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; 2542 GetArchitecture(header_arch); 2543 char dsc_path[PATH_MAX]; 2544 char dsc_path_development[PATH_MAX]; 2545 2546 snprintf( 2547 dsc_path, sizeof(dsc_path), "%s%s%s", 2548 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2549 */ 2550 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2551 header_arch.GetArchitectureName()); 2552 2553 snprintf( 2554 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2555 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2556 */ 2557 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2558 header_arch.GetArchitectureName(), ".development"); 2559 2560 FileSpec dsc_nondevelopment_filespec(dsc_path); 2561 FileSpec dsc_development_filespec(dsc_path_development); 2562 FileSpec dsc_filespec; 2563 2564 UUID dsc_uuid; 2565 UUID process_shared_cache_uuid; 2566 addr_t process_shared_cache_base_addr; 2567 2568 if (process) { 2569 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, 2570 process_shared_cache_uuid); 2571 } 2572 2573 // First see if we can find an exact match for the inferior process 2574 // shared cache UUID in the development or non-development shared caches 2575 // on disk. 2576 if (process_shared_cache_uuid.IsValid()) { 2577 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2578 UUID dsc_development_uuid = GetSharedCacheUUID( 2579 dsc_development_filespec, byte_order, addr_byte_size); 2580 if (dsc_development_uuid.IsValid() && 2581 dsc_development_uuid == process_shared_cache_uuid) { 2582 dsc_filespec = dsc_development_filespec; 2583 dsc_uuid = dsc_development_uuid; 2584 } 2585 } 2586 if (!dsc_uuid.IsValid() && 2587 FileSystem::Instance().Exists(dsc_nondevelopment_filespec)) { 2588 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2589 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2590 if (dsc_nondevelopment_uuid.IsValid() && 2591 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2592 dsc_filespec = dsc_nondevelopment_filespec; 2593 dsc_uuid = dsc_nondevelopment_uuid; 2594 } 2595 } 2596 } 2597 2598 // Failing a UUID match, prefer the development dyld_shared cache if both 2599 // are present. 2600 if (!FileSystem::Instance().Exists(dsc_filespec)) { 2601 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2602 dsc_filespec = dsc_development_filespec; 2603 } else { 2604 dsc_filespec = dsc_nondevelopment_filespec; 2605 } 2606 } 2607 2608 /* The dyld_cache_header has a pointer to the 2609 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2610 The dyld_cache_local_symbols_info structure gives us three things: 2611 1. The start and count of the nlist records in the dyld_shared_cache 2612 file 2613 2. The start and size of the strings for these nlist records 2614 3. The start and count of dyld_cache_local_symbols_entry entries 2615 2616 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2617 dyld shared cache. 2618 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2619 the dyld shared cache. 2620 The dyld_cache_local_symbols_entry also lists the start of this 2621 dylib/framework's nlist records 2622 and the count of how many nlist records there are for this 2623 dylib/framework. 2624 */ 2625 2626 // Process the dyld shared cache header to find the unmapped symbols 2627 2628 DataBufferSP dsc_data_sp = MapFileData( 2629 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2630 if (!dsc_uuid.IsValid()) { 2631 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2632 } 2633 if (dsc_data_sp) { 2634 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2635 2636 bool uuid_match = true; 2637 if (dsc_uuid.IsValid() && process) { 2638 if (process_shared_cache_uuid.IsValid() && 2639 dsc_uuid != process_shared_cache_uuid) { 2640 // The on-disk dyld_shared_cache file is not the same as the one in 2641 // this process' memory, don't use it. 2642 uuid_match = false; 2643 ModuleSP module_sp(GetModule()); 2644 if (module_sp) 2645 module_sp->ReportWarning("process shared cache does not match " 2646 "on-disk dyld_shared_cache file, some " 2647 "symbol names will be missing."); 2648 } 2649 } 2650 2651 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2652 2653 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2654 2655 // If the mappingOffset points to a location inside the header, we've 2656 // opened an old dyld shared cache, and should not proceed further. 2657 if (uuid_match && 2658 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2659 2660 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2661 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2662 mappingOffset); 2663 2664 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2665 byte_order, addr_byte_size); 2666 offset = 0; 2667 2668 // The File addresses (from the in-memory Mach-O load commands) for 2669 // the shared libraries in the shared library cache need to be 2670 // adjusted by an offset to match up with the dylibOffset identifying 2671 // field in the dyld_cache_local_symbol_entry's. This offset is 2672 // recorded in mapping_offset_value. 2673 const uint64_t mapping_offset_value = 2674 dsc_mapping_info_data.GetU64(&offset); 2675 2676 offset = 2677 offsetof(struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2678 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2679 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2680 2681 if (localSymbolsOffset && localSymbolsSize) { 2682 // Map the local symbols 2683 DataBufferSP dsc_local_symbols_data_sp = 2684 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2685 2686 if (dsc_local_symbols_data_sp) { 2687 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2688 byte_order, addr_byte_size); 2689 2690 offset = 0; 2691 2692 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 2693 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 2694 UndefinedNameToDescMap undefined_name_to_desc; 2695 SymbolIndexToName reexport_shlib_needs_fixup; 2696 2697 // Read the local_symbols_infos struct in one shot 2698 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2699 dsc_local_symbols_data.GetU32(&offset, 2700 &local_symbols_info.nlistOffset, 6); 2701 2702 SectionSP text_section_sp( 2703 section_list->FindSectionByName(GetSegmentNameTEXT())); 2704 2705 uint32_t header_file_offset = 2706 (text_section_sp->GetFileAddress() - mapping_offset_value); 2707 2708 offset = local_symbols_info.entriesOffset; 2709 for (uint32_t entry_index = 0; 2710 entry_index < local_symbols_info.entriesCount; entry_index++) { 2711 struct lldb_copy_dyld_cache_local_symbols_entry 2712 local_symbols_entry; 2713 local_symbols_entry.dylibOffset = 2714 dsc_local_symbols_data.GetU32(&offset); 2715 local_symbols_entry.nlistStartIndex = 2716 dsc_local_symbols_data.GetU32(&offset); 2717 local_symbols_entry.nlistCount = 2718 dsc_local_symbols_data.GetU32(&offset); 2719 2720 if (header_file_offset == local_symbols_entry.dylibOffset) { 2721 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2722 2723 // The normal nlist code cannot correctly size the Symbols 2724 // array, we need to allocate it here. 2725 sym = symtab->Resize( 2726 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2727 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2728 num_syms = symtab->GetNumSymbols(); 2729 2730 nlist_data_offset = 2731 local_symbols_info.nlistOffset + 2732 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2733 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2734 2735 for (uint32_t nlist_index = 0; 2736 nlist_index < local_symbols_entry.nlistCount; 2737 nlist_index++) { 2738 ///////////////////////////// 2739 { 2740 llvm::Optional<struct nlist_64> nlist_maybe = 2741 ParseNList(dsc_local_symbols_data, nlist_data_offset, 2742 nlist_byte_size); 2743 if (!nlist_maybe) 2744 break; 2745 struct nlist_64 nlist = *nlist_maybe; 2746 2747 SymbolType type = eSymbolTypeInvalid; 2748 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2749 string_table_offset + nlist.n_strx); 2750 2751 if (symbol_name == NULL) { 2752 // No symbol should be NULL, even the symbols with no 2753 // string values should have an offset zero which 2754 // points to an empty C-string 2755 Host::SystemLog( 2756 Host::eSystemLogError, 2757 "error: DSC unmapped local symbol[%u] has invalid " 2758 "string table offset 0x%x in %s, ignoring symbol\n", 2759 entry_index, nlist.n_strx, 2760 module_sp->GetFileSpec().GetPath().c_str()); 2761 continue; 2762 } 2763 if (symbol_name[0] == '\0') 2764 symbol_name = NULL; 2765 2766 const char *symbol_name_non_abi_mangled = NULL; 2767 2768 SectionSP symbol_section; 2769 uint32_t symbol_byte_size = 0; 2770 bool add_nlist = true; 2771 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2772 bool demangled_is_synthesized = false; 2773 bool is_gsym = false; 2774 bool set_value = true; 2775 2776 assert(sym_idx < num_syms); 2777 2778 sym[sym_idx].SetDebug(is_debug); 2779 2780 if (is_debug) { 2781 switch (nlist.n_type) { 2782 case N_GSYM: 2783 // global symbol: name,,NO_SECT,type,0 2784 // Sometimes the N_GSYM value contains the address. 2785 2786 // FIXME: In the .o files, we have a GSYM and a debug 2787 // symbol for all the ObjC data. They 2788 // have the same address, but we want to ensure that 2789 // we always find only the real symbol, 'cause we 2790 // don't currently correctly attribute the 2791 // GSYM one to the ObjCClass/Ivar/MetaClass 2792 // symbol type. This is a temporary hack to make 2793 // sure the ObjectiveC symbols get treated correctly. 2794 // To do this right, we should coalesce all the GSYM 2795 // & global symbols that have the same address. 2796 2797 is_gsym = true; 2798 sym[sym_idx].SetExternal(true); 2799 2800 if (symbol_name && symbol_name[0] == '_' && 2801 symbol_name[1] == 'O') { 2802 llvm::StringRef symbol_name_ref(symbol_name); 2803 if (symbol_name_ref.startswith( 2804 g_objc_v2_prefix_class)) { 2805 symbol_name_non_abi_mangled = symbol_name + 1; 2806 symbol_name = 2807 symbol_name + g_objc_v2_prefix_class.size(); 2808 type = eSymbolTypeObjCClass; 2809 demangled_is_synthesized = true; 2810 2811 } else if (symbol_name_ref.startswith( 2812 g_objc_v2_prefix_metaclass)) { 2813 symbol_name_non_abi_mangled = symbol_name + 1; 2814 symbol_name = 2815 symbol_name + g_objc_v2_prefix_metaclass.size(); 2816 type = eSymbolTypeObjCMetaClass; 2817 demangled_is_synthesized = true; 2818 } else if (symbol_name_ref.startswith( 2819 g_objc_v2_prefix_ivar)) { 2820 symbol_name_non_abi_mangled = symbol_name + 1; 2821 symbol_name = 2822 symbol_name + g_objc_v2_prefix_ivar.size(); 2823 type = eSymbolTypeObjCIVar; 2824 demangled_is_synthesized = true; 2825 } 2826 } else { 2827 if (nlist.n_value != 0) 2828 symbol_section = section_info.GetSection( 2829 nlist.n_sect, nlist.n_value); 2830 type = eSymbolTypeData; 2831 } 2832 break; 2833 2834 case N_FNAME: 2835 // procedure name (f77 kludge): name,,NO_SECT,0,0 2836 type = eSymbolTypeCompiler; 2837 break; 2838 2839 case N_FUN: 2840 // procedure: name,,n_sect,linenumber,address 2841 if (symbol_name) { 2842 type = eSymbolTypeCode; 2843 symbol_section = section_info.GetSection( 2844 nlist.n_sect, nlist.n_value); 2845 2846 N_FUN_addr_to_sym_idx.insert( 2847 std::make_pair(nlist.n_value, sym_idx)); 2848 // We use the current number of symbols in the 2849 // symbol table in lieu of using nlist_idx in case 2850 // we ever start trimming entries out 2851 N_FUN_indexes.push_back(sym_idx); 2852 } else { 2853 type = eSymbolTypeCompiler; 2854 2855 if (!N_FUN_indexes.empty()) { 2856 // Copy the size of the function into the 2857 // original 2858 // STAB entry so we don't have 2859 // to hunt for it later 2860 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2861 ->SetByteSize(nlist.n_value); 2862 N_FUN_indexes.pop_back(); 2863 // We don't really need the end function STAB as 2864 // it contains the size which we already placed 2865 // with the original symbol, so don't add it if 2866 // we want a minimal symbol table 2867 add_nlist = false; 2868 } 2869 } 2870 break; 2871 2872 case N_STSYM: 2873 // static symbol: name,,n_sect,type,address 2874 N_STSYM_addr_to_sym_idx.insert( 2875 std::make_pair(nlist.n_value, sym_idx)); 2876 symbol_section = section_info.GetSection(nlist.n_sect, 2877 nlist.n_value); 2878 if (symbol_name && symbol_name[0]) { 2879 type = ObjectFile::GetSymbolTypeFromName( 2880 symbol_name + 1, eSymbolTypeData); 2881 } 2882 break; 2883 2884 case N_LCSYM: 2885 // .lcomm symbol: name,,n_sect,type,address 2886 symbol_section = section_info.GetSection(nlist.n_sect, 2887 nlist.n_value); 2888 type = eSymbolTypeCommonBlock; 2889 break; 2890 2891 case N_BNSYM: 2892 // We use the current number of symbols in the symbol 2893 // table in lieu of using nlist_idx in case we ever 2894 // start trimming entries out Skip these if we want 2895 // minimal symbol tables 2896 add_nlist = false; 2897 break; 2898 2899 case N_ENSYM: 2900 // Set the size of the N_BNSYM to the terminating 2901 // index of this N_ENSYM so that we can always skip 2902 // the entire symbol if we need to navigate more 2903 // quickly at the source level when parsing STABS 2904 // Skip these if we want minimal symbol tables 2905 add_nlist = false; 2906 break; 2907 2908 case N_OPT: 2909 // emitted with gcc2_compiled and in gcc source 2910 type = eSymbolTypeCompiler; 2911 break; 2912 2913 case N_RSYM: 2914 // register sym: name,,NO_SECT,type,register 2915 type = eSymbolTypeVariable; 2916 break; 2917 2918 case N_SLINE: 2919 // src line: 0,,n_sect,linenumber,address 2920 symbol_section = section_info.GetSection(nlist.n_sect, 2921 nlist.n_value); 2922 type = eSymbolTypeLineEntry; 2923 break; 2924 2925 case N_SSYM: 2926 // structure elt: name,,NO_SECT,type,struct_offset 2927 type = eSymbolTypeVariableType; 2928 break; 2929 2930 case N_SO: 2931 // source file name 2932 type = eSymbolTypeSourceFile; 2933 if (symbol_name == NULL) { 2934 add_nlist = false; 2935 if (N_SO_index != UINT32_MAX) { 2936 // Set the size of the N_SO to the terminating 2937 // index of this N_SO so that we can always skip 2938 // the entire N_SO if we need to navigate more 2939 // quickly at the source level when parsing STABS 2940 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2941 symbol_ptr->SetByteSize(sym_idx); 2942 symbol_ptr->SetSizeIsSibling(true); 2943 } 2944 N_NSYM_indexes.clear(); 2945 N_INCL_indexes.clear(); 2946 N_BRAC_indexes.clear(); 2947 N_COMM_indexes.clear(); 2948 N_FUN_indexes.clear(); 2949 N_SO_index = UINT32_MAX; 2950 } else { 2951 // We use the current number of symbols in the 2952 // symbol table in lieu of using nlist_idx in case 2953 // we ever start trimming entries out 2954 const bool N_SO_has_full_path = symbol_name[0] == '/'; 2955 if (N_SO_has_full_path) { 2956 if ((N_SO_index == sym_idx - 1) && 2957 ((sym_idx - 1) < num_syms)) { 2958 // We have two consecutive N_SO entries where 2959 // the first contains a directory and the 2960 // second contains a full path. 2961 sym[sym_idx - 1].GetMangled().SetValue( 2962 ConstString(symbol_name), false); 2963 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2964 add_nlist = false; 2965 } else { 2966 // This is the first entry in a N_SO that 2967 // contains a directory or 2968 // a full path to the source file 2969 N_SO_index = sym_idx; 2970 } 2971 } else if ((N_SO_index == sym_idx - 1) && 2972 ((sym_idx - 1) < num_syms)) { 2973 // This is usually the second N_SO entry that 2974 // contains just the filename, so here we combine 2975 // it with the first one if we are minimizing the 2976 // symbol table 2977 const char *so_path = 2978 sym[sym_idx - 1] 2979 .GetMangled() 2980 .GetDemangledName( 2981 lldb::eLanguageTypeUnknown) 2982 .AsCString(); 2983 if (so_path && so_path[0]) { 2984 std::string full_so_path(so_path); 2985 const size_t double_slash_pos = 2986 full_so_path.find("//"); 2987 if (double_slash_pos != std::string::npos) { 2988 // The linker has been generating bad N_SO 2989 // entries with doubled up paths 2990 // in the format "%s%s" where the first 2991 // string in the DW_AT_comp_dir, and the 2992 // second is the directory for the source 2993 // file so you end up with a path that looks 2994 // like "/tmp/src//tmp/src/" 2995 FileSpec so_dir(so_path); 2996 if (!FileSystem::Instance().Exists(so_dir)) { 2997 so_dir.SetFile( 2998 &full_so_path[double_slash_pos + 1], 2999 FileSpec::Style::native); 3000 if (FileSystem::Instance().Exists(so_dir)) { 3001 // Trim off the incorrect path 3002 full_so_path.erase(0, double_slash_pos + 1); 3003 } 3004 } 3005 } 3006 if (*full_so_path.rbegin() != '/') 3007 full_so_path += '/'; 3008 full_so_path += symbol_name; 3009 sym[sym_idx - 1].GetMangled().SetValue( 3010 ConstString(full_so_path.c_str()), false); 3011 add_nlist = false; 3012 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3013 } 3014 } else { 3015 // This could be a relative path to a N_SO 3016 N_SO_index = sym_idx; 3017 } 3018 } 3019 break; 3020 3021 case N_OSO: 3022 // object file name: name,,0,0,st_mtime 3023 type = eSymbolTypeObjectFile; 3024 break; 3025 3026 case N_LSYM: 3027 // local sym: name,,NO_SECT,type,offset 3028 type = eSymbolTypeLocal; 3029 break; 3030 3031 // INCL scopes 3032 case N_BINCL: 3033 // include file beginning: name,,NO_SECT,0,sum We use 3034 // the current number of symbols in the symbol table 3035 // in lieu of using nlist_idx in case we ever start 3036 // trimming entries out 3037 N_INCL_indexes.push_back(sym_idx); 3038 type = eSymbolTypeScopeBegin; 3039 break; 3040 3041 case N_EINCL: 3042 // include file end: name,,NO_SECT,0,0 3043 // Set the size of the N_BINCL to the terminating 3044 // index of this N_EINCL so that we can always skip 3045 // the entire symbol if we need to navigate more 3046 // quickly at the source level when parsing STABS 3047 if (!N_INCL_indexes.empty()) { 3048 symbol_ptr = 3049 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3050 symbol_ptr->SetByteSize(sym_idx + 1); 3051 symbol_ptr->SetSizeIsSibling(true); 3052 N_INCL_indexes.pop_back(); 3053 } 3054 type = eSymbolTypeScopeEnd; 3055 break; 3056 3057 case N_SOL: 3058 // #included file name: name,,n_sect,0,address 3059 type = eSymbolTypeHeaderFile; 3060 3061 // We currently don't use the header files on darwin 3062 add_nlist = false; 3063 break; 3064 3065 case N_PARAMS: 3066 // compiler parameters: name,,NO_SECT,0,0 3067 type = eSymbolTypeCompiler; 3068 break; 3069 3070 case N_VERSION: 3071 // compiler version: name,,NO_SECT,0,0 3072 type = eSymbolTypeCompiler; 3073 break; 3074 3075 case N_OLEVEL: 3076 // compiler -O level: name,,NO_SECT,0,0 3077 type = eSymbolTypeCompiler; 3078 break; 3079 3080 case N_PSYM: 3081 // parameter: name,,NO_SECT,type,offset 3082 type = eSymbolTypeVariable; 3083 break; 3084 3085 case N_ENTRY: 3086 // alternate entry: name,,n_sect,linenumber,address 3087 symbol_section = section_info.GetSection(nlist.n_sect, 3088 nlist.n_value); 3089 type = eSymbolTypeLineEntry; 3090 break; 3091 3092 // Left and Right Braces 3093 case N_LBRAC: 3094 // left bracket: 0,,NO_SECT,nesting level,address We 3095 // use the current number of symbols in the symbol 3096 // table in lieu of using nlist_idx in case we ever 3097 // start trimming entries out 3098 symbol_section = section_info.GetSection(nlist.n_sect, 3099 nlist.n_value); 3100 N_BRAC_indexes.push_back(sym_idx); 3101 type = eSymbolTypeScopeBegin; 3102 break; 3103 3104 case N_RBRAC: 3105 // right bracket: 0,,NO_SECT,nesting level,address 3106 // Set the size of the N_LBRAC to the terminating 3107 // index of this N_RBRAC so that we can always skip 3108 // the entire symbol if we need to navigate more 3109 // quickly at the source level when parsing STABS 3110 symbol_section = section_info.GetSection(nlist.n_sect, 3111 nlist.n_value); 3112 if (!N_BRAC_indexes.empty()) { 3113 symbol_ptr = 3114 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3115 symbol_ptr->SetByteSize(sym_idx + 1); 3116 symbol_ptr->SetSizeIsSibling(true); 3117 N_BRAC_indexes.pop_back(); 3118 } 3119 type = eSymbolTypeScopeEnd; 3120 break; 3121 3122 case N_EXCL: 3123 // deleted include file: name,,NO_SECT,0,sum 3124 type = eSymbolTypeHeaderFile; 3125 break; 3126 3127 // COMM scopes 3128 case N_BCOMM: 3129 // begin common: name,,NO_SECT,0,0 3130 // We use the current number of symbols in the symbol 3131 // table in lieu of using nlist_idx in case we ever 3132 // start trimming entries out 3133 type = eSymbolTypeScopeBegin; 3134 N_COMM_indexes.push_back(sym_idx); 3135 break; 3136 3137 case N_ECOML: 3138 // end common (local name): 0,,n_sect,0,address 3139 symbol_section = section_info.GetSection(nlist.n_sect, 3140 nlist.n_value); 3141 // Fall through 3142 3143 case N_ECOMM: 3144 // end common: name,,n_sect,0,0 3145 // Set the size of the N_BCOMM to the terminating 3146 // index of this N_ECOMM/N_ECOML so that we can 3147 // always skip the entire symbol if we need to 3148 // navigate more quickly at the source level when 3149 // parsing STABS 3150 if (!N_COMM_indexes.empty()) { 3151 symbol_ptr = 3152 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3153 symbol_ptr->SetByteSize(sym_idx + 1); 3154 symbol_ptr->SetSizeIsSibling(true); 3155 N_COMM_indexes.pop_back(); 3156 } 3157 type = eSymbolTypeScopeEnd; 3158 break; 3159 3160 case N_LENG: 3161 // second stab entry with length information 3162 type = eSymbolTypeAdditional; 3163 break; 3164 3165 default: 3166 break; 3167 } 3168 } else { 3169 // uint8_t n_pext = N_PEXT & nlist.n_type; 3170 uint8_t n_type = N_TYPE & nlist.n_type; 3171 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3172 3173 switch (n_type) { 3174 case N_INDR: { 3175 const char *reexport_name_cstr = 3176 strtab_data.PeekCStr(nlist.n_value); 3177 if (reexport_name_cstr && reexport_name_cstr[0]) { 3178 type = eSymbolTypeReExported; 3179 ConstString reexport_name( 3180 reexport_name_cstr + 3181 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3182 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3183 set_value = false; 3184 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3185 indirect_symbol_names.insert(ConstString( 3186 symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3187 } else 3188 type = eSymbolTypeUndefined; 3189 } break; 3190 3191 case N_UNDF: 3192 if (symbol_name && symbol_name[0]) { 3193 ConstString undefined_name( 3194 symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3195 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3196 } 3197 // Fall through 3198 case N_PBUD: 3199 type = eSymbolTypeUndefined; 3200 break; 3201 3202 case N_ABS: 3203 type = eSymbolTypeAbsolute; 3204 break; 3205 3206 case N_SECT: { 3207 symbol_section = section_info.GetSection(nlist.n_sect, 3208 nlist.n_value); 3209 3210 if (symbol_section == NULL) { 3211 // TODO: warn about this? 3212 add_nlist = false; 3213 break; 3214 } 3215 3216 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3217 type = eSymbolTypeException; 3218 } else { 3219 uint32_t section_type = 3220 symbol_section->Get() & SECTION_TYPE; 3221 3222 switch (section_type) { 3223 case S_CSTRING_LITERALS: 3224 type = eSymbolTypeData; 3225 break; // section with only literal C strings 3226 case S_4BYTE_LITERALS: 3227 type = eSymbolTypeData; 3228 break; // section with only 4 byte literals 3229 case S_8BYTE_LITERALS: 3230 type = eSymbolTypeData; 3231 break; // section with only 8 byte literals 3232 case S_LITERAL_POINTERS: 3233 type = eSymbolTypeTrampoline; 3234 break; // section with only pointers to literals 3235 case S_NON_LAZY_SYMBOL_POINTERS: 3236 type = eSymbolTypeTrampoline; 3237 break; // section with only non-lazy symbol 3238 // pointers 3239 case S_LAZY_SYMBOL_POINTERS: 3240 type = eSymbolTypeTrampoline; 3241 break; // section with only lazy symbol pointers 3242 case S_SYMBOL_STUBS: 3243 type = eSymbolTypeTrampoline; 3244 break; // section with only symbol stubs, byte 3245 // size of stub in the reserved2 field 3246 case S_MOD_INIT_FUNC_POINTERS: 3247 type = eSymbolTypeCode; 3248 break; // section with only function pointers for 3249 // initialization 3250 case S_MOD_TERM_FUNC_POINTERS: 3251 type = eSymbolTypeCode; 3252 break; // section with only function pointers for 3253 // termination 3254 case S_INTERPOSING: 3255 type = eSymbolTypeTrampoline; 3256 break; // section with only pairs of function 3257 // pointers for interposing 3258 case S_16BYTE_LITERALS: 3259 type = eSymbolTypeData; 3260 break; // section with only 16 byte literals 3261 case S_DTRACE_DOF: 3262 type = eSymbolTypeInstrumentation; 3263 break; 3264 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3265 type = eSymbolTypeTrampoline; 3266 break; 3267 default: 3268 switch (symbol_section->GetType()) { 3269 case lldb::eSectionTypeCode: 3270 type = eSymbolTypeCode; 3271 break; 3272 case eSectionTypeData: 3273 case eSectionTypeDataCString: // Inlined C string 3274 // data 3275 case eSectionTypeDataCStringPointers: // Pointers 3276 // to C 3277 // string 3278 // data 3279 case eSectionTypeDataSymbolAddress: // Address of 3280 // a symbol in 3281 // the symbol 3282 // table 3283 case eSectionTypeData4: 3284 case eSectionTypeData8: 3285 case eSectionTypeData16: 3286 type = eSymbolTypeData; 3287 break; 3288 default: 3289 break; 3290 } 3291 break; 3292 } 3293 3294 if (type == eSymbolTypeInvalid) { 3295 const char *symbol_sect_name = 3296 symbol_section->GetName().AsCString(); 3297 if (symbol_section->IsDescendant( 3298 text_section_sp.get())) { 3299 if (symbol_section->IsClear( 3300 S_ATTR_PURE_INSTRUCTIONS | 3301 S_ATTR_SELF_MODIFYING_CODE | 3302 S_ATTR_SOME_INSTRUCTIONS)) 3303 type = eSymbolTypeData; 3304 else 3305 type = eSymbolTypeCode; 3306 } else if (symbol_section->IsDescendant( 3307 data_section_sp.get()) || 3308 symbol_section->IsDescendant( 3309 data_dirty_section_sp.get()) || 3310 symbol_section->IsDescendant( 3311 data_const_section_sp.get())) { 3312 if (symbol_sect_name && 3313 ::strstr(symbol_sect_name, "__objc") == 3314 symbol_sect_name) { 3315 type = eSymbolTypeRuntime; 3316 3317 if (symbol_name) { 3318 llvm::StringRef symbol_name_ref(symbol_name); 3319 if (symbol_name_ref.startswith("_OBJC_")) { 3320 llvm::StringRef 3321 g_objc_v2_prefix_class( 3322 "_OBJC_CLASS_$_"); 3323 llvm::StringRef 3324 g_objc_v2_prefix_metaclass( 3325 "_OBJC_METACLASS_$_"); 3326 llvm::StringRef 3327 g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 3328 if (symbol_name_ref.startswith( 3329 g_objc_v2_prefix_class)) { 3330 symbol_name_non_abi_mangled = 3331 symbol_name + 1; 3332 symbol_name = 3333 symbol_name + 3334 g_objc_v2_prefix_class.size(); 3335 type = eSymbolTypeObjCClass; 3336 demangled_is_synthesized = true; 3337 } else if ( 3338 symbol_name_ref.startswith( 3339 g_objc_v2_prefix_metaclass)) { 3340 symbol_name_non_abi_mangled = 3341 symbol_name + 1; 3342 symbol_name = 3343 symbol_name + 3344 g_objc_v2_prefix_metaclass.size(); 3345 type = eSymbolTypeObjCMetaClass; 3346 demangled_is_synthesized = true; 3347 } else if (symbol_name_ref.startswith( 3348 g_objc_v2_prefix_ivar)) { 3349 symbol_name_non_abi_mangled = 3350 symbol_name + 1; 3351 symbol_name = 3352 symbol_name + 3353 g_objc_v2_prefix_ivar.size(); 3354 type = eSymbolTypeObjCIVar; 3355 demangled_is_synthesized = true; 3356 } 3357 } 3358 } 3359 } else if (symbol_sect_name && 3360 ::strstr(symbol_sect_name, 3361 "__gcc_except_tab") == 3362 symbol_sect_name) { 3363 type = eSymbolTypeException; 3364 } else { 3365 type = eSymbolTypeData; 3366 } 3367 } else if (symbol_sect_name && 3368 ::strstr(symbol_sect_name, "__IMPORT") == 3369 symbol_sect_name) { 3370 type = eSymbolTypeTrampoline; 3371 } else if (symbol_section->IsDescendant( 3372 objc_section_sp.get())) { 3373 type = eSymbolTypeRuntime; 3374 if (symbol_name && symbol_name[0] == '.') { 3375 llvm::StringRef symbol_name_ref(symbol_name); 3376 llvm::StringRef 3377 g_objc_v1_prefix_class(".objc_class_name_"); 3378 if (symbol_name_ref.startswith( 3379 g_objc_v1_prefix_class)) { 3380 symbol_name_non_abi_mangled = symbol_name; 3381 symbol_name = symbol_name + 3382 g_objc_v1_prefix_class.size(); 3383 type = eSymbolTypeObjCClass; 3384 demangled_is_synthesized = true; 3385 } 3386 } 3387 } 3388 } 3389 } 3390 } break; 3391 } 3392 } 3393 3394 if (add_nlist) { 3395 uint64_t symbol_value = nlist.n_value; 3396 if (symbol_name_non_abi_mangled) { 3397 sym[sym_idx].GetMangled().SetMangledName( 3398 ConstString(symbol_name_non_abi_mangled)); 3399 sym[sym_idx].GetMangled().SetDemangledName( 3400 ConstString(symbol_name)); 3401 } else { 3402 bool symbol_name_is_mangled = false; 3403 3404 if (symbol_name && symbol_name[0] == '_') { 3405 symbol_name_is_mangled = symbol_name[1] == '_'; 3406 symbol_name++; // Skip the leading underscore 3407 } 3408 3409 if (symbol_name) { 3410 ConstString const_symbol_name(symbol_name); 3411 sym[sym_idx].GetMangled().SetValue( 3412 const_symbol_name, symbol_name_is_mangled); 3413 if (is_gsym && is_debug) { 3414 const char *gsym_name = 3415 sym[sym_idx] 3416 .GetMangled() 3417 .GetName(lldb::eLanguageTypeUnknown, 3418 Mangled::ePreferMangled) 3419 .GetCString(); 3420 if (gsym_name) 3421 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3422 } 3423 } 3424 } 3425 if (symbol_section) { 3426 const addr_t section_file_addr = 3427 symbol_section->GetFileAddress(); 3428 if (symbol_byte_size == 0 && 3429 function_starts_count > 0) { 3430 addr_t symbol_lookup_file_addr = nlist.n_value; 3431 // Do an exact address match for non-ARM addresses, 3432 // else get the closest since the symbol might be a 3433 // thumb symbol which has an address with bit zero 3434 // set 3435 FunctionStarts::Entry *func_start_entry = 3436 function_starts.FindEntry(symbol_lookup_file_addr, 3437 !is_arm); 3438 if (is_arm && func_start_entry) { 3439 // Verify that the function start address is the 3440 // symbol address (ARM) or the symbol address + 1 3441 // (thumb) 3442 if (func_start_entry->addr != 3443 symbol_lookup_file_addr && 3444 func_start_entry->addr != 3445 (symbol_lookup_file_addr + 1)) { 3446 // Not the right entry, NULL it out... 3447 func_start_entry = NULL; 3448 } 3449 } 3450 if (func_start_entry) { 3451 func_start_entry->data = true; 3452 3453 addr_t symbol_file_addr = func_start_entry->addr; 3454 uint32_t symbol_flags = 0; 3455 if (is_arm) { 3456 if (symbol_file_addr & 1) 3457 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3458 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3459 } 3460 3461 const FunctionStarts::Entry *next_func_start_entry = 3462 function_starts.FindNextEntry(func_start_entry); 3463 const addr_t section_end_file_addr = 3464 section_file_addr + 3465 symbol_section->GetByteSize(); 3466 if (next_func_start_entry) { 3467 addr_t next_symbol_file_addr = 3468 next_func_start_entry->addr; 3469 // Be sure the clear the Thumb address bit when 3470 // we calculate the size from the current and 3471 // next address 3472 if (is_arm) 3473 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3474 symbol_byte_size = std::min<lldb::addr_t>( 3475 next_symbol_file_addr - symbol_file_addr, 3476 section_end_file_addr - symbol_file_addr); 3477 } else { 3478 symbol_byte_size = 3479 section_end_file_addr - symbol_file_addr; 3480 } 3481 } 3482 } 3483 symbol_value -= section_file_addr; 3484 } 3485 3486 if (is_debug == false) { 3487 if (type == eSymbolTypeCode) { 3488 // See if we can find a N_FUN entry for any code 3489 // symbols. If we do find a match, and the name 3490 // matches, then we can merge the two into just the 3491 // function symbol to avoid duplicate entries in 3492 // the symbol table 3493 auto range = 3494 N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3495 if (range.first != range.second) { 3496 bool found_it = false; 3497 for (const auto pos = range.first; 3498 pos != range.second; ++pos) { 3499 if (sym[sym_idx].GetMangled().GetName( 3500 lldb::eLanguageTypeUnknown, 3501 Mangled::ePreferMangled) == 3502 sym[pos->second].GetMangled().GetName( 3503 lldb::eLanguageTypeUnknown, 3504 Mangled::ePreferMangled)) { 3505 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3506 // We just need the flags from the linker 3507 // symbol, so put these flags 3508 // into the N_FUN flags to avoid duplicate 3509 // symbols in the symbol table 3510 sym[pos->second].SetExternal( 3511 sym[sym_idx].IsExternal()); 3512 sym[pos->second].SetFlags(nlist.n_type << 16 | 3513 nlist.n_desc); 3514 if (resolver_addresses.find(nlist.n_value) != 3515 resolver_addresses.end()) 3516 sym[pos->second].SetType(eSymbolTypeResolver); 3517 sym[sym_idx].Clear(); 3518 found_it = true; 3519 break; 3520 } 3521 } 3522 if (found_it) 3523 continue; 3524 } else { 3525 if (resolver_addresses.find(nlist.n_value) != 3526 resolver_addresses.end()) 3527 type = eSymbolTypeResolver; 3528 } 3529 } else if (type == eSymbolTypeData || 3530 type == eSymbolTypeObjCClass || 3531 type == eSymbolTypeObjCMetaClass || 3532 type == eSymbolTypeObjCIVar) { 3533 // See if we can find a N_STSYM entry for any data 3534 // symbols. If we do find a match, and the name 3535 // matches, then we can merge the two into just the 3536 // Static symbol to avoid duplicate entries in the 3537 // symbol table 3538 auto range = N_STSYM_addr_to_sym_idx.equal_range( 3539 nlist.n_value); 3540 if (range.first != range.second) { 3541 bool found_it = false; 3542 for (const auto pos = range.first; 3543 pos != range.second; ++pos) { 3544 if (sym[sym_idx].GetMangled().GetName( 3545 lldb::eLanguageTypeUnknown, 3546 Mangled::ePreferMangled) == 3547 sym[pos->second].GetMangled().GetName( 3548 lldb::eLanguageTypeUnknown, 3549 Mangled::ePreferMangled)) { 3550 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3551 // We just need the flags from the linker 3552 // symbol, so put these flags 3553 // into the N_STSYM flags to avoid duplicate 3554 // symbols in the symbol table 3555 sym[pos->second].SetExternal( 3556 sym[sym_idx].IsExternal()); 3557 sym[pos->second].SetFlags(nlist.n_type << 16 | 3558 nlist.n_desc); 3559 sym[sym_idx].Clear(); 3560 found_it = true; 3561 break; 3562 } 3563 } 3564 if (found_it) 3565 continue; 3566 } else { 3567 const char *gsym_name = 3568 sym[sym_idx] 3569 .GetMangled() 3570 .GetName(lldb::eLanguageTypeUnknown, 3571 Mangled::ePreferMangled) 3572 .GetCString(); 3573 if (gsym_name) { 3574 // Combine N_GSYM stab entries with the non 3575 // stab symbol 3576 ConstNameToSymbolIndexMap::const_iterator pos = 3577 N_GSYM_name_to_sym_idx.find(gsym_name); 3578 if (pos != N_GSYM_name_to_sym_idx.end()) { 3579 const uint32_t GSYM_sym_idx = pos->second; 3580 m_nlist_idx_to_sym_idx[nlist_idx] = 3581 GSYM_sym_idx; 3582 // Copy the address, because often the N_GSYM 3583 // address has an invalid address of zero 3584 // when the global is a common symbol 3585 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3586 symbol_section); 3587 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3588 symbol_value); 3589 // We just need the flags from the linker 3590 // symbol, so put these flags 3591 // into the N_GSYM flags to avoid duplicate 3592 // symbols in the symbol table 3593 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 3594 nlist.n_desc); 3595 sym[sym_idx].Clear(); 3596 continue; 3597 } 3598 } 3599 } 3600 } 3601 } 3602 3603 sym[sym_idx].SetID(nlist_idx); 3604 sym[sym_idx].SetType(type); 3605 if (set_value) { 3606 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 3607 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3608 } 3609 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 3610 3611 if (symbol_byte_size > 0) 3612 sym[sym_idx].SetByteSize(symbol_byte_size); 3613 3614 if (demangled_is_synthesized) 3615 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3616 ++sym_idx; 3617 } else { 3618 sym[sym_idx].Clear(); 3619 } 3620 } 3621 ///////////////////////////// 3622 } 3623 break; // No more entries to consider 3624 } 3625 } 3626 3627 for (const auto &pos : reexport_shlib_needs_fixup) { 3628 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3629 if (undef_pos != undefined_name_to_desc.end()) { 3630 const uint8_t dylib_ordinal = 3631 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3632 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3633 sym[pos.first].SetReExportedSymbolSharedLibrary( 3634 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3635 } 3636 } 3637 } 3638 } 3639 } 3640 } 3641 } 3642 3643 // Must reset this in case it was mutated above! 3644 nlist_data_offset = 0; 3645 #endif 3646 3647 if (nlist_data.GetByteSize() > 0) { 3648 3649 // If the sym array was not created while parsing the DSC unmapped 3650 // symbols, create it now. 3651 if (sym == nullptr) { 3652 sym = 3653 symtab->Resize(symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3654 num_syms = symtab->GetNumSymbols(); 3655 } 3656 3657 if (unmapped_local_symbols_found) { 3658 assert(m_dysymtab.ilocalsym == 0); 3659 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3660 nlist_idx = m_dysymtab.nlocalsym; 3661 } else { 3662 nlist_idx = 0; 3663 } 3664 3665 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 3666 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 3667 UndefinedNameToDescMap undefined_name_to_desc; 3668 SymbolIndexToName reexport_shlib_needs_fixup; 3669 3670 // Symtab parsing is a huge mess. Everything is entangled and the code 3671 // requires access to a ridiculous amount of variables. LLDB depends 3672 // heavily on the proper merging of symbols and to get that right we need 3673 // to make sure we have parsed all the debug symbols first. Therefore we 3674 // invoke the lambda twice, once to parse only the debug symbols and then 3675 // once more to parse the remaining symbols. 3676 auto ParseSymbolLambda = [&](struct nlist_64 &nlist, uint32_t nlist_idx, 3677 bool debug_only) { 3678 const bool is_debug = ((nlist.n_type & N_STAB) != 0); 3679 if (is_debug != debug_only) 3680 return true; 3681 3682 const char *symbol_name_non_abi_mangled = nullptr; 3683 const char *symbol_name = nullptr; 3684 3685 if (have_strtab_data) { 3686 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3687 3688 if (symbol_name == nullptr) { 3689 // No symbol should be NULL, even the symbols with no string values 3690 // should have an offset zero which points to an empty C-string 3691 Host::SystemLog(Host::eSystemLogError, 3692 "error: symbol[%u] has invalid string table offset " 3693 "0x%x in %s, ignoring symbol\n", 3694 nlist_idx, nlist.n_strx, 3695 module_sp->GetFileSpec().GetPath().c_str()); 3696 return true; 3697 } 3698 if (symbol_name[0] == '\0') 3699 symbol_name = nullptr; 3700 } else { 3701 const addr_t str_addr = strtab_addr + nlist.n_strx; 3702 Status str_error; 3703 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3704 str_error)) 3705 symbol_name = memory_symbol_name.c_str(); 3706 } 3707 3708 SymbolType type = eSymbolTypeInvalid; 3709 SectionSP symbol_section; 3710 lldb::addr_t symbol_byte_size = 0; 3711 bool add_nlist = true; 3712 bool is_gsym = false; 3713 bool demangled_is_synthesized = false; 3714 bool set_value = true; 3715 3716 assert(sym_idx < num_syms); 3717 sym[sym_idx].SetDebug(is_debug); 3718 3719 if (is_debug) { 3720 switch (nlist.n_type) { 3721 case N_GSYM: 3722 // global symbol: name,,NO_SECT,type,0 3723 // Sometimes the N_GSYM value contains the address. 3724 3725 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3726 // the ObjC data. They 3727 // have the same address, but we want to ensure that we always find 3728 // only the real symbol, 'cause we don't currently correctly 3729 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3730 // type. This is a temporary hack to make sure the ObjectiveC 3731 // symbols get treated correctly. To do this right, we should 3732 // coalesce all the GSYM & global symbols that have the same 3733 // address. 3734 is_gsym = true; 3735 sym[sym_idx].SetExternal(true); 3736 3737 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3738 llvm::StringRef symbol_name_ref(symbol_name); 3739 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3740 symbol_name_non_abi_mangled = symbol_name + 1; 3741 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3742 type = eSymbolTypeObjCClass; 3743 demangled_is_synthesized = true; 3744 3745 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 3746 symbol_name_non_abi_mangled = symbol_name + 1; 3747 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3748 type = eSymbolTypeObjCMetaClass; 3749 demangled_is_synthesized = true; 3750 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3751 symbol_name_non_abi_mangled = symbol_name + 1; 3752 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3753 type = eSymbolTypeObjCIVar; 3754 demangled_is_synthesized = true; 3755 } 3756 } else { 3757 if (nlist.n_value != 0) 3758 symbol_section = 3759 section_info.GetSection(nlist.n_sect, nlist.n_value); 3760 type = eSymbolTypeData; 3761 } 3762 break; 3763 3764 case N_FNAME: 3765 // procedure name (f77 kludge): name,,NO_SECT,0,0 3766 type = eSymbolTypeCompiler; 3767 break; 3768 3769 case N_FUN: 3770 // procedure: name,,n_sect,linenumber,address 3771 if (symbol_name) { 3772 type = eSymbolTypeCode; 3773 symbol_section = 3774 section_info.GetSection(nlist.n_sect, nlist.n_value); 3775 3776 N_FUN_addr_to_sym_idx.insert( 3777 std::make_pair(nlist.n_value, sym_idx)); 3778 // We use the current number of symbols in the symbol table in 3779 // lieu of using nlist_idx in case we ever start trimming entries 3780 // out 3781 N_FUN_indexes.push_back(sym_idx); 3782 } else { 3783 type = eSymbolTypeCompiler; 3784 3785 if (!N_FUN_indexes.empty()) { 3786 // Copy the size of the function into the original STAB entry 3787 // so we don't have to hunt for it later 3788 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3789 ->SetByteSize(nlist.n_value); 3790 N_FUN_indexes.pop_back(); 3791 // We don't really need the end function STAB as it contains 3792 // the size which we already placed with the original symbol, 3793 // so don't add it if we want a minimal symbol table 3794 add_nlist = false; 3795 } 3796 } 3797 break; 3798 3799 case N_STSYM: 3800 // static symbol: name,,n_sect,type,address 3801 N_STSYM_addr_to_sym_idx.insert( 3802 std::make_pair(nlist.n_value, sym_idx)); 3803 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3804 if (symbol_name && symbol_name[0]) { 3805 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3806 eSymbolTypeData); 3807 } 3808 break; 3809 3810 case N_LCSYM: 3811 // .lcomm symbol: name,,n_sect,type,address 3812 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3813 type = eSymbolTypeCommonBlock; 3814 break; 3815 3816 case N_BNSYM: 3817 // We use the current number of symbols in the symbol table in lieu 3818 // of using nlist_idx in case we ever start trimming entries out 3819 // Skip these if we want minimal symbol tables 3820 add_nlist = false; 3821 break; 3822 3823 case N_ENSYM: 3824 // Set the size of the N_BNSYM to the terminating index of this 3825 // N_ENSYM so that we can always skip the entire symbol if we need 3826 // to navigate more quickly at the source level when parsing STABS 3827 // Skip these if we want minimal symbol tables 3828 add_nlist = false; 3829 break; 3830 3831 case N_OPT: 3832 // emitted with gcc2_compiled and in gcc source 3833 type = eSymbolTypeCompiler; 3834 break; 3835 3836 case N_RSYM: 3837 // register sym: name,,NO_SECT,type,register 3838 type = eSymbolTypeVariable; 3839 break; 3840 3841 case N_SLINE: 3842 // src line: 0,,n_sect,linenumber,address 3843 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3844 type = eSymbolTypeLineEntry; 3845 break; 3846 3847 case N_SSYM: 3848 // structure elt: name,,NO_SECT,type,struct_offset 3849 type = eSymbolTypeVariableType; 3850 break; 3851 3852 case N_SO: 3853 // source file name 3854 type = eSymbolTypeSourceFile; 3855 if (symbol_name == nullptr) { 3856 add_nlist = false; 3857 if (N_SO_index != UINT32_MAX) { 3858 // Set the size of the N_SO to the terminating index of this 3859 // N_SO so that we can always skip the entire N_SO if we need 3860 // to navigate more quickly at the source level when parsing 3861 // STABS 3862 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3863 symbol_ptr->SetByteSize(sym_idx); 3864 symbol_ptr->SetSizeIsSibling(true); 3865 } 3866 N_NSYM_indexes.clear(); 3867 N_INCL_indexes.clear(); 3868 N_BRAC_indexes.clear(); 3869 N_COMM_indexes.clear(); 3870 N_FUN_indexes.clear(); 3871 N_SO_index = UINT32_MAX; 3872 } else { 3873 // We use the current number of symbols in the symbol table in 3874 // lieu of using nlist_idx in case we ever start trimming entries 3875 // out 3876 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3877 if (N_SO_has_full_path) { 3878 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3879 // We have two consecutive N_SO entries where the first 3880 // contains a directory and the second contains a full path. 3881 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), 3882 false); 3883 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3884 add_nlist = false; 3885 } else { 3886 // This is the first entry in a N_SO that contains a 3887 // directory or a full path to the source file 3888 N_SO_index = sym_idx; 3889 } 3890 } else if ((N_SO_index == sym_idx - 1) && 3891 ((sym_idx - 1) < num_syms)) { 3892 // This is usually the second N_SO entry that contains just the 3893 // filename, so here we combine it with the first one if we are 3894 // minimizing the symbol table 3895 const char *so_path = 3896 sym[sym_idx - 1] 3897 .GetMangled() 3898 .GetDemangledName(lldb::eLanguageTypeUnknown) 3899 .AsCString(); 3900 if (so_path && so_path[0]) { 3901 std::string full_so_path(so_path); 3902 const size_t double_slash_pos = full_so_path.find("//"); 3903 if (double_slash_pos != std::string::npos) { 3904 // The linker has been generating bad N_SO entries with 3905 // doubled up paths in the format "%s%s" where the first 3906 // string in the DW_AT_comp_dir, and the second is the 3907 // directory for the source file so you end up with a path 3908 // that looks like "/tmp/src//tmp/src/" 3909 FileSpec so_dir(so_path); 3910 if (!FileSystem::Instance().Exists(so_dir)) { 3911 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 3912 FileSpec::Style::native); 3913 if (FileSystem::Instance().Exists(so_dir)) { 3914 // Trim off the incorrect path 3915 full_so_path.erase(0, double_slash_pos + 1); 3916 } 3917 } 3918 } 3919 if (*full_so_path.rbegin() != '/') 3920 full_so_path += '/'; 3921 full_so_path += symbol_name; 3922 sym[sym_idx - 1].GetMangled().SetValue( 3923 ConstString(full_so_path.c_str()), false); 3924 add_nlist = false; 3925 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3926 } 3927 } else { 3928 // This could be a relative path to a N_SO 3929 N_SO_index = sym_idx; 3930 } 3931 } 3932 break; 3933 3934 case N_OSO: 3935 // object file name: name,,0,0,st_mtime 3936 type = eSymbolTypeObjectFile; 3937 break; 3938 3939 case N_LSYM: 3940 // local sym: name,,NO_SECT,type,offset 3941 type = eSymbolTypeLocal; 3942 break; 3943 3944 // INCL scopes 3945 case N_BINCL: 3946 // include file beginning: name,,NO_SECT,0,sum We use the current 3947 // number of symbols in the symbol table in lieu of using nlist_idx 3948 // in case we ever start trimming entries out 3949 N_INCL_indexes.push_back(sym_idx); 3950 type = eSymbolTypeScopeBegin; 3951 break; 3952 3953 case N_EINCL: 3954 // include file end: name,,NO_SECT,0,0 3955 // Set the size of the N_BINCL to the terminating index of this 3956 // N_EINCL so that we can always skip the entire symbol if we need 3957 // to navigate more quickly at the source level when parsing STABS 3958 if (!N_INCL_indexes.empty()) { 3959 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3960 symbol_ptr->SetByteSize(sym_idx + 1); 3961 symbol_ptr->SetSizeIsSibling(true); 3962 N_INCL_indexes.pop_back(); 3963 } 3964 type = eSymbolTypeScopeEnd; 3965 break; 3966 3967 case N_SOL: 3968 // #included file name: name,,n_sect,0,address 3969 type = eSymbolTypeHeaderFile; 3970 3971 // We currently don't use the header files on darwin 3972 add_nlist = false; 3973 break; 3974 3975 case N_PARAMS: 3976 // compiler parameters: name,,NO_SECT,0,0 3977 type = eSymbolTypeCompiler; 3978 break; 3979 3980 case N_VERSION: 3981 // compiler version: name,,NO_SECT,0,0 3982 type = eSymbolTypeCompiler; 3983 break; 3984 3985 case N_OLEVEL: 3986 // compiler -O level: name,,NO_SECT,0,0 3987 type = eSymbolTypeCompiler; 3988 break; 3989 3990 case N_PSYM: 3991 // parameter: name,,NO_SECT,type,offset 3992 type = eSymbolTypeVariable; 3993 break; 3994 3995 case N_ENTRY: 3996 // alternate entry: name,,n_sect,linenumber,address 3997 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3998 type = eSymbolTypeLineEntry; 3999 break; 4000 4001 // Left and Right Braces 4002 case N_LBRAC: 4003 // left bracket: 0,,NO_SECT,nesting level,address We use the 4004 // current number of symbols in the symbol table in lieu of using 4005 // nlist_idx in case we ever start trimming entries out 4006 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4007 N_BRAC_indexes.push_back(sym_idx); 4008 type = eSymbolTypeScopeBegin; 4009 break; 4010 4011 case N_RBRAC: 4012 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4013 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4014 // can always skip the entire symbol if we need to navigate more 4015 // quickly at the source level when parsing STABS 4016 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4017 if (!N_BRAC_indexes.empty()) { 4018 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4019 symbol_ptr->SetByteSize(sym_idx + 1); 4020 symbol_ptr->SetSizeIsSibling(true); 4021 N_BRAC_indexes.pop_back(); 4022 } 4023 type = eSymbolTypeScopeEnd; 4024 break; 4025 4026 case N_EXCL: 4027 // deleted include file: name,,NO_SECT,0,sum 4028 type = eSymbolTypeHeaderFile; 4029 break; 4030 4031 // COMM scopes 4032 case N_BCOMM: 4033 // begin common: name,,NO_SECT,0,0 4034 // We use the current number of symbols in the symbol table in lieu 4035 // of using nlist_idx in case we ever start trimming entries out 4036 type = eSymbolTypeScopeBegin; 4037 N_COMM_indexes.push_back(sym_idx); 4038 break; 4039 4040 case N_ECOML: 4041 // end common (local name): 0,,n_sect,0,address 4042 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4043 LLVM_FALLTHROUGH; 4044 4045 case N_ECOMM: 4046 // end common: name,,n_sect,0,0 4047 // Set the size of the N_BCOMM to the terminating index of this 4048 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4049 // we need to navigate more quickly at the source level when 4050 // parsing STABS 4051 if (!N_COMM_indexes.empty()) { 4052 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4053 symbol_ptr->SetByteSize(sym_idx + 1); 4054 symbol_ptr->SetSizeIsSibling(true); 4055 N_COMM_indexes.pop_back(); 4056 } 4057 type = eSymbolTypeScopeEnd; 4058 break; 4059 4060 case N_LENG: 4061 // second stab entry with length information 4062 type = eSymbolTypeAdditional; 4063 break; 4064 4065 default: 4066 break; 4067 } 4068 } else { 4069 uint8_t n_type = N_TYPE & nlist.n_type; 4070 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4071 4072 switch (n_type) { 4073 case N_INDR: { 4074 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4075 if (reexport_name_cstr && reexport_name_cstr[0]) { 4076 type = eSymbolTypeReExported; 4077 ConstString reexport_name(reexport_name_cstr + 4078 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4079 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4080 set_value = false; 4081 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4082 indirect_symbol_names.insert( 4083 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4084 } else 4085 type = eSymbolTypeUndefined; 4086 } break; 4087 4088 case N_UNDF: 4089 if (symbol_name && symbol_name[0]) { 4090 ConstString undefined_name(symbol_name + 4091 ((symbol_name[0] == '_') ? 1 : 0)); 4092 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4093 } 4094 LLVM_FALLTHROUGH; 4095 4096 case N_PBUD: 4097 type = eSymbolTypeUndefined; 4098 break; 4099 4100 case N_ABS: 4101 type = eSymbolTypeAbsolute; 4102 break; 4103 4104 case N_SECT: { 4105 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4106 4107 if (!symbol_section) { 4108 // TODO: warn about this? 4109 add_nlist = false; 4110 break; 4111 } 4112 4113 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4114 type = eSymbolTypeException; 4115 } else { 4116 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4117 4118 switch (section_type) { 4119 case S_CSTRING_LITERALS: 4120 type = eSymbolTypeData; 4121 break; // section with only literal C strings 4122 case S_4BYTE_LITERALS: 4123 type = eSymbolTypeData; 4124 break; // section with only 4 byte literals 4125 case S_8BYTE_LITERALS: 4126 type = eSymbolTypeData; 4127 break; // section with only 8 byte literals 4128 case S_LITERAL_POINTERS: 4129 type = eSymbolTypeTrampoline; 4130 break; // section with only pointers to literals 4131 case S_NON_LAZY_SYMBOL_POINTERS: 4132 type = eSymbolTypeTrampoline; 4133 break; // section with only non-lazy symbol pointers 4134 case S_LAZY_SYMBOL_POINTERS: 4135 type = eSymbolTypeTrampoline; 4136 break; // section with only lazy symbol pointers 4137 case S_SYMBOL_STUBS: 4138 type = eSymbolTypeTrampoline; 4139 break; // section with only symbol stubs, byte size of stub in 4140 // the reserved2 field 4141 case S_MOD_INIT_FUNC_POINTERS: 4142 type = eSymbolTypeCode; 4143 break; // section with only function pointers for initialization 4144 case S_MOD_TERM_FUNC_POINTERS: 4145 type = eSymbolTypeCode; 4146 break; // section with only function pointers for termination 4147 case S_INTERPOSING: 4148 type = eSymbolTypeTrampoline; 4149 break; // section with only pairs of function pointers for 4150 // interposing 4151 case S_16BYTE_LITERALS: 4152 type = eSymbolTypeData; 4153 break; // section with only 16 byte literals 4154 case S_DTRACE_DOF: 4155 type = eSymbolTypeInstrumentation; 4156 break; 4157 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4158 type = eSymbolTypeTrampoline; 4159 break; 4160 default: 4161 switch (symbol_section->GetType()) { 4162 case lldb::eSectionTypeCode: 4163 type = eSymbolTypeCode; 4164 break; 4165 case eSectionTypeData: 4166 case eSectionTypeDataCString: // Inlined C string data 4167 case eSectionTypeDataCStringPointers: // Pointers to C string 4168 // data 4169 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4170 // the symbol table 4171 case eSectionTypeData4: 4172 case eSectionTypeData8: 4173 case eSectionTypeData16: 4174 type = eSymbolTypeData; 4175 break; 4176 default: 4177 break; 4178 } 4179 break; 4180 } 4181 4182 if (type == eSymbolTypeInvalid) { 4183 const char *symbol_sect_name = 4184 symbol_section->GetName().AsCString(); 4185 if (symbol_section->IsDescendant(text_section_sp.get())) { 4186 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4187 S_ATTR_SELF_MODIFYING_CODE | 4188 S_ATTR_SOME_INSTRUCTIONS)) 4189 type = eSymbolTypeData; 4190 else 4191 type = eSymbolTypeCode; 4192 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 4193 symbol_section->IsDescendant( 4194 data_dirty_section_sp.get()) || 4195 symbol_section->IsDescendant( 4196 data_const_section_sp.get())) { 4197 if (symbol_sect_name && 4198 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 4199 type = eSymbolTypeRuntime; 4200 4201 if (symbol_name) { 4202 llvm::StringRef symbol_name_ref(symbol_name); 4203 if (symbol_name_ref.startswith("_OBJC_")) { 4204 llvm::StringRef g_objc_v2_prefix_class( 4205 "_OBJC_CLASS_$_"); 4206 llvm::StringRef g_objc_v2_prefix_metaclass( 4207 "_OBJC_METACLASS_$_"); 4208 llvm::StringRef g_objc_v2_prefix_ivar( 4209 "_OBJC_IVAR_$_"); 4210 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 4211 symbol_name_non_abi_mangled = symbol_name + 1; 4212 symbol_name = 4213 symbol_name + g_objc_v2_prefix_class.size(); 4214 type = eSymbolTypeObjCClass; 4215 demangled_is_synthesized = true; 4216 } else if (symbol_name_ref.startswith( 4217 g_objc_v2_prefix_metaclass)) { 4218 symbol_name_non_abi_mangled = symbol_name + 1; 4219 symbol_name = 4220 symbol_name + g_objc_v2_prefix_metaclass.size(); 4221 type = eSymbolTypeObjCMetaClass; 4222 demangled_is_synthesized = true; 4223 } else if (symbol_name_ref.startswith( 4224 g_objc_v2_prefix_ivar)) { 4225 symbol_name_non_abi_mangled = symbol_name + 1; 4226 symbol_name = 4227 symbol_name + g_objc_v2_prefix_ivar.size(); 4228 type = eSymbolTypeObjCIVar; 4229 demangled_is_synthesized = true; 4230 } 4231 } 4232 } 4233 } else if (symbol_sect_name && 4234 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4235 symbol_sect_name) { 4236 type = eSymbolTypeException; 4237 } else { 4238 type = eSymbolTypeData; 4239 } 4240 } else if (symbol_sect_name && 4241 ::strstr(symbol_sect_name, "__IMPORT") == 4242 symbol_sect_name) { 4243 type = eSymbolTypeTrampoline; 4244 } else if (symbol_section->IsDescendant(objc_section_sp.get())) { 4245 type = eSymbolTypeRuntime; 4246 if (symbol_name && symbol_name[0] == '.') { 4247 llvm::StringRef symbol_name_ref(symbol_name); 4248 llvm::StringRef g_objc_v1_prefix_class( 4249 ".objc_class_name_"); 4250 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4251 symbol_name_non_abi_mangled = symbol_name; 4252 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4253 type = eSymbolTypeObjCClass; 4254 demangled_is_synthesized = true; 4255 } 4256 } 4257 } 4258 } 4259 } 4260 } break; 4261 } 4262 } 4263 4264 if (!add_nlist) { 4265 sym[sym_idx].Clear(); 4266 return true; 4267 } 4268 4269 uint64_t symbol_value = nlist.n_value; 4270 4271 if (symbol_name_non_abi_mangled) { 4272 sym[sym_idx].GetMangled().SetMangledName( 4273 ConstString(symbol_name_non_abi_mangled)); 4274 sym[sym_idx].GetMangled().SetDemangledName(ConstString(symbol_name)); 4275 } else { 4276 bool symbol_name_is_mangled = false; 4277 4278 if (symbol_name && symbol_name[0] == '_') { 4279 symbol_name_is_mangled = symbol_name[1] == '_'; 4280 symbol_name++; // Skip the leading underscore 4281 } 4282 4283 if (symbol_name) { 4284 ConstString const_symbol_name(symbol_name); 4285 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4286 symbol_name_is_mangled); 4287 } 4288 } 4289 4290 if (is_gsym) { 4291 const char *gsym_name = 4292 sym[sym_idx] 4293 .GetMangled() 4294 .GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled) 4295 .GetCString(); 4296 if (gsym_name) 4297 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4298 } 4299 4300 if (symbol_section) { 4301 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4302 if (symbol_byte_size == 0 && function_starts_count > 0) { 4303 addr_t symbol_lookup_file_addr = nlist.n_value; 4304 // Do an exact address match for non-ARM addresses, else get the 4305 // closest since the symbol might be a thumb symbol which has an 4306 // address with bit zero set. 4307 FunctionStarts::Entry *func_start_entry = 4308 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4309 if (is_arm && func_start_entry) { 4310 // Verify that the function start address is the symbol address 4311 // (ARM) or the symbol address + 1 (thumb). 4312 if (func_start_entry->addr != symbol_lookup_file_addr && 4313 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4314 // Not the right entry, NULL it out... 4315 func_start_entry = nullptr; 4316 } 4317 } 4318 if (func_start_entry) { 4319 func_start_entry->data = true; 4320 4321 addr_t symbol_file_addr = func_start_entry->addr; 4322 if (is_arm) 4323 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4324 4325 const FunctionStarts::Entry *next_func_start_entry = 4326 function_starts.FindNextEntry(func_start_entry); 4327 const addr_t section_end_file_addr = 4328 section_file_addr + symbol_section->GetByteSize(); 4329 if (next_func_start_entry) { 4330 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4331 // Be sure the clear the Thumb address bit when we calculate the 4332 // size from the current and next address 4333 if (is_arm) 4334 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4335 symbol_byte_size = std::min<lldb::addr_t>( 4336 next_symbol_file_addr - symbol_file_addr, 4337 section_end_file_addr - symbol_file_addr); 4338 } else { 4339 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4340 } 4341 } 4342 } 4343 symbol_value -= section_file_addr; 4344 } 4345 4346 if (!is_debug) { 4347 if (type == eSymbolTypeCode) { 4348 // See if we can find a N_FUN entry for any code symbols. If we do 4349 // find a match, and the name matches, then we can merge the two into 4350 // just the function symbol to avoid duplicate entries in the symbol 4351 // table. 4352 std::pair<ValueToSymbolIndexMap::const_iterator, 4353 ValueToSymbolIndexMap::const_iterator> 4354 range; 4355 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4356 if (range.first != range.second) { 4357 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4358 pos != range.second; ++pos) { 4359 if (sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, 4360 Mangled::ePreferMangled) == 4361 sym[pos->second].GetMangled().GetName( 4362 lldb::eLanguageTypeUnknown, Mangled::ePreferMangled)) { 4363 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4364 // We just need the flags from the linker symbol, so put these 4365 // flags into the N_FUN flags to avoid duplicate symbols in the 4366 // symbol table. 4367 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4368 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4369 if (resolver_addresses.find(nlist.n_value) != 4370 resolver_addresses.end()) 4371 sym[pos->second].SetType(eSymbolTypeResolver); 4372 sym[sym_idx].Clear(); 4373 return true; 4374 } 4375 } 4376 } else { 4377 if (resolver_addresses.find(nlist.n_value) != 4378 resolver_addresses.end()) 4379 type = eSymbolTypeResolver; 4380 } 4381 } else if (type == eSymbolTypeData || type == eSymbolTypeObjCClass || 4382 type == eSymbolTypeObjCMetaClass || 4383 type == eSymbolTypeObjCIVar) { 4384 // See if we can find a N_STSYM entry for any data symbols. If we do 4385 // find a match, and the name matches, then we can merge the two into 4386 // just the Static symbol to avoid duplicate entries in the symbol 4387 // table. 4388 std::pair<ValueToSymbolIndexMap::const_iterator, 4389 ValueToSymbolIndexMap::const_iterator> 4390 range; 4391 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4392 if (range.first != range.second) { 4393 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4394 pos != range.second; ++pos) { 4395 if (sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, 4396 Mangled::ePreferMangled) == 4397 sym[pos->second].GetMangled().GetName( 4398 lldb::eLanguageTypeUnknown, Mangled::ePreferMangled)) { 4399 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4400 // We just need the flags from the linker symbol, so put these 4401 // flags into the N_STSYM flags to avoid duplicate symbols in 4402 // the symbol table. 4403 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4404 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4405 sym[sym_idx].Clear(); 4406 return true; 4407 } 4408 } 4409 } else { 4410 // Combine N_GSYM stab entries with the non stab symbol. 4411 const char *gsym_name = sym[sym_idx] 4412 .GetMangled() 4413 .GetName(lldb::eLanguageTypeUnknown, 4414 Mangled::ePreferMangled) 4415 .GetCString(); 4416 if (gsym_name) { 4417 ConstNameToSymbolIndexMap::const_iterator pos = 4418 N_GSYM_name_to_sym_idx.find(gsym_name); 4419 if (pos != N_GSYM_name_to_sym_idx.end()) { 4420 const uint32_t GSYM_sym_idx = pos->second; 4421 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4422 // Copy the address, because often the N_GSYM address has an 4423 // invalid address of zero when the global is a common symbol. 4424 sym[GSYM_sym_idx].GetAddressRef().SetSection(symbol_section); 4425 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4426 // We just need the flags from the linker symbol, so put these 4427 // flags into the N_GSYM flags to avoid duplicate symbols in 4428 // the symbol table. 4429 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4430 sym[sym_idx].Clear(); 4431 return true; 4432 } 4433 } 4434 } 4435 } 4436 } 4437 4438 sym[sym_idx].SetID(nlist_idx); 4439 sym[sym_idx].SetType(type); 4440 if (set_value) { 4441 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4442 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4443 } 4444 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4445 if (nlist.n_desc & N_WEAK_REF) 4446 sym[sym_idx].SetIsWeak(true); 4447 4448 if (symbol_byte_size > 0) 4449 sym[sym_idx].SetByteSize(symbol_byte_size); 4450 4451 if (demangled_is_synthesized) 4452 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4453 4454 ++sym_idx; 4455 return true; 4456 }; 4457 4458 // First parse all the nlists but don't process them yet. See the next 4459 // comment for an explanation why. 4460 std::vector<struct nlist_64> nlists; 4461 nlists.reserve(symtab_load_command.nsyms); 4462 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 4463 if (auto nlist = 4464 ParseNList(nlist_data, nlist_data_offset, nlist_byte_size)) 4465 nlists.push_back(*nlist); 4466 else 4467 break; 4468 } 4469 4470 // Now parse all the debug symbols. This is needed to merge non-debug 4471 // symbols in the next step. Non-debug symbols are always coalesced into 4472 // the debug symbol. Doing this in one step would mean that some symbols 4473 // won't be merged. 4474 nlist_idx = 0; 4475 for (auto &nlist : nlists) { 4476 if (!ParseSymbolLambda(nlist, nlist_idx++, DebugSymbols)) 4477 break; 4478 } 4479 4480 // Finally parse all the non debug symbols. 4481 nlist_idx = 0; 4482 for (auto &nlist : nlists) { 4483 if (!ParseSymbolLambda(nlist, nlist_idx++, NonDebugSymbols)) 4484 break; 4485 } 4486 4487 for (const auto &pos : reexport_shlib_needs_fixup) { 4488 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4489 if (undef_pos != undefined_name_to_desc.end()) { 4490 const uint8_t dylib_ordinal = 4491 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4492 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4493 sym[pos.first].SetReExportedSymbolSharedLibrary( 4494 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4495 } 4496 } 4497 } 4498 4499 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4500 4501 if (function_starts_count > 0) { 4502 uint32_t num_synthetic_function_symbols = 0; 4503 for (i = 0; i < function_starts_count; ++i) { 4504 if (!function_starts.GetEntryRef(i).data) 4505 ++num_synthetic_function_symbols; 4506 } 4507 4508 if (num_synthetic_function_symbols > 0) { 4509 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4510 num_syms = sym_idx + num_synthetic_function_symbols; 4511 sym = symtab->Resize(num_syms); 4512 } 4513 for (i = 0; i < function_starts_count; ++i) { 4514 const FunctionStarts::Entry *func_start_entry = 4515 function_starts.GetEntryAtIndex(i); 4516 if (!func_start_entry->data) { 4517 addr_t symbol_file_addr = func_start_entry->addr; 4518 uint32_t symbol_flags = 0; 4519 if (is_arm) { 4520 if (symbol_file_addr & 1) 4521 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4522 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4523 } 4524 Address symbol_addr; 4525 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4526 SectionSP symbol_section(symbol_addr.GetSection()); 4527 uint32_t symbol_byte_size = 0; 4528 if (symbol_section) { 4529 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4530 const FunctionStarts::Entry *next_func_start_entry = 4531 function_starts.FindNextEntry(func_start_entry); 4532 const addr_t section_end_file_addr = 4533 section_file_addr + symbol_section->GetByteSize(); 4534 if (next_func_start_entry) { 4535 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4536 if (is_arm) 4537 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4538 symbol_byte_size = std::min<lldb::addr_t>( 4539 next_symbol_file_addr - symbol_file_addr, 4540 section_end_file_addr - symbol_file_addr); 4541 } else { 4542 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4543 } 4544 sym[sym_idx].SetID(synthetic_sym_id++); 4545 sym[sym_idx].GetMangled().SetDemangledName( 4546 GetNextSyntheticSymbolName()); 4547 sym[sym_idx].SetType(eSymbolTypeCode); 4548 sym[sym_idx].SetIsSynthetic(true); 4549 sym[sym_idx].GetAddressRef() = symbol_addr; 4550 if (symbol_flags) 4551 sym[sym_idx].SetFlags(symbol_flags); 4552 if (symbol_byte_size) 4553 sym[sym_idx].SetByteSize(symbol_byte_size); 4554 ++sym_idx; 4555 } 4556 } 4557 } 4558 } 4559 } 4560 } 4561 4562 // Trim our symbols down to just what we ended up with after removing any 4563 // symbols. 4564 if (sym_idx < num_syms) { 4565 num_syms = sym_idx; 4566 sym = symtab->Resize(num_syms); 4567 } 4568 4569 // Now synthesize indirect symbols 4570 if (m_dysymtab.nindirectsyms != 0) { 4571 if (indirect_symbol_index_data.GetByteSize()) { 4572 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4573 m_nlist_idx_to_sym_idx.end(); 4574 4575 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4576 ++sect_idx) { 4577 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4578 S_SYMBOL_STUBS) { 4579 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4580 if (symbol_stub_byte_size == 0) 4581 continue; 4582 4583 const uint32_t num_symbol_stubs = 4584 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4585 4586 if (num_symbol_stubs == 0) 4587 continue; 4588 4589 const uint32_t symbol_stub_index_offset = 4590 m_mach_sections[sect_idx].reserved1; 4591 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) { 4592 const uint32_t symbol_stub_index = 4593 symbol_stub_index_offset + stub_idx; 4594 const lldb::addr_t symbol_stub_addr = 4595 m_mach_sections[sect_idx].addr + 4596 (stub_idx * symbol_stub_byte_size); 4597 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4598 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4599 symbol_stub_offset, 4)) { 4600 const uint32_t stub_sym_id = 4601 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4602 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4603 continue; 4604 4605 NListIndexToSymbolIndexMap::const_iterator index_pos = 4606 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4607 Symbol *stub_symbol = nullptr; 4608 if (index_pos != end_index_pos) { 4609 // We have a remapping from the original nlist index to a 4610 // current symbol index, so just look this up by index 4611 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4612 } else { 4613 // We need to lookup a symbol using the original nlist symbol 4614 // index since this index is coming from the S_SYMBOL_STUBS 4615 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4616 } 4617 4618 if (stub_symbol) { 4619 Address so_addr(symbol_stub_addr, section_list); 4620 4621 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4622 // Change the external symbol into a trampoline that makes 4623 // sense These symbols were N_UNDF N_EXT, and are useless 4624 // to us, so we can re-use them so we don't have to make up 4625 // a synthetic symbol for no good reason. 4626 if (resolver_addresses.find(symbol_stub_addr) == 4627 resolver_addresses.end()) 4628 stub_symbol->SetType(eSymbolTypeTrampoline); 4629 else 4630 stub_symbol->SetType(eSymbolTypeResolver); 4631 stub_symbol->SetExternal(false); 4632 stub_symbol->GetAddressRef() = so_addr; 4633 stub_symbol->SetByteSize(symbol_stub_byte_size); 4634 } else { 4635 // Make a synthetic symbol to describe the trampoline stub 4636 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4637 if (sym_idx >= num_syms) { 4638 sym = symtab->Resize(++num_syms); 4639 stub_symbol = nullptr; // this pointer no longer valid 4640 } 4641 sym[sym_idx].SetID(synthetic_sym_id++); 4642 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4643 if (resolver_addresses.find(symbol_stub_addr) == 4644 resolver_addresses.end()) 4645 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4646 else 4647 sym[sym_idx].SetType(eSymbolTypeResolver); 4648 sym[sym_idx].SetIsSynthetic(true); 4649 sym[sym_idx].GetAddressRef() = so_addr; 4650 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4651 ++sym_idx; 4652 } 4653 } else { 4654 if (log) 4655 log->Warning("symbol stub referencing symbol table symbol " 4656 "%u that isn't in our minimal symbol table, " 4657 "fix this!!!", 4658 stub_sym_id); 4659 } 4660 } 4661 } 4662 } 4663 } 4664 } 4665 } 4666 4667 if (!trie_entries.empty()) { 4668 for (const auto &e : trie_entries) { 4669 if (e.entry.import_name) { 4670 // Only add indirect symbols from the Trie entries if we didn't have 4671 // a N_INDR nlist entry for this already 4672 if (indirect_symbol_names.find(e.entry.name) == 4673 indirect_symbol_names.end()) { 4674 // Make a synthetic symbol to describe re-exported symbol. 4675 if (sym_idx >= num_syms) 4676 sym = symtab->Resize(++num_syms); 4677 sym[sym_idx].SetID(synthetic_sym_id++); 4678 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4679 sym[sym_idx].SetType(eSymbolTypeReExported); 4680 sym[sym_idx].SetIsSynthetic(true); 4681 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4682 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4683 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4684 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4685 } 4686 ++sym_idx; 4687 } 4688 } 4689 } 4690 } 4691 4692 // StreamFile s(stdout, false); 4693 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4694 // symtab->Dump(&s, NULL, eSortOrderNone); 4695 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4696 // sizes 4697 symtab->CalculateSymbolSizes(); 4698 4699 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4700 // symtab->Dump(&s, NULL, eSortOrderNone); 4701 4702 return symtab->GetNumSymbols(); 4703 } 4704 4705 void ObjectFileMachO::Dump(Stream *s) { 4706 ModuleSP module_sp(GetModule()); 4707 if (module_sp) { 4708 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4709 s->Printf("%p: ", static_cast<void *>(this)); 4710 s->Indent(); 4711 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4712 s->PutCString("ObjectFileMachO64"); 4713 else 4714 s->PutCString("ObjectFileMachO32"); 4715 4716 *s << ", file = '" << m_file; 4717 ModuleSpecList all_specs; 4718 ModuleSpec base_spec; 4719 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 4720 base_spec, all_specs); 4721 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 4722 *s << "', triple"; 4723 if (e) 4724 s->Printf("[%d]", i); 4725 *s << " = "; 4726 *s << all_specs.GetModuleSpecRefAtIndex(i) 4727 .GetArchitecture() 4728 .GetTriple() 4729 .getTriple(); 4730 } 4731 *s << "\n"; 4732 SectionList *sections = GetSectionList(); 4733 if (sections) 4734 sections->Dump(s, nullptr, true, UINT32_MAX); 4735 4736 if (m_symtab_up) 4737 m_symtab_up->Dump(s, nullptr, eSortOrderNone); 4738 } 4739 } 4740 4741 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4742 const lldb_private::DataExtractor &data, 4743 lldb::offset_t lc_offset) { 4744 uint32_t i; 4745 struct uuid_command load_cmd; 4746 4747 lldb::offset_t offset = lc_offset; 4748 for (i = 0; i < header.ncmds; ++i) { 4749 const lldb::offset_t cmd_offset = offset; 4750 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4751 break; 4752 4753 if (load_cmd.cmd == LC_UUID) { 4754 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4755 4756 if (uuid_bytes) { 4757 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4758 // We pretend these object files have no UUID to prevent crashing. 4759 4760 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4761 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4762 0xbb, 0x14, 0xf0, 0x0d}; 4763 4764 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4765 return UUID(); 4766 4767 return UUID::fromOptionalData(uuid_bytes, 16); 4768 } 4769 return UUID(); 4770 } 4771 offset = cmd_offset + load_cmd.cmdsize; 4772 } 4773 return UUID(); 4774 } 4775 4776 static llvm::StringRef GetOSName(uint32_t cmd) { 4777 switch (cmd) { 4778 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4779 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4780 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4781 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4782 case llvm::MachO::LC_VERSION_MIN_TVOS: 4783 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4784 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4785 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4786 default: 4787 llvm_unreachable("unexpected LC_VERSION load command"); 4788 } 4789 } 4790 4791 namespace { 4792 struct OSEnv { 4793 llvm::StringRef os_type; 4794 llvm::StringRef environment; 4795 OSEnv(uint32_t cmd) { 4796 switch (cmd) { 4797 case llvm::MachO::PLATFORM_MACOS: 4798 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4799 return; 4800 case llvm::MachO::PLATFORM_IOS: 4801 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4802 return; 4803 case llvm::MachO::PLATFORM_TVOS: 4804 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4805 return; 4806 case llvm::MachO::PLATFORM_WATCHOS: 4807 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4808 return; 4809 // NEED_BRIDGEOS_TRIPLE case llvm::MachO::PLATFORM_BRIDGEOS: 4810 // NEED_BRIDGEOS_TRIPLE os_type = 4811 // llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4812 // NEED_BRIDGEOS_TRIPLE return; 4813 case llvm::MachO::PLATFORM_MACCATALYST: 4814 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4815 environment = llvm::Triple::getEnvironmentTypeName(llvm::Triple::MacABI); 4816 return; 4817 case llvm::MachO::PLATFORM_IOSSIMULATOR: 4818 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4819 environment = 4820 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4821 return; 4822 case llvm::MachO::PLATFORM_TVOSSIMULATOR: 4823 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4824 environment = 4825 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4826 return; 4827 case llvm::MachO::PLATFORM_WATCHOSSIMULATOR: 4828 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4829 environment = 4830 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4831 return; 4832 default: { 4833 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 4834 LIBLLDB_LOG_PROCESS)); 4835 LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION"); 4836 } 4837 } 4838 } 4839 }; 4840 4841 struct MinOS { 4842 uint32_t major_version, minor_version, patch_version; 4843 MinOS(uint32_t version) 4844 : major_version(version >> 16), minor_version((version >> 8) & 0xffu), 4845 patch_version(version & 0xffu) {} 4846 }; 4847 } // namespace 4848 4849 void ObjectFileMachO::GetAllArchSpecs(const llvm::MachO::mach_header &header, 4850 const lldb_private::DataExtractor &data, 4851 lldb::offset_t lc_offset, 4852 ModuleSpec &base_spec, 4853 lldb_private::ModuleSpecList &all_specs) { 4854 auto &base_arch = base_spec.GetArchitecture(); 4855 base_arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4856 if (!base_arch.IsValid()) 4857 return; 4858 4859 bool found_any = false; 4860 auto add_triple = [&](const llvm::Triple &triple) { 4861 auto spec = base_spec; 4862 spec.GetArchitecture().GetTriple() = triple; 4863 if (spec.GetArchitecture().IsValid()) { 4864 spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset); 4865 all_specs.Append(spec); 4866 found_any = true; 4867 } 4868 }; 4869 4870 // Set OS to an unspecified unknown or a "*" so it can match any OS 4871 llvm::Triple base_triple = base_arch.GetTriple(); 4872 base_triple.setOS(llvm::Triple::UnknownOS); 4873 base_triple.setOSName(llvm::StringRef()); 4874 4875 if (header.filetype == MH_PRELOAD) { 4876 if (header.cputype == CPU_TYPE_ARM) { 4877 // If this is a 32-bit arm binary, and it's a standalone binary, force 4878 // the Vendor to Apple so we don't accidentally pick up the generic 4879 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4880 // frame pointer register; most other armv7 ABIs use a combination of 4881 // r7 and r11. 4882 base_triple.setVendor(llvm::Triple::Apple); 4883 } else { 4884 // Set vendor to an unspecified unknown or a "*" so it can match any 4885 // vendor This is required for correct behavior of EFI debugging on 4886 // x86_64 4887 base_triple.setVendor(llvm::Triple::UnknownVendor); 4888 base_triple.setVendorName(llvm::StringRef()); 4889 } 4890 return add_triple(base_triple); 4891 } 4892 4893 struct load_command load_cmd; 4894 4895 // See if there is an LC_VERSION_MIN_* load command that can give 4896 // us the OS type. 4897 lldb::offset_t offset = lc_offset; 4898 for (uint32_t i = 0; i < header.ncmds; ++i) { 4899 const lldb::offset_t cmd_offset = offset; 4900 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4901 break; 4902 4903 struct version_min_command version_min; 4904 switch (load_cmd.cmd) { 4905 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4906 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4907 case llvm::MachO::LC_VERSION_MIN_TVOS: 4908 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 4909 if (load_cmd.cmdsize != sizeof(version_min)) 4910 break; 4911 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 4912 data.GetByteOrder(), &version_min) == 0) 4913 break; 4914 MinOS min_os(version_min.version); 4915 llvm::SmallString<32> os_name; 4916 llvm::raw_svector_ostream os(os_name); 4917 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 4918 << min_os.minor_version << '.' << min_os.patch_version; 4919 4920 auto triple = base_triple; 4921 triple.setOSName(os.str()); 4922 os_name.clear(); 4923 add_triple(triple); 4924 break; 4925 } 4926 default: 4927 break; 4928 } 4929 4930 offset = cmd_offset + load_cmd.cmdsize; 4931 } 4932 4933 // See if there are LC_BUILD_VERSION load commands that can give 4934 // us the OS type. 4935 offset = lc_offset; 4936 for (uint32_t i = 0; i < header.ncmds; ++i) { 4937 const lldb::offset_t cmd_offset = offset; 4938 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4939 break; 4940 4941 do { 4942 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 4943 struct build_version_command build_version; 4944 if (load_cmd.cmdsize < sizeof(build_version)) { 4945 // Malformed load command. 4946 break; 4947 } 4948 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 4949 data.GetByteOrder(), &build_version) == 0) 4950 break; 4951 MinOS min_os(build_version.minos); 4952 OSEnv os_env(build_version.platform); 4953 llvm::SmallString<16> os_name; 4954 llvm::raw_svector_ostream os(os_name); 4955 os << os_env.os_type << min_os.major_version << '.' 4956 << min_os.minor_version << '.' << min_os.patch_version; 4957 auto triple = base_triple; 4958 triple.setOSName(os.str()); 4959 os_name.clear(); 4960 if (!os_env.environment.empty()) 4961 triple.setEnvironmentName(os_env.environment); 4962 add_triple(triple); 4963 } 4964 } while (0); 4965 offset = cmd_offset + load_cmd.cmdsize; 4966 } 4967 4968 if (!found_any) { 4969 if (header.filetype == MH_KEXT_BUNDLE) { 4970 base_triple.setVendor(llvm::Triple::Apple); 4971 add_triple(base_triple); 4972 } else { 4973 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 4974 // so lets not say our Vendor is Apple, leave it as an unspecified 4975 // unknown. 4976 base_triple.setVendor(llvm::Triple::UnknownVendor); 4977 base_triple.setVendorName(llvm::StringRef()); 4978 add_triple(base_triple); 4979 } 4980 } 4981 } 4982 4983 ArchSpec ObjectFileMachO::GetArchitecture( 4984 ModuleSP module_sp, const llvm::MachO::mach_header &header, 4985 const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) { 4986 ModuleSpecList all_specs; 4987 ModuleSpec base_spec; 4988 GetAllArchSpecs(header, data, MachHeaderSizeFromMagic(header.magic), 4989 base_spec, all_specs); 4990 4991 // If the object file offers multiple alternative load commands, 4992 // pick the one that matches the module. 4993 if (module_sp) { 4994 const ArchSpec &module_arch = module_sp->GetArchitecture(); 4995 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 4996 ArchSpec mach_arch = 4997 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 4998 if (module_arch.IsCompatibleMatch(mach_arch)) 4999 return mach_arch; 5000 } 5001 } 5002 5003 // Return the first arch we found. 5004 if (all_specs.GetSize() == 0) 5005 return {}; 5006 return all_specs.GetModuleSpecRefAtIndex(0).GetArchitecture(); 5007 } 5008 5009 UUID ObjectFileMachO::GetUUID() { 5010 ModuleSP module_sp(GetModule()); 5011 if (module_sp) { 5012 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5013 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5014 return GetUUID(m_header, m_data, offset); 5015 } 5016 return UUID(); 5017 } 5018 5019 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5020 uint32_t count = 0; 5021 ModuleSP module_sp(GetModule()); 5022 if (module_sp) { 5023 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5024 struct load_command load_cmd; 5025 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5026 std::vector<std::string> rpath_paths; 5027 std::vector<std::string> rpath_relative_paths; 5028 std::vector<std::string> at_exec_relative_paths; 5029 uint32_t i; 5030 for (i = 0; i < m_header.ncmds; ++i) { 5031 const uint32_t cmd_offset = offset; 5032 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5033 break; 5034 5035 switch (load_cmd.cmd) { 5036 case LC_RPATH: 5037 case LC_LOAD_DYLIB: 5038 case LC_LOAD_WEAK_DYLIB: 5039 case LC_REEXPORT_DYLIB: 5040 case LC_LOAD_DYLINKER: 5041 case LC_LOADFVMLIB: 5042 case LC_LOAD_UPWARD_DYLIB: { 5043 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5044 const char *path = m_data.PeekCStr(name_offset); 5045 if (path) { 5046 if (load_cmd.cmd == LC_RPATH) 5047 rpath_paths.push_back(path); 5048 else { 5049 if (path[0] == '@') { 5050 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5051 rpath_relative_paths.push_back(path + strlen("@rpath")); 5052 else if (strncmp(path, "@executable_path", 5053 strlen("@executable_path")) == 0) 5054 at_exec_relative_paths.push_back(path + 5055 strlen("@executable_path")); 5056 } else { 5057 FileSpec file_spec(path); 5058 if (files.AppendIfUnique(file_spec)) 5059 count++; 5060 } 5061 } 5062 } 5063 } break; 5064 5065 default: 5066 break; 5067 } 5068 offset = cmd_offset + load_cmd.cmdsize; 5069 } 5070 5071 FileSpec this_file_spec(m_file); 5072 FileSystem::Instance().Resolve(this_file_spec); 5073 5074 if (!rpath_paths.empty()) { 5075 // Fixup all LC_RPATH values to be absolute paths 5076 std::string loader_path("@loader_path"); 5077 std::string executable_path("@executable_path"); 5078 for (auto &rpath : rpath_paths) { 5079 if (rpath.find(loader_path) == 0) { 5080 rpath.erase(0, loader_path.size()); 5081 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5082 } else if (rpath.find(executable_path) == 0) { 5083 rpath.erase(0, executable_path.size()); 5084 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5085 } 5086 } 5087 5088 for (const auto &rpath_relative_path : rpath_relative_paths) { 5089 for (const auto &rpath : rpath_paths) { 5090 std::string path = rpath; 5091 path += rpath_relative_path; 5092 // It is OK to resolve this path because we must find a file on disk 5093 // for us to accept it anyway if it is rpath relative. 5094 FileSpec file_spec(path); 5095 FileSystem::Instance().Resolve(file_spec); 5096 if (FileSystem::Instance().Exists(file_spec) && 5097 files.AppendIfUnique(file_spec)) { 5098 count++; 5099 break; 5100 } 5101 } 5102 } 5103 } 5104 5105 // We may have @executable_paths but no RPATHS. Figure those out here. 5106 // Only do this if this object file is the executable. We have no way to 5107 // get back to the actual executable otherwise, so we won't get the right 5108 // path. 5109 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5110 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5111 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5112 FileSpec file_spec = 5113 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5114 if (FileSystem::Instance().Exists(file_spec) && 5115 files.AppendIfUnique(file_spec)) 5116 count++; 5117 } 5118 } 5119 } 5120 return count; 5121 } 5122 5123 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5124 // If the object file is not an executable it can't hold the entry point. 5125 // m_entry_point_address is initialized to an invalid address, so we can just 5126 // return that. If m_entry_point_address is valid it means we've found it 5127 // already, so return the cached value. 5128 5129 if ((!IsExecutable() && !IsDynamicLoader()) || 5130 m_entry_point_address.IsValid()) { 5131 return m_entry_point_address; 5132 } 5133 5134 // Otherwise, look for the UnixThread or Thread command. The data for the 5135 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5136 // 5137 // uint32_t flavor - this is the flavor argument you would pass to 5138 // thread_get_state 5139 // uint32_t count - this is the count of longs in the thread state data 5140 // struct XXX_thread_state state - this is the structure from 5141 // <machine/thread_status.h> corresponding to the flavor. 5142 // <repeat this trio> 5143 // 5144 // So we just keep reading the various register flavors till we find the GPR 5145 // one, then read the PC out of there. 5146 // FIXME: We will need to have a "RegisterContext data provider" class at some 5147 // point that can get all the registers 5148 // out of data in this form & attach them to a given thread. That should 5149 // underlie the MacOS X User process plugin, and we'll also need it for the 5150 // MacOS X Core File process plugin. When we have that we can also use it 5151 // here. 5152 // 5153 // For now we hard-code the offsets and flavors we need: 5154 // 5155 // 5156 5157 ModuleSP module_sp(GetModule()); 5158 if (module_sp) { 5159 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5160 struct load_command load_cmd; 5161 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5162 uint32_t i; 5163 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5164 bool done = false; 5165 5166 for (i = 0; i < m_header.ncmds; ++i) { 5167 const lldb::offset_t cmd_offset = offset; 5168 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5169 break; 5170 5171 switch (load_cmd.cmd) { 5172 case LC_UNIXTHREAD: 5173 case LC_THREAD: { 5174 while (offset < cmd_offset + load_cmd.cmdsize) { 5175 uint32_t flavor = m_data.GetU32(&offset); 5176 uint32_t count = m_data.GetU32(&offset); 5177 if (count == 0) { 5178 // We've gotten off somehow, log and exit; 5179 return m_entry_point_address; 5180 } 5181 5182 switch (m_header.cputype) { 5183 case llvm::MachO::CPU_TYPE_ARM: 5184 if (flavor == 1 || 5185 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 5186 // from mach/arm/thread_status.h 5187 { 5188 offset += 60; // This is the offset of pc in the GPR thread state 5189 // data structure. 5190 start_address = m_data.GetU32(&offset); 5191 done = true; 5192 } 5193 break; 5194 case llvm::MachO::CPU_TYPE_ARM64: 5195 case llvm::MachO::CPU_TYPE_ARM64_32: 5196 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5197 { 5198 offset += 256; // This is the offset of pc in the GPR thread state 5199 // data structure. 5200 start_address = m_data.GetU64(&offset); 5201 done = true; 5202 } 5203 break; 5204 case llvm::MachO::CPU_TYPE_I386: 5205 if (flavor == 5206 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5207 { 5208 offset += 40; // This is the offset of eip in the GPR thread state 5209 // data structure. 5210 start_address = m_data.GetU32(&offset); 5211 done = true; 5212 } 5213 break; 5214 case llvm::MachO::CPU_TYPE_X86_64: 5215 if (flavor == 5216 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5217 { 5218 offset += 16 * 8; // This is the offset of rip in the GPR thread 5219 // state data structure. 5220 start_address = m_data.GetU64(&offset); 5221 done = true; 5222 } 5223 break; 5224 default: 5225 return m_entry_point_address; 5226 } 5227 // Haven't found the GPR flavor yet, skip over the data for this 5228 // flavor: 5229 if (done) 5230 break; 5231 offset += count * 4; 5232 } 5233 } break; 5234 case LC_MAIN: { 5235 ConstString text_segment_name("__TEXT"); 5236 uint64_t entryoffset = m_data.GetU64(&offset); 5237 SectionSP text_segment_sp = 5238 GetSectionList()->FindSectionByName(text_segment_name); 5239 if (text_segment_sp) { 5240 done = true; 5241 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5242 } 5243 } break; 5244 5245 default: 5246 break; 5247 } 5248 if (done) 5249 break; 5250 5251 // Go to the next load command: 5252 offset = cmd_offset + load_cmd.cmdsize; 5253 } 5254 5255 if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) { 5256 if (GetSymtab()) { 5257 Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType( 5258 ConstString("_dyld_start"), SymbolType::eSymbolTypeCode, 5259 Symtab::eDebugAny, Symtab::eVisibilityAny); 5260 if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) { 5261 start_address = dyld_start_sym->GetAddress().GetFileAddress(); 5262 } 5263 } 5264 } 5265 5266 if (start_address != LLDB_INVALID_ADDRESS) { 5267 // We got the start address from the load commands, so now resolve that 5268 // address in the sections of this ObjectFile: 5269 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5270 start_address, GetSectionList())) { 5271 m_entry_point_address.Clear(); 5272 } 5273 } else { 5274 // We couldn't read the UnixThread load command - maybe it wasn't there. 5275 // As a fallback look for the "start" symbol in the main executable. 5276 5277 ModuleSP module_sp(GetModule()); 5278 5279 if (module_sp) { 5280 SymbolContextList contexts; 5281 SymbolContext context; 5282 module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5283 eSymbolTypeCode, contexts); 5284 if (contexts.GetSize()) { 5285 if (contexts.GetContextAtIndex(0, context)) 5286 m_entry_point_address = context.symbol->GetAddress(); 5287 } 5288 } 5289 } 5290 } 5291 5292 return m_entry_point_address; 5293 } 5294 5295 lldb_private::Address ObjectFileMachO::GetBaseAddress() { 5296 lldb_private::Address header_addr; 5297 SectionList *section_list = GetSectionList(); 5298 if (section_list) { 5299 SectionSP text_segment_sp( 5300 section_list->FindSectionByName(GetSegmentNameTEXT())); 5301 if (text_segment_sp) { 5302 header_addr.SetSection(text_segment_sp); 5303 header_addr.SetOffset(0); 5304 } 5305 } 5306 return header_addr; 5307 } 5308 5309 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5310 ModuleSP module_sp(GetModule()); 5311 if (module_sp) { 5312 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5313 if (!m_thread_context_offsets_valid) { 5314 m_thread_context_offsets_valid = true; 5315 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5316 FileRangeArray::Entry file_range; 5317 thread_command thread_cmd; 5318 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5319 const uint32_t cmd_offset = offset; 5320 if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr) 5321 break; 5322 5323 if (thread_cmd.cmd == LC_THREAD) { 5324 file_range.SetRangeBase(offset); 5325 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5326 m_thread_context_offsets.Append(file_range); 5327 } 5328 offset = cmd_offset + thread_cmd.cmdsize; 5329 } 5330 } 5331 } 5332 return m_thread_context_offsets.GetSize(); 5333 } 5334 5335 std::string ObjectFileMachO::GetIdentifierString() { 5336 std::string result; 5337 ModuleSP module_sp(GetModule()); 5338 if (module_sp) { 5339 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5340 5341 // First, look over the load commands for an LC_NOTE load command with 5342 // data_owner string "kern ver str" & use that if found. 5343 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5344 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5345 const uint32_t cmd_offset = offset; 5346 load_command lc; 5347 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5348 break; 5349 if (lc.cmd == LC_NOTE) { 5350 char data_owner[17]; 5351 m_data.CopyData(offset, 16, data_owner); 5352 data_owner[16] = '\0'; 5353 offset += 16; 5354 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5355 uint64_t size = m_data.GetU64_unchecked(&offset); 5356 5357 // "kern ver str" has a uint32_t version and then a nul terminated 5358 // c-string. 5359 if (strcmp("kern ver str", data_owner) == 0) { 5360 offset = fileoff; 5361 uint32_t version; 5362 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5363 if (version == 1) { 5364 uint32_t strsize = size - sizeof(uint32_t); 5365 char *buf = (char *)malloc(strsize); 5366 if (buf) { 5367 m_data.CopyData(offset, strsize, buf); 5368 buf[strsize - 1] = '\0'; 5369 result = buf; 5370 if (buf) 5371 free(buf); 5372 return result; 5373 } 5374 } 5375 } 5376 } 5377 } 5378 offset = cmd_offset + lc.cmdsize; 5379 } 5380 5381 // Second, make a pass over the load commands looking for an obsolete 5382 // LC_IDENT load command. 5383 offset = MachHeaderSizeFromMagic(m_header.magic); 5384 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5385 const uint32_t cmd_offset = offset; 5386 struct ident_command ident_command; 5387 if (m_data.GetU32(&offset, &ident_command, 2) == nullptr) 5388 break; 5389 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5390 char *buf = (char *)malloc(ident_command.cmdsize); 5391 if (buf != nullptr && m_data.CopyData(offset, ident_command.cmdsize, 5392 buf) == ident_command.cmdsize) { 5393 buf[ident_command.cmdsize - 1] = '\0'; 5394 result = buf; 5395 } 5396 if (buf) 5397 free(buf); 5398 } 5399 offset = cmd_offset + ident_command.cmdsize; 5400 } 5401 } 5402 return result; 5403 } 5404 5405 bool ObjectFileMachO::GetCorefileMainBinaryInfo(addr_t &address, UUID &uuid) { 5406 address = LLDB_INVALID_ADDRESS; 5407 uuid.Clear(); 5408 ModuleSP module_sp(GetModule()); 5409 if (module_sp) { 5410 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5411 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5412 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5413 const uint32_t cmd_offset = offset; 5414 load_command lc; 5415 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5416 break; 5417 if (lc.cmd == LC_NOTE) { 5418 char data_owner[17]; 5419 memset(data_owner, 0, sizeof(data_owner)); 5420 m_data.CopyData(offset, 16, data_owner); 5421 offset += 16; 5422 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5423 uint64_t size = m_data.GetU64_unchecked(&offset); 5424 5425 // "main bin spec" (main binary specification) data payload is 5426 // formatted: 5427 // uint32_t version [currently 1] 5428 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user 5429 // process] uint64_t address [ UINT64_MAX if address not 5430 // specified ] uuid_t uuid [ all zero's if uuid not 5431 // specified ] uint32_t log2_pagesize [ process page size in log base 5432 // 2, e.g. 4k pages are 12. 0 for unspecified ] 5433 5434 if (strcmp("main bin spec", data_owner) == 0 && size >= 32) { 5435 offset = fileoff; 5436 uint32_t version; 5437 if (m_data.GetU32(&offset, &version, 1) != nullptr && version == 1) { 5438 uint32_t type = 0; 5439 uuid_t raw_uuid; 5440 memset(raw_uuid, 0, sizeof(uuid_t)); 5441 5442 if (m_data.GetU32(&offset, &type, 1) && 5443 m_data.GetU64(&offset, &address, 1) && 5444 m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5445 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5446 return true; 5447 } 5448 } 5449 } 5450 } 5451 offset = cmd_offset + lc.cmdsize; 5452 } 5453 } 5454 return false; 5455 } 5456 5457 lldb::RegisterContextSP 5458 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5459 lldb_private::Thread &thread) { 5460 lldb::RegisterContextSP reg_ctx_sp; 5461 5462 ModuleSP module_sp(GetModule()); 5463 if (module_sp) { 5464 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5465 if (!m_thread_context_offsets_valid) 5466 GetNumThreadContexts(); 5467 5468 const FileRangeArray::Entry *thread_context_file_range = 5469 m_thread_context_offsets.GetEntryAtIndex(idx); 5470 if (thread_context_file_range) { 5471 5472 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5473 thread_context_file_range->GetByteSize()); 5474 5475 switch (m_header.cputype) { 5476 case llvm::MachO::CPU_TYPE_ARM64: 5477 case llvm::MachO::CPU_TYPE_ARM64_32: 5478 reg_ctx_sp = 5479 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data); 5480 break; 5481 5482 case llvm::MachO::CPU_TYPE_ARM: 5483 reg_ctx_sp = 5484 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data); 5485 break; 5486 5487 case llvm::MachO::CPU_TYPE_I386: 5488 reg_ctx_sp = 5489 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data); 5490 break; 5491 5492 case llvm::MachO::CPU_TYPE_X86_64: 5493 reg_ctx_sp = 5494 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data); 5495 break; 5496 } 5497 } 5498 } 5499 return reg_ctx_sp; 5500 } 5501 5502 ObjectFile::Type ObjectFileMachO::CalculateType() { 5503 switch (m_header.filetype) { 5504 case MH_OBJECT: // 0x1u 5505 if (GetAddressByteSize() == 4) { 5506 // 32 bit kexts are just object files, but they do have a valid 5507 // UUID load command. 5508 if (GetUUID()) { 5509 // this checking for the UUID load command is not enough we could 5510 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5511 // this is required of kexts 5512 if (m_strata == eStrataInvalid) 5513 m_strata = eStrataKernel; 5514 return eTypeSharedLibrary; 5515 } 5516 } 5517 return eTypeObjectFile; 5518 5519 case MH_EXECUTE: 5520 return eTypeExecutable; // 0x2u 5521 case MH_FVMLIB: 5522 return eTypeSharedLibrary; // 0x3u 5523 case MH_CORE: 5524 return eTypeCoreFile; // 0x4u 5525 case MH_PRELOAD: 5526 return eTypeSharedLibrary; // 0x5u 5527 case MH_DYLIB: 5528 return eTypeSharedLibrary; // 0x6u 5529 case MH_DYLINKER: 5530 return eTypeDynamicLinker; // 0x7u 5531 case MH_BUNDLE: 5532 return eTypeSharedLibrary; // 0x8u 5533 case MH_DYLIB_STUB: 5534 return eTypeStubLibrary; // 0x9u 5535 case MH_DSYM: 5536 return eTypeDebugInfo; // 0xAu 5537 case MH_KEXT_BUNDLE: 5538 return eTypeSharedLibrary; // 0xBu 5539 default: 5540 break; 5541 } 5542 return eTypeUnknown; 5543 } 5544 5545 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5546 switch (m_header.filetype) { 5547 case MH_OBJECT: // 0x1u 5548 { 5549 // 32 bit kexts are just object files, but they do have a valid 5550 // UUID load command. 5551 if (GetUUID()) { 5552 // this checking for the UUID load command is not enough we could 5553 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5554 // this is required of kexts 5555 if (m_type == eTypeInvalid) 5556 m_type = eTypeSharedLibrary; 5557 5558 return eStrataKernel; 5559 } 5560 } 5561 return eStrataUnknown; 5562 5563 case MH_EXECUTE: // 0x2u 5564 // Check for the MH_DYLDLINK bit in the flags 5565 if (m_header.flags & MH_DYLDLINK) { 5566 return eStrataUser; 5567 } else { 5568 SectionList *section_list = GetSectionList(); 5569 if (section_list) { 5570 static ConstString g_kld_section_name("__KLD"); 5571 if (section_list->FindSectionByName(g_kld_section_name)) 5572 return eStrataKernel; 5573 } 5574 } 5575 return eStrataRawImage; 5576 5577 case MH_FVMLIB: 5578 return eStrataUser; // 0x3u 5579 case MH_CORE: 5580 return eStrataUnknown; // 0x4u 5581 case MH_PRELOAD: 5582 return eStrataRawImage; // 0x5u 5583 case MH_DYLIB: 5584 return eStrataUser; // 0x6u 5585 case MH_DYLINKER: 5586 return eStrataUser; // 0x7u 5587 case MH_BUNDLE: 5588 return eStrataUser; // 0x8u 5589 case MH_DYLIB_STUB: 5590 return eStrataUser; // 0x9u 5591 case MH_DSYM: 5592 return eStrataUnknown; // 0xAu 5593 case MH_KEXT_BUNDLE: 5594 return eStrataKernel; // 0xBu 5595 default: 5596 break; 5597 } 5598 return eStrataUnknown; 5599 } 5600 5601 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5602 ModuleSP module_sp(GetModule()); 5603 if (module_sp) { 5604 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5605 struct dylib_command load_cmd; 5606 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5607 uint32_t version_cmd = 0; 5608 uint64_t version = 0; 5609 uint32_t i; 5610 for (i = 0; i < m_header.ncmds; ++i) { 5611 const lldb::offset_t cmd_offset = offset; 5612 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5613 break; 5614 5615 if (load_cmd.cmd == LC_ID_DYLIB) { 5616 if (version_cmd == 0) { 5617 version_cmd = load_cmd.cmd; 5618 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr) 5619 break; 5620 version = load_cmd.dylib.current_version; 5621 } 5622 break; // Break for now unless there is another more complete version 5623 // number load command in the future. 5624 } 5625 offset = cmd_offset + load_cmd.cmdsize; 5626 } 5627 5628 if (version_cmd == LC_ID_DYLIB) { 5629 unsigned major = (version & 0xFFFF0000ull) >> 16; 5630 unsigned minor = (version & 0x0000FF00ull) >> 8; 5631 unsigned subminor = (version & 0x000000FFull); 5632 return llvm::VersionTuple(major, minor, subminor); 5633 } 5634 } 5635 return llvm::VersionTuple(); 5636 } 5637 5638 ArchSpec ObjectFileMachO::GetArchitecture() { 5639 ModuleSP module_sp(GetModule()); 5640 ArchSpec arch; 5641 if (module_sp) { 5642 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5643 5644 return GetArchitecture(module_sp, m_header, m_data, 5645 MachHeaderSizeFromMagic(m_header.magic)); 5646 } 5647 return arch; 5648 } 5649 5650 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, 5651 addr_t &base_addr, UUID &uuid) { 5652 uuid.Clear(); 5653 base_addr = LLDB_INVALID_ADDRESS; 5654 if (process && process->GetDynamicLoader()) { 5655 DynamicLoader *dl = process->GetDynamicLoader(); 5656 LazyBool using_shared_cache; 5657 LazyBool private_shared_cache; 5658 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5659 private_shared_cache); 5660 } 5661 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5662 LIBLLDB_LOG_PROCESS)); 5663 LLDB_LOGF( 5664 log, 5665 "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64, 5666 uuid.GetAsString().c_str(), base_addr); 5667 } 5668 5669 // From dyld SPI header dyld_process_info.h 5670 typedef void *dyld_process_info; 5671 struct lldb_copy__dyld_process_cache_info { 5672 uuid_t cacheUUID; // UUID of cache used by process 5673 uint64_t cacheBaseAddress; // load address of dyld shared cache 5674 bool noCache; // process is running without a dyld cache 5675 bool privateCache; // process is using a private copy of its dyld cache 5676 }; 5677 5678 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with 5679 // llvm enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile 5680 // errors. So we need to use the actual underlying types of task_t and 5681 // kern_return_t below. 5682 extern "C" unsigned int /*task_t*/ mach_task_self(); 5683 5684 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5685 uuid.Clear(); 5686 base_addr = LLDB_INVALID_ADDRESS; 5687 5688 #if defined(__APPLE__) && \ 5689 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5690 uint8_t *(*dyld_get_all_image_infos)(void); 5691 dyld_get_all_image_infos = 5692 (uint8_t * (*)()) dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5693 if (dyld_get_all_image_infos) { 5694 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5695 if (dyld_all_image_infos_address) { 5696 uint32_t *version = (uint32_t *) 5697 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5698 if (*version >= 13) { 5699 uuid_t *sharedCacheUUID_address = 0; 5700 int wordsize = sizeof(uint8_t *); 5701 if (wordsize == 8) { 5702 sharedCacheUUID_address = 5703 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5704 160); // sharedCacheUUID <mach-o/dyld_images.h> 5705 if (*version >= 15) 5706 base_addr = 5707 *(uint64_t 5708 *)((uint8_t *)dyld_all_image_infos_address + 5709 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5710 } else { 5711 sharedCacheUUID_address = 5712 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5713 84); // sharedCacheUUID <mach-o/dyld_images.h> 5714 if (*version >= 15) { 5715 base_addr = 0; 5716 base_addr = 5717 *(uint32_t 5718 *)((uint8_t *)dyld_all_image_infos_address + 5719 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5720 } 5721 } 5722 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5723 } 5724 } 5725 } else { 5726 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5727 dyld_process_info (*dyld_process_info_create)( 5728 unsigned int /* task_t */ task, uint64_t timestamp, 5729 unsigned int /*kern_return_t*/ *kernelError); 5730 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5731 void (*dyld_process_info_release)(dyld_process_info info); 5732 5733 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, 5734 unsigned int /*kern_return_t*/ *)) 5735 dlsym(RTLD_DEFAULT, "_dyld_process_info_create"); 5736 dyld_process_info_get_cache = (void (*)(void *, void *))dlsym( 5737 RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5738 dyld_process_info_release = 5739 (void (*)(void *))dlsym(RTLD_DEFAULT, "_dyld_process_info_release"); 5740 5741 if (dyld_process_info_create && dyld_process_info_get_cache) { 5742 unsigned int /*kern_return_t */ kern_ret; 5743 dyld_process_info process_info = 5744 dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5745 if (process_info) { 5746 struct lldb_copy__dyld_process_cache_info sc_info; 5747 memset(&sc_info, 0, sizeof(struct lldb_copy__dyld_process_cache_info)); 5748 dyld_process_info_get_cache(process_info, &sc_info); 5749 if (sc_info.cacheBaseAddress != 0) { 5750 base_addr = sc_info.cacheBaseAddress; 5751 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 5752 } 5753 dyld_process_info_release(process_info); 5754 } 5755 } 5756 } 5757 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5758 LIBLLDB_LOG_PROCESS)); 5759 if (log && uuid.IsValid()) 5760 LLDB_LOGF(log, 5761 "lldb's in-memory shared cache has a UUID of %s base address of " 5762 "0x%" PRIx64, 5763 uuid.GetAsString().c_str(), base_addr); 5764 #endif 5765 } 5766 5767 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5768 if (!m_min_os_version) { 5769 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5770 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5771 const lldb::offset_t load_cmd_offset = offset; 5772 5773 version_min_command lc; 5774 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5775 break; 5776 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5777 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5778 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5779 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5780 if (m_data.GetU32(&offset, &lc.version, 5781 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5782 const uint32_t xxxx = lc.version >> 16; 5783 const uint32_t yy = (lc.version >> 8) & 0xffu; 5784 const uint32_t zz = lc.version & 0xffu; 5785 if (xxxx) { 5786 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5787 break; 5788 } 5789 } 5790 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5791 // struct build_version_command { 5792 // uint32_t cmd; /* LC_BUILD_VERSION */ 5793 // uint32_t cmdsize; /* sizeof(struct 5794 // build_version_command) plus */ 5795 // /* ntools * sizeof(struct 5796 // build_tool_version) */ 5797 // uint32_t platform; /* platform */ 5798 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5799 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded in 5800 // nibbles xxxx.yy.zz */ uint32_t ntools; /* number of 5801 // tool entries following this */ 5802 // }; 5803 5804 offset += 4; // skip platform 5805 uint32_t minos = m_data.GetU32(&offset); 5806 5807 const uint32_t xxxx = minos >> 16; 5808 const uint32_t yy = (minos >> 8) & 0xffu; 5809 const uint32_t zz = minos & 0xffu; 5810 if (xxxx) { 5811 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5812 break; 5813 } 5814 } 5815 5816 offset = load_cmd_offset + lc.cmdsize; 5817 } 5818 5819 if (!m_min_os_version) { 5820 // Set version to an empty value so we don't keep trying to 5821 m_min_os_version = llvm::VersionTuple(); 5822 } 5823 } 5824 5825 return *m_min_os_version; 5826 } 5827 5828 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { 5829 if (!m_sdk_versions.hasValue()) { 5830 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5831 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5832 const lldb::offset_t load_cmd_offset = offset; 5833 5834 version_min_command lc; 5835 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5836 break; 5837 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5838 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5839 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5840 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5841 if (m_data.GetU32(&offset, &lc.version, 5842 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5843 const uint32_t xxxx = lc.sdk >> 16; 5844 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5845 const uint32_t zz = lc.sdk & 0xffu; 5846 if (xxxx) { 5847 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 5848 break; 5849 } else { 5850 GetModule()->ReportWarning("minimum OS version load command with " 5851 "invalid (0) version found."); 5852 } 5853 } 5854 } 5855 offset = load_cmd_offset + lc.cmdsize; 5856 } 5857 5858 if (!m_sdk_versions.hasValue()) { 5859 offset = MachHeaderSizeFromMagic(m_header.magic); 5860 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5861 const lldb::offset_t load_cmd_offset = offset; 5862 5863 version_min_command lc; 5864 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5865 break; 5866 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5867 // struct build_version_command { 5868 // uint32_t cmd; /* LC_BUILD_VERSION */ 5869 // uint32_t cmdsize; /* sizeof(struct 5870 // build_version_command) plus */ 5871 // /* ntools * sizeof(struct 5872 // build_tool_version) */ 5873 // uint32_t platform; /* platform */ 5874 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5875 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded 5876 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number 5877 // of tool entries following this */ 5878 // }; 5879 5880 offset += 4; // skip platform 5881 uint32_t minos = m_data.GetU32(&offset); 5882 5883 const uint32_t xxxx = minos >> 16; 5884 const uint32_t yy = (minos >> 8) & 0xffu; 5885 const uint32_t zz = minos & 0xffu; 5886 if (xxxx) { 5887 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 5888 break; 5889 } 5890 } 5891 offset = load_cmd_offset + lc.cmdsize; 5892 } 5893 } 5894 5895 if (!m_sdk_versions.hasValue()) 5896 m_sdk_versions = llvm::VersionTuple(); 5897 } 5898 5899 return m_sdk_versions.getValue(); 5900 } 5901 5902 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5903 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5904 } 5905 5906 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5907 return m_allow_assembly_emulation_unwind_plans; 5908 } 5909 5910 // PluginInterface protocol 5911 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5912 return GetPluginNameStatic(); 5913 } 5914 5915 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5916 5917 Section *ObjectFileMachO::GetMachHeaderSection() { 5918 // Find the first address of the mach header which is the first non-zero file 5919 // sized section whose file offset is zero. This is the base file address of 5920 // the mach-o file which can be subtracted from the vmaddr of the other 5921 // segments found in memory and added to the load address 5922 ModuleSP module_sp = GetModule(); 5923 if (!module_sp) 5924 return nullptr; 5925 SectionList *section_list = GetSectionList(); 5926 if (!section_list) 5927 return nullptr; 5928 const size_t num_sections = section_list->GetSize(); 5929 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5930 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5931 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 5932 return section; 5933 } 5934 return nullptr; 5935 } 5936 5937 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 5938 if (!section) 5939 return false; 5940 const bool is_dsym = (m_header.filetype == MH_DSYM); 5941 if (section->GetFileSize() == 0 && !is_dsym) 5942 return false; 5943 if (section->IsThreadSpecific()) 5944 return false; 5945 if (GetModule().get() != section->GetModule().get()) 5946 return false; 5947 // Be careful with __LINKEDIT and __DWARF segments 5948 if (section->GetName() == GetSegmentNameLINKEDIT() || 5949 section->GetName() == GetSegmentNameDWARF()) { 5950 // Only map __LINKEDIT and __DWARF if we have an in memory image and 5951 // this isn't a kernel binary like a kext or mach_kernel. 5952 const bool is_memory_image = (bool)m_process_wp.lock(); 5953 const Strata strata = GetStrata(); 5954 if (is_memory_image == false || strata == eStrataKernel) 5955 return false; 5956 } 5957 return true; 5958 } 5959 5960 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 5961 lldb::addr_t header_load_address, const Section *header_section, 5962 const Section *section) { 5963 ModuleSP module_sp = GetModule(); 5964 if (module_sp && header_section && section && 5965 header_load_address != LLDB_INVALID_ADDRESS) { 5966 lldb::addr_t file_addr = header_section->GetFileAddress(); 5967 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 5968 return section->GetFileAddress() - file_addr + header_load_address; 5969 } 5970 return LLDB_INVALID_ADDRESS; 5971 } 5972 5973 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 5974 bool value_is_offset) { 5975 ModuleSP module_sp = GetModule(); 5976 if (!module_sp) 5977 return false; 5978 5979 SectionList *section_list = GetSectionList(); 5980 if (!section_list) 5981 return false; 5982 5983 size_t num_loaded_sections = 0; 5984 const size_t num_sections = section_list->GetSize(); 5985 5986 if (value_is_offset) { 5987 // "value" is an offset to apply to each top level segment 5988 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5989 // Iterate through the object file sections to find all of the 5990 // sections that size on disk (to avoid __PAGEZERO) and load them 5991 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5992 if (SectionIsLoadable(section_sp.get())) 5993 if (target.GetSectionLoadList().SetSectionLoadAddress( 5994 section_sp, section_sp->GetFileAddress() + value)) 5995 ++num_loaded_sections; 5996 } 5997 } else { 5998 // "value" is the new base address of the mach_header, adjust each 5999 // section accordingly 6000 6001 Section *mach_header_section = GetMachHeaderSection(); 6002 if (mach_header_section) { 6003 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6004 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6005 6006 lldb::addr_t section_load_addr = 6007 CalculateSectionLoadAddressForMemoryImage( 6008 value, mach_header_section, section_sp.get()); 6009 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6010 if (target.GetSectionLoadList().SetSectionLoadAddress( 6011 section_sp, section_load_addr)) 6012 ++num_loaded_sections; 6013 } 6014 } 6015 } 6016 } 6017 return num_loaded_sections > 0; 6018 } 6019 6020 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6021 const FileSpec &outfile, Status &error) { 6022 if (!process_sp) 6023 return false; 6024 6025 Target &target = process_sp->GetTarget(); 6026 const ArchSpec target_arch = target.GetArchitecture(); 6027 const llvm::Triple &target_triple = target_arch.GetTriple(); 6028 if (target_triple.getVendor() == llvm::Triple::Apple && 6029 (target_triple.getOS() == llvm::Triple::MacOSX || 6030 target_triple.getOS() == llvm::Triple::IOS || 6031 target_triple.getOS() == llvm::Triple::WatchOS || 6032 target_triple.getOS() == llvm::Triple::TvOS)) { 6033 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) 6034 // { 6035 bool make_core = false; 6036 switch (target_arch.GetMachine()) { 6037 case llvm::Triple::aarch64: 6038 case llvm::Triple::aarch64_32: 6039 case llvm::Triple::arm: 6040 case llvm::Triple::thumb: 6041 case llvm::Triple::x86: 6042 case llvm::Triple::x86_64: 6043 make_core = true; 6044 break; 6045 default: 6046 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6047 target_triple.str().c_str()); 6048 break; 6049 } 6050 6051 if (make_core) { 6052 std::vector<segment_command_64> segment_load_commands; 6053 // uint32_t range_info_idx = 0; 6054 MemoryRegionInfo range_info; 6055 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6056 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6057 const ByteOrder byte_order = target_arch.GetByteOrder(); 6058 if (range_error.Success()) { 6059 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6060 const addr_t addr = range_info.GetRange().GetRangeBase(); 6061 const addr_t size = range_info.GetRange().GetByteSize(); 6062 6063 if (size == 0) 6064 break; 6065 6066 // Calculate correct protections 6067 uint32_t prot = 0; 6068 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6069 prot |= VM_PROT_READ; 6070 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6071 prot |= VM_PROT_WRITE; 6072 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6073 prot |= VM_PROT_EXECUTE; 6074 6075 if (prot != 0) { 6076 uint32_t cmd_type = LC_SEGMENT_64; 6077 uint32_t segment_size = sizeof(segment_command_64); 6078 if (addr_byte_size == 4) { 6079 cmd_type = LC_SEGMENT; 6080 segment_size = sizeof(segment_command); 6081 } 6082 segment_command_64 segment = { 6083 cmd_type, // uint32_t cmd; 6084 segment_size, // uint32_t cmdsize; 6085 {0}, // char segname[16]; 6086 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6087 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6088 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6089 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6090 prot, // uint32_t maxprot; 6091 prot, // uint32_t initprot; 6092 0, // uint32_t nsects; 6093 0}; // uint32_t flags; 6094 segment_load_commands.push_back(segment); 6095 } else { 6096 // No protections and a size of 1 used to be returned from old 6097 // debugservers when we asked about a region that was past the 6098 // last memory region and it indicates the end... 6099 if (size == 1) 6100 break; 6101 } 6102 6103 range_error = process_sp->GetMemoryRegionInfo( 6104 range_info.GetRange().GetRangeEnd(), range_info); 6105 if (range_error.Fail()) 6106 break; 6107 } 6108 6109 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6110 6111 mach_header_64 mach_header; 6112 if (addr_byte_size == 8) { 6113 mach_header.magic = MH_MAGIC_64; 6114 } else { 6115 mach_header.magic = MH_MAGIC; 6116 } 6117 mach_header.cputype = target_arch.GetMachOCPUType(); 6118 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6119 mach_header.filetype = MH_CORE; 6120 mach_header.ncmds = segment_load_commands.size(); 6121 mach_header.flags = 0; 6122 mach_header.reserved = 0; 6123 ThreadList &thread_list = process_sp->GetThreadList(); 6124 const uint32_t num_threads = thread_list.GetSize(); 6125 6126 // Make an array of LC_THREAD data items. Each one contains the 6127 // contents of the LC_THREAD load command. The data doesn't contain 6128 // the load command + load command size, we will add the load command 6129 // and load command size as we emit the data. 6130 std::vector<StreamString> LC_THREAD_datas(num_threads); 6131 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6132 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6133 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6134 LC_THREAD_data.SetByteOrder(byte_order); 6135 } 6136 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) { 6137 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6138 if (thread_sp) { 6139 switch (mach_header.cputype) { 6140 case llvm::MachO::CPU_TYPE_ARM64: 6141 case llvm::MachO::CPU_TYPE_ARM64_32: 6142 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6143 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6144 break; 6145 6146 case llvm::MachO::CPU_TYPE_ARM: 6147 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6148 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6149 break; 6150 6151 case llvm::MachO::CPU_TYPE_I386: 6152 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6153 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6154 break; 6155 6156 case llvm::MachO::CPU_TYPE_X86_64: 6157 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6158 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6159 break; 6160 } 6161 } 6162 } 6163 6164 // The size of the load command is the size of the segments... 6165 if (addr_byte_size == 8) { 6166 mach_header.sizeofcmds = 6167 segment_load_commands.size() * sizeof(struct segment_command_64); 6168 } else { 6169 mach_header.sizeofcmds = 6170 segment_load_commands.size() * sizeof(struct segment_command); 6171 } 6172 6173 // and the size of all LC_THREAD load command 6174 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6175 ++mach_header.ncmds; 6176 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6177 } 6178 6179 // Write the mach header 6180 buffer.PutHex32(mach_header.magic); 6181 buffer.PutHex32(mach_header.cputype); 6182 buffer.PutHex32(mach_header.cpusubtype); 6183 buffer.PutHex32(mach_header.filetype); 6184 buffer.PutHex32(mach_header.ncmds); 6185 buffer.PutHex32(mach_header.sizeofcmds); 6186 buffer.PutHex32(mach_header.flags); 6187 if (addr_byte_size == 8) { 6188 buffer.PutHex32(mach_header.reserved); 6189 } 6190 6191 // Skip the mach header and all load commands and align to the next 6192 // 0x1000 byte boundary 6193 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6194 if (file_offset & 0x00000fff) { 6195 file_offset += 0x00001000ull; 6196 file_offset &= (~0x00001000ull + 1); 6197 } 6198 6199 for (auto &segment : segment_load_commands) { 6200 segment.fileoff = file_offset; 6201 file_offset += segment.filesize; 6202 } 6203 6204 // Write out all of the LC_THREAD load commands 6205 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6206 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6207 buffer.PutHex32(LC_THREAD); 6208 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6209 buffer.Write(LC_THREAD_data.GetString().data(), LC_THREAD_data_size); 6210 } 6211 6212 // Write out all of the segment load commands 6213 for (const auto &segment : segment_load_commands) { 6214 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6215 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6216 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6217 segment.cmd, segment.cmdsize, segment.vmaddr, 6218 segment.vmaddr + segment.vmsize, segment.fileoff, 6219 segment.filesize, segment.maxprot, segment.initprot, 6220 segment.nsects, segment.flags); 6221 6222 buffer.PutHex32(segment.cmd); 6223 buffer.PutHex32(segment.cmdsize); 6224 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6225 if (addr_byte_size == 8) { 6226 buffer.PutHex64(segment.vmaddr); 6227 buffer.PutHex64(segment.vmsize); 6228 buffer.PutHex64(segment.fileoff); 6229 buffer.PutHex64(segment.filesize); 6230 } else { 6231 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6232 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6233 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6234 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6235 } 6236 buffer.PutHex32(segment.maxprot); 6237 buffer.PutHex32(segment.initprot); 6238 buffer.PutHex32(segment.nsects); 6239 buffer.PutHex32(segment.flags); 6240 } 6241 6242 std::string core_file_path(outfile.GetPath()); 6243 auto core_file = FileSystem::Instance().Open( 6244 outfile, File::eOpenOptionWrite | File::eOpenOptionTruncate | 6245 File::eOpenOptionCanCreate); 6246 if (!core_file) { 6247 error = core_file.takeError(); 6248 } else { 6249 // Read 1 page at a time 6250 uint8_t bytes[0x1000]; 6251 // Write the mach header and load commands out to the core file 6252 size_t bytes_written = buffer.GetString().size(); 6253 error = 6254 core_file.get()->Write(buffer.GetString().data(), bytes_written); 6255 if (error.Success()) { 6256 // Now write the file data for all memory segments in the process 6257 for (const auto &segment : segment_load_commands) { 6258 if (core_file.get()->SeekFromStart(segment.fileoff) == -1) { 6259 error.SetErrorStringWithFormat( 6260 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6261 segment.fileoff, core_file_path.c_str()); 6262 break; 6263 } 6264 6265 printf("Saving %" PRId64 6266 " bytes of data for memory region at 0x%" PRIx64 "\n", 6267 segment.vmsize, segment.vmaddr); 6268 addr_t bytes_left = segment.vmsize; 6269 addr_t addr = segment.vmaddr; 6270 Status memory_read_error; 6271 while (bytes_left > 0 && error.Success()) { 6272 const size_t bytes_to_read = 6273 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6274 6275 // In a savecore setting, we don't really care about caching, 6276 // as the data is dumped and very likely never read again, 6277 // so we call ReadMemoryFromInferior to bypass it. 6278 const size_t bytes_read = process_sp->ReadMemoryFromInferior( 6279 addr, bytes, bytes_to_read, memory_read_error); 6280 6281 if (bytes_read == bytes_to_read) { 6282 size_t bytes_written = bytes_read; 6283 error = core_file.get()->Write(bytes, bytes_written); 6284 bytes_left -= bytes_read; 6285 addr += bytes_read; 6286 } else { 6287 // Some pages within regions are not readable, those should 6288 // be zero filled 6289 memset(bytes, 0, bytes_to_read); 6290 size_t bytes_written = bytes_to_read; 6291 error = core_file.get()->Write(bytes, bytes_written); 6292 bytes_left -= bytes_to_read; 6293 addr += bytes_to_read; 6294 } 6295 } 6296 } 6297 } 6298 } 6299 } else { 6300 error.SetErrorString( 6301 "process doesn't support getting memory region info"); 6302 } 6303 } 6304 return true; // This is the right plug to handle saving core files for 6305 // this process 6306 } 6307 return false; 6308 } 6309