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; 125 ++i) 126 data.PutChar(0); 127 } 128 return; 129 } 130 } 131 // Just write zeros if all else fails 132 for (size_t i = 0; i < reg_byte_size; ++i) 133 data.PutChar(0); 134 } 135 136 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 137 public: 138 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 139 const DataExtractor &data) 140 : RegisterContextDarwin_x86_64(thread, 0) { 141 SetRegisterDataFrom_LC_THREAD(data); 142 } 143 144 void InvalidateAllRegisters() override { 145 // Do nothing... registers are always valid... 146 } 147 148 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 149 lldb::offset_t offset = 0; 150 SetError(GPRRegSet, Read, -1); 151 SetError(FPURegSet, Read, -1); 152 SetError(EXCRegSet, Read, -1); 153 bool done = false; 154 155 while (!done) { 156 int flavor = data.GetU32(&offset); 157 if (flavor == 0) 158 done = true; 159 else { 160 uint32_t i; 161 uint32_t count = data.GetU32(&offset); 162 switch (flavor) { 163 case GPRRegSet: 164 for (i = 0; i < count; ++i) 165 (&gpr.rax)[i] = data.GetU64(&offset); 166 SetError(GPRRegSet, Read, 0); 167 done = true; 168 169 break; 170 case FPURegSet: 171 // TODO: fill in FPU regs.... 172 // SetError (FPURegSet, Read, -1); 173 done = true; 174 175 break; 176 case EXCRegSet: 177 exc.trapno = data.GetU32(&offset); 178 exc.err = data.GetU32(&offset); 179 exc.faultvaddr = data.GetU64(&offset); 180 SetError(EXCRegSet, Read, 0); 181 done = true; 182 break; 183 case 7: 184 case 8: 185 case 9: 186 // fancy flavors that encapsulate of the above flavors... 187 break; 188 189 default: 190 done = true; 191 break; 192 } 193 } 194 } 195 } 196 197 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 198 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 199 if (reg_ctx_sp) { 200 RegisterContext *reg_ctx = reg_ctx_sp.get(); 201 202 data.PutHex32(GPRRegSet); // Flavor 203 data.PutHex32(GPRWordCount); 204 PrintRegisterValue(reg_ctx, "rax", nullptr, 8, data); 205 PrintRegisterValue(reg_ctx, "rbx", nullptr, 8, data); 206 PrintRegisterValue(reg_ctx, "rcx", nullptr, 8, data); 207 PrintRegisterValue(reg_ctx, "rdx", nullptr, 8, data); 208 PrintRegisterValue(reg_ctx, "rdi", nullptr, 8, data); 209 PrintRegisterValue(reg_ctx, "rsi", nullptr, 8, data); 210 PrintRegisterValue(reg_ctx, "rbp", nullptr, 8, data); 211 PrintRegisterValue(reg_ctx, "rsp", nullptr, 8, data); 212 PrintRegisterValue(reg_ctx, "r8", nullptr, 8, data); 213 PrintRegisterValue(reg_ctx, "r9", nullptr, 8, data); 214 PrintRegisterValue(reg_ctx, "r10", nullptr, 8, data); 215 PrintRegisterValue(reg_ctx, "r11", nullptr, 8, data); 216 PrintRegisterValue(reg_ctx, "r12", nullptr, 8, data); 217 PrintRegisterValue(reg_ctx, "r13", nullptr, 8, data); 218 PrintRegisterValue(reg_ctx, "r14", nullptr, 8, data); 219 PrintRegisterValue(reg_ctx, "r15", nullptr, 8, data); 220 PrintRegisterValue(reg_ctx, "rip", nullptr, 8, data); 221 PrintRegisterValue(reg_ctx, "rflags", nullptr, 8, data); 222 PrintRegisterValue(reg_ctx, "cs", nullptr, 8, data); 223 PrintRegisterValue(reg_ctx, "fs", nullptr, 8, data); 224 PrintRegisterValue(reg_ctx, "gs", nullptr, 8, data); 225 226 // // Write out the FPU registers 227 // const size_t fpu_byte_size = sizeof(FPU); 228 // size_t bytes_written = 0; 229 // data.PutHex32 (FPURegSet); 230 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 231 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 232 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 233 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 234 // data); // uint16_t fcw; // "fctrl" 235 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 236 // data); // uint16_t fsw; // "fstat" 237 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 238 // data); // uint8_t ftw; // "ftag" 239 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 240 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 241 // data); // uint16_t fop; // "fop" 242 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 243 // data); // uint32_t ip; // "fioff" 244 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 245 // data); // uint16_t cs; // "fiseg" 246 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 247 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 248 // data); // uint32_t dp; // "fooff" 249 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 250 // data); // uint16_t ds; // "foseg" 251 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 252 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 253 // data); // uint32_t mxcsr; 254 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 255 // 4, data);// uint32_t mxcsrmask; 256 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 257 // sizeof(MMSReg), data); 258 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 259 // sizeof(MMSReg), data); 260 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 261 // sizeof(MMSReg), data); 262 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 263 // sizeof(MMSReg), data); 264 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 265 // sizeof(MMSReg), data); 266 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 267 // sizeof(MMSReg), data); 268 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 269 // sizeof(MMSReg), data); 270 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 271 // sizeof(MMSReg), data); 272 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 273 // sizeof(XMMReg), data); 274 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 275 // sizeof(XMMReg), data); 276 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 277 // sizeof(XMMReg), data); 278 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 279 // sizeof(XMMReg), data); 280 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 281 // sizeof(XMMReg), data); 282 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 283 // sizeof(XMMReg), data); 284 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 285 // sizeof(XMMReg), data); 286 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 287 // sizeof(XMMReg), data); 288 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 289 // sizeof(XMMReg), data); 290 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 291 // sizeof(XMMReg), data); 292 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 293 // sizeof(XMMReg), data); 294 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 295 // sizeof(XMMReg), data); 296 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 297 // sizeof(XMMReg), data); 298 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 299 // sizeof(XMMReg), data); 300 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 301 // sizeof(XMMReg), data); 302 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 303 // sizeof(XMMReg), data); 304 // 305 // // Fill rest with zeros 306 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 307 // i) 308 // data.PutChar(0); 309 310 // Write out the EXC registers 311 data.PutHex32(EXCRegSet); 312 data.PutHex32(EXCWordCount); 313 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 314 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 315 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 8, data); 316 return true; 317 } 318 return false; 319 } 320 321 protected: 322 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 323 324 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 325 326 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 327 328 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 329 return 0; 330 } 331 332 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 333 return 0; 334 } 335 336 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 337 return 0; 338 } 339 }; 340 341 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 342 public: 343 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 344 const DataExtractor &data) 345 : RegisterContextDarwin_i386(thread, 0) { 346 SetRegisterDataFrom_LC_THREAD(data); 347 } 348 349 void InvalidateAllRegisters() override { 350 // Do nothing... registers are always valid... 351 } 352 353 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 354 lldb::offset_t offset = 0; 355 SetError(GPRRegSet, Read, -1); 356 SetError(FPURegSet, Read, -1); 357 SetError(EXCRegSet, Read, -1); 358 bool done = false; 359 360 while (!done) { 361 int flavor = data.GetU32(&offset); 362 if (flavor == 0) 363 done = true; 364 else { 365 uint32_t i; 366 uint32_t count = data.GetU32(&offset); 367 switch (flavor) { 368 case GPRRegSet: 369 for (i = 0; i < count; ++i) 370 (&gpr.eax)[i] = data.GetU32(&offset); 371 SetError(GPRRegSet, Read, 0); 372 done = true; 373 374 break; 375 case FPURegSet: 376 // TODO: fill in FPU regs.... 377 // SetError (FPURegSet, Read, -1); 378 done = true; 379 380 break; 381 case EXCRegSet: 382 exc.trapno = data.GetU32(&offset); 383 exc.err = data.GetU32(&offset); 384 exc.faultvaddr = data.GetU32(&offset); 385 SetError(EXCRegSet, Read, 0); 386 done = true; 387 break; 388 case 7: 389 case 8: 390 case 9: 391 // fancy flavors that encapsulate of the above flavors... 392 break; 393 394 default: 395 done = true; 396 break; 397 } 398 } 399 } 400 } 401 402 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 403 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 404 if (reg_ctx_sp) { 405 RegisterContext *reg_ctx = reg_ctx_sp.get(); 406 407 data.PutHex32(GPRRegSet); // Flavor 408 data.PutHex32(GPRWordCount); 409 PrintRegisterValue(reg_ctx, "eax", nullptr, 4, data); 410 PrintRegisterValue(reg_ctx, "ebx", nullptr, 4, data); 411 PrintRegisterValue(reg_ctx, "ecx", nullptr, 4, data); 412 PrintRegisterValue(reg_ctx, "edx", nullptr, 4, data); 413 PrintRegisterValue(reg_ctx, "edi", nullptr, 4, data); 414 PrintRegisterValue(reg_ctx, "esi", nullptr, 4, data); 415 PrintRegisterValue(reg_ctx, "ebp", nullptr, 4, data); 416 PrintRegisterValue(reg_ctx, "esp", nullptr, 4, data); 417 PrintRegisterValue(reg_ctx, "ss", nullptr, 4, data); 418 PrintRegisterValue(reg_ctx, "eflags", nullptr, 4, data); 419 PrintRegisterValue(reg_ctx, "eip", nullptr, 4, data); 420 PrintRegisterValue(reg_ctx, "cs", nullptr, 4, data); 421 PrintRegisterValue(reg_ctx, "ds", nullptr, 4, data); 422 PrintRegisterValue(reg_ctx, "es", nullptr, 4, data); 423 PrintRegisterValue(reg_ctx, "fs", nullptr, 4, data); 424 PrintRegisterValue(reg_ctx, "gs", nullptr, 4, data); 425 426 // Write out the EXC registers 427 data.PutHex32(EXCRegSet); 428 data.PutHex32(EXCWordCount); 429 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 430 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 431 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 4, data); 432 return true; 433 } 434 return false; 435 } 436 437 protected: 438 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 439 440 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 441 442 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 443 444 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 445 return 0; 446 } 447 448 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 449 return 0; 450 } 451 452 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 453 return 0; 454 } 455 }; 456 457 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 458 public: 459 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 460 const DataExtractor &data) 461 : RegisterContextDarwin_arm(thread, 0) { 462 SetRegisterDataFrom_LC_THREAD(data); 463 } 464 465 void InvalidateAllRegisters() override { 466 // Do nothing... registers are always valid... 467 } 468 469 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 470 lldb::offset_t offset = 0; 471 SetError(GPRRegSet, Read, -1); 472 SetError(FPURegSet, Read, -1); 473 SetError(EXCRegSet, Read, -1); 474 bool done = false; 475 476 while (!done) { 477 int flavor = data.GetU32(&offset); 478 uint32_t count = data.GetU32(&offset); 479 lldb::offset_t next_thread_state = offset + (count * 4); 480 switch (flavor) { 481 case GPRAltRegSet: 482 case GPRRegSet: 483 for (uint32_t i = 0; i < count; ++i) { 484 gpr.r[i] = data.GetU32(&offset); 485 } 486 487 // Note that gpr.cpsr is also copied by the above loop; this loop 488 // technically extends one element past the end of the gpr.r[] array. 489 490 SetError(GPRRegSet, Read, 0); 491 offset = next_thread_state; 492 break; 493 494 case FPURegSet: { 495 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 496 const int fpu_reg_buf_size = sizeof(fpu.floats); 497 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 498 fpu_reg_buf) == fpu_reg_buf_size) { 499 offset += fpu_reg_buf_size; 500 fpu.fpscr = data.GetU32(&offset); 501 SetError(FPURegSet, Read, 0); 502 } else { 503 done = true; 504 } 505 } 506 offset = next_thread_state; 507 break; 508 509 case EXCRegSet: 510 if (count == 3) { 511 exc.exception = data.GetU32(&offset); 512 exc.fsr = data.GetU32(&offset); 513 exc.far = data.GetU32(&offset); 514 SetError(EXCRegSet, Read, 0); 515 } 516 done = true; 517 offset = next_thread_state; 518 break; 519 520 // Unknown register set flavor, stop trying to parse. 521 default: 522 done = true; 523 } 524 } 525 } 526 527 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 528 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 529 if (reg_ctx_sp) { 530 RegisterContext *reg_ctx = reg_ctx_sp.get(); 531 532 data.PutHex32(GPRRegSet); // Flavor 533 data.PutHex32(GPRWordCount); 534 PrintRegisterValue(reg_ctx, "r0", nullptr, 4, data); 535 PrintRegisterValue(reg_ctx, "r1", nullptr, 4, data); 536 PrintRegisterValue(reg_ctx, "r2", nullptr, 4, data); 537 PrintRegisterValue(reg_ctx, "r3", nullptr, 4, data); 538 PrintRegisterValue(reg_ctx, "r4", nullptr, 4, data); 539 PrintRegisterValue(reg_ctx, "r5", nullptr, 4, data); 540 PrintRegisterValue(reg_ctx, "r6", nullptr, 4, data); 541 PrintRegisterValue(reg_ctx, "r7", nullptr, 4, data); 542 PrintRegisterValue(reg_ctx, "r8", nullptr, 4, data); 543 PrintRegisterValue(reg_ctx, "r9", nullptr, 4, data); 544 PrintRegisterValue(reg_ctx, "r10", nullptr, 4, data); 545 PrintRegisterValue(reg_ctx, "r11", nullptr, 4, data); 546 PrintRegisterValue(reg_ctx, "r12", nullptr, 4, data); 547 PrintRegisterValue(reg_ctx, "sp", nullptr, 4, data); 548 PrintRegisterValue(reg_ctx, "lr", nullptr, 4, data); 549 PrintRegisterValue(reg_ctx, "pc", nullptr, 4, data); 550 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 551 552 // Write out the EXC registers 553 // data.PutHex32 (EXCRegSet); 554 // data.PutHex32 (EXCWordCount); 555 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 556 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 557 // WriteRegister (reg_ctx, "far", NULL, 4, data); 558 return true; 559 } 560 return false; 561 } 562 563 protected: 564 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 565 566 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 567 568 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 569 570 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 571 572 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 573 return 0; 574 } 575 576 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 577 return 0; 578 } 579 580 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 581 return 0; 582 } 583 584 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 585 return -1; 586 } 587 }; 588 589 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 590 public: 591 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 592 const DataExtractor &data) 593 : RegisterContextDarwin_arm64(thread, 0) { 594 SetRegisterDataFrom_LC_THREAD(data); 595 } 596 597 void InvalidateAllRegisters() override { 598 // Do nothing... registers are always valid... 599 } 600 601 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 602 lldb::offset_t offset = 0; 603 SetError(GPRRegSet, Read, -1); 604 SetError(FPURegSet, Read, -1); 605 SetError(EXCRegSet, Read, -1); 606 bool done = false; 607 while (!done) { 608 int flavor = data.GetU32(&offset); 609 uint32_t count = data.GetU32(&offset); 610 lldb::offset_t next_thread_state = offset + (count * 4); 611 switch (flavor) { 612 case GPRRegSet: 613 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 614 // 32-bit register) 615 if (count >= (33 * 2) + 1) { 616 for (uint32_t i = 0; i < 29; ++i) 617 gpr.x[i] = data.GetU64(&offset); 618 gpr.fp = data.GetU64(&offset); 619 gpr.lr = data.GetU64(&offset); 620 gpr.sp = data.GetU64(&offset); 621 gpr.pc = data.GetU64(&offset); 622 gpr.cpsr = data.GetU32(&offset); 623 SetError(GPRRegSet, Read, 0); 624 } 625 offset = next_thread_state; 626 break; 627 case FPURegSet: { 628 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 629 const int fpu_reg_buf_size = sizeof(fpu); 630 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 631 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 632 fpu_reg_buf) == fpu_reg_buf_size) { 633 SetError(FPURegSet, Read, 0); 634 } else { 635 done = true; 636 } 637 } 638 offset = next_thread_state; 639 break; 640 case EXCRegSet: 641 if (count == 4) { 642 exc.far = data.GetU64(&offset); 643 exc.esr = data.GetU32(&offset); 644 exc.exception = data.GetU32(&offset); 645 SetError(EXCRegSet, Read, 0); 646 } 647 offset = next_thread_state; 648 break; 649 default: 650 done = true; 651 break; 652 } 653 } 654 } 655 656 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 657 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 658 if (reg_ctx_sp) { 659 RegisterContext *reg_ctx = reg_ctx_sp.get(); 660 661 data.PutHex32(GPRRegSet); // Flavor 662 data.PutHex32(GPRWordCount); 663 PrintRegisterValue(reg_ctx, "x0", nullptr, 8, data); 664 PrintRegisterValue(reg_ctx, "x1", nullptr, 8, data); 665 PrintRegisterValue(reg_ctx, "x2", nullptr, 8, data); 666 PrintRegisterValue(reg_ctx, "x3", nullptr, 8, data); 667 PrintRegisterValue(reg_ctx, "x4", nullptr, 8, data); 668 PrintRegisterValue(reg_ctx, "x5", nullptr, 8, data); 669 PrintRegisterValue(reg_ctx, "x6", nullptr, 8, data); 670 PrintRegisterValue(reg_ctx, "x7", nullptr, 8, data); 671 PrintRegisterValue(reg_ctx, "x8", nullptr, 8, data); 672 PrintRegisterValue(reg_ctx, "x9", nullptr, 8, data); 673 PrintRegisterValue(reg_ctx, "x10", nullptr, 8, data); 674 PrintRegisterValue(reg_ctx, "x11", nullptr, 8, data); 675 PrintRegisterValue(reg_ctx, "x12", nullptr, 8, data); 676 PrintRegisterValue(reg_ctx, "x13", nullptr, 8, data); 677 PrintRegisterValue(reg_ctx, "x14", nullptr, 8, data); 678 PrintRegisterValue(reg_ctx, "x15", nullptr, 8, data); 679 PrintRegisterValue(reg_ctx, "x16", nullptr, 8, data); 680 PrintRegisterValue(reg_ctx, "x17", nullptr, 8, data); 681 PrintRegisterValue(reg_ctx, "x18", nullptr, 8, data); 682 PrintRegisterValue(reg_ctx, "x19", nullptr, 8, data); 683 PrintRegisterValue(reg_ctx, "x20", nullptr, 8, data); 684 PrintRegisterValue(reg_ctx, "x21", nullptr, 8, data); 685 PrintRegisterValue(reg_ctx, "x22", nullptr, 8, data); 686 PrintRegisterValue(reg_ctx, "x23", nullptr, 8, data); 687 PrintRegisterValue(reg_ctx, "x24", nullptr, 8, data); 688 PrintRegisterValue(reg_ctx, "x25", nullptr, 8, data); 689 PrintRegisterValue(reg_ctx, "x26", nullptr, 8, data); 690 PrintRegisterValue(reg_ctx, "x27", nullptr, 8, data); 691 PrintRegisterValue(reg_ctx, "x28", nullptr, 8, data); 692 PrintRegisterValue(reg_ctx, "fp", nullptr, 8, data); 693 PrintRegisterValue(reg_ctx, "lr", nullptr, 8, data); 694 PrintRegisterValue(reg_ctx, "sp", nullptr, 8, data); 695 PrintRegisterValue(reg_ctx, "pc", nullptr, 8, data); 696 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 697 698 // Write out the EXC registers 699 // data.PutHex32 (EXCRegSet); 700 // data.PutHex32 (EXCWordCount); 701 // WriteRegister (reg_ctx, "far", NULL, 8, data); 702 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 703 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 704 return true; 705 } 706 return false; 707 } 708 709 protected: 710 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 711 712 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 713 714 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 715 716 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 717 718 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 719 return 0; 720 } 721 722 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 723 return 0; 724 } 725 726 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 727 return 0; 728 } 729 730 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 731 return -1; 732 } 733 }; 734 735 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 736 switch (magic) { 737 case MH_MAGIC: 738 case MH_CIGAM: 739 return sizeof(struct mach_header); 740 741 case MH_MAGIC_64: 742 case MH_CIGAM_64: 743 return sizeof(struct mach_header_64); 744 break; 745 746 default: 747 break; 748 } 749 return 0; 750 } 751 752 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 753 754 char ObjectFileMachO::ID; 755 756 void ObjectFileMachO::Initialize() { 757 PluginManager::RegisterPlugin( 758 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 759 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 760 } 761 762 void ObjectFileMachO::Terminate() { 763 PluginManager::UnregisterPlugin(CreateInstance); 764 } 765 766 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 767 static ConstString g_name("mach-o"); 768 return g_name; 769 } 770 771 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 772 return "Mach-o object file reader (32 and 64 bit)"; 773 } 774 775 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 776 DataBufferSP &data_sp, 777 lldb::offset_t data_offset, 778 const FileSpec *file, 779 lldb::offset_t file_offset, 780 lldb::offset_t length) { 781 if (!data_sp) { 782 data_sp = MapFileData(*file, length, file_offset); 783 if (!data_sp) 784 return nullptr; 785 data_offset = 0; 786 } 787 788 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 789 return nullptr; 790 791 // Update the data to contain the entire file if it doesn't already 792 if (data_sp->GetByteSize() < length) { 793 data_sp = MapFileData(*file, length, file_offset); 794 if (!data_sp) 795 return nullptr; 796 data_offset = 0; 797 } 798 auto objfile_up = std::make_unique<ObjectFileMachO>( 799 module_sp, data_sp, data_offset, file, file_offset, length); 800 if (!objfile_up || !objfile_up->ParseHeader()) 801 return nullptr; 802 803 return objfile_up.release(); 804 } 805 806 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 807 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 808 const ProcessSP &process_sp, lldb::addr_t header_addr) { 809 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 810 std::unique_ptr<ObjectFile> objfile_up( 811 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 812 if (objfile_up.get() && objfile_up->ParseHeader()) 813 return objfile_up.release(); 814 } 815 return nullptr; 816 } 817 818 size_t ObjectFileMachO::GetModuleSpecifications( 819 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 820 lldb::offset_t data_offset, lldb::offset_t file_offset, 821 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 822 const size_t initial_count = specs.GetSize(); 823 824 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 825 DataExtractor data; 826 data.SetData(data_sp); 827 llvm::MachO::mach_header header; 828 if (ParseHeader(data, &data_offset, header)) { 829 size_t header_and_load_cmds = 830 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 831 if (header_and_load_cmds >= data_sp->GetByteSize()) { 832 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 833 data.SetData(data_sp); 834 data_offset = MachHeaderSizeFromMagic(header.magic); 835 } 836 if (data_sp) { 837 ModuleSpec base_spec; 838 base_spec.GetFileSpec() = file; 839 base_spec.SetObjectOffset(file_offset); 840 base_spec.SetObjectSize(length); 841 GetAllArchSpecs(header, data, data_offset, base_spec, specs); 842 } 843 } 844 } 845 return specs.GetSize() - initial_count; 846 } 847 848 ConstString ObjectFileMachO::GetSegmentNameTEXT() { 849 static ConstString g_segment_name_TEXT("__TEXT"); 850 return g_segment_name_TEXT; 851 } 852 853 ConstString ObjectFileMachO::GetSegmentNameDATA() { 854 static ConstString g_segment_name_DATA("__DATA"); 855 return g_segment_name_DATA; 856 } 857 858 ConstString ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 859 static ConstString g_segment_name("__DATA_DIRTY"); 860 return g_segment_name; 861 } 862 863 ConstString ObjectFileMachO::GetSegmentNameDATA_CONST() { 864 static ConstString g_segment_name("__DATA_CONST"); 865 return g_segment_name; 866 } 867 868 ConstString ObjectFileMachO::GetSegmentNameOBJC() { 869 static ConstString g_segment_name_OBJC("__OBJC"); 870 return g_segment_name_OBJC; 871 } 872 873 ConstString ObjectFileMachO::GetSegmentNameLINKEDIT() { 874 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 875 return g_section_name_LINKEDIT; 876 } 877 878 ConstString ObjectFileMachO::GetSegmentNameDWARF() { 879 static ConstString g_section_name("__DWARF"); 880 return g_section_name; 881 } 882 883 ConstString ObjectFileMachO::GetSectionNameEHFrame() { 884 static ConstString g_section_name_eh_frame("__eh_frame"); 885 return g_section_name_eh_frame; 886 } 887 888 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 889 lldb::addr_t data_offset, 890 lldb::addr_t data_length) { 891 DataExtractor data; 892 data.SetData(data_sp, data_offset, data_length); 893 lldb::offset_t offset = 0; 894 uint32_t magic = data.GetU32(&offset); 895 return MachHeaderSizeFromMagic(magic) != 0; 896 } 897 898 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 899 DataBufferSP &data_sp, 900 lldb::offset_t data_offset, 901 const FileSpec *file, 902 lldb::offset_t file_offset, 903 lldb::offset_t length) 904 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 905 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 906 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 907 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 908 ::memset(&m_header, 0, sizeof(m_header)); 909 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 910 } 911 912 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 913 lldb::DataBufferSP &header_data_sp, 914 const lldb::ProcessSP &process_sp, 915 lldb::addr_t header_addr) 916 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 917 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 918 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 919 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 920 ::memset(&m_header, 0, sizeof(m_header)); 921 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 922 } 923 924 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 925 lldb::offset_t *data_offset_ptr, 926 llvm::MachO::mach_header &header) { 927 data.SetByteOrder(endian::InlHostByteOrder()); 928 // Leave magic in the original byte order 929 header.magic = data.GetU32(data_offset_ptr); 930 bool can_parse = false; 931 bool is_64_bit = false; 932 switch (header.magic) { 933 case MH_MAGIC: 934 data.SetByteOrder(endian::InlHostByteOrder()); 935 data.SetAddressByteSize(4); 936 can_parse = true; 937 break; 938 939 case MH_MAGIC_64: 940 data.SetByteOrder(endian::InlHostByteOrder()); 941 data.SetAddressByteSize(8); 942 can_parse = true; 943 is_64_bit = true; 944 break; 945 946 case MH_CIGAM: 947 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 948 ? eByteOrderLittle 949 : eByteOrderBig); 950 data.SetAddressByteSize(4); 951 can_parse = true; 952 break; 953 954 case MH_CIGAM_64: 955 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 956 ? eByteOrderLittle 957 : eByteOrderBig); 958 data.SetAddressByteSize(8); 959 is_64_bit = true; 960 can_parse = true; 961 break; 962 963 default: 964 break; 965 } 966 967 if (can_parse) { 968 data.GetU32(data_offset_ptr, &header.cputype, 6); 969 if (is_64_bit) 970 *data_offset_ptr += 4; 971 return true; 972 } else { 973 memset(&header, 0, sizeof(header)); 974 } 975 return false; 976 } 977 978 bool ObjectFileMachO::ParseHeader() { 979 ModuleSP module_sp(GetModule()); 980 if (module_sp) { 981 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 982 bool can_parse = false; 983 lldb::offset_t offset = 0; 984 m_data.SetByteOrder(endian::InlHostByteOrder()); 985 // Leave magic in the original byte order 986 m_header.magic = m_data.GetU32(&offset); 987 switch (m_header.magic) { 988 case MH_MAGIC: 989 m_data.SetByteOrder(endian::InlHostByteOrder()); 990 m_data.SetAddressByteSize(4); 991 can_parse = true; 992 break; 993 994 case MH_MAGIC_64: 995 m_data.SetByteOrder(endian::InlHostByteOrder()); 996 m_data.SetAddressByteSize(8); 997 can_parse = true; 998 break; 999 1000 case MH_CIGAM: 1001 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1002 ? eByteOrderLittle 1003 : eByteOrderBig); 1004 m_data.SetAddressByteSize(4); 1005 can_parse = true; 1006 break; 1007 1008 case MH_CIGAM_64: 1009 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1010 ? eByteOrderLittle 1011 : eByteOrderBig); 1012 m_data.SetAddressByteSize(8); 1013 can_parse = true; 1014 break; 1015 1016 default: 1017 break; 1018 } 1019 1020 if (can_parse) { 1021 m_data.GetU32(&offset, &m_header.cputype, 6); 1022 1023 ModuleSpecList all_specs; 1024 ModuleSpec base_spec; 1025 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 1026 base_spec, all_specs); 1027 1028 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 1029 ArchSpec mach_arch = 1030 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 1031 1032 // Check if the module has a required architecture 1033 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1034 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1035 continue; 1036 1037 if (SetModulesArchitecture(mach_arch)) { 1038 const size_t header_and_lc_size = 1039 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1040 if (m_data.GetByteSize() < header_and_lc_size) { 1041 DataBufferSP data_sp; 1042 ProcessSP process_sp(m_process_wp.lock()); 1043 if (process_sp) { 1044 data_sp = 1045 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 } 1064 return false; 1065 } 1066 1067 ByteOrder ObjectFileMachO::GetByteOrder() const { 1068 return m_data.GetByteOrder(); 1069 } 1070 1071 bool ObjectFileMachO::IsExecutable() const { 1072 return m_header.filetype == MH_EXECUTE; 1073 } 1074 1075 bool ObjectFileMachO::IsDynamicLoader() const { 1076 return m_header.filetype == MH_DYLINKER; 1077 } 1078 1079 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1080 return m_data.GetAddressByteSize(); 1081 } 1082 1083 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1084 Symtab *symtab = GetSymtab(); 1085 if (symtab) { 1086 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1087 if (symbol) { 1088 if (symbol->ValueIsAddress()) { 1089 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1090 if (section_sp) { 1091 const lldb::SectionType section_type = section_sp->GetType(); 1092 switch (section_type) { 1093 case eSectionTypeInvalid: 1094 return AddressClass::eUnknown; 1095 1096 case eSectionTypeCode: 1097 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1098 // For ARM we have a bit in the n_desc field of the symbol that 1099 // tells us ARM/Thumb which is bit 0x0008. 1100 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1101 return AddressClass::eCodeAlternateISA; 1102 } 1103 return AddressClass::eCode; 1104 1105 case eSectionTypeContainer: 1106 return AddressClass::eUnknown; 1107 1108 case eSectionTypeData: 1109 case eSectionTypeDataCString: 1110 case eSectionTypeDataCStringPointers: 1111 case eSectionTypeDataSymbolAddress: 1112 case eSectionTypeData4: 1113 case eSectionTypeData8: 1114 case eSectionTypeData16: 1115 case eSectionTypeDataPointers: 1116 case eSectionTypeZeroFill: 1117 case eSectionTypeDataObjCMessageRefs: 1118 case eSectionTypeDataObjCCFStrings: 1119 case eSectionTypeGoSymtab: 1120 return AddressClass::eData; 1121 1122 case eSectionTypeDebug: 1123 case eSectionTypeDWARFDebugAbbrev: 1124 case eSectionTypeDWARFDebugAbbrevDwo: 1125 case eSectionTypeDWARFDebugAddr: 1126 case eSectionTypeDWARFDebugAranges: 1127 case eSectionTypeDWARFDebugCuIndex: 1128 case eSectionTypeDWARFDebugFrame: 1129 case eSectionTypeDWARFDebugInfo: 1130 case eSectionTypeDWARFDebugInfoDwo: 1131 case eSectionTypeDWARFDebugLine: 1132 case eSectionTypeDWARFDebugLineStr: 1133 case eSectionTypeDWARFDebugLoc: 1134 case eSectionTypeDWARFDebugLocLists: 1135 case eSectionTypeDWARFDebugMacInfo: 1136 case eSectionTypeDWARFDebugMacro: 1137 case eSectionTypeDWARFDebugNames: 1138 case eSectionTypeDWARFDebugPubNames: 1139 case eSectionTypeDWARFDebugPubTypes: 1140 case eSectionTypeDWARFDebugRanges: 1141 case eSectionTypeDWARFDebugRngLists: 1142 case eSectionTypeDWARFDebugStr: 1143 case eSectionTypeDWARFDebugStrDwo: 1144 case eSectionTypeDWARFDebugStrOffsets: 1145 case eSectionTypeDWARFDebugStrOffsetsDwo: 1146 case eSectionTypeDWARFDebugTypes: 1147 case eSectionTypeDWARFDebugTypesDwo: 1148 case eSectionTypeDWARFAppleNames: 1149 case eSectionTypeDWARFAppleTypes: 1150 case eSectionTypeDWARFAppleNamespaces: 1151 case eSectionTypeDWARFAppleObjC: 1152 case eSectionTypeDWARFGNUDebugAltLink: 1153 return AddressClass::eDebug; 1154 1155 case eSectionTypeEHFrame: 1156 case eSectionTypeARMexidx: 1157 case eSectionTypeARMextab: 1158 case eSectionTypeCompactUnwind: 1159 return AddressClass::eRuntime; 1160 1161 case eSectionTypeAbsoluteAddress: 1162 case eSectionTypeELFSymbolTable: 1163 case eSectionTypeELFDynamicSymbols: 1164 case eSectionTypeELFRelocationEntries: 1165 case eSectionTypeELFDynamicLinkInfo: 1166 case eSectionTypeOther: 1167 return AddressClass::eUnknown; 1168 } 1169 } 1170 } 1171 1172 const SymbolType symbol_type = symbol->GetType(); 1173 switch (symbol_type) { 1174 case eSymbolTypeAny: 1175 return AddressClass::eUnknown; 1176 case eSymbolTypeAbsolute: 1177 return AddressClass::eUnknown; 1178 1179 case eSymbolTypeCode: 1180 case eSymbolTypeTrampoline: 1181 case eSymbolTypeResolver: 1182 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1183 // For ARM we have a bit in the n_desc field of the symbol that tells 1184 // us ARM/Thumb which is bit 0x0008. 1185 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1186 return AddressClass::eCodeAlternateISA; 1187 } 1188 return AddressClass::eCode; 1189 1190 case eSymbolTypeData: 1191 return AddressClass::eData; 1192 case eSymbolTypeRuntime: 1193 return AddressClass::eRuntime; 1194 case eSymbolTypeException: 1195 return AddressClass::eRuntime; 1196 case eSymbolTypeSourceFile: 1197 return AddressClass::eDebug; 1198 case eSymbolTypeHeaderFile: 1199 return AddressClass::eDebug; 1200 case eSymbolTypeObjectFile: 1201 return AddressClass::eDebug; 1202 case eSymbolTypeCommonBlock: 1203 return AddressClass::eDebug; 1204 case eSymbolTypeBlock: 1205 return AddressClass::eDebug; 1206 case eSymbolTypeLocal: 1207 return AddressClass::eData; 1208 case eSymbolTypeParam: 1209 return AddressClass::eData; 1210 case eSymbolTypeVariable: 1211 return AddressClass::eData; 1212 case eSymbolTypeVariableType: 1213 return AddressClass::eDebug; 1214 case eSymbolTypeLineEntry: 1215 return AddressClass::eDebug; 1216 case eSymbolTypeLineHeader: 1217 return AddressClass::eDebug; 1218 case eSymbolTypeScopeBegin: 1219 return AddressClass::eDebug; 1220 case eSymbolTypeScopeEnd: 1221 return AddressClass::eDebug; 1222 case eSymbolTypeAdditional: 1223 return AddressClass::eUnknown; 1224 case eSymbolTypeCompiler: 1225 return AddressClass::eDebug; 1226 case eSymbolTypeInstrumentation: 1227 return AddressClass::eDebug; 1228 case eSymbolTypeUndefined: 1229 return AddressClass::eUnknown; 1230 case eSymbolTypeObjCClass: 1231 return AddressClass::eRuntime; 1232 case eSymbolTypeObjCMetaClass: 1233 return AddressClass::eRuntime; 1234 case eSymbolTypeObjCIVar: 1235 return AddressClass::eRuntime; 1236 case eSymbolTypeReExported: 1237 return AddressClass::eRuntime; 1238 } 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 TrieEntry() 1899 : name(), address(LLDB_INVALID_ADDRESS), flags(0), other(0), 1900 import_name() {} 1901 1902 void Clear() { 1903 name.Clear(); 1904 address = LLDB_INVALID_ADDRESS; 1905 flags = 0; 1906 other = 0; 1907 import_name.Clear(); 1908 } 1909 1910 void Dump() const { 1911 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1912 static_cast<unsigned long long>(address), 1913 static_cast<unsigned long long>(flags), 1914 static_cast<unsigned long long>(other), name.GetCString()); 1915 if (import_name) 1916 printf(" -> \"%s\"\n", import_name.GetCString()); 1917 else 1918 printf("\n"); 1919 } 1920 ConstString name; 1921 uint64_t address; 1922 uint64_t flags; 1923 uint64_t other; 1924 ConstString import_name; 1925 }; 1926 1927 struct TrieEntryWithOffset { 1928 lldb::offset_t nodeOffset; 1929 TrieEntry entry; 1930 1931 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1932 1933 void Dump(uint32_t idx) const { 1934 printf("[%3u] 0x%16.16llx: ", idx, 1935 static_cast<unsigned long long>(nodeOffset)); 1936 entry.Dump(); 1937 } 1938 1939 bool operator<(const TrieEntryWithOffset &other) const { 1940 return (nodeOffset < other.nodeOffset); 1941 } 1942 }; 1943 1944 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 1945 const bool is_arm, 1946 std::vector<llvm::StringRef> &nameSlices, 1947 std::set<lldb::addr_t> &resolver_addresses, 1948 std::vector<TrieEntryWithOffset> &output) { 1949 if (!data.ValidOffset(offset)) 1950 return true; 1951 1952 const uint64_t terminalSize = data.GetULEB128(&offset); 1953 lldb::offset_t children_offset = offset + terminalSize; 1954 if (terminalSize != 0) { 1955 TrieEntryWithOffset e(offset); 1956 e.entry.flags = data.GetULEB128(&offset); 1957 const char *import_name = nullptr; 1958 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 1959 e.entry.address = 0; 1960 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 1961 import_name = data.GetCStr(&offset); 1962 } else { 1963 e.entry.address = data.GetULEB128(&offset); 1964 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 1965 e.entry.other = data.GetULEB128(&offset); 1966 uint64_t resolver_addr = e.entry.other; 1967 if (is_arm) 1968 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 1969 resolver_addresses.insert(resolver_addr); 1970 } else 1971 e.entry.other = 0; 1972 } 1973 // Only add symbols that are reexport symbols with a valid import name 1974 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && 1975 import_name[0]) { 1976 std::string name; 1977 if (!nameSlices.empty()) { 1978 for (auto name_slice : nameSlices) 1979 name.append(name_slice.data(), name_slice.size()); 1980 } 1981 if (name.size() > 1) { 1982 // Skip the leading '_' 1983 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 1984 } 1985 if (import_name) { 1986 // Skip the leading '_' 1987 e.entry.import_name.SetCString(import_name + 1); 1988 } 1989 output.push_back(e); 1990 } 1991 } 1992 1993 const uint8_t childrenCount = data.GetU8(&children_offset); 1994 for (uint8_t i = 0; i < childrenCount; ++i) { 1995 const char *cstr = data.GetCStr(&children_offset); 1996 if (cstr) 1997 nameSlices.push_back(llvm::StringRef(cstr)); 1998 else 1999 return false; // Corrupt data 2000 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2001 if (childNodeOffset) { 2002 if (!ParseTrieEntries(data, childNodeOffset, is_arm, nameSlices, 2003 resolver_addresses, output)) { 2004 return false; 2005 } 2006 } 2007 nameSlices.pop_back(); 2008 } 2009 return true; 2010 } 2011 2012 // Read the UUID out of a dyld_shared_cache file on-disk. 2013 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2014 const ByteOrder byte_order, 2015 const uint32_t addr_byte_size) { 2016 UUID dsc_uuid; 2017 DataBufferSP DscData = MapFileData( 2018 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2019 if (!DscData) 2020 return dsc_uuid; 2021 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2022 2023 char version_str[7]; 2024 lldb::offset_t offset = 0; 2025 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2026 version_str[6] = '\0'; 2027 if (strcmp(version_str, "dyld_v") == 0) { 2028 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2029 dsc_uuid = UUID::fromOptionalData( 2030 dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t)); 2031 } 2032 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2033 if (log && dsc_uuid.IsValid()) { 2034 LLDB_LOGF(log, "Shared cache %s has UUID %s", 2035 dyld_shared_cache.GetPath().c_str(), 2036 dsc_uuid.GetAsString().c_str()); 2037 } 2038 return dsc_uuid; 2039 } 2040 2041 size_t ObjectFileMachO::ParseSymtab() { 2042 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2043 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2044 m_file.GetFilename().AsCString("")); 2045 ModuleSP module_sp(GetModule()); 2046 if (!module_sp) 2047 return 0; 2048 2049 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2050 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2051 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2052 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2053 FunctionStarts function_starts; 2054 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2055 uint32_t i; 2056 FileSpecList dylib_files; 2057 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2058 static const llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2059 static const llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2060 static const llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2061 2062 for (i = 0; i < m_header.ncmds; ++i) { 2063 const lldb::offset_t cmd_offset = offset; 2064 // Read in the load command and load command size 2065 struct load_command lc; 2066 if (m_data.GetU32(&offset, &lc, 2) == nullptr) 2067 break; 2068 // Watch for the symbol table load command 2069 switch (lc.cmd) { 2070 case LC_SYMTAB: 2071 symtab_load_command.cmd = lc.cmd; 2072 symtab_load_command.cmdsize = lc.cmdsize; 2073 // Read in the rest of the symtab load command 2074 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2075 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2076 return 0; 2077 if (symtab_load_command.symoff == 0) { 2078 if (log) 2079 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2080 return 0; 2081 } 2082 2083 if (symtab_load_command.stroff == 0) { 2084 if (log) 2085 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2086 return 0; 2087 } 2088 2089 if (symtab_load_command.nsyms == 0) { 2090 if (log) 2091 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2092 return 0; 2093 } 2094 2095 if (symtab_load_command.strsize == 0) { 2096 if (log) 2097 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2098 return 0; 2099 } 2100 break; 2101 2102 case LC_DYLD_INFO: 2103 case LC_DYLD_INFO_ONLY: 2104 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2105 dyld_info.cmd = lc.cmd; 2106 dyld_info.cmdsize = lc.cmdsize; 2107 } else { 2108 memset(&dyld_info, 0, sizeof(dyld_info)); 2109 } 2110 break; 2111 2112 case LC_LOAD_DYLIB: 2113 case LC_LOAD_WEAK_DYLIB: 2114 case LC_REEXPORT_DYLIB: 2115 case LC_LOADFVMLIB: 2116 case LC_LOAD_UPWARD_DYLIB: { 2117 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2118 const char *path = m_data.PeekCStr(name_offset); 2119 if (path) { 2120 FileSpec file_spec(path); 2121 // Strip the path if there is @rpath, @executable, etc so we just use 2122 // the basename 2123 if (path[0] == '@') 2124 file_spec.GetDirectory().Clear(); 2125 2126 if (lc.cmd == LC_REEXPORT_DYLIB) { 2127 m_reexported_dylibs.AppendIfUnique(file_spec); 2128 } 2129 2130 dylib_files.Append(file_spec); 2131 } 2132 } break; 2133 2134 case LC_FUNCTION_STARTS: 2135 function_starts_load_command.cmd = lc.cmd; 2136 function_starts_load_command.cmdsize = lc.cmdsize; 2137 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2138 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2139 memset(&function_starts_load_command, 0, 2140 sizeof(function_starts_load_command)); 2141 break; 2142 2143 default: 2144 break; 2145 } 2146 offset = cmd_offset + lc.cmdsize; 2147 } 2148 2149 if (symtab_load_command.cmd) { 2150 Symtab *symtab = m_symtab_up.get(); 2151 SectionList *section_list = GetSectionList(); 2152 if (section_list == nullptr) 2153 return 0; 2154 2155 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2156 const ByteOrder byte_order = m_data.GetByteOrder(); 2157 bool bit_width_32 = addr_byte_size == 4; 2158 const size_t nlist_byte_size = 2159 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2160 2161 DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size); 2162 DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size); 2163 DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size); 2164 DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order, 2165 addr_byte_size); 2166 DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size); 2167 2168 const addr_t nlist_data_byte_size = 2169 symtab_load_command.nsyms * nlist_byte_size; 2170 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2171 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2172 2173 ProcessSP process_sp(m_process_wp.lock()); 2174 Process *process = process_sp.get(); 2175 2176 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2177 2178 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2179 Target &target = process->GetTarget(); 2180 2181 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2182 2183 SectionSP linkedit_section_sp( 2184 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2185 // Reading mach file from memory in a process or core file... 2186 2187 if (linkedit_section_sp) { 2188 addr_t linkedit_load_addr = 2189 linkedit_section_sp->GetLoadBaseAddress(&target); 2190 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2191 // We might be trying to access the symbol table before the 2192 // __LINKEDIT's load address has been set in the target. We can't 2193 // fail to read the symbol table, so calculate the right address 2194 // manually 2195 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2196 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2197 } 2198 2199 const addr_t linkedit_file_offset = 2200 linkedit_section_sp->GetFileOffset(); 2201 const addr_t symoff_addr = linkedit_load_addr + 2202 symtab_load_command.symoff - 2203 linkedit_file_offset; 2204 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2205 linkedit_file_offset; 2206 2207 bool data_was_read = false; 2208 2209 #if defined(__APPLE__) && \ 2210 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2211 if (m_header.flags & 0x80000000u && 2212 process->GetAddressByteSize() == sizeof(void *)) { 2213 // This mach-o memory file is in the dyld shared cache. If this 2214 // program is not remote and this is iOS, then this process will 2215 // share the same shared cache as the process we are debugging and we 2216 // can read the entire __LINKEDIT from the address space in this 2217 // process. This is a needed optimization that is used for local iOS 2218 // debugging only since all shared libraries in the shared cache do 2219 // not have corresponding files that exist in the file system of the 2220 // device. They have been combined into a single file. This means we 2221 // always have to load these files from memory. All of the symbol and 2222 // string tables from all of the __LINKEDIT sections from the shared 2223 // libraries in the shared cache have been merged into a single large 2224 // symbol and string table. Reading all of this symbol and string 2225 // table data across can slow down debug launch times, so we optimize 2226 // this by reading the memory for the __LINKEDIT section from this 2227 // process. 2228 2229 UUID lldb_shared_cache; 2230 addr_t lldb_shared_cache_addr; 2231 GetLLDBSharedCacheUUID (lldb_shared_cache_addr, lldb_shared_cache); 2232 UUID process_shared_cache; 2233 addr_t process_shared_cache_addr; 2234 GetProcessSharedCacheUUID(process, process_shared_cache_addr, process_shared_cache); 2235 bool use_lldb_cache = true; 2236 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2237 (lldb_shared_cache != process_shared_cache 2238 || process_shared_cache_addr != lldb_shared_cache_addr)) { 2239 use_lldb_cache = false; 2240 } 2241 2242 PlatformSP platform_sp(target.GetPlatform()); 2243 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2244 data_was_read = true; 2245 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2246 eByteOrderLittle); 2247 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2248 eByteOrderLittle); 2249 if (function_starts_load_command.cmd) { 2250 const addr_t func_start_addr = 2251 linkedit_load_addr + function_starts_load_command.dataoff - 2252 linkedit_file_offset; 2253 function_starts_data.SetData( 2254 (void *)func_start_addr, 2255 function_starts_load_command.datasize, eByteOrderLittle); 2256 } 2257 } 2258 } 2259 #endif 2260 2261 if (!data_was_read) { 2262 // Always load dyld - the dynamic linker - from memory if we didn't 2263 // find a binary anywhere else. lldb will not register 2264 // dylib/framework/bundle loads/unloads if we don't have the dyld 2265 // symbols, we force dyld to load from memory despite the user's 2266 // target.memory-module-load-level setting. 2267 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2268 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2269 DataBufferSP nlist_data_sp( 2270 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2271 if (nlist_data_sp) 2272 nlist_data.SetData(nlist_data_sp, 0, 2273 nlist_data_sp->GetByteSize()); 2274 if (m_dysymtab.nindirectsyms != 0) { 2275 const addr_t indirect_syms_addr = linkedit_load_addr + 2276 m_dysymtab.indirectsymoff - 2277 linkedit_file_offset; 2278 DataBufferSP indirect_syms_data_sp( 2279 ReadMemory(process_sp, indirect_syms_addr, 2280 m_dysymtab.nindirectsyms * 4)); 2281 if (indirect_syms_data_sp) 2282 indirect_symbol_index_data.SetData( 2283 indirect_syms_data_sp, 0, 2284 indirect_syms_data_sp->GetByteSize()); 2285 // If this binary is outside the shared cache, 2286 // cache the string table. 2287 // Binaries in the shared cache all share a giant string table, and 2288 // we can't share the string tables across multiple ObjectFileMachO's, 2289 // so we'd end up re-reading this mega-strtab for every binary 2290 // in the shared cache - it would be a big perf problem. 2291 // For binaries outside the shared cache, it's faster to read the 2292 // entire strtab at once instead of piece-by-piece as we process 2293 // the nlist records. 2294 if ((m_header.flags & 0x80000000u) == 0) { 2295 DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, 2296 strtab_data_byte_size)); 2297 if (strtab_data_sp) { 2298 strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 2299 } 2300 } 2301 } 2302 } 2303 if (memory_module_load_level >= 2304 eMemoryModuleLoadLevelPartial) { 2305 if (function_starts_load_command.cmd) { 2306 const addr_t func_start_addr = 2307 linkedit_load_addr + function_starts_load_command.dataoff - 2308 linkedit_file_offset; 2309 DataBufferSP func_start_data_sp( 2310 ReadMemory(process_sp, func_start_addr, 2311 function_starts_load_command.datasize)); 2312 if (func_start_data_sp) 2313 function_starts_data.SetData(func_start_data_sp, 0, 2314 func_start_data_sp->GetByteSize()); 2315 } 2316 } 2317 } 2318 } 2319 } else { 2320 nlist_data.SetData(m_data, symtab_load_command.symoff, 2321 nlist_data_byte_size); 2322 strtab_data.SetData(m_data, symtab_load_command.stroff, 2323 strtab_data_byte_size); 2324 2325 if (dyld_info.export_size > 0) { 2326 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2327 dyld_info.export_size); 2328 } 2329 2330 if (m_dysymtab.nindirectsyms != 0) { 2331 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2332 m_dysymtab.nindirectsyms * 4); 2333 } 2334 if (function_starts_load_command.cmd) { 2335 function_starts_data.SetData(m_data, 2336 function_starts_load_command.dataoff, 2337 function_starts_load_command.datasize); 2338 } 2339 } 2340 2341 if (nlist_data.GetByteSize() == 0 && 2342 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2343 if (log) 2344 module_sp->LogMessage(log, "failed to read nlist data"); 2345 return 0; 2346 } 2347 2348 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2349 if (!have_strtab_data) { 2350 if (process) { 2351 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2352 if (log) 2353 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2354 return 0; 2355 } 2356 } else { 2357 if (log) 2358 module_sp->LogMessage(log, "failed to read strtab data"); 2359 return 0; 2360 } 2361 } 2362 2363 ConstString g_segment_name_TEXT = GetSegmentNameTEXT(); 2364 ConstString g_segment_name_DATA = GetSegmentNameDATA(); 2365 ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2366 ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2367 ConstString g_segment_name_OBJC = GetSegmentNameOBJC(); 2368 ConstString g_section_name_eh_frame = GetSectionNameEHFrame(); 2369 SectionSP text_section_sp( 2370 section_list->FindSectionByName(g_segment_name_TEXT)); 2371 SectionSP data_section_sp( 2372 section_list->FindSectionByName(g_segment_name_DATA)); 2373 SectionSP data_dirty_section_sp( 2374 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2375 SectionSP data_const_section_sp( 2376 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2377 SectionSP objc_section_sp( 2378 section_list->FindSectionByName(g_segment_name_OBJC)); 2379 SectionSP eh_frame_section_sp; 2380 if (text_section_sp.get()) 2381 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2382 g_section_name_eh_frame); 2383 else 2384 eh_frame_section_sp = 2385 section_list->FindSectionByName(g_section_name_eh_frame); 2386 2387 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2388 2389 // lldb works best if it knows the start address of all functions in a 2390 // module. Linker symbols or debug info are normally the best source of 2391 // information for start addr / size but they may be stripped in a released 2392 // binary. Two additional sources of information exist in Mach-O binaries: 2393 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2394 // function's start address in the 2395 // binary, relative to the text section. 2396 // eh_frame - the eh_frame FDEs have the start addr & size of 2397 // each function 2398 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2399 // all modern binaries. 2400 // Binaries built to run on older releases may need to use eh_frame 2401 // information. 2402 2403 if (text_section_sp && function_starts_data.GetByteSize()) { 2404 FunctionStarts::Entry function_start_entry; 2405 function_start_entry.data = false; 2406 lldb::offset_t function_start_offset = 0; 2407 function_start_entry.addr = text_section_sp->GetFileAddress(); 2408 uint64_t delta; 2409 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2410 0) { 2411 // Now append the current entry 2412 function_start_entry.addr += delta; 2413 function_starts.Append(function_start_entry); 2414 } 2415 } else { 2416 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2417 // load command claiming an eh_frame but it doesn't actually have the 2418 // eh_frame content. And if we have a dSYM, we don't need to do any of 2419 // this fill-in-the-missing-symbols works anyway - the debug info should 2420 // give us all the functions in the module. 2421 if (text_section_sp.get() && eh_frame_section_sp.get() && 2422 m_type != eTypeDebugInfo) { 2423 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2424 DWARFCallFrameInfo::EH); 2425 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2426 eh_frame.GetFunctionAddressAndSizeVector(functions); 2427 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2428 size_t count = functions.GetSize(); 2429 for (size_t i = 0; i < count; ++i) { 2430 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2431 functions.GetEntryAtIndex(i); 2432 if (func) { 2433 FunctionStarts::Entry function_start_entry; 2434 function_start_entry.addr = func->base - text_base_addr; 2435 function_starts.Append(function_start_entry); 2436 } 2437 } 2438 } 2439 } 2440 2441 const size_t function_starts_count = function_starts.GetSize(); 2442 2443 // For user process binaries (executables, dylibs, frameworks, bundles), if 2444 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2445 // going to assume the binary has been stripped. Don't allow assembly 2446 // language instruction emulation because we don't know proper function 2447 // start boundaries. 2448 // 2449 // For all other types of binaries (kernels, stand-alone bare board 2450 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2451 // sections - we should not make any assumptions about them based on that. 2452 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2453 m_allow_assembly_emulation_unwind_plans = false; 2454 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2455 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2456 2457 if (unwind_or_symbol_log) 2458 module_sp->LogMessage( 2459 unwind_or_symbol_log, 2460 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2461 } 2462 2463 const user_id_t TEXT_eh_frame_sectID = 2464 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2465 : static_cast<user_id_t>(NO_SECT); 2466 2467 lldb::offset_t nlist_data_offset = 0; 2468 2469 uint32_t N_SO_index = UINT32_MAX; 2470 2471 MachSymtabSectionInfo section_info(section_list); 2472 std::vector<uint32_t> N_FUN_indexes; 2473 std::vector<uint32_t> N_NSYM_indexes; 2474 std::vector<uint32_t> N_INCL_indexes; 2475 std::vector<uint32_t> N_BRAC_indexes; 2476 std::vector<uint32_t> N_COMM_indexes; 2477 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2478 typedef std::map<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2479 typedef std::map<const char *, uint32_t> ConstNameToSymbolIndexMap; 2480 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2481 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2482 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2483 // Any symbols that get merged into another will get an entry in this map 2484 // so we know 2485 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2486 uint32_t nlist_idx = 0; 2487 Symbol *symbol_ptr = nullptr; 2488 2489 uint32_t sym_idx = 0; 2490 Symbol *sym = nullptr; 2491 size_t num_syms = 0; 2492 std::string memory_symbol_name; 2493 uint32_t unmapped_local_symbols_found = 0; 2494 2495 std::vector<TrieEntryWithOffset> trie_entries; 2496 std::set<lldb::addr_t> resolver_addresses; 2497 2498 if (dyld_trie_data.GetByteSize() > 0) { 2499 std::vector<llvm::StringRef> nameSlices; 2500 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, 2501 resolver_addresses, trie_entries); 2502 2503 ConstString text_segment_name("__TEXT"); 2504 SectionSP text_segment_sp = 2505 GetSectionList()->FindSectionByName(text_segment_name); 2506 if (text_segment_sp) { 2507 const lldb::addr_t text_segment_file_addr = 2508 text_segment_sp->GetFileAddress(); 2509 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2510 for (auto &e : trie_entries) 2511 e.entry.address += text_segment_file_addr; 2512 } 2513 } 2514 } 2515 2516 typedef std::set<ConstString> IndirectSymbols; 2517 IndirectSymbols indirect_symbol_names; 2518 2519 #if defined(__APPLE__) && \ 2520 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2521 2522 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2523 // optimized by moving LOCAL symbols out of the memory mapped portion of 2524 // the DSC. The symbol information has all been retained, but it isn't 2525 // available in the normal nlist data. However, there *are* duplicate 2526 // entries of *some* 2527 // LOCAL symbols in the normal nlist data. To handle this situation 2528 // correctly, we must first attempt 2529 // to parse any DSC unmapped symbol information. If we find any, we set a 2530 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2531 2532 if (m_header.flags & 0x80000000u) { 2533 // Before we can start mapping the DSC, we need to make certain the 2534 // target process is actually using the cache we can find. 2535 2536 // Next we need to determine the correct path for the dyld shared cache. 2537 2538 ArchSpec header_arch; 2539 GetArchitecture(header_arch); 2540 char dsc_path[PATH_MAX]; 2541 char dsc_path_development[PATH_MAX]; 2542 2543 snprintf( 2544 dsc_path, sizeof(dsc_path), "%s%s%s", 2545 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2546 */ 2547 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2548 header_arch.GetArchitectureName()); 2549 2550 snprintf( 2551 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2552 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2553 */ 2554 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2555 header_arch.GetArchitectureName(), ".development"); 2556 2557 FileSpec dsc_nondevelopment_filespec(dsc_path, false); 2558 FileSpec dsc_development_filespec(dsc_path_development, false); 2559 FileSpec dsc_filespec; 2560 2561 UUID dsc_uuid; 2562 UUID process_shared_cache_uuid; 2563 addr_t process_shared_cache_base_addr; 2564 2565 if (process) { 2566 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, process_shared_cache_uuid); 2567 } 2568 2569 // First see if we can find an exact match for the inferior process 2570 // shared cache UUID in the development or non-development shared caches 2571 // on disk. 2572 if (process_shared_cache_uuid.IsValid()) { 2573 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2574 UUID dsc_development_uuid = GetSharedCacheUUID( 2575 dsc_development_filespec, byte_order, addr_byte_size); 2576 if (dsc_development_uuid.IsValid() && 2577 dsc_development_uuid == process_shared_cache_uuid) { 2578 dsc_filespec = dsc_development_filespec; 2579 dsc_uuid = dsc_development_uuid; 2580 } 2581 } 2582 if (!dsc_uuid.IsValid() && 2583 FileSystem::Instance().Exists(dsc_nondevelopment_filespec)) { 2584 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2585 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2586 if (dsc_nondevelopment_uuid.IsValid() && 2587 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2588 dsc_filespec = dsc_nondevelopment_filespec; 2589 dsc_uuid = dsc_nondevelopment_uuid; 2590 } 2591 } 2592 } 2593 2594 // Failing a UUID match, prefer the development dyld_shared cache if both 2595 // are present. 2596 if (!FileSystem::Instance().Exists(dsc_filespec)) { 2597 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2598 dsc_filespec = dsc_development_filespec; 2599 } else { 2600 dsc_filespec = dsc_nondevelopment_filespec; 2601 } 2602 } 2603 2604 /* The dyld_cache_header has a pointer to the 2605 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2606 The dyld_cache_local_symbols_info structure gives us three things: 2607 1. The start and count of the nlist records in the dyld_shared_cache 2608 file 2609 2. The start and size of the strings for these nlist records 2610 3. The start and count of dyld_cache_local_symbols_entry entries 2611 2612 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2613 dyld shared cache. 2614 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2615 the dyld shared cache. 2616 The dyld_cache_local_symbols_entry also lists the start of this 2617 dylib/framework's nlist records 2618 and the count of how many nlist records there are for this 2619 dylib/framework. 2620 */ 2621 2622 // Process the dyld shared cache header to find the unmapped symbols 2623 2624 DataBufferSP dsc_data_sp = MapFileData( 2625 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2626 if (!dsc_uuid.IsValid()) { 2627 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2628 } 2629 if (dsc_data_sp) { 2630 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2631 2632 bool uuid_match = true; 2633 if (dsc_uuid.IsValid() && process) { 2634 if (process_shared_cache_uuid.IsValid() && 2635 dsc_uuid != process_shared_cache_uuid) { 2636 // The on-disk dyld_shared_cache file is not the same as the one in 2637 // this process' memory, don't use it. 2638 uuid_match = false; 2639 ModuleSP module_sp(GetModule()); 2640 if (module_sp) 2641 module_sp->ReportWarning("process shared cache does not match " 2642 "on-disk dyld_shared_cache file, some " 2643 "symbol names will be missing."); 2644 } 2645 } 2646 2647 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2648 2649 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2650 2651 // If the mappingOffset points to a location inside the header, we've 2652 // opened an old dyld shared cache, and should not proceed further. 2653 if (uuid_match && 2654 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2655 2656 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2657 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2658 mappingOffset); 2659 2660 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2661 byte_order, addr_byte_size); 2662 offset = 0; 2663 2664 // The File addresses (from the in-memory Mach-O load commands) for 2665 // the shared libraries in the shared library cache need to be 2666 // adjusted by an offset to match up with the dylibOffset identifying 2667 // field in the dyld_cache_local_symbol_entry's. This offset is 2668 // recorded in mapping_offset_value. 2669 const uint64_t mapping_offset_value = 2670 dsc_mapping_info_data.GetU64(&offset); 2671 2672 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, 2673 localSymbolsOffset); 2674 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2675 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2676 2677 if (localSymbolsOffset && localSymbolsSize) { 2678 // Map the local symbols 2679 DataBufferSP dsc_local_symbols_data_sp = 2680 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2681 2682 if (dsc_local_symbols_data_sp) { 2683 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2684 byte_order, addr_byte_size); 2685 2686 offset = 0; 2687 2688 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2689 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2690 UndefinedNameToDescMap undefined_name_to_desc; 2691 SymbolIndexToName reexport_shlib_needs_fixup; 2692 2693 // Read the local_symbols_infos struct in one shot 2694 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2695 dsc_local_symbols_data.GetU32(&offset, 2696 &local_symbols_info.nlistOffset, 6); 2697 2698 SectionSP text_section_sp( 2699 section_list->FindSectionByName(GetSegmentNameTEXT())); 2700 2701 uint32_t header_file_offset = 2702 (text_section_sp->GetFileAddress() - mapping_offset_value); 2703 2704 offset = local_symbols_info.entriesOffset; 2705 for (uint32_t entry_index = 0; 2706 entry_index < local_symbols_info.entriesCount; 2707 entry_index++) { 2708 struct lldb_copy_dyld_cache_local_symbols_entry 2709 local_symbols_entry; 2710 local_symbols_entry.dylibOffset = 2711 dsc_local_symbols_data.GetU32(&offset); 2712 local_symbols_entry.nlistStartIndex = 2713 dsc_local_symbols_data.GetU32(&offset); 2714 local_symbols_entry.nlistCount = 2715 dsc_local_symbols_data.GetU32(&offset); 2716 2717 if (header_file_offset == local_symbols_entry.dylibOffset) { 2718 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2719 2720 // The normal nlist code cannot correctly size the Symbols 2721 // array, we need to allocate it here. 2722 sym = symtab->Resize( 2723 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2724 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2725 num_syms = symtab->GetNumSymbols(); 2726 2727 nlist_data_offset = 2728 local_symbols_info.nlistOffset + 2729 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2730 uint32_t string_table_offset = 2731 local_symbols_info.stringsOffset; 2732 2733 for (uint32_t nlist_index = 0; 2734 nlist_index < local_symbols_entry.nlistCount; 2735 nlist_index++) { 2736 ///////////////////////////// 2737 { 2738 struct nlist_64 nlist; 2739 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize( 2740 nlist_data_offset, nlist_byte_size)) 2741 break; 2742 2743 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked( 2744 &nlist_data_offset); 2745 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked( 2746 &nlist_data_offset); 2747 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked( 2748 &nlist_data_offset); 2749 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked( 2750 &nlist_data_offset); 2751 nlist.n_value = 2752 dsc_local_symbols_data.GetAddress_unchecked( 2753 &nlist_data_offset); 2754 2755 SymbolType type = eSymbolTypeInvalid; 2756 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2757 string_table_offset + nlist.n_strx); 2758 2759 if (symbol_name == NULL) { 2760 // No symbol should be NULL, even the symbols with no 2761 // string values should have an offset zero which 2762 // points to an empty C-string 2763 Host::SystemLog( 2764 Host::eSystemLogError, 2765 "error: DSC unmapped local symbol[%u] has invalid " 2766 "string table offset 0x%x in %s, ignoring symbol\n", 2767 entry_index, nlist.n_strx, 2768 module_sp->GetFileSpec().GetPath().c_str()); 2769 continue; 2770 } 2771 if (symbol_name[0] == '\0') 2772 symbol_name = NULL; 2773 2774 const char *symbol_name_non_abi_mangled = NULL; 2775 2776 SectionSP symbol_section; 2777 uint32_t symbol_byte_size = 0; 2778 bool add_nlist = true; 2779 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2780 bool demangled_is_synthesized = false; 2781 bool is_gsym = false; 2782 bool set_value = true; 2783 2784 assert(sym_idx < num_syms); 2785 2786 sym[sym_idx].SetDebug(is_debug); 2787 2788 if (is_debug) { 2789 switch (nlist.n_type) { 2790 case N_GSYM: 2791 // global symbol: name,,NO_SECT,type,0 2792 // Sometimes the N_GSYM value contains the address. 2793 2794 // FIXME: In the .o files, we have a GSYM and a debug 2795 // symbol for all the ObjC data. They 2796 // have the same address, but we want to ensure that 2797 // we always find only the real symbol, 'cause we 2798 // don't currently correctly attribute the 2799 // GSYM one to the ObjCClass/Ivar/MetaClass 2800 // symbol type. This is a temporary hack to make 2801 // sure the ObjectiveC symbols get treated correctly. 2802 // To do this right, we should coalesce all the GSYM 2803 // & global symbols that have the same address. 2804 2805 is_gsym = true; 2806 sym[sym_idx].SetExternal(true); 2807 2808 if (symbol_name && symbol_name[0] == '_' && 2809 symbol_name[1] == 'O') { 2810 llvm::StringRef symbol_name_ref(symbol_name); 2811 if (symbol_name_ref.startswith( 2812 g_objc_v2_prefix_class)) { 2813 symbol_name_non_abi_mangled = symbol_name + 1; 2814 symbol_name = 2815 symbol_name + g_objc_v2_prefix_class.size(); 2816 type = eSymbolTypeObjCClass; 2817 demangled_is_synthesized = true; 2818 2819 } else if (symbol_name_ref.startswith( 2820 g_objc_v2_prefix_metaclass)) { 2821 symbol_name_non_abi_mangled = symbol_name + 1; 2822 symbol_name = symbol_name + 2823 g_objc_v2_prefix_metaclass.size(); 2824 type = eSymbolTypeObjCMetaClass; 2825 demangled_is_synthesized = true; 2826 } else if (symbol_name_ref.startswith( 2827 g_objc_v2_prefix_ivar)) { 2828 symbol_name_non_abi_mangled = symbol_name + 1; 2829 symbol_name = 2830 symbol_name + g_objc_v2_prefix_ivar.size(); 2831 type = eSymbolTypeObjCIVar; 2832 demangled_is_synthesized = true; 2833 } 2834 } else { 2835 if (nlist.n_value != 0) 2836 symbol_section = section_info.GetSection( 2837 nlist.n_sect, nlist.n_value); 2838 type = eSymbolTypeData; 2839 } 2840 break; 2841 2842 case N_FNAME: 2843 // procedure name (f77 kludge): name,,NO_SECT,0,0 2844 type = eSymbolTypeCompiler; 2845 break; 2846 2847 case N_FUN: 2848 // procedure: name,,n_sect,linenumber,address 2849 if (symbol_name) { 2850 type = eSymbolTypeCode; 2851 symbol_section = section_info.GetSection( 2852 nlist.n_sect, nlist.n_value); 2853 2854 N_FUN_addr_to_sym_idx.insert( 2855 std::make_pair(nlist.n_value, sym_idx)); 2856 // We use the current number of symbols in the 2857 // symbol table in lieu of using nlist_idx in case 2858 // we ever start trimming entries out 2859 N_FUN_indexes.push_back(sym_idx); 2860 } else { 2861 type = eSymbolTypeCompiler; 2862 2863 if (!N_FUN_indexes.empty()) { 2864 // Copy the size of the function into the 2865 // original 2866 // STAB entry so we don't have 2867 // to hunt for it later 2868 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2869 ->SetByteSize(nlist.n_value); 2870 N_FUN_indexes.pop_back(); 2871 // We don't really need the end function STAB as 2872 // it contains the size which we already placed 2873 // with the original symbol, so don't add it if 2874 // we want a minimal symbol table 2875 add_nlist = false; 2876 } 2877 } 2878 break; 2879 2880 case N_STSYM: 2881 // static symbol: name,,n_sect,type,address 2882 N_STSYM_addr_to_sym_idx.insert( 2883 std::make_pair(nlist.n_value, sym_idx)); 2884 symbol_section = section_info.GetSection( 2885 nlist.n_sect, nlist.n_value); 2886 if (symbol_name && symbol_name[0]) { 2887 type = ObjectFile::GetSymbolTypeFromName( 2888 symbol_name + 1, eSymbolTypeData); 2889 } 2890 break; 2891 2892 case N_LCSYM: 2893 // .lcomm symbol: name,,n_sect,type,address 2894 symbol_section = section_info.GetSection( 2895 nlist.n_sect, nlist.n_value); 2896 type = eSymbolTypeCommonBlock; 2897 break; 2898 2899 case N_BNSYM: 2900 // We use the current number of symbols in the symbol 2901 // table in lieu of using nlist_idx in case we ever 2902 // start trimming entries out Skip these if we want 2903 // minimal symbol tables 2904 add_nlist = false; 2905 break; 2906 2907 case N_ENSYM: 2908 // Set the size of the N_BNSYM to the terminating 2909 // index of this N_ENSYM so that we can always skip 2910 // the entire symbol if we need to navigate more 2911 // quickly at the source level when parsing STABS 2912 // Skip these if we want minimal symbol tables 2913 add_nlist = false; 2914 break; 2915 2916 case N_OPT: 2917 // emitted with gcc2_compiled and in gcc source 2918 type = eSymbolTypeCompiler; 2919 break; 2920 2921 case N_RSYM: 2922 // register sym: name,,NO_SECT,type,register 2923 type = eSymbolTypeVariable; 2924 break; 2925 2926 case N_SLINE: 2927 // src line: 0,,n_sect,linenumber,address 2928 symbol_section = section_info.GetSection( 2929 nlist.n_sect, nlist.n_value); 2930 type = eSymbolTypeLineEntry; 2931 break; 2932 2933 case N_SSYM: 2934 // structure elt: name,,NO_SECT,type,struct_offset 2935 type = eSymbolTypeVariableType; 2936 break; 2937 2938 case N_SO: 2939 // source file name 2940 type = eSymbolTypeSourceFile; 2941 if (symbol_name == NULL) { 2942 add_nlist = false; 2943 if (N_SO_index != UINT32_MAX) { 2944 // Set the size of the N_SO to the terminating 2945 // index of this N_SO so that we can always skip 2946 // the entire N_SO if we need to navigate more 2947 // quickly at the source level when parsing STABS 2948 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2949 symbol_ptr->SetByteSize(sym_idx); 2950 symbol_ptr->SetSizeIsSibling(true); 2951 } 2952 N_NSYM_indexes.clear(); 2953 N_INCL_indexes.clear(); 2954 N_BRAC_indexes.clear(); 2955 N_COMM_indexes.clear(); 2956 N_FUN_indexes.clear(); 2957 N_SO_index = UINT32_MAX; 2958 } else { 2959 // We use the current number of symbols in the 2960 // symbol table in lieu of using nlist_idx in case 2961 // we ever start trimming entries out 2962 const bool N_SO_has_full_path = 2963 symbol_name[0] == '/'; 2964 if (N_SO_has_full_path) { 2965 if ((N_SO_index == sym_idx - 1) && 2966 ((sym_idx - 1) < num_syms)) { 2967 // We have two consecutive N_SO entries where 2968 // the first contains a directory and the 2969 // second contains a full path. 2970 sym[sym_idx - 1].GetMangled().SetValue( 2971 ConstString(symbol_name), false); 2972 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2973 add_nlist = false; 2974 } else { 2975 // This is the first entry in a N_SO that 2976 // contains a directory or 2977 // a full path to the source file 2978 N_SO_index = sym_idx; 2979 } 2980 } else if ((N_SO_index == sym_idx - 1) && 2981 ((sym_idx - 1) < num_syms)) { 2982 // This is usually the second N_SO entry that 2983 // contains just the filename, so here we combine 2984 // it with the first one if we are minimizing the 2985 // symbol table 2986 const char *so_path = 2987 sym[sym_idx - 1] 2988 .GetMangled() 2989 .GetDemangledName( 2990 lldb::eLanguageTypeUnknown) 2991 .AsCString(); 2992 if (so_path && so_path[0]) { 2993 std::string full_so_path(so_path); 2994 const size_t double_slash_pos = 2995 full_so_path.find("//"); 2996 if (double_slash_pos != std::string::npos) { 2997 // The linker has been generating bad N_SO 2998 // entries with doubled up paths 2999 // in the format "%s%s" where the first 3000 // string in the DW_AT_comp_dir, and the 3001 // second is the directory for the source 3002 // file so you end up with a path that looks 3003 // like "/tmp/src//tmp/src/" 3004 FileSpec so_dir(so_path, false); 3005 if (!FileSystem::Instance().Exists(so_dir)) { 3006 so_dir.SetFile( 3007 &full_so_path[double_slash_pos + 1], 3008 false); 3009 if (FileSystem::Instance().Exists(so_dir)) { 3010 // Trim off the incorrect path 3011 full_so_path.erase(0, 3012 double_slash_pos + 1); 3013 } 3014 } 3015 } 3016 if (*full_so_path.rbegin() != '/') 3017 full_so_path += '/'; 3018 full_so_path += symbol_name; 3019 sym[sym_idx - 1].GetMangled().SetValue( 3020 ConstString(full_so_path.c_str()), false); 3021 add_nlist = false; 3022 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3023 } 3024 } else { 3025 // This could be a relative path to a N_SO 3026 N_SO_index = sym_idx; 3027 } 3028 } 3029 break; 3030 3031 case N_OSO: 3032 // object file name: name,,0,0,st_mtime 3033 type = eSymbolTypeObjectFile; 3034 break; 3035 3036 case N_LSYM: 3037 // local sym: name,,NO_SECT,type,offset 3038 type = eSymbolTypeLocal; 3039 break; 3040 3041 // INCL scopes 3042 case N_BINCL: 3043 // include file beginning: name,,NO_SECT,0,sum We use 3044 // the current number of symbols in the symbol table 3045 // in lieu of using nlist_idx in case we ever start 3046 // trimming entries out 3047 N_INCL_indexes.push_back(sym_idx); 3048 type = eSymbolTypeScopeBegin; 3049 break; 3050 3051 case N_EINCL: 3052 // include file end: name,,NO_SECT,0,0 3053 // Set the size of the N_BINCL to the terminating 3054 // index of this N_EINCL so that we can always skip 3055 // the entire symbol if we need to navigate more 3056 // quickly at the source level when parsing STABS 3057 if (!N_INCL_indexes.empty()) { 3058 symbol_ptr = 3059 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3060 symbol_ptr->SetByteSize(sym_idx + 1); 3061 symbol_ptr->SetSizeIsSibling(true); 3062 N_INCL_indexes.pop_back(); 3063 } 3064 type = eSymbolTypeScopeEnd; 3065 break; 3066 3067 case N_SOL: 3068 // #included file name: name,,n_sect,0,address 3069 type = eSymbolTypeHeaderFile; 3070 3071 // We currently don't use the header files on darwin 3072 add_nlist = false; 3073 break; 3074 3075 case N_PARAMS: 3076 // compiler parameters: name,,NO_SECT,0,0 3077 type = eSymbolTypeCompiler; 3078 break; 3079 3080 case N_VERSION: 3081 // compiler version: name,,NO_SECT,0,0 3082 type = eSymbolTypeCompiler; 3083 break; 3084 3085 case N_OLEVEL: 3086 // compiler -O level: name,,NO_SECT,0,0 3087 type = eSymbolTypeCompiler; 3088 break; 3089 3090 case N_PSYM: 3091 // parameter: name,,NO_SECT,type,offset 3092 type = eSymbolTypeVariable; 3093 break; 3094 3095 case N_ENTRY: 3096 // alternate entry: name,,n_sect,linenumber,address 3097 symbol_section = section_info.GetSection( 3098 nlist.n_sect, nlist.n_value); 3099 type = eSymbolTypeLineEntry; 3100 break; 3101 3102 // Left and Right Braces 3103 case N_LBRAC: 3104 // left bracket: 0,,NO_SECT,nesting level,address We 3105 // use the current number of symbols in the symbol 3106 // table in lieu of using nlist_idx in case we ever 3107 // start trimming entries out 3108 symbol_section = section_info.GetSection( 3109 nlist.n_sect, nlist.n_value); 3110 N_BRAC_indexes.push_back(sym_idx); 3111 type = eSymbolTypeScopeBegin; 3112 break; 3113 3114 case N_RBRAC: 3115 // right bracket: 0,,NO_SECT,nesting level,address 3116 // Set the size of the N_LBRAC to the terminating 3117 // index of this N_RBRAC so that we can always skip 3118 // the entire symbol if we need to navigate more 3119 // quickly at the source level when parsing STABS 3120 symbol_section = section_info.GetSection( 3121 nlist.n_sect, nlist.n_value); 3122 if (!N_BRAC_indexes.empty()) { 3123 symbol_ptr = 3124 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3125 symbol_ptr->SetByteSize(sym_idx + 1); 3126 symbol_ptr->SetSizeIsSibling(true); 3127 N_BRAC_indexes.pop_back(); 3128 } 3129 type = eSymbolTypeScopeEnd; 3130 break; 3131 3132 case N_EXCL: 3133 // deleted include file: name,,NO_SECT,0,sum 3134 type = eSymbolTypeHeaderFile; 3135 break; 3136 3137 // COMM scopes 3138 case N_BCOMM: 3139 // begin common: name,,NO_SECT,0,0 3140 // We use the current number of symbols in the symbol 3141 // table in lieu of using nlist_idx in case we ever 3142 // start trimming entries out 3143 type = eSymbolTypeScopeBegin; 3144 N_COMM_indexes.push_back(sym_idx); 3145 break; 3146 3147 case N_ECOML: 3148 // end common (local name): 0,,n_sect,0,address 3149 symbol_section = section_info.GetSection( 3150 nlist.n_sect, nlist.n_value); 3151 // Fall through 3152 3153 case N_ECOMM: 3154 // end common: name,,n_sect,0,0 3155 // Set the size of the N_BCOMM to the terminating 3156 // index of this N_ECOMM/N_ECOML so that we can 3157 // always skip the entire symbol if we need to 3158 // navigate more quickly at the source level when 3159 // parsing STABS 3160 if (!N_COMM_indexes.empty()) { 3161 symbol_ptr = 3162 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3163 symbol_ptr->SetByteSize(sym_idx + 1); 3164 symbol_ptr->SetSizeIsSibling(true); 3165 N_COMM_indexes.pop_back(); 3166 } 3167 type = eSymbolTypeScopeEnd; 3168 break; 3169 3170 case N_LENG: 3171 // second stab entry with length information 3172 type = eSymbolTypeAdditional; 3173 break; 3174 3175 default: 3176 break; 3177 } 3178 } else { 3179 // uint8_t n_pext = N_PEXT & nlist.n_type; 3180 uint8_t n_type = N_TYPE & nlist.n_type; 3181 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3182 3183 switch (n_type) { 3184 case N_INDR: { 3185 const char *reexport_name_cstr = 3186 strtab_data.PeekCStr(nlist.n_value); 3187 if (reexport_name_cstr && reexport_name_cstr[0]) { 3188 type = eSymbolTypeReExported; 3189 ConstString reexport_name( 3190 reexport_name_cstr + 3191 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3192 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3193 set_value = false; 3194 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3195 indirect_symbol_names.insert( 3196 ConstString(symbol_name + 3197 ((symbol_name[0] == '_') ? 1 : 0))); 3198 } else 3199 type = eSymbolTypeUndefined; 3200 } break; 3201 3202 case N_UNDF: 3203 if (symbol_name && symbol_name[0]) { 3204 ConstString undefined_name( 3205 symbol_name + 3206 ((symbol_name[0] == '_') ? 1 : 0)); 3207 undefined_name_to_desc[undefined_name] = 3208 nlist.n_desc; 3209 } 3210 // Fall through 3211 case N_PBUD: 3212 type = eSymbolTypeUndefined; 3213 break; 3214 3215 case N_ABS: 3216 type = eSymbolTypeAbsolute; 3217 break; 3218 3219 case N_SECT: { 3220 symbol_section = section_info.GetSection( 3221 nlist.n_sect, nlist.n_value); 3222 3223 if (symbol_section == NULL) { 3224 // TODO: warn about this? 3225 add_nlist = false; 3226 break; 3227 } 3228 3229 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3230 type = eSymbolTypeException; 3231 } else { 3232 uint32_t section_type = 3233 symbol_section->Get() & SECTION_TYPE; 3234 3235 switch (section_type) { 3236 case S_CSTRING_LITERALS: 3237 type = eSymbolTypeData; 3238 break; // section with only literal C strings 3239 case S_4BYTE_LITERALS: 3240 type = eSymbolTypeData; 3241 break; // section with only 4 byte literals 3242 case S_8BYTE_LITERALS: 3243 type = eSymbolTypeData; 3244 break; // section with only 8 byte literals 3245 case S_LITERAL_POINTERS: 3246 type = eSymbolTypeTrampoline; 3247 break; // section with only pointers to literals 3248 case S_NON_LAZY_SYMBOL_POINTERS: 3249 type = eSymbolTypeTrampoline; 3250 break; // section with only non-lazy symbol 3251 // pointers 3252 case S_LAZY_SYMBOL_POINTERS: 3253 type = eSymbolTypeTrampoline; 3254 break; // section with only lazy symbol pointers 3255 case S_SYMBOL_STUBS: 3256 type = eSymbolTypeTrampoline; 3257 break; // section with only symbol stubs, byte 3258 // size of stub in the reserved2 field 3259 case S_MOD_INIT_FUNC_POINTERS: 3260 type = eSymbolTypeCode; 3261 break; // section with only function pointers for 3262 // initialization 3263 case S_MOD_TERM_FUNC_POINTERS: 3264 type = eSymbolTypeCode; 3265 break; // section with only function pointers for 3266 // termination 3267 case S_INTERPOSING: 3268 type = eSymbolTypeTrampoline; 3269 break; // section with only pairs of function 3270 // pointers for interposing 3271 case S_16BYTE_LITERALS: 3272 type = eSymbolTypeData; 3273 break; // section with only 16 byte literals 3274 case S_DTRACE_DOF: 3275 type = eSymbolTypeInstrumentation; 3276 break; 3277 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3278 type = eSymbolTypeTrampoline; 3279 break; 3280 default: 3281 switch (symbol_section->GetType()) { 3282 case lldb::eSectionTypeCode: 3283 type = eSymbolTypeCode; 3284 break; 3285 case eSectionTypeData: 3286 case eSectionTypeDataCString: // Inlined C string 3287 // data 3288 case eSectionTypeDataCStringPointers: // Pointers 3289 // to C 3290 // string 3291 // data 3292 case eSectionTypeDataSymbolAddress: // Address of 3293 // a symbol in 3294 // the symbol 3295 // table 3296 case eSectionTypeData4: 3297 case eSectionTypeData8: 3298 case eSectionTypeData16: 3299 type = eSymbolTypeData; 3300 break; 3301 default: 3302 break; 3303 } 3304 break; 3305 } 3306 3307 if (type == eSymbolTypeInvalid) { 3308 const char *symbol_sect_name = 3309 symbol_section->GetName().AsCString(); 3310 if (symbol_section->IsDescendant( 3311 text_section_sp.get())) { 3312 if (symbol_section->IsClear( 3313 S_ATTR_PURE_INSTRUCTIONS | 3314 S_ATTR_SELF_MODIFYING_CODE | 3315 S_ATTR_SOME_INSTRUCTIONS)) 3316 type = eSymbolTypeData; 3317 else 3318 type = eSymbolTypeCode; 3319 } else if (symbol_section->IsDescendant( 3320 data_section_sp.get()) || 3321 symbol_section->IsDescendant( 3322 data_dirty_section_sp.get()) || 3323 symbol_section->IsDescendant( 3324 data_const_section_sp.get())) { 3325 if (symbol_sect_name && 3326 ::strstr(symbol_sect_name, "__objc") == 3327 symbol_sect_name) { 3328 type = eSymbolTypeRuntime; 3329 3330 if (symbol_name) { 3331 llvm::StringRef symbol_name_ref( 3332 symbol_name); 3333 if (symbol_name_ref.startswith("_OBJC_")) { 3334 static const llvm::StringRef 3335 g_objc_v2_prefix_class( 3336 "_OBJC_CLASS_$_"); 3337 static const llvm::StringRef 3338 g_objc_v2_prefix_metaclass( 3339 "_OBJC_METACLASS_$_"); 3340 static const llvm::StringRef 3341 g_objc_v2_prefix_ivar( 3342 "_OBJC_IVAR_$_"); 3343 if (symbol_name_ref.startswith( 3344 g_objc_v2_prefix_class)) { 3345 symbol_name_non_abi_mangled = 3346 symbol_name + 1; 3347 symbol_name = 3348 symbol_name + 3349 g_objc_v2_prefix_class.size(); 3350 type = eSymbolTypeObjCClass; 3351 demangled_is_synthesized = true; 3352 } else if ( 3353 symbol_name_ref.startswith( 3354 g_objc_v2_prefix_metaclass)) { 3355 symbol_name_non_abi_mangled = 3356 symbol_name + 1; 3357 symbol_name = 3358 symbol_name + 3359 g_objc_v2_prefix_metaclass.size(); 3360 type = eSymbolTypeObjCMetaClass; 3361 demangled_is_synthesized = true; 3362 } else if (symbol_name_ref.startswith( 3363 g_objc_v2_prefix_ivar)) { 3364 symbol_name_non_abi_mangled = 3365 symbol_name + 1; 3366 symbol_name = 3367 symbol_name + 3368 g_objc_v2_prefix_ivar.size(); 3369 type = eSymbolTypeObjCIVar; 3370 demangled_is_synthesized = true; 3371 } 3372 } 3373 } 3374 } else if (symbol_sect_name && 3375 ::strstr(symbol_sect_name, 3376 "__gcc_except_tab") == 3377 symbol_sect_name) { 3378 type = eSymbolTypeException; 3379 } else { 3380 type = eSymbolTypeData; 3381 } 3382 } else if (symbol_sect_name && 3383 ::strstr(symbol_sect_name, 3384 "__IMPORT") == 3385 symbol_sect_name) { 3386 type = eSymbolTypeTrampoline; 3387 } else if (symbol_section->IsDescendant( 3388 objc_section_sp.get())) { 3389 type = eSymbolTypeRuntime; 3390 if (symbol_name && symbol_name[0] == '.') { 3391 llvm::StringRef symbol_name_ref(symbol_name); 3392 static const llvm::StringRef 3393 g_objc_v1_prefix_class( 3394 ".objc_class_name_"); 3395 if (symbol_name_ref.startswith( 3396 g_objc_v1_prefix_class)) { 3397 symbol_name_non_abi_mangled = symbol_name; 3398 symbol_name = symbol_name + 3399 g_objc_v1_prefix_class.size(); 3400 type = eSymbolTypeObjCClass; 3401 demangled_is_synthesized = true; 3402 } 3403 } 3404 } 3405 } 3406 } 3407 } break; 3408 } 3409 } 3410 3411 if (add_nlist) { 3412 uint64_t symbol_value = nlist.n_value; 3413 if (symbol_name_non_abi_mangled) { 3414 sym[sym_idx].GetMangled().SetMangledName( 3415 ConstString(symbol_name_non_abi_mangled)); 3416 sym[sym_idx].GetMangled().SetDemangledName( 3417 ConstString(symbol_name)); 3418 } else { 3419 bool symbol_name_is_mangled = false; 3420 3421 if (symbol_name && symbol_name[0] == '_') { 3422 symbol_name_is_mangled = symbol_name[1] == '_'; 3423 symbol_name++; // Skip the leading underscore 3424 } 3425 3426 if (symbol_name) { 3427 ConstString const_symbol_name(symbol_name); 3428 sym[sym_idx].GetMangled().SetValue( 3429 const_symbol_name, symbol_name_is_mangled); 3430 if (is_gsym && is_debug) { 3431 const char *gsym_name = 3432 sym[sym_idx] 3433 .GetMangled() 3434 .GetName(lldb::eLanguageTypeUnknown, 3435 Mangled::ePreferMangled) 3436 .GetCString(); 3437 if (gsym_name) 3438 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3439 } 3440 } 3441 } 3442 if (symbol_section) { 3443 const addr_t section_file_addr = 3444 symbol_section->GetFileAddress(); 3445 if (symbol_byte_size == 0 && 3446 function_starts_count > 0) { 3447 addr_t symbol_lookup_file_addr = nlist.n_value; 3448 // Do an exact address match for non-ARM addresses, 3449 // else get the closest since the symbol might be a 3450 // thumb symbol which has an address with bit zero 3451 // set 3452 FunctionStarts::Entry *func_start_entry = 3453 function_starts.FindEntry( 3454 symbol_lookup_file_addr, !is_arm); 3455 if (is_arm && func_start_entry) { 3456 // Verify that the function start address is the 3457 // symbol address (ARM) or the symbol address + 1 3458 // (thumb) 3459 if (func_start_entry->addr != 3460 symbol_lookup_file_addr && 3461 func_start_entry->addr != 3462 (symbol_lookup_file_addr + 1)) { 3463 // Not the right entry, NULL it out... 3464 func_start_entry = NULL; 3465 } 3466 } 3467 if (func_start_entry) { 3468 func_start_entry->data = true; 3469 3470 addr_t symbol_file_addr = func_start_entry->addr; 3471 uint32_t symbol_flags = 0; 3472 if (is_arm) { 3473 if (symbol_file_addr & 1) 3474 symbol_flags = 3475 MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3476 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3477 } 3478 3479 const FunctionStarts::Entry 3480 *next_func_start_entry = 3481 function_starts.FindNextEntry( 3482 func_start_entry); 3483 const addr_t section_end_file_addr = 3484 section_file_addr + 3485 symbol_section->GetByteSize(); 3486 if (next_func_start_entry) { 3487 addr_t next_symbol_file_addr = 3488 next_func_start_entry->addr; 3489 // Be sure the clear the Thumb address bit when 3490 // we calculate the size from the current and 3491 // next address 3492 if (is_arm) 3493 next_symbol_file_addr &= 3494 THUMB_ADDRESS_BIT_MASK; 3495 symbol_byte_size = std::min<lldb::addr_t>( 3496 next_symbol_file_addr - symbol_file_addr, 3497 section_end_file_addr - symbol_file_addr); 3498 } else { 3499 symbol_byte_size = 3500 section_end_file_addr - symbol_file_addr; 3501 } 3502 } 3503 } 3504 symbol_value -= section_file_addr; 3505 } 3506 3507 if (is_debug == false) { 3508 if (type == eSymbolTypeCode) { 3509 // See if we can find a N_FUN entry for any code 3510 // symbols. If we do find a match, and the name 3511 // matches, then we can merge the two into just the 3512 // function symbol to avoid duplicate entries in 3513 // the symbol table 3514 std::pair<ValueToSymbolIndexMap::const_iterator, 3515 ValueToSymbolIndexMap::const_iterator> 3516 range; 3517 range = N_FUN_addr_to_sym_idx.equal_range( 3518 nlist.n_value); 3519 if (range.first != range.second) { 3520 bool found_it = false; 3521 for (ValueToSymbolIndexMap::const_iterator pos = 3522 range.first; 3523 pos != range.second; ++pos) { 3524 if (sym[sym_idx].GetMangled().GetName( 3525 lldb::eLanguageTypeUnknown, 3526 Mangled::ePreferMangled) == 3527 sym[pos->second].GetMangled().GetName( 3528 lldb::eLanguageTypeUnknown, 3529 Mangled::ePreferMangled)) { 3530 m_nlist_idx_to_sym_idx[nlist_idx] = 3531 pos->second; 3532 // We just need the flags from the linker 3533 // symbol, so put these flags 3534 // into the N_FUN flags to avoid duplicate 3535 // symbols in the symbol table 3536 sym[pos->second].SetExternal( 3537 sym[sym_idx].IsExternal()); 3538 sym[pos->second].SetFlags(nlist.n_type << 16 | 3539 nlist.n_desc); 3540 if (resolver_addresses.find(nlist.n_value) != 3541 resolver_addresses.end()) 3542 sym[pos->second].SetType( 3543 eSymbolTypeResolver); 3544 sym[sym_idx].Clear(); 3545 found_it = true; 3546 break; 3547 } 3548 } 3549 if (found_it) 3550 continue; 3551 } else { 3552 if (resolver_addresses.find(nlist.n_value) != 3553 resolver_addresses.end()) 3554 type = eSymbolTypeResolver; 3555 } 3556 } else if (type == eSymbolTypeData || 3557 type == eSymbolTypeObjCClass || 3558 type == eSymbolTypeObjCMetaClass || 3559 type == eSymbolTypeObjCIVar) { 3560 // See if we can find a N_STSYM entry for any data 3561 // symbols. If we do find a match, and the name 3562 // matches, then we can merge the two into just the 3563 // Static symbol to avoid duplicate entries in the 3564 // symbol table 3565 std::pair<ValueToSymbolIndexMap::const_iterator, 3566 ValueToSymbolIndexMap::const_iterator> 3567 range; 3568 range = N_STSYM_addr_to_sym_idx.equal_range( 3569 nlist.n_value); 3570 if (range.first != range.second) { 3571 bool found_it = false; 3572 for (ValueToSymbolIndexMap::const_iterator pos = 3573 range.first; 3574 pos != range.second; ++pos) { 3575 if (sym[sym_idx].GetMangled().GetName( 3576 lldb::eLanguageTypeUnknown, 3577 Mangled::ePreferMangled) == 3578 sym[pos->second].GetMangled().GetName( 3579 lldb::eLanguageTypeUnknown, 3580 Mangled::ePreferMangled)) { 3581 m_nlist_idx_to_sym_idx[nlist_idx] = 3582 pos->second; 3583 // We just need the flags from the linker 3584 // symbol, so put these flags 3585 // into the N_STSYM flags to avoid duplicate 3586 // symbols in the symbol table 3587 sym[pos->second].SetExternal( 3588 sym[sym_idx].IsExternal()); 3589 sym[pos->second].SetFlags(nlist.n_type << 16 | 3590 nlist.n_desc); 3591 sym[sym_idx].Clear(); 3592 found_it = true; 3593 break; 3594 } 3595 } 3596 if (found_it) 3597 continue; 3598 } else { 3599 const char *gsym_name = 3600 sym[sym_idx] 3601 .GetMangled() 3602 .GetName(lldb::eLanguageTypeUnknown, 3603 Mangled::ePreferMangled) 3604 .GetCString(); 3605 if (gsym_name) { 3606 // Combine N_GSYM stab entries with the non 3607 // stab symbol 3608 ConstNameToSymbolIndexMap::const_iterator pos = 3609 N_GSYM_name_to_sym_idx.find(gsym_name); 3610 if (pos != N_GSYM_name_to_sym_idx.end()) { 3611 const uint32_t GSYM_sym_idx = pos->second; 3612 m_nlist_idx_to_sym_idx[nlist_idx] = 3613 GSYM_sym_idx; 3614 // Copy the address, because often the N_GSYM 3615 // address has an invalid address of zero 3616 // when the global is a common symbol 3617 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3618 symbol_section); 3619 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3620 symbol_value); 3621 // We just need the flags from the linker 3622 // symbol, so put these flags 3623 // into the N_GSYM flags to avoid duplicate 3624 // symbols in the symbol table 3625 sym[GSYM_sym_idx].SetFlags( 3626 nlist.n_type << 16 | nlist.n_desc); 3627 sym[sym_idx].Clear(); 3628 continue; 3629 } 3630 } 3631 } 3632 } 3633 } 3634 3635 sym[sym_idx].SetID(nlist_idx); 3636 sym[sym_idx].SetType(type); 3637 if (set_value) { 3638 sym[sym_idx].GetAddressRef().SetSection( 3639 symbol_section); 3640 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3641 } 3642 sym[sym_idx].SetFlags(nlist.n_type << 16 | 3643 nlist.n_desc); 3644 3645 if (symbol_byte_size > 0) 3646 sym[sym_idx].SetByteSize(symbol_byte_size); 3647 3648 if (demangled_is_synthesized) 3649 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3650 ++sym_idx; 3651 } else { 3652 sym[sym_idx].Clear(); 3653 } 3654 } 3655 ///////////////////////////// 3656 } 3657 break; // No more entries to consider 3658 } 3659 } 3660 3661 for (const auto &pos : reexport_shlib_needs_fixup) { 3662 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3663 if (undef_pos != undefined_name_to_desc.end()) { 3664 const uint8_t dylib_ordinal = 3665 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3666 if (dylib_ordinal > 0 && 3667 dylib_ordinal < dylib_files.GetSize()) 3668 sym[pos.first].SetReExportedSymbolSharedLibrary( 3669 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3670 } 3671 } 3672 } 3673 } 3674 } 3675 } 3676 } 3677 3678 // Must reset this in case it was mutated above! 3679 nlist_data_offset = 0; 3680 #endif 3681 3682 if (nlist_data.GetByteSize() > 0) { 3683 3684 // If the sym array was not created while parsing the DSC unmapped 3685 // symbols, create it now. 3686 if (sym == nullptr) { 3687 sym = symtab->Resize(symtab_load_command.nsyms + 3688 m_dysymtab.nindirectsyms); 3689 num_syms = symtab->GetNumSymbols(); 3690 } 3691 3692 if (unmapped_local_symbols_found) { 3693 assert(m_dysymtab.ilocalsym == 0); 3694 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3695 nlist_idx = m_dysymtab.nlocalsym; 3696 } else { 3697 nlist_idx = 0; 3698 } 3699 3700 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3701 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3702 UndefinedNameToDescMap undefined_name_to_desc; 3703 SymbolIndexToName reexport_shlib_needs_fixup; 3704 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 3705 struct nlist_64 nlist; 3706 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, 3707 nlist_byte_size)) 3708 break; 3709 3710 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3711 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 3712 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 3713 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 3714 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 3715 3716 SymbolType type = eSymbolTypeInvalid; 3717 const char *symbol_name = nullptr; 3718 3719 if (have_strtab_data) { 3720 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3721 3722 if (symbol_name == nullptr) { 3723 // No symbol should be NULL, even the symbols with no string values 3724 // should have an offset zero which points to an empty C-string 3725 Host::SystemLog(Host::eSystemLogError, 3726 "error: symbol[%u] has invalid string table offset " 3727 "0x%x in %s, ignoring symbol\n", 3728 nlist_idx, nlist.n_strx, 3729 module_sp->GetFileSpec().GetPath().c_str()); 3730 continue; 3731 } 3732 if (symbol_name[0] == '\0') 3733 symbol_name = nullptr; 3734 } else { 3735 const addr_t str_addr = strtab_addr + nlist.n_strx; 3736 Status str_error; 3737 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3738 str_error)) 3739 symbol_name = memory_symbol_name.c_str(); 3740 } 3741 const char *symbol_name_non_abi_mangled = nullptr; 3742 3743 SectionSP symbol_section; 3744 lldb::addr_t symbol_byte_size = 0; 3745 bool add_nlist = true; 3746 bool is_gsym = false; 3747 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3748 bool demangled_is_synthesized = false; 3749 bool set_value = true; 3750 assert(sym_idx < num_syms); 3751 3752 sym[sym_idx].SetDebug(is_debug); 3753 3754 if (is_debug) { 3755 switch (nlist.n_type) { 3756 case N_GSYM: 3757 // global symbol: name,,NO_SECT,type,0 3758 // Sometimes the N_GSYM value contains the address. 3759 3760 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3761 // the ObjC data. They 3762 // have the same address, but we want to ensure that we always find 3763 // only the real symbol, 'cause we don't currently correctly 3764 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3765 // type. This is a temporary hack to make sure the ObjectiveC 3766 // symbols get treated correctly. To do this right, we should 3767 // coalesce all the GSYM & global symbols that have the same 3768 // address. 3769 is_gsym = true; 3770 sym[sym_idx].SetExternal(true); 3771 3772 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3773 llvm::StringRef symbol_name_ref(symbol_name); 3774 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3775 symbol_name_non_abi_mangled = symbol_name + 1; 3776 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3777 type = eSymbolTypeObjCClass; 3778 demangled_is_synthesized = true; 3779 3780 } else if (symbol_name_ref.startswith( 3781 g_objc_v2_prefix_metaclass)) { 3782 symbol_name_non_abi_mangled = symbol_name + 1; 3783 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3784 type = eSymbolTypeObjCMetaClass; 3785 demangled_is_synthesized = true; 3786 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3787 symbol_name_non_abi_mangled = symbol_name + 1; 3788 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3789 type = eSymbolTypeObjCIVar; 3790 demangled_is_synthesized = true; 3791 } 3792 } else { 3793 if (nlist.n_value != 0) 3794 symbol_section = 3795 section_info.GetSection(nlist.n_sect, nlist.n_value); 3796 type = eSymbolTypeData; 3797 } 3798 break; 3799 3800 case N_FNAME: 3801 // procedure name (f77 kludge): name,,NO_SECT,0,0 3802 type = eSymbolTypeCompiler; 3803 break; 3804 3805 case N_FUN: 3806 // procedure: name,,n_sect,linenumber,address 3807 if (symbol_name) { 3808 type = eSymbolTypeCode; 3809 symbol_section = 3810 section_info.GetSection(nlist.n_sect, nlist.n_value); 3811 3812 N_FUN_addr_to_sym_idx.insert( 3813 std::make_pair(nlist.n_value, sym_idx)); 3814 // We use the current number of symbols in the symbol table in 3815 // lieu of using nlist_idx in case we ever start trimming entries 3816 // out 3817 N_FUN_indexes.push_back(sym_idx); 3818 } else { 3819 type = eSymbolTypeCompiler; 3820 3821 if (!N_FUN_indexes.empty()) { 3822 // Copy the size of the function into the original STAB entry 3823 // so we don't have to hunt for it later 3824 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3825 ->SetByteSize(nlist.n_value); 3826 N_FUN_indexes.pop_back(); 3827 // We don't really need the end function STAB as it contains 3828 // the size which we already placed with the original symbol, 3829 // so don't add it if we want a minimal symbol table 3830 add_nlist = false; 3831 } 3832 } 3833 break; 3834 3835 case N_STSYM: 3836 // static symbol: name,,n_sect,type,address 3837 N_STSYM_addr_to_sym_idx.insert( 3838 std::make_pair(nlist.n_value, sym_idx)); 3839 symbol_section = 3840 section_info.GetSection(nlist.n_sect, nlist.n_value); 3841 if (symbol_name && symbol_name[0]) { 3842 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3843 eSymbolTypeData); 3844 } 3845 break; 3846 3847 case N_LCSYM: 3848 // .lcomm symbol: name,,n_sect,type,address 3849 symbol_section = 3850 section_info.GetSection(nlist.n_sect, nlist.n_value); 3851 type = eSymbolTypeCommonBlock; 3852 break; 3853 3854 case N_BNSYM: 3855 // We use the current number of symbols in the symbol table in lieu 3856 // of using nlist_idx in case we ever start trimming entries out 3857 // Skip these if we want minimal symbol tables 3858 add_nlist = false; 3859 break; 3860 3861 case N_ENSYM: 3862 // Set the size of the N_BNSYM to the terminating index of this 3863 // N_ENSYM so that we can always skip the entire symbol if we need 3864 // to navigate more quickly at the source level when parsing STABS 3865 // Skip these if we want minimal symbol tables 3866 add_nlist = false; 3867 break; 3868 3869 case N_OPT: 3870 // emitted with gcc2_compiled and in gcc source 3871 type = eSymbolTypeCompiler; 3872 break; 3873 3874 case N_RSYM: 3875 // register sym: name,,NO_SECT,type,register 3876 type = eSymbolTypeVariable; 3877 break; 3878 3879 case N_SLINE: 3880 // src line: 0,,n_sect,linenumber,address 3881 symbol_section = 3882 section_info.GetSection(nlist.n_sect, nlist.n_value); 3883 type = eSymbolTypeLineEntry; 3884 break; 3885 3886 case N_SSYM: 3887 // structure elt: name,,NO_SECT,type,struct_offset 3888 type = eSymbolTypeVariableType; 3889 break; 3890 3891 case N_SO: 3892 // source file name 3893 type = eSymbolTypeSourceFile; 3894 if (symbol_name == nullptr) { 3895 add_nlist = false; 3896 if (N_SO_index != UINT32_MAX) { 3897 // Set the size of the N_SO to the terminating index of this 3898 // N_SO so that we can always skip the entire N_SO if we need 3899 // to navigate more quickly at the source level when parsing 3900 // STABS 3901 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3902 symbol_ptr->SetByteSize(sym_idx); 3903 symbol_ptr->SetSizeIsSibling(true); 3904 } 3905 N_NSYM_indexes.clear(); 3906 N_INCL_indexes.clear(); 3907 N_BRAC_indexes.clear(); 3908 N_COMM_indexes.clear(); 3909 N_FUN_indexes.clear(); 3910 N_SO_index = UINT32_MAX; 3911 } else { 3912 // We use the current number of symbols in the symbol table in 3913 // lieu of using nlist_idx in case we ever start trimming entries 3914 // out 3915 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3916 if (N_SO_has_full_path) { 3917 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3918 // We have two consecutive N_SO entries where the first 3919 // contains a directory and the second contains a full path. 3920 sym[sym_idx - 1].GetMangled().SetValue( 3921 ConstString(symbol_name), false); 3922 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3923 add_nlist = false; 3924 } else { 3925 // This is the first entry in a N_SO that contains a 3926 // directory or a full path to the source file 3927 N_SO_index = sym_idx; 3928 } 3929 } else if ((N_SO_index == sym_idx - 1) && 3930 ((sym_idx - 1) < num_syms)) { 3931 // This is usually the second N_SO entry that contains just the 3932 // filename, so here we combine it with the first one if we are 3933 // minimizing the symbol table 3934 const char *so_path = 3935 sym[sym_idx - 1] 3936 .GetMangled() 3937 .GetDemangledName(lldb::eLanguageTypeUnknown) 3938 .AsCString(); 3939 if (so_path && so_path[0]) { 3940 std::string full_so_path(so_path); 3941 const size_t double_slash_pos = full_so_path.find("//"); 3942 if (double_slash_pos != std::string::npos) { 3943 // The linker has been generating bad N_SO entries with 3944 // doubled up paths in the format "%s%s" where the first 3945 // string in the DW_AT_comp_dir, and the second is the 3946 // directory for the source file so you end up with a path 3947 // that looks like "/tmp/src//tmp/src/" 3948 FileSpec so_dir(so_path); 3949 if (!FileSystem::Instance().Exists(so_dir)) { 3950 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 3951 FileSpec::Style::native); 3952 if (FileSystem::Instance().Exists(so_dir)) { 3953 // Trim off the incorrect path 3954 full_so_path.erase(0, double_slash_pos + 1); 3955 } 3956 } 3957 } 3958 if (*full_so_path.rbegin() != '/') 3959 full_so_path += '/'; 3960 full_so_path += symbol_name; 3961 sym[sym_idx - 1].GetMangled().SetValue( 3962 ConstString(full_so_path.c_str()), false); 3963 add_nlist = false; 3964 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3965 } 3966 } else { 3967 // This could be a relative path to a N_SO 3968 N_SO_index = sym_idx; 3969 } 3970 } 3971 break; 3972 3973 case N_OSO: 3974 // object file name: name,,0,0,st_mtime 3975 type = eSymbolTypeObjectFile; 3976 break; 3977 3978 case N_LSYM: 3979 // local sym: name,,NO_SECT,type,offset 3980 type = eSymbolTypeLocal; 3981 break; 3982 3983 // INCL scopes 3984 case N_BINCL: 3985 // include file beginning: name,,NO_SECT,0,sum We use the current 3986 // number of symbols in the symbol table in lieu of using nlist_idx 3987 // in case we ever start trimming entries out 3988 N_INCL_indexes.push_back(sym_idx); 3989 type = eSymbolTypeScopeBegin; 3990 break; 3991 3992 case N_EINCL: 3993 // include file end: name,,NO_SECT,0,0 3994 // Set the size of the N_BINCL to the terminating index of this 3995 // N_EINCL so that we can always skip the entire symbol if we need 3996 // to navigate more quickly at the source level when parsing STABS 3997 if (!N_INCL_indexes.empty()) { 3998 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3999 symbol_ptr->SetByteSize(sym_idx + 1); 4000 symbol_ptr->SetSizeIsSibling(true); 4001 N_INCL_indexes.pop_back(); 4002 } 4003 type = eSymbolTypeScopeEnd; 4004 break; 4005 4006 case N_SOL: 4007 // #included file name: name,,n_sect,0,address 4008 type = eSymbolTypeHeaderFile; 4009 4010 // We currently don't use the header files on darwin 4011 add_nlist = false; 4012 break; 4013 4014 case N_PARAMS: 4015 // compiler parameters: name,,NO_SECT,0,0 4016 type = eSymbolTypeCompiler; 4017 break; 4018 4019 case N_VERSION: 4020 // compiler version: name,,NO_SECT,0,0 4021 type = eSymbolTypeCompiler; 4022 break; 4023 4024 case N_OLEVEL: 4025 // compiler -O level: name,,NO_SECT,0,0 4026 type = eSymbolTypeCompiler; 4027 break; 4028 4029 case N_PSYM: 4030 // parameter: name,,NO_SECT,type,offset 4031 type = eSymbolTypeVariable; 4032 break; 4033 4034 case N_ENTRY: 4035 // alternate entry: name,,n_sect,linenumber,address 4036 symbol_section = 4037 section_info.GetSection(nlist.n_sect, nlist.n_value); 4038 type = eSymbolTypeLineEntry; 4039 break; 4040 4041 // Left and Right Braces 4042 case N_LBRAC: 4043 // left bracket: 0,,NO_SECT,nesting level,address We use the 4044 // current number of symbols in the symbol table in lieu of using 4045 // nlist_idx in case we ever start trimming entries out 4046 symbol_section = 4047 section_info.GetSection(nlist.n_sect, nlist.n_value); 4048 N_BRAC_indexes.push_back(sym_idx); 4049 type = eSymbolTypeScopeBegin; 4050 break; 4051 4052 case N_RBRAC: 4053 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4054 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4055 // can always skip the entire symbol if we need to navigate more 4056 // quickly at the source level when parsing STABS 4057 symbol_section = 4058 section_info.GetSection(nlist.n_sect, nlist.n_value); 4059 if (!N_BRAC_indexes.empty()) { 4060 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4061 symbol_ptr->SetByteSize(sym_idx + 1); 4062 symbol_ptr->SetSizeIsSibling(true); 4063 N_BRAC_indexes.pop_back(); 4064 } 4065 type = eSymbolTypeScopeEnd; 4066 break; 4067 4068 case N_EXCL: 4069 // deleted include file: name,,NO_SECT,0,sum 4070 type = eSymbolTypeHeaderFile; 4071 break; 4072 4073 // COMM scopes 4074 case N_BCOMM: 4075 // begin common: name,,NO_SECT,0,0 4076 // We use the current number of symbols in the symbol table in lieu 4077 // of using nlist_idx in case we ever start trimming entries out 4078 type = eSymbolTypeScopeBegin; 4079 N_COMM_indexes.push_back(sym_idx); 4080 break; 4081 4082 case N_ECOML: 4083 // end common (local name): 0,,n_sect,0,address 4084 symbol_section = 4085 section_info.GetSection(nlist.n_sect, nlist.n_value); 4086 LLVM_FALLTHROUGH; 4087 4088 case N_ECOMM: 4089 // end common: name,,n_sect,0,0 4090 // Set the size of the N_BCOMM to the terminating index of this 4091 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4092 // we need to navigate more quickly at the source level when 4093 // parsing STABS 4094 if (!N_COMM_indexes.empty()) { 4095 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4096 symbol_ptr->SetByteSize(sym_idx + 1); 4097 symbol_ptr->SetSizeIsSibling(true); 4098 N_COMM_indexes.pop_back(); 4099 } 4100 type = eSymbolTypeScopeEnd; 4101 break; 4102 4103 case N_LENG: 4104 // second stab entry with length information 4105 type = eSymbolTypeAdditional; 4106 break; 4107 4108 default: 4109 break; 4110 } 4111 } else { 4112 // uint8_t n_pext = N_PEXT & nlist.n_type; 4113 uint8_t n_type = N_TYPE & nlist.n_type; 4114 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4115 4116 switch (n_type) { 4117 case N_INDR: { 4118 const char *reexport_name_cstr = 4119 strtab_data.PeekCStr(nlist.n_value); 4120 if (reexport_name_cstr && reexport_name_cstr[0]) { 4121 type = eSymbolTypeReExported; 4122 ConstString reexport_name( 4123 reexport_name_cstr + 4124 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4125 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4126 set_value = false; 4127 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4128 indirect_symbol_names.insert( 4129 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4130 } else 4131 type = eSymbolTypeUndefined; 4132 } break; 4133 4134 case N_UNDF: 4135 if (symbol_name && symbol_name[0]) { 4136 ConstString undefined_name(symbol_name + 4137 ((symbol_name[0] == '_') ? 1 : 0)); 4138 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4139 } 4140 LLVM_FALLTHROUGH; 4141 4142 case N_PBUD: 4143 type = eSymbolTypeUndefined; 4144 break; 4145 4146 case N_ABS: 4147 type = eSymbolTypeAbsolute; 4148 break; 4149 4150 case N_SECT: { 4151 symbol_section = 4152 section_info.GetSection(nlist.n_sect, nlist.n_value); 4153 4154 if (!symbol_section) { 4155 // TODO: warn about this? 4156 add_nlist = false; 4157 break; 4158 } 4159 4160 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4161 type = eSymbolTypeException; 4162 } else { 4163 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4164 4165 switch (section_type) { 4166 case S_CSTRING_LITERALS: 4167 type = eSymbolTypeData; 4168 break; // section with only literal C strings 4169 case S_4BYTE_LITERALS: 4170 type = eSymbolTypeData; 4171 break; // section with only 4 byte literals 4172 case S_8BYTE_LITERALS: 4173 type = eSymbolTypeData; 4174 break; // section with only 8 byte literals 4175 case S_LITERAL_POINTERS: 4176 type = eSymbolTypeTrampoline; 4177 break; // section with only pointers to literals 4178 case S_NON_LAZY_SYMBOL_POINTERS: 4179 type = eSymbolTypeTrampoline; 4180 break; // section with only non-lazy symbol pointers 4181 case S_LAZY_SYMBOL_POINTERS: 4182 type = eSymbolTypeTrampoline; 4183 break; // section with only lazy symbol pointers 4184 case S_SYMBOL_STUBS: 4185 type = eSymbolTypeTrampoline; 4186 break; // section with only symbol stubs, byte size of stub in 4187 // the reserved2 field 4188 case S_MOD_INIT_FUNC_POINTERS: 4189 type = eSymbolTypeCode; 4190 break; // section with only function pointers for initialization 4191 case S_MOD_TERM_FUNC_POINTERS: 4192 type = eSymbolTypeCode; 4193 break; // section with only function pointers for termination 4194 case S_INTERPOSING: 4195 type = eSymbolTypeTrampoline; 4196 break; // section with only pairs of function pointers for 4197 // interposing 4198 case S_16BYTE_LITERALS: 4199 type = eSymbolTypeData; 4200 break; // section with only 16 byte literals 4201 case S_DTRACE_DOF: 4202 type = eSymbolTypeInstrumentation; 4203 break; 4204 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4205 type = eSymbolTypeTrampoline; 4206 break; 4207 default: 4208 switch (symbol_section->GetType()) { 4209 case lldb::eSectionTypeCode: 4210 type = eSymbolTypeCode; 4211 break; 4212 case eSectionTypeData: 4213 case eSectionTypeDataCString: // Inlined C string data 4214 case eSectionTypeDataCStringPointers: // Pointers to C string 4215 // data 4216 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4217 // the symbol table 4218 case eSectionTypeData4: 4219 case eSectionTypeData8: 4220 case eSectionTypeData16: 4221 type = eSymbolTypeData; 4222 break; 4223 default: 4224 break; 4225 } 4226 break; 4227 } 4228 4229 if (type == eSymbolTypeInvalid) { 4230 const char *symbol_sect_name = 4231 symbol_section->GetName().AsCString(); 4232 if (symbol_section->IsDescendant(text_section_sp.get())) { 4233 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4234 S_ATTR_SELF_MODIFYING_CODE | 4235 S_ATTR_SOME_INSTRUCTIONS)) 4236 type = eSymbolTypeData; 4237 else 4238 type = eSymbolTypeCode; 4239 } else if (symbol_section->IsDescendant( 4240 data_section_sp.get()) || 4241 symbol_section->IsDescendant( 4242 data_dirty_section_sp.get()) || 4243 symbol_section->IsDescendant( 4244 data_const_section_sp.get())) { 4245 if (symbol_sect_name && 4246 ::strstr(symbol_sect_name, "__objc") == 4247 symbol_sect_name) { 4248 type = eSymbolTypeRuntime; 4249 4250 if (symbol_name) { 4251 llvm::StringRef symbol_name_ref(symbol_name); 4252 if (symbol_name_ref.startswith("_OBJC_")) { 4253 static const llvm::StringRef g_objc_v2_prefix_class( 4254 "_OBJC_CLASS_$_"); 4255 static const llvm::StringRef g_objc_v2_prefix_metaclass( 4256 "_OBJC_METACLASS_$_"); 4257 static const llvm::StringRef g_objc_v2_prefix_ivar( 4258 "_OBJC_IVAR_$_"); 4259 if (symbol_name_ref.startswith( 4260 g_objc_v2_prefix_class)) { 4261 symbol_name_non_abi_mangled = symbol_name + 1; 4262 symbol_name = 4263 symbol_name + g_objc_v2_prefix_class.size(); 4264 type = eSymbolTypeObjCClass; 4265 demangled_is_synthesized = true; 4266 } else if (symbol_name_ref.startswith( 4267 g_objc_v2_prefix_metaclass)) { 4268 symbol_name_non_abi_mangled = symbol_name + 1; 4269 symbol_name = 4270 symbol_name + g_objc_v2_prefix_metaclass.size(); 4271 type = eSymbolTypeObjCMetaClass; 4272 demangled_is_synthesized = true; 4273 } else if (symbol_name_ref.startswith( 4274 g_objc_v2_prefix_ivar)) { 4275 symbol_name_non_abi_mangled = symbol_name + 1; 4276 symbol_name = 4277 symbol_name + g_objc_v2_prefix_ivar.size(); 4278 type = eSymbolTypeObjCIVar; 4279 demangled_is_synthesized = true; 4280 } 4281 } 4282 } 4283 } else if (symbol_sect_name && 4284 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4285 symbol_sect_name) { 4286 type = eSymbolTypeException; 4287 } else { 4288 type = eSymbolTypeData; 4289 } 4290 } else if (symbol_sect_name && 4291 ::strstr(symbol_sect_name, "__IMPORT") == 4292 symbol_sect_name) { 4293 type = eSymbolTypeTrampoline; 4294 } else if (symbol_section->IsDescendant( 4295 objc_section_sp.get())) { 4296 type = eSymbolTypeRuntime; 4297 if (symbol_name && symbol_name[0] == '.') { 4298 llvm::StringRef symbol_name_ref(symbol_name); 4299 static const llvm::StringRef g_objc_v1_prefix_class( 4300 ".objc_class_name_"); 4301 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4302 symbol_name_non_abi_mangled = symbol_name; 4303 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4304 type = eSymbolTypeObjCClass; 4305 demangled_is_synthesized = true; 4306 } 4307 } 4308 } 4309 } 4310 } 4311 } break; 4312 } 4313 } 4314 4315 if (add_nlist) { 4316 uint64_t symbol_value = nlist.n_value; 4317 4318 if (symbol_name_non_abi_mangled) { 4319 sym[sym_idx].GetMangled().SetMangledName( 4320 ConstString(symbol_name_non_abi_mangled)); 4321 sym[sym_idx].GetMangled().SetDemangledName( 4322 ConstString(symbol_name)); 4323 } else { 4324 bool symbol_name_is_mangled = false; 4325 4326 if (symbol_name && symbol_name[0] == '_') { 4327 symbol_name_is_mangled = symbol_name[1] == '_'; 4328 symbol_name++; // Skip the leading underscore 4329 } 4330 4331 if (symbol_name) { 4332 ConstString const_symbol_name(symbol_name); 4333 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4334 symbol_name_is_mangled); 4335 } 4336 } 4337 4338 if (is_gsym) { 4339 const char *gsym_name = sym[sym_idx] 4340 .GetMangled() 4341 .GetName(lldb::eLanguageTypeUnknown, 4342 Mangled::ePreferMangled) 4343 .GetCString(); 4344 if (gsym_name) 4345 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4346 } 4347 4348 if (symbol_section) { 4349 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4350 if (symbol_byte_size == 0 && function_starts_count > 0) { 4351 addr_t symbol_lookup_file_addr = nlist.n_value; 4352 // Do an exact address match for non-ARM addresses, else get the 4353 // closest since the symbol might be a thumb symbol which has an 4354 // address with bit zero set 4355 FunctionStarts::Entry *func_start_entry = 4356 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4357 if (is_arm && func_start_entry) { 4358 // Verify that the function start address is the symbol address 4359 // (ARM) or the symbol address + 1 (thumb) 4360 if (func_start_entry->addr != symbol_lookup_file_addr && 4361 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4362 // Not the right entry, NULL it out... 4363 func_start_entry = nullptr; 4364 } 4365 } 4366 if (func_start_entry) { 4367 func_start_entry->data = true; 4368 4369 addr_t symbol_file_addr = func_start_entry->addr; 4370 if (is_arm) 4371 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4372 4373 const FunctionStarts::Entry *next_func_start_entry = 4374 function_starts.FindNextEntry(func_start_entry); 4375 const addr_t section_end_file_addr = 4376 section_file_addr + symbol_section->GetByteSize(); 4377 if (next_func_start_entry) { 4378 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4379 // Be sure the clear the Thumb address bit when we calculate 4380 // the size from the current and next address 4381 if (is_arm) 4382 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4383 symbol_byte_size = std::min<lldb::addr_t>( 4384 next_symbol_file_addr - symbol_file_addr, 4385 section_end_file_addr - symbol_file_addr); 4386 } else { 4387 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4388 } 4389 } 4390 } 4391 symbol_value -= section_file_addr; 4392 } 4393 4394 if (!is_debug) { 4395 if (type == eSymbolTypeCode) { 4396 // See if we can find a N_FUN entry for any code symbols. If we 4397 // do find a match, and the name matches, then we can merge the 4398 // two into just the function symbol to avoid duplicate entries 4399 // in the symbol table 4400 std::pair<ValueToSymbolIndexMap::const_iterator, 4401 ValueToSymbolIndexMap::const_iterator> 4402 range; 4403 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4404 if (range.first != range.second) { 4405 bool found_it = false; 4406 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4407 pos != range.second; ++pos) { 4408 if (sym[sym_idx].GetMangled().GetName( 4409 lldb::eLanguageTypeUnknown, 4410 Mangled::ePreferMangled) == 4411 sym[pos->second].GetMangled().GetName( 4412 lldb::eLanguageTypeUnknown, 4413 Mangled::ePreferMangled)) { 4414 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4415 // We just need the flags from the linker symbol, so put 4416 // these flags into the N_FUN flags to avoid duplicate 4417 // symbols in the symbol table 4418 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4419 sym[pos->second].SetFlags(nlist.n_type << 16 | 4420 nlist.n_desc); 4421 if (resolver_addresses.find(nlist.n_value) != 4422 resolver_addresses.end()) 4423 sym[pos->second].SetType(eSymbolTypeResolver); 4424 sym[sym_idx].Clear(); 4425 found_it = true; 4426 break; 4427 } 4428 } 4429 if (found_it) 4430 continue; 4431 } else { 4432 if (resolver_addresses.find(nlist.n_value) != 4433 resolver_addresses.end()) 4434 type = eSymbolTypeResolver; 4435 } 4436 } else if (type == eSymbolTypeData || 4437 type == eSymbolTypeObjCClass || 4438 type == eSymbolTypeObjCMetaClass || 4439 type == eSymbolTypeObjCIVar) { 4440 // See if we can find a N_STSYM entry for any data symbols. If we 4441 // do find a match, and the name matches, then we can merge the 4442 // two into just the Static symbol to avoid duplicate entries in 4443 // the symbol table 4444 std::pair<ValueToSymbolIndexMap::const_iterator, 4445 ValueToSymbolIndexMap::const_iterator> 4446 range; 4447 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4448 if (range.first != range.second) { 4449 bool found_it = false; 4450 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4451 pos != range.second; ++pos) { 4452 if (sym[sym_idx].GetMangled().GetName( 4453 lldb::eLanguageTypeUnknown, 4454 Mangled::ePreferMangled) == 4455 sym[pos->second].GetMangled().GetName( 4456 lldb::eLanguageTypeUnknown, 4457 Mangled::ePreferMangled)) { 4458 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4459 // We just need the flags from the linker symbol, so put 4460 // these flags into the N_STSYM flags to avoid duplicate 4461 // symbols in the symbol table 4462 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4463 sym[pos->second].SetFlags(nlist.n_type << 16 | 4464 nlist.n_desc); 4465 sym[sym_idx].Clear(); 4466 found_it = true; 4467 break; 4468 } 4469 } 4470 if (found_it) 4471 continue; 4472 } else { 4473 // Combine N_GSYM stab entries with the non stab symbol 4474 const char *gsym_name = sym[sym_idx] 4475 .GetMangled() 4476 .GetName(lldb::eLanguageTypeUnknown, 4477 Mangled::ePreferMangled) 4478 .GetCString(); 4479 if (gsym_name) { 4480 ConstNameToSymbolIndexMap::const_iterator pos = 4481 N_GSYM_name_to_sym_idx.find(gsym_name); 4482 if (pos != N_GSYM_name_to_sym_idx.end()) { 4483 const uint32_t GSYM_sym_idx = pos->second; 4484 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4485 // Copy the address, because often the N_GSYM address has 4486 // an invalid address of zero when the global is a common 4487 // symbol 4488 sym[GSYM_sym_idx].GetAddressRef().SetSection( 4489 symbol_section); 4490 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4491 // We just need the flags from the linker symbol, so put 4492 // these flags into the N_GSYM flags to avoid duplicate 4493 // symbols in the symbol table 4494 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 4495 nlist.n_desc); 4496 sym[sym_idx].Clear(); 4497 continue; 4498 } 4499 } 4500 } 4501 } 4502 } 4503 4504 sym[sym_idx].SetID(nlist_idx); 4505 sym[sym_idx].SetType(type); 4506 if (set_value) { 4507 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4508 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4509 } 4510 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4511 if (nlist.n_desc & N_WEAK_REF) 4512 sym[sym_idx].SetIsWeak(true); 4513 4514 if (symbol_byte_size > 0) 4515 sym[sym_idx].SetByteSize(symbol_byte_size); 4516 4517 if (demangled_is_synthesized) 4518 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4519 4520 ++sym_idx; 4521 } else { 4522 sym[sym_idx].Clear(); 4523 } 4524 } 4525 4526 for (const auto &pos : reexport_shlib_needs_fixup) { 4527 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4528 if (undef_pos != undefined_name_to_desc.end()) { 4529 const uint8_t dylib_ordinal = 4530 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4531 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4532 sym[pos.first].SetReExportedSymbolSharedLibrary( 4533 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4534 } 4535 } 4536 } 4537 4538 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4539 4540 if (function_starts_count > 0) { 4541 uint32_t num_synthetic_function_symbols = 0; 4542 for (i = 0; i < function_starts_count; ++i) { 4543 if (!function_starts.GetEntryRef(i).data) 4544 ++num_synthetic_function_symbols; 4545 } 4546 4547 if (num_synthetic_function_symbols > 0) { 4548 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4549 num_syms = sym_idx + num_synthetic_function_symbols; 4550 sym = symtab->Resize(num_syms); 4551 } 4552 for (i = 0; i < function_starts_count; ++i) { 4553 const FunctionStarts::Entry *func_start_entry = 4554 function_starts.GetEntryAtIndex(i); 4555 if (!func_start_entry->data) { 4556 addr_t symbol_file_addr = func_start_entry->addr; 4557 uint32_t symbol_flags = 0; 4558 if (is_arm) { 4559 if (symbol_file_addr & 1) 4560 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4561 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4562 } 4563 Address symbol_addr; 4564 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4565 SectionSP symbol_section(symbol_addr.GetSection()); 4566 uint32_t symbol_byte_size = 0; 4567 if (symbol_section) { 4568 const addr_t section_file_addr = 4569 symbol_section->GetFileAddress(); 4570 const FunctionStarts::Entry *next_func_start_entry = 4571 function_starts.FindNextEntry(func_start_entry); 4572 const addr_t section_end_file_addr = 4573 section_file_addr + symbol_section->GetByteSize(); 4574 if (next_func_start_entry) { 4575 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4576 if (is_arm) 4577 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4578 symbol_byte_size = std::min<lldb::addr_t>( 4579 next_symbol_file_addr - symbol_file_addr, 4580 section_end_file_addr - symbol_file_addr); 4581 } else { 4582 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4583 } 4584 sym[sym_idx].SetID(synthetic_sym_id++); 4585 sym[sym_idx].GetMangled().SetDemangledName( 4586 GetNextSyntheticSymbolName()); 4587 sym[sym_idx].SetType(eSymbolTypeCode); 4588 sym[sym_idx].SetIsSynthetic(true); 4589 sym[sym_idx].GetAddressRef() = symbol_addr; 4590 if (symbol_flags) 4591 sym[sym_idx].SetFlags(symbol_flags); 4592 if (symbol_byte_size) 4593 sym[sym_idx].SetByteSize(symbol_byte_size); 4594 ++sym_idx; 4595 } 4596 } 4597 } 4598 } 4599 } 4600 } 4601 4602 // Trim our symbols down to just what we ended up with after removing any 4603 // symbols. 4604 if (sym_idx < num_syms) { 4605 num_syms = sym_idx; 4606 sym = symtab->Resize(num_syms); 4607 } 4608 4609 // Now synthesize indirect symbols 4610 if (m_dysymtab.nindirectsyms != 0) { 4611 if (indirect_symbol_index_data.GetByteSize()) { 4612 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4613 m_nlist_idx_to_sym_idx.end(); 4614 4615 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4616 ++sect_idx) { 4617 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4618 S_SYMBOL_STUBS) { 4619 uint32_t symbol_stub_byte_size = 4620 m_mach_sections[sect_idx].reserved2; 4621 if (symbol_stub_byte_size == 0) 4622 continue; 4623 4624 const uint32_t num_symbol_stubs = 4625 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4626 4627 if (num_symbol_stubs == 0) 4628 continue; 4629 4630 const uint32_t symbol_stub_index_offset = 4631 m_mach_sections[sect_idx].reserved1; 4632 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; 4633 ++stub_idx) { 4634 const uint32_t symbol_stub_index = 4635 symbol_stub_index_offset + stub_idx; 4636 const lldb::addr_t symbol_stub_addr = 4637 m_mach_sections[sect_idx].addr + 4638 (stub_idx * symbol_stub_byte_size); 4639 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4640 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4641 symbol_stub_offset, 4)) { 4642 const uint32_t stub_sym_id = 4643 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4644 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4645 continue; 4646 4647 NListIndexToSymbolIndexMap::const_iterator index_pos = 4648 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4649 Symbol *stub_symbol = nullptr; 4650 if (index_pos != end_index_pos) { 4651 // We have a remapping from the original nlist index to a 4652 // current symbol index, so just look this up by index 4653 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4654 } else { 4655 // We need to lookup a symbol using the original nlist symbol 4656 // index since this index is coming from the S_SYMBOL_STUBS 4657 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4658 } 4659 4660 if (stub_symbol) { 4661 Address so_addr(symbol_stub_addr, section_list); 4662 4663 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4664 // Change the external symbol into a trampoline that makes 4665 // sense These symbols were N_UNDF N_EXT, and are useless 4666 // to us, so we can re-use them so we don't have to make up 4667 // a synthetic symbol for no good reason. 4668 if (resolver_addresses.find(symbol_stub_addr) == 4669 resolver_addresses.end()) 4670 stub_symbol->SetType(eSymbolTypeTrampoline); 4671 else 4672 stub_symbol->SetType(eSymbolTypeResolver); 4673 stub_symbol->SetExternal(false); 4674 stub_symbol->GetAddressRef() = so_addr; 4675 stub_symbol->SetByteSize(symbol_stub_byte_size); 4676 } else { 4677 // Make a synthetic symbol to describe the trampoline stub 4678 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4679 if (sym_idx >= num_syms) { 4680 sym = symtab->Resize(++num_syms); 4681 stub_symbol = nullptr; // this pointer no longer valid 4682 } 4683 sym[sym_idx].SetID(synthetic_sym_id++); 4684 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4685 if (resolver_addresses.find(symbol_stub_addr) == 4686 resolver_addresses.end()) 4687 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4688 else 4689 sym[sym_idx].SetType(eSymbolTypeResolver); 4690 sym[sym_idx].SetIsSynthetic(true); 4691 sym[sym_idx].GetAddressRef() = so_addr; 4692 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4693 ++sym_idx; 4694 } 4695 } else { 4696 if (log) 4697 log->Warning("symbol stub referencing symbol table symbol " 4698 "%u that isn't in our minimal symbol table, " 4699 "fix this!!!", 4700 stub_sym_id); 4701 } 4702 } 4703 } 4704 } 4705 } 4706 } 4707 } 4708 4709 if (!trie_entries.empty()) { 4710 for (const auto &e : trie_entries) { 4711 if (e.entry.import_name) { 4712 // Only add indirect symbols from the Trie entries if we didn't have 4713 // a N_INDR nlist entry for this already 4714 if (indirect_symbol_names.find(e.entry.name) == 4715 indirect_symbol_names.end()) { 4716 // Make a synthetic symbol to describe re-exported symbol. 4717 if (sym_idx >= num_syms) 4718 sym = symtab->Resize(++num_syms); 4719 sym[sym_idx].SetID(synthetic_sym_id++); 4720 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4721 sym[sym_idx].SetType(eSymbolTypeReExported); 4722 sym[sym_idx].SetIsSynthetic(true); 4723 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4724 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4725 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4726 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4727 } 4728 ++sym_idx; 4729 } 4730 } 4731 } 4732 } 4733 4734 // StreamFile s(stdout, false); 4735 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4736 // symtab->Dump(&s, NULL, eSortOrderNone); 4737 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4738 // sizes 4739 symtab->CalculateSymbolSizes(); 4740 4741 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4742 // symtab->Dump(&s, NULL, eSortOrderNone); 4743 4744 return symtab->GetNumSymbols(); 4745 } 4746 return 0; 4747 } 4748 4749 void ObjectFileMachO::Dump(Stream *s) { 4750 ModuleSP module_sp(GetModule()); 4751 if (module_sp) { 4752 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4753 s->Printf("%p: ", static_cast<void *>(this)); 4754 s->Indent(); 4755 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4756 s->PutCString("ObjectFileMachO64"); 4757 else 4758 s->PutCString("ObjectFileMachO32"); 4759 4760 *s << ", file = '" << m_file; 4761 ModuleSpecList all_specs; 4762 ModuleSpec base_spec; 4763 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 4764 base_spec, all_specs); 4765 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 4766 *s << "', triple"; 4767 if (e) 4768 s->Printf("[%d]", i); 4769 *s << " = "; 4770 *s << all_specs.GetModuleSpecRefAtIndex(i) 4771 .GetArchitecture() 4772 .GetTriple() 4773 .getTriple(); 4774 } 4775 *s << "\n"; 4776 SectionList *sections = GetSectionList(); 4777 if (sections) 4778 sections->Dump(s, nullptr, true, UINT32_MAX); 4779 4780 if (m_symtab_up) 4781 m_symtab_up->Dump(s, nullptr, eSortOrderNone); 4782 } 4783 } 4784 4785 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4786 const lldb_private::DataExtractor &data, 4787 lldb::offset_t lc_offset) { 4788 uint32_t i; 4789 struct uuid_command load_cmd; 4790 4791 lldb::offset_t offset = lc_offset; 4792 for (i = 0; i < header.ncmds; ++i) { 4793 const lldb::offset_t cmd_offset = offset; 4794 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4795 break; 4796 4797 if (load_cmd.cmd == LC_UUID) { 4798 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4799 4800 if (uuid_bytes) { 4801 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4802 // We pretend these object files have no UUID to prevent crashing. 4803 4804 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4805 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4806 0xbb, 0x14, 0xf0, 0x0d}; 4807 4808 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4809 return UUID(); 4810 4811 return UUID::fromOptionalData(uuid_bytes, 16); 4812 } 4813 return UUID(); 4814 } 4815 offset = cmd_offset + load_cmd.cmdsize; 4816 } 4817 return UUID(); 4818 } 4819 4820 static llvm::StringRef GetOSName(uint32_t cmd) { 4821 switch (cmd) { 4822 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4823 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4824 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4825 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4826 case llvm::MachO::LC_VERSION_MIN_TVOS: 4827 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4828 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4829 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4830 default: 4831 llvm_unreachable("unexpected LC_VERSION load command"); 4832 } 4833 } 4834 4835 namespace { 4836 struct OSEnv { 4837 llvm::StringRef os_type; 4838 llvm::StringRef environment; 4839 OSEnv(uint32_t cmd) { 4840 switch (cmd) { 4841 case llvm::MachO::PLATFORM_MACOS: 4842 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4843 return; 4844 case llvm::MachO::PLATFORM_IOS: 4845 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4846 return; 4847 case llvm::MachO::PLATFORM_TVOS: 4848 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4849 return; 4850 case llvm::MachO::PLATFORM_WATCHOS: 4851 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4852 return; 4853 // NEED_BRIDGEOS_TRIPLE case llvm::MachO::PLATFORM_BRIDGEOS: 4854 // NEED_BRIDGEOS_TRIPLE os_type = llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4855 // NEED_BRIDGEOS_TRIPLE return; 4856 case llvm::MachO::PLATFORM_MACCATALYST: 4857 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4858 environment = 4859 llvm::Triple::getEnvironmentTypeName(llvm::Triple::MacABI); 4860 return; 4861 case llvm::MachO::PLATFORM_IOSSIMULATOR: 4862 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4863 environment = 4864 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4865 return; 4866 case llvm::MachO::PLATFORM_TVOSSIMULATOR: 4867 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4868 environment = 4869 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4870 return; 4871 case llvm::MachO::PLATFORM_WATCHOSSIMULATOR: 4872 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4873 environment = 4874 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4875 return; 4876 default: { 4877 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 4878 LIBLLDB_LOG_PROCESS)); 4879 LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION"); 4880 } 4881 } 4882 } 4883 }; 4884 4885 struct MinOS { 4886 uint32_t major_version, minor_version, patch_version; 4887 MinOS(uint32_t version) 4888 : major_version(version >> 16), 4889 minor_version((version >> 8) & 0xffu), 4890 patch_version(version & 0xffu) {} 4891 }; 4892 } // namespace 4893 4894 void ObjectFileMachO::GetAllArchSpecs(const llvm::MachO::mach_header &header, 4895 const lldb_private::DataExtractor &data, 4896 lldb::offset_t lc_offset, 4897 ModuleSpec &base_spec, 4898 lldb_private::ModuleSpecList &all_specs) { 4899 auto &base_arch = base_spec.GetArchitecture(); 4900 base_arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4901 if (!base_arch.IsValid()) 4902 return; 4903 4904 bool found_any = false; 4905 auto add_triple = [&](const llvm::Triple &triple) { 4906 auto spec = base_spec; 4907 spec.GetArchitecture().GetTriple() = triple; 4908 if (spec.GetArchitecture().IsValid()) { 4909 spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset); 4910 all_specs.Append(spec); 4911 found_any = true; 4912 } 4913 }; 4914 4915 // Set OS to an unspecified unknown or a "*" so it can match any OS 4916 llvm::Triple base_triple = base_arch.GetTriple(); 4917 base_triple.setOS(llvm::Triple::UnknownOS); 4918 base_triple.setOSName(llvm::StringRef()); 4919 4920 if (header.filetype == MH_PRELOAD) { 4921 if (header.cputype == CPU_TYPE_ARM) { 4922 // If this is a 32-bit arm binary, and it's a standalone binary, force 4923 // the Vendor to Apple so we don't accidentally pick up the generic 4924 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4925 // frame pointer register; most other armv7 ABIs use a combination of 4926 // r7 and r11. 4927 base_triple.setVendor(llvm::Triple::Apple); 4928 } else { 4929 // Set vendor to an unspecified unknown or a "*" so it can match any 4930 // vendor This is required for correct behavior of EFI debugging on 4931 // x86_64 4932 base_triple.setVendor(llvm::Triple::UnknownVendor); 4933 base_triple.setVendorName(llvm::StringRef()); 4934 } 4935 return add_triple(base_triple); 4936 } 4937 4938 struct load_command load_cmd; 4939 4940 // See if there is an LC_VERSION_MIN_* load command that can give 4941 // us the OS type. 4942 lldb::offset_t offset = lc_offset; 4943 for (uint32_t i = 0; i < header.ncmds; ++i) { 4944 const lldb::offset_t cmd_offset = offset; 4945 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4946 break; 4947 4948 struct version_min_command version_min; 4949 switch (load_cmd.cmd) { 4950 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4951 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4952 case llvm::MachO::LC_VERSION_MIN_TVOS: 4953 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 4954 if (load_cmd.cmdsize != sizeof(version_min)) 4955 break; 4956 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 4957 data.GetByteOrder(), &version_min) == 0) 4958 break; 4959 MinOS min_os(version_min.version); 4960 llvm::SmallString<32> os_name; 4961 llvm::raw_svector_ostream os(os_name); 4962 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 4963 << min_os.minor_version << '.' << min_os.patch_version; 4964 4965 auto triple = base_triple; 4966 triple.setOSName(os.str()); 4967 os_name.clear(); 4968 add_triple(triple); 4969 break; 4970 } 4971 default: 4972 break; 4973 } 4974 4975 offset = cmd_offset + load_cmd.cmdsize; 4976 } 4977 4978 // See if there are LC_BUILD_VERSION load commands that can give 4979 // us the OS type. 4980 offset = lc_offset; 4981 for (uint32_t i = 0; i < header.ncmds; ++i) { 4982 const lldb::offset_t cmd_offset = offset; 4983 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4984 break; 4985 4986 do { 4987 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 4988 struct build_version_command build_version; 4989 if (load_cmd.cmdsize < sizeof(build_version)) { 4990 // Malformed load command. 4991 break; 4992 } 4993 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 4994 data.GetByteOrder(), &build_version) == 0) 4995 break; 4996 MinOS min_os(build_version.minos); 4997 OSEnv os_env(build_version.platform); 4998 llvm::SmallString<16> os_name; 4999 llvm::raw_svector_ostream os(os_name); 5000 os << os_env.os_type << min_os.major_version << '.' 5001 << min_os.minor_version << '.' << min_os.patch_version; 5002 auto triple = base_triple; 5003 triple.setOSName(os.str()); 5004 os_name.clear(); 5005 if (!os_env.environment.empty()) 5006 triple.setEnvironmentName(os_env.environment); 5007 add_triple(triple); 5008 } 5009 } while (0); 5010 offset = cmd_offset + load_cmd.cmdsize; 5011 } 5012 5013 if (!found_any) { 5014 if (header.filetype == MH_KEXT_BUNDLE) { 5015 base_triple.setVendor(llvm::Triple::Apple); 5016 add_triple (base_triple); 5017 } else { 5018 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 5019 // so lets not say our Vendor is Apple, leave it as an unspecified 5020 // unknown. 5021 base_triple.setVendor(llvm::Triple::UnknownVendor); 5022 base_triple.setVendorName(llvm::StringRef()); 5023 add_triple(base_triple); 5024 } 5025 } 5026 } 5027 5028 ArchSpec ObjectFileMachO::GetArchitecture( 5029 ModuleSP module_sp, const llvm::MachO::mach_header &header, 5030 const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) { 5031 ModuleSpecList all_specs; 5032 ModuleSpec base_spec; 5033 GetAllArchSpecs(header, data, MachHeaderSizeFromMagic(header.magic), 5034 base_spec, all_specs); 5035 5036 // If the object file offers multiple alternative load commands, 5037 // pick the one that matches the module. 5038 if (module_sp) { 5039 const ArchSpec &module_arch = module_sp->GetArchitecture(); 5040 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 5041 ArchSpec mach_arch = 5042 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 5043 if (module_arch.IsCompatibleMatch(mach_arch)) 5044 return mach_arch; 5045 } 5046 } 5047 5048 // Return the first arch we found. 5049 if (all_specs.GetSize() == 0) 5050 return {}; 5051 return all_specs.GetModuleSpecRefAtIndex(0).GetArchitecture(); 5052 } 5053 5054 UUID ObjectFileMachO::GetUUID() { 5055 ModuleSP module_sp(GetModule()); 5056 if (module_sp) { 5057 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5058 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5059 return GetUUID(m_header, m_data, offset); 5060 } 5061 return UUID(); 5062 } 5063 5064 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5065 uint32_t count = 0; 5066 ModuleSP module_sp(GetModule()); 5067 if (module_sp) { 5068 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5069 struct load_command load_cmd; 5070 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5071 std::vector<std::string> rpath_paths; 5072 std::vector<std::string> rpath_relative_paths; 5073 std::vector<std::string> at_exec_relative_paths; 5074 uint32_t i; 5075 for (i = 0; i < m_header.ncmds; ++i) { 5076 const uint32_t cmd_offset = offset; 5077 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5078 break; 5079 5080 switch (load_cmd.cmd) { 5081 case LC_RPATH: 5082 case LC_LOAD_DYLIB: 5083 case LC_LOAD_WEAK_DYLIB: 5084 case LC_REEXPORT_DYLIB: 5085 case LC_LOAD_DYLINKER: 5086 case LC_LOADFVMLIB: 5087 case LC_LOAD_UPWARD_DYLIB: { 5088 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5089 const char *path = m_data.PeekCStr(name_offset); 5090 if (path) { 5091 if (load_cmd.cmd == LC_RPATH) 5092 rpath_paths.push_back(path); 5093 else { 5094 if (path[0] == '@') { 5095 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5096 rpath_relative_paths.push_back(path + strlen("@rpath")); 5097 else if (strncmp(path, "@executable_path", 5098 strlen("@executable_path")) == 0) 5099 at_exec_relative_paths.push_back(path 5100 + strlen("@executable_path")); 5101 } else { 5102 FileSpec file_spec(path); 5103 if (files.AppendIfUnique(file_spec)) 5104 count++; 5105 } 5106 } 5107 } 5108 } break; 5109 5110 default: 5111 break; 5112 } 5113 offset = cmd_offset + load_cmd.cmdsize; 5114 } 5115 5116 FileSpec this_file_spec(m_file); 5117 FileSystem::Instance().Resolve(this_file_spec); 5118 5119 if (!rpath_paths.empty()) { 5120 // Fixup all LC_RPATH values to be absolute paths 5121 std::string loader_path("@loader_path"); 5122 std::string executable_path("@executable_path"); 5123 for (auto &rpath : rpath_paths) { 5124 if (rpath.find(loader_path) == 0) { 5125 rpath.erase(0, loader_path.size()); 5126 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5127 } else if (rpath.find(executable_path) == 0) { 5128 rpath.erase(0, executable_path.size()); 5129 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5130 } 5131 } 5132 5133 for (const auto &rpath_relative_path : rpath_relative_paths) { 5134 for (const auto &rpath : rpath_paths) { 5135 std::string path = rpath; 5136 path += rpath_relative_path; 5137 // It is OK to resolve this path because we must find a file on disk 5138 // for us to accept it anyway if it is rpath relative. 5139 FileSpec file_spec(path); 5140 FileSystem::Instance().Resolve(file_spec); 5141 if (FileSystem::Instance().Exists(file_spec) && 5142 files.AppendIfUnique(file_spec)) { 5143 count++; 5144 break; 5145 } 5146 } 5147 } 5148 } 5149 5150 // We may have @executable_paths but no RPATHS. Figure those out here. 5151 // Only do this if this object file is the executable. We have no way to 5152 // get back to the actual executable otherwise, so we won't get the right 5153 // path. 5154 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5155 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5156 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5157 FileSpec file_spec = 5158 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5159 if (FileSystem::Instance().Exists(file_spec) && 5160 files.AppendIfUnique(file_spec)) 5161 count++; 5162 } 5163 } 5164 } 5165 return count; 5166 } 5167 5168 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5169 // If the object file is not an executable it can't hold the entry point. 5170 // m_entry_point_address is initialized to an invalid address, so we can just 5171 // return that. If m_entry_point_address is valid it means we've found it 5172 // already, so return the cached value. 5173 5174 if ((!IsExecutable() && !IsDynamicLoader()) || 5175 m_entry_point_address.IsValid()) { 5176 return m_entry_point_address; 5177 } 5178 5179 // Otherwise, look for the UnixThread or Thread command. The data for the 5180 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5181 // 5182 // uint32_t flavor - this is the flavor argument you would pass to 5183 // thread_get_state 5184 // uint32_t count - this is the count of longs in the thread state data 5185 // struct XXX_thread_state state - this is the structure from 5186 // <machine/thread_status.h> corresponding to the flavor. 5187 // <repeat this trio> 5188 // 5189 // So we just keep reading the various register flavors till we find the GPR 5190 // one, then read the PC out of there. 5191 // FIXME: We will need to have a "RegisterContext data provider" class at some 5192 // point that can get all the registers 5193 // out of data in this form & attach them to a given thread. That should 5194 // underlie the MacOS X User process plugin, and we'll also need it for the 5195 // MacOS X Core File process plugin. When we have that we can also use it 5196 // here. 5197 // 5198 // For now we hard-code the offsets and flavors we need: 5199 // 5200 // 5201 5202 ModuleSP module_sp(GetModule()); 5203 if (module_sp) { 5204 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5205 struct load_command load_cmd; 5206 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5207 uint32_t i; 5208 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5209 bool done = false; 5210 5211 for (i = 0; i < m_header.ncmds; ++i) { 5212 const lldb::offset_t cmd_offset = offset; 5213 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5214 break; 5215 5216 switch (load_cmd.cmd) { 5217 case LC_UNIXTHREAD: 5218 case LC_THREAD: { 5219 while (offset < cmd_offset + load_cmd.cmdsize) { 5220 uint32_t flavor = m_data.GetU32(&offset); 5221 uint32_t count = m_data.GetU32(&offset); 5222 if (count == 0) { 5223 // We've gotten off somehow, log and exit; 5224 return m_entry_point_address; 5225 } 5226 5227 switch (m_header.cputype) { 5228 case llvm::MachO::CPU_TYPE_ARM: 5229 if (flavor == 1 || 5230 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 from 5231 // mach/arm/thread_status.h 5232 { 5233 offset += 60; // This is the offset of pc in the GPR thread state 5234 // data structure. 5235 start_address = m_data.GetU32(&offset); 5236 done = true; 5237 } 5238 break; 5239 case llvm::MachO::CPU_TYPE_ARM64: 5240 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5241 { 5242 offset += 256; // This is the offset of pc in the GPR thread state 5243 // data structure. 5244 start_address = m_data.GetU64(&offset); 5245 done = true; 5246 } 5247 break; 5248 case llvm::MachO::CPU_TYPE_I386: 5249 if (flavor == 5250 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5251 { 5252 offset += 40; // This is the offset of eip in the GPR thread state 5253 // data structure. 5254 start_address = m_data.GetU32(&offset); 5255 done = true; 5256 } 5257 break; 5258 case llvm::MachO::CPU_TYPE_X86_64: 5259 if (flavor == 5260 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5261 { 5262 offset += 16 * 8; // This is the offset of rip in the GPR thread 5263 // state data structure. 5264 start_address = m_data.GetU64(&offset); 5265 done = true; 5266 } 5267 break; 5268 default: 5269 return m_entry_point_address; 5270 } 5271 // Haven't found the GPR flavor yet, skip over the data for this 5272 // flavor: 5273 if (done) 5274 break; 5275 offset += count * 4; 5276 } 5277 } break; 5278 case LC_MAIN: { 5279 ConstString text_segment_name("__TEXT"); 5280 uint64_t entryoffset = m_data.GetU64(&offset); 5281 SectionSP text_segment_sp = 5282 GetSectionList()->FindSectionByName(text_segment_name); 5283 if (text_segment_sp) { 5284 done = true; 5285 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5286 } 5287 } break; 5288 5289 default: 5290 break; 5291 } 5292 if (done) 5293 break; 5294 5295 // Go to the next load command: 5296 offset = cmd_offset + load_cmd.cmdsize; 5297 } 5298 5299 if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) { 5300 if (GetSymtab()) { 5301 Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType( 5302 ConstString("_dyld_start"), SymbolType::eSymbolTypeCode, 5303 Symtab::eDebugAny, Symtab::eVisibilityAny); 5304 if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) { 5305 start_address = dyld_start_sym->GetAddress().GetFileAddress(); 5306 } 5307 } 5308 } 5309 5310 if (start_address != LLDB_INVALID_ADDRESS) { 5311 // We got the start address from the load commands, so now resolve that 5312 // address in the sections of this ObjectFile: 5313 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5314 start_address, GetSectionList())) { 5315 m_entry_point_address.Clear(); 5316 } 5317 } else { 5318 // We couldn't read the UnixThread load command - maybe it wasn't there. 5319 // As a fallback look for the "start" symbol in the main executable. 5320 5321 ModuleSP module_sp(GetModule()); 5322 5323 if (module_sp) { 5324 SymbolContextList contexts; 5325 SymbolContext context; 5326 if (module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5327 eSymbolTypeCode, contexts)) { 5328 if (contexts.GetContextAtIndex(0, context)) 5329 m_entry_point_address = context.symbol->GetAddress(); 5330 } 5331 } 5332 } 5333 } 5334 5335 return m_entry_point_address; 5336 } 5337 5338 lldb_private::Address ObjectFileMachO::GetBaseAddress() { 5339 lldb_private::Address header_addr; 5340 SectionList *section_list = GetSectionList(); 5341 if (section_list) { 5342 SectionSP text_segment_sp( 5343 section_list->FindSectionByName(GetSegmentNameTEXT())); 5344 if (text_segment_sp) { 5345 header_addr.SetSection(text_segment_sp); 5346 header_addr.SetOffset(0); 5347 } 5348 } 5349 return header_addr; 5350 } 5351 5352 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5353 ModuleSP module_sp(GetModule()); 5354 if (module_sp) { 5355 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5356 if (!m_thread_context_offsets_valid) { 5357 m_thread_context_offsets_valid = true; 5358 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5359 FileRangeArray::Entry file_range; 5360 thread_command thread_cmd; 5361 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5362 const uint32_t cmd_offset = offset; 5363 if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr) 5364 break; 5365 5366 if (thread_cmd.cmd == LC_THREAD) { 5367 file_range.SetRangeBase(offset); 5368 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5369 m_thread_context_offsets.Append(file_range); 5370 } 5371 offset = cmd_offset + thread_cmd.cmdsize; 5372 } 5373 } 5374 } 5375 return m_thread_context_offsets.GetSize(); 5376 } 5377 5378 std::string ObjectFileMachO::GetIdentifierString() { 5379 std::string result; 5380 ModuleSP module_sp(GetModule()); 5381 if (module_sp) { 5382 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5383 5384 // First, look over the load commands for an LC_NOTE load command with 5385 // data_owner string "kern ver str" & use that if found. 5386 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5387 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5388 const uint32_t cmd_offset = offset; 5389 load_command lc; 5390 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5391 break; 5392 if (lc.cmd == LC_NOTE) 5393 { 5394 char data_owner[17]; 5395 m_data.CopyData (offset, 16, data_owner); 5396 data_owner[16] = '\0'; 5397 offset += 16; 5398 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5399 uint64_t size = m_data.GetU64_unchecked (&offset); 5400 5401 // "kern ver str" has a uint32_t version and then a nul terminated 5402 // c-string. 5403 if (strcmp ("kern ver str", data_owner) == 0) 5404 { 5405 offset = fileoff; 5406 uint32_t version; 5407 if (m_data.GetU32 (&offset, &version, 1) != nullptr) 5408 { 5409 if (version == 1) 5410 { 5411 uint32_t strsize = size - sizeof (uint32_t); 5412 char *buf = (char*) malloc (strsize); 5413 if (buf) 5414 { 5415 m_data.CopyData (offset, strsize, buf); 5416 buf[strsize - 1] = '\0'; 5417 result = buf; 5418 if (buf) 5419 free (buf); 5420 return result; 5421 } 5422 } 5423 } 5424 } 5425 } 5426 offset = cmd_offset + lc.cmdsize; 5427 } 5428 5429 // Second, make a pass over the load commands looking for an obsolete 5430 // LC_IDENT load command. 5431 offset = MachHeaderSizeFromMagic(m_header.magic); 5432 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5433 const uint32_t cmd_offset = offset; 5434 struct ident_command ident_command; 5435 if (m_data.GetU32(&offset, &ident_command, 2) == nullptr) 5436 break; 5437 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5438 char *buf = (char *) malloc (ident_command.cmdsize); 5439 if (buf != nullptr 5440 && m_data.CopyData (offset, ident_command.cmdsize, buf) == ident_command.cmdsize) { 5441 buf[ident_command.cmdsize - 1] = '\0'; 5442 result = buf; 5443 } 5444 if (buf) 5445 free (buf); 5446 } 5447 offset = cmd_offset + ident_command.cmdsize; 5448 } 5449 5450 } 5451 return result; 5452 } 5453 5454 bool ObjectFileMachO::GetCorefileMainBinaryInfo (addr_t &address, UUID &uuid) { 5455 address = LLDB_INVALID_ADDRESS; 5456 uuid.Clear(); 5457 ModuleSP module_sp(GetModule()); 5458 if (module_sp) { 5459 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5460 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5461 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5462 const uint32_t cmd_offset = offset; 5463 load_command lc; 5464 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5465 break; 5466 if (lc.cmd == LC_NOTE) 5467 { 5468 char data_owner[17]; 5469 memset (data_owner, 0, sizeof (data_owner)); 5470 m_data.CopyData (offset, 16, data_owner); 5471 offset += 16; 5472 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5473 uint64_t size = m_data.GetU64_unchecked (&offset); 5474 5475 // "main bin spec" (main binary specification) data payload is 5476 // formatted: 5477 // uint32_t version [currently 1] 5478 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user process] 5479 // uint64_t address [ UINT64_MAX if address not specified ] 5480 // uuid_t uuid [ all zero's if uuid not specified ] 5481 // uint32_t log2_pagesize [ process page size in log base 2, e.g. 4k pages are 12. 0 for unspecified ] 5482 5483 if (strcmp ("main bin spec", data_owner) == 0 && size >= 32) 5484 { 5485 offset = fileoff; 5486 uint32_t version; 5487 if (m_data.GetU32 (&offset, &version, 1) != nullptr && version == 1) 5488 { 5489 uint32_t type = 0; 5490 uuid_t raw_uuid; 5491 memset (raw_uuid, 0, sizeof (uuid_t)); 5492 5493 if (m_data.GetU32(&offset, &type, 1) && 5494 m_data.GetU64(&offset, &address, 1) && 5495 m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5496 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5497 return true; 5498 } 5499 } 5500 } 5501 } 5502 offset = cmd_offset + lc.cmdsize; 5503 } 5504 } 5505 return false; 5506 } 5507 5508 lldb::RegisterContextSP 5509 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5510 lldb_private::Thread &thread) { 5511 lldb::RegisterContextSP reg_ctx_sp; 5512 5513 ModuleSP module_sp(GetModule()); 5514 if (module_sp) { 5515 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5516 if (!m_thread_context_offsets_valid) 5517 GetNumThreadContexts(); 5518 5519 const FileRangeArray::Entry *thread_context_file_range = 5520 m_thread_context_offsets.GetEntryAtIndex(idx); 5521 if (thread_context_file_range) { 5522 5523 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5524 thread_context_file_range->GetByteSize()); 5525 5526 switch (m_header.cputype) { 5527 case llvm::MachO::CPU_TYPE_ARM64: 5528 reg_ctx_sp = 5529 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data); 5530 break; 5531 5532 case llvm::MachO::CPU_TYPE_ARM: 5533 reg_ctx_sp = 5534 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data); 5535 break; 5536 5537 case llvm::MachO::CPU_TYPE_I386: 5538 reg_ctx_sp = 5539 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data); 5540 break; 5541 5542 case llvm::MachO::CPU_TYPE_X86_64: 5543 reg_ctx_sp = 5544 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data); 5545 break; 5546 } 5547 } 5548 } 5549 return reg_ctx_sp; 5550 } 5551 5552 ObjectFile::Type ObjectFileMachO::CalculateType() { 5553 switch (m_header.filetype) { 5554 case MH_OBJECT: // 0x1u 5555 if (GetAddressByteSize() == 4) { 5556 // 32 bit kexts are just object files, but they do have a valid 5557 // UUID load command. 5558 if (GetUUID()) { 5559 // this checking for the UUID load command is not enough we could 5560 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5561 // this is required of kexts 5562 if (m_strata == eStrataInvalid) 5563 m_strata = eStrataKernel; 5564 return eTypeSharedLibrary; 5565 } 5566 } 5567 return eTypeObjectFile; 5568 5569 case MH_EXECUTE: 5570 return eTypeExecutable; // 0x2u 5571 case MH_FVMLIB: 5572 return eTypeSharedLibrary; // 0x3u 5573 case MH_CORE: 5574 return eTypeCoreFile; // 0x4u 5575 case MH_PRELOAD: 5576 return eTypeSharedLibrary; // 0x5u 5577 case MH_DYLIB: 5578 return eTypeSharedLibrary; // 0x6u 5579 case MH_DYLINKER: 5580 return eTypeDynamicLinker; // 0x7u 5581 case MH_BUNDLE: 5582 return eTypeSharedLibrary; // 0x8u 5583 case MH_DYLIB_STUB: 5584 return eTypeStubLibrary; // 0x9u 5585 case MH_DSYM: 5586 return eTypeDebugInfo; // 0xAu 5587 case MH_KEXT_BUNDLE: 5588 return eTypeSharedLibrary; // 0xBu 5589 default: 5590 break; 5591 } 5592 return eTypeUnknown; 5593 } 5594 5595 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5596 switch (m_header.filetype) { 5597 case MH_OBJECT: // 0x1u 5598 { 5599 // 32 bit kexts are just object files, but they do have a valid 5600 // UUID load command. 5601 if (GetUUID()) { 5602 // this checking for the UUID load command is not enough we could 5603 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5604 // this is required of kexts 5605 if (m_type == eTypeInvalid) 5606 m_type = eTypeSharedLibrary; 5607 5608 return eStrataKernel; 5609 } 5610 } 5611 return eStrataUnknown; 5612 5613 case MH_EXECUTE: // 0x2u 5614 // Check for the MH_DYLDLINK bit in the flags 5615 if (m_header.flags & MH_DYLDLINK) { 5616 return eStrataUser; 5617 } else { 5618 SectionList *section_list = GetSectionList(); 5619 if (section_list) { 5620 static ConstString g_kld_section_name("__KLD"); 5621 if (section_list->FindSectionByName(g_kld_section_name)) 5622 return eStrataKernel; 5623 } 5624 } 5625 return eStrataRawImage; 5626 5627 case MH_FVMLIB: 5628 return eStrataUser; // 0x3u 5629 case MH_CORE: 5630 return eStrataUnknown; // 0x4u 5631 case MH_PRELOAD: 5632 return eStrataRawImage; // 0x5u 5633 case MH_DYLIB: 5634 return eStrataUser; // 0x6u 5635 case MH_DYLINKER: 5636 return eStrataUser; // 0x7u 5637 case MH_BUNDLE: 5638 return eStrataUser; // 0x8u 5639 case MH_DYLIB_STUB: 5640 return eStrataUser; // 0x9u 5641 case MH_DSYM: 5642 return eStrataUnknown; // 0xAu 5643 case MH_KEXT_BUNDLE: 5644 return eStrataKernel; // 0xBu 5645 default: 5646 break; 5647 } 5648 return eStrataUnknown; 5649 } 5650 5651 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5652 ModuleSP module_sp(GetModule()); 5653 if (module_sp) { 5654 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5655 struct dylib_command load_cmd; 5656 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5657 uint32_t version_cmd = 0; 5658 uint64_t version = 0; 5659 uint32_t i; 5660 for (i = 0; i < m_header.ncmds; ++i) { 5661 const lldb::offset_t cmd_offset = offset; 5662 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5663 break; 5664 5665 if (load_cmd.cmd == LC_ID_DYLIB) { 5666 if (version_cmd == 0) { 5667 version_cmd = load_cmd.cmd; 5668 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr) 5669 break; 5670 version = load_cmd.dylib.current_version; 5671 } 5672 break; // Break for now unless there is another more complete version 5673 // number load command in the future. 5674 } 5675 offset = cmd_offset + load_cmd.cmdsize; 5676 } 5677 5678 if (version_cmd == LC_ID_DYLIB) { 5679 unsigned major = (version & 0xFFFF0000ull) >> 16; 5680 unsigned minor = (version & 0x0000FF00ull) >> 8; 5681 unsigned subminor = (version & 0x000000FFull); 5682 return llvm::VersionTuple(major, minor, subminor); 5683 } 5684 } 5685 return llvm::VersionTuple(); 5686 } 5687 5688 ArchSpec ObjectFileMachO::GetArchitecture() { 5689 ModuleSP module_sp(GetModule()); 5690 ArchSpec arch; 5691 if (module_sp) { 5692 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5693 5694 return GetArchitecture(module_sp, m_header, m_data, 5695 MachHeaderSizeFromMagic(m_header.magic)); 5696 } 5697 return arch; 5698 } 5699 5700 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, addr_t &base_addr, UUID &uuid) { 5701 uuid.Clear(); 5702 base_addr = LLDB_INVALID_ADDRESS; 5703 if (process && process->GetDynamicLoader()) { 5704 DynamicLoader *dl = process->GetDynamicLoader(); 5705 LazyBool using_shared_cache; 5706 LazyBool private_shared_cache; 5707 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5708 private_shared_cache); 5709 } 5710 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5711 LLDB_LOGF( 5712 log, 5713 "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64, 5714 uuid.GetAsString().c_str(), base_addr); 5715 } 5716 5717 // From dyld SPI header dyld_process_info.h 5718 typedef void *dyld_process_info; 5719 struct lldb_copy__dyld_process_cache_info { 5720 uuid_t cacheUUID; // UUID of cache used by process 5721 uint64_t cacheBaseAddress; // load address of dyld shared cache 5722 bool noCache; // process is running without a dyld cache 5723 bool privateCache; // process is using a private copy of its dyld cache 5724 }; 5725 5726 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with llvm 5727 // enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile errors. 5728 // So we need to use the actual underlying types of task_t and kern_return_t 5729 // below. 5730 extern "C" unsigned int /*task_t*/ mach_task_self(); 5731 5732 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5733 uuid.Clear(); 5734 base_addr = LLDB_INVALID_ADDRESS; 5735 5736 #if defined(__APPLE__) && \ 5737 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5738 uint8_t *(*dyld_get_all_image_infos)(void); 5739 dyld_get_all_image_infos = 5740 (uint8_t * (*)())dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5741 if (dyld_get_all_image_infos) { 5742 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5743 if (dyld_all_image_infos_address) { 5744 uint32_t *version = (uint32_t *) 5745 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5746 if (*version >= 13) { 5747 uuid_t *sharedCacheUUID_address = 0; 5748 int wordsize = sizeof(uint8_t *); 5749 if (wordsize == 8) { 5750 sharedCacheUUID_address = 5751 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5752 160); // sharedCacheUUID <mach-o/dyld_images.h> 5753 if (*version >= 15) 5754 base_addr = *(uint64_t *) ((uint8_t *) dyld_all_image_infos_address 5755 + 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5756 } else { 5757 sharedCacheUUID_address = 5758 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5759 84); // sharedCacheUUID <mach-o/dyld_images.h> 5760 if (*version >= 15) { 5761 base_addr = 0; 5762 base_addr = *(uint32_t *) ((uint8_t *) dyld_all_image_infos_address 5763 + 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5764 } 5765 } 5766 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5767 } 5768 } 5769 } else { 5770 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5771 dyld_process_info (*dyld_process_info_create)(unsigned int /* task_t */ task, uint64_t timestamp, unsigned int /*kern_return_t*/ *kernelError); 5772 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5773 void (*dyld_process_info_release)(dyld_process_info info); 5774 5775 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, unsigned int /*kern_return_t*/ *)) 5776 dlsym (RTLD_DEFAULT, "_dyld_process_info_create"); 5777 dyld_process_info_get_cache = (void (*)(void *, void *)) 5778 dlsym (RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5779 dyld_process_info_release = (void (*)(void *)) 5780 dlsym (RTLD_DEFAULT, "_dyld_process_info_release"); 5781 5782 if (dyld_process_info_create && dyld_process_info_get_cache) { 5783 unsigned int /*kern_return_t */ kern_ret; 5784 dyld_process_info process_info = dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5785 if (process_info) { 5786 struct lldb_copy__dyld_process_cache_info sc_info; 5787 memset (&sc_info, 0, sizeof (struct lldb_copy__dyld_process_cache_info)); 5788 dyld_process_info_get_cache (process_info, &sc_info); 5789 if (sc_info.cacheBaseAddress != 0) { 5790 base_addr = sc_info.cacheBaseAddress; 5791 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 5792 } 5793 dyld_process_info_release (process_info); 5794 } 5795 } 5796 } 5797 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5798 if (log && uuid.IsValid()) 5799 LLDB_LOGF(log, 5800 "lldb's in-memory shared cache has a UUID of %s base address of " 5801 "0x%" PRIx64, 5802 uuid.GetAsString().c_str(), base_addr); 5803 #endif 5804 } 5805 5806 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5807 if (!m_min_os_version) { 5808 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5809 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5810 const lldb::offset_t load_cmd_offset = offset; 5811 5812 version_min_command lc; 5813 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5814 break; 5815 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5816 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5817 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5818 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5819 if (m_data.GetU32(&offset, &lc.version, 5820 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5821 const uint32_t xxxx = lc.version >> 16; 5822 const uint32_t yy = (lc.version >> 8) & 0xffu; 5823 const uint32_t zz = lc.version & 0xffu; 5824 if (xxxx) { 5825 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5826 break; 5827 } 5828 } 5829 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5830 // struct build_version_command { 5831 // uint32_t cmd; /* LC_BUILD_VERSION */ 5832 // uint32_t cmdsize; /* sizeof(struct build_version_command) plus */ 5833 // /* ntools * sizeof(struct build_tool_version) */ 5834 // uint32_t platform; /* platform */ 5835 // uint32_t minos; /* X.Y.Z is encoded in nibbles xxxx.yy.zz */ 5836 // uint32_t sdk; /* X.Y.Z is encoded in nibbles xxxx.yy.zz */ 5837 // uint32_t ntools; /* number of tool entries following this */ 5838 // }; 5839 5840 offset += 4; // skip platform 5841 uint32_t minos = m_data.GetU32(&offset); 5842 5843 const uint32_t xxxx = minos >> 16; 5844 const uint32_t yy = (minos >> 8) & 0xffu; 5845 const uint32_t zz = minos & 0xffu; 5846 if (xxxx) { 5847 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5848 break; 5849 } 5850 } 5851 5852 offset = load_cmd_offset + lc.cmdsize; 5853 } 5854 5855 if (!m_min_os_version) { 5856 // Set version to an empty value so we don't keep trying to 5857 m_min_os_version = llvm::VersionTuple(); 5858 } 5859 } 5860 5861 return *m_min_os_version; 5862 } 5863 5864 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { 5865 if (!m_sdk_versions.hasValue()) { 5866 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5867 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5868 const lldb::offset_t load_cmd_offset = offset; 5869 5870 version_min_command lc; 5871 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5872 break; 5873 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5874 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5875 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5876 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5877 if (m_data.GetU32(&offset, &lc.version, 5878 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5879 const uint32_t xxxx = lc.sdk >> 16; 5880 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5881 const uint32_t zz = lc.sdk & 0xffu; 5882 if (xxxx) { 5883 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 5884 break; 5885 } else { 5886 GetModule()->ReportWarning( 5887 "minimum OS version load command with invalid (0) version found."); 5888 } 5889 } 5890 } 5891 offset = load_cmd_offset + lc.cmdsize; 5892 } 5893 5894 if (!m_sdk_versions.hasValue()) { 5895 offset = MachHeaderSizeFromMagic(m_header.magic); 5896 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5897 const lldb::offset_t load_cmd_offset = offset; 5898 5899 version_min_command lc; 5900 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5901 break; 5902 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5903 // struct build_version_command { 5904 // uint32_t cmd; /* LC_BUILD_VERSION */ 5905 // uint32_t cmdsize; /* sizeof(struct 5906 // build_version_command) plus */ 5907 // /* ntools * sizeof(struct 5908 // build_tool_version) */ 5909 // uint32_t platform; /* platform */ 5910 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5911 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded 5912 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number 5913 // of tool entries following this */ 5914 // }; 5915 5916 offset += 4; // skip platform 5917 uint32_t minos = m_data.GetU32(&offset); 5918 5919 const uint32_t xxxx = minos >> 16; 5920 const uint32_t yy = (minos >> 8) & 0xffu; 5921 const uint32_t zz = minos & 0xffu; 5922 if (xxxx) { 5923 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 5924 break; 5925 } 5926 } 5927 offset = load_cmd_offset + lc.cmdsize; 5928 } 5929 } 5930 5931 if (!m_sdk_versions.hasValue()) 5932 m_sdk_versions = llvm::VersionTuple(); 5933 } 5934 5935 return m_sdk_versions.getValue(); 5936 } 5937 5938 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5939 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5940 } 5941 5942 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5943 return m_allow_assembly_emulation_unwind_plans; 5944 } 5945 5946 // PluginInterface protocol 5947 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5948 return GetPluginNameStatic(); 5949 } 5950 5951 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5952 5953 Section *ObjectFileMachO::GetMachHeaderSection() { 5954 // Find the first address of the mach header which is the first non-zero file 5955 // sized section whose file offset is zero. This is the base file address of 5956 // the mach-o file which can be subtracted from the vmaddr of the other 5957 // segments found in memory and added to the load address 5958 ModuleSP module_sp = GetModule(); 5959 if (!module_sp) 5960 return nullptr; 5961 SectionList *section_list = GetSectionList(); 5962 if (!section_list) 5963 return nullptr; 5964 const size_t num_sections = section_list->GetSize(); 5965 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5966 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5967 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 5968 return section; 5969 } 5970 return nullptr; 5971 } 5972 5973 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 5974 if (!section) 5975 return false; 5976 const bool is_dsym = (m_header.filetype == MH_DSYM); 5977 if (section->GetFileSize() == 0 && !is_dsym) 5978 return false; 5979 if (section->IsThreadSpecific()) 5980 return false; 5981 if (GetModule().get() != section->GetModule().get()) 5982 return false; 5983 // Be careful with __LINKEDIT and __DWARF segments 5984 if (section->GetName() == GetSegmentNameLINKEDIT() || 5985 section->GetName() == GetSegmentNameDWARF()) { 5986 // Only map __LINKEDIT and __DWARF if we have an in memory image and 5987 // this isn't a kernel binary like a kext or mach_kernel. 5988 const bool is_memory_image = (bool)m_process_wp.lock(); 5989 const Strata strata = GetStrata(); 5990 if (is_memory_image == false || strata == eStrataKernel) 5991 return false; 5992 } 5993 return true; 5994 } 5995 5996 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 5997 lldb::addr_t header_load_address, const Section *header_section, 5998 const Section *section) { 5999 ModuleSP module_sp = GetModule(); 6000 if (module_sp && header_section && section && 6001 header_load_address != LLDB_INVALID_ADDRESS) { 6002 lldb::addr_t file_addr = header_section->GetFileAddress(); 6003 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 6004 return section->GetFileAddress() - file_addr + header_load_address; 6005 } 6006 return LLDB_INVALID_ADDRESS; 6007 } 6008 6009 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 6010 bool value_is_offset) { 6011 ModuleSP module_sp = GetModule(); 6012 if (module_sp) { 6013 size_t num_loaded_sections = 0; 6014 SectionList *section_list = GetSectionList(); 6015 if (section_list) { 6016 const size_t num_sections = section_list->GetSize(); 6017 6018 if (value_is_offset) { 6019 // "value" is an offset to apply to each top level segment 6020 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6021 // Iterate through the object file sections to find all of the 6022 // sections that size on disk (to avoid __PAGEZERO) and load them 6023 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6024 if (SectionIsLoadable(section_sp.get())) 6025 if (target.GetSectionLoadList().SetSectionLoadAddress( 6026 section_sp, section_sp->GetFileAddress() + value)) 6027 ++num_loaded_sections; 6028 } 6029 } else { 6030 // "value" is the new base address of the mach_header, adjust each 6031 // section accordingly 6032 6033 Section *mach_header_section = GetMachHeaderSection(); 6034 if (mach_header_section) { 6035 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6036 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6037 6038 lldb::addr_t section_load_addr = 6039 CalculateSectionLoadAddressForMemoryImage( 6040 value, mach_header_section, section_sp.get()); 6041 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6042 if (target.GetSectionLoadList().SetSectionLoadAddress( 6043 section_sp, section_load_addr)) 6044 ++num_loaded_sections; 6045 } 6046 } 6047 } 6048 } 6049 } 6050 return num_loaded_sections > 0; 6051 } 6052 return false; 6053 } 6054 6055 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6056 const FileSpec &outfile, Status &error) { 6057 if (process_sp) { 6058 Target &target = process_sp->GetTarget(); 6059 const ArchSpec target_arch = target.GetArchitecture(); 6060 const llvm::Triple &target_triple = target_arch.GetTriple(); 6061 if (target_triple.getVendor() == llvm::Triple::Apple && 6062 (target_triple.getOS() == llvm::Triple::MacOSX || 6063 target_triple.getOS() == llvm::Triple::IOS || 6064 target_triple.getOS() == llvm::Triple::WatchOS || 6065 target_triple.getOS() == llvm::Triple::TvOS)) { 6066 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) { 6067 bool make_core = false; 6068 switch (target_arch.GetMachine()) { 6069 case llvm::Triple::aarch64: 6070 case llvm::Triple::arm: 6071 case llvm::Triple::thumb: 6072 case llvm::Triple::x86: 6073 case llvm::Triple::x86_64: 6074 make_core = true; 6075 break; 6076 default: 6077 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6078 target_triple.str().c_str()); 6079 break; 6080 } 6081 6082 if (make_core) { 6083 std::vector<segment_command_64> segment_load_commands; 6084 // uint32_t range_info_idx = 0; 6085 MemoryRegionInfo range_info; 6086 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6087 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6088 const ByteOrder byte_order = target_arch.GetByteOrder(); 6089 if (range_error.Success()) { 6090 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6091 const addr_t addr = range_info.GetRange().GetRangeBase(); 6092 const addr_t size = range_info.GetRange().GetByteSize(); 6093 6094 if (size == 0) 6095 break; 6096 6097 // Calculate correct protections 6098 uint32_t prot = 0; 6099 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6100 prot |= VM_PROT_READ; 6101 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6102 prot |= VM_PROT_WRITE; 6103 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6104 prot |= VM_PROT_EXECUTE; 6105 6106 if (prot != 0) { 6107 uint32_t cmd_type = LC_SEGMENT_64; 6108 uint32_t segment_size = sizeof(segment_command_64); 6109 if (addr_byte_size == 4) { 6110 cmd_type = LC_SEGMENT; 6111 segment_size = sizeof(segment_command); 6112 } 6113 segment_command_64 segment = { 6114 cmd_type, // uint32_t cmd; 6115 segment_size, // uint32_t cmdsize; 6116 {0}, // char segname[16]; 6117 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6118 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6119 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6120 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6121 prot, // uint32_t maxprot; 6122 prot, // uint32_t initprot; 6123 0, // uint32_t nsects; 6124 0}; // uint32_t flags; 6125 segment_load_commands.push_back(segment); 6126 } else { 6127 // No protections and a size of 1 used to be returned from old 6128 // debugservers when we asked about a region that was past the 6129 // last memory region and it indicates the end... 6130 if (size == 1) 6131 break; 6132 } 6133 6134 range_error = process_sp->GetMemoryRegionInfo( 6135 range_info.GetRange().GetRangeEnd(), range_info); 6136 if (range_error.Fail()) 6137 break; 6138 } 6139 6140 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6141 6142 mach_header_64 mach_header; 6143 if (addr_byte_size == 8) { 6144 mach_header.magic = MH_MAGIC_64; 6145 } else { 6146 mach_header.magic = MH_MAGIC; 6147 } 6148 mach_header.cputype = target_arch.GetMachOCPUType(); 6149 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6150 mach_header.filetype = MH_CORE; 6151 mach_header.ncmds = segment_load_commands.size(); 6152 mach_header.flags = 0; 6153 mach_header.reserved = 0; 6154 ThreadList &thread_list = process_sp->GetThreadList(); 6155 const uint32_t num_threads = thread_list.GetSize(); 6156 6157 // Make an array of LC_THREAD data items. Each one contains the 6158 // contents of the LC_THREAD load command. The data doesn't contain 6159 // the load command + load command size, we will add the load command 6160 // and load command size as we emit the data. 6161 std::vector<StreamString> LC_THREAD_datas(num_threads); 6162 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6163 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6164 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6165 LC_THREAD_data.SetByteOrder(byte_order); 6166 } 6167 for (uint32_t thread_idx = 0; thread_idx < num_threads; 6168 ++thread_idx) { 6169 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6170 if (thread_sp) { 6171 switch (mach_header.cputype) { 6172 case llvm::MachO::CPU_TYPE_ARM64: 6173 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6174 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6175 break; 6176 6177 case llvm::MachO::CPU_TYPE_ARM: 6178 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6179 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6180 break; 6181 6182 case llvm::MachO::CPU_TYPE_I386: 6183 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6184 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6185 break; 6186 6187 case llvm::MachO::CPU_TYPE_X86_64: 6188 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6189 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6190 break; 6191 } 6192 } 6193 } 6194 6195 // The size of the load command is the size of the segments... 6196 if (addr_byte_size == 8) { 6197 mach_header.sizeofcmds = segment_load_commands.size() * 6198 sizeof(struct segment_command_64); 6199 } else { 6200 mach_header.sizeofcmds = 6201 segment_load_commands.size() * sizeof(struct segment_command); 6202 } 6203 6204 // and the size of all LC_THREAD load command 6205 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6206 ++mach_header.ncmds; 6207 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6208 } 6209 6210 // Write the mach header 6211 buffer.PutHex32(mach_header.magic); 6212 buffer.PutHex32(mach_header.cputype); 6213 buffer.PutHex32(mach_header.cpusubtype); 6214 buffer.PutHex32(mach_header.filetype); 6215 buffer.PutHex32(mach_header.ncmds); 6216 buffer.PutHex32(mach_header.sizeofcmds); 6217 buffer.PutHex32(mach_header.flags); 6218 if (addr_byte_size == 8) { 6219 buffer.PutHex32(mach_header.reserved); 6220 } 6221 6222 // Skip the mach header and all load commands and align to the next 6223 // 0x1000 byte boundary 6224 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6225 if (file_offset & 0x00000fff) { 6226 file_offset += 0x00001000ull; 6227 file_offset &= (~0x00001000ull + 1); 6228 } 6229 6230 for (auto &segment : segment_load_commands) { 6231 segment.fileoff = file_offset; 6232 file_offset += segment.filesize; 6233 } 6234 6235 // Write out all of the LC_THREAD load commands 6236 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6237 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6238 buffer.PutHex32(LC_THREAD); 6239 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6240 buffer.Write(LC_THREAD_data.GetString().data(), 6241 LC_THREAD_data_size); 6242 } 6243 6244 // Write out all of the segment load commands 6245 for (const auto &segment : segment_load_commands) { 6246 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6247 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6248 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6249 segment.cmd, segment.cmdsize, segment.vmaddr, 6250 segment.vmaddr + segment.vmsize, segment.fileoff, 6251 segment.filesize, segment.maxprot, segment.initprot, 6252 segment.nsects, segment.flags); 6253 6254 buffer.PutHex32(segment.cmd); 6255 buffer.PutHex32(segment.cmdsize); 6256 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6257 if (addr_byte_size == 8) { 6258 buffer.PutHex64(segment.vmaddr); 6259 buffer.PutHex64(segment.vmsize); 6260 buffer.PutHex64(segment.fileoff); 6261 buffer.PutHex64(segment.filesize); 6262 } else { 6263 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6264 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6265 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6266 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6267 } 6268 buffer.PutHex32(segment.maxprot); 6269 buffer.PutHex32(segment.initprot); 6270 buffer.PutHex32(segment.nsects); 6271 buffer.PutHex32(segment.flags); 6272 } 6273 6274 File core_file; 6275 std::string core_file_path(outfile.GetPath()); 6276 error = FileSystem::Instance().Open(core_file, outfile, 6277 File::eOpenOptionWrite | 6278 File::eOpenOptionTruncate | 6279 File::eOpenOptionCanCreate); 6280 if (error.Success()) { 6281 // Read 1 page at a time 6282 uint8_t bytes[0x1000]; 6283 // Write the mach header and load commands out to the core file 6284 size_t bytes_written = buffer.GetString().size(); 6285 error = core_file.Write(buffer.GetString().data(), bytes_written); 6286 if (error.Success()) { 6287 // Now write the file data for all memory segments in the process 6288 for (const auto &segment : segment_load_commands) { 6289 if (core_file.SeekFromStart(segment.fileoff) == -1) { 6290 error.SetErrorStringWithFormat( 6291 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6292 segment.fileoff, core_file_path.c_str()); 6293 break; 6294 } 6295 6296 printf("Saving %" PRId64 6297 " bytes of data for memory region at 0x%" PRIx64 "\n", 6298 segment.vmsize, segment.vmaddr); 6299 addr_t bytes_left = segment.vmsize; 6300 addr_t addr = segment.vmaddr; 6301 Status memory_read_error; 6302 while (bytes_left > 0 && error.Success()) { 6303 const size_t bytes_to_read = 6304 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6305 6306 // In a savecore setting, we don't really care about caching, 6307 // as the data is dumped and very likely never read again, 6308 // so we call ReadMemoryFromInferior to bypass it. 6309 const size_t bytes_read = process_sp->ReadMemoryFromInferior( 6310 addr, bytes, bytes_to_read, memory_read_error); 6311 6312 if (bytes_read == bytes_to_read) { 6313 size_t bytes_written = bytes_read; 6314 error = core_file.Write(bytes, bytes_written); 6315 bytes_left -= bytes_read; 6316 addr += bytes_read; 6317 } else { 6318 // Some pages within regions are not readable, those should 6319 // be zero filled 6320 memset(bytes, 0, bytes_to_read); 6321 size_t bytes_written = bytes_to_read; 6322 error = core_file.Write(bytes, bytes_written); 6323 bytes_left -= bytes_to_read; 6324 addr += bytes_to_read; 6325 } 6326 } 6327 } 6328 } 6329 } 6330 } else { 6331 error.SetErrorString( 6332 "process doesn't support getting memory region info"); 6333 } 6334 } 6335 return true; // This is the right plug to handle saving core files for 6336 // this process 6337 } 6338 } 6339 return false; 6340 } 6341