]) { 1964 std::string name; 1965 if (!nameSlices.empty()) { 1966 for (auto name_slice : nameSlices) 1967 name.append(name_slice.data(), name_slice.size()); 1968 } 1969 if (name.size() > 1) { 1970 // Skip the leading '_' 1971 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 1972 } 1973 if (import_name) { 1974 // Skip the leading '_' 1975 e.entry.import_name.SetCString(import_name + 1); 1976 } 1977 output.push_back(e); 1978 } 1979 } 1980 1981 const uint8_t childrenCount = data.GetU8(&children_offset); 1982 for (uint8_t i = 0; i < childrenCount; ++i) { 1983 const char *cstr = data.GetCStr(&children_offset); 1984 if (cstr) 1985 nameSlices.push_back(llvm::StringRef(cstr)); 1986 else 1987 return false; // Corrupt data 1988 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 1989 if (childNodeOffset) { 1990 if (!ParseTrieEntries(data, childNodeOffset, is_arm, nameSlices, 1991 resolver_addresses, output)) { 1992 return false; 1993 } 1994 } 1995 nameSlices.pop_back(); 1996 } 1997 return true; 1998 } 1999 2000 // Read the UUID out of a dyld_shared_cache file on-disk. 2001 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2002 const ByteOrder byte_order, 2003 const uint32_t addr_byte_size) { 2004 UUID dsc_uuid; 2005 DataBufferSP DscData = MapFileData( 2006 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2007 if (!DscData) 2008 return dsc_uuid; 2009 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2010 2011 char version_str[7]; 2012 lldb::offset_t offset = 0; 2013 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2014 version_str[6] = '\0'; 2015 if (strcmp(version_str, "dyld_v") == 0) { 2016 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2017 dsc_uuid = UUID::fromOptionalData( 2018 dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t)); 2019 } 2020 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2021 if (log && dsc_uuid.IsValid()) { 2022 LLDB_LOGF(log, "Shared cache %s has UUID %s", 2023 dyld_shared_cache.GetPath().c_str(), 2024 dsc_uuid.GetAsString().c_str()); 2025 } 2026 return dsc_uuid; 2027 } 2028 2029 static llvm::Optional<struct nlist_64> 2030 ParseNList(DataExtractor &nlist_data, lldb::offset_t &nlist_data_offset, 2031 size_t nlist_byte_size) { 2032 struct nlist_64 nlist; 2033 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2034 return {}; 2035 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 2036 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 2037 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 2038 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 2039 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 2040 return nlist; 2041 } 2042 2043 enum { DebugSymbols = true, NonDebugSymbols = false }; 2044 2045 size_t ObjectFileMachO::ParseSymtab() { 2046 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2047 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2048 m_file.GetFilename().AsCString("")); 2049 ModuleSP module_sp(GetModule()); 2050 if (!module_sp) 2051 return 0; 2052 2053 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2054 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2055 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2056 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2057 FunctionStarts function_starts; 2058 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2059 uint32_t i; 2060 FileSpecList dylib_files; 2061 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2062 llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2063 llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2064 llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2065 2066 for (i = 0; i < m_header.ncmds; ++i) { 2067 const lldb::offset_t cmd_offset = offset; 2068 // Read in the load command and load command size 2069 struct load_command lc; 2070 if (m_data.GetU32(&offset, &lc, 2) == nullptr) 2071 break; 2072 // Watch for the symbol table load command 2073 switch (lc.cmd) { 2074 case LC_SYMTAB: 2075 symtab_load_command.cmd = lc.cmd; 2076 symtab_load_command.cmdsize = lc.cmdsize; 2077 // Read in the rest of the symtab load command 2078 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2079 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2080 return 0; 2081 if (symtab_load_command.symoff == 0) { 2082 if (log) 2083 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2084 return 0; 2085 } 2086 2087 if (symtab_load_command.stroff == 0) { 2088 if (log) 2089 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2090 return 0; 2091 } 2092 2093 if (symtab_load_command.nsyms == 0) { 2094 if (log) 2095 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2096 return 0; 2097 } 2098 2099 if (symtab_load_command.strsize == 0) { 2100 if (log) 2101 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2102 return 0; 2103 } 2104 break; 2105 2106 case LC_DYLD_INFO: 2107 case LC_DYLD_INFO_ONLY: 2108 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2109 dyld_info.cmd = lc.cmd; 2110 dyld_info.cmdsize = lc.cmdsize; 2111 } else { 2112 memset(&dyld_info, 0, sizeof(dyld_info)); 2113 } 2114 break; 2115 2116 case LC_LOAD_DYLIB: 2117 case LC_LOAD_WEAK_DYLIB: 2118 case LC_REEXPORT_DYLIB: 2119 case LC_LOADFVMLIB: 2120 case LC_LOAD_UPWARD_DYLIB: { 2121 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2122 const char *path = m_data.PeekCStr(name_offset); 2123 if (path) { 2124 FileSpec file_spec(path); 2125 // Strip the path if there is @rpath, @executable, etc so we just use 2126 // the basename 2127 if (path[0] == '@') 2128 file_spec.GetDirectory().Clear(); 2129 2130 if (lc.cmd == LC_REEXPORT_DYLIB) { 2131 m_reexported_dylibs.AppendIfUnique(file_spec); 2132 } 2133 2134 dylib_files.Append(file_spec); 2135 } 2136 } break; 2137 2138 case LC_FUNCTION_STARTS: 2139 function_starts_load_command.cmd = lc.cmd; 2140 function_starts_load_command.cmdsize = lc.cmdsize; 2141 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2142 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2143 memset(&function_starts_load_command, 0, 2144 sizeof(function_starts_load_command)); 2145 break; 2146 2147 default: 2148 break; 2149 } 2150 offset = cmd_offset + lc.cmdsize; 2151 } 2152 2153 if (!symtab_load_command.cmd) 2154 return 0; 2155 2156 Symtab *symtab = m_symtab_up.get(); 2157 SectionList *section_list = GetSectionList(); 2158 if (section_list == nullptr) 2159 return 0; 2160 2161 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2162 const ByteOrder byte_order = m_data.GetByteOrder(); 2163 bool bit_width_32 = addr_byte_size == 4; 2164 const size_t nlist_byte_size = 2165 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2166 2167 DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size); 2168 DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size); 2169 DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size); 2170 DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order, 2171 addr_byte_size); 2172 DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size); 2173 2174 const addr_t nlist_data_byte_size = 2175 symtab_load_command.nsyms * nlist_byte_size; 2176 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2177 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2178 2179 ProcessSP process_sp(m_process_wp.lock()); 2180 Process *process = process_sp.get(); 2181 2182 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2183 2184 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2185 Target &target = process->GetTarget(); 2186 2187 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2188 2189 SectionSP linkedit_section_sp( 2190 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2191 // Reading mach file from memory in a process or core file... 2192 2193 if (linkedit_section_sp) { 2194 addr_t linkedit_load_addr = 2195 linkedit_section_sp->GetLoadBaseAddress(&target); 2196 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2197 // We might be trying to access the symbol table before the 2198 // __LINKEDIT's load address has been set in the target. We can't 2199 // fail to read the symbol table, so calculate the right address 2200 // manually 2201 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2202 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2203 } 2204 2205 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2206 const addr_t symoff_addr = linkedit_load_addr + 2207 symtab_load_command.symoff - 2208 linkedit_file_offset; 2209 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2210 linkedit_file_offset; 2211 2212 bool data_was_read = false; 2213 2214 #if defined(__APPLE__) && \ 2215 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2216 if (m_header.flags & 0x80000000u && 2217 process->GetAddressByteSize() == sizeof(void *)) { 2218 // This mach-o memory file is in the dyld shared cache. If this 2219 // program is not remote and this is iOS, then this process will 2220 // share the same shared cache as the process we are debugging and we 2221 // can read the entire __LINKEDIT from the address space in this 2222 // process. This is a needed optimization that is used for local iOS 2223 // debugging only since all shared libraries in the shared cache do 2224 // not have corresponding files that exist in the file system of the 2225 // device. They have been combined into a single file. This means we 2226 // always have to load these files from memory. All of the symbol and 2227 // string tables from all of the __LINKEDIT sections from the shared 2228 // libraries in the shared cache have been merged into a single large 2229 // symbol and string table. Reading all of this symbol and string 2230 // table data across can slow down debug launch times, so we optimize 2231 // this by reading the memory for the __LINKEDIT section from this 2232 // process. 2233 2234 UUID lldb_shared_cache; 2235 addr_t lldb_shared_cache_addr; 2236 GetLLDBSharedCacheUUID(lldb_shared_cache_addr, lldb_shared_cache); 2237 UUID process_shared_cache; 2238 addr_t process_shared_cache_addr; 2239 GetProcessSharedCacheUUID(process, process_shared_cache_addr, 2240 process_shared_cache); 2241 bool use_lldb_cache = true; 2242 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2243 (lldb_shared_cache != process_shared_cache || 2244 process_shared_cache_addr != lldb_shared_cache_addr)) { 2245 use_lldb_cache = false; 2246 } 2247 2248 PlatformSP platform_sp(target.GetPlatform()); 2249 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2250 data_was_read = true; 2251 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2252 eByteOrderLittle); 2253 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2254 eByteOrderLittle); 2255 if (function_starts_load_command.cmd) { 2256 const addr_t func_start_addr = 2257 linkedit_load_addr + function_starts_load_command.dataoff - 2258 linkedit_file_offset; 2259 function_starts_data.SetData((void *)func_start_addr, 2260 function_starts_load_command.datasize, 2261 eByteOrderLittle); 2262 } 2263 } 2264 } 2265 #endif 2266 2267 if (!data_was_read) { 2268 // Always load dyld - the dynamic linker - from memory if we didn't 2269 // find a binary anywhere else. lldb will not register 2270 // dylib/framework/bundle loads/unloads if we don't have the dyld 2271 // symbols, we force dyld to load from memory despite the user's 2272 // target.memory-module-load-level setting. 2273 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2274 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2275 DataBufferSP nlist_data_sp( 2276 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2277 if (nlist_data_sp) 2278 nlist_data.SetData(nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2279 if (m_dysymtab.nindirectsyms != 0) { 2280 const addr_t indirect_syms_addr = linkedit_load_addr + 2281 m_dysymtab.indirectsymoff - 2282 linkedit_file_offset; 2283 DataBufferSP indirect_syms_data_sp(ReadMemory( 2284 process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 2285 if (indirect_syms_data_sp) 2286 indirect_symbol_index_data.SetData( 2287 indirect_syms_data_sp, 0, 2288 indirect_syms_data_sp->GetByteSize()); 2289 // If this binary is outside the shared cache, 2290 // cache the string table. 2291 // Binaries in the shared cache all share a giant string table, 2292 // and we can't share the string tables across multiple 2293 // ObjectFileMachO's, so we'd end up re-reading this mega-strtab 2294 // for every binary in the shared cache - it would be a big perf 2295 // problem. For binaries outside the shared cache, it's faster to 2296 // read the entire strtab at once instead of piece-by-piece as we 2297 // process the nlist records. 2298 if ((m_header.flags & 0x80000000u) == 0) { 2299 DataBufferSP strtab_data_sp( 2300 ReadMemory(process_sp, strtab_addr, strtab_data_byte_size)); 2301 if (strtab_data_sp) { 2302 strtab_data.SetData(strtab_data_sp, 0, 2303 strtab_data_sp->GetByteSize()); 2304 } 2305 } 2306 } 2307 } 2308 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) { 2309 if (function_starts_load_command.cmd) { 2310 const addr_t func_start_addr = 2311 linkedit_load_addr + function_starts_load_command.dataoff - 2312 linkedit_file_offset; 2313 DataBufferSP func_start_data_sp( 2314 ReadMemory(process_sp, func_start_addr, 2315 function_starts_load_command.datasize)); 2316 if (func_start_data_sp) 2317 function_starts_data.SetData(func_start_data_sp, 0, 2318 func_start_data_sp->GetByteSize()); 2319 } 2320 } 2321 } 2322 } 2323 } else { 2324 nlist_data.SetData(m_data, symtab_load_command.symoff, 2325 nlist_data_byte_size); 2326 strtab_data.SetData(m_data, symtab_load_command.stroff, 2327 strtab_data_byte_size); 2328 2329 if (dyld_info.export_size > 0) { 2330 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2331 dyld_info.export_size); 2332 } 2333 2334 if (m_dysymtab.nindirectsyms != 0) { 2335 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2336 m_dysymtab.nindirectsyms * 4); 2337 } 2338 if (function_starts_load_command.cmd) { 2339 function_starts_data.SetData(m_data, function_starts_load_command.dataoff, 2340 function_starts_load_command.datasize); 2341 } 2342 } 2343 2344 if (nlist_data.GetByteSize() == 0 && 2345 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2346 if (log) 2347 module_sp->LogMessage(log, "failed to read nlist data"); 2348 return 0; 2349 } 2350 2351 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2352 if (!have_strtab_data) { 2353 if (process) { 2354 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2355 if (log) 2356 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2357 return 0; 2358 } 2359 } else { 2360 if (log) 2361 module_sp->LogMessage(log, "failed to read strtab data"); 2362 return 0; 2363 } 2364 } 2365 2366 ConstString g_segment_name_TEXT = GetSegmentNameTEXT(); 2367 ConstString g_segment_name_DATA = GetSegmentNameDATA(); 2368 ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2369 ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2370 ConstString g_segment_name_OBJC = GetSegmentNameOBJC(); 2371 ConstString g_section_name_eh_frame = GetSectionNameEHFrame(); 2372 SectionSP text_section_sp( 2373 section_list->FindSectionByName(g_segment_name_TEXT)); 2374 SectionSP data_section_sp( 2375 section_list->FindSectionByName(g_segment_name_DATA)); 2376 SectionSP data_dirty_section_sp( 2377 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2378 SectionSP data_const_section_sp( 2379 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2380 SectionSP objc_section_sp( 2381 section_list->FindSectionByName(g_segment_name_OBJC)); 2382 SectionSP eh_frame_section_sp; 2383 if (text_section_sp.get()) 2384 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2385 g_section_name_eh_frame); 2386 else 2387 eh_frame_section_sp = 2388 section_list->FindSectionByName(g_section_name_eh_frame); 2389 2390 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2391 2392 // lldb works best if it knows the start address of all functions in a 2393 // module. Linker symbols or debug info are normally the best source of 2394 // information for start addr / size but they may be stripped in a released 2395 // binary. Two additional sources of information exist in Mach-O binaries: 2396 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2397 // function's start address in the 2398 // binary, relative to the text section. 2399 // eh_frame - the eh_frame FDEs have the start addr & size of 2400 // each function 2401 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2402 // all modern binaries. 2403 // Binaries built to run on older releases may need to use eh_frame 2404 // information. 2405 2406 if (text_section_sp && function_starts_data.GetByteSize()) { 2407 FunctionStarts::Entry function_start_entry; 2408 function_start_entry.data = false; 2409 lldb::offset_t function_start_offset = 0; 2410 function_start_entry.addr = text_section_sp->GetFileAddress(); 2411 uint64_t delta; 2412 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2413 0) { 2414 // Now append the current entry 2415 function_start_entry.addr += delta; 2416 function_starts.Append(function_start_entry); 2417 } 2418 } else { 2419 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2420 // load command claiming an eh_frame but it doesn't actually have the 2421 // eh_frame content. And if we have a dSYM, we don't need to do any of 2422 // this fill-in-the-missing-symbols works anyway - the debug info should 2423 // give us all the functions in the module. 2424 if (text_section_sp.get() && eh_frame_section_sp.get() && 2425 m_type != eTypeDebugInfo) { 2426 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2427 DWARFCallFrameInfo::EH); 2428 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2429 eh_frame.GetFunctionAddressAndSizeVector(functions); 2430 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2431 size_t count = functions.GetSize(); 2432 for (size_t i = 0; i < count; ++i) { 2433 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2434 functions.GetEntryAtIndex(i); 2435 if (func) { 2436 FunctionStarts::Entry function_start_entry; 2437 function_start_entry.addr = func->base - text_base_addr; 2438 function_starts.Append(function_start_entry); 2439 } 2440 } 2441 } 2442 } 2443 2444 const size_t function_starts_count = function_starts.GetSize(); 2445 2446 // For user process binaries (executables, dylibs, frameworks, bundles), if 2447 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2448 // going to assume the binary has been stripped. Don't allow assembly 2449 // language instruction emulation because we don't know proper function 2450 // start boundaries. 2451 // 2452 // For all other types of binaries (kernels, stand-alone bare board 2453 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2454 // sections - we should not make any assumptions about them based on that. 2455 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2456 m_allow_assembly_emulation_unwind_plans = false; 2457 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2458 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2459 2460 if (unwind_or_symbol_log) 2461 module_sp->LogMessage( 2462 unwind_or_symbol_log, 2463 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2464 } 2465 2466 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() 2467 ? eh_frame_section_sp->GetID() 2468 : static_cast<user_id_t>(NO_SECT); 2469 2470 lldb::offset_t nlist_data_offset = 0; 2471 2472 uint32_t N_SO_index = UINT32_MAX; 2473 2474 MachSymtabSectionInfo section_info(section_list); 2475 std::vector<uint32_t> N_FUN_indexes; 2476 std::vector<uint32_t> N_NSYM_indexes; 2477 std::vector<uint32_t> N_INCL_indexes; 2478 std::vector<uint32_t> N_BRAC_indexes; 2479 std::vector<uint32_t> N_COMM_indexes; 2480 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2481 typedef llvm::DenseMap<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2482 typedef llvm::DenseMap<const char *, uint32_t> ConstNameToSymbolIndexMap; 2483 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2484 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2485 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2486 // Any symbols that get merged into another will get an entry in this map 2487 // so we know 2488 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2489 uint32_t nlist_idx = 0; 2490 Symbol *symbol_ptr = nullptr; 2491 2492 uint32_t sym_idx = 0; 2493 Symbol *sym = nullptr; 2494 size_t num_syms = 0; 2495 std::string memory_symbol_name; 2496 uint32_t unmapped_local_symbols_found = 0; 2497 2498 std::vector<TrieEntryWithOffset> trie_entries; 2499 std::set<lldb::addr_t> resolver_addresses; 2500 2501 if (dyld_trie_data.GetByteSize() > 0) { 2502 std::vector<llvm::StringRef> nameSlices; 2503 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, resolver_addresses, 2504 trie_entries); 2505 2506 ConstString text_segment_name("__TEXT"); 2507 SectionSP text_segment_sp = 2508 GetSectionList()->FindSectionByName(text_segment_name); 2509 if (text_segment_sp) { 2510 const lldb::addr_t text_segment_file_addr = 2511 text_segment_sp->GetFileAddress(); 2512 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2513 for (auto &e : trie_entries) 2514 e.entry.address += text_segment_file_addr; 2515 } 2516 } 2517 } 2518 2519 typedef std::set<ConstString> IndirectSymbols; 2520 IndirectSymbols indirect_symbol_names; 2521 2522 #if defined(__APPLE__) && \ 2523 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2524 2525 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2526 // optimized by moving LOCAL symbols out of the memory mapped portion of 2527 // the DSC. The symbol information has all been retained, but it isn't 2528 // available in the normal nlist data. However, there *are* duplicate 2529 // entries of *some* 2530 // LOCAL symbols in the normal nlist data. To handle this situation 2531 // correctly, we must first attempt 2532 // to parse any DSC unmapped symbol information. If we find any, we set a 2533 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2534 2535 if (m_header.flags & 0x80000000u) { 2536 // Before we can start mapping the DSC, we need to make certain the 2537 // target process is actually using the cache we can find. 2538 2539 // Next we need to determine the correct path for the dyld shared cache. 2540 2541 ArchSpec header_arch = GetArchitecture(); 2542 char dsc_path[PATH_MAX]; 2543 char dsc_path_development[PATH_MAX]; 2544 2545 snprintf( 2546 dsc_path, sizeof(dsc_path), "%s%s%s", 2547 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2548 */ 2549 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2550 header_arch.GetArchitectureName()); 2551 2552 snprintf( 2553 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2554 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2555 */ 2556 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2557 header_arch.GetArchitectureName(), ".development"); 2558 2559 FileSpec dsc_nondevelopment_filespec(dsc_path); 2560 FileSpec dsc_development_filespec(dsc_path_development); 2561 FileSpec dsc_filespec; 2562 2563 UUID dsc_uuid; 2564 UUID process_shared_cache_uuid; 2565 addr_t process_shared_cache_base_addr; 2566 2567 if (process) { 2568 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, 2569 process_shared_cache_uuid); 2570 } 2571 2572 // First see if we can find an exact match for the inferior process 2573 // shared cache UUID in the development or non-development shared caches 2574 // on disk. 2575 if (process_shared_cache_uuid.IsValid()) { 2576 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2577 UUID dsc_development_uuid = GetSharedCacheUUID( 2578 dsc_development_filespec, byte_order, addr_byte_size); 2579 if (dsc_development_uuid.IsValid() && 2580 dsc_development_uuid == process_shared_cache_uuid) { 2581 dsc_filespec = dsc_development_filespec; 2582 dsc_uuid = dsc_development_uuid; 2583 } 2584 } 2585 if (!dsc_uuid.IsValid() && 2586 FileSystem::Instance().Exists(dsc_nondevelopment_filespec)) { 2587 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2588 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2589 if (dsc_nondevelopment_uuid.IsValid() && 2590 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2591 dsc_filespec = dsc_nondevelopment_filespec; 2592 dsc_uuid = dsc_nondevelopment_uuid; 2593 } 2594 } 2595 } 2596 2597 // Failing a UUID match, prefer the development dyld_shared cache if both 2598 // are present. 2599 if (!FileSystem::Instance().Exists(dsc_filespec)) { 2600 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2601 dsc_filespec = dsc_development_filespec; 2602 } else { 2603 dsc_filespec = dsc_nondevelopment_filespec; 2604 } 2605 } 2606 2607 /* The dyld_cache_header has a pointer to the 2608 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2609 The dyld_cache_local_symbols_info structure gives us three things: 2610 1. The start and count of the nlist records in the dyld_shared_cache 2611 file 2612 2. The start and size of the strings for these nlist records 2613 3. The start and count of dyld_cache_local_symbols_entry entries 2614 2615 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2616 dyld shared cache. 2617 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2618 the dyld shared cache. 2619 The dyld_cache_local_symbols_entry also lists the start of this 2620 dylib/framework's nlist records 2621 and the count of how many nlist records there are for this 2622 dylib/framework. 2623 */ 2624 2625 // Process the dyld shared cache header to find the unmapped symbols 2626 2627 DataBufferSP dsc_data_sp = MapFileData( 2628 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2629 if (!dsc_uuid.IsValid()) { 2630 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2631 } 2632 if (dsc_data_sp) { 2633 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2634 2635 bool uuid_match = true; 2636 if (dsc_uuid.IsValid() && process) { 2637 if (process_shared_cache_uuid.IsValid() && 2638 dsc_uuid != process_shared_cache_uuid) { 2639 // The on-disk dyld_shared_cache file is not the same as the one in 2640 // this process' memory, don't use it. 2641 uuid_match = false; 2642 ModuleSP module_sp(GetModule()); 2643 if (module_sp) 2644 module_sp->ReportWarning("process shared cache does not match " 2645 "on-disk dyld_shared_cache file, some " 2646 "symbol names will be missing."); 2647 } 2648 } 2649 2650 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2651 2652 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2653 2654 // If the mappingOffset points to a location inside the header, we've 2655 // opened an old dyld shared cache, and should not proceed further. 2656 if (uuid_match && 2657 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2658 2659 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2660 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2661 mappingOffset); 2662 2663 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2664 byte_order, addr_byte_size); 2665 offset = 0; 2666 2667 // The File addresses (from the in-memory Mach-O load commands) for 2668 // the shared libraries in the shared library cache need to be 2669 // adjusted by an offset to match up with the dylibOffset identifying 2670 // field in the dyld_cache_local_symbol_entry's. This offset is 2671 // recorded in mapping_offset_value. 2672 const uint64_t mapping_offset_value = 2673 dsc_mapping_info_data.GetU64(&offset); 2674 2675 offset = 2676 offsetof(struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2677 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2678 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2679 2680 if (localSymbolsOffset && localSymbolsSize) { 2681 // Map the local symbols 2682 DataBufferSP dsc_local_symbols_data_sp = 2683 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2684 2685 if (dsc_local_symbols_data_sp) { 2686 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2687 byte_order, addr_byte_size); 2688 2689 offset = 0; 2690 2691 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 2692 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 2693 UndefinedNameToDescMap undefined_name_to_desc; 2694 SymbolIndexToName reexport_shlib_needs_fixup; 2695 2696 // Read the local_symbols_infos struct in one shot 2697 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2698 dsc_local_symbols_data.GetU32(&offset, 2699 &local_symbols_info.nlistOffset, 6); 2700 2701 SectionSP text_section_sp( 2702 section_list->FindSectionByName(GetSegmentNameTEXT())); 2703 2704 uint32_t header_file_offset = 2705 (text_section_sp->GetFileAddress() - mapping_offset_value); 2706 2707 offset = local_symbols_info.entriesOffset; 2708 for (uint32_t entry_index = 0; 2709 entry_index < local_symbols_info.entriesCount; entry_index++) { 2710 struct lldb_copy_dyld_cache_local_symbols_entry 2711 local_symbols_entry; 2712 local_symbols_entry.dylibOffset = 2713 dsc_local_symbols_data.GetU32(&offset); 2714 local_symbols_entry.nlistStartIndex = 2715 dsc_local_symbols_data.GetU32(&offset); 2716 local_symbols_entry.nlistCount = 2717 dsc_local_symbols_data.GetU32(&offset); 2718 2719 if (header_file_offset == local_symbols_entry.dylibOffset) { 2720 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2721 2722 // The normal nlist code cannot correctly size the Symbols 2723 // array, we need to allocate it here. 2724 sym = symtab->Resize( 2725 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2726 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2727 num_syms = symtab->GetNumSymbols(); 2728 2729 nlist_data_offset = 2730 local_symbols_info.nlistOffset + 2731 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2732 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2733 2734 for (uint32_t nlist_index = 0; 2735 nlist_index < local_symbols_entry.nlistCount; 2736 nlist_index++) { 2737 ///////////////////////////// 2738 { 2739 llvm::Optional<struct nlist_64> nlist_maybe = 2740 ParseNList(dsc_local_symbols_data, nlist_data_offset, 2741 nlist_byte_size); 2742 if (!nlist_maybe) 2743 break; 2744 struct nlist_64 nlist = *nlist_maybe; 2745 2746 SymbolType type = eSymbolTypeInvalid; 2747 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2748 string_table_offset + nlist.n_strx); 2749 2750 if (symbol_name == NULL) { 2751 // No symbol should be NULL, even the symbols with no 2752 // string values should have an offset zero which 2753 // points to an empty C-string 2754 Host::SystemLog( 2755 Host::eSystemLogError, 2756 "error: DSC unmapped local symbol[%u] has invalid " 2757 "string table offset 0x%x in %s, ignoring symbol\n", 2758 entry_index, nlist.n_strx, 2759 module_sp->GetFileSpec().GetPath().c_str()); 2760 continue; 2761 } 2762 if (symbol_name[0] == '\0') 2763 symbol_name = NULL; 2764 2765 const char *symbol_name_non_abi_mangled = NULL; 2766 2767 SectionSP symbol_section; 2768 uint32_t symbol_byte_size = 0; 2769 bool add_nlist = true; 2770 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2771 bool demangled_is_synthesized = false; 2772 bool is_gsym = false; 2773 bool set_value = true; 2774 2775 assert(sym_idx < num_syms); 2776 2777 sym[sym_idx].SetDebug(is_debug); 2778 2779 if (is_debug) { 2780 switch (nlist.n_type) { 2781 case N_GSYM: 2782 // global symbol: name,,NO_SECT,type,0 2783 // Sometimes the N_GSYM value contains the address. 2784 2785 // FIXME: In the .o files, we have a GSYM and a debug 2786 // symbol for all the ObjC data. They 2787 // have the same address, but we want to ensure that 2788 // we always find only the real symbol, 'cause we 2789 // don't currently correctly attribute the 2790 // GSYM one to the ObjCClass/Ivar/MetaClass 2791 // symbol type. This is a temporary hack to make 2792 // sure the ObjectiveC symbols get treated correctly. 2793 // To do this right, we should coalesce all the GSYM 2794 // & global symbols that have the same address. 2795 2796 is_gsym = true; 2797 sym[sym_idx].SetExternal(true); 2798 2799 if (symbol_name && symbol_name[0] == '_' && 2800 symbol_name[1] == 'O') { 2801 llvm::StringRef symbol_name_ref(symbol_name); 2802 if (symbol_name_ref.startswith( 2803 g_objc_v2_prefix_class)) { 2804 symbol_name_non_abi_mangled = symbol_name + 1; 2805 symbol_name = 2806 symbol_name + g_objc_v2_prefix_class.size(); 2807 type = eSymbolTypeObjCClass; 2808 demangled_is_synthesized = true; 2809 2810 } else if (symbol_name_ref.startswith( 2811 g_objc_v2_prefix_metaclass)) { 2812 symbol_name_non_abi_mangled = symbol_name + 1; 2813 symbol_name = 2814 symbol_name + g_objc_v2_prefix_metaclass.size(); 2815 type = eSymbolTypeObjCMetaClass; 2816 demangled_is_synthesized = true; 2817 } else if (symbol_name_ref.startswith( 2818 g_objc_v2_prefix_ivar)) { 2819 symbol_name_non_abi_mangled = symbol_name + 1; 2820 symbol_name = 2821 symbol_name + g_objc_v2_prefix_ivar.size(); 2822 type = eSymbolTypeObjCIVar; 2823 demangled_is_synthesized = true; 2824 } 2825 } else { 2826 if (nlist.n_value != 0) 2827 symbol_section = section_info.GetSection( 2828 nlist.n_sect, nlist.n_value); 2829 type = eSymbolTypeData; 2830 } 2831 break; 2832 2833 case N_FNAME: 2834 // procedure name (f77 kludge): name,,NO_SECT,0,0 2835 type = eSymbolTypeCompiler; 2836 break; 2837 2838 case N_FUN: 2839 // procedure: name,,n_sect,linenumber,address 2840 if (symbol_name) { 2841 type = eSymbolTypeCode; 2842 symbol_section = section_info.GetSection( 2843 nlist.n_sect, nlist.n_value); 2844 2845 N_FUN_addr_to_sym_idx.insert( 2846 std::make_pair(nlist.n_value, sym_idx)); 2847 // We use the current number of symbols in the 2848 // symbol table in lieu of using nlist_idx in case 2849 // we ever start trimming entries out 2850 N_FUN_indexes.push_back(sym_idx); 2851 } else { 2852 type = eSymbolTypeCompiler; 2853 2854 if (!N_FUN_indexes.empty()) { 2855 // Copy the size of the function into the 2856 // original 2857 // STAB entry so we don't have 2858 // to hunt for it later 2859 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2860 ->SetByteSize(nlist.n_value); 2861 N_FUN_indexes.pop_back(); 2862 // We don't really need the end function STAB as 2863 // it contains the size which we already placed 2864 // with the original symbol, so don't add it if 2865 // we want a minimal symbol table 2866 add_nlist = false; 2867 } 2868 } 2869 break; 2870 2871 case N_STSYM: 2872 // static symbol: name,,n_sect,type,address 2873 N_STSYM_addr_to_sym_idx.insert( 2874 std::make_pair(nlist.n_value, sym_idx)); 2875 symbol_section = section_info.GetSection(nlist.n_sect, 2876 nlist.n_value); 2877 if (symbol_name && symbol_name[0]) { 2878 type = ObjectFile::GetSymbolTypeFromName( 2879 symbol_name + 1, eSymbolTypeData); 2880 } 2881 break; 2882 2883 case N_LCSYM: 2884 // .lcomm symbol: name,,n_sect,type,address 2885 symbol_section = section_info.GetSection(nlist.n_sect, 2886 nlist.n_value); 2887 type = eSymbolTypeCommonBlock; 2888 break; 2889 2890 case N_BNSYM: 2891 // We use the current number of symbols in the symbol 2892 // table in lieu of using nlist_idx in case we ever 2893 // start trimming entries out Skip these if we want 2894 // minimal symbol tables 2895 add_nlist = false; 2896 break; 2897 2898 case N_ENSYM: 2899 // Set the size of the N_BNSYM to the terminating 2900 // index of this N_ENSYM so that we can always skip 2901 // the entire symbol if we need to navigate more 2902 // quickly at the source level when parsing STABS 2903 // Skip these if we want minimal symbol tables 2904 add_nlist = false; 2905 break; 2906 2907 case N_OPT: 2908 // emitted with gcc2_compiled and in gcc source 2909 type = eSymbolTypeCompiler; 2910 break; 2911 2912 case N_RSYM: 2913 // register sym: name,,NO_SECT,type,register 2914 type = eSymbolTypeVariable; 2915 break; 2916 2917 case N_SLINE: 2918 // src line: 0,,n_sect,linenumber,address 2919 symbol_section = section_info.GetSection(nlist.n_sect, 2920 nlist.n_value); 2921 type = eSymbolTypeLineEntry; 2922 break; 2923 2924 case N_SSYM: 2925 // structure elt: name,,NO_SECT,type,struct_offset 2926 type = eSymbolTypeVariableType; 2927 break; 2928 2929 case N_SO: 2930 // source file name 2931 type = eSymbolTypeSourceFile; 2932 if (symbol_name == NULL) { 2933 add_nlist = false; 2934 if (N_SO_index != UINT32_MAX) { 2935 // Set the size of the N_SO to the terminating 2936 // index of this N_SO so that we can always skip 2937 // the entire N_SO if we need to navigate more 2938 // quickly at the source level when parsing STABS 2939 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2940 symbol_ptr->SetByteSize(sym_idx); 2941 symbol_ptr->SetSizeIsSibling(true); 2942 } 2943 N_NSYM_indexes.clear(); 2944 N_INCL_indexes.clear(); 2945 N_BRAC_indexes.clear(); 2946 N_COMM_indexes.clear(); 2947 N_FUN_indexes.clear(); 2948 N_SO_index = UINT32_MAX; 2949 } else { 2950 // We use the current number of symbols in the 2951 // symbol table in lieu of using nlist_idx in case 2952 // we ever start trimming entries out 2953 const bool N_SO_has_full_path = symbol_name[0] == '/'; 2954 if (N_SO_has_full_path) { 2955 if ((N_SO_index == sym_idx - 1) && 2956 ((sym_idx - 1) < num_syms)) { 2957 // We have two consecutive N_SO entries where 2958 // the first contains a directory and the 2959 // second contains a full path. 2960 sym[sym_idx - 1].GetMangled().SetValue( 2961 ConstString(symbol_name), false); 2962 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2963 add_nlist = false; 2964 } else { 2965 // This is the first entry in a N_SO that 2966 // contains a directory or 2967 // a full path to the source file 2968 N_SO_index = sym_idx; 2969 } 2970 } else if ((N_SO_index == sym_idx - 1) && 2971 ((sym_idx - 1) < num_syms)) { 2972 // This is usually the second N_SO entry that 2973 // contains just the filename, so here we combine 2974 // it with the first one if we are minimizing the 2975 // symbol table 2976 const char *so_path = 2977 sym[sym_idx - 1] 2978 .GetMangled() 2979 .GetDemangledName( 2980 lldb::eLanguageTypeUnknown) 2981 .AsCString(); 2982 if (so_path && so_path[0]) { 2983 std::string full_so_path(so_path); 2984 const size_t double_slash_pos = 2985 full_so_path.find("//"); 2986 if (double_slash_pos != std::string::npos) { 2987 // The linker has been generating bad N_SO 2988 // entries with doubled up paths 2989 // in the format "%s%s" where the first 2990 // string in the DW_AT_comp_dir, and the 2991 // second is the directory for the source 2992 // file so you end up with a path that looks 2993 // like "/tmp/src//tmp/src/" 2994 FileSpec so_dir(so_path); 2995 if (!FileSystem::Instance().Exists(so_dir)) { 2996 so_dir.SetFile( 2997 &full_so_path[double_slash_pos + 1], 2998 FileSpec::Style::native); 2999 if (FileSystem::Instance().Exists(so_dir)) { 3000 // Trim off the incorrect path 3001 full_so_path.erase(0, double_slash_pos + 1); 3002 } 3003 } 3004 } 3005 if (*full_so_path.rbegin() != '/') 3006 full_so_path += '/'; 3007 full_so_path += symbol_name; 3008 sym[sym_idx - 1].GetMangled().SetValue( 3009 ConstString(full_so_path.c_str()), false); 3010 add_nlist = false; 3011 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3012 } 3013 } else { 3014 // This could be a relative path to a N_SO 3015 N_SO_index = sym_idx; 3016 } 3017 } 3018 break; 3019 3020 case N_OSO: 3021 // object file name: name,,0,0,st_mtime 3022 type = eSymbolTypeObjectFile; 3023 break; 3024 3025 case N_LSYM: 3026 // local sym: name,,NO_SECT,type,offset 3027 type = eSymbolTypeLocal; 3028 break; 3029 3030 // INCL scopes 3031 case N_BINCL: 3032 // include file beginning: name,,NO_SECT,0,sum We use 3033 // the current number of symbols in the symbol table 3034 // in lieu of using nlist_idx in case we ever start 3035 // trimming entries out 3036 N_INCL_indexes.push_back(sym_idx); 3037 type = eSymbolTypeScopeBegin; 3038 break; 3039 3040 case N_EINCL: 3041 // include file end: name,,NO_SECT,0,0 3042 // Set the size of the N_BINCL to the terminating 3043 // index of this N_EINCL so that we can always skip 3044 // the entire symbol if we need to navigate more 3045 // quickly at the source level when parsing STABS 3046 if (!N_INCL_indexes.empty()) { 3047 symbol_ptr = 3048 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3049 symbol_ptr->SetByteSize(sym_idx + 1); 3050 symbol_ptr->SetSizeIsSibling(true); 3051 N_INCL_indexes.pop_back(); 3052 } 3053 type = eSymbolTypeScopeEnd; 3054 break; 3055 3056 case N_SOL: 3057 // #included file name: name,,n_sect,0,address 3058 type = eSymbolTypeHeaderFile; 3059 3060 // We currently don't use the header files on darwin 3061 add_nlist = false; 3062 break; 3063 3064 case N_PARAMS: 3065 // compiler parameters: name,,NO_SECT,0,0 3066 type = eSymbolTypeCompiler; 3067 break; 3068 3069 case N_VERSION: 3070 // compiler version: name,,NO_SECT,0,0 3071 type = eSymbolTypeCompiler; 3072 break; 3073 3074 case N_OLEVEL: 3075 // compiler -O level: name,,NO_SECT,0,0 3076 type = eSymbolTypeCompiler; 3077 break; 3078 3079 case N_PSYM: 3080 // parameter: name,,NO_SECT,type,offset 3081 type = eSymbolTypeVariable; 3082 break; 3083 3084 case N_ENTRY: 3085 // alternate entry: name,,n_sect,linenumber,address 3086 symbol_section = section_info.GetSection(nlist.n_sect, 3087 nlist.n_value); 3088 type = eSymbolTypeLineEntry; 3089 break; 3090 3091 // Left and Right Braces 3092 case N_LBRAC: 3093 // left bracket: 0,,NO_SECT,nesting level,address We 3094 // use the current number of symbols in the symbol 3095 // table in lieu of using nlist_idx in case we ever 3096 // start trimming entries out 3097 symbol_section = section_info.GetSection(nlist.n_sect, 3098 nlist.n_value); 3099 N_BRAC_indexes.push_back(sym_idx); 3100 type = eSymbolTypeScopeBegin; 3101 break; 3102 3103 case N_RBRAC: 3104 // right bracket: 0,,NO_SECT,nesting level,address 3105 // Set the size of the N_LBRAC to the terminating 3106 // index of this N_RBRAC so that we can always skip 3107 // the entire symbol if we need to navigate more 3108 // quickly at the source level when parsing STABS 3109 symbol_section = section_info.GetSection(nlist.n_sect, 3110 nlist.n_value); 3111 if (!N_BRAC_indexes.empty()) { 3112 symbol_ptr = 3113 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3114 symbol_ptr->SetByteSize(sym_idx + 1); 3115 symbol_ptr->SetSizeIsSibling(true); 3116 N_BRAC_indexes.pop_back(); 3117 } 3118 type = eSymbolTypeScopeEnd; 3119 break; 3120 3121 case N_EXCL: 3122 // deleted include file: name,,NO_SECT,0,sum 3123 type = eSymbolTypeHeaderFile; 3124 break; 3125 3126 // COMM scopes 3127 case N_BCOMM: 3128 // begin common: name,,NO_SECT,0,0 3129 // We use the current number of symbols in the symbol 3130 // table in lieu of using nlist_idx in case we ever 3131 // start trimming entries out 3132 type = eSymbolTypeScopeBegin; 3133 N_COMM_indexes.push_back(sym_idx); 3134 break; 3135 3136 case N_ECOML: 3137 // end common (local name): 0,,n_sect,0,address 3138 symbol_section = section_info.GetSection(nlist.n_sect, 3139 nlist.n_value); 3140 // Fall through 3141 3142 case N_ECOMM: 3143 // end common: name,,n_sect,0,0 3144 // Set the size of the N_BCOMM to the terminating 3145 // index of this N_ECOMM/N_ECOML so that we can 3146 // always skip the entire symbol if we need to 3147 // navigate more quickly at the source level when 3148 // parsing STABS 3149 if (!N_COMM_indexes.empty()) { 3150 symbol_ptr = 3151 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3152 symbol_ptr->SetByteSize(sym_idx + 1); 3153 symbol_ptr->SetSizeIsSibling(true); 3154 N_COMM_indexes.pop_back(); 3155 } 3156 type = eSymbolTypeScopeEnd; 3157 break; 3158 3159 case N_LENG: 3160 // second stab entry with length information 3161 type = eSymbolTypeAdditional; 3162 break; 3163 3164 default: 3165 break; 3166 } 3167 } else { 3168 // uint8_t n_pext = N_PEXT & nlist.n_type; 3169 uint8_t n_type = N_TYPE & nlist.n_type; 3170 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3171 3172 switch (n_type) { 3173 case N_INDR: { 3174 const char *reexport_name_cstr = 3175 strtab_data.PeekCStr(nlist.n_value); 3176 if (reexport_name_cstr && reexport_name_cstr[0]) { 3177 type = eSymbolTypeReExported; 3178 ConstString reexport_name( 3179 reexport_name_cstr + 3180 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3181 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3182 set_value = false; 3183 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3184 indirect_symbol_names.insert(ConstString( 3185 symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3186 } else 3187 type = eSymbolTypeUndefined; 3188 } break; 3189 3190 case N_UNDF: 3191 if (symbol_name && symbol_name[0]) { 3192 ConstString undefined_name( 3193 symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3194 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3195 } 3196 // Fall through 3197 case N_PBUD: 3198 type = eSymbolTypeUndefined; 3199 break; 3200 3201 case N_ABS: 3202 type = eSymbolTypeAbsolute; 3203 break; 3204 3205 case N_SECT: { 3206 symbol_section = section_info.GetSection(nlist.n_sect, 3207 nlist.n_value); 3208 3209 if (symbol_section == NULL) { 3210 // TODO: warn about this? 3211 add_nlist = false; 3212 break; 3213 } 3214 3215 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3216 type = eSymbolTypeException; 3217 } else { 3218 uint32_t section_type = 3219 symbol_section->Get() & SECTION_TYPE; 3220 3221 switch (section_type) { 3222 case S_CSTRING_LITERALS: 3223 type = eSymbolTypeData; 3224 break; // section with only literal C strings 3225 case S_4BYTE_LITERALS: 3226 type = eSymbolTypeData; 3227 break; // section with only 4 byte literals 3228 case S_8BYTE_LITERALS: 3229 type = eSymbolTypeData; 3230 break; // section with only 8 byte literals 3231 case S_LITERAL_POINTERS: 3232 type = eSymbolTypeTrampoline; 3233 break; // section with only pointers to literals 3234 case S_NON_LAZY_SYMBOL_POINTERS: 3235 type = eSymbolTypeTrampoline; 3236 break; // section with only non-lazy symbol 3237 // pointers 3238 case S_LAZY_SYMBOL_POINTERS: 3239 type = eSymbolTypeTrampoline; 3240 break; // section with only lazy symbol pointers 3241 case S_SYMBOL_STUBS: 3242 type = eSymbolTypeTrampoline; 3243 break; // section with only symbol stubs, byte 3244 // size of stub in the reserved2 field 3245 case S_MOD_INIT_FUNC_POINTERS: 3246 type = eSymbolTypeCode; 3247 break; // section with only function pointers for 3248 // initialization 3249 case S_MOD_TERM_FUNC_POINTERS: 3250 type = eSymbolTypeCode; 3251 break; // section with only function pointers for 3252 // termination 3253 case S_INTERPOSING: 3254 type = eSymbolTypeTrampoline; 3255 break; // section with only pairs of function 3256 // pointers for interposing 3257 case S_16BYTE_LITERALS: 3258 type = eSymbolTypeData; 3259 break; // section with only 16 byte literals 3260 case S_DTRACE_DOF: 3261 type = eSymbolTypeInstrumentation; 3262 break; 3263 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3264 type = eSymbolTypeTrampoline; 3265 break; 3266 default: 3267 switch (symbol_section->GetType()) { 3268 case lldb::eSectionTypeCode: 3269 type = eSymbolTypeCode; 3270 break; 3271 case eSectionTypeData: 3272 case eSectionTypeDataCString: // Inlined C string 3273 // data 3274 case eSectionTypeDataCStringPointers: // Pointers 3275 // to C 3276 // string 3277 // data 3278 case eSectionTypeDataSymbolAddress: // Address of 3279 // a symbol in 3280 // the symbol 3281 // table 3282 case eSectionTypeData4: 3283 case eSectionTypeData8: 3284 case eSectionTypeData16: 3285 type = eSymbolTypeData; 3286 break; 3287 default: 3288 break; 3289 } 3290 break; 3291 } 3292 3293 if (type == eSymbolTypeInvalid) { 3294 const char *symbol_sect_name = 3295 symbol_section->GetName().AsCString(); 3296 if (symbol_section->IsDescendant( 3297 text_section_sp.get())) { 3298 if (symbol_section->IsClear( 3299 S_ATTR_PURE_INSTRUCTIONS | 3300 S_ATTR_SELF_MODIFYING_CODE | 3301 S_ATTR_SOME_INSTRUCTIONS)) 3302 type = eSymbolTypeData; 3303 else 3304 type = eSymbolTypeCode; 3305 } else if (symbol_section->IsDescendant( 3306 data_section_sp.get()) || 3307 symbol_section->IsDescendant( 3308 data_dirty_section_sp.get()) || 3309 symbol_section->IsDescendant( 3310 data_const_section_sp.get())) { 3311 if (symbol_sect_name && 3312 ::strstr(symbol_sect_name, "__objc") == 3313 symbol_sect_name) { 3314 type = eSymbolTypeRuntime; 3315 3316 if (symbol_name) { 3317 llvm::StringRef symbol_name_ref(symbol_name); 3318 if (symbol_name_ref.startswith("_OBJC_")) { 3319 llvm::StringRef 3320 g_objc_v2_prefix_class( 3321 "_OBJC_CLASS_$_"); 3322 llvm::StringRef 3323 g_objc_v2_prefix_metaclass( 3324 "_OBJC_METACLASS_$_"); 3325 llvm::StringRef 3326 g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 3327 if (symbol_name_ref.startswith( 3328 g_objc_v2_prefix_class)) { 3329 symbol_name_non_abi_mangled = 3330 symbol_name + 1; 3331 symbol_name = 3332 symbol_name + 3333 g_objc_v2_prefix_class.size(); 3334 type = eSymbolTypeObjCClass; 3335 demangled_is_synthesized = true; 3336 } else if ( 3337 symbol_name_ref.startswith( 3338 g_objc_v2_prefix_metaclass)) { 3339 symbol_name_non_abi_mangled = 3340 symbol_name + 1; 3341 symbol_name = 3342 symbol_name + 3343 g_objc_v2_prefix_metaclass.size(); 3344 type = eSymbolTypeObjCMetaClass; 3345 demangled_is_synthesized = true; 3346 } else if (symbol_name_ref.startswith( 3347 g_objc_v2_prefix_ivar)) { 3348 symbol_name_non_abi_mangled = 3349 symbol_name + 1; 3350 symbol_name = 3351 symbol_name + 3352 g_objc_v2_prefix_ivar.size(); 3353 type = eSymbolTypeObjCIVar; 3354 demangled_is_synthesized = true; 3355 } 3356 } 3357 } 3358 } else if (symbol_sect_name && 3359 ::strstr(symbol_sect_name, 3360 "__gcc_except_tab") == 3361 symbol_sect_name) { 3362 type = eSymbolTypeException; 3363 } else { 3364 type = eSymbolTypeData; 3365 } 3366 } else if (symbol_sect_name && 3367 ::strstr(symbol_sect_name, "__IMPORT") == 3368 symbol_sect_name) { 3369 type = eSymbolTypeTrampoline; 3370 } else if (symbol_section->IsDescendant( 3371 objc_section_sp.get())) { 3372 type = eSymbolTypeRuntime; 3373 if (symbol_name && symbol_name[0] == '.') { 3374 llvm::StringRef symbol_name_ref(symbol_name); 3375 llvm::StringRef 3376 g_objc_v1_prefix_class(".objc_class_name_"); 3377 if (symbol_name_ref.startswith( 3378 g_objc_v1_prefix_class)) { 3379 symbol_name_non_abi_mangled = symbol_name; 3380 symbol_name = symbol_name + 3381 g_objc_v1_prefix_class.size(); 3382 type = eSymbolTypeObjCClass; 3383 demangled_is_synthesized = true; 3384 } 3385 } 3386 } 3387 } 3388 } 3389 } break; 3390 } 3391 } 3392 3393 if (add_nlist) { 3394 uint64_t symbol_value = nlist.n_value; 3395 if (symbol_name_non_abi_mangled) { 3396 sym[sym_idx].GetMangled().SetMangledName( 3397 ConstString(symbol_name_non_abi_mangled)); 3398 sym[sym_idx].GetMangled().SetDemangledName( 3399 ConstString(symbol_name)); 3400 } else { 3401 bool symbol_name_is_mangled = false; 3402 3403 if (symbol_name && symbol_name[0] == '_') { 3404 symbol_name_is_mangled = symbol_name[1] == '_'; 3405 symbol_name++; // Skip the leading underscore 3406 } 3407 3408 if (symbol_name) { 3409 ConstString const_symbol_name(symbol_name); 3410 sym[sym_idx].GetMangled().SetValue( 3411 const_symbol_name, symbol_name_is_mangled); 3412 if (is_gsym && is_debug) { 3413 const char *gsym_name = 3414 sym[sym_idx] 3415 .GetMangled() 3416 .GetName(lldb::eLanguageTypeUnknown, 3417 Mangled::ePreferMangled) 3418 .GetCString(); 3419 if (gsym_name) 3420 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3421 } 3422 } 3423 } 3424 if (symbol_section) { 3425 const addr_t section_file_addr = 3426 symbol_section->GetFileAddress(); 3427 if (symbol_byte_size == 0 && 3428 function_starts_count > 0) { 3429 addr_t symbol_lookup_file_addr = nlist.n_value; 3430 // Do an exact address match for non-ARM addresses, 3431 // else get the closest since the symbol might be a 3432 // thumb symbol which has an address with bit zero 3433 // set 3434 FunctionStarts::Entry *func_start_entry = 3435 function_starts.FindEntry(symbol_lookup_file_addr, 3436 !is_arm); 3437 if (is_arm && func_start_entry) { 3438 // Verify that the function start address is the 3439 // symbol address (ARM) or the symbol address + 1 3440 // (thumb) 3441 if (func_start_entry->addr != 3442 symbol_lookup_file_addr && 3443 func_start_entry->addr != 3444 (symbol_lookup_file_addr + 1)) { 3445 // Not the right entry, NULL it out... 3446 func_start_entry = NULL; 3447 } 3448 } 3449 if (func_start_entry) { 3450 func_start_entry->data = true; 3451 3452 addr_t symbol_file_addr = func_start_entry->addr; 3453 uint32_t symbol_flags = 0; 3454 if (is_arm) { 3455 if (symbol_file_addr & 1) 3456 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3457 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3458 } 3459 3460 const FunctionStarts::Entry *next_func_start_entry = 3461 function_starts.FindNextEntry(func_start_entry); 3462 const addr_t section_end_file_addr = 3463 section_file_addr + 3464 symbol_section->GetByteSize(); 3465 if (next_func_start_entry) { 3466 addr_t next_symbol_file_addr = 3467 next_func_start_entry->addr; 3468 // Be sure the clear the Thumb address bit when 3469 // we calculate the size from the current and 3470 // next address 3471 if (is_arm) 3472 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3473 symbol_byte_size = std::min<lldb::addr_t>( 3474 next_symbol_file_addr - symbol_file_addr, 3475 section_end_file_addr - symbol_file_addr); 3476 } else { 3477 symbol_byte_size = 3478 section_end_file_addr - symbol_file_addr; 3479 } 3480 } 3481 } 3482 symbol_value -= section_file_addr; 3483 } 3484 3485 if (is_debug == false) { 3486 if (type == eSymbolTypeCode) { 3487 // See if we can find a N_FUN entry for any code 3488 // symbols. If we do find a match, and the name 3489 // matches, then we can merge the two into just the 3490 // function symbol to avoid duplicate entries in 3491 // the symbol table 3492 auto range = 3493 N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3494 if (range.first != range.second) { 3495 bool found_it = false; 3496 for (const auto pos = range.first; 3497 pos != range.second; ++pos) { 3498 if (sym[sym_idx].GetMangled().GetName( 3499 lldb::eLanguageTypeUnknown, 3500 Mangled::ePreferMangled) == 3501 sym[pos->second].GetMangled().GetName( 3502 lldb::eLanguageTypeUnknown, 3503 Mangled::ePreferMangled)) { 3504 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3505 // We just need the flags from the linker 3506 // symbol, so put these flags 3507 // into the N_FUN flags to avoid duplicate 3508 // symbols in the symbol table 3509 sym[pos->second].SetExternal( 3510 sym[sym_idx].IsExternal()); 3511 sym[pos->second].SetFlags(nlist.n_type << 16 | 3512 nlist.n_desc); 3513 if (resolver_addresses.find(nlist.n_value) != 3514 resolver_addresses.end()) 3515 sym[pos->second].SetType(eSymbolTypeResolver); 3516 sym[sym_idx].Clear(); 3517 found_it = true; 3518 break; 3519 } 3520 } 3521 if (found_it) 3522 continue; 3523 } else { 3524 if (resolver_addresses.find(nlist.n_value) != 3525 resolver_addresses.end()) 3526 type = eSymbolTypeResolver; 3527 } 3528 } else if (type == eSymbolTypeData || 3529 type == eSymbolTypeObjCClass || 3530 type == eSymbolTypeObjCMetaClass || 3531 type == eSymbolTypeObjCIVar) { 3532 // See if we can find a N_STSYM entry for any data 3533 // symbols. If we do find a match, and the name 3534 // matches, then we can merge the two into just the 3535 // Static symbol to avoid duplicate entries in the 3536 // symbol table 3537 auto range = N_STSYM_addr_to_sym_idx.equal_range( 3538 nlist.n_value); 3539 if (range.first != range.second) { 3540 bool found_it = false; 3541 for (const auto pos = range.first; 3542 pos != range.second; ++pos) { 3543 if (sym[sym_idx].GetMangled().GetName( 3544 lldb::eLanguageTypeUnknown, 3545 Mangled::ePreferMangled) == 3546 sym[pos->second].GetMangled().GetName( 3547 lldb::eLanguageTypeUnknown, 3548 Mangled::ePreferMangled)) { 3549 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3550 // We just need the flags from the linker 3551 // symbol, so put these flags 3552 // into the N_STSYM flags to avoid duplicate 3553 // symbols in the symbol table 3554 sym[pos->second].SetExternal( 3555 sym[sym_idx].IsExternal()); 3556 sym[pos->second].SetFlags(nlist.n_type << 16 | 3557 nlist.n_desc); 3558 sym[sym_idx].Clear(); 3559 found_it = true; 3560 break; 3561 } 3562 } 3563 if (found_it) 3564 continue; 3565 } else { 3566 const char *gsym_name = 3567 sym[sym_idx] 3568 .GetMangled() 3569 .GetName(lldb::eLanguageTypeUnknown, 3570 Mangled::ePreferMangled) 3571 .GetCString(); 3572 if (gsym_name) { 3573 // Combine N_GSYM stab entries with the non 3574 // stab symbol 3575 ConstNameToSymbolIndexMap::const_iterator pos = 3576 N_GSYM_name_to_sym_idx.find(gsym_name); 3577 if (pos != N_GSYM_name_to_sym_idx.end()) { 3578 const uint32_t GSYM_sym_idx = pos->second; 3579 m_nlist_idx_to_sym_idx[nlist_idx] = 3580 GSYM_sym_idx; 3581 // Copy the address, because often the N_GSYM 3582 // address has an invalid address of zero 3583 // when the global is a common symbol 3584 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3585 symbol_section); 3586 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3587 symbol_value); 3588 // We just need the flags from the linker 3589 // symbol, so put these flags 3590 // into the N_GSYM flags to avoid duplicate 3591 // symbols in the symbol table 3592 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 3593 nlist.n_desc); 3594 sym[sym_idx].Clear(); 3595 continue; 3596 } 3597 } 3598 } 3599 } 3600 } 3601 3602 sym[sym_idx].SetID(nlist_idx); 3603 sym[sym_idx].SetType(type); 3604 if (set_value) { 3605 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 3606 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3607 } 3608 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 3609 3610 if (symbol_byte_size > 0) 3611 sym[sym_idx].SetByteSize(symbol_byte_size); 3612 3613 if (demangled_is_synthesized) 3614 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3615 ++sym_idx; 3616 } else { 3617 sym[sym_idx].Clear(); 3618 } 3619 } 3620 ///////////////////////////// 3621 } 3622 break; // No more entries to consider 3623 } 3624 } 3625 3626 for (const auto &pos : reexport_shlib_needs_fixup) { 3627 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3628 if (undef_pos != undefined_name_to_desc.end()) { 3629 const uint8_t dylib_ordinal = 3630 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3631 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3632 sym[pos.first].SetReExportedSymbolSharedLibrary( 3633 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3634 } 3635 } 3636 } 3637 } 3638 } 3639 } 3640 } 3641 3642 // Must reset this in case it was mutated above! 3643 nlist_data_offset = 0; 3644 #endif 3645 3646 if (nlist_data.GetByteSize() > 0) { 3647 3648 // If the sym array was not created while parsing the DSC unmapped 3649 // symbols, create it now. 3650 if (sym == nullptr) { 3651 sym = 3652 symtab->Resize(symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3653 num_syms = symtab->GetNumSymbols(); 3654 } 3655 3656 if (unmapped_local_symbols_found) { 3657 assert(m_dysymtab.ilocalsym == 0); 3658 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3659 nlist_idx = m_dysymtab.nlocalsym; 3660 } else { 3661 nlist_idx = 0; 3662 } 3663 3664 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 3665 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 3666 UndefinedNameToDescMap undefined_name_to_desc; 3667 SymbolIndexToName reexport_shlib_needs_fixup; 3668 3669 // Symtab parsing is a huge mess. Everything is entangled and the code 3670 // requires access to a ridiculous amount of variables. LLDB depends 3671 // heavily on the proper merging of symbols and to get that right we need 3672 // to make sure we have parsed all the debug symbols first. Therefore we 3673 // invoke the lambda twice, once to parse only the debug symbols and then 3674 // once more to parse the remaining symbols. 3675 auto ParseSymbolLambda = [&](struct nlist_64 &nlist, uint32_t nlist_idx, 3676 bool debug_only) { 3677 const bool is_debug = ((nlist.n_type & N_STAB) != 0); 3678 if (is_debug != debug_only) 3679 return true; 3680 3681 const char *symbol_name_non_abi_mangled = nullptr; 3682 const char *symbol_name = nullptr; 3683 3684 if (have_strtab_data) { 3685 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3686 3687 if (symbol_name == nullptr) { 3688 // No symbol should be NULL, even the symbols with no string values 3689 // should have an offset zero which points to an empty C-string 3690 Host::SystemLog(Host::eSystemLogError, 3691 "error: symbol[%u] has invalid string table offset " 3692 "0x%x in %s, ignoring symbol\n", 3693 nlist_idx, nlist.n_strx, 3694 module_sp->GetFileSpec().GetPath().c_str()); 3695 return true; 3696 } 3697 if (symbol_name[0] == '\0') 3698 symbol_name = nullptr; 3699 } else { 3700 const addr_t str_addr = strtab_addr + nlist.n_strx; 3701 Status str_error; 3702 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3703 str_error)) 3704 symbol_name = memory_symbol_name.c_str(); 3705 } 3706 3707 SymbolType type = eSymbolTypeInvalid; 3708 SectionSP symbol_section; 3709 lldb::addr_t symbol_byte_size = 0; 3710 bool add_nlist = true; 3711 bool is_gsym = false; 3712 bool demangled_is_synthesized = false; 3713 bool set_value = true; 3714 3715 assert(sym_idx < num_syms); 3716 sym[sym_idx].SetDebug(is_debug); 3717 3718 if (is_debug) { 3719 switch (nlist.n_type) { 3720 case N_GSYM: 3721 // global symbol: name,,NO_SECT,type,0 3722 // Sometimes the N_GSYM value contains the address. 3723 3724 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3725 // the ObjC data. They 3726 // have the same address, but we want to ensure that we always find 3727 // only the real symbol, 'cause we don't currently correctly 3728 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3729 // type. This is a temporary hack to make sure the ObjectiveC 3730 // symbols get treated correctly. To do this right, we should 3731 // coalesce all the GSYM & global symbols that have the same 3732 // address. 3733 is_gsym = true; 3734 sym[sym_idx].SetExternal(true); 3735 3736 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3737 llvm::StringRef symbol_name_ref(symbol_name); 3738 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3739 symbol_name_non_abi_mangled = symbol_name + 1; 3740 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3741 type = eSymbolTypeObjCClass; 3742 demangled_is_synthesized = true; 3743 3744 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 3745 symbol_name_non_abi_mangled = symbol_name + 1; 3746 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3747 type = eSymbolTypeObjCMetaClass; 3748 demangled_is_synthesized = true; 3749 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3750 symbol_name_non_abi_mangled = symbol_name + 1; 3751 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3752 type = eSymbolTypeObjCIVar; 3753 demangled_is_synthesized = true; 3754 } 3755 } else { 3756 if (nlist.n_value != 0) 3757 symbol_section = 3758 section_info.GetSection(nlist.n_sect, nlist.n_value); 3759 type = eSymbolTypeData; 3760 } 3761 break; 3762 3763 case N_FNAME: 3764 // procedure name (f77 kludge): name,,NO_SECT,0,0 3765 type = eSymbolTypeCompiler; 3766 break; 3767 3768 case N_FUN: 3769 // procedure: name,,n_sect,linenumber,address 3770 if (symbol_name) { 3771 type = eSymbolTypeCode; 3772 symbol_section = 3773 section_info.GetSection(nlist.n_sect, nlist.n_value); 3774 3775 N_FUN_addr_to_sym_idx.insert( 3776 std::make_pair(nlist.n_value, sym_idx)); 3777 // We use the current number of symbols in the symbol table in 3778 // lieu of using nlist_idx in case we ever start trimming entries 3779 // out 3780 N_FUN_indexes.push_back(sym_idx); 3781 } else { 3782 type = eSymbolTypeCompiler; 3783 3784 if (!N_FUN_indexes.empty()) { 3785 // Copy the size of the function into the original STAB entry 3786 // so we don't have to hunt for it later 3787 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3788 ->SetByteSize(nlist.n_value); 3789 N_FUN_indexes.pop_back(); 3790 // We don't really need the end function STAB as it contains 3791 // the size which we already placed with the original symbol, 3792 // so don't add it if we want a minimal symbol table 3793 add_nlist = false; 3794 } 3795 } 3796 break; 3797 3798 case N_STSYM: 3799 // static symbol: name,,n_sect,type,address 3800 N_STSYM_addr_to_sym_idx.insert( 3801 std::make_pair(nlist.n_value, sym_idx)); 3802 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3803 if (symbol_name && symbol_name[0]) { 3804 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3805 eSymbolTypeData); 3806 } 3807 break; 3808 3809 case N_LCSYM: 3810 // .lcomm symbol: name,,n_sect,type,address 3811 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3812 type = eSymbolTypeCommonBlock; 3813 break; 3814 3815 case N_BNSYM: 3816 // We use the current number of symbols in the symbol table in lieu 3817 // of using nlist_idx in case we ever start trimming entries out 3818 // Skip these if we want minimal symbol tables 3819 add_nlist = false; 3820 break; 3821 3822 case N_ENSYM: 3823 // Set the size of the N_BNSYM to the terminating index of this 3824 // N_ENSYM so that we can always skip the entire symbol if we need 3825 // to navigate more quickly at the source level when parsing STABS 3826 // Skip these if we want minimal symbol tables 3827 add_nlist = false; 3828 break; 3829 3830 case N_OPT: 3831 // emitted with gcc2_compiled and in gcc source 3832 type = eSymbolTypeCompiler; 3833 break; 3834 3835 case N_RSYM: 3836 // register sym: name,,NO_SECT,type,register 3837 type = eSymbolTypeVariable; 3838 break; 3839 3840 case N_SLINE: 3841 // src line: 0,,n_sect,linenumber,address 3842 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3843 type = eSymbolTypeLineEntry; 3844 break; 3845 3846 case N_SSYM: 3847 // structure elt: name,,NO_SECT,type,struct_offset 3848 type = eSymbolTypeVariableType; 3849 break; 3850 3851 case N_SO: 3852 // source file name 3853 type = eSymbolTypeSourceFile; 3854 if (symbol_name == nullptr) { 3855 add_nlist = false; 3856 if (N_SO_index != UINT32_MAX) { 3857 // Set the size of the N_SO to the terminating index of this 3858 // N_SO so that we can always skip the entire N_SO if we need 3859 // to navigate more quickly at the source level when parsing 3860 // STABS 3861 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3862 symbol_ptr->SetByteSize(sym_idx); 3863 symbol_ptr->SetSizeIsSibling(true); 3864 } 3865 N_NSYM_indexes.clear(); 3866 N_INCL_indexes.clear(); 3867 N_BRAC_indexes.clear(); 3868 N_COMM_indexes.clear(); 3869 N_FUN_indexes.clear(); 3870 N_SO_index = UINT32_MAX; 3871 } else { 3872 // We use the current number of symbols in the symbol table in 3873 // lieu of using nlist_idx in case we ever start trimming entries 3874 // out 3875 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3876 if (N_SO_has_full_path) { 3877 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3878 // We have two consecutive N_SO entries where the first 3879 // contains a directory and the second contains a full path. 3880 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), 3881 false); 3882 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3883 add_nlist = false; 3884 } else { 3885 // This is the first entry in a N_SO that contains a 3886 // directory or a full path to the source file 3887 N_SO_index = sym_idx; 3888 } 3889 } else if ((N_SO_index == sym_idx - 1) && 3890 ((sym_idx - 1) < num_syms)) { 3891 // This is usually the second N_SO entry that contains just the 3892 // filename, so here we combine it with the first one if we are 3893 // minimizing the symbol table 3894 const char *so_path = 3895 sym[sym_idx - 1] 3896 .GetMangled() 3897 .GetDemangledName(lldb::eLanguageTypeUnknown) 3898 .AsCString(); 3899 if (so_path && so_path[0]) { 3900 std::string full_so_path(so_path); 3901 const size_t double_slash_pos = full_so_path.find("//"); 3902 if (double_slash_pos != std::string::npos) { 3903 // The linker has been generating bad N_SO entries with 3904 // doubled up paths in the format "%s%s" where the first 3905 // string in the DW_AT_comp_dir, and the second is the 3906 // directory for the source file so you end up with a path 3907 // that looks like "/tmp/src//tmp/src/" 3908 FileSpec so_dir(so_path); 3909 if (!FileSystem::Instance().Exists(so_dir)) { 3910 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 3911 FileSpec::Style::native); 3912 if (FileSystem::Instance().Exists(so_dir)) { 3913 // Trim off the incorrect path 3914 full_so_path.erase(0, double_slash_pos + 1); 3915 } 3916 } 3917 } 3918 if (*full_so_path.rbegin() != '/') 3919 full_so_path += '/'; 3920 full_so_path += symbol_name; 3921 sym[sym_idx - 1].GetMangled().SetValue( 3922 ConstString(full_so_path.c_str()), false); 3923 add_nlist = false; 3924 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3925 } 3926 } else { 3927 // This could be a relative path to a N_SO 3928 N_SO_index = sym_idx; 3929 } 3930 } 3931 break; 3932 3933 case N_OSO: 3934 // object file name: name,,0,0,st_mtime 3935 type = eSymbolTypeObjectFile; 3936 break; 3937 3938 case N_LSYM: 3939 // local sym: name,,NO_SECT,type,offset 3940 type = eSymbolTypeLocal; 3941 break; 3942 3943 // INCL scopes 3944 case N_BINCL: 3945 // include file beginning: name,,NO_SECT,0,sum We use the current 3946 // number of symbols in the symbol table in lieu of using nlist_idx 3947 // in case we ever start trimming entries out 3948 N_INCL_indexes.push_back(sym_idx); 3949 type = eSymbolTypeScopeBegin; 3950 break; 3951 3952 case N_EINCL: 3953 // include file end: name,,NO_SECT,0,0 3954 // Set the size of the N_BINCL to the terminating index of this 3955 // N_EINCL so that we can always skip the entire symbol if we need 3956 // to navigate more quickly at the source level when parsing STABS 3957 if (!N_INCL_indexes.empty()) { 3958 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3959 symbol_ptr->SetByteSize(sym_idx + 1); 3960 symbol_ptr->SetSizeIsSibling(true); 3961 N_INCL_indexes.pop_back(); 3962 } 3963 type = eSymbolTypeScopeEnd; 3964 break; 3965 3966 case N_SOL: 3967 // #included file name: name,,n_sect,0,address 3968 type = eSymbolTypeHeaderFile; 3969 3970 // We currently don't use the header files on darwin 3971 add_nlist = false; 3972 break; 3973 3974 case N_PARAMS: 3975 // compiler parameters: name,,NO_SECT,0,0 3976 type = eSymbolTypeCompiler; 3977 break; 3978 3979 case N_VERSION: 3980 // compiler version: name,,NO_SECT,0,0 3981 type = eSymbolTypeCompiler; 3982 break; 3983 3984 case N_OLEVEL: 3985 // compiler -O level: name,,NO_SECT,0,0 3986 type = eSymbolTypeCompiler; 3987 break; 3988 3989 case N_PSYM: 3990 // parameter: name,,NO_SECT,type,offset 3991 type = eSymbolTypeVariable; 3992 break; 3993 3994 case N_ENTRY: 3995 // alternate entry: name,,n_sect,linenumber,address 3996 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3997 type = eSymbolTypeLineEntry; 3998 break; 3999 4000 // Left and Right Braces 4001 case N_LBRAC: 4002 // left bracket: 0,,NO_SECT,nesting level,address We use the 4003 // current number of symbols in the symbol table in lieu of using 4004 // nlist_idx in case we ever start trimming entries out 4005 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4006 N_BRAC_indexes.push_back(sym_idx); 4007 type = eSymbolTypeScopeBegin; 4008 break; 4009 4010 case N_RBRAC: 4011 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4012 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4013 // can always skip the entire symbol if we need to navigate more 4014 // quickly at the source level when parsing STABS 4015 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4016 if (!N_BRAC_indexes.empty()) { 4017 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4018 symbol_ptr->SetByteSize(sym_idx + 1); 4019 symbol_ptr->SetSizeIsSibling(true); 4020 N_BRAC_indexes.pop_back(); 4021 } 4022 type = eSymbolTypeScopeEnd; 4023 break; 4024 4025 case N_EXCL: 4026 // deleted include file: name,,NO_SECT,0,sum 4027 type = eSymbolTypeHeaderFile; 4028 break; 4029 4030 // COMM scopes 4031 case N_BCOMM: 4032 // begin common: name,,NO_SECT,0,0 4033 // We use the current number of symbols in the symbol table in lieu 4034 // of using nlist_idx in case we ever start trimming entries out 4035 type = eSymbolTypeScopeBegin; 4036 N_COMM_indexes.push_back(sym_idx); 4037 break; 4038 4039 case N_ECOML: 4040 // end common (local name): 0,,n_sect,0,address 4041 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4042 LLVM_FALLTHROUGH; 4043 4044 case N_ECOMM: 4045 // end common: name,,n_sect,0,0 4046 // Set the size of the N_BCOMM to the terminating index of this 4047 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4048 // we need to navigate more quickly at the source level when 4049 // parsing STABS 4050 if (!N_COMM_indexes.empty()) { 4051 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4052 symbol_ptr->SetByteSize(sym_idx + 1); 4053 symbol_ptr->SetSizeIsSibling(true); 4054 N_COMM_indexes.pop_back(); 4055 } 4056 type = eSymbolTypeScopeEnd; 4057 break; 4058 4059 case N_LENG: 4060 // second stab entry with length information 4061 type = eSymbolTypeAdditional; 4062 break; 4063 4064 default: 4065 break; 4066 } 4067 } else { 4068 uint8_t n_type = N_TYPE & nlist.n_type; 4069 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4070 4071 switch (n_type) { 4072 case N_INDR: { 4073 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4074 if (reexport_name_cstr && reexport_name_cstr[0]) { 4075 type = eSymbolTypeReExported; 4076 ConstString reexport_name(reexport_name_cstr + 4077 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4078 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4079 set_value = false; 4080 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4081 indirect_symbol_names.insert( 4082 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4083 } else 4084 type = eSymbolTypeUndefined; 4085 } break; 4086 4087 case N_UNDF: 4088 if (symbol_name && symbol_name[0]) { 4089 ConstString undefined_name(symbol_name + 4090 ((symbol_name[0] == '_') ? 1 : 0)); 4091 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4092 } 4093 LLVM_FALLTHROUGH; 4094 4095 case N_PBUD: 4096 type = eSymbolTypeUndefined; 4097 break; 4098 4099 case N_ABS: 4100 type = eSymbolTypeAbsolute; 4101 break; 4102 4103 case N_SECT: { 4104 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4105 4106 if (!symbol_section) { 4107 // TODO: warn about this? 4108 add_nlist = false; 4109 break; 4110 } 4111 4112 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4113 type = eSymbolTypeException; 4114 } else { 4115 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4116 4117 switch (section_type) { 4118 case S_CSTRING_LITERALS: 4119 type = eSymbolTypeData; 4120 break; // section with only literal C strings 4121 case S_4BYTE_LITERALS: 4122 type = eSymbolTypeData; 4123 break; // section with only 4 byte literals 4124 case S_8BYTE_LITERALS: 4125 type = eSymbolTypeData; 4126 break; // section with only 8 byte literals 4127 case S_LITERAL_POINTERS: 4128 type = eSymbolTypeTrampoline; 4129 break; // section with only pointers to literals 4130 case S_NON_LAZY_SYMBOL_POINTERS: 4131 type = eSymbolTypeTrampoline; 4132 break; // section with only non-lazy symbol pointers 4133 case S_LAZY_SYMBOL_POINTERS: 4134 type = eSymbolTypeTrampoline; 4135 break; // section with only lazy symbol pointers 4136 case S_SYMBOL_STUBS: 4137 type = eSymbolTypeTrampoline; 4138 break; // section with only symbol stubs, byte size of stub in 4139 // the reserved2 field 4140 case S_MOD_INIT_FUNC_POINTERS: 4141 type = eSymbolTypeCode; 4142 break; // section with only function pointers for initialization 4143 case S_MOD_TERM_FUNC_POINTERS: 4144 type = eSymbolTypeCode; 4145 break; // section with only function pointers for termination 4146 case S_INTERPOSING: 4147 type = eSymbolTypeTrampoline; 4148 break; // section with only pairs of function pointers for 4149 // interposing 4150 case S_16BYTE_LITERALS: 4151 type = eSymbolTypeData; 4152 break; // section with only 16 byte literals 4153 case S_DTRACE_DOF: 4154 type = eSymbolTypeInstrumentation; 4155 break; 4156 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4157 type = eSymbolTypeTrampoline; 4158 break; 4159 default: 4160 switch (symbol_section->GetType()) { 4161 case lldb::eSectionTypeCode: 4162 type = eSymbolTypeCode; 4163 break; 4164 case eSectionTypeData: 4165 case eSectionTypeDataCString: // Inlined C string data 4166 case eSectionTypeDataCStringPointers: // Pointers to C string 4167 // data 4168 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4169 // the symbol table 4170 case eSectionTypeData4: 4171 case eSectionTypeData8: 4172 case eSectionTypeData16: 4173 type = eSymbolTypeData; 4174 break; 4175 default: 4176 break; 4177 } 4178 break; 4179 } 4180 4181 if (type == eSymbolTypeInvalid) { 4182 const char *symbol_sect_name = 4183 symbol_section->GetName().AsCString(); 4184 if (symbol_section->IsDescendant(text_section_sp.get())) { 4185 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4186 S_ATTR_SELF_MODIFYING_CODE | 4187 S_ATTR_SOME_INSTRUCTIONS)) 4188 type = eSymbolTypeData; 4189 else 4190 type = eSymbolTypeCode; 4191 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 4192 symbol_section->IsDescendant( 4193 data_dirty_section_sp.get()) || 4194 symbol_section->IsDescendant( 4195 data_const_section_sp.get())) { 4196 if (symbol_sect_name && 4197 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 4198 type = eSymbolTypeRuntime; 4199 4200 if (symbol_name) { 4201 llvm::StringRef symbol_name_ref(symbol_name); 4202 if (symbol_name_ref.startswith("_OBJC_")) { 4203 llvm::StringRef g_objc_v2_prefix_class( 4204 "_OBJC_CLASS_$_"); 4205 llvm::StringRef g_objc_v2_prefix_metaclass( 4206 "_OBJC_METACLASS_$_"); 4207 llvm::StringRef g_objc_v2_prefix_ivar( 4208 "_OBJC_IVAR_$_"); 4209 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 4210 symbol_name_non_abi_mangled = symbol_name + 1; 4211 symbol_name = 4212 symbol_name + g_objc_v2_prefix_class.size(); 4213 type = eSymbolTypeObjCClass; 4214 demangled_is_synthesized = true; 4215 } else if (symbol_name_ref.startswith( 4216 g_objc_v2_prefix_metaclass)) { 4217 symbol_name_non_abi_mangled = symbol_name + 1; 4218 symbol_name = 4219 symbol_name + g_objc_v2_prefix_metaclass.size(); 4220 type = eSymbolTypeObjCMetaClass; 4221 demangled_is_synthesized = true; 4222 } else if (symbol_name_ref.startswith( 4223 g_objc_v2_prefix_ivar)) { 4224 symbol_name_non_abi_mangled = symbol_name + 1; 4225 symbol_name = 4226 symbol_name + g_objc_v2_prefix_ivar.size(); 4227 type = eSymbolTypeObjCIVar; 4228 demangled_is_synthesized = true; 4229 } 4230 } 4231 } 4232 } else if (symbol_sect_name && 4233 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4234 symbol_sect_name) { 4235 type = eSymbolTypeException; 4236 } else { 4237 type = eSymbolTypeData; 4238 } 4239 } else if (symbol_sect_name && 4240 ::strstr(symbol_sect_name, "__IMPORT") == 4241 symbol_sect_name) { 4242 type = eSymbolTypeTrampoline; 4243 } else if (symbol_section->IsDescendant(objc_section_sp.get())) { 4244 type = eSymbolTypeRuntime; 4245 if (symbol_name && symbol_name[0] == '.') { 4246 llvm::StringRef symbol_name_ref(symbol_name); 4247 llvm::StringRef g_objc_v1_prefix_class( 4248 ".objc_class_name_"); 4249 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4250 symbol_name_non_abi_mangled = symbol_name; 4251 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4252 type = eSymbolTypeObjCClass; 4253 demangled_is_synthesized = true; 4254 } 4255 } 4256 } 4257 } 4258 } 4259 } break; 4260 } 4261 } 4262 4263 if (!add_nlist) { 4264 sym[sym_idx].Clear(); 4265 return true; 4266 } 4267 4268 uint64_t symbol_value = nlist.n_value; 4269 4270 if (symbol_name_non_abi_mangled) { 4271 sym[sym_idx].GetMangled().SetMangledName( 4272 ConstString(symbol_name_non_abi_mangled)); 4273 sym[sym_idx].GetMangled().SetDemangledName(ConstString(symbol_name)); 4274 } else { 4275 bool symbol_name_is_mangled = false; 4276 4277 if (symbol_name && symbol_name[0] == '_') { 4278 symbol_name_is_mangled = symbol_name[1] == '_'; 4279 symbol_name++; // Skip the leading underscore 4280 } 4281 4282 if (symbol_name) { 4283 ConstString const_symbol_name(symbol_name); 4284 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4285 symbol_name_is_mangled); 4286 } 4287 } 4288 4289 if (is_gsym) { 4290 const char *gsym_name = 4291 sym[sym_idx] 4292 .GetMangled() 4293 .GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled) 4294 .GetCString(); 4295 if (gsym_name) 4296 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4297 } 4298 4299 if (symbol_section) { 4300 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4301 if (symbol_byte_size == 0 && function_starts_count > 0) { 4302 addr_t symbol_lookup_file_addr = nlist.n_value; 4303 // Do an exact address match for non-ARM addresses, else get the 4304 // closest since the symbol might be a thumb symbol which has an 4305 // address with bit zero set. 4306 FunctionStarts::Entry *func_start_entry = 4307 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4308 if (is_arm && func_start_entry) { 4309 // Verify that the function start address is the symbol address 4310 // (ARM) or the symbol address + 1 (thumb). 4311 if (func_start_entry->addr != symbol_lookup_file_addr && 4312 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4313 // Not the right entry, NULL it out... 4314 func_start_entry = nullptr; 4315 } 4316 } 4317 if (func_start_entry) { 4318 func_start_entry->data = true; 4319 4320 addr_t symbol_file_addr = func_start_entry->addr; 4321 if (is_arm) 4322 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4323 4324 const FunctionStarts::Entry *next_func_start_entry = 4325 function_starts.FindNextEntry(func_start_entry); 4326 const addr_t section_end_file_addr = 4327 section_file_addr + symbol_section->GetByteSize(); 4328 if (next_func_start_entry) { 4329 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4330 // Be sure the clear the Thumb address bit when we calculate the 4331 // size from the current and next address 4332 if (is_arm) 4333 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4334 symbol_byte_size = std::min<lldb::addr_t>( 4335 next_symbol_file_addr - symbol_file_addr, 4336 section_end_file_addr - symbol_file_addr); 4337 } else { 4338 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4339 } 4340 } 4341 } 4342 symbol_value -= section_file_addr; 4343 } 4344 4345 if (!is_debug) { 4346 if (type == eSymbolTypeCode) { 4347 // See if we can find a N_FUN entry for any code symbols. If we do 4348 // find a match, and the name matches, then we can merge the two into 4349 // just the function symbol to avoid duplicate entries in the symbol 4350 // table. 4351 std::pair<ValueToSymbolIndexMap::const_iterator, 4352 ValueToSymbolIndexMap::const_iterator> 4353 range; 4354 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4355 if (range.first != range.second) { 4356 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4357 pos != range.second; ++pos) { 4358 if (sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, 4359 Mangled::ePreferMangled) == 4360 sym[pos->second].GetMangled().GetName( 4361 lldb::eLanguageTypeUnknown, Mangled::ePreferMangled)) { 4362 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4363 // We just need the flags from the linker symbol, so put these 4364 // flags into the N_FUN flags to avoid duplicate symbols in the 4365 // symbol table. 4366 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4367 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4368 if (resolver_addresses.find(nlist.n_value) != 4369 resolver_addresses.end()) 4370 sym[pos->second].SetType(eSymbolTypeResolver); 4371 sym[sym_idx].Clear(); 4372 return true; 4373 } 4374 } 4375 } else { 4376 if (resolver_addresses.find(nlist.n_value) != 4377 resolver_addresses.end()) 4378 type = eSymbolTypeResolver; 4379 } 4380 } else if (type == eSymbolTypeData || type == eSymbolTypeObjCClass || 4381 type == eSymbolTypeObjCMetaClass || 4382 type == eSymbolTypeObjCIVar) { 4383 // See if we can find a N_STSYM entry for any data symbols. If we do 4384 // find a match, and the name matches, then we can merge the two into 4385 // just the Static symbol to avoid duplicate entries in the symbol 4386 // table. 4387 std::pair<ValueToSymbolIndexMap::const_iterator, 4388 ValueToSymbolIndexMap::const_iterator> 4389 range; 4390 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4391 if (range.first != range.second) { 4392 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4393 pos != range.second; ++pos) { 4394 if (sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, 4395 Mangled::ePreferMangled) == 4396 sym[pos->second].GetMangled().GetName( 4397 lldb::eLanguageTypeUnknown, Mangled::ePreferMangled)) { 4398 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4399 // We just need the flags from the linker symbol, so put these 4400 // flags into the N_STSYM flags to avoid duplicate symbols in 4401 // the symbol table. 4402 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4403 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4404 sym[sym_idx].Clear(); 4405 return true; 4406 } 4407 } 4408 } else { 4409 // Combine N_GSYM stab entries with the non stab symbol. 4410 const char *gsym_name = sym[sym_idx] 4411 .GetMangled() 4412 .GetName(lldb::eLanguageTypeUnknown, 4413 Mangled::ePreferMangled) 4414 .GetCString(); 4415 if (gsym_name) { 4416 ConstNameToSymbolIndexMap::const_iterator pos = 4417 N_GSYM_name_to_sym_idx.find(gsym_name); 4418 if (pos != N_GSYM_name_to_sym_idx.end()) { 4419 const uint32_t GSYM_sym_idx = pos->second; 4420 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4421 // Copy the address, because often the N_GSYM address has an 4422 // invalid address of zero when the global is a common symbol. 4423 sym[GSYM_sym_idx].GetAddressRef().SetSection(symbol_section); 4424 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4425 // We just need the flags from the linker symbol, so put these 4426 // flags into the N_GSYM flags to avoid duplicate symbols in 4427 // the symbol table. 4428 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4429 sym[sym_idx].Clear(); 4430 return true; 4431 } 4432 } 4433 } 4434 } 4435 } 4436 4437 sym[sym_idx].SetID(nlist_idx); 4438 sym[sym_idx].SetType(type); 4439 if (set_value) { 4440 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4441 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4442 } 4443 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4444 if (nlist.n_desc & N_WEAK_REF) 4445 sym[sym_idx].SetIsWeak(true); 4446 4447 if (symbol_byte_size > 0) 4448 sym[sym_idx].SetByteSize(symbol_byte_size); 4449 4450 if (demangled_is_synthesized) 4451 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4452 4453 ++sym_idx; 4454 return true; 4455 }; 4456 4457 // First parse all the nlists but don't process them yet. See the next 4458 // comment for an explanation why. 4459 std::vector<struct nlist_64> nlists; 4460 nlists.reserve(symtab_load_command.nsyms); 4461 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 4462 if (auto nlist = 4463 ParseNList(nlist_data, nlist_data_offset, nlist_byte_size)) 4464 nlists.push_back(*nlist); 4465 else 4466 break; 4467 } 4468 4469 // Now parse all the debug symbols. This is needed to merge non-debug 4470 // symbols in the next step. Non-debug symbols are always coalesced into 4471 // the debug symbol. Doing this in one step would mean that some symbols 4472 // won't be merged. 4473 nlist_idx = 0; 4474 for (auto &nlist : nlists) { 4475 if (!ParseSymbolLambda(nlist, nlist_idx++, DebugSymbols)) 4476 break; 4477 } 4478 4479 // Finally parse all the non debug symbols. 4480 nlist_idx = 0; 4481 for (auto &nlist : nlists) { 4482 if (!ParseSymbolLambda(nlist, nlist_idx++, NonDebugSymbols)) 4483 break; 4484 } 4485 4486 for (const auto &pos : reexport_shlib_needs_fixup) { 4487 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4488 if (undef_pos != undefined_name_to_desc.end()) { 4489 const uint8_t dylib_ordinal = 4490 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4491 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4492 sym[pos.first].SetReExportedSymbolSharedLibrary( 4493 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4494 } 4495 } 4496 } 4497 4498 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4499 4500 if (function_starts_count > 0) { 4501 uint32_t num_synthetic_function_symbols = 0; 4502 for (i = 0; i < function_starts_count; ++i) { 4503 if (!function_starts.GetEntryRef(i).data) 4504 ++num_synthetic_function_symbols; 4505 } 4506 4507 if (num_synthetic_function_symbols > 0) { 4508 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4509 num_syms = sym_idx + num_synthetic_function_symbols; 4510 sym = symtab->Resize(num_syms); 4511 } 4512 for (i = 0; i < function_starts_count; ++i) { 4513 const FunctionStarts::Entry *func_start_entry = 4514 function_starts.GetEntryAtIndex(i); 4515 if (!func_start_entry->data) { 4516 addr_t symbol_file_addr = func_start_entry->addr; 4517 uint32_t symbol_flags = 0; 4518 if (is_arm) { 4519 if (symbol_file_addr & 1) 4520 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4521 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4522 } 4523 Address symbol_addr; 4524 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4525 SectionSP symbol_section(symbol_addr.GetSection()); 4526 uint32_t symbol_byte_size = 0; 4527 if (symbol_section) { 4528 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4529 const FunctionStarts::Entry *next_func_start_entry = 4530 function_starts.FindNextEntry(func_start_entry); 4531 const addr_t section_end_file_addr = 4532 section_file_addr + symbol_section->GetByteSize(); 4533 if (next_func_start_entry) { 4534 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4535 if (is_arm) 4536 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4537 symbol_byte_size = std::min<lldb::addr_t>( 4538 next_symbol_file_addr - symbol_file_addr, 4539 section_end_file_addr - symbol_file_addr); 4540 } else { 4541 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4542 } 4543 sym[sym_idx].SetID(synthetic_sym_id++); 4544 sym[sym_idx].GetMangled().SetDemangledName( 4545 GetNextSyntheticSymbolName()); 4546 sym[sym_idx].SetType(eSymbolTypeCode); 4547 sym[sym_idx].SetIsSynthetic(true); 4548 sym[sym_idx].GetAddressRef() = symbol_addr; 4549 if (symbol_flags) 4550 sym[sym_idx].SetFlags(symbol_flags); 4551 if (symbol_byte_size) 4552 sym[sym_idx].SetByteSize(symbol_byte_size); 4553 ++sym_idx; 4554 } 4555 } 4556 } 4557 } 4558 } 4559 } 4560 4561 // Trim our symbols down to just what we ended up with after removing any 4562 // symbols. 4563 if (sym_idx < num_syms) { 4564 num_syms = sym_idx; 4565 sym = symtab->Resize(num_syms); 4566 } 4567 4568 // Now synthesize indirect symbols 4569 if (m_dysymtab.nindirectsyms != 0) { 4570 if (indirect_symbol_index_data.GetByteSize()) { 4571 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4572 m_nlist_idx_to_sym_idx.end(); 4573 4574 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4575 ++sect_idx) { 4576 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4577 S_SYMBOL_STUBS) { 4578 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4579 if (symbol_stub_byte_size == 0) 4580 continue; 4581 4582 const uint32_t num_symbol_stubs = 4583 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4584 4585 if (num_symbol_stubs == 0) 4586 continue; 4587 4588 const uint32_t symbol_stub_index_offset = 4589 m_mach_sections[sect_idx].reserved1; 4590 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) { 4591 const uint32_t symbol_stub_index = 4592 symbol_stub_index_offset + stub_idx; 4593 const lldb::addr_t symbol_stub_addr = 4594 m_mach_sections[sect_idx].addr + 4595 (stub_idx * symbol_stub_byte_size); 4596 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4597 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4598 symbol_stub_offset, 4)) { 4599 const uint32_t stub_sym_id = 4600 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4601 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4602 continue; 4603 4604 NListIndexToSymbolIndexMap::const_iterator index_pos = 4605 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4606 Symbol *stub_symbol = nullptr; 4607 if (index_pos != end_index_pos) { 4608 // We have a remapping from the original nlist index to a 4609 // current symbol index, so just look this up by index 4610 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4611 } else { 4612 // We need to lookup a symbol using the original nlist symbol 4613 // index since this index is coming from the S_SYMBOL_STUBS 4614 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4615 } 4616 4617 if (stub_symbol) { 4618 Address so_addr(symbol_stub_addr, section_list); 4619 4620 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4621 // Change the external symbol into a trampoline that makes 4622 // sense These symbols were N_UNDF N_EXT, and are useless 4623 // to us, so we can re-use them so we don't have to make up 4624 // a synthetic symbol for no good reason. 4625 if (resolver_addresses.find(symbol_stub_addr) == 4626 resolver_addresses.end()) 4627 stub_symbol->SetType(eSymbolTypeTrampoline); 4628 else 4629 stub_symbol->SetType(eSymbolTypeResolver); 4630 stub_symbol->SetExternal(false); 4631 stub_symbol->GetAddressRef() = so_addr; 4632 stub_symbol->SetByteSize(symbol_stub_byte_size); 4633 } else { 4634 // Make a synthetic symbol to describe the trampoline stub 4635 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4636 if (sym_idx >= num_syms) { 4637 sym = symtab->Resize(++num_syms); 4638 stub_symbol = nullptr; // this pointer no longer valid 4639 } 4640 sym[sym_idx].SetID(synthetic_sym_id++); 4641 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4642 if (resolver_addresses.find(symbol_stub_addr) == 4643 resolver_addresses.end()) 4644 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4645 else 4646 sym[sym_idx].SetType(eSymbolTypeResolver); 4647 sym[sym_idx].SetIsSynthetic(true); 4648 sym[sym_idx].GetAddressRef() = so_addr; 4649 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4650 ++sym_idx; 4651 } 4652 } else { 4653 if (log) 4654 log->Warning("symbol stub referencing symbol table symbol " 4655 "%u that isn't in our minimal symbol table, " 4656 "fix this!!!", 4657 stub_sym_id); 4658 } 4659 } 4660 } 4661 } 4662 } 4663 } 4664 } 4665 4666 if (!trie_entries.empty()) { 4667 for (const auto &e : trie_entries) { 4668 if (e.entry.import_name) { 4669 // Only add indirect symbols from the Trie entries if we didn't have 4670 // a N_INDR nlist entry for this already 4671 if (indirect_symbol_names.find(e.entry.name) == 4672 indirect_symbol_names.end()) { 4673 // Make a synthetic symbol to describe re-exported symbol. 4674 if (sym_idx >= num_syms) 4675 sym = symtab->Resize(++num_syms); 4676 sym[sym_idx].SetID(synthetic_sym_id++); 4677 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4678 sym[sym_idx].SetType(eSymbolTypeReExported); 4679 sym[sym_idx].SetIsSynthetic(true); 4680 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4681 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4682 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4683 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4684 } 4685 ++sym_idx; 4686 } 4687 } 4688 } 4689 } 4690 4691 // StreamFile s(stdout, false); 4692 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4693 // symtab->Dump(&s, NULL, eSortOrderNone); 4694 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4695 // sizes 4696 symtab->CalculateSymbolSizes(); 4697 4698 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4699 // symtab->Dump(&s, NULL, eSortOrderNone); 4700 4701 return symtab->GetNumSymbols(); 4702 } 4703 4704 void ObjectFileMachO::Dump(Stream *s) { 4705 ModuleSP module_sp(GetModule()); 4706 if (module_sp) { 4707 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4708 s->Printf("%p: ", static_cast<void *>(this)); 4709 s->Indent(); 4710 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4711 s->PutCString("ObjectFileMachO64"); 4712 else 4713 s->PutCString("ObjectFileMachO32"); 4714 4715 *s << ", file = '" << m_file; 4716 ModuleSpecList all_specs; 4717 ModuleSpec base_spec; 4718 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 4719 base_spec, all_specs); 4720 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 4721 *s << "', triple"; 4722 if (e) 4723 s->Printf("[%d]", i); 4724 *s << " = "; 4725 *s << all_specs.GetModuleSpecRefAtIndex(i) 4726 .GetArchitecture() 4727 .GetTriple() 4728 .getTriple(); 4729 } 4730 *s << "\n"; 4731 SectionList *sections = GetSectionList(); 4732 if (sections) 4733 sections->Dump(s, nullptr, true, UINT32_MAX); 4734 4735 if (m_symtab_up) 4736 m_symtab_up->Dump(s, nullptr, eSortOrderNone); 4737 } 4738 } 4739 4740 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4741 const lldb_private::DataExtractor &data, 4742 lldb::offset_t lc_offset) { 4743 uint32_t i; 4744 struct uuid_command load_cmd; 4745 4746 lldb::offset_t offset = lc_offset; 4747 for (i = 0; i < header.ncmds; ++i) { 4748 const lldb::offset_t cmd_offset = offset; 4749 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4750 break; 4751 4752 if (load_cmd.cmd == LC_UUID) { 4753 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4754 4755 if (uuid_bytes) { 4756 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4757 // We pretend these object files have no UUID to prevent crashing. 4758 4759 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4760 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4761 0xbb, 0x14, 0xf0, 0x0d}; 4762 4763 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4764 return UUID(); 4765 4766 return UUID::fromOptionalData(uuid_bytes, 16); 4767 } 4768 return UUID(); 4769 } 4770 offset = cmd_offset + load_cmd.cmdsize; 4771 } 4772 return UUID(); 4773 } 4774 4775 static llvm::StringRef GetOSName(uint32_t cmd) { 4776 switch (cmd) { 4777 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4778 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4779 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4780 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4781 case llvm::MachO::LC_VERSION_MIN_TVOS: 4782 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4783 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4784 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4785 default: 4786 llvm_unreachable("unexpected LC_VERSION load command"); 4787 } 4788 } 4789 4790 namespace { 4791 struct OSEnv { 4792 llvm::StringRef os_type; 4793 llvm::StringRef environment; 4794 OSEnv(uint32_t cmd) { 4795 switch (cmd) { 4796 case llvm::MachO::PLATFORM_MACOS: 4797 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4798 return; 4799 case llvm::MachO::PLATFORM_IOS: 4800 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4801 return; 4802 case llvm::MachO::PLATFORM_TVOS: 4803 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4804 return; 4805 case llvm::MachO::PLATFORM_WATCHOS: 4806 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4807 return; 4808 // NEED_BRIDGEOS_TRIPLE case llvm::MachO::PLATFORM_BRIDGEOS: 4809 // NEED_BRIDGEOS_TRIPLE os_type = 4810 // llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4811 // NEED_BRIDGEOS_TRIPLE return; 4812 case llvm::MachO::PLATFORM_MACCATALYST: 4813 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4814 environment = llvm::Triple::getEnvironmentTypeName(llvm::Triple::MacABI); 4815 return; 4816 case llvm::MachO::PLATFORM_IOSSIMULATOR: 4817 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4818 environment = 4819 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4820 return; 4821 case llvm::MachO::PLATFORM_TVOSSIMULATOR: 4822 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4823 environment = 4824 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4825 return; 4826 case llvm::MachO::PLATFORM_WATCHOSSIMULATOR: 4827 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4828 environment = 4829 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4830 return; 4831 default: { 4832 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 4833 LIBLLDB_LOG_PROCESS)); 4834 LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION"); 4835 } 4836 } 4837 } 4838 }; 4839 4840 struct MinOS { 4841 uint32_t major_version, minor_version, patch_version; 4842 MinOS(uint32_t version) 4843 : major_version(version >> 16), minor_version((version >> 8) & 0xffu), 4844 patch_version(version & 0xffu) {} 4845 }; 4846 } // namespace 4847 4848 void ObjectFileMachO::GetAllArchSpecs(const llvm::MachO::mach_header &header, 4849 const lldb_private::DataExtractor &data, 4850 lldb::offset_t lc_offset, 4851 ModuleSpec &base_spec, 4852 lldb_private::ModuleSpecList &all_specs) { 4853 auto &base_arch = base_spec.GetArchitecture(); 4854 base_arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4855 if (!base_arch.IsValid()) 4856 return; 4857 4858 bool found_any = false; 4859 auto add_triple = [&](const llvm::Triple &triple) { 4860 auto spec = base_spec; 4861 spec.GetArchitecture().GetTriple() = triple; 4862 if (spec.GetArchitecture().IsValid()) { 4863 spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset); 4864 all_specs.Append(spec); 4865 found_any = true; 4866 } 4867 }; 4868 4869 // Set OS to an unspecified unknown or a "*" so it can match any OS 4870 llvm::Triple base_triple = base_arch.GetTriple(); 4871 base_triple.setOS(llvm::Triple::UnknownOS); 4872 base_triple.setOSName(llvm::StringRef()); 4873 4874 if (header.filetype == MH_PRELOAD) { 4875 if (header.cputype == CPU_TYPE_ARM) { 4876 // If this is a 32-bit arm binary, and it's a standalone binary, force 4877 // the Vendor to Apple so we don't accidentally pick up the generic 4878 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4879 // frame pointer register; most other armv7 ABIs use a combination of 4880 // r7 and r11. 4881 base_triple.setVendor(llvm::Triple::Apple); 4882 } else { 4883 // Set vendor to an unspecified unknown or a "*" so it can match any 4884 // vendor This is required for correct behavior of EFI debugging on 4885 // x86_64 4886 base_triple.setVendor(llvm::Triple::UnknownVendor); 4887 base_triple.setVendorName(llvm::StringRef()); 4888 } 4889 return add_triple(base_triple); 4890 } 4891 4892 struct load_command load_cmd; 4893 4894 // See if there is an LC_VERSION_MIN_* load command that can give 4895 // us the OS type. 4896 lldb::offset_t offset = lc_offset; 4897 for (uint32_t i = 0; i < header.ncmds; ++i) { 4898 const lldb::offset_t cmd_offset = offset; 4899 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4900 break; 4901 4902 struct version_min_command version_min; 4903 switch (load_cmd.cmd) { 4904 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4905 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4906 case llvm::MachO::LC_VERSION_MIN_TVOS: 4907 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 4908 if (load_cmd.cmdsize != sizeof(version_min)) 4909 break; 4910 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 4911 data.GetByteOrder(), &version_min) == 0) 4912 break; 4913 MinOS min_os(version_min.version); 4914 llvm::SmallString<32> os_name; 4915 llvm::raw_svector_ostream os(os_name); 4916 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 4917 << min_os.minor_version << '.' << min_os.patch_version; 4918 4919 auto triple = base_triple; 4920 triple.setOSName(os.str()); 4921 os_name.clear(); 4922 add_triple(triple); 4923 break; 4924 } 4925 default: 4926 break; 4927 } 4928 4929 offset = cmd_offset + load_cmd.cmdsize; 4930 } 4931 4932 // See if there are LC_BUILD_VERSION load commands that can give 4933 // us the OS type. 4934 offset = lc_offset; 4935 for (uint32_t i = 0; i < header.ncmds; ++i) { 4936 const lldb::offset_t cmd_offset = offset; 4937 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4938 break; 4939 4940 do { 4941 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 4942 struct build_version_command build_version; 4943 if (load_cmd.cmdsize < sizeof(build_version)) { 4944 // Malformed load command. 4945 break; 4946 } 4947 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 4948 data.GetByteOrder(), &build_version) == 0) 4949 break; 4950 MinOS min_os(build_version.minos); 4951 OSEnv os_env(build_version.platform); 4952 llvm::SmallString<16> os_name; 4953 llvm::raw_svector_ostream os(os_name); 4954 os << os_env.os_type << min_os.major_version << '.' 4955 << min_os.minor_version << '.' << min_os.patch_version; 4956 auto triple = base_triple; 4957 triple.setOSName(os.str()); 4958 os_name.clear(); 4959 if (!os_env.environment.empty()) 4960 triple.setEnvironmentName(os_env.environment); 4961 add_triple(triple); 4962 } 4963 } while (0); 4964 offset = cmd_offset + load_cmd.cmdsize; 4965 } 4966 4967 if (!found_any) { 4968 if (header.filetype == MH_KEXT_BUNDLE) { 4969 base_triple.setVendor(llvm::Triple::Apple); 4970 add_triple(base_triple); 4971 } else { 4972 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 4973 // so lets not say our Vendor is Apple, leave it as an unspecified 4974 // unknown. 4975 base_triple.setVendor(llvm::Triple::UnknownVendor); 4976 base_triple.setVendorName(llvm::StringRef()); 4977 add_triple(base_triple); 4978 } 4979 } 4980 } 4981 4982 ArchSpec ObjectFileMachO::GetArchitecture( 4983 ModuleSP module_sp, const llvm::MachO::mach_header &header, 4984 const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) { 4985 ModuleSpecList all_specs; 4986 ModuleSpec base_spec; 4987 GetAllArchSpecs(header, data, MachHeaderSizeFromMagic(header.magic), 4988 base_spec, all_specs); 4989 4990 // If the object file offers multiple alternative load commands, 4991 // pick the one that matches the module. 4992 if (module_sp) { 4993 const ArchSpec &module_arch = module_sp->GetArchitecture(); 4994 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 4995 ArchSpec mach_arch = 4996 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 4997 if (module_arch.IsCompatibleMatch(mach_arch)) 4998 return mach_arch; 4999 } 5000 } 5001 5002 // Return the first arch we found. 5003 if (all_specs.GetSize() == 0) 5004 return {}; 5005 return all_specs.GetModuleSpecRefAtIndex(0).GetArchitecture(); 5006 } 5007 5008 UUID ObjectFileMachO::GetUUID() { 5009 ModuleSP module_sp(GetModule()); 5010 if (module_sp) { 5011 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5012 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5013 return GetUUID(m_header, m_data, offset); 5014 } 5015 return UUID(); 5016 } 5017 5018 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5019 uint32_t count = 0; 5020 ModuleSP module_sp(GetModule()); 5021 if (module_sp) { 5022 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5023 struct load_command load_cmd; 5024 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5025 std::vector<std::string> rpath_paths; 5026 std::vector<std::string> rpath_relative_paths; 5027 std::vector<std::string> at_exec_relative_paths; 5028 uint32_t i; 5029 for (i = 0; i < m_header.ncmds; ++i) { 5030 const uint32_t cmd_offset = offset; 5031 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5032 break; 5033 5034 switch (load_cmd.cmd) { 5035 case LC_RPATH: 5036 case LC_LOAD_DYLIB: 5037 case LC_LOAD_WEAK_DYLIB: 5038 case LC_REEXPORT_DYLIB: 5039 case LC_LOAD_DYLINKER: 5040 case LC_LOADFVMLIB: 5041 case LC_LOAD_UPWARD_DYLIB: { 5042 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5043 const char *path = m_data.PeekCStr(name_offset); 5044 if (path) { 5045 if (load_cmd.cmd == LC_RPATH) 5046 rpath_paths.push_back(path); 5047 else { 5048 if (path[0] == '@') { 5049 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5050 rpath_relative_paths.push_back(path + strlen("@rpath")); 5051 else if (strncmp(path, "@executable_path", 5052 strlen("@executable_path")) == 0) 5053 at_exec_relative_paths.push_back(path + 5054 strlen("@executable_path")); 5055 } else { 5056 FileSpec file_spec(path); 5057 if (files.AppendIfUnique(file_spec)) 5058 count++; 5059 } 5060 } 5061 } 5062 } break; 5063 5064 default: 5065 break; 5066 } 5067 offset = cmd_offset + load_cmd.cmdsize; 5068 } 5069 5070 FileSpec this_file_spec(m_file); 5071 FileSystem::Instance().Resolve(this_file_spec); 5072 5073 if (!rpath_paths.empty()) { 5074 // Fixup all LC_RPATH values to be absolute paths 5075 std::string loader_path("@loader_path"); 5076 std::string executable_path("@executable_path"); 5077 for (auto &rpath : rpath_paths) { 5078 if (rpath.find(loader_path) == 0) { 5079 rpath.erase(0, loader_path.size()); 5080 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5081 } else if (rpath.find(executable_path) == 0) { 5082 rpath.erase(0, executable_path.size()); 5083 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5084 } 5085 } 5086 5087 for (const auto &rpath_relative_path : rpath_relative_paths) { 5088 for (const auto &rpath : rpath_paths) { 5089 std::string path = rpath; 5090 path += rpath_relative_path; 5091 // It is OK to resolve this path because we must find a file on disk 5092 // for us to accept it anyway if it is rpath relative. 5093 FileSpec file_spec(path); 5094 FileSystem::Instance().Resolve(file_spec); 5095 if (FileSystem::Instance().Exists(file_spec) && 5096 files.AppendIfUnique(file_spec)) { 5097 count++; 5098 break; 5099 } 5100 } 5101 } 5102 } 5103 5104 // We may have @executable_paths but no RPATHS. Figure those out here. 5105 // Only do this if this object file is the executable. We have no way to 5106 // get back to the actual executable otherwise, so we won't get the right 5107 // path. 5108 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5109 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5110 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5111 FileSpec file_spec = 5112 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5113 if (FileSystem::Instance().Exists(file_spec) && 5114 files.AppendIfUnique(file_spec)) 5115 count++; 5116 } 5117 } 5118 } 5119 return count; 5120 } 5121 5122 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5123 // If the object file is not an executable it can't hold the entry point. 5124 // m_entry_point_address is initialized to an invalid address, so we can just 5125 // return that. If m_entry_point_address is valid it means we've found it 5126 // already, so return the cached value. 5127 5128 if ((!IsExecutable() && !IsDynamicLoader()) || 5129 m_entry_point_address.IsValid()) { 5130 return m_entry_point_address; 5131 } 5132 5133 // Otherwise, look for the UnixThread or Thread command. The data for the 5134 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5135 // 5136 // uint32_t flavor - this is the flavor argument you would pass to 5137 // thread_get_state 5138 // uint32_t count - this is the count of longs in the thread state data 5139 // struct XXX_thread_state state - this is the structure from 5140 // <machine/thread_status.h> corresponding to the flavor. 5141 // <repeat this trio> 5142 // 5143 // So we just keep reading the various register flavors till we find the GPR 5144 // one, then read the PC out of there. 5145 // FIXME: We will need to have a "RegisterContext data provider" class at some 5146 // point that can get all the registers 5147 // out of data in this form & attach them to a given thread. That should 5148 // underlie the MacOS X User process plugin, and we'll also need it for the 5149 // MacOS X Core File process plugin. When we have that we can also use it 5150 // here. 5151 // 5152 // For now we hard-code the offsets and flavors we need: 5153 // 5154 // 5155 5156 ModuleSP module_sp(GetModule()); 5157 if (module_sp) { 5158 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5159 struct load_command load_cmd; 5160 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5161 uint32_t i; 5162 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5163 bool done = false; 5164 5165 for (i = 0; i < m_header.ncmds; ++i) { 5166 const lldb::offset_t cmd_offset = offset; 5167 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5168 break; 5169 5170 switch (load_cmd.cmd) { 5171 case LC_UNIXTHREAD: 5172 case LC_THREAD: { 5173 while (offset < cmd_offset + load_cmd.cmdsize) { 5174 uint32_t flavor = m_data.GetU32(&offset); 5175 uint32_t count = m_data.GetU32(&offset); 5176 if (count == 0) { 5177 // We've gotten off somehow, log and exit; 5178 return m_entry_point_address; 5179 } 5180 5181 switch (m_header.cputype) { 5182 case llvm::MachO::CPU_TYPE_ARM: 5183 if (flavor == 1 || 5184 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 5185 // from mach/arm/thread_status.h 5186 { 5187 offset += 60; // This is the offset of pc in the GPR thread state 5188 // data structure. 5189 start_address = m_data.GetU32(&offset); 5190 done = true; 5191 } 5192 break; 5193 case llvm::MachO::CPU_TYPE_ARM64: 5194 case llvm::MachO::CPU_TYPE_ARM64_32: 5195 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5196 { 5197 offset += 256; // This is the offset of pc in the GPR thread state 5198 // data structure. 5199 start_address = m_data.GetU64(&offset); 5200 done = true; 5201 } 5202 break; 5203 case llvm::MachO::CPU_TYPE_I386: 5204 if (flavor == 5205 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5206 { 5207 offset += 40; // This is the offset of eip in the GPR thread state 5208 // data structure. 5209 start_address = m_data.GetU32(&offset); 5210 done = true; 5211 } 5212 break; 5213 case llvm::MachO::CPU_TYPE_X86_64: 5214 if (flavor == 5215 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5216 { 5217 offset += 16 * 8; // This is the offset of rip in the GPR thread 5218 // state data structure. 5219 start_address = m_data.GetU64(&offset); 5220 done = true; 5221 } 5222 break; 5223 default: 5224 return m_entry_point_address; 5225 } 5226 // Haven't found the GPR flavor yet, skip over the data for this 5227 // flavor: 5228 if (done) 5229 break; 5230 offset += count * 4; 5231 } 5232 } break; 5233 case LC_MAIN: { 5234 ConstString text_segment_name("__TEXT"); 5235 uint64_t entryoffset = m_data.GetU64(&offset); 5236 SectionSP text_segment_sp = 5237 GetSectionList()->FindSectionByName(text_segment_name); 5238 if (text_segment_sp) { 5239 done = true; 5240 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5241 } 5242 } break; 5243 5244 default: 5245 break; 5246 } 5247 if (done) 5248 break; 5249 5250 // Go to the next load command: 5251 offset = cmd_offset + load_cmd.cmdsize; 5252 } 5253 5254 if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) { 5255 if (GetSymtab()) { 5256 Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType( 5257 ConstString("_dyld_start"), SymbolType::eSymbolTypeCode, 5258 Symtab::eDebugAny, Symtab::eVisibilityAny); 5259 if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) { 5260 start_address = dyld_start_sym->GetAddress().GetFileAddress(); 5261 } 5262 } 5263 } 5264 5265 if (start_address != LLDB_INVALID_ADDRESS) { 5266 // We got the start address from the load commands, so now resolve that 5267 // address in the sections of this ObjectFile: 5268 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5269 start_address, GetSectionList())) { 5270 m_entry_point_address.Clear(); 5271 } 5272 } else { 5273 // We couldn't read the UnixThread load command - maybe it wasn't there. 5274 // As a fallback look for the "start" symbol in the main executable. 5275 5276 ModuleSP module_sp(GetModule()); 5277 5278 if (module_sp) { 5279 SymbolContextList contexts; 5280 SymbolContext context; 5281 module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5282 eSymbolTypeCode, contexts); 5283 if (contexts.GetSize()) { 5284 if (contexts.GetContextAtIndex(0, context)) 5285 m_entry_point_address = context.symbol->GetAddress(); 5286 } 5287 } 5288 } 5289 } 5290 5291 return m_entry_point_address; 5292 } 5293 5294 lldb_private::Address ObjectFileMachO::GetBaseAddress() { 5295 lldb_private::Address header_addr; 5296 SectionList *section_list = GetSectionList(); 5297 if (section_list) { 5298 SectionSP text_segment_sp( 5299 section_list->FindSectionByName(GetSegmentNameTEXT())); 5300 if (text_segment_sp) { 5301 header_addr.SetSection(text_segment_sp); 5302 header_addr.SetOffset(0); 5303 } 5304 } 5305 return header_addr; 5306 } 5307 5308 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5309 ModuleSP module_sp(GetModule()); 5310 if (module_sp) { 5311 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5312 if (!m_thread_context_offsets_valid) { 5313 m_thread_context_offsets_valid = true; 5314 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5315 FileRangeArray::Entry file_range; 5316 thread_command thread_cmd; 5317 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5318 const uint32_t cmd_offset = offset; 5319 if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr) 5320 break; 5321 5322 if (thread_cmd.cmd == LC_THREAD) { 5323 file_range.SetRangeBase(offset); 5324 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5325 m_thread_context_offsets.Append(file_range); 5326 } 5327 offset = cmd_offset + thread_cmd.cmdsize; 5328 } 5329 } 5330 } 5331 return m_thread_context_offsets.GetSize(); 5332 } 5333 5334 std::string ObjectFileMachO::GetIdentifierString() { 5335 std::string result; 5336 ModuleSP module_sp(GetModule()); 5337 if (module_sp) { 5338 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5339 5340 // First, look over the load commands for an LC_NOTE load command with 5341 // data_owner string "kern ver str" & use that if found. 5342 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5343 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5344 const uint32_t cmd_offset = offset; 5345 load_command lc; 5346 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5347 break; 5348 if (lc.cmd == LC_NOTE) { 5349 char data_owner[17]; 5350 m_data.CopyData(offset, 16, data_owner); 5351 data_owner[16] = '\0'; 5352 offset += 16; 5353 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5354 uint64_t size = m_data.GetU64_unchecked(&offset); 5355 5356 // "kern ver str" has a uint32_t version and then a nul terminated 5357 // c-string. 5358 if (strcmp("kern ver str", data_owner) == 0) { 5359 offset = fileoff; 5360 uint32_t version; 5361 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5362 if (version == 1) { 5363 uint32_t strsize = size - sizeof(uint32_t); 5364 char *buf = (char *)malloc(strsize); 5365 if (buf) { 5366 m_data.CopyData(offset, strsize, buf); 5367 buf[strsize - 1] = '\0'; 5368 result = buf; 5369 if (buf) 5370 free(buf); 5371 return result; 5372 } 5373 } 5374 } 5375 } 5376 } 5377 offset = cmd_offset + lc.cmdsize; 5378 } 5379 5380 // Second, make a pass over the load commands looking for an obsolete 5381 // LC_IDENT load command. 5382 offset = MachHeaderSizeFromMagic(m_header.magic); 5383 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5384 const uint32_t cmd_offset = offset; 5385 struct ident_command ident_command; 5386 if (m_data.GetU32(&offset, &ident_command, 2) == nullptr) 5387 break; 5388 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5389 char *buf = (char *)malloc(ident_command.cmdsize); 5390 if (buf != nullptr && m_data.CopyData(offset, ident_command.cmdsize, 5391 buf) == ident_command.cmdsize) { 5392 buf[ident_command.cmdsize - 1] = '\0'; 5393 result = buf; 5394 } 5395 if (buf) 5396 free(buf); 5397 } 5398 offset = cmd_offset + ident_command.cmdsize; 5399 } 5400 } 5401 return result; 5402 } 5403 5404 bool ObjectFileMachO::GetCorefileMainBinaryInfo(addr_t &address, UUID &uuid) { 5405 address = LLDB_INVALID_ADDRESS; 5406 uuid.Clear(); 5407 ModuleSP module_sp(GetModule()); 5408 if (module_sp) { 5409 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5410 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5411 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5412 const uint32_t cmd_offset = offset; 5413 load_command lc; 5414 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5415 break; 5416 if (lc.cmd == LC_NOTE) { 5417 char data_owner[17]; 5418 memset(data_owner, 0, sizeof(data_owner)); 5419 m_data.CopyData(offset, 16, data_owner); 5420 offset += 16; 5421 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5422 uint64_t size = m_data.GetU64_unchecked(&offset); 5423 5424 // "main bin spec" (main binary specification) data payload is 5425 // formatted: 5426 // uint32_t version [currently 1] 5427 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user 5428 // process] uint64_t address [ UINT64_MAX if address not 5429 // specified ] uuid_t uuid [ all zero's if uuid not 5430 // specified ] uint32_t log2_pagesize [ process page size in log base 5431 // 2, e.g. 4k pages are 12. 0 for unspecified ] 5432 5433 if (strcmp("main bin spec", data_owner) == 0 && size >= 32) { 5434 offset = fileoff; 5435 uint32_t version; 5436 if (m_data.GetU32(&offset, &version, 1) != nullptr && version == 1) { 5437 uint32_t type = 0; 5438 uuid_t raw_uuid; 5439 memset(raw_uuid, 0, sizeof(uuid_t)); 5440 5441 if (m_data.GetU32(&offset, &type, 1) && 5442 m_data.GetU64(&offset, &address, 1) && 5443 m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5444 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5445 return true; 5446 } 5447 } 5448 } 5449 } 5450 offset = cmd_offset + lc.cmdsize; 5451 } 5452 } 5453 return false; 5454 } 5455 5456 lldb::RegisterContextSP 5457 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5458 lldb_private::Thread &thread) { 5459 lldb::RegisterContextSP reg_ctx_sp; 5460 5461 ModuleSP module_sp(GetModule()); 5462 if (module_sp) { 5463 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5464 if (!m_thread_context_offsets_valid) 5465 GetNumThreadContexts(); 5466 5467 const FileRangeArray::Entry *thread_context_file_range = 5468 m_thread_context_offsets.GetEntryAtIndex(idx); 5469 if (thread_context_file_range) { 5470 5471 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5472 thread_context_file_range->GetByteSize()); 5473 5474 switch (m_header.cputype) { 5475 case llvm::MachO::CPU_TYPE_ARM64: 5476 case llvm::MachO::CPU_TYPE_ARM64_32: 5477 reg_ctx_sp = 5478 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data); 5479 break; 5480 5481 case llvm::MachO::CPU_TYPE_ARM: 5482 reg_ctx_sp = 5483 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data); 5484 break; 5485 5486 case llvm::MachO::CPU_TYPE_I386: 5487 reg_ctx_sp = 5488 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data); 5489 break; 5490 5491 case llvm::MachO::CPU_TYPE_X86_64: 5492 reg_ctx_sp = 5493 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data); 5494 break; 5495 } 5496 } 5497 } 5498 return reg_ctx_sp; 5499 } 5500 5501 ObjectFile::Type ObjectFileMachO::CalculateType() { 5502 switch (m_header.filetype) { 5503 case MH_OBJECT: // 0x1u 5504 if (GetAddressByteSize() == 4) { 5505 // 32 bit kexts are just object files, but they do have a valid 5506 // UUID load command. 5507 if (GetUUID()) { 5508 // this checking for the UUID load command is not enough we could 5509 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5510 // this is required of kexts 5511 if (m_strata == eStrataInvalid) 5512 m_strata = eStrataKernel; 5513 return eTypeSharedLibrary; 5514 } 5515 } 5516 return eTypeObjectFile; 5517 5518 case MH_EXECUTE: 5519 return eTypeExecutable; // 0x2u 5520 case MH_FVMLIB: 5521 return eTypeSharedLibrary; // 0x3u 5522 case MH_CORE: 5523 return eTypeCoreFile; // 0x4u 5524 case MH_PRELOAD: 5525 return eTypeSharedLibrary; // 0x5u 5526 case MH_DYLIB: 5527 return eTypeSharedLibrary; // 0x6u 5528 case MH_DYLINKER: 5529 return eTypeDynamicLinker; // 0x7u 5530 case MH_BUNDLE: 5531 return eTypeSharedLibrary; // 0x8u 5532 case MH_DYLIB_STUB: 5533 return eTypeStubLibrary; // 0x9u 5534 case MH_DSYM: 5535 return eTypeDebugInfo; // 0xAu 5536 case MH_KEXT_BUNDLE: 5537 return eTypeSharedLibrary; // 0xBu 5538 default: 5539 break; 5540 } 5541 return eTypeUnknown; 5542 } 5543 5544 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5545 switch (m_header.filetype) { 5546 case MH_OBJECT: // 0x1u 5547 { 5548 // 32 bit kexts are just object files, but they do have a valid 5549 // UUID load command. 5550 if (GetUUID()) { 5551 // this checking for the UUID load command is not enough we could 5552 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5553 // this is required of kexts 5554 if (m_type == eTypeInvalid) 5555 m_type = eTypeSharedLibrary; 5556 5557 return eStrataKernel; 5558 } 5559 } 5560 return eStrataUnknown; 5561 5562 case MH_EXECUTE: // 0x2u 5563 // Check for the MH_DYLDLINK bit in the flags 5564 if (m_header.flags & MH_DYLDLINK) { 5565 return eStrataUser; 5566 } else { 5567 SectionList *section_list = GetSectionList(); 5568 if (section_list) { 5569 static ConstString g_kld_section_name("__KLD"); 5570 if (section_list->FindSectionByName(g_kld_section_name)) 5571 return eStrataKernel; 5572 } 5573 } 5574 return eStrataRawImage; 5575 5576 case MH_FVMLIB: 5577 return eStrataUser; // 0x3u 5578 case MH_CORE: 5579 return eStrataUnknown; // 0x4u 5580 case MH_PRELOAD: 5581 return eStrataRawImage; // 0x5u 5582 case MH_DYLIB: 5583 return eStrataUser; // 0x6u 5584 case MH_DYLINKER: 5585 return eStrataUser; // 0x7u 5586 case MH_BUNDLE: 5587 return eStrataUser; // 0x8u 5588 case MH_DYLIB_STUB: 5589 return eStrataUser; // 0x9u 5590 case MH_DSYM: 5591 return eStrataUnknown; // 0xAu 5592 case MH_KEXT_BUNDLE: 5593 return eStrataKernel; // 0xBu 5594 default: 5595 break; 5596 } 5597 return eStrataUnknown; 5598 } 5599 5600 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5601 ModuleSP module_sp(GetModule()); 5602 if (module_sp) { 5603 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5604 struct dylib_command load_cmd; 5605 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5606 uint32_t version_cmd = 0; 5607 uint64_t version = 0; 5608 uint32_t i; 5609 for (i = 0; i < m_header.ncmds; ++i) { 5610 const lldb::offset_t cmd_offset = offset; 5611 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5612 break; 5613 5614 if (load_cmd.cmd == LC_ID_DYLIB) { 5615 if (version_cmd == 0) { 5616 version_cmd = load_cmd.cmd; 5617 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr) 5618 break; 5619 version = load_cmd.dylib.current_version; 5620 } 5621 break; // Break for now unless there is another more complete version 5622 // number load command in the future. 5623 } 5624 offset = cmd_offset + load_cmd.cmdsize; 5625 } 5626 5627 if (version_cmd == LC_ID_DYLIB) { 5628 unsigned major = (version & 0xFFFF0000ull) >> 16; 5629 unsigned minor = (version & 0x0000FF00ull) >> 8; 5630 unsigned subminor = (version & 0x000000FFull); 5631 return llvm::VersionTuple(major, minor, subminor); 5632 } 5633 } 5634 return llvm::VersionTuple(); 5635 } 5636 5637 ArchSpec ObjectFileMachO::GetArchitecture() { 5638 ModuleSP module_sp(GetModule()); 5639 ArchSpec arch; 5640 if (module_sp) { 5641 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5642 5643 return GetArchitecture(module_sp, m_header, m_data, 5644 MachHeaderSizeFromMagic(m_header.magic)); 5645 } 5646 return arch; 5647 } 5648 5649 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, 5650 addr_t &base_addr, UUID &uuid) { 5651 uuid.Clear(); 5652 base_addr = LLDB_INVALID_ADDRESS; 5653 if (process && process->GetDynamicLoader()) { 5654 DynamicLoader *dl = process->GetDynamicLoader(); 5655 LazyBool using_shared_cache; 5656 LazyBool private_shared_cache; 5657 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5658 private_shared_cache); 5659 } 5660 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5661 LIBLLDB_LOG_PROCESS)); 5662 LLDB_LOGF( 5663 log, 5664 "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64, 5665 uuid.GetAsString().c_str(), base_addr); 5666 } 5667 5668 // From dyld SPI header dyld_process_info.h 5669 typedef void *dyld_process_info; 5670 struct lldb_copy__dyld_process_cache_info { 5671 uuid_t cacheUUID; // UUID of cache used by process 5672 uint64_t cacheBaseAddress; // load address of dyld shared cache 5673 bool noCache; // process is running without a dyld cache 5674 bool privateCache; // process is using a private copy of its dyld cache 5675 }; 5676 5677 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with 5678 // llvm enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile 5679 // errors. So we need to use the actual underlying types of task_t and 5680 // kern_return_t below. 5681 extern "C" unsigned int /*task_t*/ mach_task_self(); 5682 5683 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5684 uuid.Clear(); 5685 base_addr = LLDB_INVALID_ADDRESS; 5686 5687 #if defined(__APPLE__) && \ 5688 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5689 uint8_t *(*dyld_get_all_image_infos)(void); 5690 dyld_get_all_image_infos = 5691 (uint8_t * (*)()) dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5692 if (dyld_get_all_image_infos) { 5693 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5694 if (dyld_all_image_infos_address) { 5695 uint32_t *version = (uint32_t *) 5696 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5697 if (*version >= 13) { 5698 uuid_t *sharedCacheUUID_address = 0; 5699 int wordsize = sizeof(uint8_t *); 5700 if (wordsize == 8) { 5701 sharedCacheUUID_address = 5702 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5703 160); // sharedCacheUUID <mach-o/dyld_images.h> 5704 if (*version >= 15) 5705 base_addr = 5706 *(uint64_t 5707 *)((uint8_t *)dyld_all_image_infos_address + 5708 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5709 } else { 5710 sharedCacheUUID_address = 5711 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5712 84); // sharedCacheUUID <mach-o/dyld_images.h> 5713 if (*version >= 15) { 5714 base_addr = 0; 5715 base_addr = 5716 *(uint32_t 5717 *)((uint8_t *)dyld_all_image_infos_address + 5718 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5719 } 5720 } 5721 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5722 } 5723 } 5724 } else { 5725 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5726 dyld_process_info (*dyld_process_info_create)( 5727 unsigned int /* task_t */ task, uint64_t timestamp, 5728 unsigned int /*kern_return_t*/ *kernelError); 5729 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5730 void (*dyld_process_info_release)(dyld_process_info info); 5731 5732 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, 5733 unsigned int /*kern_return_t*/ *)) 5734 dlsym(RTLD_DEFAULT, "_dyld_process_info_create"); 5735 dyld_process_info_get_cache = (void (*)(void *, void *))dlsym( 5736 RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5737 dyld_process_info_release = 5738 (void (*)(void *))dlsym(RTLD_DEFAULT, "_dyld_process_info_release"); 5739 5740 if (dyld_process_info_create && dyld_process_info_get_cache) { 5741 unsigned int /*kern_return_t */ kern_ret; 5742 dyld_process_info process_info = 5743 dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5744 if (process_info) { 5745 struct lldb_copy__dyld_process_cache_info sc_info; 5746 memset(&sc_info, 0, sizeof(struct lldb_copy__dyld_process_cache_info)); 5747 dyld_process_info_get_cache(process_info, &sc_info); 5748 if (sc_info.cacheBaseAddress != 0) { 5749 base_addr = sc_info.cacheBaseAddress; 5750 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 5751 } 5752 dyld_process_info_release(process_info); 5753 } 5754 } 5755 } 5756 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5757 LIBLLDB_LOG_PROCESS)); 5758 if (log && uuid.IsValid()) 5759 LLDB_LOGF(log, 5760 "lldb's in-memory shared cache has a UUID of %s base address of " 5761 "0x%" PRIx64, 5762 uuid.GetAsString().c_str(), base_addr); 5763 #endif 5764 } 5765 5766 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5767 if (!m_min_os_version) { 5768 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5769 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5770 const lldb::offset_t load_cmd_offset = offset; 5771 5772 version_min_command lc; 5773 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5774 break; 5775 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5776 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5777 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5778 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5779 if (m_data.GetU32(&offset, &lc.version, 5780 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5781 const uint32_t xxxx = lc.version >> 16; 5782 const uint32_t yy = (lc.version >> 8) & 0xffu; 5783 const uint32_t zz = lc.version & 0xffu; 5784 if (xxxx) { 5785 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5786 break; 5787 } 5788 } 5789 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5790 // struct build_version_command { 5791 // uint32_t cmd; /* LC_BUILD_VERSION */ 5792 // uint32_t cmdsize; /* sizeof(struct 5793 // build_version_command) plus */ 5794 // /* ntools * sizeof(struct 5795 // build_tool_version) */ 5796 // uint32_t platform; /* platform */ 5797 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5798 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded in 5799 // nibbles xxxx.yy.zz */ uint32_t ntools; /* number of 5800 // tool entries following this */ 5801 // }; 5802 5803 offset += 4; // skip platform 5804 uint32_t minos = m_data.GetU32(&offset); 5805 5806 const uint32_t xxxx = minos >> 16; 5807 const uint32_t yy = (minos >> 8) & 0xffu; 5808 const uint32_t zz = minos & 0xffu; 5809 if (xxxx) { 5810 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5811 break; 5812 } 5813 } 5814 5815 offset = load_cmd_offset + lc.cmdsize; 5816 } 5817 5818 if (!m_min_os_version) { 5819 // Set version to an empty value so we don't keep trying to 5820 m_min_os_version = llvm::VersionTuple(); 5821 } 5822 } 5823 5824 return *m_min_os_version; 5825 } 5826 5827 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { 5828 if (!m_sdk_versions.hasValue()) { 5829 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5830 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5831 const lldb::offset_t load_cmd_offset = offset; 5832 5833 version_min_command lc; 5834 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5835 break; 5836 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5837 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5838 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5839 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5840 if (m_data.GetU32(&offset, &lc.version, 5841 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5842 const uint32_t xxxx = lc.sdk >> 16; 5843 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5844 const uint32_t zz = lc.sdk & 0xffu; 5845 if (xxxx) { 5846 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 5847 break; 5848 } else { 5849 GetModule()->ReportWarning("minimum OS version load command with " 5850 "invalid (0) version found."); 5851 } 5852 } 5853 } 5854 offset = load_cmd_offset + lc.cmdsize; 5855 } 5856 5857 if (!m_sdk_versions.hasValue()) { 5858 offset = MachHeaderSizeFromMagic(m_header.magic); 5859 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5860 const lldb::offset_t load_cmd_offset = offset; 5861 5862 version_min_command lc; 5863 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5864 break; 5865 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5866 // struct build_version_command { 5867 // uint32_t cmd; /* LC_BUILD_VERSION */ 5868 // uint32_t cmdsize; /* sizeof(struct 5869 // build_version_command) plus */ 5870 // /* ntools * sizeof(struct 5871 // build_tool_version) */ 5872 // uint32_t platform; /* platform */ 5873 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5874 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded 5875 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number 5876 // of tool entries following this */ 5877 // }; 5878 5879 offset += 4; // skip platform 5880 uint32_t minos = m_data.GetU32(&offset); 5881 5882 const uint32_t xxxx = minos >> 16; 5883 const uint32_t yy = (minos >> 8) & 0xffu; 5884 const uint32_t zz = minos & 0xffu; 5885 if (xxxx) { 5886 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 5887 break; 5888 } 5889 } 5890 offset = load_cmd_offset + lc.cmdsize; 5891 } 5892 } 5893 5894 if (!m_sdk_versions.hasValue()) 5895 m_sdk_versions = llvm::VersionTuple(); 5896 } 5897 5898 return m_sdk_versions.getValue(); 5899 } 5900 5901 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5902 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5903 } 5904 5905 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5906 return m_allow_assembly_emulation_unwind_plans; 5907 } 5908 5909 // PluginInterface protocol 5910 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5911 return GetPluginNameStatic(); 5912 } 5913 5914 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5915 5916 Section *ObjectFileMachO::GetMachHeaderSection() { 5917 // Find the first address of the mach header which is the first non-zero file 5918 // sized section whose file offset is zero. This is the base file address of 5919 // the mach-o file which can be subtracted from the vmaddr of the other 5920 // segments found in memory and added to the load address 5921 ModuleSP module_sp = GetModule(); 5922 if (!module_sp) 5923 return nullptr; 5924 SectionList *section_list = GetSectionList(); 5925 if (!section_list) 5926 return nullptr; 5927 const size_t num_sections = section_list->GetSize(); 5928 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5929 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5930 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 5931 return section; 5932 } 5933 return nullptr; 5934 } 5935 5936 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 5937 if (!section) 5938 return false; 5939 const bool is_dsym = (m_header.filetype == MH_DSYM); 5940 if (section->GetFileSize() == 0 && !is_dsym) 5941 return false; 5942 if (section->IsThreadSpecific()) 5943 return false; 5944 if (GetModule().get() != section->GetModule().get()) 5945 return false; 5946 // Be careful with __LINKEDIT and __DWARF segments 5947 if (section->GetName() == GetSegmentNameLINKEDIT() || 5948 section->GetName() == GetSegmentNameDWARF()) { 5949 // Only map __LINKEDIT and __DWARF if we have an in memory image and 5950 // this isn't a kernel binary like a kext or mach_kernel. 5951 const bool is_memory_image = (bool)m_process_wp.lock(); 5952 const Strata strata = GetStrata(); 5953 if (is_memory_image == false || strata == eStrataKernel) 5954 return false; 5955 } 5956 return true; 5957 } 5958 5959 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 5960 lldb::addr_t header_load_address, const Section *header_section, 5961 const Section *section) { 5962 ModuleSP module_sp = GetModule(); 5963 if (module_sp && header_section && section && 5964 header_load_address != LLDB_INVALID_ADDRESS) { 5965 lldb::addr_t file_addr = header_section->GetFileAddress(); 5966 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 5967 return section->GetFileAddress() - file_addr + header_load_address; 5968 } 5969 return LLDB_INVALID_ADDRESS; 5970 } 5971 5972 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 5973 bool value_is_offset) { 5974 ModuleSP module_sp = GetModule(); 5975 if (!module_sp) 5976 return false; 5977 5978 SectionList *section_list = GetSectionList(); 5979 if (!section_list) 5980 return false; 5981 5982 size_t num_loaded_sections = 0; 5983 const size_t num_sections = section_list->GetSize(); 5984 5985 if (value_is_offset) { 5986 // "value" is an offset to apply to each top level segment 5987 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5988 // Iterate through the object file sections to find all of the 5989 // sections that size on disk (to avoid __PAGEZERO) and load them 5990 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5991 if (SectionIsLoadable(section_sp.get())) 5992 if (target.GetSectionLoadList().SetSectionLoadAddress( 5993 section_sp, section_sp->GetFileAddress() + value)) 5994 ++num_loaded_sections; 5995 } 5996 } else { 5997 // "value" is the new base address of the mach_header, adjust each 5998 // section accordingly 5999 6000 Section *mach_header_section = GetMachHeaderSection(); 6001 if (mach_header_section) { 6002 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6003 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6004 6005 lldb::addr_t section_load_addr = 6006 CalculateSectionLoadAddressForMemoryImage( 6007 value, mach_header_section, section_sp.get()); 6008 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6009 if (target.GetSectionLoadList().SetSectionLoadAddress( 6010 section_sp, section_load_addr)) 6011 ++num_loaded_sections; 6012 } 6013 } 6014 } 6015 } 6016 return num_loaded_sections > 0; 6017 } 6018 6019 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6020 const FileSpec &outfile, Status &error) { 6021 if (!process_sp) 6022 return false; 6023 6024 Target &target = process_sp->GetTarget(); 6025 const ArchSpec target_arch = target.GetArchitecture(); 6026 const llvm::Triple &target_triple = target_arch.GetTriple(); 6027 if (target_triple.getVendor() == llvm::Triple::Apple && 6028 (target_triple.getOS() == llvm::Triple::MacOSX || 6029 target_triple.getOS() == llvm::Triple::IOS || 6030 target_triple.getOS() == llvm::Triple::WatchOS || 6031 target_triple.getOS() == llvm::Triple::TvOS)) { 6032 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) 6033 // { 6034 bool make_core = false; 6035 switch (target_arch.GetMachine()) { 6036 case llvm::Triple::aarch64: 6037 case llvm::Triple::aarch64_32: 6038 case llvm::Triple::arm: 6039 case llvm::Triple::thumb: 6040 case llvm::Triple::x86: 6041 case llvm::Triple::x86_64: 6042 make_core = true; 6043 break; 6044 default: 6045 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6046 target_triple.str().c_str()); 6047 break; 6048 } 6049 6050 if (make_core) { 6051 std::vector<segment_command_64> segment_load_commands; 6052 // uint32_t range_info_idx = 0; 6053 MemoryRegionInfo range_info; 6054 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6055 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6056 const ByteOrder byte_order = target_arch.GetByteOrder(); 6057 if (range_error.Success()) { 6058 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6059 const addr_t addr = range_info.GetRange().GetRangeBase(); 6060 const addr_t size = range_info.GetRange().GetByteSize(); 6061 6062 if (size == 0) 6063 break; 6064 6065 // Calculate correct protections 6066 uint32_t prot = 0; 6067 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6068 prot |= VM_PROT_READ; 6069 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6070 prot |= VM_PROT_WRITE; 6071 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6072 prot |= VM_PROT_EXECUTE; 6073 6074 if (prot != 0) { 6075 uint32_t cmd_type = LC_SEGMENT_64; 6076 uint32_t segment_size = sizeof(segment_command_64); 6077 if (addr_byte_size == 4) { 6078 cmd_type = LC_SEGMENT; 6079 segment_size = sizeof(segment_command); 6080 } 6081 segment_command_64 segment = { 6082 cmd_type, // uint32_t cmd; 6083 segment_size, // uint32_t cmdsize; 6084 {0}, // char segname[16]; 6085 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6086 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6087 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6088 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6089 prot, // uint32_t maxprot; 6090 prot, // uint32_t initprot; 6091 0, // uint32_t nsects; 6092 0}; // uint32_t flags; 6093 segment_load_commands.push_back(segment); 6094 } else { 6095 // No protections and a size of 1 used to be returned from old 6096 // debugservers when we asked about a region that was past the 6097 // last memory region and it indicates the end... 6098 if (size == 1) 6099 break; 6100 } 6101 6102 range_error = process_sp->GetMemoryRegionInfo( 6103 range_info.GetRange().GetRangeEnd(), range_info); 6104 if (range_error.Fail()) 6105 break; 6106 } 6107 6108 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6109 6110 mach_header_64 mach_header; 6111 if (addr_byte_size == 8) { 6112 mach_header.magic = MH_MAGIC_64; 6113 } else { 6114 mach_header.magic = MH_MAGIC; 6115 } 6116 mach_header.cputype = target_arch.GetMachOCPUType(); 6117 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6118 mach_header.filetype = MH_CORE; 6119 mach_header.ncmds = segment_load_commands.size(); 6120 mach_header.flags = 0; 6121 mach_header.reserved = 0; 6122 ThreadList &thread_list = process_sp->GetThreadList(); 6123 const uint32_t num_threads = thread_list.GetSize(); 6124 6125 // Make an array of LC_THREAD data items. Each one contains the 6126 // contents of the LC_THREAD load command. The data doesn't contain 6127 // the load command + load command size, we will add the load command 6128 // and load command size as we emit the data. 6129 std::vector<StreamString> LC_THREAD_datas(num_threads); 6130 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6131 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6132 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6133 LC_THREAD_data.SetByteOrder(byte_order); 6134 } 6135 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) { 6136 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6137 if (thread_sp) { 6138 switch (mach_header.cputype) { 6139 case llvm::MachO::CPU_TYPE_ARM64: 6140 case llvm::MachO::CPU_TYPE_ARM64_32: 6141 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6142 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6143 break; 6144 6145 case llvm::MachO::CPU_TYPE_ARM: 6146 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6147 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6148 break; 6149 6150 case llvm::MachO::CPU_TYPE_I386: 6151 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6152 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6153 break; 6154 6155 case llvm::MachO::CPU_TYPE_X86_64: 6156 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6157 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6158 break; 6159 } 6160 } 6161 } 6162 6163 // The size of the load command is the size of the segments... 6164 if (addr_byte_size == 8) { 6165 mach_header.sizeofcmds = 6166 segment_load_commands.size() * sizeof(struct segment_command_64); 6167 } else { 6168 mach_header.sizeofcmds = 6169 segment_load_commands.size() * sizeof(struct segment_command); 6170 } 6171 6172 // and the size of all LC_THREAD load command 6173 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6174 ++mach_header.ncmds; 6175 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6176 } 6177 6178 // Write the mach header 6179 buffer.PutHex32(mach_header.magic); 6180 buffer.PutHex32(mach_header.cputype); 6181 buffer.PutHex32(mach_header.cpusubtype); 6182 buffer.PutHex32(mach_header.filetype); 6183 buffer.PutHex32(mach_header.ncmds); 6184 buffer.PutHex32(mach_header.sizeofcmds); 6185 buffer.PutHex32(mach_header.flags); 6186 if (addr_byte_size == 8) { 6187 buffer.PutHex32(mach_header.reserved); 6188 } 6189 6190 // Skip the mach header and all load commands and align to the next 6191 // 0x1000 byte boundary 6192 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6193 if (file_offset & 0x00000fff) { 6194 file_offset += 0x00001000ull; 6195 file_offset &= (~0x00001000ull + 1); 6196 } 6197 6198 for (auto &segment : segment_load_commands) { 6199 segment.fileoff = file_offset; 6200 file_offset += segment.filesize; 6201 } 6202 6203 // Write out all of the LC_THREAD load commands 6204 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6205 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6206 buffer.PutHex32(LC_THREAD); 6207 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6208 buffer.Write(LC_THREAD_data.GetString().data(), LC_THREAD_data_size); 6209 } 6210 6211 // Write out all of the segment load commands 6212 for (const auto &segment : segment_load_commands) { 6213 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6214 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6215 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6216 segment.cmd, segment.cmdsize, segment.vmaddr, 6217 segment.vmaddr + segment.vmsize, segment.fileoff, 6218 segment.filesize, segment.maxprot, segment.initprot, 6219 segment.nsects, segment.flags); 6220 6221 buffer.PutHex32(segment.cmd); 6222 buffer.PutHex32(segment.cmdsize); 6223 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6224 if (addr_byte_size == 8) { 6225 buffer.PutHex64(segment.vmaddr); 6226 buffer.PutHex64(segment.vmsize); 6227 buffer.PutHex64(segment.fileoff); 6228 buffer.PutHex64(segment.filesize); 6229 } else { 6230 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6231 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6232 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6233 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6234 } 6235 buffer.PutHex32(segment.maxprot); 6236 buffer.PutHex32(segment.initprot); 6237 buffer.PutHex32(segment.nsects); 6238 buffer.PutHex32(segment.flags); 6239 } 6240 6241 std::string core_file_path(outfile.GetPath()); 6242 auto core_file = FileSystem::Instance().Open( 6243 outfile, File::eOpenOptionWrite | File::eOpenOptionTruncate | 6244 File::eOpenOptionCanCreate); 6245 if (!core_file) { 6246 error = core_file.takeError(); 6247 } else { 6248 // Read 1 page at a time 6249 uint8_t bytes[0x1000]; 6250 // Write the mach header and load commands out to the core file 6251 size_t bytes_written = buffer.GetString().size(); 6252 error = 6253 core_file.get()->Write(buffer.GetString().data(), bytes_written); 6254 if (error.Success()) { 6255 // Now write the file data for all memory segments in the process 6256 for (const auto &segment : segment_load_commands) { 6257 if (core_file.get()->SeekFromStart(segment.fileoff) == -1) { 6258 error.SetErrorStringWithFormat( 6259 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6260 segment.fileoff, core_file_path.c_str()); 6261 break; 6262 } 6263 6264 printf("Saving %" PRId64 6265 " bytes of data for memory region at 0x%" PRIx64 "\n", 6266 segment.vmsize, segment.vmaddr); 6267 addr_t bytes_left = segment.vmsize; 6268 addr_t addr = segment.vmaddr; 6269 Status memory_read_error; 6270 while (bytes_left > 0 && error.Success()) { 6271 const size_t bytes_to_read = 6272 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6273 6274 // In a savecore setting, we don't really care about caching, 6275 // as the data is dumped and very likely never read again, 6276 // so we call ReadMemoryFromInferior to bypass it. 6277 const size_t bytes_read = process_sp->ReadMemoryFromInferior( 6278 addr, bytes, bytes_to_read, memory_read_error); 6279 6280 if (bytes_read == bytes_to_read) { 6281 size_t bytes_written = bytes_read; 6282 error = core_file.get()->Write(bytes, bytes_written); 6283 bytes_left -= bytes_read; 6284 addr += bytes_read; 6285 } else { 6286 // Some pages within regions are not readable, those should 6287 // be zero filled 6288 memset(bytes, 0, bytes_to_read); 6289 size_t bytes_written = bytes_to_read; 6290 error = core_file.get()->Write(bytes, bytes_written); 6291 bytes_left -= bytes_to_read; 6292 addr += bytes_to_read; 6293 } 6294 } 6295 } 6296 } 6297 } 6298 } else { 6299 error.SetErrorString( 6300 "process doesn't support getting memory region info"); 6301 } 6302 } 6303 return true; // This is the right plug to handle saving core files for 6304 // this process 6305 } 6306 return false; 6307 } 6308