1 //===-- ObjectFileMachO.cpp -----------------------------------------------===// 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/Progress.h" 21 #include "lldb/Core/Section.h" 22 #include "lldb/Core/StreamFile.h" 23 #include "lldb/Host/Host.h" 24 #include "lldb/Symbol/DWARFCallFrameInfo.h" 25 #include "lldb/Symbol/ObjectFile.h" 26 #include "lldb/Target/DynamicLoader.h" 27 #include "lldb/Target/MemoryRegionInfo.h" 28 #include "lldb/Target/Platform.h" 29 #include "lldb/Target/Process.h" 30 #include "lldb/Target/SectionLoadList.h" 31 #include "lldb/Target/Target.h" 32 #include "lldb/Target/Thread.h" 33 #include "lldb/Target/ThreadList.h" 34 #include "lldb/Utility/ArchSpec.h" 35 #include "lldb/Utility/DataBuffer.h" 36 #include "lldb/Utility/FileSpec.h" 37 #include "lldb/Utility/Log.h" 38 #include "lldb/Utility/RangeMap.h" 39 #include "lldb/Utility/RegisterValue.h" 40 #include "lldb/Utility/Status.h" 41 #include "lldb/Utility/StreamString.h" 42 #include "lldb/Utility/Timer.h" 43 #include "lldb/Utility/UUID.h" 44 45 #include "lldb/Host/SafeMachO.h" 46 47 #include "llvm/Support/FormatVariadic.h" 48 #include "llvm/Support/MemoryBuffer.h" 49 50 #include "ObjectFileMachO.h" 51 52 #if defined(__APPLE__) 53 #include <TargetConditionals.h> 54 // GetLLDBSharedCacheUUID() needs to call dlsym() 55 #include <dlfcn.h> 56 #endif 57 58 #ifndef __APPLE__ 59 #include "Utility/UuidCompatibility.h" 60 #else 61 #include <uuid/uuid.h> 62 #endif 63 64 #include <memory> 65 66 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 67 using namespace lldb; 68 using namespace lldb_private; 69 using namespace llvm::MachO; 70 71 LLDB_PLUGIN_DEFINE(ObjectFileMachO) 72 73 // Some structure definitions needed for parsing the dyld shared cache files 74 // found on iOS devices. 75 76 struct lldb_copy_dyld_cache_header_v1 { 77 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 78 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 79 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 80 uint32_t imagesOffset; 81 uint32_t imagesCount; 82 uint64_t dyldBaseAddress; 83 uint64_t codeSignatureOffset; 84 uint64_t codeSignatureSize; 85 uint64_t slideInfoOffset; 86 uint64_t slideInfoSize; 87 uint64_t localSymbolsOffset; 88 uint64_t localSymbolsSize; 89 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 90 // and later 91 }; 92 93 struct lldb_copy_dyld_cache_mapping_info { 94 uint64_t address; 95 uint64_t size; 96 uint64_t fileOffset; 97 uint32_t maxProt; 98 uint32_t initProt; 99 }; 100 101 struct lldb_copy_dyld_cache_local_symbols_info { 102 uint32_t nlistOffset; 103 uint32_t nlistCount; 104 uint32_t stringsOffset; 105 uint32_t stringsSize; 106 uint32_t entriesOffset; 107 uint32_t entriesCount; 108 }; 109 struct lldb_copy_dyld_cache_local_symbols_entry { 110 uint32_t dylibOffset; 111 uint32_t nlistStartIndex; 112 uint32_t nlistCount; 113 }; 114 115 static void PrintRegisterValue(RegisterContext *reg_ctx, const char *name, 116 const char *alt_name, size_t reg_byte_size, 117 Stream &data) { 118 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 119 if (reg_info == nullptr) 120 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 121 if (reg_info) { 122 lldb_private::RegisterValue reg_value; 123 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 124 if (reg_info->byte_size >= reg_byte_size) 125 data.Write(reg_value.GetBytes(), reg_byte_size); 126 else { 127 data.Write(reg_value.GetBytes(), reg_info->byte_size); 128 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; ++i) 129 data.PutChar(0); 130 } 131 return; 132 } 133 } 134 // Just write zeros if all else fails 135 for (size_t i = 0; i < reg_byte_size; ++i) 136 data.PutChar(0); 137 } 138 139 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 140 public: 141 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 142 const DataExtractor &data) 143 : RegisterContextDarwin_x86_64(thread, 0) { 144 SetRegisterDataFrom_LC_THREAD(data); 145 } 146 147 void InvalidateAllRegisters() override { 148 // Do nothing... registers are always valid... 149 } 150 151 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 152 lldb::offset_t offset = 0; 153 SetError(GPRRegSet, Read, -1); 154 SetError(FPURegSet, Read, -1); 155 SetError(EXCRegSet, Read, -1); 156 bool done = false; 157 158 while (!done) { 159 int flavor = data.GetU32(&offset); 160 if (flavor == 0) 161 done = true; 162 else { 163 uint32_t i; 164 uint32_t count = data.GetU32(&offset); 165 switch (flavor) { 166 case GPRRegSet: 167 for (i = 0; i < count; ++i) 168 (&gpr.rax)[i] = data.GetU64(&offset); 169 SetError(GPRRegSet, Read, 0); 170 done = true; 171 172 break; 173 case FPURegSet: 174 // TODO: fill in FPU regs.... 175 // SetError (FPURegSet, Read, -1); 176 done = true; 177 178 break; 179 case EXCRegSet: 180 exc.trapno = data.GetU32(&offset); 181 exc.err = data.GetU32(&offset); 182 exc.faultvaddr = data.GetU64(&offset); 183 SetError(EXCRegSet, Read, 0); 184 done = true; 185 break; 186 case 7: 187 case 8: 188 case 9: 189 // fancy flavors that encapsulate of the above flavors... 190 break; 191 192 default: 193 done = true; 194 break; 195 } 196 } 197 } 198 } 199 200 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 201 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 202 if (reg_ctx_sp) { 203 RegisterContext *reg_ctx = reg_ctx_sp.get(); 204 205 data.PutHex32(GPRRegSet); // Flavor 206 data.PutHex32(GPRWordCount); 207 PrintRegisterValue(reg_ctx, "rax", nullptr, 8, data); 208 PrintRegisterValue(reg_ctx, "rbx", nullptr, 8, data); 209 PrintRegisterValue(reg_ctx, "rcx", nullptr, 8, data); 210 PrintRegisterValue(reg_ctx, "rdx", nullptr, 8, data); 211 PrintRegisterValue(reg_ctx, "rdi", nullptr, 8, data); 212 PrintRegisterValue(reg_ctx, "rsi", nullptr, 8, data); 213 PrintRegisterValue(reg_ctx, "rbp", nullptr, 8, data); 214 PrintRegisterValue(reg_ctx, "rsp", nullptr, 8, data); 215 PrintRegisterValue(reg_ctx, "r8", nullptr, 8, data); 216 PrintRegisterValue(reg_ctx, "r9", nullptr, 8, data); 217 PrintRegisterValue(reg_ctx, "r10", nullptr, 8, data); 218 PrintRegisterValue(reg_ctx, "r11", nullptr, 8, data); 219 PrintRegisterValue(reg_ctx, "r12", nullptr, 8, data); 220 PrintRegisterValue(reg_ctx, "r13", nullptr, 8, data); 221 PrintRegisterValue(reg_ctx, "r14", nullptr, 8, data); 222 PrintRegisterValue(reg_ctx, "r15", nullptr, 8, data); 223 PrintRegisterValue(reg_ctx, "rip", nullptr, 8, data); 224 PrintRegisterValue(reg_ctx, "rflags", nullptr, 8, data); 225 PrintRegisterValue(reg_ctx, "cs", nullptr, 8, data); 226 PrintRegisterValue(reg_ctx, "fs", nullptr, 8, data); 227 PrintRegisterValue(reg_ctx, "gs", nullptr, 8, data); 228 229 // // Write out the FPU registers 230 // const size_t fpu_byte_size = sizeof(FPU); 231 // size_t bytes_written = 0; 232 // data.PutHex32 (FPURegSet); 233 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 234 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 235 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 236 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 237 // data); // uint16_t fcw; // "fctrl" 238 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 239 // data); // uint16_t fsw; // "fstat" 240 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 241 // data); // uint8_t ftw; // "ftag" 242 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 243 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 244 // data); // uint16_t fop; // "fop" 245 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 246 // data); // uint32_t ip; // "fioff" 247 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 248 // data); // uint16_t cs; // "fiseg" 249 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 250 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 251 // data); // uint32_t dp; // "fooff" 252 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 253 // data); // uint16_t ds; // "foseg" 254 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 255 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 256 // data); // uint32_t mxcsr; 257 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 258 // 4, data);// uint32_t mxcsrmask; 259 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 260 // sizeof(MMSReg), data); 261 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 262 // sizeof(MMSReg), data); 263 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 264 // sizeof(MMSReg), data); 265 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 266 // sizeof(MMSReg), data); 267 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 268 // sizeof(MMSReg), data); 269 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 270 // sizeof(MMSReg), data); 271 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 272 // sizeof(MMSReg), data); 273 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 274 // sizeof(MMSReg), data); 275 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 276 // sizeof(XMMReg), data); 277 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 278 // sizeof(XMMReg), data); 279 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 280 // sizeof(XMMReg), data); 281 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 282 // sizeof(XMMReg), data); 283 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 284 // sizeof(XMMReg), data); 285 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 286 // sizeof(XMMReg), data); 287 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 288 // sizeof(XMMReg), data); 289 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 290 // sizeof(XMMReg), data); 291 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 292 // sizeof(XMMReg), data); 293 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 294 // sizeof(XMMReg), data); 295 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 296 // sizeof(XMMReg), data); 297 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 298 // sizeof(XMMReg), data); 299 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 300 // sizeof(XMMReg), data); 301 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 302 // sizeof(XMMReg), data); 303 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 304 // sizeof(XMMReg), data); 305 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 306 // sizeof(XMMReg), data); 307 // 308 // // Fill rest with zeros 309 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 310 // i) 311 // data.PutChar(0); 312 313 // Write out the EXC registers 314 data.PutHex32(EXCRegSet); 315 data.PutHex32(EXCWordCount); 316 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 317 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 318 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 8, data); 319 return true; 320 } 321 return false; 322 } 323 324 protected: 325 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 326 327 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 328 329 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 330 331 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 332 return 0; 333 } 334 335 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 336 return 0; 337 } 338 339 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 340 return 0; 341 } 342 }; 343 344 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 345 public: 346 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 347 const DataExtractor &data) 348 : RegisterContextDarwin_i386(thread, 0) { 349 SetRegisterDataFrom_LC_THREAD(data); 350 } 351 352 void InvalidateAllRegisters() override { 353 // Do nothing... registers are always valid... 354 } 355 356 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 357 lldb::offset_t offset = 0; 358 SetError(GPRRegSet, Read, -1); 359 SetError(FPURegSet, Read, -1); 360 SetError(EXCRegSet, Read, -1); 361 bool done = false; 362 363 while (!done) { 364 int flavor = data.GetU32(&offset); 365 if (flavor == 0) 366 done = true; 367 else { 368 uint32_t i; 369 uint32_t count = data.GetU32(&offset); 370 switch (flavor) { 371 case GPRRegSet: 372 for (i = 0; i < count; ++i) 373 (&gpr.eax)[i] = data.GetU32(&offset); 374 SetError(GPRRegSet, Read, 0); 375 done = true; 376 377 break; 378 case FPURegSet: 379 // TODO: fill in FPU regs.... 380 // SetError (FPURegSet, Read, -1); 381 done = true; 382 383 break; 384 case EXCRegSet: 385 exc.trapno = data.GetU32(&offset); 386 exc.err = data.GetU32(&offset); 387 exc.faultvaddr = data.GetU32(&offset); 388 SetError(EXCRegSet, Read, 0); 389 done = true; 390 break; 391 case 7: 392 case 8: 393 case 9: 394 // fancy flavors that encapsulate of the above flavors... 395 break; 396 397 default: 398 done = true; 399 break; 400 } 401 } 402 } 403 } 404 405 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 406 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 407 if (reg_ctx_sp) { 408 RegisterContext *reg_ctx = reg_ctx_sp.get(); 409 410 data.PutHex32(GPRRegSet); // Flavor 411 data.PutHex32(GPRWordCount); 412 PrintRegisterValue(reg_ctx, "eax", nullptr, 4, data); 413 PrintRegisterValue(reg_ctx, "ebx", nullptr, 4, data); 414 PrintRegisterValue(reg_ctx, "ecx", nullptr, 4, data); 415 PrintRegisterValue(reg_ctx, "edx", nullptr, 4, data); 416 PrintRegisterValue(reg_ctx, "edi", nullptr, 4, data); 417 PrintRegisterValue(reg_ctx, "esi", nullptr, 4, data); 418 PrintRegisterValue(reg_ctx, "ebp", nullptr, 4, data); 419 PrintRegisterValue(reg_ctx, "esp", nullptr, 4, data); 420 PrintRegisterValue(reg_ctx, "ss", nullptr, 4, data); 421 PrintRegisterValue(reg_ctx, "eflags", nullptr, 4, data); 422 PrintRegisterValue(reg_ctx, "eip", nullptr, 4, data); 423 PrintRegisterValue(reg_ctx, "cs", nullptr, 4, data); 424 PrintRegisterValue(reg_ctx, "ds", nullptr, 4, data); 425 PrintRegisterValue(reg_ctx, "es", nullptr, 4, data); 426 PrintRegisterValue(reg_ctx, "fs", nullptr, 4, data); 427 PrintRegisterValue(reg_ctx, "gs", nullptr, 4, data); 428 429 // Write out the EXC registers 430 data.PutHex32(EXCRegSet); 431 data.PutHex32(EXCWordCount); 432 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 433 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 434 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 4, data); 435 return true; 436 } 437 return false; 438 } 439 440 protected: 441 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 442 443 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 444 445 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 446 447 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 448 return 0; 449 } 450 451 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 452 return 0; 453 } 454 455 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 456 return 0; 457 } 458 }; 459 460 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 461 public: 462 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 463 const DataExtractor &data) 464 : RegisterContextDarwin_arm(thread, 0) { 465 SetRegisterDataFrom_LC_THREAD(data); 466 } 467 468 void InvalidateAllRegisters() override { 469 // Do nothing... registers are always valid... 470 } 471 472 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 473 lldb::offset_t offset = 0; 474 SetError(GPRRegSet, Read, -1); 475 SetError(FPURegSet, Read, -1); 476 SetError(EXCRegSet, Read, -1); 477 bool done = false; 478 479 while (!done) { 480 int flavor = data.GetU32(&offset); 481 uint32_t count = data.GetU32(&offset); 482 lldb::offset_t next_thread_state = offset + (count * 4); 483 switch (flavor) { 484 case GPRAltRegSet: 485 case GPRRegSet: 486 // On ARM, the CPSR register is also included in the count but it is 487 // not included in gpr.r so loop until (count-1). 488 for (uint32_t i = 0; i < (count - 1); ++i) { 489 gpr.r[i] = data.GetU32(&offset); 490 } 491 // Save cpsr explicitly. 492 gpr.cpsr = data.GetU32(&offset); 493 494 SetError(GPRRegSet, Read, 0); 495 offset = next_thread_state; 496 break; 497 498 case FPURegSet: { 499 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 500 const int fpu_reg_buf_size = sizeof(fpu.floats); 501 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 502 fpu_reg_buf) == fpu_reg_buf_size) { 503 offset += fpu_reg_buf_size; 504 fpu.fpscr = data.GetU32(&offset); 505 SetError(FPURegSet, Read, 0); 506 } else { 507 done = true; 508 } 509 } 510 offset = next_thread_state; 511 break; 512 513 case EXCRegSet: 514 if (count == 3) { 515 exc.exception = data.GetU32(&offset); 516 exc.fsr = data.GetU32(&offset); 517 exc.far = data.GetU32(&offset); 518 SetError(EXCRegSet, Read, 0); 519 } 520 done = true; 521 offset = next_thread_state; 522 break; 523 524 // Unknown register set flavor, stop trying to parse. 525 default: 526 done = true; 527 } 528 } 529 } 530 531 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 532 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 533 if (reg_ctx_sp) { 534 RegisterContext *reg_ctx = reg_ctx_sp.get(); 535 536 data.PutHex32(GPRRegSet); // Flavor 537 data.PutHex32(GPRWordCount); 538 PrintRegisterValue(reg_ctx, "r0", nullptr, 4, data); 539 PrintRegisterValue(reg_ctx, "r1", nullptr, 4, data); 540 PrintRegisterValue(reg_ctx, "r2", nullptr, 4, data); 541 PrintRegisterValue(reg_ctx, "r3", nullptr, 4, data); 542 PrintRegisterValue(reg_ctx, "r4", nullptr, 4, data); 543 PrintRegisterValue(reg_ctx, "r5", nullptr, 4, data); 544 PrintRegisterValue(reg_ctx, "r6", nullptr, 4, data); 545 PrintRegisterValue(reg_ctx, "r7", nullptr, 4, data); 546 PrintRegisterValue(reg_ctx, "r8", nullptr, 4, data); 547 PrintRegisterValue(reg_ctx, "r9", nullptr, 4, data); 548 PrintRegisterValue(reg_ctx, "r10", nullptr, 4, data); 549 PrintRegisterValue(reg_ctx, "r11", nullptr, 4, data); 550 PrintRegisterValue(reg_ctx, "r12", nullptr, 4, data); 551 PrintRegisterValue(reg_ctx, "sp", nullptr, 4, data); 552 PrintRegisterValue(reg_ctx, "lr", nullptr, 4, data); 553 PrintRegisterValue(reg_ctx, "pc", nullptr, 4, data); 554 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 555 556 // Write out the EXC registers 557 // data.PutHex32 (EXCRegSet); 558 // data.PutHex32 (EXCWordCount); 559 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 560 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 561 // WriteRegister (reg_ctx, "far", NULL, 4, data); 562 return true; 563 } 564 return false; 565 } 566 567 protected: 568 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 569 570 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 571 572 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 573 574 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 575 576 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 577 return 0; 578 } 579 580 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 581 return 0; 582 } 583 584 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 585 return 0; 586 } 587 588 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 589 return -1; 590 } 591 }; 592 593 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 594 public: 595 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 596 const DataExtractor &data) 597 : RegisterContextDarwin_arm64(thread, 0) { 598 SetRegisterDataFrom_LC_THREAD(data); 599 } 600 601 void InvalidateAllRegisters() override { 602 // Do nothing... registers are always valid... 603 } 604 605 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 606 lldb::offset_t offset = 0; 607 SetError(GPRRegSet, Read, -1); 608 SetError(FPURegSet, Read, -1); 609 SetError(EXCRegSet, Read, -1); 610 bool done = false; 611 while (!done) { 612 int flavor = data.GetU32(&offset); 613 uint32_t count = data.GetU32(&offset); 614 lldb::offset_t next_thread_state = offset + (count * 4); 615 switch (flavor) { 616 case GPRRegSet: 617 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 618 // 32-bit register) 619 if (count >= (33 * 2) + 1) { 620 for (uint32_t i = 0; i < 29; ++i) 621 gpr.x[i] = data.GetU64(&offset); 622 gpr.fp = data.GetU64(&offset); 623 gpr.lr = data.GetU64(&offset); 624 gpr.sp = data.GetU64(&offset); 625 gpr.pc = data.GetU64(&offset); 626 gpr.cpsr = data.GetU32(&offset); 627 SetError(GPRRegSet, Read, 0); 628 } 629 offset = next_thread_state; 630 break; 631 case FPURegSet: { 632 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 633 const int fpu_reg_buf_size = sizeof(fpu); 634 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 635 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 636 fpu_reg_buf) == fpu_reg_buf_size) { 637 SetError(FPURegSet, Read, 0); 638 } else { 639 done = true; 640 } 641 } 642 offset = next_thread_state; 643 break; 644 case EXCRegSet: 645 if (count == 4) { 646 exc.far = data.GetU64(&offset); 647 exc.esr = data.GetU32(&offset); 648 exc.exception = data.GetU32(&offset); 649 SetError(EXCRegSet, Read, 0); 650 } 651 offset = next_thread_state; 652 break; 653 default: 654 done = true; 655 break; 656 } 657 } 658 } 659 660 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 661 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 662 if (reg_ctx_sp) { 663 RegisterContext *reg_ctx = reg_ctx_sp.get(); 664 665 data.PutHex32(GPRRegSet); // Flavor 666 data.PutHex32(GPRWordCount); 667 PrintRegisterValue(reg_ctx, "x0", nullptr, 8, data); 668 PrintRegisterValue(reg_ctx, "x1", nullptr, 8, data); 669 PrintRegisterValue(reg_ctx, "x2", nullptr, 8, data); 670 PrintRegisterValue(reg_ctx, "x3", nullptr, 8, data); 671 PrintRegisterValue(reg_ctx, "x4", nullptr, 8, data); 672 PrintRegisterValue(reg_ctx, "x5", nullptr, 8, data); 673 PrintRegisterValue(reg_ctx, "x6", nullptr, 8, data); 674 PrintRegisterValue(reg_ctx, "x7", nullptr, 8, data); 675 PrintRegisterValue(reg_ctx, "x8", nullptr, 8, data); 676 PrintRegisterValue(reg_ctx, "x9", nullptr, 8, data); 677 PrintRegisterValue(reg_ctx, "x10", nullptr, 8, data); 678 PrintRegisterValue(reg_ctx, "x11", nullptr, 8, data); 679 PrintRegisterValue(reg_ctx, "x12", nullptr, 8, data); 680 PrintRegisterValue(reg_ctx, "x13", nullptr, 8, data); 681 PrintRegisterValue(reg_ctx, "x14", nullptr, 8, data); 682 PrintRegisterValue(reg_ctx, "x15", nullptr, 8, data); 683 PrintRegisterValue(reg_ctx, "x16", nullptr, 8, data); 684 PrintRegisterValue(reg_ctx, "x17", nullptr, 8, data); 685 PrintRegisterValue(reg_ctx, "x18", nullptr, 8, data); 686 PrintRegisterValue(reg_ctx, "x19", nullptr, 8, data); 687 PrintRegisterValue(reg_ctx, "x20", nullptr, 8, data); 688 PrintRegisterValue(reg_ctx, "x21", nullptr, 8, data); 689 PrintRegisterValue(reg_ctx, "x22", nullptr, 8, data); 690 PrintRegisterValue(reg_ctx, "x23", nullptr, 8, data); 691 PrintRegisterValue(reg_ctx, "x24", nullptr, 8, data); 692 PrintRegisterValue(reg_ctx, "x25", nullptr, 8, data); 693 PrintRegisterValue(reg_ctx, "x26", nullptr, 8, data); 694 PrintRegisterValue(reg_ctx, "x27", nullptr, 8, data); 695 PrintRegisterValue(reg_ctx, "x28", nullptr, 8, data); 696 PrintRegisterValue(reg_ctx, "fp", nullptr, 8, data); 697 PrintRegisterValue(reg_ctx, "lr", nullptr, 8, data); 698 PrintRegisterValue(reg_ctx, "sp", nullptr, 8, data); 699 PrintRegisterValue(reg_ctx, "pc", nullptr, 8, data); 700 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 701 702 // Write out the EXC registers 703 // data.PutHex32 (EXCRegSet); 704 // data.PutHex32 (EXCWordCount); 705 // WriteRegister (reg_ctx, "far", NULL, 8, data); 706 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 707 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 708 return true; 709 } 710 return false; 711 } 712 713 protected: 714 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 715 716 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 717 718 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 719 720 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 721 722 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 723 return 0; 724 } 725 726 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 727 return 0; 728 } 729 730 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 731 return 0; 732 } 733 734 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 735 return -1; 736 } 737 }; 738 739 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 740 switch (magic) { 741 case MH_MAGIC: 742 case MH_CIGAM: 743 return sizeof(struct mach_header); 744 745 case MH_MAGIC_64: 746 case MH_CIGAM_64: 747 return sizeof(struct mach_header_64); 748 break; 749 750 default: 751 break; 752 } 753 return 0; 754 } 755 756 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 757 758 char ObjectFileMachO::ID; 759 760 void ObjectFileMachO::Initialize() { 761 PluginManager::RegisterPlugin( 762 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 763 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 764 } 765 766 void ObjectFileMachO::Terminate() { 767 PluginManager::UnregisterPlugin(CreateInstance); 768 } 769 770 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 771 static ConstString g_name("mach-o"); 772 return g_name; 773 } 774 775 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 776 return "Mach-o object file reader (32 and 64 bit)"; 777 } 778 779 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 780 DataBufferSP &data_sp, 781 lldb::offset_t data_offset, 782 const FileSpec *file, 783 lldb::offset_t file_offset, 784 lldb::offset_t length) { 785 if (!data_sp) { 786 data_sp = MapFileData(*file, length, file_offset); 787 if (!data_sp) 788 return nullptr; 789 data_offset = 0; 790 } 791 792 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 793 return nullptr; 794 795 // Update the data to contain the entire file if it doesn't already 796 if (data_sp->GetByteSize() < length) { 797 data_sp = MapFileData(*file, length, file_offset); 798 if (!data_sp) 799 return nullptr; 800 data_offset = 0; 801 } 802 auto objfile_up = std::make_unique<ObjectFileMachO>( 803 module_sp, data_sp, data_offset, file, file_offset, length); 804 if (!objfile_up || !objfile_up->ParseHeader()) 805 return nullptr; 806 807 return objfile_up.release(); 808 } 809 810 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 811 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 812 const ProcessSP &process_sp, lldb::addr_t header_addr) { 813 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 814 std::unique_ptr<ObjectFile> objfile_up( 815 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 816 if (objfile_up.get() && objfile_up->ParseHeader()) 817 return objfile_up.release(); 818 } 819 return nullptr; 820 } 821 822 size_t ObjectFileMachO::GetModuleSpecifications( 823 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 824 lldb::offset_t data_offset, lldb::offset_t file_offset, 825 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 826 const size_t initial_count = specs.GetSize(); 827 828 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 829 DataExtractor data; 830 data.SetData(data_sp); 831 llvm::MachO::mach_header header; 832 if (ParseHeader(data, &data_offset, header)) { 833 size_t header_and_load_cmds = 834 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 835 if (header_and_load_cmds >= data_sp->GetByteSize()) { 836 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 837 data.SetData(data_sp); 838 data_offset = MachHeaderSizeFromMagic(header.magic); 839 } 840 if (data_sp) { 841 ModuleSpec base_spec; 842 base_spec.GetFileSpec() = file; 843 base_spec.SetObjectOffset(file_offset); 844 base_spec.SetObjectSize(length); 845 GetAllArchSpecs(header, data, data_offset, base_spec, specs); 846 } 847 } 848 } 849 return specs.GetSize() - initial_count; 850 } 851 852 ConstString ObjectFileMachO::GetSegmentNameTEXT() { 853 static ConstString g_segment_name_TEXT("__TEXT"); 854 return g_segment_name_TEXT; 855 } 856 857 ConstString ObjectFileMachO::GetSegmentNameDATA() { 858 static ConstString g_segment_name_DATA("__DATA"); 859 return g_segment_name_DATA; 860 } 861 862 ConstString ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 863 static ConstString g_segment_name("__DATA_DIRTY"); 864 return g_segment_name; 865 } 866 867 ConstString ObjectFileMachO::GetSegmentNameDATA_CONST() { 868 static ConstString g_segment_name("__DATA_CONST"); 869 return g_segment_name; 870 } 871 872 ConstString ObjectFileMachO::GetSegmentNameOBJC() { 873 static ConstString g_segment_name_OBJC("__OBJC"); 874 return g_segment_name_OBJC; 875 } 876 877 ConstString ObjectFileMachO::GetSegmentNameLINKEDIT() { 878 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 879 return g_section_name_LINKEDIT; 880 } 881 882 ConstString ObjectFileMachO::GetSegmentNameDWARF() { 883 static ConstString g_section_name("__DWARF"); 884 return g_section_name; 885 } 886 887 ConstString ObjectFileMachO::GetSectionNameEHFrame() { 888 static ConstString g_section_name_eh_frame("__eh_frame"); 889 return g_section_name_eh_frame; 890 } 891 892 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 893 lldb::addr_t data_offset, 894 lldb::addr_t data_length) { 895 DataExtractor data; 896 data.SetData(data_sp, data_offset, data_length); 897 lldb::offset_t offset = 0; 898 uint32_t magic = data.GetU32(&offset); 899 return MachHeaderSizeFromMagic(magic) != 0; 900 } 901 902 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 903 DataBufferSP &data_sp, 904 lldb::offset_t data_offset, 905 const FileSpec *file, 906 lldb::offset_t file_offset, 907 lldb::offset_t length) 908 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 909 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 910 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 911 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 912 ::memset(&m_header, 0, sizeof(m_header)); 913 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 914 } 915 916 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 917 lldb::DataBufferSP &header_data_sp, 918 const lldb::ProcessSP &process_sp, 919 lldb::addr_t header_addr) 920 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 921 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 922 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 923 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 924 ::memset(&m_header, 0, sizeof(m_header)); 925 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 926 } 927 928 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 929 lldb::offset_t *data_offset_ptr, 930 llvm::MachO::mach_header &header) { 931 data.SetByteOrder(endian::InlHostByteOrder()); 932 // Leave magic in the original byte order 933 header.magic = data.GetU32(data_offset_ptr); 934 bool can_parse = false; 935 bool is_64_bit = false; 936 switch (header.magic) { 937 case MH_MAGIC: 938 data.SetByteOrder(endian::InlHostByteOrder()); 939 data.SetAddressByteSize(4); 940 can_parse = true; 941 break; 942 943 case MH_MAGIC_64: 944 data.SetByteOrder(endian::InlHostByteOrder()); 945 data.SetAddressByteSize(8); 946 can_parse = true; 947 is_64_bit = true; 948 break; 949 950 case MH_CIGAM: 951 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 952 ? eByteOrderLittle 953 : eByteOrderBig); 954 data.SetAddressByteSize(4); 955 can_parse = true; 956 break; 957 958 case MH_CIGAM_64: 959 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 960 ? eByteOrderLittle 961 : eByteOrderBig); 962 data.SetAddressByteSize(8); 963 is_64_bit = true; 964 can_parse = true; 965 break; 966 967 default: 968 break; 969 } 970 971 if (can_parse) { 972 data.GetU32(data_offset_ptr, &header.cputype, 6); 973 if (is_64_bit) 974 *data_offset_ptr += 4; 975 return true; 976 } else { 977 memset(&header, 0, sizeof(header)); 978 } 979 return false; 980 } 981 982 bool ObjectFileMachO::ParseHeader() { 983 ModuleSP module_sp(GetModule()); 984 if (!module_sp) 985 return false; 986 987 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 988 bool can_parse = false; 989 lldb::offset_t offset = 0; 990 m_data.SetByteOrder(endian::InlHostByteOrder()); 991 // Leave magic in the original byte order 992 m_header.magic = m_data.GetU32(&offset); 993 switch (m_header.magic) { 994 case MH_MAGIC: 995 m_data.SetByteOrder(endian::InlHostByteOrder()); 996 m_data.SetAddressByteSize(4); 997 can_parse = true; 998 break; 999 1000 case MH_MAGIC_64: 1001 m_data.SetByteOrder(endian::InlHostByteOrder()); 1002 m_data.SetAddressByteSize(8); 1003 can_parse = true; 1004 break; 1005 1006 case MH_CIGAM: 1007 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1008 ? eByteOrderLittle 1009 : eByteOrderBig); 1010 m_data.SetAddressByteSize(4); 1011 can_parse = true; 1012 break; 1013 1014 case MH_CIGAM_64: 1015 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1016 ? eByteOrderLittle 1017 : eByteOrderBig); 1018 m_data.SetAddressByteSize(8); 1019 can_parse = true; 1020 break; 1021 1022 default: 1023 break; 1024 } 1025 1026 if (can_parse) { 1027 m_data.GetU32(&offset, &m_header.cputype, 6); 1028 1029 ModuleSpecList all_specs; 1030 ModuleSpec base_spec; 1031 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 1032 base_spec, all_specs); 1033 1034 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 1035 ArchSpec mach_arch = 1036 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 1037 1038 // Check if the module has a required architecture 1039 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1040 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1041 continue; 1042 1043 if (SetModulesArchitecture(mach_arch)) { 1044 const size_t header_and_lc_size = 1045 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1046 if (m_data.GetByteSize() < header_and_lc_size) { 1047 DataBufferSP data_sp; 1048 ProcessSP process_sp(m_process_wp.lock()); 1049 if (process_sp) { 1050 data_sp = ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1051 } else { 1052 // Read in all only the load command data from the file on disk 1053 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset); 1054 if (data_sp->GetByteSize() != header_and_lc_size) 1055 continue; 1056 } 1057 if (data_sp) 1058 m_data.SetData(data_sp); 1059 } 1060 } 1061 return true; 1062 } 1063 // None found. 1064 return false; 1065 } else { 1066 memset(&m_header, 0, sizeof(struct mach_header)); 1067 } 1068 return false; 1069 } 1070 1071 ByteOrder ObjectFileMachO::GetByteOrder() const { 1072 return m_data.GetByteOrder(); 1073 } 1074 1075 bool ObjectFileMachO::IsExecutable() const { 1076 return m_header.filetype == MH_EXECUTE; 1077 } 1078 1079 bool ObjectFileMachO::IsDynamicLoader() const { 1080 return m_header.filetype == MH_DYLINKER; 1081 } 1082 1083 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1084 return m_data.GetAddressByteSize(); 1085 } 1086 1087 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1088 Symtab *symtab = GetSymtab(); 1089 if (!symtab) 1090 return AddressClass::eUnknown; 1091 1092 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1093 if (symbol) { 1094 if (symbol->ValueIsAddress()) { 1095 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1096 if (section_sp) { 1097 const lldb::SectionType section_type = section_sp->GetType(); 1098 switch (section_type) { 1099 case eSectionTypeInvalid: 1100 return AddressClass::eUnknown; 1101 1102 case eSectionTypeCode: 1103 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1104 // For ARM we have a bit in the n_desc field of the symbol that 1105 // tells us ARM/Thumb which is bit 0x0008. 1106 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1107 return AddressClass::eCodeAlternateISA; 1108 } 1109 return AddressClass::eCode; 1110 1111 case eSectionTypeContainer: 1112 return AddressClass::eUnknown; 1113 1114 case eSectionTypeData: 1115 case eSectionTypeDataCString: 1116 case eSectionTypeDataCStringPointers: 1117 case eSectionTypeDataSymbolAddress: 1118 case eSectionTypeData4: 1119 case eSectionTypeData8: 1120 case eSectionTypeData16: 1121 case eSectionTypeDataPointers: 1122 case eSectionTypeZeroFill: 1123 case eSectionTypeDataObjCMessageRefs: 1124 case eSectionTypeDataObjCCFStrings: 1125 case eSectionTypeGoSymtab: 1126 return AddressClass::eData; 1127 1128 case eSectionTypeDebug: 1129 case eSectionTypeDWARFDebugAbbrev: 1130 case eSectionTypeDWARFDebugAbbrevDwo: 1131 case eSectionTypeDWARFDebugAddr: 1132 case eSectionTypeDWARFDebugAranges: 1133 case eSectionTypeDWARFDebugCuIndex: 1134 case eSectionTypeDWARFDebugFrame: 1135 case eSectionTypeDWARFDebugInfo: 1136 case eSectionTypeDWARFDebugInfoDwo: 1137 case eSectionTypeDWARFDebugLine: 1138 case eSectionTypeDWARFDebugLineStr: 1139 case eSectionTypeDWARFDebugLoc: 1140 case eSectionTypeDWARFDebugLocDwo: 1141 case eSectionTypeDWARFDebugLocLists: 1142 case eSectionTypeDWARFDebugLocListsDwo: 1143 case eSectionTypeDWARFDebugMacInfo: 1144 case eSectionTypeDWARFDebugMacro: 1145 case eSectionTypeDWARFDebugNames: 1146 case eSectionTypeDWARFDebugPubNames: 1147 case eSectionTypeDWARFDebugPubTypes: 1148 case eSectionTypeDWARFDebugRanges: 1149 case eSectionTypeDWARFDebugRngLists: 1150 case eSectionTypeDWARFDebugRngListsDwo: 1151 case eSectionTypeDWARFDebugStr: 1152 case eSectionTypeDWARFDebugStrDwo: 1153 case eSectionTypeDWARFDebugStrOffsets: 1154 case eSectionTypeDWARFDebugStrOffsetsDwo: 1155 case eSectionTypeDWARFDebugTuIndex: 1156 case eSectionTypeDWARFDebugTypes: 1157 case eSectionTypeDWARFDebugTypesDwo: 1158 case eSectionTypeDWARFAppleNames: 1159 case eSectionTypeDWARFAppleTypes: 1160 case eSectionTypeDWARFAppleNamespaces: 1161 case eSectionTypeDWARFAppleObjC: 1162 case eSectionTypeDWARFGNUDebugAltLink: 1163 return AddressClass::eDebug; 1164 1165 case eSectionTypeEHFrame: 1166 case eSectionTypeARMexidx: 1167 case eSectionTypeARMextab: 1168 case eSectionTypeCompactUnwind: 1169 return AddressClass::eRuntime; 1170 1171 case eSectionTypeAbsoluteAddress: 1172 case eSectionTypeELFSymbolTable: 1173 case eSectionTypeELFDynamicSymbols: 1174 case eSectionTypeELFRelocationEntries: 1175 case eSectionTypeELFDynamicLinkInfo: 1176 case eSectionTypeOther: 1177 return AddressClass::eUnknown; 1178 } 1179 } 1180 } 1181 1182 const SymbolType symbol_type = symbol->GetType(); 1183 switch (symbol_type) { 1184 case eSymbolTypeAny: 1185 return AddressClass::eUnknown; 1186 case eSymbolTypeAbsolute: 1187 return AddressClass::eUnknown; 1188 1189 case eSymbolTypeCode: 1190 case eSymbolTypeTrampoline: 1191 case eSymbolTypeResolver: 1192 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1193 // For ARM we have a bit in the n_desc field of the symbol that tells 1194 // us ARM/Thumb which is bit 0x0008. 1195 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1196 return AddressClass::eCodeAlternateISA; 1197 } 1198 return AddressClass::eCode; 1199 1200 case eSymbolTypeData: 1201 return AddressClass::eData; 1202 case eSymbolTypeRuntime: 1203 return AddressClass::eRuntime; 1204 case eSymbolTypeException: 1205 return AddressClass::eRuntime; 1206 case eSymbolTypeSourceFile: 1207 return AddressClass::eDebug; 1208 case eSymbolTypeHeaderFile: 1209 return AddressClass::eDebug; 1210 case eSymbolTypeObjectFile: 1211 return AddressClass::eDebug; 1212 case eSymbolTypeCommonBlock: 1213 return AddressClass::eDebug; 1214 case eSymbolTypeBlock: 1215 return AddressClass::eDebug; 1216 case eSymbolTypeLocal: 1217 return AddressClass::eData; 1218 case eSymbolTypeParam: 1219 return AddressClass::eData; 1220 case eSymbolTypeVariable: 1221 return AddressClass::eData; 1222 case eSymbolTypeVariableType: 1223 return AddressClass::eDebug; 1224 case eSymbolTypeLineEntry: 1225 return AddressClass::eDebug; 1226 case eSymbolTypeLineHeader: 1227 return AddressClass::eDebug; 1228 case eSymbolTypeScopeBegin: 1229 return AddressClass::eDebug; 1230 case eSymbolTypeScopeEnd: 1231 return AddressClass::eDebug; 1232 case eSymbolTypeAdditional: 1233 return AddressClass::eUnknown; 1234 case eSymbolTypeCompiler: 1235 return AddressClass::eDebug; 1236 case eSymbolTypeInstrumentation: 1237 return AddressClass::eDebug; 1238 case eSymbolTypeUndefined: 1239 return AddressClass::eUnknown; 1240 case eSymbolTypeObjCClass: 1241 return AddressClass::eRuntime; 1242 case eSymbolTypeObjCMetaClass: 1243 return AddressClass::eRuntime; 1244 case eSymbolTypeObjCIVar: 1245 return AddressClass::eRuntime; 1246 case eSymbolTypeReExported: 1247 return AddressClass::eRuntime; 1248 } 1249 } 1250 return AddressClass::eUnknown; 1251 } 1252 1253 Symtab *ObjectFileMachO::GetSymtab() { 1254 ModuleSP module_sp(GetModule()); 1255 if (module_sp) { 1256 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1257 if (m_symtab_up == nullptr) { 1258 m_symtab_up = std::make_unique<Symtab>(this); 1259 std::lock_guard<std::recursive_mutex> symtab_guard( 1260 m_symtab_up->GetMutex()); 1261 ParseSymtab(); 1262 m_symtab_up->Finalize(); 1263 } 1264 } 1265 return m_symtab_up.get(); 1266 } 1267 1268 bool ObjectFileMachO::IsStripped() { 1269 if (m_dysymtab.cmd == 0) { 1270 ModuleSP module_sp(GetModule()); 1271 if (module_sp) { 1272 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1273 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1274 const lldb::offset_t load_cmd_offset = offset; 1275 1276 load_command lc; 1277 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 1278 break; 1279 if (lc.cmd == LC_DYSYMTAB) { 1280 m_dysymtab.cmd = lc.cmd; 1281 m_dysymtab.cmdsize = lc.cmdsize; 1282 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1283 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1284 nullptr) { 1285 // Clear m_dysymtab if we were unable to read all items from the 1286 // load command 1287 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1288 } 1289 } 1290 offset = load_cmd_offset + lc.cmdsize; 1291 } 1292 } 1293 } 1294 if (m_dysymtab.cmd) 1295 return m_dysymtab.nlocalsym <= 1; 1296 return false; 1297 } 1298 1299 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1300 EncryptedFileRanges result; 1301 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1302 1303 encryption_info_command encryption_cmd; 1304 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1305 const lldb::offset_t load_cmd_offset = offset; 1306 if (m_data.GetU32(&offset, &encryption_cmd, 2) == nullptr) 1307 break; 1308 1309 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the 1310 // 3 fields we care about, so treat them the same. 1311 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1312 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1313 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1314 if (encryption_cmd.cryptid != 0) { 1315 EncryptedFileRanges::Entry entry; 1316 entry.SetRangeBase(encryption_cmd.cryptoff); 1317 entry.SetByteSize(encryption_cmd.cryptsize); 1318 result.Append(entry); 1319 } 1320 } 1321 } 1322 offset = load_cmd_offset + encryption_cmd.cmdsize; 1323 } 1324 1325 return result; 1326 } 1327 1328 void ObjectFileMachO::SanitizeSegmentCommand(segment_command_64 &seg_cmd, 1329 uint32_t cmd_idx) { 1330 if (m_length == 0 || seg_cmd.filesize == 0) 1331 return; 1332 1333 if ((m_header.flags & MH_DYLIB_IN_CACHE) && !IsInMemory()) { 1334 // In shared cache images, the load commands are relative to the 1335 // shared cache file, and not the the specific image we are 1336 // examining. Let's fix this up so that it looks like a normal 1337 // image. 1338 if (strncmp(seg_cmd.segname, "__TEXT", sizeof(seg_cmd.segname)) == 0) 1339 m_text_address = seg_cmd.vmaddr; 1340 if (strncmp(seg_cmd.segname, "__LINKEDIT", sizeof(seg_cmd.segname)) == 0) 1341 m_linkedit_original_offset = seg_cmd.fileoff; 1342 1343 seg_cmd.fileoff = seg_cmd.vmaddr - m_text_address; 1344 } 1345 1346 if (seg_cmd.fileoff > m_length) { 1347 // We have a load command that says it extends past the end of the file. 1348 // This is likely a corrupt file. We don't have any way to return an error 1349 // condition here (this method was likely invoked from something like 1350 // ObjectFile::GetSectionList()), so we just null out the section contents, 1351 // and dump a message to stdout. The most common case here is core file 1352 // debugging with a truncated file. 1353 const char *lc_segment_name = 1354 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1355 GetModule()->ReportWarning( 1356 "load command %u %s has a fileoff (0x%" PRIx64 1357 ") that extends beyond the end of the file (0x%" PRIx64 1358 "), ignoring this section", 1359 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1360 1361 seg_cmd.fileoff = 0; 1362 seg_cmd.filesize = 0; 1363 } 1364 1365 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1366 // We have a load command that says it extends past the end of the file. 1367 // This is likely a corrupt file. We don't have any way to return an error 1368 // condition here (this method was likely invoked from something like 1369 // ObjectFile::GetSectionList()), so we just null out the section contents, 1370 // and dump a message to stdout. The most common case here is core file 1371 // debugging with a truncated file. 1372 const char *lc_segment_name = 1373 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1374 GetModule()->ReportWarning( 1375 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1376 ") that extends beyond the end of the file (0x%" PRIx64 1377 "), the segment will be truncated to match", 1378 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1379 1380 // Truncate the length 1381 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1382 } 1383 } 1384 1385 static uint32_t GetSegmentPermissions(const segment_command_64 &seg_cmd) { 1386 uint32_t result = 0; 1387 if (seg_cmd.initprot & VM_PROT_READ) 1388 result |= ePermissionsReadable; 1389 if (seg_cmd.initprot & VM_PROT_WRITE) 1390 result |= ePermissionsWritable; 1391 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1392 result |= ePermissionsExecutable; 1393 return result; 1394 } 1395 1396 static lldb::SectionType GetSectionType(uint32_t flags, 1397 ConstString section_name) { 1398 1399 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1400 return eSectionTypeCode; 1401 1402 uint32_t mach_sect_type = flags & SECTION_TYPE; 1403 static ConstString g_sect_name_objc_data("__objc_data"); 1404 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1405 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1406 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1407 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1408 static ConstString g_sect_name_objc_const("__objc_const"); 1409 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1410 static ConstString g_sect_name_cfstring("__cfstring"); 1411 1412 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1413 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1414 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1415 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1416 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1417 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1418 static ConstString g_sect_name_dwarf_debug_loclists("__debug_loclists"); 1419 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1420 static ConstString g_sect_name_dwarf_debug_names("__debug_names"); 1421 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1422 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1423 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1424 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1425 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1426 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1427 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1428 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1429 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1430 static ConstString g_sect_name_eh_frame("__eh_frame"); 1431 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1432 static ConstString g_sect_name_text("__text"); 1433 static ConstString g_sect_name_data("__data"); 1434 static ConstString g_sect_name_go_symtab("__gosymtab"); 1435 1436 if (section_name == g_sect_name_dwarf_debug_abbrev) 1437 return eSectionTypeDWARFDebugAbbrev; 1438 if (section_name == g_sect_name_dwarf_debug_aranges) 1439 return eSectionTypeDWARFDebugAranges; 1440 if (section_name == g_sect_name_dwarf_debug_frame) 1441 return eSectionTypeDWARFDebugFrame; 1442 if (section_name == g_sect_name_dwarf_debug_info) 1443 return eSectionTypeDWARFDebugInfo; 1444 if (section_name == g_sect_name_dwarf_debug_line) 1445 return eSectionTypeDWARFDebugLine; 1446 if (section_name == g_sect_name_dwarf_debug_loc) 1447 return eSectionTypeDWARFDebugLoc; 1448 if (section_name == g_sect_name_dwarf_debug_loclists) 1449 return eSectionTypeDWARFDebugLocLists; 1450 if (section_name == g_sect_name_dwarf_debug_macinfo) 1451 return eSectionTypeDWARFDebugMacInfo; 1452 if (section_name == g_sect_name_dwarf_debug_names) 1453 return eSectionTypeDWARFDebugNames; 1454 if (section_name == g_sect_name_dwarf_debug_pubnames) 1455 return eSectionTypeDWARFDebugPubNames; 1456 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1457 return eSectionTypeDWARFDebugPubTypes; 1458 if (section_name == g_sect_name_dwarf_debug_ranges) 1459 return eSectionTypeDWARFDebugRanges; 1460 if (section_name == g_sect_name_dwarf_debug_str) 1461 return eSectionTypeDWARFDebugStr; 1462 if (section_name == g_sect_name_dwarf_debug_types) 1463 return eSectionTypeDWARFDebugTypes; 1464 if (section_name == g_sect_name_dwarf_apple_names) 1465 return eSectionTypeDWARFAppleNames; 1466 if (section_name == g_sect_name_dwarf_apple_types) 1467 return eSectionTypeDWARFAppleTypes; 1468 if (section_name == g_sect_name_dwarf_apple_namespaces) 1469 return eSectionTypeDWARFAppleNamespaces; 1470 if (section_name == g_sect_name_dwarf_apple_objc) 1471 return eSectionTypeDWARFAppleObjC; 1472 if (section_name == g_sect_name_objc_selrefs) 1473 return eSectionTypeDataCStringPointers; 1474 if (section_name == g_sect_name_objc_msgrefs) 1475 return eSectionTypeDataObjCMessageRefs; 1476 if (section_name == g_sect_name_eh_frame) 1477 return eSectionTypeEHFrame; 1478 if (section_name == g_sect_name_compact_unwind) 1479 return eSectionTypeCompactUnwind; 1480 if (section_name == g_sect_name_cfstring) 1481 return eSectionTypeDataObjCCFStrings; 1482 if (section_name == g_sect_name_go_symtab) 1483 return eSectionTypeGoSymtab; 1484 if (section_name == g_sect_name_objc_data || 1485 section_name == g_sect_name_objc_classrefs || 1486 section_name == g_sect_name_objc_superrefs || 1487 section_name == g_sect_name_objc_const || 1488 section_name == g_sect_name_objc_classlist) { 1489 return eSectionTypeDataPointers; 1490 } 1491 1492 switch (mach_sect_type) { 1493 // TODO: categorize sections by other flags for regular sections 1494 case S_REGULAR: 1495 if (section_name == g_sect_name_text) 1496 return eSectionTypeCode; 1497 if (section_name == g_sect_name_data) 1498 return eSectionTypeData; 1499 return eSectionTypeOther; 1500 case S_ZEROFILL: 1501 return eSectionTypeZeroFill; 1502 case S_CSTRING_LITERALS: // section with only literal C strings 1503 return eSectionTypeDataCString; 1504 case S_4BYTE_LITERALS: // section with only 4 byte literals 1505 return eSectionTypeData4; 1506 case S_8BYTE_LITERALS: // section with only 8 byte literals 1507 return eSectionTypeData8; 1508 case S_LITERAL_POINTERS: // section with only pointers to literals 1509 return eSectionTypeDataPointers; 1510 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1511 return eSectionTypeDataPointers; 1512 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1513 return eSectionTypeDataPointers; 1514 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1515 // the reserved2 field 1516 return eSectionTypeCode; 1517 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1518 // initialization 1519 return eSectionTypeDataPointers; 1520 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1521 // termination 1522 return eSectionTypeDataPointers; 1523 case S_COALESCED: 1524 return eSectionTypeOther; 1525 case S_GB_ZEROFILL: 1526 return eSectionTypeZeroFill; 1527 case S_INTERPOSING: // section with only pairs of function pointers for 1528 // interposing 1529 return eSectionTypeCode; 1530 case S_16BYTE_LITERALS: // section with only 16 byte literals 1531 return eSectionTypeData16; 1532 case S_DTRACE_DOF: 1533 return eSectionTypeDebug; 1534 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1535 return eSectionTypeDataPointers; 1536 default: 1537 return eSectionTypeOther; 1538 } 1539 } 1540 1541 struct ObjectFileMachO::SegmentParsingContext { 1542 const EncryptedFileRanges EncryptedRanges; 1543 lldb_private::SectionList &UnifiedList; 1544 uint32_t NextSegmentIdx = 0; 1545 uint32_t NextSectionIdx = 0; 1546 bool FileAddressesChanged = false; 1547 1548 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1549 lldb_private::SectionList &UnifiedList) 1550 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1551 }; 1552 1553 void ObjectFileMachO::ProcessSegmentCommand(const load_command &load_cmd_, 1554 lldb::offset_t offset, 1555 uint32_t cmd_idx, 1556 SegmentParsingContext &context) { 1557 segment_command_64 load_cmd; 1558 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1559 1560 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1561 return; 1562 1563 ModuleSP module_sp = GetModule(); 1564 const bool is_core = GetType() == eTypeCoreFile; 1565 const bool is_dsym = (m_header.filetype == MH_DSYM); 1566 bool add_section = true; 1567 bool add_to_unified = true; 1568 ConstString const_segname( 1569 load_cmd.segname, strnlen(load_cmd.segname, sizeof(load_cmd.segname))); 1570 1571 SectionSP unified_section_sp( 1572 context.UnifiedList.FindSectionByName(const_segname)); 1573 if (is_dsym && unified_section_sp) { 1574 if (const_segname == GetSegmentNameLINKEDIT()) { 1575 // We need to keep the __LINKEDIT segment private to this object file 1576 // only 1577 add_to_unified = false; 1578 } else { 1579 // This is the dSYM file and this section has already been created by the 1580 // object file, no need to create it. 1581 add_section = false; 1582 } 1583 } 1584 load_cmd.vmaddr = m_data.GetAddress(&offset); 1585 load_cmd.vmsize = m_data.GetAddress(&offset); 1586 load_cmd.fileoff = m_data.GetAddress(&offset); 1587 load_cmd.filesize = m_data.GetAddress(&offset); 1588 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1589 return; 1590 1591 SanitizeSegmentCommand(load_cmd, cmd_idx); 1592 1593 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1594 const bool segment_is_encrypted = 1595 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1596 1597 // Keep a list of mach segments around in case we need to get at data that 1598 // isn't stored in the abstracted Sections. 1599 m_mach_segments.push_back(load_cmd); 1600 1601 // Use a segment ID of the segment index shifted left by 8 so they never 1602 // conflict with any of the sections. 1603 SectionSP segment_sp; 1604 if (add_section && (const_segname || is_core)) { 1605 segment_sp = std::make_shared<Section>( 1606 module_sp, // Module to which this section belongs 1607 this, // Object file to which this sections belongs 1608 ++context.NextSegmentIdx 1609 << 8, // Section ID is the 1 based segment index 1610 // shifted right by 8 bits as not to collide with any of the 256 1611 // section IDs that are possible 1612 const_segname, // Name of this section 1613 eSectionTypeContainer, // This section is a container of other 1614 // sections. 1615 load_cmd.vmaddr, // File VM address == addresses as they are 1616 // found in the object file 1617 load_cmd.vmsize, // VM size in bytes of this section 1618 load_cmd.fileoff, // Offset to the data for this section in 1619 // the file 1620 load_cmd.filesize, // Size in bytes of this section as found 1621 // in the file 1622 0, // Segments have no alignment information 1623 load_cmd.flags); // Flags for this section 1624 1625 segment_sp->SetIsEncrypted(segment_is_encrypted); 1626 m_sections_up->AddSection(segment_sp); 1627 segment_sp->SetPermissions(segment_permissions); 1628 if (add_to_unified) 1629 context.UnifiedList.AddSection(segment_sp); 1630 } else if (unified_section_sp) { 1631 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1632 // Check to see if the module was read from memory? 1633 if (module_sp->GetObjectFile()->IsInMemory()) { 1634 // We have a module that is in memory and needs to have its file 1635 // address adjusted. We need to do this because when we load a file 1636 // from memory, its addresses will be slid already, yet the addresses 1637 // in the new symbol file will still be unslid. Since everything is 1638 // stored as section offset, this shouldn't cause any problems. 1639 1640 // Make sure we've parsed the symbol table from the ObjectFile before 1641 // we go around changing its Sections. 1642 module_sp->GetObjectFile()->GetSymtab(); 1643 // eh_frame would present the same problems but we parse that on a per- 1644 // function basis as-needed so it's more difficult to remove its use of 1645 // the Sections. Realistically, the environments where this code path 1646 // will be taken will not have eh_frame sections. 1647 1648 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1649 1650 // Notify the module that the section addresses have been changed once 1651 // we're done so any file-address caches can be updated. 1652 context.FileAddressesChanged = true; 1653 } 1654 } 1655 m_sections_up->AddSection(unified_section_sp); 1656 } 1657 1658 struct section_64 sect64; 1659 ::memset(§64, 0, sizeof(sect64)); 1660 // Push a section into our mach sections for the section at index zero 1661 // (NO_SECT) if we don't have any mach sections yet... 1662 if (m_mach_sections.empty()) 1663 m_mach_sections.push_back(sect64); 1664 uint32_t segment_sect_idx; 1665 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1666 1667 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1668 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1669 ++segment_sect_idx) { 1670 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1671 sizeof(sect64.sectname)) == nullptr) 1672 break; 1673 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1674 sizeof(sect64.segname)) == nullptr) 1675 break; 1676 sect64.addr = m_data.GetAddress(&offset); 1677 sect64.size = m_data.GetAddress(&offset); 1678 1679 if (m_data.GetU32(&offset, §64.offset, num_u32s) == nullptr) 1680 break; 1681 1682 if ((m_header.flags & MH_DYLIB_IN_CACHE) && !IsInMemory()) { 1683 sect64.offset = sect64.addr - m_text_address; 1684 } 1685 1686 // Keep a list of mach sections around in case we need to get at data that 1687 // isn't stored in the abstracted Sections. 1688 m_mach_sections.push_back(sect64); 1689 1690 if (add_section) { 1691 ConstString section_name( 1692 sect64.sectname, strnlen(sect64.sectname, sizeof(sect64.sectname))); 1693 if (!const_segname) { 1694 // We have a segment with no name so we need to conjure up segments 1695 // that correspond to the section's segname if there isn't already such 1696 // a section. If there is such a section, we resize the section so that 1697 // it spans all sections. We also mark these sections as fake so 1698 // address matches don't hit if they land in the gaps between the child 1699 // sections. 1700 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1701 sizeof(sect64.segname)); 1702 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1703 if (segment_sp.get()) { 1704 Section *segment = segment_sp.get(); 1705 // Grow the section size as needed. 1706 const lldb::addr_t sect64_min_addr = sect64.addr; 1707 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1708 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1709 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1710 const lldb::addr_t curr_seg_max_addr = 1711 curr_seg_min_addr + curr_seg_byte_size; 1712 if (sect64_min_addr >= curr_seg_min_addr) { 1713 const lldb::addr_t new_seg_byte_size = 1714 sect64_max_addr - curr_seg_min_addr; 1715 // Only grow the section size if needed 1716 if (new_seg_byte_size > curr_seg_byte_size) 1717 segment->SetByteSize(new_seg_byte_size); 1718 } else { 1719 // We need to change the base address of the segment and adjust the 1720 // child section offsets for all existing children. 1721 const lldb::addr_t slide_amount = 1722 sect64_min_addr - curr_seg_min_addr; 1723 segment->Slide(slide_amount, false); 1724 segment->GetChildren().Slide(-slide_amount, false); 1725 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1726 } 1727 1728 // Grow the section size as needed. 1729 if (sect64.offset) { 1730 const lldb::addr_t segment_min_file_offset = 1731 segment->GetFileOffset(); 1732 const lldb::addr_t segment_max_file_offset = 1733 segment_min_file_offset + segment->GetFileSize(); 1734 1735 const lldb::addr_t section_min_file_offset = sect64.offset; 1736 const lldb::addr_t section_max_file_offset = 1737 section_min_file_offset + sect64.size; 1738 const lldb::addr_t new_file_offset = 1739 std::min(section_min_file_offset, segment_min_file_offset); 1740 const lldb::addr_t new_file_size = 1741 std::max(section_max_file_offset, segment_max_file_offset) - 1742 new_file_offset; 1743 segment->SetFileOffset(new_file_offset); 1744 segment->SetFileSize(new_file_size); 1745 } 1746 } else { 1747 // Create a fake section for the section's named segment 1748 segment_sp = std::make_shared<Section>( 1749 segment_sp, // Parent section 1750 module_sp, // Module to which this section belongs 1751 this, // Object file to which this section belongs 1752 ++context.NextSegmentIdx 1753 << 8, // Section ID is the 1 based segment index 1754 // shifted right by 8 bits as not to 1755 // collide with any of the 256 section IDs 1756 // that are possible 1757 const_segname, // Name of this section 1758 eSectionTypeContainer, // This section is a container of 1759 // other sections. 1760 sect64.addr, // File VM address == addresses as they are 1761 // found in the object file 1762 sect64.size, // VM size in bytes of this section 1763 sect64.offset, // Offset to the data for this section in 1764 // the file 1765 sect64.offset ? sect64.size : 0, // Size in bytes of 1766 // this section as 1767 // found in the file 1768 sect64.align, 1769 load_cmd.flags); // Flags for this section 1770 segment_sp->SetIsFake(true); 1771 segment_sp->SetPermissions(segment_permissions); 1772 m_sections_up->AddSection(segment_sp); 1773 if (add_to_unified) 1774 context.UnifiedList.AddSection(segment_sp); 1775 segment_sp->SetIsEncrypted(segment_is_encrypted); 1776 } 1777 } 1778 assert(segment_sp.get()); 1779 1780 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1781 1782 SectionSP section_sp(new Section( 1783 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1784 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1785 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1786 sect64.flags)); 1787 // Set the section to be encrypted to match the segment 1788 1789 bool section_is_encrypted = false; 1790 if (!segment_is_encrypted && load_cmd.filesize != 0) 1791 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1792 sect64.offset) != nullptr; 1793 1794 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1795 section_sp->SetPermissions(segment_permissions); 1796 segment_sp->GetChildren().AddSection(section_sp); 1797 1798 if (segment_sp->IsFake()) { 1799 segment_sp.reset(); 1800 const_segname.Clear(); 1801 } 1802 } 1803 } 1804 if (segment_sp && is_dsym) { 1805 if (first_segment_sectID <= context.NextSectionIdx) { 1806 lldb::user_id_t sect_uid; 1807 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1808 ++sect_uid) { 1809 SectionSP curr_section_sp( 1810 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1811 SectionSP next_section_sp; 1812 if (sect_uid + 1 <= context.NextSectionIdx) 1813 next_section_sp = 1814 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1815 1816 if (curr_section_sp.get()) { 1817 if (curr_section_sp->GetByteSize() == 0) { 1818 if (next_section_sp.get() != nullptr) 1819 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1820 curr_section_sp->GetFileAddress()); 1821 else 1822 curr_section_sp->SetByteSize(load_cmd.vmsize); 1823 } 1824 } 1825 } 1826 } 1827 } 1828 } 1829 1830 void ObjectFileMachO::ProcessDysymtabCommand(const load_command &load_cmd, 1831 lldb::offset_t offset) { 1832 m_dysymtab.cmd = load_cmd.cmd; 1833 m_dysymtab.cmdsize = load_cmd.cmdsize; 1834 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1835 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1836 } 1837 1838 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1839 if (m_sections_up) 1840 return; 1841 1842 m_sections_up = std::make_unique<SectionList>(); 1843 1844 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1845 // bool dump_sections = false; 1846 ModuleSP module_sp(GetModule()); 1847 1848 offset = MachHeaderSizeFromMagic(m_header.magic); 1849 1850 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1851 struct load_command load_cmd; 1852 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1853 const lldb::offset_t load_cmd_offset = offset; 1854 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 1855 break; 1856 1857 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1858 ProcessSegmentCommand(load_cmd, offset, i, context); 1859 else if (load_cmd.cmd == LC_DYSYMTAB) 1860 ProcessDysymtabCommand(load_cmd, offset); 1861 1862 offset = load_cmd_offset + load_cmd.cmdsize; 1863 } 1864 1865 if (context.FileAddressesChanged && module_sp) 1866 module_sp->SectionFileAddressesChanged(); 1867 } 1868 1869 class MachSymtabSectionInfo { 1870 public: 1871 MachSymtabSectionInfo(SectionList *section_list) 1872 : m_section_list(section_list), m_section_infos() { 1873 // Get the number of sections down to a depth of 1 to include all segments 1874 // and their sections, but no other sections that may be added for debug 1875 // map or 1876 m_section_infos.resize(section_list->GetNumSections(1)); 1877 } 1878 1879 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1880 if (n_sect == 0) 1881 return SectionSP(); 1882 if (n_sect < m_section_infos.size()) { 1883 if (!m_section_infos[n_sect].section_sp) { 1884 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1885 m_section_infos[n_sect].section_sp = section_sp; 1886 if (section_sp) { 1887 m_section_infos[n_sect].vm_range.SetBaseAddress( 1888 section_sp->GetFileAddress()); 1889 m_section_infos[n_sect].vm_range.SetByteSize( 1890 section_sp->GetByteSize()); 1891 } else { 1892 std::string filename = "<unknown>"; 1893 SectionSP first_section_sp(m_section_list->GetSectionAtIndex(0)); 1894 if (first_section_sp) 1895 filename = first_section_sp->GetObjectFile()->GetFileSpec().GetPath(); 1896 1897 Host::SystemLog(Host::eSystemLogError, 1898 "error: unable to find section %d for a symbol in %s, corrupt file?\n", 1899 n_sect, 1900 filename.c_str()); 1901 } 1902 } 1903 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1904 // Symbol is in section. 1905 return m_section_infos[n_sect].section_sp; 1906 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1907 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1908 file_addr) { 1909 // Symbol is in section with zero size, but has the same start address 1910 // as the section. This can happen with linker symbols (symbols that 1911 // start with the letter 'l' or 'L'. 1912 return m_section_infos[n_sect].section_sp; 1913 } 1914 } 1915 return m_section_list->FindSectionContainingFileAddress(file_addr); 1916 } 1917 1918 protected: 1919 struct SectionInfo { 1920 SectionInfo() : vm_range(), section_sp() {} 1921 1922 VMRange vm_range; 1923 SectionSP section_sp; 1924 }; 1925 SectionList *m_section_list; 1926 std::vector<SectionInfo> m_section_infos; 1927 }; 1928 1929 #define TRIE_SYMBOL_IS_THUMB (1ULL << 63) 1930 struct TrieEntry { 1931 void Dump() const { 1932 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1933 static_cast<unsigned long long>(address), 1934 static_cast<unsigned long long>(flags), 1935 static_cast<unsigned long long>(other), name.GetCString()); 1936 if (import_name) 1937 printf(" -> \"%s\"\n", import_name.GetCString()); 1938 else 1939 printf("\n"); 1940 } 1941 ConstString name; 1942 uint64_t address = LLDB_INVALID_ADDRESS; 1943 uint64_t flags = 1944 0; // EXPORT_SYMBOL_FLAGS_REEXPORT, EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER, 1945 // TRIE_SYMBOL_IS_THUMB 1946 uint64_t other = 0; 1947 ConstString import_name; 1948 }; 1949 1950 struct TrieEntryWithOffset { 1951 lldb::offset_t nodeOffset; 1952 TrieEntry entry; 1953 1954 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1955 1956 void Dump(uint32_t idx) const { 1957 printf("[%3u] 0x%16.16llx: ", idx, 1958 static_cast<unsigned long long>(nodeOffset)); 1959 entry.Dump(); 1960 } 1961 1962 bool operator<(const TrieEntryWithOffset &other) const { 1963 return (nodeOffset < other.nodeOffset); 1964 } 1965 }; 1966 1967 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 1968 const bool is_arm, addr_t text_seg_base_addr, 1969 std::vector<llvm::StringRef> &nameSlices, 1970 std::set<lldb::addr_t> &resolver_addresses, 1971 std::vector<TrieEntryWithOffset> &reexports, 1972 std::vector<TrieEntryWithOffset> &ext_symbols) { 1973 if (!data.ValidOffset(offset)) 1974 return true; 1975 1976 // Terminal node -- end of a branch, possibly add this to 1977 // the symbol table or resolver table. 1978 const uint64_t terminalSize = data.GetULEB128(&offset); 1979 lldb::offset_t children_offset = offset + terminalSize; 1980 if (terminalSize != 0) { 1981 TrieEntryWithOffset e(offset); 1982 e.entry.flags = data.GetULEB128(&offset); 1983 const char *import_name = nullptr; 1984 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 1985 e.entry.address = 0; 1986 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 1987 import_name = data.GetCStr(&offset); 1988 } else { 1989 e.entry.address = data.GetULEB128(&offset); 1990 if (text_seg_base_addr != LLDB_INVALID_ADDRESS) 1991 e.entry.address += text_seg_base_addr; 1992 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 1993 e.entry.other = data.GetULEB128(&offset); 1994 uint64_t resolver_addr = e.entry.other; 1995 if (text_seg_base_addr != LLDB_INVALID_ADDRESS) 1996 resolver_addr += text_seg_base_addr; 1997 if (is_arm) 1998 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 1999 resolver_addresses.insert(resolver_addr); 2000 } else 2001 e.entry.other = 0; 2002 } 2003 bool add_this_entry = false; 2004 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT) && 2005 import_name && import_name[0]) { 2006 // add symbols that are reexport symbols with a valid import name. 2007 add_this_entry = true; 2008 } else if (e.entry.flags == 0 && 2009 (import_name == nullptr || import_name[0] == '\0')) { 2010 // add externally visible symbols, in case the nlist record has 2011 // been stripped/omitted. 2012 add_this_entry = true; 2013 } 2014 if (add_this_entry) { 2015 std::string name; 2016 if (!nameSlices.empty()) { 2017 for (auto name_slice : nameSlices) 2018 name.append(name_slice.data(), name_slice.size()); 2019 } 2020 if (name.size() > 1) { 2021 // Skip the leading '_' 2022 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2023 } 2024 if (import_name) { 2025 // Skip the leading '_' 2026 e.entry.import_name.SetCString(import_name + 1); 2027 } 2028 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT)) { 2029 reexports.push_back(e); 2030 } else { 2031 if (is_arm && (e.entry.address & 1)) { 2032 e.entry.flags |= TRIE_SYMBOL_IS_THUMB; 2033 e.entry.address &= THUMB_ADDRESS_BIT_MASK; 2034 } 2035 ext_symbols.push_back(e); 2036 } 2037 } 2038 } 2039 2040 const uint8_t childrenCount = data.GetU8(&children_offset); 2041 for (uint8_t i = 0; i < childrenCount; ++i) { 2042 const char *cstr = data.GetCStr(&children_offset); 2043 if (cstr) 2044 nameSlices.push_back(llvm::StringRef(cstr)); 2045 else 2046 return false; // Corrupt data 2047 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2048 if (childNodeOffset) { 2049 if (!ParseTrieEntries(data, childNodeOffset, is_arm, text_seg_base_addr, 2050 nameSlices, resolver_addresses, reexports, 2051 ext_symbols)) { 2052 return false; 2053 } 2054 } 2055 nameSlices.pop_back(); 2056 } 2057 return true; 2058 } 2059 2060 static SymbolType GetSymbolType(const char *&symbol_name, 2061 bool &demangled_is_synthesized, 2062 const SectionSP &text_section_sp, 2063 const SectionSP &data_section_sp, 2064 const SectionSP &data_dirty_section_sp, 2065 const SectionSP &data_const_section_sp, 2066 const SectionSP &symbol_section) { 2067 SymbolType type = eSymbolTypeInvalid; 2068 2069 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2070 if (symbol_section->IsDescendant(text_section_sp.get())) { 2071 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 2072 S_ATTR_SELF_MODIFYING_CODE | 2073 S_ATTR_SOME_INSTRUCTIONS)) 2074 type = eSymbolTypeData; 2075 else 2076 type = eSymbolTypeCode; 2077 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 2078 symbol_section->IsDescendant(data_dirty_section_sp.get()) || 2079 symbol_section->IsDescendant(data_const_section_sp.get())) { 2080 if (symbol_sect_name && 2081 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 2082 type = eSymbolTypeRuntime; 2083 2084 if (symbol_name) { 2085 llvm::StringRef symbol_name_ref(symbol_name); 2086 if (symbol_name_ref.startswith("OBJC_")) { 2087 static const llvm::StringRef g_objc_v2_prefix_class("OBJC_CLASS_$_"); 2088 static const llvm::StringRef g_objc_v2_prefix_metaclass( 2089 "OBJC_METACLASS_$_"); 2090 static const llvm::StringRef g_objc_v2_prefix_ivar("OBJC_IVAR_$_"); 2091 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 2092 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2093 type = eSymbolTypeObjCClass; 2094 demangled_is_synthesized = true; 2095 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 2096 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2097 type = eSymbolTypeObjCMetaClass; 2098 demangled_is_synthesized = true; 2099 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 2100 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2101 type = eSymbolTypeObjCIVar; 2102 demangled_is_synthesized = true; 2103 } 2104 } 2105 } 2106 } else if (symbol_sect_name && 2107 ::strstr(symbol_sect_name, "__gcc_except_tab") == 2108 symbol_sect_name) { 2109 type = eSymbolTypeException; 2110 } else { 2111 type = eSymbolTypeData; 2112 } 2113 } else if (symbol_sect_name && 2114 ::strstr(symbol_sect_name, "__IMPORT") == symbol_sect_name) { 2115 type = eSymbolTypeTrampoline; 2116 } 2117 return type; 2118 } 2119 2120 // Read the UUID out of a dyld_shared_cache file on-disk. 2121 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2122 const ByteOrder byte_order, 2123 const uint32_t addr_byte_size) { 2124 UUID dsc_uuid; 2125 DataBufferSP DscData = MapFileData( 2126 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2127 if (!DscData) 2128 return dsc_uuid; 2129 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2130 2131 char version_str[7]; 2132 lldb::offset_t offset = 0; 2133 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2134 version_str[6] = '\0'; 2135 if (strcmp(version_str, "dyld_v") == 0) { 2136 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2137 dsc_uuid = UUID::fromOptionalData( 2138 dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t)); 2139 } 2140 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2141 if (log && dsc_uuid.IsValid()) { 2142 LLDB_LOGF(log, "Shared cache %s has UUID %s", 2143 dyld_shared_cache.GetPath().c_str(), 2144 dsc_uuid.GetAsString().c_str()); 2145 } 2146 return dsc_uuid; 2147 } 2148 2149 static llvm::Optional<struct nlist_64> 2150 ParseNList(DataExtractor &nlist_data, lldb::offset_t &nlist_data_offset, 2151 size_t nlist_byte_size) { 2152 struct nlist_64 nlist; 2153 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2154 return {}; 2155 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 2156 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 2157 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 2158 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 2159 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 2160 return nlist; 2161 } 2162 2163 enum { DebugSymbols = true, NonDebugSymbols = false }; 2164 2165 size_t ObjectFileMachO::ParseSymtab() { 2166 LLDB_SCOPED_TIMERF("ObjectFileMachO::ParseSymtab () module = %s", 2167 m_file.GetFilename().AsCString("")); 2168 ModuleSP module_sp(GetModule()); 2169 if (!module_sp) 2170 return 0; 2171 2172 Progress progress(llvm::formatv("Parsing symbol table for {0}", 2173 m_file.GetFilename().AsCString("<Unknown>"))); 2174 2175 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2176 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2177 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2178 // The data element of type bool indicates that this entry is thumb 2179 // code. 2180 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2181 2182 // Record the address of every function/data that we add to the symtab. 2183 // We add symbols to the table in the order of most information (nlist 2184 // records) to least (function starts), and avoid duplicating symbols 2185 // via this set. 2186 std::set<addr_t> symbols_added; 2187 FunctionStarts function_starts; 2188 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2189 uint32_t i; 2190 FileSpecList dylib_files; 2191 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2192 llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2193 llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2194 llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2195 2196 for (i = 0; i < m_header.ncmds; ++i) { 2197 const lldb::offset_t cmd_offset = offset; 2198 // Read in the load command and load command size 2199 struct load_command lc; 2200 if (m_data.GetU32(&offset, &lc, 2) == nullptr) 2201 break; 2202 // Watch for the symbol table load command 2203 switch (lc.cmd) { 2204 case LC_SYMTAB: 2205 symtab_load_command.cmd = lc.cmd; 2206 symtab_load_command.cmdsize = lc.cmdsize; 2207 // Read in the rest of the symtab load command 2208 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2209 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2210 return 0; 2211 break; 2212 2213 case LC_DYLD_INFO: 2214 case LC_DYLD_INFO_ONLY: 2215 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2216 dyld_info.cmd = lc.cmd; 2217 dyld_info.cmdsize = lc.cmdsize; 2218 } else { 2219 memset(&dyld_info, 0, sizeof(dyld_info)); 2220 } 2221 break; 2222 2223 case LC_LOAD_DYLIB: 2224 case LC_LOAD_WEAK_DYLIB: 2225 case LC_REEXPORT_DYLIB: 2226 case LC_LOADFVMLIB: 2227 case LC_LOAD_UPWARD_DYLIB: { 2228 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2229 const char *path = m_data.PeekCStr(name_offset); 2230 if (path) { 2231 FileSpec file_spec(path); 2232 // Strip the path if there is @rpath, @executable, etc so we just use 2233 // the basename 2234 if (path[0] == '@') 2235 file_spec.GetDirectory().Clear(); 2236 2237 if (lc.cmd == LC_REEXPORT_DYLIB) { 2238 m_reexported_dylibs.AppendIfUnique(file_spec); 2239 } 2240 2241 dylib_files.Append(file_spec); 2242 } 2243 } break; 2244 2245 case LC_FUNCTION_STARTS: 2246 function_starts_load_command.cmd = lc.cmd; 2247 function_starts_load_command.cmdsize = lc.cmdsize; 2248 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2249 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2250 memset(&function_starts_load_command, 0, 2251 sizeof(function_starts_load_command)); 2252 break; 2253 2254 default: 2255 break; 2256 } 2257 offset = cmd_offset + lc.cmdsize; 2258 } 2259 2260 if (!symtab_load_command.cmd) 2261 return 0; 2262 2263 Symtab *symtab = m_symtab_up.get(); 2264 SectionList *section_list = GetSectionList(); 2265 if (section_list == nullptr) 2266 return 0; 2267 2268 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2269 const ByteOrder byte_order = m_data.GetByteOrder(); 2270 bool bit_width_32 = addr_byte_size == 4; 2271 const size_t nlist_byte_size = 2272 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2273 2274 DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size); 2275 DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size); 2276 DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size); 2277 DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order, 2278 addr_byte_size); 2279 DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size); 2280 2281 const addr_t nlist_data_byte_size = 2282 symtab_load_command.nsyms * nlist_byte_size; 2283 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2284 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2285 2286 ProcessSP process_sp(m_process_wp.lock()); 2287 Process *process = process_sp.get(); 2288 2289 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2290 bool is_shared_cache_image = m_header.flags & MH_DYLIB_IN_CACHE; 2291 bool is_local_shared_cache_image = is_shared_cache_image && !IsInMemory(); 2292 SectionSP linkedit_section_sp( 2293 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2294 2295 if (process && m_header.filetype != llvm::MachO::MH_OBJECT && 2296 !is_local_shared_cache_image) { 2297 Target &target = process->GetTarget(); 2298 2299 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2300 2301 // Reading mach file from memory in a process or core file... 2302 2303 if (linkedit_section_sp) { 2304 addr_t linkedit_load_addr = 2305 linkedit_section_sp->GetLoadBaseAddress(&target); 2306 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2307 // We might be trying to access the symbol table before the 2308 // __LINKEDIT's load address has been set in the target. We can't 2309 // fail to read the symbol table, so calculate the right address 2310 // manually 2311 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2312 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2313 } 2314 2315 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2316 const addr_t symoff_addr = linkedit_load_addr + 2317 symtab_load_command.symoff - 2318 linkedit_file_offset; 2319 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2320 linkedit_file_offset; 2321 2322 // Always load dyld - the dynamic linker - from memory if we didn't 2323 // find a binary anywhere else. lldb will not register 2324 // dylib/framework/bundle loads/unloads if we don't have the dyld 2325 // symbols, we force dyld to load from memory despite the user's 2326 // target.memory-module-load-level setting. 2327 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2328 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2329 DataBufferSP nlist_data_sp( 2330 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2331 if (nlist_data_sp) 2332 nlist_data.SetData(nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2333 if (m_dysymtab.nindirectsyms != 0) { 2334 const addr_t indirect_syms_addr = linkedit_load_addr + 2335 m_dysymtab.indirectsymoff - 2336 linkedit_file_offset; 2337 DataBufferSP indirect_syms_data_sp(ReadMemory( 2338 process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 2339 if (indirect_syms_data_sp) 2340 indirect_symbol_index_data.SetData( 2341 indirect_syms_data_sp, 0, 2342 indirect_syms_data_sp->GetByteSize()); 2343 // If this binary is outside the shared cache, 2344 // cache the string table. 2345 // Binaries in the shared cache all share a giant string table, 2346 // and we can't share the string tables across multiple 2347 // ObjectFileMachO's, so we'd end up re-reading this mega-strtab 2348 // for every binary in the shared cache - it would be a big perf 2349 // problem. For binaries outside the shared cache, it's faster to 2350 // read the entire strtab at once instead of piece-by-piece as we 2351 // process the nlist records. 2352 if (!is_shared_cache_image) { 2353 DataBufferSP strtab_data_sp( 2354 ReadMemory(process_sp, strtab_addr, strtab_data_byte_size)); 2355 if (strtab_data_sp) { 2356 strtab_data.SetData(strtab_data_sp, 0, 2357 strtab_data_sp->GetByteSize()); 2358 } 2359 } 2360 } 2361 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) { 2362 if (function_starts_load_command.cmd) { 2363 const addr_t func_start_addr = 2364 linkedit_load_addr + function_starts_load_command.dataoff - 2365 linkedit_file_offset; 2366 DataBufferSP func_start_data_sp( 2367 ReadMemory(process_sp, func_start_addr, 2368 function_starts_load_command.datasize)); 2369 if (func_start_data_sp) 2370 function_starts_data.SetData(func_start_data_sp, 0, 2371 func_start_data_sp->GetByteSize()); 2372 } 2373 } 2374 } 2375 } 2376 } else { 2377 if (is_local_shared_cache_image) { 2378 // The load commands in shared cache images are relative to the 2379 // beginning of the shared cache, not the library image. The 2380 // data we get handed when creating the ObjectFileMachO starts 2381 // at the beginning of a specific library and spans to the end 2382 // of the cache to be able to reach the shared LINKEDIT 2383 // segments. We need to convert the load command offsets to be 2384 // relative to the beginning of our specific image. 2385 lldb::addr_t linkedit_offset = linkedit_section_sp->GetFileOffset(); 2386 lldb::offset_t linkedit_slide = 2387 linkedit_offset - m_linkedit_original_offset; 2388 symtab_load_command.symoff += linkedit_slide; 2389 symtab_load_command.stroff += linkedit_slide; 2390 dyld_info.export_off += linkedit_slide; 2391 m_dysymtab.indirectsymoff += linkedit_slide; 2392 function_starts_load_command.dataoff += linkedit_slide; 2393 } 2394 2395 nlist_data.SetData(m_data, symtab_load_command.symoff, 2396 nlist_data_byte_size); 2397 strtab_data.SetData(m_data, symtab_load_command.stroff, 2398 strtab_data_byte_size); 2399 2400 if (dyld_info.export_size > 0) { 2401 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2402 dyld_info.export_size); 2403 } 2404 2405 if (m_dysymtab.nindirectsyms != 0) { 2406 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2407 m_dysymtab.nindirectsyms * 4); 2408 } 2409 if (function_starts_load_command.cmd) { 2410 function_starts_data.SetData(m_data, function_starts_load_command.dataoff, 2411 function_starts_load_command.datasize); 2412 } 2413 } 2414 2415 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2416 2417 ConstString g_segment_name_TEXT = GetSegmentNameTEXT(); 2418 ConstString g_segment_name_DATA = GetSegmentNameDATA(); 2419 ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2420 ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2421 ConstString g_segment_name_OBJC = GetSegmentNameOBJC(); 2422 ConstString g_section_name_eh_frame = GetSectionNameEHFrame(); 2423 SectionSP text_section_sp( 2424 section_list->FindSectionByName(g_segment_name_TEXT)); 2425 SectionSP data_section_sp( 2426 section_list->FindSectionByName(g_segment_name_DATA)); 2427 SectionSP data_dirty_section_sp( 2428 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2429 SectionSP data_const_section_sp( 2430 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2431 SectionSP objc_section_sp( 2432 section_list->FindSectionByName(g_segment_name_OBJC)); 2433 SectionSP eh_frame_section_sp; 2434 if (text_section_sp.get()) 2435 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2436 g_section_name_eh_frame); 2437 else 2438 eh_frame_section_sp = 2439 section_list->FindSectionByName(g_section_name_eh_frame); 2440 2441 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2442 const bool always_thumb = GetArchitecture().IsAlwaysThumbInstructions(); 2443 2444 // lldb works best if it knows the start address of all functions in a 2445 // module. Linker symbols or debug info are normally the best source of 2446 // information for start addr / size but they may be stripped in a released 2447 // binary. Two additional sources of information exist in Mach-O binaries: 2448 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2449 // function's start address in the 2450 // binary, relative to the text section. 2451 // eh_frame - the eh_frame FDEs have the start addr & size of 2452 // each function 2453 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2454 // all modern binaries. 2455 // Binaries built to run on older releases may need to use eh_frame 2456 // information. 2457 2458 if (text_section_sp && function_starts_data.GetByteSize()) { 2459 FunctionStarts::Entry function_start_entry; 2460 function_start_entry.data = false; 2461 lldb::offset_t function_start_offset = 0; 2462 function_start_entry.addr = text_section_sp->GetFileAddress(); 2463 uint64_t delta; 2464 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2465 0) { 2466 // Now append the current entry 2467 function_start_entry.addr += delta; 2468 if (is_arm) { 2469 if (function_start_entry.addr & 1) { 2470 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2471 function_start_entry.data = true; 2472 } else if (always_thumb) { 2473 function_start_entry.data = true; 2474 } 2475 } 2476 function_starts.Append(function_start_entry); 2477 } 2478 } else { 2479 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2480 // load command claiming an eh_frame but it doesn't actually have the 2481 // eh_frame content. And if we have a dSYM, we don't need to do any of 2482 // this fill-in-the-missing-symbols works anyway - the debug info should 2483 // give us all the functions in the module. 2484 if (text_section_sp.get() && eh_frame_section_sp.get() && 2485 m_type != eTypeDebugInfo) { 2486 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2487 DWARFCallFrameInfo::EH); 2488 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2489 eh_frame.GetFunctionAddressAndSizeVector(functions); 2490 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2491 size_t count = functions.GetSize(); 2492 for (size_t i = 0; i < count; ++i) { 2493 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2494 functions.GetEntryAtIndex(i); 2495 if (func) { 2496 FunctionStarts::Entry function_start_entry; 2497 function_start_entry.addr = func->base - text_base_addr; 2498 if (is_arm) { 2499 if (function_start_entry.addr & 1) { 2500 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2501 function_start_entry.data = true; 2502 } else if (always_thumb) { 2503 function_start_entry.data = true; 2504 } 2505 } 2506 function_starts.Append(function_start_entry); 2507 } 2508 } 2509 } 2510 } 2511 2512 const size_t function_starts_count = function_starts.GetSize(); 2513 2514 // For user process binaries (executables, dylibs, frameworks, bundles), if 2515 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2516 // going to assume the binary has been stripped. Don't allow assembly 2517 // language instruction emulation because we don't know proper function 2518 // start boundaries. 2519 // 2520 // For all other types of binaries (kernels, stand-alone bare board 2521 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2522 // sections - we should not make any assumptions about them based on that. 2523 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2524 m_allow_assembly_emulation_unwind_plans = false; 2525 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2526 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2527 2528 if (unwind_or_symbol_log) 2529 module_sp->LogMessage( 2530 unwind_or_symbol_log, 2531 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2532 } 2533 2534 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() 2535 ? eh_frame_section_sp->GetID() 2536 : static_cast<user_id_t>(NO_SECT); 2537 2538 lldb::offset_t nlist_data_offset = 0; 2539 2540 uint32_t N_SO_index = UINT32_MAX; 2541 2542 MachSymtabSectionInfo section_info(section_list); 2543 std::vector<uint32_t> N_FUN_indexes; 2544 std::vector<uint32_t> N_NSYM_indexes; 2545 std::vector<uint32_t> N_INCL_indexes; 2546 std::vector<uint32_t> N_BRAC_indexes; 2547 std::vector<uint32_t> N_COMM_indexes; 2548 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2549 typedef llvm::DenseMap<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2550 typedef llvm::DenseMap<const char *, uint32_t> ConstNameToSymbolIndexMap; 2551 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2552 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2553 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2554 // Any symbols that get merged into another will get an entry in this map 2555 // so we know 2556 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2557 uint32_t nlist_idx = 0; 2558 Symbol *symbol_ptr = nullptr; 2559 2560 uint32_t sym_idx = 0; 2561 Symbol *sym = nullptr; 2562 size_t num_syms = 0; 2563 std::string memory_symbol_name; 2564 uint32_t unmapped_local_symbols_found = 0; 2565 2566 std::vector<TrieEntryWithOffset> reexport_trie_entries; 2567 std::vector<TrieEntryWithOffset> external_sym_trie_entries; 2568 std::set<lldb::addr_t> resolver_addresses; 2569 2570 if (dyld_trie_data.GetByteSize() > 0) { 2571 ConstString text_segment_name("__TEXT"); 2572 SectionSP text_segment_sp = 2573 GetSectionList()->FindSectionByName(text_segment_name); 2574 lldb::addr_t text_segment_file_addr = LLDB_INVALID_ADDRESS; 2575 if (text_segment_sp) 2576 text_segment_file_addr = text_segment_sp->GetFileAddress(); 2577 std::vector<llvm::StringRef> nameSlices; 2578 ParseTrieEntries(dyld_trie_data, 0, is_arm, text_segment_file_addr, 2579 nameSlices, resolver_addresses, reexport_trie_entries, 2580 external_sym_trie_entries); 2581 } 2582 2583 typedef std::set<ConstString> IndirectSymbols; 2584 IndirectSymbols indirect_symbol_names; 2585 2586 #if TARGET_OS_IPHONE 2587 2588 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2589 // optimized by moving LOCAL symbols out of the memory mapped portion of 2590 // the DSC. The symbol information has all been retained, but it isn't 2591 // available in the normal nlist data. However, there *are* duplicate 2592 // entries of *some* 2593 // LOCAL symbols in the normal nlist data. To handle this situation 2594 // correctly, we must first attempt 2595 // to parse any DSC unmapped symbol information. If we find any, we set a 2596 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2597 2598 if (m_header.flags & MH_DYLIB_IN_CACHE) { 2599 // Before we can start mapping the DSC, we need to make certain the 2600 // target process is actually using the cache we can find. 2601 2602 // Next we need to determine the correct path for the dyld shared cache. 2603 2604 ArchSpec header_arch = GetArchitecture(); 2605 char dsc_path[PATH_MAX]; 2606 char dsc_path_development[PATH_MAX]; 2607 2608 snprintf( 2609 dsc_path, sizeof(dsc_path), "%s%s%s", 2610 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2611 */ 2612 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2613 header_arch.GetArchitectureName()); 2614 2615 snprintf( 2616 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2617 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2618 */ 2619 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2620 header_arch.GetArchitectureName(), ".development"); 2621 2622 FileSpec dsc_nondevelopment_filespec(dsc_path); 2623 FileSpec dsc_development_filespec(dsc_path_development); 2624 FileSpec dsc_filespec; 2625 2626 UUID dsc_uuid; 2627 UUID process_shared_cache_uuid; 2628 addr_t process_shared_cache_base_addr; 2629 2630 if (process) { 2631 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, 2632 process_shared_cache_uuid); 2633 } 2634 2635 // First see if we can find an exact match for the inferior process 2636 // shared cache UUID in the development or non-development shared caches 2637 // on disk. 2638 if (process_shared_cache_uuid.IsValid()) { 2639 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2640 UUID dsc_development_uuid = GetSharedCacheUUID( 2641 dsc_development_filespec, byte_order, addr_byte_size); 2642 if (dsc_development_uuid.IsValid() && 2643 dsc_development_uuid == process_shared_cache_uuid) { 2644 dsc_filespec = dsc_development_filespec; 2645 dsc_uuid = dsc_development_uuid; 2646 } 2647 } 2648 if (!dsc_uuid.IsValid() && 2649 FileSystem::Instance().Exists(dsc_nondevelopment_filespec)) { 2650 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2651 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2652 if (dsc_nondevelopment_uuid.IsValid() && 2653 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2654 dsc_filespec = dsc_nondevelopment_filespec; 2655 dsc_uuid = dsc_nondevelopment_uuid; 2656 } 2657 } 2658 } 2659 2660 // Failing a UUID match, prefer the development dyld_shared cache if both 2661 // are present. 2662 if (!FileSystem::Instance().Exists(dsc_filespec)) { 2663 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2664 dsc_filespec = dsc_development_filespec; 2665 } else { 2666 dsc_filespec = dsc_nondevelopment_filespec; 2667 } 2668 } 2669 2670 /* The dyld_cache_header has a pointer to the 2671 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2672 The dyld_cache_local_symbols_info structure gives us three things: 2673 1. The start and count of the nlist records in the dyld_shared_cache 2674 file 2675 2. The start and size of the strings for these nlist records 2676 3. The start and count of dyld_cache_local_symbols_entry entries 2677 2678 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2679 dyld shared cache. 2680 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2681 the dyld shared cache. 2682 The dyld_cache_local_symbols_entry also lists the start of this 2683 dylib/framework's nlist records 2684 and the count of how many nlist records there are for this 2685 dylib/framework. 2686 */ 2687 2688 // Process the dyld shared cache header to find the unmapped symbols 2689 2690 DataBufferSP dsc_data_sp = MapFileData( 2691 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2692 if (!dsc_uuid.IsValid()) { 2693 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2694 } 2695 if (dsc_data_sp) { 2696 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2697 2698 bool uuid_match = true; 2699 if (dsc_uuid.IsValid() && process) { 2700 if (process_shared_cache_uuid.IsValid() && 2701 dsc_uuid != process_shared_cache_uuid) { 2702 // The on-disk dyld_shared_cache file is not the same as the one in 2703 // this process' memory, don't use it. 2704 uuid_match = false; 2705 ModuleSP module_sp(GetModule()); 2706 if (module_sp) 2707 module_sp->ReportWarning("process shared cache does not match " 2708 "on-disk dyld_shared_cache file, some " 2709 "symbol names will be missing."); 2710 } 2711 } 2712 2713 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2714 2715 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2716 2717 // If the mappingOffset points to a location inside the header, we've 2718 // opened an old dyld shared cache, and should not proceed further. 2719 if (uuid_match && 2720 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2721 2722 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2723 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2724 mappingOffset); 2725 2726 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2727 byte_order, addr_byte_size); 2728 offset = 0; 2729 2730 // The File addresses (from the in-memory Mach-O load commands) for 2731 // the shared libraries in the shared library cache need to be 2732 // adjusted by an offset to match up with the dylibOffset identifying 2733 // field in the dyld_cache_local_symbol_entry's. This offset is 2734 // recorded in mapping_offset_value. 2735 const uint64_t mapping_offset_value = 2736 dsc_mapping_info_data.GetU64(&offset); 2737 2738 offset = 2739 offsetof(struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2740 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2741 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2742 2743 if (localSymbolsOffset && localSymbolsSize) { 2744 // Map the local symbols 2745 DataBufferSP dsc_local_symbols_data_sp = 2746 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2747 2748 if (dsc_local_symbols_data_sp) { 2749 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2750 byte_order, addr_byte_size); 2751 2752 offset = 0; 2753 2754 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 2755 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 2756 UndefinedNameToDescMap undefined_name_to_desc; 2757 SymbolIndexToName reexport_shlib_needs_fixup; 2758 2759 // Read the local_symbols_infos struct in one shot 2760 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2761 dsc_local_symbols_data.GetU32(&offset, 2762 &local_symbols_info.nlistOffset, 6); 2763 2764 SectionSP text_section_sp( 2765 section_list->FindSectionByName(GetSegmentNameTEXT())); 2766 2767 uint32_t header_file_offset = 2768 (text_section_sp->GetFileAddress() - mapping_offset_value); 2769 2770 offset = local_symbols_info.entriesOffset; 2771 for (uint32_t entry_index = 0; 2772 entry_index < local_symbols_info.entriesCount; entry_index++) { 2773 struct lldb_copy_dyld_cache_local_symbols_entry 2774 local_symbols_entry; 2775 local_symbols_entry.dylibOffset = 2776 dsc_local_symbols_data.GetU32(&offset); 2777 local_symbols_entry.nlistStartIndex = 2778 dsc_local_symbols_data.GetU32(&offset); 2779 local_symbols_entry.nlistCount = 2780 dsc_local_symbols_data.GetU32(&offset); 2781 2782 if (header_file_offset == local_symbols_entry.dylibOffset) { 2783 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2784 2785 // The normal nlist code cannot correctly size the Symbols 2786 // array, we need to allocate it here. 2787 sym = symtab->Resize( 2788 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2789 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2790 num_syms = symtab->GetNumSymbols(); 2791 2792 nlist_data_offset = 2793 local_symbols_info.nlistOffset + 2794 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2795 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2796 2797 for (uint32_t nlist_index = 0; 2798 nlist_index < local_symbols_entry.nlistCount; 2799 nlist_index++) { 2800 ///////////////////////////// 2801 { 2802 llvm::Optional<struct nlist_64> nlist_maybe = 2803 ParseNList(dsc_local_symbols_data, nlist_data_offset, 2804 nlist_byte_size); 2805 if (!nlist_maybe) 2806 break; 2807 struct nlist_64 nlist = *nlist_maybe; 2808 2809 SymbolType type = eSymbolTypeInvalid; 2810 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2811 string_table_offset + nlist.n_strx); 2812 2813 if (symbol_name == NULL) { 2814 // No symbol should be NULL, even the symbols with no 2815 // string values should have an offset zero which 2816 // points to an empty C-string 2817 Host::SystemLog( 2818 Host::eSystemLogError, 2819 "error: DSC unmapped local symbol[%u] has invalid " 2820 "string table offset 0x%x in %s, ignoring symbol\n", 2821 entry_index, nlist.n_strx, 2822 module_sp->GetFileSpec().GetPath().c_str()); 2823 continue; 2824 } 2825 if (symbol_name[0] == '\0') 2826 symbol_name = NULL; 2827 2828 const char *symbol_name_non_abi_mangled = NULL; 2829 2830 SectionSP symbol_section; 2831 uint32_t symbol_byte_size = 0; 2832 bool add_nlist = true; 2833 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2834 bool demangled_is_synthesized = false; 2835 bool is_gsym = false; 2836 bool set_value = true; 2837 2838 assert(sym_idx < num_syms); 2839 2840 sym[sym_idx].SetDebug(is_debug); 2841 2842 if (is_debug) { 2843 switch (nlist.n_type) { 2844 case N_GSYM: 2845 // global symbol: name,,NO_SECT,type,0 2846 // Sometimes the N_GSYM value contains the address. 2847 2848 // FIXME: In the .o files, we have a GSYM and a debug 2849 // symbol for all the ObjC data. They 2850 // have the same address, but we want to ensure that 2851 // we always find only the real symbol, 'cause we 2852 // don't currently correctly attribute the 2853 // GSYM one to the ObjCClass/Ivar/MetaClass 2854 // symbol type. This is a temporary hack to make 2855 // sure the ObjectiveC symbols get treated correctly. 2856 // To do this right, we should coalesce all the GSYM 2857 // & global symbols that have the same address. 2858 2859 is_gsym = true; 2860 sym[sym_idx].SetExternal(true); 2861 2862 if (symbol_name && symbol_name[0] == '_' && 2863 symbol_name[1] == 'O') { 2864 llvm::StringRef symbol_name_ref(symbol_name); 2865 if (symbol_name_ref.startswith( 2866 g_objc_v2_prefix_class)) { 2867 symbol_name_non_abi_mangled = symbol_name + 1; 2868 symbol_name = 2869 symbol_name + g_objc_v2_prefix_class.size(); 2870 type = eSymbolTypeObjCClass; 2871 demangled_is_synthesized = true; 2872 2873 } else if (symbol_name_ref.startswith( 2874 g_objc_v2_prefix_metaclass)) { 2875 symbol_name_non_abi_mangled = symbol_name + 1; 2876 symbol_name = 2877 symbol_name + g_objc_v2_prefix_metaclass.size(); 2878 type = eSymbolTypeObjCMetaClass; 2879 demangled_is_synthesized = true; 2880 } else if (symbol_name_ref.startswith( 2881 g_objc_v2_prefix_ivar)) { 2882 symbol_name_non_abi_mangled = symbol_name + 1; 2883 symbol_name = 2884 symbol_name + g_objc_v2_prefix_ivar.size(); 2885 type = eSymbolTypeObjCIVar; 2886 demangled_is_synthesized = true; 2887 } 2888 } else { 2889 if (nlist.n_value != 0) 2890 symbol_section = section_info.GetSection( 2891 nlist.n_sect, nlist.n_value); 2892 type = eSymbolTypeData; 2893 } 2894 break; 2895 2896 case N_FNAME: 2897 // procedure name (f77 kludge): name,,NO_SECT,0,0 2898 type = eSymbolTypeCompiler; 2899 break; 2900 2901 case N_FUN: 2902 // procedure: name,,n_sect,linenumber,address 2903 if (symbol_name) { 2904 type = eSymbolTypeCode; 2905 symbol_section = section_info.GetSection( 2906 nlist.n_sect, nlist.n_value); 2907 2908 N_FUN_addr_to_sym_idx.insert( 2909 std::make_pair(nlist.n_value, sym_idx)); 2910 // We use the current number of symbols in the 2911 // symbol table in lieu of using nlist_idx in case 2912 // we ever start trimming entries out 2913 N_FUN_indexes.push_back(sym_idx); 2914 } else { 2915 type = eSymbolTypeCompiler; 2916 2917 if (!N_FUN_indexes.empty()) { 2918 // Copy the size of the function into the 2919 // original 2920 // STAB entry so we don't have 2921 // to hunt for it later 2922 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2923 ->SetByteSize(nlist.n_value); 2924 N_FUN_indexes.pop_back(); 2925 // We don't really need the end function STAB as 2926 // it contains the size which we already placed 2927 // with the original symbol, so don't add it if 2928 // we want a minimal symbol table 2929 add_nlist = false; 2930 } 2931 } 2932 break; 2933 2934 case N_STSYM: 2935 // static symbol: name,,n_sect,type,address 2936 N_STSYM_addr_to_sym_idx.insert( 2937 std::make_pair(nlist.n_value, sym_idx)); 2938 symbol_section = section_info.GetSection(nlist.n_sect, 2939 nlist.n_value); 2940 if (symbol_name && symbol_name[0]) { 2941 type = ObjectFile::GetSymbolTypeFromName( 2942 symbol_name + 1, eSymbolTypeData); 2943 } 2944 break; 2945 2946 case N_LCSYM: 2947 // .lcomm symbol: name,,n_sect,type,address 2948 symbol_section = section_info.GetSection(nlist.n_sect, 2949 nlist.n_value); 2950 type = eSymbolTypeCommonBlock; 2951 break; 2952 2953 case N_BNSYM: 2954 // We use the current number of symbols in the symbol 2955 // table in lieu of using nlist_idx in case we ever 2956 // start trimming entries out Skip these if we want 2957 // minimal symbol tables 2958 add_nlist = false; 2959 break; 2960 2961 case N_ENSYM: 2962 // Set the size of the N_BNSYM to the terminating 2963 // index of this N_ENSYM so that we can always skip 2964 // the entire symbol if we need to navigate more 2965 // quickly at the source level when parsing STABS 2966 // Skip these if we want minimal symbol tables 2967 add_nlist = false; 2968 break; 2969 2970 case N_OPT: 2971 // emitted with gcc2_compiled and in gcc source 2972 type = eSymbolTypeCompiler; 2973 break; 2974 2975 case N_RSYM: 2976 // register sym: name,,NO_SECT,type,register 2977 type = eSymbolTypeVariable; 2978 break; 2979 2980 case N_SLINE: 2981 // src line: 0,,n_sect,linenumber,address 2982 symbol_section = section_info.GetSection(nlist.n_sect, 2983 nlist.n_value); 2984 type = eSymbolTypeLineEntry; 2985 break; 2986 2987 case N_SSYM: 2988 // structure elt: name,,NO_SECT,type,struct_offset 2989 type = eSymbolTypeVariableType; 2990 break; 2991 2992 case N_SO: 2993 // source file name 2994 type = eSymbolTypeSourceFile; 2995 if (symbol_name == NULL) { 2996 add_nlist = false; 2997 if (N_SO_index != UINT32_MAX) { 2998 // Set the size of the N_SO to the terminating 2999 // index of this N_SO so that we can always skip 3000 // the entire N_SO if we need to navigate more 3001 // quickly at the source level when parsing STABS 3002 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3003 symbol_ptr->SetByteSize(sym_idx); 3004 symbol_ptr->SetSizeIsSibling(true); 3005 } 3006 N_NSYM_indexes.clear(); 3007 N_INCL_indexes.clear(); 3008 N_BRAC_indexes.clear(); 3009 N_COMM_indexes.clear(); 3010 N_FUN_indexes.clear(); 3011 N_SO_index = UINT32_MAX; 3012 } else { 3013 // We use the current number of symbols in the 3014 // symbol table in lieu of using nlist_idx in case 3015 // we ever start trimming entries out 3016 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3017 if (N_SO_has_full_path) { 3018 if ((N_SO_index == sym_idx - 1) && 3019 ((sym_idx - 1) < num_syms)) { 3020 // We have two consecutive N_SO entries where 3021 // the first contains a directory and the 3022 // second contains a full path. 3023 sym[sym_idx - 1].GetMangled().SetValue( 3024 ConstString(symbol_name), false); 3025 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3026 add_nlist = false; 3027 } else { 3028 // This is the first entry in a N_SO that 3029 // contains a directory or 3030 // a full path to the source file 3031 N_SO_index = sym_idx; 3032 } 3033 } else if ((N_SO_index == sym_idx - 1) && 3034 ((sym_idx - 1) < num_syms)) { 3035 // This is usually the second N_SO entry that 3036 // contains just the filename, so here we combine 3037 // it with the first one if we are minimizing the 3038 // symbol table 3039 const char *so_path = sym[sym_idx - 1] 3040 .GetMangled() 3041 .GetDemangledName() 3042 .AsCString(); 3043 if (so_path && so_path[0]) { 3044 std::string full_so_path(so_path); 3045 const size_t double_slash_pos = 3046 full_so_path.find("//"); 3047 if (double_slash_pos != std::string::npos) { 3048 // The linker has been generating bad N_SO 3049 // entries with doubled up paths 3050 // in the format "%s%s" where the first 3051 // string in the DW_AT_comp_dir, and the 3052 // second is the directory for the source 3053 // file so you end up with a path that looks 3054 // like "/tmp/src//tmp/src/" 3055 FileSpec so_dir(so_path); 3056 if (!FileSystem::Instance().Exists(so_dir)) { 3057 so_dir.SetFile( 3058 &full_so_path[double_slash_pos + 1], 3059 FileSpec::Style::native); 3060 if (FileSystem::Instance().Exists(so_dir)) { 3061 // Trim off the incorrect path 3062 full_so_path.erase(0, double_slash_pos + 1); 3063 } 3064 } 3065 } 3066 if (*full_so_path.rbegin() != '/') 3067 full_so_path += '/'; 3068 full_so_path += symbol_name; 3069 sym[sym_idx - 1].GetMangled().SetValue( 3070 ConstString(full_so_path.c_str()), false); 3071 add_nlist = false; 3072 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3073 } 3074 } else { 3075 // This could be a relative path to a N_SO 3076 N_SO_index = sym_idx; 3077 } 3078 } 3079 break; 3080 3081 case N_OSO: 3082 // object file name: name,,0,0,st_mtime 3083 type = eSymbolTypeObjectFile; 3084 break; 3085 3086 case N_LSYM: 3087 // local sym: name,,NO_SECT,type,offset 3088 type = eSymbolTypeLocal; 3089 break; 3090 3091 // INCL scopes 3092 case N_BINCL: 3093 // include file beginning: name,,NO_SECT,0,sum We use 3094 // the current number of symbols in the symbol table 3095 // in lieu of using nlist_idx in case we ever start 3096 // trimming entries out 3097 N_INCL_indexes.push_back(sym_idx); 3098 type = eSymbolTypeScopeBegin; 3099 break; 3100 3101 case N_EINCL: 3102 // include file end: name,,NO_SECT,0,0 3103 // Set the size of the N_BINCL to the terminating 3104 // index of this N_EINCL so that we can always skip 3105 // the entire symbol if we need to navigate more 3106 // quickly at the source level when parsing STABS 3107 if (!N_INCL_indexes.empty()) { 3108 symbol_ptr = 3109 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3110 symbol_ptr->SetByteSize(sym_idx + 1); 3111 symbol_ptr->SetSizeIsSibling(true); 3112 N_INCL_indexes.pop_back(); 3113 } 3114 type = eSymbolTypeScopeEnd; 3115 break; 3116 3117 case N_SOL: 3118 // #included file name: name,,n_sect,0,address 3119 type = eSymbolTypeHeaderFile; 3120 3121 // We currently don't use the header files on darwin 3122 add_nlist = false; 3123 break; 3124 3125 case N_PARAMS: 3126 // compiler parameters: name,,NO_SECT,0,0 3127 type = eSymbolTypeCompiler; 3128 break; 3129 3130 case N_VERSION: 3131 // compiler version: name,,NO_SECT,0,0 3132 type = eSymbolTypeCompiler; 3133 break; 3134 3135 case N_OLEVEL: 3136 // compiler -O level: name,,NO_SECT,0,0 3137 type = eSymbolTypeCompiler; 3138 break; 3139 3140 case N_PSYM: 3141 // parameter: name,,NO_SECT,type,offset 3142 type = eSymbolTypeVariable; 3143 break; 3144 3145 case N_ENTRY: 3146 // alternate entry: name,,n_sect,linenumber,address 3147 symbol_section = section_info.GetSection(nlist.n_sect, 3148 nlist.n_value); 3149 type = eSymbolTypeLineEntry; 3150 break; 3151 3152 // Left and Right Braces 3153 case N_LBRAC: 3154 // left bracket: 0,,NO_SECT,nesting level,address We 3155 // use the current number of symbols in the symbol 3156 // table in lieu of using nlist_idx in case we ever 3157 // start trimming entries out 3158 symbol_section = section_info.GetSection(nlist.n_sect, 3159 nlist.n_value); 3160 N_BRAC_indexes.push_back(sym_idx); 3161 type = eSymbolTypeScopeBegin; 3162 break; 3163 3164 case N_RBRAC: 3165 // right bracket: 0,,NO_SECT,nesting level,address 3166 // Set the size of the N_LBRAC to the terminating 3167 // index of this N_RBRAC so that we can always skip 3168 // the entire symbol if we need to navigate more 3169 // quickly at the source level when parsing STABS 3170 symbol_section = section_info.GetSection(nlist.n_sect, 3171 nlist.n_value); 3172 if (!N_BRAC_indexes.empty()) { 3173 symbol_ptr = 3174 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3175 symbol_ptr->SetByteSize(sym_idx + 1); 3176 symbol_ptr->SetSizeIsSibling(true); 3177 N_BRAC_indexes.pop_back(); 3178 } 3179 type = eSymbolTypeScopeEnd; 3180 break; 3181 3182 case N_EXCL: 3183 // deleted include file: name,,NO_SECT,0,sum 3184 type = eSymbolTypeHeaderFile; 3185 break; 3186 3187 // COMM scopes 3188 case N_BCOMM: 3189 // begin common: name,,NO_SECT,0,0 3190 // We use the current number of symbols in the symbol 3191 // table in lieu of using nlist_idx in case we ever 3192 // start trimming entries out 3193 type = eSymbolTypeScopeBegin; 3194 N_COMM_indexes.push_back(sym_idx); 3195 break; 3196 3197 case N_ECOML: 3198 // end common (local name): 0,,n_sect,0,address 3199 symbol_section = section_info.GetSection(nlist.n_sect, 3200 nlist.n_value); 3201 // Fall through 3202 3203 case N_ECOMM: 3204 // end common: name,,n_sect,0,0 3205 // Set the size of the N_BCOMM to the terminating 3206 // index of this N_ECOMM/N_ECOML so that we can 3207 // always skip the entire symbol if we need to 3208 // navigate more quickly at the source level when 3209 // parsing STABS 3210 if (!N_COMM_indexes.empty()) { 3211 symbol_ptr = 3212 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3213 symbol_ptr->SetByteSize(sym_idx + 1); 3214 symbol_ptr->SetSizeIsSibling(true); 3215 N_COMM_indexes.pop_back(); 3216 } 3217 type = eSymbolTypeScopeEnd; 3218 break; 3219 3220 case N_LENG: 3221 // second stab entry with length information 3222 type = eSymbolTypeAdditional; 3223 break; 3224 3225 default: 3226 break; 3227 } 3228 } else { 3229 // uint8_t n_pext = N_PEXT & nlist.n_type; 3230 uint8_t n_type = N_TYPE & nlist.n_type; 3231 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3232 3233 switch (n_type) { 3234 case N_INDR: { 3235 const char *reexport_name_cstr = 3236 strtab_data.PeekCStr(nlist.n_value); 3237 if (reexport_name_cstr && reexport_name_cstr[0]) { 3238 type = eSymbolTypeReExported; 3239 ConstString reexport_name( 3240 reexport_name_cstr + 3241 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3242 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3243 set_value = false; 3244 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3245 indirect_symbol_names.insert(ConstString( 3246 symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3247 } else 3248 type = eSymbolTypeUndefined; 3249 } break; 3250 3251 case N_UNDF: 3252 if (symbol_name && symbol_name[0]) { 3253 ConstString undefined_name( 3254 symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3255 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3256 } 3257 // Fall through 3258 case N_PBUD: 3259 type = eSymbolTypeUndefined; 3260 break; 3261 3262 case N_ABS: 3263 type = eSymbolTypeAbsolute; 3264 break; 3265 3266 case N_SECT: { 3267 symbol_section = section_info.GetSection(nlist.n_sect, 3268 nlist.n_value); 3269 3270 if (symbol_section == NULL) { 3271 // TODO: warn about this? 3272 add_nlist = false; 3273 break; 3274 } 3275 3276 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3277 type = eSymbolTypeException; 3278 } else { 3279 uint32_t section_type = 3280 symbol_section->Get() & SECTION_TYPE; 3281 3282 switch (section_type) { 3283 case S_CSTRING_LITERALS: 3284 type = eSymbolTypeData; 3285 break; // section with only literal C strings 3286 case S_4BYTE_LITERALS: 3287 type = eSymbolTypeData; 3288 break; // section with only 4 byte literals 3289 case S_8BYTE_LITERALS: 3290 type = eSymbolTypeData; 3291 break; // section with only 8 byte literals 3292 case S_LITERAL_POINTERS: 3293 type = eSymbolTypeTrampoline; 3294 break; // section with only pointers to literals 3295 case S_NON_LAZY_SYMBOL_POINTERS: 3296 type = eSymbolTypeTrampoline; 3297 break; // section with only non-lazy symbol 3298 // pointers 3299 case S_LAZY_SYMBOL_POINTERS: 3300 type = eSymbolTypeTrampoline; 3301 break; // section with only lazy symbol pointers 3302 case S_SYMBOL_STUBS: 3303 type = eSymbolTypeTrampoline; 3304 break; // section with only symbol stubs, byte 3305 // size of stub in the reserved2 field 3306 case S_MOD_INIT_FUNC_POINTERS: 3307 type = eSymbolTypeCode; 3308 break; // section with only function pointers for 3309 // initialization 3310 case S_MOD_TERM_FUNC_POINTERS: 3311 type = eSymbolTypeCode; 3312 break; // section with only function pointers for 3313 // termination 3314 case S_INTERPOSING: 3315 type = eSymbolTypeTrampoline; 3316 break; // section with only pairs of function 3317 // pointers for interposing 3318 case S_16BYTE_LITERALS: 3319 type = eSymbolTypeData; 3320 break; // section with only 16 byte literals 3321 case S_DTRACE_DOF: 3322 type = eSymbolTypeInstrumentation; 3323 break; 3324 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3325 type = eSymbolTypeTrampoline; 3326 break; 3327 default: 3328 switch (symbol_section->GetType()) { 3329 case lldb::eSectionTypeCode: 3330 type = eSymbolTypeCode; 3331 break; 3332 case eSectionTypeData: 3333 case eSectionTypeDataCString: // Inlined C string 3334 // data 3335 case eSectionTypeDataCStringPointers: // Pointers 3336 // to C 3337 // string 3338 // data 3339 case eSectionTypeDataSymbolAddress: // Address of 3340 // a symbol in 3341 // the symbol 3342 // table 3343 case eSectionTypeData4: 3344 case eSectionTypeData8: 3345 case eSectionTypeData16: 3346 type = eSymbolTypeData; 3347 break; 3348 default: 3349 break; 3350 } 3351 break; 3352 } 3353 3354 if (type == eSymbolTypeInvalid) { 3355 const char *symbol_sect_name = 3356 symbol_section->GetName().AsCString(); 3357 if (symbol_section->IsDescendant( 3358 text_section_sp.get())) { 3359 if (symbol_section->IsClear( 3360 S_ATTR_PURE_INSTRUCTIONS | 3361 S_ATTR_SELF_MODIFYING_CODE | 3362 S_ATTR_SOME_INSTRUCTIONS)) 3363 type = eSymbolTypeData; 3364 else 3365 type = eSymbolTypeCode; 3366 } else if (symbol_section->IsDescendant( 3367 data_section_sp.get()) || 3368 symbol_section->IsDescendant( 3369 data_dirty_section_sp.get()) || 3370 symbol_section->IsDescendant( 3371 data_const_section_sp.get())) { 3372 if (symbol_sect_name && 3373 ::strstr(symbol_sect_name, "__objc") == 3374 symbol_sect_name) { 3375 type = eSymbolTypeRuntime; 3376 3377 if (symbol_name) { 3378 llvm::StringRef symbol_name_ref(symbol_name); 3379 if (symbol_name_ref.startswith("_OBJC_")) { 3380 llvm::StringRef 3381 g_objc_v2_prefix_class( 3382 "_OBJC_CLASS_$_"); 3383 llvm::StringRef 3384 g_objc_v2_prefix_metaclass( 3385 "_OBJC_METACLASS_$_"); 3386 llvm::StringRef 3387 g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 3388 if (symbol_name_ref.startswith( 3389 g_objc_v2_prefix_class)) { 3390 symbol_name_non_abi_mangled = 3391 symbol_name + 1; 3392 symbol_name = 3393 symbol_name + 3394 g_objc_v2_prefix_class.size(); 3395 type = eSymbolTypeObjCClass; 3396 demangled_is_synthesized = true; 3397 } else if ( 3398 symbol_name_ref.startswith( 3399 g_objc_v2_prefix_metaclass)) { 3400 symbol_name_non_abi_mangled = 3401 symbol_name + 1; 3402 symbol_name = 3403 symbol_name + 3404 g_objc_v2_prefix_metaclass.size(); 3405 type = eSymbolTypeObjCMetaClass; 3406 demangled_is_synthesized = true; 3407 } else if (symbol_name_ref.startswith( 3408 g_objc_v2_prefix_ivar)) { 3409 symbol_name_non_abi_mangled = 3410 symbol_name + 1; 3411 symbol_name = 3412 symbol_name + 3413 g_objc_v2_prefix_ivar.size(); 3414 type = eSymbolTypeObjCIVar; 3415 demangled_is_synthesized = true; 3416 } 3417 } 3418 } 3419 } else if (symbol_sect_name && 3420 ::strstr(symbol_sect_name, 3421 "__gcc_except_tab") == 3422 symbol_sect_name) { 3423 type = eSymbolTypeException; 3424 } else { 3425 type = eSymbolTypeData; 3426 } 3427 } else if (symbol_sect_name && 3428 ::strstr(symbol_sect_name, "__IMPORT") == 3429 symbol_sect_name) { 3430 type = eSymbolTypeTrampoline; 3431 } else if (symbol_section->IsDescendant( 3432 objc_section_sp.get())) { 3433 type = eSymbolTypeRuntime; 3434 if (symbol_name && symbol_name[0] == '.') { 3435 llvm::StringRef symbol_name_ref(symbol_name); 3436 llvm::StringRef 3437 g_objc_v1_prefix_class(".objc_class_name_"); 3438 if (symbol_name_ref.startswith( 3439 g_objc_v1_prefix_class)) { 3440 symbol_name_non_abi_mangled = symbol_name; 3441 symbol_name = symbol_name + 3442 g_objc_v1_prefix_class.size(); 3443 type = eSymbolTypeObjCClass; 3444 demangled_is_synthesized = true; 3445 } 3446 } 3447 } 3448 } 3449 } 3450 } break; 3451 } 3452 } 3453 3454 if (add_nlist) { 3455 uint64_t symbol_value = nlist.n_value; 3456 if (symbol_name_non_abi_mangled) { 3457 sym[sym_idx].GetMangled().SetMangledName( 3458 ConstString(symbol_name_non_abi_mangled)); 3459 sym[sym_idx].GetMangled().SetDemangledName( 3460 ConstString(symbol_name)); 3461 } else { 3462 bool symbol_name_is_mangled = false; 3463 3464 if (symbol_name && symbol_name[0] == '_') { 3465 symbol_name_is_mangled = symbol_name[1] == '_'; 3466 symbol_name++; // Skip the leading underscore 3467 } 3468 3469 if (symbol_name) { 3470 ConstString const_symbol_name(symbol_name); 3471 sym[sym_idx].GetMangled().SetValue( 3472 const_symbol_name, symbol_name_is_mangled); 3473 if (is_gsym && is_debug) { 3474 const char *gsym_name = 3475 sym[sym_idx] 3476 .GetMangled() 3477 .GetName(Mangled::ePreferMangled) 3478 .GetCString(); 3479 if (gsym_name) 3480 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3481 } 3482 } 3483 } 3484 if (symbol_section) { 3485 const addr_t section_file_addr = 3486 symbol_section->GetFileAddress(); 3487 if (symbol_byte_size == 0 && 3488 function_starts_count > 0) { 3489 addr_t symbol_lookup_file_addr = nlist.n_value; 3490 // Do an exact address match for non-ARM addresses, 3491 // else get the closest since the symbol might be a 3492 // thumb symbol which has an address with bit zero 3493 // set 3494 FunctionStarts::Entry *func_start_entry = 3495 function_starts.FindEntry(symbol_lookup_file_addr, 3496 !is_arm); 3497 if (is_arm && func_start_entry) { 3498 // Verify that the function start address is the 3499 // symbol address (ARM) or the symbol address + 1 3500 // (thumb) 3501 if (func_start_entry->addr != 3502 symbol_lookup_file_addr && 3503 func_start_entry->addr != 3504 (symbol_lookup_file_addr + 1)) { 3505 // Not the right entry, NULL it out... 3506 func_start_entry = NULL; 3507 } 3508 } 3509 if (func_start_entry) { 3510 func_start_entry->data = true; 3511 3512 addr_t symbol_file_addr = func_start_entry->addr; 3513 uint32_t symbol_flags = 0; 3514 if (is_arm) { 3515 if (symbol_file_addr & 1) 3516 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3517 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3518 } 3519 3520 const FunctionStarts::Entry *next_func_start_entry = 3521 function_starts.FindNextEntry(func_start_entry); 3522 const addr_t section_end_file_addr = 3523 section_file_addr + 3524 symbol_section->GetByteSize(); 3525 if (next_func_start_entry) { 3526 addr_t next_symbol_file_addr = 3527 next_func_start_entry->addr; 3528 // Be sure the clear the Thumb address bit when 3529 // we calculate the size from the current and 3530 // next address 3531 if (is_arm) 3532 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3533 symbol_byte_size = std::min<lldb::addr_t>( 3534 next_symbol_file_addr - symbol_file_addr, 3535 section_end_file_addr - symbol_file_addr); 3536 } else { 3537 symbol_byte_size = 3538 section_end_file_addr - symbol_file_addr; 3539 } 3540 } 3541 } 3542 symbol_value -= section_file_addr; 3543 } 3544 3545 if (is_debug == false) { 3546 if (type == eSymbolTypeCode) { 3547 // See if we can find a N_FUN entry for any code 3548 // symbols. If we do find a match, and the name 3549 // matches, then we can merge the two into just the 3550 // function symbol to avoid duplicate entries in 3551 // the symbol table 3552 auto range = 3553 N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3554 if (range.first != range.second) { 3555 bool found_it = false; 3556 for (auto pos = range.first; pos != range.second; 3557 ++pos) { 3558 if (sym[sym_idx].GetMangled().GetName( 3559 Mangled::ePreferMangled) == 3560 sym[pos->second].GetMangled().GetName( 3561 Mangled::ePreferMangled)) { 3562 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3563 // We just need the flags from the linker 3564 // symbol, so put these flags 3565 // into the N_FUN flags to avoid duplicate 3566 // symbols in the symbol table 3567 sym[pos->second].SetExternal( 3568 sym[sym_idx].IsExternal()); 3569 sym[pos->second].SetFlags(nlist.n_type << 16 | 3570 nlist.n_desc); 3571 if (resolver_addresses.find(nlist.n_value) != 3572 resolver_addresses.end()) 3573 sym[pos->second].SetType(eSymbolTypeResolver); 3574 sym[sym_idx].Clear(); 3575 found_it = true; 3576 break; 3577 } 3578 } 3579 if (found_it) 3580 continue; 3581 } else { 3582 if (resolver_addresses.find(nlist.n_value) != 3583 resolver_addresses.end()) 3584 type = eSymbolTypeResolver; 3585 } 3586 } else if (type == eSymbolTypeData || 3587 type == eSymbolTypeObjCClass || 3588 type == eSymbolTypeObjCMetaClass || 3589 type == eSymbolTypeObjCIVar) { 3590 // See if we can find a N_STSYM entry for any data 3591 // symbols. If we do find a match, and the name 3592 // matches, then we can merge the two into just the 3593 // Static symbol to avoid duplicate entries in the 3594 // symbol table 3595 auto range = N_STSYM_addr_to_sym_idx.equal_range( 3596 nlist.n_value); 3597 if (range.first != range.second) { 3598 bool found_it = false; 3599 for (auto pos = range.first; pos != range.second; 3600 ++pos) { 3601 if (sym[sym_idx].GetMangled().GetName( 3602 Mangled::ePreferMangled) == 3603 sym[pos->second].GetMangled().GetName( 3604 Mangled::ePreferMangled)) { 3605 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3606 // We just need the flags from the linker 3607 // symbol, so put these flags 3608 // into the N_STSYM flags to avoid duplicate 3609 // symbols in the symbol table 3610 sym[pos->second].SetExternal( 3611 sym[sym_idx].IsExternal()); 3612 sym[pos->second].SetFlags(nlist.n_type << 16 | 3613 nlist.n_desc); 3614 sym[sym_idx].Clear(); 3615 found_it = true; 3616 break; 3617 } 3618 } 3619 if (found_it) 3620 continue; 3621 } else { 3622 const char *gsym_name = 3623 sym[sym_idx] 3624 .GetMangled() 3625 .GetName(Mangled::ePreferMangled) 3626 .GetCString(); 3627 if (gsym_name) { 3628 // Combine N_GSYM stab entries with the non 3629 // stab symbol 3630 ConstNameToSymbolIndexMap::const_iterator pos = 3631 N_GSYM_name_to_sym_idx.find(gsym_name); 3632 if (pos != N_GSYM_name_to_sym_idx.end()) { 3633 const uint32_t GSYM_sym_idx = pos->second; 3634 m_nlist_idx_to_sym_idx[nlist_idx] = 3635 GSYM_sym_idx; 3636 // Copy the address, because often the N_GSYM 3637 // address has an invalid address of zero 3638 // when the global is a common symbol 3639 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3640 symbol_section); 3641 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3642 symbol_value); 3643 symbols_added.insert(sym[GSYM_sym_idx] 3644 .GetAddress() 3645 .GetFileAddress()); 3646 // We just need the flags from the linker 3647 // symbol, so put these flags 3648 // into the N_GSYM flags to avoid duplicate 3649 // symbols in the symbol table 3650 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 3651 nlist.n_desc); 3652 sym[sym_idx].Clear(); 3653 continue; 3654 } 3655 } 3656 } 3657 } 3658 } 3659 3660 sym[sym_idx].SetID(nlist_idx); 3661 sym[sym_idx].SetType(type); 3662 if (set_value) { 3663 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 3664 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3665 symbols_added.insert( 3666 sym[sym_idx].GetAddress().GetFileAddress()); 3667 } 3668 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 3669 3670 if (symbol_byte_size > 0) 3671 sym[sym_idx].SetByteSize(symbol_byte_size); 3672 3673 if (demangled_is_synthesized) 3674 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3675 ++sym_idx; 3676 } else { 3677 sym[sym_idx].Clear(); 3678 } 3679 } 3680 ///////////////////////////// 3681 } 3682 break; // No more entries to consider 3683 } 3684 } 3685 3686 for (const auto &pos : reexport_shlib_needs_fixup) { 3687 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3688 if (undef_pos != undefined_name_to_desc.end()) { 3689 const uint8_t dylib_ordinal = 3690 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3691 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3692 sym[pos.first].SetReExportedSymbolSharedLibrary( 3693 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3694 } 3695 } 3696 } 3697 } 3698 } 3699 } 3700 } 3701 3702 // Must reset this in case it was mutated above! 3703 nlist_data_offset = 0; 3704 #endif 3705 3706 if (nlist_data.GetByteSize() > 0) { 3707 3708 // If the sym array was not created while parsing the DSC unmapped 3709 // symbols, create it now. 3710 if (sym == nullptr) { 3711 sym = 3712 symtab->Resize(symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3713 num_syms = symtab->GetNumSymbols(); 3714 } 3715 3716 if (unmapped_local_symbols_found) { 3717 assert(m_dysymtab.ilocalsym == 0); 3718 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3719 nlist_idx = m_dysymtab.nlocalsym; 3720 } else { 3721 nlist_idx = 0; 3722 } 3723 3724 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 3725 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 3726 UndefinedNameToDescMap undefined_name_to_desc; 3727 SymbolIndexToName reexport_shlib_needs_fixup; 3728 3729 // Symtab parsing is a huge mess. Everything is entangled and the code 3730 // requires access to a ridiculous amount of variables. LLDB depends 3731 // heavily on the proper merging of symbols and to get that right we need 3732 // to make sure we have parsed all the debug symbols first. Therefore we 3733 // invoke the lambda twice, once to parse only the debug symbols and then 3734 // once more to parse the remaining symbols. 3735 auto ParseSymbolLambda = [&](struct nlist_64 &nlist, uint32_t nlist_idx, 3736 bool debug_only) { 3737 const bool is_debug = ((nlist.n_type & N_STAB) != 0); 3738 if (is_debug != debug_only) 3739 return true; 3740 3741 const char *symbol_name_non_abi_mangled = nullptr; 3742 const char *symbol_name = nullptr; 3743 3744 if (have_strtab_data) { 3745 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3746 3747 if (symbol_name == nullptr) { 3748 // No symbol should be NULL, even the symbols with no string values 3749 // should have an offset zero which points to an empty C-string 3750 Host::SystemLog(Host::eSystemLogError, 3751 "error: symbol[%u] has invalid string table offset " 3752 "0x%x in %s, ignoring symbol\n", 3753 nlist_idx, nlist.n_strx, 3754 module_sp->GetFileSpec().GetPath().c_str()); 3755 return true; 3756 } 3757 if (symbol_name[0] == '\0') 3758 symbol_name = nullptr; 3759 } else { 3760 const addr_t str_addr = strtab_addr + nlist.n_strx; 3761 Status str_error; 3762 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3763 str_error)) 3764 symbol_name = memory_symbol_name.c_str(); 3765 } 3766 3767 SymbolType type = eSymbolTypeInvalid; 3768 SectionSP symbol_section; 3769 lldb::addr_t symbol_byte_size = 0; 3770 bool add_nlist = true; 3771 bool is_gsym = false; 3772 bool demangled_is_synthesized = false; 3773 bool set_value = true; 3774 3775 assert(sym_idx < num_syms); 3776 sym[sym_idx].SetDebug(is_debug); 3777 3778 if (is_debug) { 3779 switch (nlist.n_type) { 3780 case N_GSYM: 3781 // global symbol: name,,NO_SECT,type,0 3782 // Sometimes the N_GSYM value contains the address. 3783 3784 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3785 // the ObjC data. They 3786 // have the same address, but we want to ensure that we always find 3787 // only the real symbol, 'cause we don't currently correctly 3788 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3789 // type. This is a temporary hack to make sure the ObjectiveC 3790 // symbols get treated correctly. To do this right, we should 3791 // coalesce all the GSYM & global symbols that have the same 3792 // address. 3793 is_gsym = true; 3794 sym[sym_idx].SetExternal(true); 3795 3796 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3797 llvm::StringRef symbol_name_ref(symbol_name); 3798 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3799 symbol_name_non_abi_mangled = symbol_name + 1; 3800 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3801 type = eSymbolTypeObjCClass; 3802 demangled_is_synthesized = true; 3803 3804 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 3805 symbol_name_non_abi_mangled = symbol_name + 1; 3806 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3807 type = eSymbolTypeObjCMetaClass; 3808 demangled_is_synthesized = true; 3809 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3810 symbol_name_non_abi_mangled = symbol_name + 1; 3811 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3812 type = eSymbolTypeObjCIVar; 3813 demangled_is_synthesized = true; 3814 } 3815 } else { 3816 if (nlist.n_value != 0) 3817 symbol_section = 3818 section_info.GetSection(nlist.n_sect, nlist.n_value); 3819 type = eSymbolTypeData; 3820 } 3821 break; 3822 3823 case N_FNAME: 3824 // procedure name (f77 kludge): name,,NO_SECT,0,0 3825 type = eSymbolTypeCompiler; 3826 break; 3827 3828 case N_FUN: 3829 // procedure: name,,n_sect,linenumber,address 3830 if (symbol_name) { 3831 type = eSymbolTypeCode; 3832 symbol_section = 3833 section_info.GetSection(nlist.n_sect, nlist.n_value); 3834 3835 N_FUN_addr_to_sym_idx.insert( 3836 std::make_pair(nlist.n_value, sym_idx)); 3837 // We use the current number of symbols in the symbol table in 3838 // lieu of using nlist_idx in case we ever start trimming entries 3839 // out 3840 N_FUN_indexes.push_back(sym_idx); 3841 } else { 3842 type = eSymbolTypeCompiler; 3843 3844 if (!N_FUN_indexes.empty()) { 3845 // Copy the size of the function into the original STAB entry 3846 // so we don't have to hunt for it later 3847 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3848 ->SetByteSize(nlist.n_value); 3849 N_FUN_indexes.pop_back(); 3850 // We don't really need the end function STAB as it contains 3851 // the size which we already placed with the original symbol, 3852 // so don't add it if we want a minimal symbol table 3853 add_nlist = false; 3854 } 3855 } 3856 break; 3857 3858 case N_STSYM: 3859 // static symbol: name,,n_sect,type,address 3860 N_STSYM_addr_to_sym_idx.insert( 3861 std::make_pair(nlist.n_value, sym_idx)); 3862 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3863 if (symbol_name && symbol_name[0]) { 3864 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3865 eSymbolTypeData); 3866 } 3867 break; 3868 3869 case N_LCSYM: 3870 // .lcomm symbol: name,,n_sect,type,address 3871 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3872 type = eSymbolTypeCommonBlock; 3873 break; 3874 3875 case N_BNSYM: 3876 // We use the current number of symbols in the symbol table in lieu 3877 // of using nlist_idx in case we ever start trimming entries out 3878 // Skip these if we want minimal symbol tables 3879 add_nlist = false; 3880 break; 3881 3882 case N_ENSYM: 3883 // Set the size of the N_BNSYM to the terminating index of this 3884 // N_ENSYM so that we can always skip the entire symbol if we need 3885 // to navigate more quickly at the source level when parsing STABS 3886 // Skip these if we want minimal symbol tables 3887 add_nlist = false; 3888 break; 3889 3890 case N_OPT: 3891 // emitted with gcc2_compiled and in gcc source 3892 type = eSymbolTypeCompiler; 3893 break; 3894 3895 case N_RSYM: 3896 // register sym: name,,NO_SECT,type,register 3897 type = eSymbolTypeVariable; 3898 break; 3899 3900 case N_SLINE: 3901 // src line: 0,,n_sect,linenumber,address 3902 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3903 type = eSymbolTypeLineEntry; 3904 break; 3905 3906 case N_SSYM: 3907 // structure elt: name,,NO_SECT,type,struct_offset 3908 type = eSymbolTypeVariableType; 3909 break; 3910 3911 case N_SO: 3912 // source file name 3913 type = eSymbolTypeSourceFile; 3914 if (symbol_name == nullptr) { 3915 add_nlist = false; 3916 if (N_SO_index != UINT32_MAX) { 3917 // Set the size of the N_SO to the terminating index of this 3918 // N_SO so that we can always skip the entire N_SO if we need 3919 // to navigate more quickly at the source level when parsing 3920 // STABS 3921 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3922 symbol_ptr->SetByteSize(sym_idx); 3923 symbol_ptr->SetSizeIsSibling(true); 3924 } 3925 N_NSYM_indexes.clear(); 3926 N_INCL_indexes.clear(); 3927 N_BRAC_indexes.clear(); 3928 N_COMM_indexes.clear(); 3929 N_FUN_indexes.clear(); 3930 N_SO_index = UINT32_MAX; 3931 } else { 3932 // We use the current number of symbols in the symbol table in 3933 // lieu of using nlist_idx in case we ever start trimming entries 3934 // out 3935 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3936 if (N_SO_has_full_path) { 3937 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3938 // We have two consecutive N_SO entries where the first 3939 // contains a directory and the second contains a full path. 3940 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), 3941 false); 3942 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3943 add_nlist = false; 3944 } else { 3945 // This is the first entry in a N_SO that contains a 3946 // directory or a full path to the source file 3947 N_SO_index = sym_idx; 3948 } 3949 } else if ((N_SO_index == sym_idx - 1) && 3950 ((sym_idx - 1) < num_syms)) { 3951 // This is usually the second N_SO entry that contains just the 3952 // filename, so here we combine it with the first one if we are 3953 // minimizing the symbol table 3954 const char *so_path = 3955 sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3956 if (so_path && so_path[0]) { 3957 std::string full_so_path(so_path); 3958 const size_t double_slash_pos = full_so_path.find("//"); 3959 if (double_slash_pos != std::string::npos) { 3960 // The linker has been generating bad N_SO entries with 3961 // doubled up paths in the format "%s%s" where the first 3962 // string in the DW_AT_comp_dir, and the second is the 3963 // directory for the source file so you end up with a path 3964 // that looks like "/tmp/src//tmp/src/" 3965 FileSpec so_dir(so_path); 3966 if (!FileSystem::Instance().Exists(so_dir)) { 3967 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 3968 FileSpec::Style::native); 3969 if (FileSystem::Instance().Exists(so_dir)) { 3970 // Trim off the incorrect path 3971 full_so_path.erase(0, double_slash_pos + 1); 3972 } 3973 } 3974 } 3975 if (*full_so_path.rbegin() != '/') 3976 full_so_path += '/'; 3977 full_so_path += symbol_name; 3978 sym[sym_idx - 1].GetMangled().SetValue( 3979 ConstString(full_so_path.c_str()), false); 3980 add_nlist = false; 3981 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3982 } 3983 } else { 3984 // This could be a relative path to a N_SO 3985 N_SO_index = sym_idx; 3986 } 3987 } 3988 break; 3989 3990 case N_OSO: 3991 // object file name: name,,0,0,st_mtime 3992 type = eSymbolTypeObjectFile; 3993 break; 3994 3995 case N_LSYM: 3996 // local sym: name,,NO_SECT,type,offset 3997 type = eSymbolTypeLocal; 3998 break; 3999 4000 // INCL scopes 4001 case N_BINCL: 4002 // include file beginning: name,,NO_SECT,0,sum We use the current 4003 // number of symbols in the symbol table in lieu of using nlist_idx 4004 // in case we ever start trimming entries out 4005 N_INCL_indexes.push_back(sym_idx); 4006 type = eSymbolTypeScopeBegin; 4007 break; 4008 4009 case N_EINCL: 4010 // include file end: name,,NO_SECT,0,0 4011 // Set the size of the N_BINCL to the terminating index of this 4012 // N_EINCL so that we can always skip the entire symbol if we need 4013 // to navigate more quickly at the source level when parsing STABS 4014 if (!N_INCL_indexes.empty()) { 4015 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4016 symbol_ptr->SetByteSize(sym_idx + 1); 4017 symbol_ptr->SetSizeIsSibling(true); 4018 N_INCL_indexes.pop_back(); 4019 } 4020 type = eSymbolTypeScopeEnd; 4021 break; 4022 4023 case N_SOL: 4024 // #included file name: name,,n_sect,0,address 4025 type = eSymbolTypeHeaderFile; 4026 4027 // We currently don't use the header files on darwin 4028 add_nlist = false; 4029 break; 4030 4031 case N_PARAMS: 4032 // compiler parameters: name,,NO_SECT,0,0 4033 type = eSymbolTypeCompiler; 4034 break; 4035 4036 case N_VERSION: 4037 // compiler version: name,,NO_SECT,0,0 4038 type = eSymbolTypeCompiler; 4039 break; 4040 4041 case N_OLEVEL: 4042 // compiler -O level: name,,NO_SECT,0,0 4043 type = eSymbolTypeCompiler; 4044 break; 4045 4046 case N_PSYM: 4047 // parameter: name,,NO_SECT,type,offset 4048 type = eSymbolTypeVariable; 4049 break; 4050 4051 case N_ENTRY: 4052 // alternate entry: name,,n_sect,linenumber,address 4053 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4054 type = eSymbolTypeLineEntry; 4055 break; 4056 4057 // Left and Right Braces 4058 case N_LBRAC: 4059 // left bracket: 0,,NO_SECT,nesting level,address We use the 4060 // current number of symbols in the symbol table in lieu of using 4061 // nlist_idx in case we ever start trimming entries out 4062 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4063 N_BRAC_indexes.push_back(sym_idx); 4064 type = eSymbolTypeScopeBegin; 4065 break; 4066 4067 case N_RBRAC: 4068 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4069 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4070 // can always skip the entire symbol if we need to navigate more 4071 // quickly at the source level when parsing STABS 4072 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4073 if (!N_BRAC_indexes.empty()) { 4074 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4075 symbol_ptr->SetByteSize(sym_idx + 1); 4076 symbol_ptr->SetSizeIsSibling(true); 4077 N_BRAC_indexes.pop_back(); 4078 } 4079 type = eSymbolTypeScopeEnd; 4080 break; 4081 4082 case N_EXCL: 4083 // deleted include file: name,,NO_SECT,0,sum 4084 type = eSymbolTypeHeaderFile; 4085 break; 4086 4087 // COMM scopes 4088 case N_BCOMM: 4089 // begin common: name,,NO_SECT,0,0 4090 // We use the current number of symbols in the symbol table in lieu 4091 // of using nlist_idx in case we ever start trimming entries out 4092 type = eSymbolTypeScopeBegin; 4093 N_COMM_indexes.push_back(sym_idx); 4094 break; 4095 4096 case N_ECOML: 4097 // end common (local name): 0,,n_sect,0,address 4098 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4099 LLVM_FALLTHROUGH; 4100 4101 case N_ECOMM: 4102 // end common: name,,n_sect,0,0 4103 // Set the size of the N_BCOMM to the terminating index of this 4104 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4105 // we need to navigate more quickly at the source level when 4106 // parsing STABS 4107 if (!N_COMM_indexes.empty()) { 4108 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4109 symbol_ptr->SetByteSize(sym_idx + 1); 4110 symbol_ptr->SetSizeIsSibling(true); 4111 N_COMM_indexes.pop_back(); 4112 } 4113 type = eSymbolTypeScopeEnd; 4114 break; 4115 4116 case N_LENG: 4117 // second stab entry with length information 4118 type = eSymbolTypeAdditional; 4119 break; 4120 4121 default: 4122 break; 4123 } 4124 } else { 4125 uint8_t n_type = N_TYPE & nlist.n_type; 4126 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4127 4128 switch (n_type) { 4129 case N_INDR: { 4130 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4131 if (reexport_name_cstr && reexport_name_cstr[0]) { 4132 type = eSymbolTypeReExported; 4133 ConstString reexport_name(reexport_name_cstr + 4134 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4135 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4136 set_value = false; 4137 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4138 indirect_symbol_names.insert( 4139 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4140 } else 4141 type = eSymbolTypeUndefined; 4142 } break; 4143 4144 case N_UNDF: 4145 if (symbol_name && symbol_name[0]) { 4146 ConstString undefined_name(symbol_name + 4147 ((symbol_name[0] == '_') ? 1 : 0)); 4148 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4149 } 4150 LLVM_FALLTHROUGH; 4151 4152 case N_PBUD: 4153 type = eSymbolTypeUndefined; 4154 break; 4155 4156 case N_ABS: 4157 type = eSymbolTypeAbsolute; 4158 break; 4159 4160 case N_SECT: { 4161 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4162 4163 if (!symbol_section) { 4164 // TODO: warn about this? 4165 add_nlist = false; 4166 break; 4167 } 4168 4169 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4170 type = eSymbolTypeException; 4171 } else { 4172 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4173 4174 switch (section_type) { 4175 case S_CSTRING_LITERALS: 4176 type = eSymbolTypeData; 4177 break; // section with only literal C strings 4178 case S_4BYTE_LITERALS: 4179 type = eSymbolTypeData; 4180 break; // section with only 4 byte literals 4181 case S_8BYTE_LITERALS: 4182 type = eSymbolTypeData; 4183 break; // section with only 8 byte literals 4184 case S_LITERAL_POINTERS: 4185 type = eSymbolTypeTrampoline; 4186 break; // section with only pointers to literals 4187 case S_NON_LAZY_SYMBOL_POINTERS: 4188 type = eSymbolTypeTrampoline; 4189 break; // section with only non-lazy symbol pointers 4190 case S_LAZY_SYMBOL_POINTERS: 4191 type = eSymbolTypeTrampoline; 4192 break; // section with only lazy symbol pointers 4193 case S_SYMBOL_STUBS: 4194 type = eSymbolTypeTrampoline; 4195 break; // section with only symbol stubs, byte size of stub in 4196 // the reserved2 field 4197 case S_MOD_INIT_FUNC_POINTERS: 4198 type = eSymbolTypeCode; 4199 break; // section with only function pointers for initialization 4200 case S_MOD_TERM_FUNC_POINTERS: 4201 type = eSymbolTypeCode; 4202 break; // section with only function pointers for termination 4203 case S_INTERPOSING: 4204 type = eSymbolTypeTrampoline; 4205 break; // section with only pairs of function pointers for 4206 // interposing 4207 case S_16BYTE_LITERALS: 4208 type = eSymbolTypeData; 4209 break; // section with only 16 byte literals 4210 case S_DTRACE_DOF: 4211 type = eSymbolTypeInstrumentation; 4212 break; 4213 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4214 type = eSymbolTypeTrampoline; 4215 break; 4216 default: 4217 switch (symbol_section->GetType()) { 4218 case lldb::eSectionTypeCode: 4219 type = eSymbolTypeCode; 4220 break; 4221 case eSectionTypeData: 4222 case eSectionTypeDataCString: // Inlined C string data 4223 case eSectionTypeDataCStringPointers: // Pointers to C string 4224 // data 4225 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4226 // the symbol table 4227 case eSectionTypeData4: 4228 case eSectionTypeData8: 4229 case eSectionTypeData16: 4230 type = eSymbolTypeData; 4231 break; 4232 default: 4233 break; 4234 } 4235 break; 4236 } 4237 4238 if (type == eSymbolTypeInvalid) { 4239 const char *symbol_sect_name = 4240 symbol_section->GetName().AsCString(); 4241 if (symbol_section->IsDescendant(text_section_sp.get())) { 4242 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4243 S_ATTR_SELF_MODIFYING_CODE | 4244 S_ATTR_SOME_INSTRUCTIONS)) 4245 type = eSymbolTypeData; 4246 else 4247 type = eSymbolTypeCode; 4248 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 4249 symbol_section->IsDescendant( 4250 data_dirty_section_sp.get()) || 4251 symbol_section->IsDescendant( 4252 data_const_section_sp.get())) { 4253 if (symbol_sect_name && 4254 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 4255 type = eSymbolTypeRuntime; 4256 4257 if (symbol_name) { 4258 llvm::StringRef symbol_name_ref(symbol_name); 4259 if (symbol_name_ref.startswith("_OBJC_")) { 4260 llvm::StringRef g_objc_v2_prefix_class( 4261 "_OBJC_CLASS_$_"); 4262 llvm::StringRef g_objc_v2_prefix_metaclass( 4263 "_OBJC_METACLASS_$_"); 4264 llvm::StringRef g_objc_v2_prefix_ivar( 4265 "_OBJC_IVAR_$_"); 4266 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 4267 symbol_name_non_abi_mangled = symbol_name + 1; 4268 symbol_name = 4269 symbol_name + g_objc_v2_prefix_class.size(); 4270 type = eSymbolTypeObjCClass; 4271 demangled_is_synthesized = true; 4272 } else if (symbol_name_ref.startswith( 4273 g_objc_v2_prefix_metaclass)) { 4274 symbol_name_non_abi_mangled = symbol_name + 1; 4275 symbol_name = 4276 symbol_name + g_objc_v2_prefix_metaclass.size(); 4277 type = eSymbolTypeObjCMetaClass; 4278 demangled_is_synthesized = true; 4279 } else if (symbol_name_ref.startswith( 4280 g_objc_v2_prefix_ivar)) { 4281 symbol_name_non_abi_mangled = symbol_name + 1; 4282 symbol_name = 4283 symbol_name + g_objc_v2_prefix_ivar.size(); 4284 type = eSymbolTypeObjCIVar; 4285 demangled_is_synthesized = true; 4286 } 4287 } 4288 } 4289 } else if (symbol_sect_name && 4290 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4291 symbol_sect_name) { 4292 type = eSymbolTypeException; 4293 } else { 4294 type = eSymbolTypeData; 4295 } 4296 } else if (symbol_sect_name && 4297 ::strstr(symbol_sect_name, "__IMPORT") == 4298 symbol_sect_name) { 4299 type = eSymbolTypeTrampoline; 4300 } else if (symbol_section->IsDescendant(objc_section_sp.get())) { 4301 type = eSymbolTypeRuntime; 4302 if (symbol_name && symbol_name[0] == '.') { 4303 llvm::StringRef symbol_name_ref(symbol_name); 4304 llvm::StringRef g_objc_v1_prefix_class( 4305 ".objc_class_name_"); 4306 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4307 symbol_name_non_abi_mangled = symbol_name; 4308 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4309 type = eSymbolTypeObjCClass; 4310 demangled_is_synthesized = true; 4311 } 4312 } 4313 } 4314 } 4315 } 4316 } break; 4317 } 4318 } 4319 4320 if (!add_nlist) { 4321 sym[sym_idx].Clear(); 4322 return true; 4323 } 4324 4325 uint64_t symbol_value = nlist.n_value; 4326 4327 if (symbol_name_non_abi_mangled) { 4328 sym[sym_idx].GetMangled().SetMangledName( 4329 ConstString(symbol_name_non_abi_mangled)); 4330 sym[sym_idx].GetMangled().SetDemangledName(ConstString(symbol_name)); 4331 } else { 4332 bool symbol_name_is_mangled = false; 4333 4334 if (symbol_name && symbol_name[0] == '_') { 4335 symbol_name_is_mangled = symbol_name[1] == '_'; 4336 symbol_name++; // Skip the leading underscore 4337 } 4338 4339 if (symbol_name) { 4340 ConstString const_symbol_name(symbol_name); 4341 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4342 symbol_name_is_mangled); 4343 } 4344 } 4345 4346 if (is_gsym) { 4347 const char *gsym_name = sym[sym_idx] 4348 .GetMangled() 4349 .GetName(Mangled::ePreferMangled) 4350 .GetCString(); 4351 if (gsym_name) 4352 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4353 } 4354 4355 if (symbol_section) { 4356 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4357 if (symbol_byte_size == 0 && function_starts_count > 0) { 4358 addr_t symbol_lookup_file_addr = nlist.n_value; 4359 // Do an exact address match for non-ARM addresses, else get the 4360 // closest since the symbol might be a thumb symbol which has an 4361 // address with bit zero set. 4362 FunctionStarts::Entry *func_start_entry = 4363 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4364 if (is_arm && func_start_entry) { 4365 // Verify that the function start address is the symbol address 4366 // (ARM) or the symbol address + 1 (thumb). 4367 if (func_start_entry->addr != symbol_lookup_file_addr && 4368 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4369 // Not the right entry, NULL it out... 4370 func_start_entry = nullptr; 4371 } 4372 } 4373 if (func_start_entry) { 4374 func_start_entry->data = true; 4375 4376 addr_t symbol_file_addr = func_start_entry->addr; 4377 if (is_arm) 4378 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4379 4380 const FunctionStarts::Entry *next_func_start_entry = 4381 function_starts.FindNextEntry(func_start_entry); 4382 const addr_t section_end_file_addr = 4383 section_file_addr + symbol_section->GetByteSize(); 4384 if (next_func_start_entry) { 4385 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4386 // Be sure the clear the Thumb address bit when we calculate the 4387 // size from the current and next address 4388 if (is_arm) 4389 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4390 symbol_byte_size = std::min<lldb::addr_t>( 4391 next_symbol_file_addr - symbol_file_addr, 4392 section_end_file_addr - symbol_file_addr); 4393 } else { 4394 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4395 } 4396 } 4397 } 4398 symbol_value -= section_file_addr; 4399 } 4400 4401 if (!is_debug) { 4402 if (type == eSymbolTypeCode) { 4403 // See if we can find a N_FUN entry for any code symbols. If we do 4404 // find a match, and the name matches, then we can merge the two into 4405 // just the function symbol to avoid duplicate entries in the symbol 4406 // table. 4407 std::pair<ValueToSymbolIndexMap::const_iterator, 4408 ValueToSymbolIndexMap::const_iterator> 4409 range; 4410 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4411 if (range.first != range.second) { 4412 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4413 pos != range.second; ++pos) { 4414 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4415 sym[pos->second].GetMangled().GetName( 4416 Mangled::ePreferMangled)) { 4417 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4418 // We just need the flags from the linker symbol, so put these 4419 // flags into the N_FUN flags to avoid duplicate symbols in the 4420 // symbol table. 4421 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4422 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4423 if (resolver_addresses.find(nlist.n_value) != 4424 resolver_addresses.end()) 4425 sym[pos->second].SetType(eSymbolTypeResolver); 4426 sym[sym_idx].Clear(); 4427 return true; 4428 } 4429 } 4430 } else { 4431 if (resolver_addresses.find(nlist.n_value) != 4432 resolver_addresses.end()) 4433 type = eSymbolTypeResolver; 4434 } 4435 } else if (type == eSymbolTypeData || type == eSymbolTypeObjCClass || 4436 type == eSymbolTypeObjCMetaClass || 4437 type == eSymbolTypeObjCIVar) { 4438 // See if we can find a N_STSYM entry for any data symbols. If we do 4439 // find a match, and the name matches, then we can merge the two into 4440 // just the Static symbol to avoid duplicate entries in the symbol 4441 // table. 4442 std::pair<ValueToSymbolIndexMap::const_iterator, 4443 ValueToSymbolIndexMap::const_iterator> 4444 range; 4445 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4446 if (range.first != range.second) { 4447 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4448 pos != range.second; ++pos) { 4449 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4450 sym[pos->second].GetMangled().GetName( 4451 Mangled::ePreferMangled)) { 4452 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4453 // We just need the flags from the linker symbol, so put these 4454 // flags into the N_STSYM flags to avoid duplicate symbols in 4455 // the symbol table. 4456 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4457 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4458 sym[sym_idx].Clear(); 4459 return true; 4460 } 4461 } 4462 } else { 4463 // Combine N_GSYM stab entries with the non stab symbol. 4464 const char *gsym_name = sym[sym_idx] 4465 .GetMangled() 4466 .GetName(Mangled::ePreferMangled) 4467 .GetCString(); 4468 if (gsym_name) { 4469 ConstNameToSymbolIndexMap::const_iterator pos = 4470 N_GSYM_name_to_sym_idx.find(gsym_name); 4471 if (pos != N_GSYM_name_to_sym_idx.end()) { 4472 const uint32_t GSYM_sym_idx = pos->second; 4473 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4474 // Copy the address, because often the N_GSYM address has an 4475 // invalid address of zero when the global is a common symbol. 4476 sym[GSYM_sym_idx].GetAddressRef().SetSection(symbol_section); 4477 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4478 symbols_added.insert( 4479 sym[GSYM_sym_idx].GetAddress().GetFileAddress()); 4480 // We just need the flags from the linker symbol, so put these 4481 // flags into the N_GSYM flags to avoid duplicate symbols in 4482 // the symbol table. 4483 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4484 sym[sym_idx].Clear(); 4485 return true; 4486 } 4487 } 4488 } 4489 } 4490 } 4491 4492 sym[sym_idx].SetID(nlist_idx); 4493 sym[sym_idx].SetType(type); 4494 if (set_value) { 4495 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4496 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4497 symbols_added.insert(sym[sym_idx].GetAddress().GetFileAddress()); 4498 } 4499 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4500 if (nlist.n_desc & N_WEAK_REF) 4501 sym[sym_idx].SetIsWeak(true); 4502 4503 if (symbol_byte_size > 0) 4504 sym[sym_idx].SetByteSize(symbol_byte_size); 4505 4506 if (demangled_is_synthesized) 4507 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4508 4509 ++sym_idx; 4510 return true; 4511 }; 4512 4513 // First parse all the nlists but don't process them yet. See the next 4514 // comment for an explanation why. 4515 std::vector<struct nlist_64> nlists; 4516 nlists.reserve(symtab_load_command.nsyms); 4517 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 4518 if (auto nlist = 4519 ParseNList(nlist_data, nlist_data_offset, nlist_byte_size)) 4520 nlists.push_back(*nlist); 4521 else 4522 break; 4523 } 4524 4525 // Now parse all the debug symbols. This is needed to merge non-debug 4526 // symbols in the next step. Non-debug symbols are always coalesced into 4527 // the debug symbol. Doing this in one step would mean that some symbols 4528 // won't be merged. 4529 nlist_idx = 0; 4530 for (auto &nlist : nlists) { 4531 if (!ParseSymbolLambda(nlist, nlist_idx++, DebugSymbols)) 4532 break; 4533 } 4534 4535 // Finally parse all the non debug symbols. 4536 nlist_idx = 0; 4537 for (auto &nlist : nlists) { 4538 if (!ParseSymbolLambda(nlist, nlist_idx++, NonDebugSymbols)) 4539 break; 4540 } 4541 4542 for (const auto &pos : reexport_shlib_needs_fixup) { 4543 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4544 if (undef_pos != undefined_name_to_desc.end()) { 4545 const uint8_t dylib_ordinal = 4546 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4547 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4548 sym[pos.first].SetReExportedSymbolSharedLibrary( 4549 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4550 } 4551 } 4552 } 4553 4554 // Count how many trie symbols we'll add to the symbol table 4555 int trie_symbol_table_augment_count = 0; 4556 for (auto &e : external_sym_trie_entries) { 4557 if (symbols_added.find(e.entry.address) == symbols_added.end()) 4558 trie_symbol_table_augment_count++; 4559 } 4560 4561 if (num_syms < sym_idx + trie_symbol_table_augment_count) { 4562 num_syms = sym_idx + trie_symbol_table_augment_count; 4563 sym = symtab->Resize(num_syms); 4564 } 4565 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4566 4567 // Add symbols from the trie to the symbol table. 4568 for (auto &e : external_sym_trie_entries) { 4569 if (symbols_added.find(e.entry.address) != symbols_added.end()) 4570 continue; 4571 4572 // Find the section that this trie address is in, use that to annotate 4573 // symbol type as we add the trie address and name to the symbol table. 4574 Address symbol_addr; 4575 if (module_sp->ResolveFileAddress(e.entry.address, symbol_addr)) { 4576 SectionSP symbol_section(symbol_addr.GetSection()); 4577 const char *symbol_name = e.entry.name.GetCString(); 4578 bool demangled_is_synthesized = false; 4579 SymbolType type = 4580 GetSymbolType(symbol_name, demangled_is_synthesized, text_section_sp, 4581 data_section_sp, data_dirty_section_sp, 4582 data_const_section_sp, symbol_section); 4583 4584 sym[sym_idx].SetType(type); 4585 if (symbol_section) { 4586 sym[sym_idx].SetID(synthetic_sym_id++); 4587 sym[sym_idx].GetMangled().SetMangledName(ConstString(symbol_name)); 4588 if (demangled_is_synthesized) 4589 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4590 sym[sym_idx].SetIsSynthetic(true); 4591 sym[sym_idx].SetExternal(true); 4592 sym[sym_idx].GetAddressRef() = symbol_addr; 4593 symbols_added.insert(symbol_addr.GetFileAddress()); 4594 if (e.entry.flags & TRIE_SYMBOL_IS_THUMB) 4595 sym[sym_idx].SetFlags(MACHO_NLIST_ARM_SYMBOL_IS_THUMB); 4596 ++sym_idx; 4597 } 4598 } 4599 } 4600 4601 if (function_starts_count > 0) { 4602 uint32_t num_synthetic_function_symbols = 0; 4603 for (i = 0; i < function_starts_count; ++i) { 4604 if (symbols_added.find(function_starts.GetEntryRef(i).addr) == 4605 symbols_added.end()) 4606 ++num_synthetic_function_symbols; 4607 } 4608 4609 if (num_synthetic_function_symbols > 0) { 4610 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4611 num_syms = sym_idx + num_synthetic_function_symbols; 4612 sym = symtab->Resize(num_syms); 4613 } 4614 for (i = 0; i < function_starts_count; ++i) { 4615 const FunctionStarts::Entry *func_start_entry = 4616 function_starts.GetEntryAtIndex(i); 4617 if (symbols_added.find(func_start_entry->addr) == symbols_added.end()) { 4618 addr_t symbol_file_addr = func_start_entry->addr; 4619 uint32_t symbol_flags = 0; 4620 if (func_start_entry->data) 4621 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4622 Address symbol_addr; 4623 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4624 SectionSP symbol_section(symbol_addr.GetSection()); 4625 uint32_t symbol_byte_size = 0; 4626 if (symbol_section) { 4627 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4628 const FunctionStarts::Entry *next_func_start_entry = 4629 function_starts.FindNextEntry(func_start_entry); 4630 const addr_t section_end_file_addr = 4631 section_file_addr + symbol_section->GetByteSize(); 4632 if (next_func_start_entry) { 4633 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4634 if (is_arm) 4635 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4636 symbol_byte_size = std::min<lldb::addr_t>( 4637 next_symbol_file_addr - symbol_file_addr, 4638 section_end_file_addr - symbol_file_addr); 4639 } else { 4640 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4641 } 4642 sym[sym_idx].SetID(synthetic_sym_id++); 4643 sym[sym_idx].GetMangled().SetDemangledName( 4644 GetNextSyntheticSymbolName()); 4645 sym[sym_idx].SetType(eSymbolTypeCode); 4646 sym[sym_idx].SetIsSynthetic(true); 4647 sym[sym_idx].GetAddressRef() = symbol_addr; 4648 symbols_added.insert(symbol_addr.GetFileAddress()); 4649 if (symbol_flags) 4650 sym[sym_idx].SetFlags(symbol_flags); 4651 if (symbol_byte_size) 4652 sym[sym_idx].SetByteSize(symbol_byte_size); 4653 ++sym_idx; 4654 } 4655 } 4656 } 4657 } 4658 } 4659 } 4660 4661 // Trim our symbols down to just what we ended up with after removing any 4662 // symbols. 4663 if (sym_idx < num_syms) { 4664 num_syms = sym_idx; 4665 sym = symtab->Resize(num_syms); 4666 } 4667 4668 // Now synthesize indirect symbols 4669 if (m_dysymtab.nindirectsyms != 0) { 4670 if (indirect_symbol_index_data.GetByteSize()) { 4671 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4672 m_nlist_idx_to_sym_idx.end(); 4673 4674 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4675 ++sect_idx) { 4676 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4677 S_SYMBOL_STUBS) { 4678 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4679 if (symbol_stub_byte_size == 0) 4680 continue; 4681 4682 const uint32_t num_symbol_stubs = 4683 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4684 4685 if (num_symbol_stubs == 0) 4686 continue; 4687 4688 const uint32_t symbol_stub_index_offset = 4689 m_mach_sections[sect_idx].reserved1; 4690 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) { 4691 const uint32_t symbol_stub_index = 4692 symbol_stub_index_offset + stub_idx; 4693 const lldb::addr_t symbol_stub_addr = 4694 m_mach_sections[sect_idx].addr + 4695 (stub_idx * symbol_stub_byte_size); 4696 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4697 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4698 symbol_stub_offset, 4)) { 4699 const uint32_t stub_sym_id = 4700 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4701 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4702 continue; 4703 4704 NListIndexToSymbolIndexMap::const_iterator index_pos = 4705 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4706 Symbol *stub_symbol = nullptr; 4707 if (index_pos != end_index_pos) { 4708 // We have a remapping from the original nlist index to a 4709 // current symbol index, so just look this up by index 4710 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4711 } else { 4712 // We need to lookup a symbol using the original nlist symbol 4713 // index since this index is coming from the S_SYMBOL_STUBS 4714 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4715 } 4716 4717 if (stub_symbol) { 4718 Address so_addr(symbol_stub_addr, section_list); 4719 4720 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4721 // Change the external symbol into a trampoline that makes 4722 // sense These symbols were N_UNDF N_EXT, and are useless 4723 // to us, so we can re-use them so we don't have to make up 4724 // a synthetic symbol for no good reason. 4725 if (resolver_addresses.find(symbol_stub_addr) == 4726 resolver_addresses.end()) 4727 stub_symbol->SetType(eSymbolTypeTrampoline); 4728 else 4729 stub_symbol->SetType(eSymbolTypeResolver); 4730 stub_symbol->SetExternal(false); 4731 stub_symbol->GetAddressRef() = so_addr; 4732 stub_symbol->SetByteSize(symbol_stub_byte_size); 4733 } else { 4734 // Make a synthetic symbol to describe the trampoline stub 4735 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4736 if (sym_idx >= num_syms) { 4737 sym = symtab->Resize(++num_syms); 4738 stub_symbol = nullptr; // this pointer no longer valid 4739 } 4740 sym[sym_idx].SetID(synthetic_sym_id++); 4741 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4742 if (resolver_addresses.find(symbol_stub_addr) == 4743 resolver_addresses.end()) 4744 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4745 else 4746 sym[sym_idx].SetType(eSymbolTypeResolver); 4747 sym[sym_idx].SetIsSynthetic(true); 4748 sym[sym_idx].GetAddressRef() = so_addr; 4749 symbols_added.insert(so_addr.GetFileAddress()); 4750 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4751 ++sym_idx; 4752 } 4753 } else { 4754 if (log) 4755 log->Warning("symbol stub referencing symbol table symbol " 4756 "%u that isn't in our minimal symbol table, " 4757 "fix this!!!", 4758 stub_sym_id); 4759 } 4760 } 4761 } 4762 } 4763 } 4764 } 4765 } 4766 4767 if (!reexport_trie_entries.empty()) { 4768 for (const auto &e : reexport_trie_entries) { 4769 if (e.entry.import_name) { 4770 // Only add indirect symbols from the Trie entries if we didn't have 4771 // a N_INDR nlist entry for this already 4772 if (indirect_symbol_names.find(e.entry.name) == 4773 indirect_symbol_names.end()) { 4774 // Make a synthetic symbol to describe re-exported symbol. 4775 if (sym_idx >= num_syms) 4776 sym = symtab->Resize(++num_syms); 4777 sym[sym_idx].SetID(synthetic_sym_id++); 4778 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4779 sym[sym_idx].SetType(eSymbolTypeReExported); 4780 sym[sym_idx].SetIsSynthetic(true); 4781 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4782 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4783 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4784 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4785 } 4786 ++sym_idx; 4787 } 4788 } 4789 } 4790 } 4791 4792 // StreamFile s(stdout, false); 4793 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4794 // symtab->Dump(&s, NULL, eSortOrderNone); 4795 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4796 // sizes 4797 symtab->CalculateSymbolSizes(); 4798 4799 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4800 // symtab->Dump(&s, NULL, eSortOrderNone); 4801 4802 return symtab->GetNumSymbols(); 4803 } 4804 4805 void ObjectFileMachO::Dump(Stream *s) { 4806 ModuleSP module_sp(GetModule()); 4807 if (module_sp) { 4808 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4809 s->Printf("%p: ", static_cast<void *>(this)); 4810 s->Indent(); 4811 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4812 s->PutCString("ObjectFileMachO64"); 4813 else 4814 s->PutCString("ObjectFileMachO32"); 4815 4816 *s << ", file = '" << m_file; 4817 ModuleSpecList all_specs; 4818 ModuleSpec base_spec; 4819 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 4820 base_spec, all_specs); 4821 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 4822 *s << "', triple"; 4823 if (e) 4824 s->Printf("[%d]", i); 4825 *s << " = "; 4826 *s << all_specs.GetModuleSpecRefAtIndex(i) 4827 .GetArchitecture() 4828 .GetTriple() 4829 .getTriple(); 4830 } 4831 *s << "\n"; 4832 SectionList *sections = GetSectionList(); 4833 if (sections) 4834 sections->Dump(s->AsRawOstream(), s->GetIndentLevel(), nullptr, true, 4835 UINT32_MAX); 4836 4837 if (m_symtab_up) 4838 m_symtab_up->Dump(s, nullptr, eSortOrderNone); 4839 } 4840 } 4841 4842 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4843 const lldb_private::DataExtractor &data, 4844 lldb::offset_t lc_offset) { 4845 uint32_t i; 4846 struct uuid_command load_cmd; 4847 4848 lldb::offset_t offset = lc_offset; 4849 for (i = 0; i < header.ncmds; ++i) { 4850 const lldb::offset_t cmd_offset = offset; 4851 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4852 break; 4853 4854 if (load_cmd.cmd == LC_UUID) { 4855 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4856 4857 if (uuid_bytes) { 4858 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4859 // We pretend these object files have no UUID to prevent crashing. 4860 4861 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4862 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4863 0xbb, 0x14, 0xf0, 0x0d}; 4864 4865 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4866 return UUID(); 4867 4868 return UUID::fromOptionalData(uuid_bytes, 16); 4869 } 4870 return UUID(); 4871 } 4872 offset = cmd_offset + load_cmd.cmdsize; 4873 } 4874 return UUID(); 4875 } 4876 4877 static llvm::StringRef GetOSName(uint32_t cmd) { 4878 switch (cmd) { 4879 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4880 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4881 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4882 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4883 case llvm::MachO::LC_VERSION_MIN_TVOS: 4884 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4885 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4886 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4887 default: 4888 llvm_unreachable("unexpected LC_VERSION load command"); 4889 } 4890 } 4891 4892 namespace { 4893 struct OSEnv { 4894 llvm::StringRef os_type; 4895 llvm::StringRef environment; 4896 OSEnv(uint32_t cmd) { 4897 switch (cmd) { 4898 case llvm::MachO::PLATFORM_MACOS: 4899 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4900 return; 4901 case llvm::MachO::PLATFORM_IOS: 4902 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4903 return; 4904 case llvm::MachO::PLATFORM_TVOS: 4905 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4906 return; 4907 case llvm::MachO::PLATFORM_WATCHOS: 4908 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4909 return; 4910 // NEED_BRIDGEOS_TRIPLE case llvm::MachO::PLATFORM_BRIDGEOS: 4911 // NEED_BRIDGEOS_TRIPLE os_type = 4912 // llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4913 // NEED_BRIDGEOS_TRIPLE return; 4914 case llvm::MachO::PLATFORM_MACCATALYST: 4915 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4916 environment = llvm::Triple::getEnvironmentTypeName(llvm::Triple::MacABI); 4917 return; 4918 case llvm::MachO::PLATFORM_IOSSIMULATOR: 4919 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4920 environment = 4921 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4922 return; 4923 case llvm::MachO::PLATFORM_TVOSSIMULATOR: 4924 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4925 environment = 4926 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4927 return; 4928 case llvm::MachO::PLATFORM_WATCHOSSIMULATOR: 4929 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4930 environment = 4931 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4932 return; 4933 default: { 4934 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 4935 LIBLLDB_LOG_PROCESS)); 4936 LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION"); 4937 } 4938 } 4939 } 4940 }; 4941 4942 struct MinOS { 4943 uint32_t major_version, minor_version, patch_version; 4944 MinOS(uint32_t version) 4945 : major_version(version >> 16), minor_version((version >> 8) & 0xffu), 4946 patch_version(version & 0xffu) {} 4947 }; 4948 } // namespace 4949 4950 void ObjectFileMachO::GetAllArchSpecs(const llvm::MachO::mach_header &header, 4951 const lldb_private::DataExtractor &data, 4952 lldb::offset_t lc_offset, 4953 ModuleSpec &base_spec, 4954 lldb_private::ModuleSpecList &all_specs) { 4955 auto &base_arch = base_spec.GetArchitecture(); 4956 base_arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4957 if (!base_arch.IsValid()) 4958 return; 4959 4960 bool found_any = false; 4961 auto add_triple = [&](const llvm::Triple &triple) { 4962 auto spec = base_spec; 4963 spec.GetArchitecture().GetTriple() = triple; 4964 if (spec.GetArchitecture().IsValid()) { 4965 spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset); 4966 all_specs.Append(spec); 4967 found_any = true; 4968 } 4969 }; 4970 4971 // Set OS to an unspecified unknown or a "*" so it can match any OS 4972 llvm::Triple base_triple = base_arch.GetTriple(); 4973 base_triple.setOS(llvm::Triple::UnknownOS); 4974 base_triple.setOSName(llvm::StringRef()); 4975 4976 if (header.filetype == MH_PRELOAD) { 4977 if (header.cputype == CPU_TYPE_ARM) { 4978 // If this is a 32-bit arm binary, and it's a standalone binary, force 4979 // the Vendor to Apple so we don't accidentally pick up the generic 4980 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4981 // frame pointer register; most other armv7 ABIs use a combination of 4982 // r7 and r11. 4983 base_triple.setVendor(llvm::Triple::Apple); 4984 } else { 4985 // Set vendor to an unspecified unknown or a "*" so it can match any 4986 // vendor This is required for correct behavior of EFI debugging on 4987 // x86_64 4988 base_triple.setVendor(llvm::Triple::UnknownVendor); 4989 base_triple.setVendorName(llvm::StringRef()); 4990 } 4991 return add_triple(base_triple); 4992 } 4993 4994 struct load_command load_cmd; 4995 4996 // See if there is an LC_VERSION_MIN_* load command that can give 4997 // us the OS type. 4998 lldb::offset_t offset = lc_offset; 4999 for (uint32_t i = 0; i < header.ncmds; ++i) { 5000 const lldb::offset_t cmd_offset = offset; 5001 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 5002 break; 5003 5004 struct version_min_command version_min; 5005 switch (load_cmd.cmd) { 5006 case llvm::MachO::LC_VERSION_MIN_MACOSX: 5007 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 5008 case llvm::MachO::LC_VERSION_MIN_TVOS: 5009 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 5010 if (load_cmd.cmdsize != sizeof(version_min)) 5011 break; 5012 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 5013 data.GetByteOrder(), &version_min) == 0) 5014 break; 5015 MinOS min_os(version_min.version); 5016 llvm::SmallString<32> os_name; 5017 llvm::raw_svector_ostream os(os_name); 5018 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 5019 << min_os.minor_version << '.' << min_os.patch_version; 5020 5021 auto triple = base_triple; 5022 triple.setOSName(os.str()); 5023 5024 // Disambiguate legacy simulator platforms. 5025 if (load_cmd.cmd != llvm::MachO::LC_VERSION_MIN_MACOSX && 5026 (base_triple.getArch() == llvm::Triple::x86_64 || 5027 base_triple.getArch() == llvm::Triple::x86)) { 5028 // The combination of legacy LC_VERSION_MIN load command and 5029 // x86 architecture always indicates a simulator environment. 5030 // The combination of LC_VERSION_MIN and arm architecture only 5031 // appears for native binaries. Back-deploying simulator 5032 // binaries on Apple Silicon Macs use the modern unambigous 5033 // LC_BUILD_VERSION load commands; no special handling required. 5034 triple.setEnvironment(llvm::Triple::Simulator); 5035 } 5036 add_triple(triple); 5037 break; 5038 } 5039 default: 5040 break; 5041 } 5042 5043 offset = cmd_offset + load_cmd.cmdsize; 5044 } 5045 5046 // See if there are LC_BUILD_VERSION load commands that can give 5047 // us the OS type. 5048 offset = lc_offset; 5049 for (uint32_t i = 0; i < header.ncmds; ++i) { 5050 const lldb::offset_t cmd_offset = offset; 5051 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 5052 break; 5053 5054 do { 5055 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 5056 struct build_version_command build_version; 5057 if (load_cmd.cmdsize < sizeof(build_version)) { 5058 // Malformed load command. 5059 break; 5060 } 5061 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 5062 data.GetByteOrder(), &build_version) == 0) 5063 break; 5064 MinOS min_os(build_version.minos); 5065 OSEnv os_env(build_version.platform); 5066 llvm::SmallString<16> os_name; 5067 llvm::raw_svector_ostream os(os_name); 5068 os << os_env.os_type << min_os.major_version << '.' 5069 << min_os.minor_version << '.' << min_os.patch_version; 5070 auto triple = base_triple; 5071 triple.setOSName(os.str()); 5072 os_name.clear(); 5073 if (!os_env.environment.empty()) 5074 triple.setEnvironmentName(os_env.environment); 5075 add_triple(triple); 5076 } 5077 } while (0); 5078 offset = cmd_offset + load_cmd.cmdsize; 5079 } 5080 5081 if (!found_any) { 5082 if (header.filetype == MH_KEXT_BUNDLE) { 5083 base_triple.setVendor(llvm::Triple::Apple); 5084 add_triple(base_triple); 5085 } else { 5086 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 5087 // so lets not say our Vendor is Apple, leave it as an unspecified 5088 // unknown. 5089 base_triple.setVendor(llvm::Triple::UnknownVendor); 5090 base_triple.setVendorName(llvm::StringRef()); 5091 add_triple(base_triple); 5092 } 5093 } 5094 } 5095 5096 ArchSpec ObjectFileMachO::GetArchitecture( 5097 ModuleSP module_sp, const llvm::MachO::mach_header &header, 5098 const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) { 5099 ModuleSpecList all_specs; 5100 ModuleSpec base_spec; 5101 GetAllArchSpecs(header, data, MachHeaderSizeFromMagic(header.magic), 5102 base_spec, all_specs); 5103 5104 // If the object file offers multiple alternative load commands, 5105 // pick the one that matches the module. 5106 if (module_sp) { 5107 const ArchSpec &module_arch = module_sp->GetArchitecture(); 5108 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 5109 ArchSpec mach_arch = 5110 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 5111 if (module_arch.IsCompatibleMatch(mach_arch)) 5112 return mach_arch; 5113 } 5114 } 5115 5116 // Return the first arch we found. 5117 if (all_specs.GetSize() == 0) 5118 return {}; 5119 return all_specs.GetModuleSpecRefAtIndex(0).GetArchitecture(); 5120 } 5121 5122 UUID ObjectFileMachO::GetUUID() { 5123 ModuleSP module_sp(GetModule()); 5124 if (module_sp) { 5125 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5126 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5127 return GetUUID(m_header, m_data, offset); 5128 } 5129 return UUID(); 5130 } 5131 5132 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5133 uint32_t count = 0; 5134 ModuleSP module_sp(GetModule()); 5135 if (module_sp) { 5136 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5137 struct load_command load_cmd; 5138 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5139 std::vector<std::string> rpath_paths; 5140 std::vector<std::string> rpath_relative_paths; 5141 std::vector<std::string> at_exec_relative_paths; 5142 uint32_t i; 5143 for (i = 0; i < m_header.ncmds; ++i) { 5144 const uint32_t cmd_offset = offset; 5145 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5146 break; 5147 5148 switch (load_cmd.cmd) { 5149 case LC_RPATH: 5150 case LC_LOAD_DYLIB: 5151 case LC_LOAD_WEAK_DYLIB: 5152 case LC_REEXPORT_DYLIB: 5153 case LC_LOAD_DYLINKER: 5154 case LC_LOADFVMLIB: 5155 case LC_LOAD_UPWARD_DYLIB: { 5156 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5157 const char *path = m_data.PeekCStr(name_offset); 5158 if (path) { 5159 if (load_cmd.cmd == LC_RPATH) 5160 rpath_paths.push_back(path); 5161 else { 5162 if (path[0] == '@') { 5163 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5164 rpath_relative_paths.push_back(path + strlen("@rpath")); 5165 else if (strncmp(path, "@executable_path", 5166 strlen("@executable_path")) == 0) 5167 at_exec_relative_paths.push_back(path + 5168 strlen("@executable_path")); 5169 } else { 5170 FileSpec file_spec(path); 5171 if (files.AppendIfUnique(file_spec)) 5172 count++; 5173 } 5174 } 5175 } 5176 } break; 5177 5178 default: 5179 break; 5180 } 5181 offset = cmd_offset + load_cmd.cmdsize; 5182 } 5183 5184 FileSpec this_file_spec(m_file); 5185 FileSystem::Instance().Resolve(this_file_spec); 5186 5187 if (!rpath_paths.empty()) { 5188 // Fixup all LC_RPATH values to be absolute paths 5189 std::string loader_path("@loader_path"); 5190 std::string executable_path("@executable_path"); 5191 for (auto &rpath : rpath_paths) { 5192 if (llvm::StringRef(rpath).startswith(loader_path)) { 5193 rpath.erase(0, loader_path.size()); 5194 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5195 } else if (llvm::StringRef(rpath).startswith(executable_path)) { 5196 rpath.erase(0, executable_path.size()); 5197 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5198 } 5199 } 5200 5201 for (const auto &rpath_relative_path : rpath_relative_paths) { 5202 for (const auto &rpath : rpath_paths) { 5203 std::string path = rpath; 5204 path += rpath_relative_path; 5205 // It is OK to resolve this path because we must find a file on disk 5206 // for us to accept it anyway if it is rpath relative. 5207 FileSpec file_spec(path); 5208 FileSystem::Instance().Resolve(file_spec); 5209 if (FileSystem::Instance().Exists(file_spec) && 5210 files.AppendIfUnique(file_spec)) { 5211 count++; 5212 break; 5213 } 5214 } 5215 } 5216 } 5217 5218 // We may have @executable_paths but no RPATHS. Figure those out here. 5219 // Only do this if this object file is the executable. We have no way to 5220 // get back to the actual executable otherwise, so we won't get the right 5221 // path. 5222 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5223 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5224 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5225 FileSpec file_spec = 5226 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5227 if (FileSystem::Instance().Exists(file_spec) && 5228 files.AppendIfUnique(file_spec)) 5229 count++; 5230 } 5231 } 5232 } 5233 return count; 5234 } 5235 5236 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5237 // If the object file is not an executable it can't hold the entry point. 5238 // m_entry_point_address is initialized to an invalid address, so we can just 5239 // return that. If m_entry_point_address is valid it means we've found it 5240 // already, so return the cached value. 5241 5242 if ((!IsExecutable() && !IsDynamicLoader()) || 5243 m_entry_point_address.IsValid()) { 5244 return m_entry_point_address; 5245 } 5246 5247 // Otherwise, look for the UnixThread or Thread command. The data for the 5248 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5249 // 5250 // uint32_t flavor - this is the flavor argument you would pass to 5251 // thread_get_state 5252 // uint32_t count - this is the count of longs in the thread state data 5253 // struct XXX_thread_state state - this is the structure from 5254 // <machine/thread_status.h> corresponding to the flavor. 5255 // <repeat this trio> 5256 // 5257 // So we just keep reading the various register flavors till we find the GPR 5258 // one, then read the PC out of there. 5259 // FIXME: We will need to have a "RegisterContext data provider" class at some 5260 // point that can get all the registers 5261 // out of data in this form & attach them to a given thread. That should 5262 // underlie the MacOS X User process plugin, and we'll also need it for the 5263 // MacOS X Core File process plugin. When we have that we can also use it 5264 // here. 5265 // 5266 // For now we hard-code the offsets and flavors we need: 5267 // 5268 // 5269 5270 ModuleSP module_sp(GetModule()); 5271 if (module_sp) { 5272 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5273 struct load_command load_cmd; 5274 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5275 uint32_t i; 5276 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5277 bool done = false; 5278 5279 for (i = 0; i < m_header.ncmds; ++i) { 5280 const lldb::offset_t cmd_offset = offset; 5281 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5282 break; 5283 5284 switch (load_cmd.cmd) { 5285 case LC_UNIXTHREAD: 5286 case LC_THREAD: { 5287 while (offset < cmd_offset + load_cmd.cmdsize) { 5288 uint32_t flavor = m_data.GetU32(&offset); 5289 uint32_t count = m_data.GetU32(&offset); 5290 if (count == 0) { 5291 // We've gotten off somehow, log and exit; 5292 return m_entry_point_address; 5293 } 5294 5295 switch (m_header.cputype) { 5296 case llvm::MachO::CPU_TYPE_ARM: 5297 if (flavor == 1 || 5298 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 5299 // from mach/arm/thread_status.h 5300 { 5301 offset += 60; // This is the offset of pc in the GPR thread state 5302 // data structure. 5303 start_address = m_data.GetU32(&offset); 5304 done = true; 5305 } 5306 break; 5307 case llvm::MachO::CPU_TYPE_ARM64: 5308 case llvm::MachO::CPU_TYPE_ARM64_32: 5309 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5310 { 5311 offset += 256; // This is the offset of pc in the GPR thread state 5312 // data structure. 5313 start_address = m_data.GetU64(&offset); 5314 done = true; 5315 } 5316 break; 5317 case llvm::MachO::CPU_TYPE_I386: 5318 if (flavor == 5319 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5320 { 5321 offset += 40; // This is the offset of eip in the GPR thread state 5322 // data structure. 5323 start_address = m_data.GetU32(&offset); 5324 done = true; 5325 } 5326 break; 5327 case llvm::MachO::CPU_TYPE_X86_64: 5328 if (flavor == 5329 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5330 { 5331 offset += 16 * 8; // This is the offset of rip in the GPR thread 5332 // state data structure. 5333 start_address = m_data.GetU64(&offset); 5334 done = true; 5335 } 5336 break; 5337 default: 5338 return m_entry_point_address; 5339 } 5340 // Haven't found the GPR flavor yet, skip over the data for this 5341 // flavor: 5342 if (done) 5343 break; 5344 offset += count * 4; 5345 } 5346 } break; 5347 case LC_MAIN: { 5348 ConstString text_segment_name("__TEXT"); 5349 uint64_t entryoffset = m_data.GetU64(&offset); 5350 SectionSP text_segment_sp = 5351 GetSectionList()->FindSectionByName(text_segment_name); 5352 if (text_segment_sp) { 5353 done = true; 5354 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5355 } 5356 } break; 5357 5358 default: 5359 break; 5360 } 5361 if (done) 5362 break; 5363 5364 // Go to the next load command: 5365 offset = cmd_offset + load_cmd.cmdsize; 5366 } 5367 5368 if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) { 5369 if (GetSymtab()) { 5370 Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType( 5371 ConstString("_dyld_start"), SymbolType::eSymbolTypeCode, 5372 Symtab::eDebugAny, Symtab::eVisibilityAny); 5373 if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) { 5374 start_address = dyld_start_sym->GetAddress().GetFileAddress(); 5375 } 5376 } 5377 } 5378 5379 if (start_address != LLDB_INVALID_ADDRESS) { 5380 // We got the start address from the load commands, so now resolve that 5381 // address in the sections of this ObjectFile: 5382 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5383 start_address, GetSectionList())) { 5384 m_entry_point_address.Clear(); 5385 } 5386 } else { 5387 // We couldn't read the UnixThread load command - maybe it wasn't there. 5388 // As a fallback look for the "start" symbol in the main executable. 5389 5390 ModuleSP module_sp(GetModule()); 5391 5392 if (module_sp) { 5393 SymbolContextList contexts; 5394 SymbolContext context; 5395 module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5396 eSymbolTypeCode, contexts); 5397 if (contexts.GetSize()) { 5398 if (contexts.GetContextAtIndex(0, context)) 5399 m_entry_point_address = context.symbol->GetAddress(); 5400 } 5401 } 5402 } 5403 } 5404 5405 return m_entry_point_address; 5406 } 5407 5408 lldb_private::Address ObjectFileMachO::GetBaseAddress() { 5409 lldb_private::Address header_addr; 5410 SectionList *section_list = GetSectionList(); 5411 if (section_list) { 5412 SectionSP text_segment_sp( 5413 section_list->FindSectionByName(GetSegmentNameTEXT())); 5414 if (text_segment_sp) { 5415 header_addr.SetSection(text_segment_sp); 5416 header_addr.SetOffset(0); 5417 } 5418 } 5419 return header_addr; 5420 } 5421 5422 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5423 ModuleSP module_sp(GetModule()); 5424 if (module_sp) { 5425 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5426 if (!m_thread_context_offsets_valid) { 5427 m_thread_context_offsets_valid = true; 5428 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5429 FileRangeArray::Entry file_range; 5430 thread_command thread_cmd; 5431 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5432 const uint32_t cmd_offset = offset; 5433 if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr) 5434 break; 5435 5436 if (thread_cmd.cmd == LC_THREAD) { 5437 file_range.SetRangeBase(offset); 5438 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5439 m_thread_context_offsets.Append(file_range); 5440 } 5441 offset = cmd_offset + thread_cmd.cmdsize; 5442 } 5443 } 5444 } 5445 return m_thread_context_offsets.GetSize(); 5446 } 5447 5448 std::string ObjectFileMachO::GetIdentifierString() { 5449 std::string result; 5450 ModuleSP module_sp(GetModule()); 5451 if (module_sp) { 5452 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5453 5454 // First, look over the load commands for an LC_NOTE load command with 5455 // data_owner string "kern ver str" & use that if found. 5456 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5457 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5458 const uint32_t cmd_offset = offset; 5459 load_command lc; 5460 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5461 break; 5462 if (lc.cmd == LC_NOTE) { 5463 char data_owner[17]; 5464 m_data.CopyData(offset, 16, data_owner); 5465 data_owner[16] = '\0'; 5466 offset += 16; 5467 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5468 uint64_t size = m_data.GetU64_unchecked(&offset); 5469 5470 // "kern ver str" has a uint32_t version and then a nul terminated 5471 // c-string. 5472 if (strcmp("kern ver str", data_owner) == 0) { 5473 offset = fileoff; 5474 uint32_t version; 5475 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5476 if (version == 1) { 5477 uint32_t strsize = size - sizeof(uint32_t); 5478 char *buf = (char *)malloc(strsize); 5479 if (buf) { 5480 m_data.CopyData(offset, strsize, buf); 5481 buf[strsize - 1] = '\0'; 5482 result = buf; 5483 if (buf) 5484 free(buf); 5485 return result; 5486 } 5487 } 5488 } 5489 } 5490 } 5491 offset = cmd_offset + lc.cmdsize; 5492 } 5493 5494 // Second, make a pass over the load commands looking for an obsolete 5495 // LC_IDENT load command. 5496 offset = MachHeaderSizeFromMagic(m_header.magic); 5497 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5498 const uint32_t cmd_offset = offset; 5499 struct ident_command ident_command; 5500 if (m_data.GetU32(&offset, &ident_command, 2) == nullptr) 5501 break; 5502 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5503 char *buf = (char *)malloc(ident_command.cmdsize); 5504 if (buf != nullptr && m_data.CopyData(offset, ident_command.cmdsize, 5505 buf) == ident_command.cmdsize) { 5506 buf[ident_command.cmdsize - 1] = '\0'; 5507 result = buf; 5508 } 5509 if (buf) 5510 free(buf); 5511 } 5512 offset = cmd_offset + ident_command.cmdsize; 5513 } 5514 } 5515 return result; 5516 } 5517 5518 bool ObjectFileMachO::GetCorefileMainBinaryInfo(addr_t &address, UUID &uuid, 5519 ObjectFile::BinaryType &type) { 5520 address = LLDB_INVALID_ADDRESS; 5521 uuid.Clear(); 5522 ModuleSP module_sp(GetModule()); 5523 if (module_sp) { 5524 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5525 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5526 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5527 const uint32_t cmd_offset = offset; 5528 load_command lc; 5529 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5530 break; 5531 if (lc.cmd == LC_NOTE) { 5532 char data_owner[17]; 5533 memset(data_owner, 0, sizeof(data_owner)); 5534 m_data.CopyData(offset, 16, data_owner); 5535 offset += 16; 5536 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5537 uint64_t size = m_data.GetU64_unchecked(&offset); 5538 5539 // "main bin spec" (main binary specification) data payload is 5540 // formatted: 5541 // uint32_t version [currently 1] 5542 // uint32_t type [0 == unspecified, 1 == kernel, 5543 // 2 == user process, 3 == firmware ] 5544 // uint64_t address [ UINT64_MAX if address not specified ] 5545 // uuid_t uuid [ all zero's if uuid not specified ] 5546 // uint32_t log2_pagesize [ process page size in log base 5547 // 2, e.g. 4k pages are 12. 5548 // 0 for unspecified ] 5549 // uint32_t unused [ for alignment ] 5550 5551 if (strcmp("main bin spec", data_owner) == 0 && size >= 32) { 5552 offset = fileoff; 5553 uint32_t version; 5554 if (m_data.GetU32(&offset, &version, 1) != nullptr && version == 1) { 5555 uint32_t binspec_type = 0; 5556 uuid_t raw_uuid; 5557 memset(raw_uuid, 0, sizeof(uuid_t)); 5558 5559 if (m_data.GetU32(&offset, &binspec_type, 1) && 5560 m_data.GetU64(&offset, &address, 1) && 5561 m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5562 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5563 // convert the "main bin spec" type into our 5564 // ObjectFile::BinaryType enum 5565 switch (binspec_type) { 5566 case 0: 5567 type = eBinaryTypeUnknown; 5568 break; 5569 case 1: 5570 type = eBinaryTypeKernel; 5571 break; 5572 case 2: 5573 type = eBinaryTypeUser; 5574 break; 5575 case 3: 5576 type = eBinaryTypeStandalone; 5577 break; 5578 } 5579 return true; 5580 } 5581 } 5582 } 5583 } 5584 offset = cmd_offset + lc.cmdsize; 5585 } 5586 } 5587 return false; 5588 } 5589 5590 lldb::RegisterContextSP 5591 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5592 lldb_private::Thread &thread) { 5593 lldb::RegisterContextSP reg_ctx_sp; 5594 5595 ModuleSP module_sp(GetModule()); 5596 if (module_sp) { 5597 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5598 if (!m_thread_context_offsets_valid) 5599 GetNumThreadContexts(); 5600 5601 const FileRangeArray::Entry *thread_context_file_range = 5602 m_thread_context_offsets.GetEntryAtIndex(idx); 5603 if (thread_context_file_range) { 5604 5605 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5606 thread_context_file_range->GetByteSize()); 5607 5608 switch (m_header.cputype) { 5609 case llvm::MachO::CPU_TYPE_ARM64: 5610 case llvm::MachO::CPU_TYPE_ARM64_32: 5611 reg_ctx_sp = 5612 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data); 5613 break; 5614 5615 case llvm::MachO::CPU_TYPE_ARM: 5616 reg_ctx_sp = 5617 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data); 5618 break; 5619 5620 case llvm::MachO::CPU_TYPE_I386: 5621 reg_ctx_sp = 5622 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data); 5623 break; 5624 5625 case llvm::MachO::CPU_TYPE_X86_64: 5626 reg_ctx_sp = 5627 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data); 5628 break; 5629 } 5630 } 5631 } 5632 return reg_ctx_sp; 5633 } 5634 5635 ObjectFile::Type ObjectFileMachO::CalculateType() { 5636 switch (m_header.filetype) { 5637 case MH_OBJECT: // 0x1u 5638 if (GetAddressByteSize() == 4) { 5639 // 32 bit kexts are just object files, but they do have a valid 5640 // UUID load command. 5641 if (GetUUID()) { 5642 // this checking for the UUID load command is not enough we could 5643 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5644 // this is required of kexts 5645 if (m_strata == eStrataInvalid) 5646 m_strata = eStrataKernel; 5647 return eTypeSharedLibrary; 5648 } 5649 } 5650 return eTypeObjectFile; 5651 5652 case MH_EXECUTE: 5653 return eTypeExecutable; // 0x2u 5654 case MH_FVMLIB: 5655 return eTypeSharedLibrary; // 0x3u 5656 case MH_CORE: 5657 return eTypeCoreFile; // 0x4u 5658 case MH_PRELOAD: 5659 return eTypeSharedLibrary; // 0x5u 5660 case MH_DYLIB: 5661 return eTypeSharedLibrary; // 0x6u 5662 case MH_DYLINKER: 5663 return eTypeDynamicLinker; // 0x7u 5664 case MH_BUNDLE: 5665 return eTypeSharedLibrary; // 0x8u 5666 case MH_DYLIB_STUB: 5667 return eTypeStubLibrary; // 0x9u 5668 case MH_DSYM: 5669 return eTypeDebugInfo; // 0xAu 5670 case MH_KEXT_BUNDLE: 5671 return eTypeSharedLibrary; // 0xBu 5672 default: 5673 break; 5674 } 5675 return eTypeUnknown; 5676 } 5677 5678 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5679 switch (m_header.filetype) { 5680 case MH_OBJECT: // 0x1u 5681 { 5682 // 32 bit kexts are just object files, but they do have a valid 5683 // UUID load command. 5684 if (GetUUID()) { 5685 // this checking for the UUID load command is not enough we could 5686 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5687 // this is required of kexts 5688 if (m_type == eTypeInvalid) 5689 m_type = eTypeSharedLibrary; 5690 5691 return eStrataKernel; 5692 } 5693 } 5694 return eStrataUnknown; 5695 5696 case MH_EXECUTE: // 0x2u 5697 // Check for the MH_DYLDLINK bit in the flags 5698 if (m_header.flags & MH_DYLDLINK) { 5699 return eStrataUser; 5700 } else { 5701 SectionList *section_list = GetSectionList(); 5702 if (section_list) { 5703 static ConstString g_kld_section_name("__KLD"); 5704 if (section_list->FindSectionByName(g_kld_section_name)) 5705 return eStrataKernel; 5706 } 5707 } 5708 return eStrataRawImage; 5709 5710 case MH_FVMLIB: 5711 return eStrataUser; // 0x3u 5712 case MH_CORE: 5713 return eStrataUnknown; // 0x4u 5714 case MH_PRELOAD: 5715 return eStrataRawImage; // 0x5u 5716 case MH_DYLIB: 5717 return eStrataUser; // 0x6u 5718 case MH_DYLINKER: 5719 return eStrataUser; // 0x7u 5720 case MH_BUNDLE: 5721 return eStrataUser; // 0x8u 5722 case MH_DYLIB_STUB: 5723 return eStrataUser; // 0x9u 5724 case MH_DSYM: 5725 return eStrataUnknown; // 0xAu 5726 case MH_KEXT_BUNDLE: 5727 return eStrataKernel; // 0xBu 5728 default: 5729 break; 5730 } 5731 return eStrataUnknown; 5732 } 5733 5734 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5735 ModuleSP module_sp(GetModule()); 5736 if (module_sp) { 5737 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5738 struct dylib_command load_cmd; 5739 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5740 uint32_t version_cmd = 0; 5741 uint64_t version = 0; 5742 uint32_t i; 5743 for (i = 0; i < m_header.ncmds; ++i) { 5744 const lldb::offset_t cmd_offset = offset; 5745 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5746 break; 5747 5748 if (load_cmd.cmd == LC_ID_DYLIB) { 5749 if (version_cmd == 0) { 5750 version_cmd = load_cmd.cmd; 5751 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr) 5752 break; 5753 version = load_cmd.dylib.current_version; 5754 } 5755 break; // Break for now unless there is another more complete version 5756 // number load command in the future. 5757 } 5758 offset = cmd_offset + load_cmd.cmdsize; 5759 } 5760 5761 if (version_cmd == LC_ID_DYLIB) { 5762 unsigned major = (version & 0xFFFF0000ull) >> 16; 5763 unsigned minor = (version & 0x0000FF00ull) >> 8; 5764 unsigned subminor = (version & 0x000000FFull); 5765 return llvm::VersionTuple(major, minor, subminor); 5766 } 5767 } 5768 return llvm::VersionTuple(); 5769 } 5770 5771 ArchSpec ObjectFileMachO::GetArchitecture() { 5772 ModuleSP module_sp(GetModule()); 5773 ArchSpec arch; 5774 if (module_sp) { 5775 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5776 5777 return GetArchitecture(module_sp, m_header, m_data, 5778 MachHeaderSizeFromMagic(m_header.magic)); 5779 } 5780 return arch; 5781 } 5782 5783 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, 5784 addr_t &base_addr, UUID &uuid) { 5785 uuid.Clear(); 5786 base_addr = LLDB_INVALID_ADDRESS; 5787 if (process && process->GetDynamicLoader()) { 5788 DynamicLoader *dl = process->GetDynamicLoader(); 5789 LazyBool using_shared_cache; 5790 LazyBool private_shared_cache; 5791 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5792 private_shared_cache); 5793 } 5794 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5795 LIBLLDB_LOG_PROCESS)); 5796 LLDB_LOGF( 5797 log, 5798 "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64, 5799 uuid.GetAsString().c_str(), base_addr); 5800 } 5801 5802 // From dyld SPI header dyld_process_info.h 5803 typedef void *dyld_process_info; 5804 struct lldb_copy__dyld_process_cache_info { 5805 uuid_t cacheUUID; // UUID of cache used by process 5806 uint64_t cacheBaseAddress; // load address of dyld shared cache 5807 bool noCache; // process is running without a dyld cache 5808 bool privateCache; // process is using a private copy of its dyld cache 5809 }; 5810 5811 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with 5812 // llvm enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile 5813 // errors. So we need to use the actual underlying types of task_t and 5814 // kern_return_t below. 5815 extern "C" unsigned int /*task_t*/ mach_task_self(); 5816 5817 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5818 uuid.Clear(); 5819 base_addr = LLDB_INVALID_ADDRESS; 5820 5821 #if defined(__APPLE__) 5822 uint8_t *(*dyld_get_all_image_infos)(void); 5823 dyld_get_all_image_infos = 5824 (uint8_t * (*)()) dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5825 if (dyld_get_all_image_infos) { 5826 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5827 if (dyld_all_image_infos_address) { 5828 uint32_t *version = (uint32_t *) 5829 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5830 if (*version >= 13) { 5831 uuid_t *sharedCacheUUID_address = 0; 5832 int wordsize = sizeof(uint8_t *); 5833 if (wordsize == 8) { 5834 sharedCacheUUID_address = 5835 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5836 160); // sharedCacheUUID <mach-o/dyld_images.h> 5837 if (*version >= 15) 5838 base_addr = 5839 *(uint64_t 5840 *)((uint8_t *)dyld_all_image_infos_address + 5841 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5842 } else { 5843 sharedCacheUUID_address = 5844 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5845 84); // sharedCacheUUID <mach-o/dyld_images.h> 5846 if (*version >= 15) { 5847 base_addr = 0; 5848 base_addr = 5849 *(uint32_t 5850 *)((uint8_t *)dyld_all_image_infos_address + 5851 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5852 } 5853 } 5854 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5855 } 5856 } 5857 } else { 5858 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5859 dyld_process_info (*dyld_process_info_create)( 5860 unsigned int /* task_t */ task, uint64_t timestamp, 5861 unsigned int /*kern_return_t*/ *kernelError); 5862 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5863 void (*dyld_process_info_release)(dyld_process_info info); 5864 5865 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, 5866 unsigned int /*kern_return_t*/ *)) 5867 dlsym(RTLD_DEFAULT, "_dyld_process_info_create"); 5868 dyld_process_info_get_cache = (void (*)(void *, void *))dlsym( 5869 RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5870 dyld_process_info_release = 5871 (void (*)(void *))dlsym(RTLD_DEFAULT, "_dyld_process_info_release"); 5872 5873 if (dyld_process_info_create && dyld_process_info_get_cache) { 5874 unsigned int /*kern_return_t */ kern_ret; 5875 dyld_process_info process_info = 5876 dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5877 if (process_info) { 5878 struct lldb_copy__dyld_process_cache_info sc_info; 5879 memset(&sc_info, 0, sizeof(struct lldb_copy__dyld_process_cache_info)); 5880 dyld_process_info_get_cache(process_info, &sc_info); 5881 if (sc_info.cacheBaseAddress != 0) { 5882 base_addr = sc_info.cacheBaseAddress; 5883 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 5884 } 5885 dyld_process_info_release(process_info); 5886 } 5887 } 5888 } 5889 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5890 LIBLLDB_LOG_PROCESS)); 5891 if (log && uuid.IsValid()) 5892 LLDB_LOGF(log, 5893 "lldb's in-memory shared cache has a UUID of %s base address of " 5894 "0x%" PRIx64, 5895 uuid.GetAsString().c_str(), base_addr); 5896 #endif 5897 } 5898 5899 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5900 if (!m_min_os_version) { 5901 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5902 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5903 const lldb::offset_t load_cmd_offset = offset; 5904 5905 version_min_command lc; 5906 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5907 break; 5908 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5909 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5910 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5911 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5912 if (m_data.GetU32(&offset, &lc.version, 5913 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5914 const uint32_t xxxx = lc.version >> 16; 5915 const uint32_t yy = (lc.version >> 8) & 0xffu; 5916 const uint32_t zz = lc.version & 0xffu; 5917 if (xxxx) { 5918 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5919 break; 5920 } 5921 } 5922 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5923 // struct build_version_command { 5924 // uint32_t cmd; /* LC_BUILD_VERSION */ 5925 // uint32_t cmdsize; /* sizeof(struct 5926 // build_version_command) plus */ 5927 // /* ntools * sizeof(struct 5928 // build_tool_version) */ 5929 // uint32_t platform; /* platform */ 5930 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5931 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded in 5932 // nibbles xxxx.yy.zz */ uint32_t ntools; /* number of 5933 // tool entries following this */ 5934 // }; 5935 5936 offset += 4; // skip platform 5937 uint32_t minos = m_data.GetU32(&offset); 5938 5939 const uint32_t xxxx = minos >> 16; 5940 const uint32_t yy = (minos >> 8) & 0xffu; 5941 const uint32_t zz = minos & 0xffu; 5942 if (xxxx) { 5943 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5944 break; 5945 } 5946 } 5947 5948 offset = load_cmd_offset + lc.cmdsize; 5949 } 5950 5951 if (!m_min_os_version) { 5952 // Set version to an empty value so we don't keep trying to 5953 m_min_os_version = llvm::VersionTuple(); 5954 } 5955 } 5956 5957 return *m_min_os_version; 5958 } 5959 5960 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { 5961 if (!m_sdk_versions.hasValue()) { 5962 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5963 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5964 const lldb::offset_t load_cmd_offset = offset; 5965 5966 version_min_command lc; 5967 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5968 break; 5969 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5970 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5971 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5972 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5973 if (m_data.GetU32(&offset, &lc.version, 5974 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5975 const uint32_t xxxx = lc.sdk >> 16; 5976 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5977 const uint32_t zz = lc.sdk & 0xffu; 5978 if (xxxx) { 5979 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 5980 break; 5981 } else { 5982 GetModule()->ReportWarning("minimum OS version load command with " 5983 "invalid (0) version found."); 5984 } 5985 } 5986 } 5987 offset = load_cmd_offset + lc.cmdsize; 5988 } 5989 5990 if (!m_sdk_versions.hasValue()) { 5991 offset = MachHeaderSizeFromMagic(m_header.magic); 5992 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5993 const lldb::offset_t load_cmd_offset = offset; 5994 5995 version_min_command lc; 5996 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5997 break; 5998 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5999 // struct build_version_command { 6000 // uint32_t cmd; /* LC_BUILD_VERSION */ 6001 // uint32_t cmdsize; /* sizeof(struct 6002 // build_version_command) plus */ 6003 // /* ntools * sizeof(struct 6004 // build_tool_version) */ 6005 // uint32_t platform; /* platform */ 6006 // uint32_t minos; /* X.Y.Z is encoded in nibbles 6007 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded 6008 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number 6009 // of tool entries following this */ 6010 // }; 6011 6012 offset += 4; // skip platform 6013 uint32_t minos = m_data.GetU32(&offset); 6014 6015 const uint32_t xxxx = minos >> 16; 6016 const uint32_t yy = (minos >> 8) & 0xffu; 6017 const uint32_t zz = minos & 0xffu; 6018 if (xxxx) { 6019 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 6020 break; 6021 } 6022 } 6023 offset = load_cmd_offset + lc.cmdsize; 6024 } 6025 } 6026 6027 if (!m_sdk_versions.hasValue()) 6028 m_sdk_versions = llvm::VersionTuple(); 6029 } 6030 6031 return m_sdk_versions.getValue(); 6032 } 6033 6034 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 6035 return m_header.filetype == llvm::MachO::MH_DYLINKER; 6036 } 6037 6038 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 6039 return m_allow_assembly_emulation_unwind_plans; 6040 } 6041 6042 // PluginInterface protocol 6043 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 6044 return GetPluginNameStatic(); 6045 } 6046 6047 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 6048 6049 Section *ObjectFileMachO::GetMachHeaderSection() { 6050 // Find the first address of the mach header which is the first non-zero file 6051 // sized section whose file offset is zero. This is the base file address of 6052 // the mach-o file which can be subtracted from the vmaddr of the other 6053 // segments found in memory and added to the load address 6054 ModuleSP module_sp = GetModule(); 6055 if (!module_sp) 6056 return nullptr; 6057 SectionList *section_list = GetSectionList(); 6058 if (!section_list) 6059 return nullptr; 6060 const size_t num_sections = section_list->GetSize(); 6061 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6062 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 6063 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 6064 return section; 6065 } 6066 return nullptr; 6067 } 6068 6069 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 6070 if (!section) 6071 return false; 6072 const bool is_dsym = (m_header.filetype == MH_DSYM); 6073 if (section->GetFileSize() == 0 && !is_dsym) 6074 return false; 6075 if (section->IsThreadSpecific()) 6076 return false; 6077 if (GetModule().get() != section->GetModule().get()) 6078 return false; 6079 // Be careful with __LINKEDIT and __DWARF segments 6080 if (section->GetName() == GetSegmentNameLINKEDIT() || 6081 section->GetName() == GetSegmentNameDWARF()) { 6082 // Only map __LINKEDIT and __DWARF if we have an in memory image and 6083 // this isn't a kernel binary like a kext or mach_kernel. 6084 const bool is_memory_image = (bool)m_process_wp.lock(); 6085 const Strata strata = GetStrata(); 6086 if (is_memory_image == false || strata == eStrataKernel) 6087 return false; 6088 } 6089 return true; 6090 } 6091 6092 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 6093 lldb::addr_t header_load_address, const Section *header_section, 6094 const Section *section) { 6095 ModuleSP module_sp = GetModule(); 6096 if (module_sp && header_section && section && 6097 header_load_address != LLDB_INVALID_ADDRESS) { 6098 lldb::addr_t file_addr = header_section->GetFileAddress(); 6099 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 6100 return section->GetFileAddress() - file_addr + header_load_address; 6101 } 6102 return LLDB_INVALID_ADDRESS; 6103 } 6104 6105 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 6106 bool value_is_offset) { 6107 ModuleSP module_sp = GetModule(); 6108 if (!module_sp) 6109 return false; 6110 6111 SectionList *section_list = GetSectionList(); 6112 if (!section_list) 6113 return false; 6114 6115 size_t num_loaded_sections = 0; 6116 const size_t num_sections = section_list->GetSize(); 6117 6118 if (value_is_offset) { 6119 // "value" is an offset to apply to each top level segment 6120 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6121 // Iterate through the object file sections to find all of the 6122 // sections that size on disk (to avoid __PAGEZERO) and load them 6123 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6124 if (SectionIsLoadable(section_sp.get())) 6125 if (target.GetSectionLoadList().SetSectionLoadAddress( 6126 section_sp, section_sp->GetFileAddress() + value)) 6127 ++num_loaded_sections; 6128 } 6129 } else { 6130 // "value" is the new base address of the mach_header, adjust each 6131 // section accordingly 6132 6133 Section *mach_header_section = GetMachHeaderSection(); 6134 if (mach_header_section) { 6135 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6136 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6137 6138 lldb::addr_t section_load_addr = 6139 CalculateSectionLoadAddressForMemoryImage( 6140 value, mach_header_section, section_sp.get()); 6141 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6142 if (target.GetSectionLoadList().SetSectionLoadAddress( 6143 section_sp, section_load_addr)) 6144 ++num_loaded_sections; 6145 } 6146 } 6147 } 6148 } 6149 return num_loaded_sections > 0; 6150 } 6151 6152 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6153 const FileSpec &outfile, Status &error) { 6154 if (!process_sp) 6155 return false; 6156 6157 Target &target = process_sp->GetTarget(); 6158 const ArchSpec target_arch = target.GetArchitecture(); 6159 const llvm::Triple &target_triple = target_arch.GetTriple(); 6160 if (target_triple.getVendor() == llvm::Triple::Apple && 6161 (target_triple.getOS() == llvm::Triple::MacOSX || 6162 target_triple.getOS() == llvm::Triple::IOS || 6163 target_triple.getOS() == llvm::Triple::WatchOS || 6164 target_triple.getOS() == llvm::Triple::TvOS)) { 6165 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) 6166 // { 6167 bool make_core = false; 6168 switch (target_arch.GetMachine()) { 6169 case llvm::Triple::aarch64: 6170 case llvm::Triple::aarch64_32: 6171 case llvm::Triple::arm: 6172 case llvm::Triple::thumb: 6173 case llvm::Triple::x86: 6174 case llvm::Triple::x86_64: 6175 make_core = true; 6176 break; 6177 default: 6178 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6179 target_triple.str().c_str()); 6180 break; 6181 } 6182 6183 if (make_core) { 6184 std::vector<segment_command_64> segment_load_commands; 6185 // uint32_t range_info_idx = 0; 6186 MemoryRegionInfo range_info; 6187 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6188 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6189 const ByteOrder byte_order = target_arch.GetByteOrder(); 6190 if (range_error.Success()) { 6191 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6192 const addr_t addr = range_info.GetRange().GetRangeBase(); 6193 const addr_t size = range_info.GetRange().GetByteSize(); 6194 6195 if (size == 0) 6196 break; 6197 6198 // Calculate correct protections 6199 uint32_t prot = 0; 6200 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6201 prot |= VM_PROT_READ; 6202 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6203 prot |= VM_PROT_WRITE; 6204 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6205 prot |= VM_PROT_EXECUTE; 6206 6207 if (prot != 0) { 6208 uint32_t cmd_type = LC_SEGMENT_64; 6209 uint32_t segment_size = sizeof(segment_command_64); 6210 if (addr_byte_size == 4) { 6211 cmd_type = LC_SEGMENT; 6212 segment_size = sizeof(segment_command); 6213 } 6214 segment_command_64 segment = { 6215 cmd_type, // uint32_t cmd; 6216 segment_size, // uint32_t cmdsize; 6217 {0}, // char segname[16]; 6218 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6219 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6220 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6221 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6222 prot, // uint32_t maxprot; 6223 prot, // uint32_t initprot; 6224 0, // uint32_t nsects; 6225 0}; // uint32_t flags; 6226 segment_load_commands.push_back(segment); 6227 } else { 6228 // No protections and a size of 1 used to be returned from old 6229 // debugservers when we asked about a region that was past the 6230 // last memory region and it indicates the end... 6231 if (size == 1) 6232 break; 6233 } 6234 6235 range_error = process_sp->GetMemoryRegionInfo( 6236 range_info.GetRange().GetRangeEnd(), range_info); 6237 if (range_error.Fail()) 6238 break; 6239 } 6240 6241 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6242 6243 mach_header_64 mach_header; 6244 if (addr_byte_size == 8) { 6245 mach_header.magic = MH_MAGIC_64; 6246 } else { 6247 mach_header.magic = MH_MAGIC; 6248 } 6249 mach_header.cputype = target_arch.GetMachOCPUType(); 6250 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6251 mach_header.filetype = MH_CORE; 6252 mach_header.ncmds = segment_load_commands.size(); 6253 mach_header.flags = 0; 6254 mach_header.reserved = 0; 6255 ThreadList &thread_list = process_sp->GetThreadList(); 6256 const uint32_t num_threads = thread_list.GetSize(); 6257 6258 // Make an array of LC_THREAD data items. Each one contains the 6259 // contents of the LC_THREAD load command. The data doesn't contain 6260 // the load command + load command size, we will add the load command 6261 // and load command size as we emit the data. 6262 std::vector<StreamString> LC_THREAD_datas(num_threads); 6263 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6264 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6265 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6266 LC_THREAD_data.SetByteOrder(byte_order); 6267 } 6268 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) { 6269 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6270 if (thread_sp) { 6271 switch (mach_header.cputype) { 6272 case llvm::MachO::CPU_TYPE_ARM64: 6273 case llvm::MachO::CPU_TYPE_ARM64_32: 6274 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6275 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6276 break; 6277 6278 case llvm::MachO::CPU_TYPE_ARM: 6279 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6280 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6281 break; 6282 6283 case llvm::MachO::CPU_TYPE_I386: 6284 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6285 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6286 break; 6287 6288 case llvm::MachO::CPU_TYPE_X86_64: 6289 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6290 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6291 break; 6292 } 6293 } 6294 } 6295 6296 // The size of the load command is the size of the segments... 6297 if (addr_byte_size == 8) { 6298 mach_header.sizeofcmds = 6299 segment_load_commands.size() * sizeof(struct segment_command_64); 6300 } else { 6301 mach_header.sizeofcmds = 6302 segment_load_commands.size() * sizeof(struct segment_command); 6303 } 6304 6305 // and the size of all LC_THREAD load command 6306 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6307 ++mach_header.ncmds; 6308 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6309 } 6310 6311 // Write the mach header 6312 buffer.PutHex32(mach_header.magic); 6313 buffer.PutHex32(mach_header.cputype); 6314 buffer.PutHex32(mach_header.cpusubtype); 6315 buffer.PutHex32(mach_header.filetype); 6316 buffer.PutHex32(mach_header.ncmds); 6317 buffer.PutHex32(mach_header.sizeofcmds); 6318 buffer.PutHex32(mach_header.flags); 6319 if (addr_byte_size == 8) { 6320 buffer.PutHex32(mach_header.reserved); 6321 } 6322 6323 // Skip the mach header and all load commands and align to the next 6324 // 0x1000 byte boundary 6325 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6326 if (file_offset & 0x00000fff) { 6327 file_offset += 0x00001000ull; 6328 file_offset &= (~0x00001000ull + 1); 6329 } 6330 6331 for (auto &segment : segment_load_commands) { 6332 segment.fileoff = file_offset; 6333 file_offset += segment.filesize; 6334 } 6335 6336 // Write out all of the LC_THREAD load commands 6337 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6338 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6339 buffer.PutHex32(LC_THREAD); 6340 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6341 buffer.Write(LC_THREAD_data.GetString().data(), LC_THREAD_data_size); 6342 } 6343 6344 // Write out all of the segment load commands 6345 for (const auto &segment : segment_load_commands) { 6346 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6347 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6348 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6349 segment.cmd, segment.cmdsize, segment.vmaddr, 6350 segment.vmaddr + segment.vmsize, segment.fileoff, 6351 segment.filesize, segment.maxprot, segment.initprot, 6352 segment.nsects, segment.flags); 6353 6354 buffer.PutHex32(segment.cmd); 6355 buffer.PutHex32(segment.cmdsize); 6356 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6357 if (addr_byte_size == 8) { 6358 buffer.PutHex64(segment.vmaddr); 6359 buffer.PutHex64(segment.vmsize); 6360 buffer.PutHex64(segment.fileoff); 6361 buffer.PutHex64(segment.filesize); 6362 } else { 6363 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6364 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6365 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6366 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6367 } 6368 buffer.PutHex32(segment.maxprot); 6369 buffer.PutHex32(segment.initprot); 6370 buffer.PutHex32(segment.nsects); 6371 buffer.PutHex32(segment.flags); 6372 } 6373 6374 std::string core_file_path(outfile.GetPath()); 6375 auto core_file = FileSystem::Instance().Open( 6376 outfile, File::eOpenOptionWrite | File::eOpenOptionTruncate | 6377 File::eOpenOptionCanCreate); 6378 if (!core_file) { 6379 error = core_file.takeError(); 6380 } else { 6381 // Read 1 page at a time 6382 uint8_t bytes[0x1000]; 6383 // Write the mach header and load commands out to the core file 6384 size_t bytes_written = buffer.GetString().size(); 6385 error = 6386 core_file.get()->Write(buffer.GetString().data(), bytes_written); 6387 if (error.Success()) { 6388 // Now write the file data for all memory segments in the process 6389 for (const auto &segment : segment_load_commands) { 6390 if (core_file.get()->SeekFromStart(segment.fileoff) == -1) { 6391 error.SetErrorStringWithFormat( 6392 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6393 segment.fileoff, core_file_path.c_str()); 6394 break; 6395 } 6396 6397 printf("Saving %" PRId64 6398 " bytes of data for memory region at 0x%" PRIx64 "\n", 6399 segment.vmsize, segment.vmaddr); 6400 addr_t bytes_left = segment.vmsize; 6401 addr_t addr = segment.vmaddr; 6402 Status memory_read_error; 6403 while (bytes_left > 0 && error.Success()) { 6404 const size_t bytes_to_read = 6405 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6406 6407 // In a savecore setting, we don't really care about caching, 6408 // as the data is dumped and very likely never read again, 6409 // so we call ReadMemoryFromInferior to bypass it. 6410 const size_t bytes_read = process_sp->ReadMemoryFromInferior( 6411 addr, bytes, bytes_to_read, memory_read_error); 6412 6413 if (bytes_read == bytes_to_read) { 6414 size_t bytes_written = bytes_read; 6415 error = core_file.get()->Write(bytes, bytes_written); 6416 bytes_left -= bytes_read; 6417 addr += bytes_read; 6418 } else { 6419 // Some pages within regions are not readable, those should 6420 // be zero filled 6421 memset(bytes, 0, bytes_to_read); 6422 size_t bytes_written = bytes_to_read; 6423 error = core_file.get()->Write(bytes, bytes_written); 6424 bytes_left -= bytes_to_read; 6425 addr += bytes_to_read; 6426 } 6427 } 6428 } 6429 } 6430 } 6431 } else { 6432 error.SetErrorString( 6433 "process doesn't support getting memory region info"); 6434 } 6435 } 6436 return true; // This is the right plug to handle saving core files for 6437 // this process 6438 } 6439 return false; 6440 } 6441