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/Section.h" 21 #include "lldb/Core/StreamFile.h" 22 #include "lldb/Host/Host.h" 23 #include "lldb/Symbol/DWARFCallFrameInfo.h" 24 #include "lldb/Symbol/ObjectFile.h" 25 #include "lldb/Target/DynamicLoader.h" 26 #include "lldb/Target/MemoryRegionInfo.h" 27 #include "lldb/Target/Platform.h" 28 #include "lldb/Target/Process.h" 29 #include "lldb/Target/SectionLoadList.h" 30 #include "lldb/Target/Target.h" 31 #include "lldb/Target/Thread.h" 32 #include "lldb/Target/ThreadList.h" 33 #include "lldb/Utility/ArchSpec.h" 34 #include "lldb/Utility/DataBuffer.h" 35 #include "lldb/Utility/FileSpec.h" 36 #include "lldb/Utility/Log.h" 37 #include "lldb/Utility/RangeMap.h" 38 #include "lldb/Utility/RegisterValue.h" 39 #include "lldb/Utility/Status.h" 40 #include "lldb/Utility/StreamString.h" 41 #include "lldb/Utility/Timer.h" 42 #include "lldb/Utility/UUID.h" 43 44 #include "lldb/Host/SafeMachO.h" 45 46 #include "llvm/Support/MemoryBuffer.h" 47 48 #include "ObjectFileMachO.h" 49 50 #if defined(__APPLE__) && \ 51 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 52 // GetLLDBSharedCacheUUID() needs to call dlsym() 53 #include <dlfcn.h> 54 #endif 55 56 #ifndef __APPLE__ 57 #include "Utility/UuidCompatibility.h" 58 #else 59 #include <uuid/uuid.h> 60 #endif 61 62 #include <memory> 63 64 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 65 using namespace lldb; 66 using namespace lldb_private; 67 using namespace llvm::MachO; 68 69 LLDB_PLUGIN_DEFINE(ObjectFileMachO) 70 71 // Some structure definitions needed for parsing the dyld shared cache files 72 // found on iOS devices. 73 74 struct lldb_copy_dyld_cache_header_v1 { 75 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 76 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 77 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 78 uint32_t imagesOffset; 79 uint32_t imagesCount; 80 uint64_t dyldBaseAddress; 81 uint64_t codeSignatureOffset; 82 uint64_t codeSignatureSize; 83 uint64_t slideInfoOffset; 84 uint64_t slideInfoSize; 85 uint64_t localSymbolsOffset; 86 uint64_t localSymbolsSize; 87 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 88 // and later 89 }; 90 91 struct lldb_copy_dyld_cache_mapping_info { 92 uint64_t address; 93 uint64_t size; 94 uint64_t fileOffset; 95 uint32_t maxProt; 96 uint32_t initProt; 97 }; 98 99 struct lldb_copy_dyld_cache_local_symbols_info { 100 uint32_t nlistOffset; 101 uint32_t nlistCount; 102 uint32_t stringsOffset; 103 uint32_t stringsSize; 104 uint32_t entriesOffset; 105 uint32_t entriesCount; 106 }; 107 struct lldb_copy_dyld_cache_local_symbols_entry { 108 uint32_t dylibOffset; 109 uint32_t nlistStartIndex; 110 uint32_t nlistCount; 111 }; 112 113 static void PrintRegisterValue(RegisterContext *reg_ctx, const char *name, 114 const char *alt_name, size_t reg_byte_size, 115 Stream &data) { 116 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 117 if (reg_info == nullptr) 118 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 119 if (reg_info) { 120 lldb_private::RegisterValue reg_value; 121 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 122 if (reg_info->byte_size >= reg_byte_size) 123 data.Write(reg_value.GetBytes(), reg_byte_size); 124 else { 125 data.Write(reg_value.GetBytes(), reg_info->byte_size); 126 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; ++i) 127 data.PutChar(0); 128 } 129 return; 130 } 131 } 132 // Just write zeros if all else fails 133 for (size_t i = 0; i < reg_byte_size; ++i) 134 data.PutChar(0); 135 } 136 137 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 138 public: 139 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 140 const DataExtractor &data) 141 : RegisterContextDarwin_x86_64(thread, 0) { 142 SetRegisterDataFrom_LC_THREAD(data); 143 } 144 145 void InvalidateAllRegisters() override { 146 // Do nothing... registers are always valid... 147 } 148 149 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 150 lldb::offset_t offset = 0; 151 SetError(GPRRegSet, Read, -1); 152 SetError(FPURegSet, Read, -1); 153 SetError(EXCRegSet, Read, -1); 154 bool done = false; 155 156 while (!done) { 157 int flavor = data.GetU32(&offset); 158 if (flavor == 0) 159 done = true; 160 else { 161 uint32_t i; 162 uint32_t count = data.GetU32(&offset); 163 switch (flavor) { 164 case GPRRegSet: 165 for (i = 0; i < count; ++i) 166 (&gpr.rax)[i] = data.GetU64(&offset); 167 SetError(GPRRegSet, Read, 0); 168 done = true; 169 170 break; 171 case FPURegSet: 172 // TODO: fill in FPU regs.... 173 // SetError (FPURegSet, Read, -1); 174 done = true; 175 176 break; 177 case EXCRegSet: 178 exc.trapno = data.GetU32(&offset); 179 exc.err = data.GetU32(&offset); 180 exc.faultvaddr = data.GetU64(&offset); 181 SetError(EXCRegSet, Read, 0); 182 done = true; 183 break; 184 case 7: 185 case 8: 186 case 9: 187 // fancy flavors that encapsulate of the above flavors... 188 break; 189 190 default: 191 done = true; 192 break; 193 } 194 } 195 } 196 } 197 198 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 199 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 200 if (reg_ctx_sp) { 201 RegisterContext *reg_ctx = reg_ctx_sp.get(); 202 203 data.PutHex32(GPRRegSet); // Flavor 204 data.PutHex32(GPRWordCount); 205 PrintRegisterValue(reg_ctx, "rax", nullptr, 8, data); 206 PrintRegisterValue(reg_ctx, "rbx", nullptr, 8, data); 207 PrintRegisterValue(reg_ctx, "rcx", nullptr, 8, data); 208 PrintRegisterValue(reg_ctx, "rdx", nullptr, 8, data); 209 PrintRegisterValue(reg_ctx, "rdi", nullptr, 8, data); 210 PrintRegisterValue(reg_ctx, "rsi", nullptr, 8, data); 211 PrintRegisterValue(reg_ctx, "rbp", nullptr, 8, data); 212 PrintRegisterValue(reg_ctx, "rsp", nullptr, 8, data); 213 PrintRegisterValue(reg_ctx, "r8", nullptr, 8, data); 214 PrintRegisterValue(reg_ctx, "r9", nullptr, 8, data); 215 PrintRegisterValue(reg_ctx, "r10", nullptr, 8, data); 216 PrintRegisterValue(reg_ctx, "r11", nullptr, 8, data); 217 PrintRegisterValue(reg_ctx, "r12", nullptr, 8, data); 218 PrintRegisterValue(reg_ctx, "r13", nullptr, 8, data); 219 PrintRegisterValue(reg_ctx, "r14", nullptr, 8, data); 220 PrintRegisterValue(reg_ctx, "r15", nullptr, 8, data); 221 PrintRegisterValue(reg_ctx, "rip", nullptr, 8, data); 222 PrintRegisterValue(reg_ctx, "rflags", nullptr, 8, data); 223 PrintRegisterValue(reg_ctx, "cs", nullptr, 8, data); 224 PrintRegisterValue(reg_ctx, "fs", nullptr, 8, data); 225 PrintRegisterValue(reg_ctx, "gs", nullptr, 8, data); 226 227 // // Write out the FPU registers 228 // const size_t fpu_byte_size = sizeof(FPU); 229 // size_t bytes_written = 0; 230 // data.PutHex32 (FPURegSet); 231 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 232 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 233 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 234 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 235 // data); // uint16_t fcw; // "fctrl" 236 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 237 // data); // uint16_t fsw; // "fstat" 238 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 239 // data); // uint8_t ftw; // "ftag" 240 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 241 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 242 // data); // uint16_t fop; // "fop" 243 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 244 // data); // uint32_t ip; // "fioff" 245 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 246 // data); // uint16_t cs; // "fiseg" 247 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 248 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 249 // data); // uint32_t dp; // "fooff" 250 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 251 // data); // uint16_t ds; // "foseg" 252 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 253 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 254 // data); // uint32_t mxcsr; 255 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 256 // 4, data);// uint32_t mxcsrmask; 257 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 258 // sizeof(MMSReg), data); 259 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 260 // sizeof(MMSReg), data); 261 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 262 // sizeof(MMSReg), data); 263 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 264 // sizeof(MMSReg), data); 265 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 266 // sizeof(MMSReg), data); 267 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 268 // sizeof(MMSReg), data); 269 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 270 // sizeof(MMSReg), data); 271 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 272 // sizeof(MMSReg), data); 273 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 274 // sizeof(XMMReg), data); 275 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 276 // sizeof(XMMReg), data); 277 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 278 // sizeof(XMMReg), data); 279 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 280 // sizeof(XMMReg), data); 281 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 282 // sizeof(XMMReg), data); 283 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 284 // sizeof(XMMReg), data); 285 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 286 // sizeof(XMMReg), data); 287 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 288 // sizeof(XMMReg), data); 289 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 290 // sizeof(XMMReg), data); 291 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 292 // sizeof(XMMReg), data); 293 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 294 // sizeof(XMMReg), data); 295 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 296 // sizeof(XMMReg), data); 297 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 298 // sizeof(XMMReg), data); 299 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 300 // sizeof(XMMReg), data); 301 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 302 // sizeof(XMMReg), data); 303 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 304 // sizeof(XMMReg), data); 305 // 306 // // Fill rest with zeros 307 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 308 // i) 309 // data.PutChar(0); 310 311 // Write out the EXC registers 312 data.PutHex32(EXCRegSet); 313 data.PutHex32(EXCWordCount); 314 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 315 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 316 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 8, data); 317 return true; 318 } 319 return false; 320 } 321 322 protected: 323 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 324 325 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 326 327 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 328 329 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 330 return 0; 331 } 332 333 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 334 return 0; 335 } 336 337 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 338 return 0; 339 } 340 }; 341 342 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 343 public: 344 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 345 const DataExtractor &data) 346 : RegisterContextDarwin_i386(thread, 0) { 347 SetRegisterDataFrom_LC_THREAD(data); 348 } 349 350 void InvalidateAllRegisters() override { 351 // Do nothing... registers are always valid... 352 } 353 354 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 355 lldb::offset_t offset = 0; 356 SetError(GPRRegSet, Read, -1); 357 SetError(FPURegSet, Read, -1); 358 SetError(EXCRegSet, Read, -1); 359 bool done = false; 360 361 while (!done) { 362 int flavor = data.GetU32(&offset); 363 if (flavor == 0) 364 done = true; 365 else { 366 uint32_t i; 367 uint32_t count = data.GetU32(&offset); 368 switch (flavor) { 369 case GPRRegSet: 370 for (i = 0; i < count; ++i) 371 (&gpr.eax)[i] = data.GetU32(&offset); 372 SetError(GPRRegSet, Read, 0); 373 done = true; 374 375 break; 376 case FPURegSet: 377 // TODO: fill in FPU regs.... 378 // SetError (FPURegSet, Read, -1); 379 done = true; 380 381 break; 382 case EXCRegSet: 383 exc.trapno = data.GetU32(&offset); 384 exc.err = data.GetU32(&offset); 385 exc.faultvaddr = data.GetU32(&offset); 386 SetError(EXCRegSet, Read, 0); 387 done = true; 388 break; 389 case 7: 390 case 8: 391 case 9: 392 // fancy flavors that encapsulate of the above flavors... 393 break; 394 395 default: 396 done = true; 397 break; 398 } 399 } 400 } 401 } 402 403 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 404 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 405 if (reg_ctx_sp) { 406 RegisterContext *reg_ctx = reg_ctx_sp.get(); 407 408 data.PutHex32(GPRRegSet); // Flavor 409 data.PutHex32(GPRWordCount); 410 PrintRegisterValue(reg_ctx, "eax", nullptr, 4, data); 411 PrintRegisterValue(reg_ctx, "ebx", nullptr, 4, data); 412 PrintRegisterValue(reg_ctx, "ecx", nullptr, 4, data); 413 PrintRegisterValue(reg_ctx, "edx", nullptr, 4, data); 414 PrintRegisterValue(reg_ctx, "edi", nullptr, 4, data); 415 PrintRegisterValue(reg_ctx, "esi", nullptr, 4, data); 416 PrintRegisterValue(reg_ctx, "ebp", nullptr, 4, data); 417 PrintRegisterValue(reg_ctx, "esp", nullptr, 4, data); 418 PrintRegisterValue(reg_ctx, "ss", nullptr, 4, data); 419 PrintRegisterValue(reg_ctx, "eflags", nullptr, 4, data); 420 PrintRegisterValue(reg_ctx, "eip", nullptr, 4, data); 421 PrintRegisterValue(reg_ctx, "cs", nullptr, 4, data); 422 PrintRegisterValue(reg_ctx, "ds", nullptr, 4, data); 423 PrintRegisterValue(reg_ctx, "es", nullptr, 4, data); 424 PrintRegisterValue(reg_ctx, "fs", nullptr, 4, data); 425 PrintRegisterValue(reg_ctx, "gs", nullptr, 4, data); 426 427 // Write out the EXC registers 428 data.PutHex32(EXCRegSet); 429 data.PutHex32(EXCWordCount); 430 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 431 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 432 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 4, data); 433 return true; 434 } 435 return false; 436 } 437 438 protected: 439 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 440 441 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 442 443 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 444 445 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 446 return 0; 447 } 448 449 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 450 return 0; 451 } 452 453 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 454 return 0; 455 } 456 }; 457 458 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 459 public: 460 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 461 const DataExtractor &data) 462 : RegisterContextDarwin_arm(thread, 0) { 463 SetRegisterDataFrom_LC_THREAD(data); 464 } 465 466 void InvalidateAllRegisters() override { 467 // Do nothing... registers are always valid... 468 } 469 470 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 471 lldb::offset_t offset = 0; 472 SetError(GPRRegSet, Read, -1); 473 SetError(FPURegSet, Read, -1); 474 SetError(EXCRegSet, Read, -1); 475 bool done = false; 476 477 while (!done) { 478 int flavor = data.GetU32(&offset); 479 uint32_t count = data.GetU32(&offset); 480 lldb::offset_t next_thread_state = offset + (count * 4); 481 switch (flavor) { 482 case GPRAltRegSet: 483 case GPRRegSet: 484 // On ARM, the CPSR register is also included in the count but it is 485 // not included in gpr.r so loop until (count-1). 486 for (uint32_t i = 0; i < (count - 1); ++i) { 487 gpr.r[i] = data.GetU32(&offset); 488 } 489 // Save cpsr explicitly. 490 gpr.cpsr = data.GetU32(&offset); 491 492 SetError(GPRRegSet, Read, 0); 493 offset = next_thread_state; 494 break; 495 496 case FPURegSet: { 497 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 498 const int fpu_reg_buf_size = sizeof(fpu.floats); 499 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 500 fpu_reg_buf) == fpu_reg_buf_size) { 501 offset += fpu_reg_buf_size; 502 fpu.fpscr = data.GetU32(&offset); 503 SetError(FPURegSet, Read, 0); 504 } else { 505 done = true; 506 } 507 } 508 offset = next_thread_state; 509 break; 510 511 case EXCRegSet: 512 if (count == 3) { 513 exc.exception = data.GetU32(&offset); 514 exc.fsr = data.GetU32(&offset); 515 exc.far = data.GetU32(&offset); 516 SetError(EXCRegSet, Read, 0); 517 } 518 done = true; 519 offset = next_thread_state; 520 break; 521 522 // Unknown register set flavor, stop trying to parse. 523 default: 524 done = true; 525 } 526 } 527 } 528 529 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 530 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 531 if (reg_ctx_sp) { 532 RegisterContext *reg_ctx = reg_ctx_sp.get(); 533 534 data.PutHex32(GPRRegSet); // Flavor 535 data.PutHex32(GPRWordCount); 536 PrintRegisterValue(reg_ctx, "r0", nullptr, 4, data); 537 PrintRegisterValue(reg_ctx, "r1", nullptr, 4, data); 538 PrintRegisterValue(reg_ctx, "r2", nullptr, 4, data); 539 PrintRegisterValue(reg_ctx, "r3", nullptr, 4, data); 540 PrintRegisterValue(reg_ctx, "r4", nullptr, 4, data); 541 PrintRegisterValue(reg_ctx, "r5", nullptr, 4, data); 542 PrintRegisterValue(reg_ctx, "r6", nullptr, 4, data); 543 PrintRegisterValue(reg_ctx, "r7", nullptr, 4, data); 544 PrintRegisterValue(reg_ctx, "r8", nullptr, 4, data); 545 PrintRegisterValue(reg_ctx, "r9", nullptr, 4, data); 546 PrintRegisterValue(reg_ctx, "r10", nullptr, 4, data); 547 PrintRegisterValue(reg_ctx, "r11", nullptr, 4, data); 548 PrintRegisterValue(reg_ctx, "r12", nullptr, 4, data); 549 PrintRegisterValue(reg_ctx, "sp", nullptr, 4, data); 550 PrintRegisterValue(reg_ctx, "lr", nullptr, 4, data); 551 PrintRegisterValue(reg_ctx, "pc", nullptr, 4, data); 552 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 553 554 // Write out the EXC registers 555 // data.PutHex32 (EXCRegSet); 556 // data.PutHex32 (EXCWordCount); 557 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 558 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 559 // WriteRegister (reg_ctx, "far", NULL, 4, data); 560 return true; 561 } 562 return false; 563 } 564 565 protected: 566 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 567 568 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 569 570 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 571 572 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 573 574 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 575 return 0; 576 } 577 578 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 579 return 0; 580 } 581 582 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 583 return 0; 584 } 585 586 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 587 return -1; 588 } 589 }; 590 591 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 592 public: 593 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 594 const DataExtractor &data) 595 : RegisterContextDarwin_arm64(thread, 0) { 596 SetRegisterDataFrom_LC_THREAD(data); 597 } 598 599 void InvalidateAllRegisters() override { 600 // Do nothing... registers are always valid... 601 } 602 603 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 604 lldb::offset_t offset = 0; 605 SetError(GPRRegSet, Read, -1); 606 SetError(FPURegSet, Read, -1); 607 SetError(EXCRegSet, Read, -1); 608 bool done = false; 609 while (!done) { 610 int flavor = data.GetU32(&offset); 611 uint32_t count = data.GetU32(&offset); 612 lldb::offset_t next_thread_state = offset + (count * 4); 613 switch (flavor) { 614 case GPRRegSet: 615 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 616 // 32-bit register) 617 if (count >= (33 * 2) + 1) { 618 for (uint32_t i = 0; i < 29; ++i) 619 gpr.x[i] = data.GetU64(&offset); 620 gpr.fp = data.GetU64(&offset); 621 gpr.lr = data.GetU64(&offset); 622 gpr.sp = data.GetU64(&offset); 623 gpr.pc = data.GetU64(&offset); 624 gpr.cpsr = data.GetU32(&offset); 625 SetError(GPRRegSet, Read, 0); 626 } 627 offset = next_thread_state; 628 break; 629 case FPURegSet: { 630 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 631 const int fpu_reg_buf_size = sizeof(fpu); 632 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 633 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 634 fpu_reg_buf) == fpu_reg_buf_size) { 635 SetError(FPURegSet, Read, 0); 636 } else { 637 done = true; 638 } 639 } 640 offset = next_thread_state; 641 break; 642 case EXCRegSet: 643 if (count == 4) { 644 exc.far = data.GetU64(&offset); 645 exc.esr = data.GetU32(&offset); 646 exc.exception = data.GetU32(&offset); 647 SetError(EXCRegSet, Read, 0); 648 } 649 offset = next_thread_state; 650 break; 651 default: 652 done = true; 653 break; 654 } 655 } 656 } 657 658 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 659 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 660 if (reg_ctx_sp) { 661 RegisterContext *reg_ctx = reg_ctx_sp.get(); 662 663 data.PutHex32(GPRRegSet); // Flavor 664 data.PutHex32(GPRWordCount); 665 PrintRegisterValue(reg_ctx, "x0", nullptr, 8, data); 666 PrintRegisterValue(reg_ctx, "x1", nullptr, 8, data); 667 PrintRegisterValue(reg_ctx, "x2", nullptr, 8, data); 668 PrintRegisterValue(reg_ctx, "x3", nullptr, 8, data); 669 PrintRegisterValue(reg_ctx, "x4", nullptr, 8, data); 670 PrintRegisterValue(reg_ctx, "x5", nullptr, 8, data); 671 PrintRegisterValue(reg_ctx, "x6", nullptr, 8, data); 672 PrintRegisterValue(reg_ctx, "x7", nullptr, 8, data); 673 PrintRegisterValue(reg_ctx, "x8", nullptr, 8, data); 674 PrintRegisterValue(reg_ctx, "x9", nullptr, 8, data); 675 PrintRegisterValue(reg_ctx, "x10", nullptr, 8, data); 676 PrintRegisterValue(reg_ctx, "x11", nullptr, 8, data); 677 PrintRegisterValue(reg_ctx, "x12", nullptr, 8, data); 678 PrintRegisterValue(reg_ctx, "x13", nullptr, 8, data); 679 PrintRegisterValue(reg_ctx, "x14", nullptr, 8, data); 680 PrintRegisterValue(reg_ctx, "x15", nullptr, 8, data); 681 PrintRegisterValue(reg_ctx, "x16", nullptr, 8, data); 682 PrintRegisterValue(reg_ctx, "x17", nullptr, 8, data); 683 PrintRegisterValue(reg_ctx, "x18", nullptr, 8, data); 684 PrintRegisterValue(reg_ctx, "x19", nullptr, 8, data); 685 PrintRegisterValue(reg_ctx, "x20", nullptr, 8, data); 686 PrintRegisterValue(reg_ctx, "x21", nullptr, 8, data); 687 PrintRegisterValue(reg_ctx, "x22", nullptr, 8, data); 688 PrintRegisterValue(reg_ctx, "x23", nullptr, 8, data); 689 PrintRegisterValue(reg_ctx, "x24", nullptr, 8, data); 690 PrintRegisterValue(reg_ctx, "x25", nullptr, 8, data); 691 PrintRegisterValue(reg_ctx, "x26", nullptr, 8, data); 692 PrintRegisterValue(reg_ctx, "x27", nullptr, 8, data); 693 PrintRegisterValue(reg_ctx, "x28", nullptr, 8, data); 694 PrintRegisterValue(reg_ctx, "fp", nullptr, 8, data); 695 PrintRegisterValue(reg_ctx, "lr", nullptr, 8, data); 696 PrintRegisterValue(reg_ctx, "sp", nullptr, 8, data); 697 PrintRegisterValue(reg_ctx, "pc", nullptr, 8, data); 698 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 699 700 // Write out the EXC registers 701 // data.PutHex32 (EXCRegSet); 702 // data.PutHex32 (EXCWordCount); 703 // WriteRegister (reg_ctx, "far", NULL, 8, data); 704 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 705 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 706 return true; 707 } 708 return false; 709 } 710 711 protected: 712 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 713 714 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 715 716 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 717 718 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 719 720 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 721 return 0; 722 } 723 724 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 725 return 0; 726 } 727 728 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 729 return 0; 730 } 731 732 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 733 return -1; 734 } 735 }; 736 737 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 738 switch (magic) { 739 case MH_MAGIC: 740 case MH_CIGAM: 741 return sizeof(struct mach_header); 742 743 case MH_MAGIC_64: 744 case MH_CIGAM_64: 745 return sizeof(struct mach_header_64); 746 break; 747 748 default: 749 break; 750 } 751 return 0; 752 } 753 754 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 755 756 char ObjectFileMachO::ID; 757 758 void ObjectFileMachO::Initialize() { 759 PluginManager::RegisterPlugin( 760 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 761 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 762 } 763 764 void ObjectFileMachO::Terminate() { 765 PluginManager::UnregisterPlugin(CreateInstance); 766 } 767 768 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 769 static ConstString g_name("mach-o"); 770 return g_name; 771 } 772 773 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 774 return "Mach-o object file reader (32 and 64 bit)"; 775 } 776 777 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 778 DataBufferSP &data_sp, 779 lldb::offset_t data_offset, 780 const FileSpec *file, 781 lldb::offset_t file_offset, 782 lldb::offset_t length) { 783 if (!data_sp) { 784 data_sp = MapFileData(*file, length, file_offset); 785 if (!data_sp) 786 return nullptr; 787 data_offset = 0; 788 } 789 790 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 791 return nullptr; 792 793 // Update the data to contain the entire file if it doesn't already 794 if (data_sp->GetByteSize() < length) { 795 data_sp = MapFileData(*file, length, file_offset); 796 if (!data_sp) 797 return nullptr; 798 data_offset = 0; 799 } 800 auto objfile_up = std::make_unique<ObjectFileMachO>( 801 module_sp, data_sp, data_offset, file, file_offset, length); 802 if (!objfile_up || !objfile_up->ParseHeader()) 803 return nullptr; 804 805 return objfile_up.release(); 806 } 807 808 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 809 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 810 const ProcessSP &process_sp, lldb::addr_t header_addr) { 811 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 812 std::unique_ptr<ObjectFile> objfile_up( 813 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 814 if (objfile_up.get() && objfile_up->ParseHeader()) 815 return objfile_up.release(); 816 } 817 return nullptr; 818 } 819 820 size_t ObjectFileMachO::GetModuleSpecifications( 821 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 822 lldb::offset_t data_offset, lldb::offset_t file_offset, 823 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 824 const size_t initial_count = specs.GetSize(); 825 826 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 827 DataExtractor data; 828 data.SetData(data_sp); 829 llvm::MachO::mach_header header; 830 if (ParseHeader(data, &data_offset, header)) { 831 size_t header_and_load_cmds = 832 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 833 if (header_and_load_cmds >= data_sp->GetByteSize()) { 834 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 835 data.SetData(data_sp); 836 data_offset = MachHeaderSizeFromMagic(header.magic); 837 } 838 if (data_sp) { 839 ModuleSpec base_spec; 840 base_spec.GetFileSpec() = file; 841 base_spec.SetObjectOffset(file_offset); 842 base_spec.SetObjectSize(length); 843 GetAllArchSpecs(header, data, data_offset, base_spec, specs); 844 } 845 } 846 } 847 return specs.GetSize() - initial_count; 848 } 849 850 ConstString ObjectFileMachO::GetSegmentNameTEXT() { 851 static ConstString g_segment_name_TEXT("__TEXT"); 852 return g_segment_name_TEXT; 853 } 854 855 ConstString ObjectFileMachO::GetSegmentNameDATA() { 856 static ConstString g_segment_name_DATA("__DATA"); 857 return g_segment_name_DATA; 858 } 859 860 ConstString ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 861 static ConstString g_segment_name("__DATA_DIRTY"); 862 return g_segment_name; 863 } 864 865 ConstString ObjectFileMachO::GetSegmentNameDATA_CONST() { 866 static ConstString g_segment_name("__DATA_CONST"); 867 return g_segment_name; 868 } 869 870 ConstString ObjectFileMachO::GetSegmentNameOBJC() { 871 static ConstString g_segment_name_OBJC("__OBJC"); 872 return g_segment_name_OBJC; 873 } 874 875 ConstString ObjectFileMachO::GetSegmentNameLINKEDIT() { 876 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 877 return g_section_name_LINKEDIT; 878 } 879 880 ConstString ObjectFileMachO::GetSegmentNameDWARF() { 881 static ConstString g_section_name("__DWARF"); 882 return g_section_name; 883 } 884 885 ConstString ObjectFileMachO::GetSectionNameEHFrame() { 886 static ConstString g_section_name_eh_frame("__eh_frame"); 887 return g_section_name_eh_frame; 888 } 889 890 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 891 lldb::addr_t data_offset, 892 lldb::addr_t data_length) { 893 DataExtractor data; 894 data.SetData(data_sp, data_offset, data_length); 895 lldb::offset_t offset = 0; 896 uint32_t magic = data.GetU32(&offset); 897 return MachHeaderSizeFromMagic(magic) != 0; 898 } 899 900 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 901 DataBufferSP &data_sp, 902 lldb::offset_t data_offset, 903 const FileSpec *file, 904 lldb::offset_t file_offset, 905 lldb::offset_t length) 906 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 907 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 908 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 909 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 910 ::memset(&m_header, 0, sizeof(m_header)); 911 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 912 } 913 914 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 915 lldb::DataBufferSP &header_data_sp, 916 const lldb::ProcessSP &process_sp, 917 lldb::addr_t header_addr) 918 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 919 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 920 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 921 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 922 ::memset(&m_header, 0, sizeof(m_header)); 923 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 924 } 925 926 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 927 lldb::offset_t *data_offset_ptr, 928 llvm::MachO::mach_header &header) { 929 data.SetByteOrder(endian::InlHostByteOrder()); 930 // Leave magic in the original byte order 931 header.magic = data.GetU32(data_offset_ptr); 932 bool can_parse = false; 933 bool is_64_bit = false; 934 switch (header.magic) { 935 case MH_MAGIC: 936 data.SetByteOrder(endian::InlHostByteOrder()); 937 data.SetAddressByteSize(4); 938 can_parse = true; 939 break; 940 941 case MH_MAGIC_64: 942 data.SetByteOrder(endian::InlHostByteOrder()); 943 data.SetAddressByteSize(8); 944 can_parse = true; 945 is_64_bit = true; 946 break; 947 948 case MH_CIGAM: 949 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 950 ? eByteOrderLittle 951 : eByteOrderBig); 952 data.SetAddressByteSize(4); 953 can_parse = true; 954 break; 955 956 case MH_CIGAM_64: 957 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 958 ? eByteOrderLittle 959 : eByteOrderBig); 960 data.SetAddressByteSize(8); 961 is_64_bit = true; 962 can_parse = true; 963 break; 964 965 default: 966 break; 967 } 968 969 if (can_parse) { 970 data.GetU32(data_offset_ptr, &header.cputype, 6); 971 if (is_64_bit) 972 *data_offset_ptr += 4; 973 return true; 974 } else { 975 memset(&header, 0, sizeof(header)); 976 } 977 return false; 978 } 979 980 bool ObjectFileMachO::ParseHeader() { 981 ModuleSP module_sp(GetModule()); 982 if (!module_sp) 983 return false; 984 985 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 986 bool can_parse = false; 987 lldb::offset_t offset = 0; 988 m_data.SetByteOrder(endian::InlHostByteOrder()); 989 // Leave magic in the original byte order 990 m_header.magic = m_data.GetU32(&offset); 991 switch (m_header.magic) { 992 case MH_MAGIC: 993 m_data.SetByteOrder(endian::InlHostByteOrder()); 994 m_data.SetAddressByteSize(4); 995 can_parse = true; 996 break; 997 998 case MH_MAGIC_64: 999 m_data.SetByteOrder(endian::InlHostByteOrder()); 1000 m_data.SetAddressByteSize(8); 1001 can_parse = true; 1002 break; 1003 1004 case MH_CIGAM: 1005 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1006 ? eByteOrderLittle 1007 : eByteOrderBig); 1008 m_data.SetAddressByteSize(4); 1009 can_parse = true; 1010 break; 1011 1012 case MH_CIGAM_64: 1013 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1014 ? eByteOrderLittle 1015 : eByteOrderBig); 1016 m_data.SetAddressByteSize(8); 1017 can_parse = true; 1018 break; 1019 1020 default: 1021 break; 1022 } 1023 1024 if (can_parse) { 1025 m_data.GetU32(&offset, &m_header.cputype, 6); 1026 1027 ModuleSpecList all_specs; 1028 ModuleSpec base_spec; 1029 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 1030 base_spec, all_specs); 1031 1032 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 1033 ArchSpec mach_arch = 1034 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 1035 1036 // Check if the module has a required architecture 1037 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1038 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1039 continue; 1040 1041 if (SetModulesArchitecture(mach_arch)) { 1042 const size_t header_and_lc_size = 1043 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1044 if (m_data.GetByteSize() < header_and_lc_size) { 1045 DataBufferSP data_sp; 1046 ProcessSP process_sp(m_process_wp.lock()); 1047 if (process_sp) { 1048 data_sp = ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1049 } else { 1050 // Read in all only the load command data from the file on disk 1051 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset); 1052 if (data_sp->GetByteSize() != header_and_lc_size) 1053 continue; 1054 } 1055 if (data_sp) 1056 m_data.SetData(data_sp); 1057 } 1058 } 1059 return true; 1060 } 1061 // None found. 1062 return false; 1063 } else { 1064 memset(&m_header, 0, sizeof(struct mach_header)); 1065 } 1066 return false; 1067 } 1068 1069 ByteOrder ObjectFileMachO::GetByteOrder() const { 1070 return m_data.GetByteOrder(); 1071 } 1072 1073 bool ObjectFileMachO::IsExecutable() const { 1074 return m_header.filetype == MH_EXECUTE; 1075 } 1076 1077 bool ObjectFileMachO::IsDynamicLoader() const { 1078 return m_header.filetype == MH_DYLINKER; 1079 } 1080 1081 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1082 return m_data.GetAddressByteSize(); 1083 } 1084 1085 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1086 Symtab *symtab = GetSymtab(); 1087 if (!symtab) 1088 return AddressClass::eUnknown; 1089 1090 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1091 if (symbol) { 1092 if (symbol->ValueIsAddress()) { 1093 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1094 if (section_sp) { 1095 const lldb::SectionType section_type = section_sp->GetType(); 1096 switch (section_type) { 1097 case eSectionTypeInvalid: 1098 return AddressClass::eUnknown; 1099 1100 case eSectionTypeCode: 1101 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1102 // For ARM we have a bit in the n_desc field of the symbol that 1103 // tells us ARM/Thumb which is bit 0x0008. 1104 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1105 return AddressClass::eCodeAlternateISA; 1106 } 1107 return AddressClass::eCode; 1108 1109 case eSectionTypeContainer: 1110 return AddressClass::eUnknown; 1111 1112 case eSectionTypeData: 1113 case eSectionTypeDataCString: 1114 case eSectionTypeDataCStringPointers: 1115 case eSectionTypeDataSymbolAddress: 1116 case eSectionTypeData4: 1117 case eSectionTypeData8: 1118 case eSectionTypeData16: 1119 case eSectionTypeDataPointers: 1120 case eSectionTypeZeroFill: 1121 case eSectionTypeDataObjCMessageRefs: 1122 case eSectionTypeDataObjCCFStrings: 1123 case eSectionTypeGoSymtab: 1124 return AddressClass::eData; 1125 1126 case eSectionTypeDebug: 1127 case eSectionTypeDWARFDebugAbbrev: 1128 case eSectionTypeDWARFDebugAbbrevDwo: 1129 case eSectionTypeDWARFDebugAddr: 1130 case eSectionTypeDWARFDebugAranges: 1131 case eSectionTypeDWARFDebugCuIndex: 1132 case eSectionTypeDWARFDebugFrame: 1133 case eSectionTypeDWARFDebugInfo: 1134 case eSectionTypeDWARFDebugInfoDwo: 1135 case eSectionTypeDWARFDebugLine: 1136 case eSectionTypeDWARFDebugLineStr: 1137 case eSectionTypeDWARFDebugLoc: 1138 case eSectionTypeDWARFDebugLocDwo: 1139 case eSectionTypeDWARFDebugLocLists: 1140 case eSectionTypeDWARFDebugLocListsDwo: 1141 case eSectionTypeDWARFDebugMacInfo: 1142 case eSectionTypeDWARFDebugMacro: 1143 case eSectionTypeDWARFDebugNames: 1144 case eSectionTypeDWARFDebugPubNames: 1145 case eSectionTypeDWARFDebugPubTypes: 1146 case eSectionTypeDWARFDebugRanges: 1147 case eSectionTypeDWARFDebugRngLists: 1148 case eSectionTypeDWARFDebugRngListsDwo: 1149 case eSectionTypeDWARFDebugStr: 1150 case eSectionTypeDWARFDebugStrDwo: 1151 case eSectionTypeDWARFDebugStrOffsets: 1152 case eSectionTypeDWARFDebugStrOffsetsDwo: 1153 case eSectionTypeDWARFDebugTuIndex: 1154 case eSectionTypeDWARFDebugTypes: 1155 case eSectionTypeDWARFDebugTypesDwo: 1156 case eSectionTypeDWARFAppleNames: 1157 case eSectionTypeDWARFAppleTypes: 1158 case eSectionTypeDWARFAppleNamespaces: 1159 case eSectionTypeDWARFAppleObjC: 1160 case eSectionTypeDWARFGNUDebugAltLink: 1161 return AddressClass::eDebug; 1162 1163 case eSectionTypeEHFrame: 1164 case eSectionTypeARMexidx: 1165 case eSectionTypeARMextab: 1166 case eSectionTypeCompactUnwind: 1167 return AddressClass::eRuntime; 1168 1169 case eSectionTypeAbsoluteAddress: 1170 case eSectionTypeELFSymbolTable: 1171 case eSectionTypeELFDynamicSymbols: 1172 case eSectionTypeELFRelocationEntries: 1173 case eSectionTypeELFDynamicLinkInfo: 1174 case eSectionTypeOther: 1175 return AddressClass::eUnknown; 1176 } 1177 } 1178 } 1179 1180 const SymbolType symbol_type = symbol->GetType(); 1181 switch (symbol_type) { 1182 case eSymbolTypeAny: 1183 return AddressClass::eUnknown; 1184 case eSymbolTypeAbsolute: 1185 return AddressClass::eUnknown; 1186 1187 case eSymbolTypeCode: 1188 case eSymbolTypeTrampoline: 1189 case eSymbolTypeResolver: 1190 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1191 // For ARM we have a bit in the n_desc field of the symbol that tells 1192 // us ARM/Thumb which is bit 0x0008. 1193 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1194 return AddressClass::eCodeAlternateISA; 1195 } 1196 return AddressClass::eCode; 1197 1198 case eSymbolTypeData: 1199 return AddressClass::eData; 1200 case eSymbolTypeRuntime: 1201 return AddressClass::eRuntime; 1202 case eSymbolTypeException: 1203 return AddressClass::eRuntime; 1204 case eSymbolTypeSourceFile: 1205 return AddressClass::eDebug; 1206 case eSymbolTypeHeaderFile: 1207 return AddressClass::eDebug; 1208 case eSymbolTypeObjectFile: 1209 return AddressClass::eDebug; 1210 case eSymbolTypeCommonBlock: 1211 return AddressClass::eDebug; 1212 case eSymbolTypeBlock: 1213 return AddressClass::eDebug; 1214 case eSymbolTypeLocal: 1215 return AddressClass::eData; 1216 case eSymbolTypeParam: 1217 return AddressClass::eData; 1218 case eSymbolTypeVariable: 1219 return AddressClass::eData; 1220 case eSymbolTypeVariableType: 1221 return AddressClass::eDebug; 1222 case eSymbolTypeLineEntry: 1223 return AddressClass::eDebug; 1224 case eSymbolTypeLineHeader: 1225 return AddressClass::eDebug; 1226 case eSymbolTypeScopeBegin: 1227 return AddressClass::eDebug; 1228 case eSymbolTypeScopeEnd: 1229 return AddressClass::eDebug; 1230 case eSymbolTypeAdditional: 1231 return AddressClass::eUnknown; 1232 case eSymbolTypeCompiler: 1233 return AddressClass::eDebug; 1234 case eSymbolTypeInstrumentation: 1235 return AddressClass::eDebug; 1236 case eSymbolTypeUndefined: 1237 return AddressClass::eUnknown; 1238 case eSymbolTypeObjCClass: 1239 return AddressClass::eRuntime; 1240 case eSymbolTypeObjCMetaClass: 1241 return AddressClass::eRuntime; 1242 case eSymbolTypeObjCIVar: 1243 return AddressClass::eRuntime; 1244 case eSymbolTypeReExported: 1245 return AddressClass::eRuntime; 1246 } 1247 } 1248 return AddressClass::eUnknown; 1249 } 1250 1251 Symtab *ObjectFileMachO::GetSymtab() { 1252 ModuleSP module_sp(GetModule()); 1253 if (module_sp) { 1254 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1255 if (m_symtab_up == nullptr) { 1256 m_symtab_up.reset(new Symtab(this)); 1257 std::lock_guard<std::recursive_mutex> symtab_guard( 1258 m_symtab_up->GetMutex()); 1259 ParseSymtab(); 1260 m_symtab_up->Finalize(); 1261 } 1262 } 1263 return m_symtab_up.get(); 1264 } 1265 1266 bool ObjectFileMachO::IsStripped() { 1267 if (m_dysymtab.cmd == 0) { 1268 ModuleSP module_sp(GetModule()); 1269 if (module_sp) { 1270 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1271 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1272 const lldb::offset_t load_cmd_offset = offset; 1273 1274 load_command lc; 1275 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 1276 break; 1277 if (lc.cmd == LC_DYSYMTAB) { 1278 m_dysymtab.cmd = lc.cmd; 1279 m_dysymtab.cmdsize = lc.cmdsize; 1280 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1281 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1282 nullptr) { 1283 // Clear m_dysymtab if we were unable to read all items from the 1284 // load command 1285 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1286 } 1287 } 1288 offset = load_cmd_offset + lc.cmdsize; 1289 } 1290 } 1291 } 1292 if (m_dysymtab.cmd) 1293 return m_dysymtab.nlocalsym <= 1; 1294 return false; 1295 } 1296 1297 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1298 EncryptedFileRanges result; 1299 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1300 1301 encryption_info_command encryption_cmd; 1302 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1303 const lldb::offset_t load_cmd_offset = offset; 1304 if (m_data.GetU32(&offset, &encryption_cmd, 2) == nullptr) 1305 break; 1306 1307 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the 1308 // 3 fields we care about, so treat them the same. 1309 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1310 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1311 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1312 if (encryption_cmd.cryptid != 0) { 1313 EncryptedFileRanges::Entry entry; 1314 entry.SetRangeBase(encryption_cmd.cryptoff); 1315 entry.SetByteSize(encryption_cmd.cryptsize); 1316 result.Append(entry); 1317 } 1318 } 1319 } 1320 offset = load_cmd_offset + encryption_cmd.cmdsize; 1321 } 1322 1323 return result; 1324 } 1325 1326 void ObjectFileMachO::SanitizeSegmentCommand(segment_command_64 &seg_cmd, 1327 uint32_t cmd_idx) { 1328 if (m_length == 0 || seg_cmd.filesize == 0) 1329 return; 1330 1331 if (seg_cmd.fileoff > m_length) { 1332 // We have a load command that says it extends past the end of the file. 1333 // This is likely a corrupt file. We don't have any way to return an error 1334 // condition here (this method was likely invoked from something like 1335 // ObjectFile::GetSectionList()), so we just null out the section contents, 1336 // and dump a message to stdout. The most common case here is core file 1337 // debugging with a truncated file. 1338 const char *lc_segment_name = 1339 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1340 GetModule()->ReportWarning( 1341 "load command %u %s has a fileoff (0x%" PRIx64 1342 ") that extends beyond the end of the file (0x%" PRIx64 1343 "), ignoring this section", 1344 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1345 1346 seg_cmd.fileoff = 0; 1347 seg_cmd.filesize = 0; 1348 } 1349 1350 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1351 // We have a load command that says it extends past the end of the file. 1352 // This is likely a corrupt file. We don't have any way to return an error 1353 // condition here (this method was likely invoked from something like 1354 // ObjectFile::GetSectionList()), so we just null out the section contents, 1355 // and dump a message to stdout. The most common case here is core file 1356 // debugging with a truncated file. 1357 const char *lc_segment_name = 1358 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1359 GetModule()->ReportWarning( 1360 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1361 ") that extends beyond the end of the file (0x%" PRIx64 1362 "), the segment will be truncated to match", 1363 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1364 1365 // Truncate the length 1366 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1367 } 1368 } 1369 1370 static uint32_t GetSegmentPermissions(const segment_command_64 &seg_cmd) { 1371 uint32_t result = 0; 1372 if (seg_cmd.initprot & VM_PROT_READ) 1373 result |= ePermissionsReadable; 1374 if (seg_cmd.initprot & VM_PROT_WRITE) 1375 result |= ePermissionsWritable; 1376 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1377 result |= ePermissionsExecutable; 1378 return result; 1379 } 1380 1381 static lldb::SectionType GetSectionType(uint32_t flags, 1382 ConstString section_name) { 1383 1384 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1385 return eSectionTypeCode; 1386 1387 uint32_t mach_sect_type = flags & SECTION_TYPE; 1388 static ConstString g_sect_name_objc_data("__objc_data"); 1389 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1390 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1391 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1392 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1393 static ConstString g_sect_name_objc_const("__objc_const"); 1394 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1395 static ConstString g_sect_name_cfstring("__cfstring"); 1396 1397 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1398 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1399 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1400 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1401 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1402 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1403 static ConstString g_sect_name_dwarf_debug_loclists("__debug_loclists"); 1404 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1405 static ConstString g_sect_name_dwarf_debug_names("__debug_names"); 1406 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1407 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1408 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1409 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1410 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1411 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1412 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1413 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1414 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1415 static ConstString g_sect_name_eh_frame("__eh_frame"); 1416 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1417 static ConstString g_sect_name_text("__text"); 1418 static ConstString g_sect_name_data("__data"); 1419 static ConstString g_sect_name_go_symtab("__gosymtab"); 1420 1421 if (section_name == g_sect_name_dwarf_debug_abbrev) 1422 return eSectionTypeDWARFDebugAbbrev; 1423 if (section_name == g_sect_name_dwarf_debug_aranges) 1424 return eSectionTypeDWARFDebugAranges; 1425 if (section_name == g_sect_name_dwarf_debug_frame) 1426 return eSectionTypeDWARFDebugFrame; 1427 if (section_name == g_sect_name_dwarf_debug_info) 1428 return eSectionTypeDWARFDebugInfo; 1429 if (section_name == g_sect_name_dwarf_debug_line) 1430 return eSectionTypeDWARFDebugLine; 1431 if (section_name == g_sect_name_dwarf_debug_loc) 1432 return eSectionTypeDWARFDebugLoc; 1433 if (section_name == g_sect_name_dwarf_debug_loclists) 1434 return eSectionTypeDWARFDebugLocLists; 1435 if (section_name == g_sect_name_dwarf_debug_macinfo) 1436 return eSectionTypeDWARFDebugMacInfo; 1437 if (section_name == g_sect_name_dwarf_debug_names) 1438 return eSectionTypeDWARFDebugNames; 1439 if (section_name == g_sect_name_dwarf_debug_pubnames) 1440 return eSectionTypeDWARFDebugPubNames; 1441 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1442 return eSectionTypeDWARFDebugPubTypes; 1443 if (section_name == g_sect_name_dwarf_debug_ranges) 1444 return eSectionTypeDWARFDebugRanges; 1445 if (section_name == g_sect_name_dwarf_debug_str) 1446 return eSectionTypeDWARFDebugStr; 1447 if (section_name == g_sect_name_dwarf_debug_types) 1448 return eSectionTypeDWARFDebugTypes; 1449 if (section_name == g_sect_name_dwarf_apple_names) 1450 return eSectionTypeDWARFAppleNames; 1451 if (section_name == g_sect_name_dwarf_apple_types) 1452 return eSectionTypeDWARFAppleTypes; 1453 if (section_name == g_sect_name_dwarf_apple_namespaces) 1454 return eSectionTypeDWARFAppleNamespaces; 1455 if (section_name == g_sect_name_dwarf_apple_objc) 1456 return eSectionTypeDWARFAppleObjC; 1457 if (section_name == g_sect_name_objc_selrefs) 1458 return eSectionTypeDataCStringPointers; 1459 if (section_name == g_sect_name_objc_msgrefs) 1460 return eSectionTypeDataObjCMessageRefs; 1461 if (section_name == g_sect_name_eh_frame) 1462 return eSectionTypeEHFrame; 1463 if (section_name == g_sect_name_compact_unwind) 1464 return eSectionTypeCompactUnwind; 1465 if (section_name == g_sect_name_cfstring) 1466 return eSectionTypeDataObjCCFStrings; 1467 if (section_name == g_sect_name_go_symtab) 1468 return eSectionTypeGoSymtab; 1469 if (section_name == g_sect_name_objc_data || 1470 section_name == g_sect_name_objc_classrefs || 1471 section_name == g_sect_name_objc_superrefs || 1472 section_name == g_sect_name_objc_const || 1473 section_name == g_sect_name_objc_classlist) { 1474 return eSectionTypeDataPointers; 1475 } 1476 1477 switch (mach_sect_type) { 1478 // TODO: categorize sections by other flags for regular sections 1479 case S_REGULAR: 1480 if (section_name == g_sect_name_text) 1481 return eSectionTypeCode; 1482 if (section_name == g_sect_name_data) 1483 return eSectionTypeData; 1484 return eSectionTypeOther; 1485 case S_ZEROFILL: 1486 return eSectionTypeZeroFill; 1487 case S_CSTRING_LITERALS: // section with only literal C strings 1488 return eSectionTypeDataCString; 1489 case S_4BYTE_LITERALS: // section with only 4 byte literals 1490 return eSectionTypeData4; 1491 case S_8BYTE_LITERALS: // section with only 8 byte literals 1492 return eSectionTypeData8; 1493 case S_LITERAL_POINTERS: // section with only pointers to literals 1494 return eSectionTypeDataPointers; 1495 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1496 return eSectionTypeDataPointers; 1497 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1498 return eSectionTypeDataPointers; 1499 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1500 // the reserved2 field 1501 return eSectionTypeCode; 1502 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1503 // initialization 1504 return eSectionTypeDataPointers; 1505 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1506 // termination 1507 return eSectionTypeDataPointers; 1508 case S_COALESCED: 1509 return eSectionTypeOther; 1510 case S_GB_ZEROFILL: 1511 return eSectionTypeZeroFill; 1512 case S_INTERPOSING: // section with only pairs of function pointers for 1513 // interposing 1514 return eSectionTypeCode; 1515 case S_16BYTE_LITERALS: // section with only 16 byte literals 1516 return eSectionTypeData16; 1517 case S_DTRACE_DOF: 1518 return eSectionTypeDebug; 1519 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1520 return eSectionTypeDataPointers; 1521 default: 1522 return eSectionTypeOther; 1523 } 1524 } 1525 1526 struct ObjectFileMachO::SegmentParsingContext { 1527 const EncryptedFileRanges EncryptedRanges; 1528 lldb_private::SectionList &UnifiedList; 1529 uint32_t NextSegmentIdx = 0; 1530 uint32_t NextSectionIdx = 0; 1531 bool FileAddressesChanged = false; 1532 1533 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1534 lldb_private::SectionList &UnifiedList) 1535 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1536 }; 1537 1538 void ObjectFileMachO::ProcessSegmentCommand(const load_command &load_cmd_, 1539 lldb::offset_t offset, 1540 uint32_t cmd_idx, 1541 SegmentParsingContext &context) { 1542 segment_command_64 load_cmd; 1543 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1544 1545 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1546 return; 1547 1548 ModuleSP module_sp = GetModule(); 1549 const bool is_core = GetType() == eTypeCoreFile; 1550 const bool is_dsym = (m_header.filetype == MH_DSYM); 1551 bool add_section = true; 1552 bool add_to_unified = true; 1553 ConstString const_segname( 1554 load_cmd.segname, strnlen(load_cmd.segname, sizeof(load_cmd.segname))); 1555 1556 SectionSP unified_section_sp( 1557 context.UnifiedList.FindSectionByName(const_segname)); 1558 if (is_dsym && unified_section_sp) { 1559 if (const_segname == GetSegmentNameLINKEDIT()) { 1560 // We need to keep the __LINKEDIT segment private to this object file 1561 // only 1562 add_to_unified = false; 1563 } else { 1564 // This is the dSYM file and this section has already been created by the 1565 // object file, no need to create it. 1566 add_section = false; 1567 } 1568 } 1569 load_cmd.vmaddr = m_data.GetAddress(&offset); 1570 load_cmd.vmsize = m_data.GetAddress(&offset); 1571 load_cmd.fileoff = m_data.GetAddress(&offset); 1572 load_cmd.filesize = m_data.GetAddress(&offset); 1573 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1574 return; 1575 1576 SanitizeSegmentCommand(load_cmd, cmd_idx); 1577 1578 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1579 const bool segment_is_encrypted = 1580 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1581 1582 // Keep a list of mach segments around in case we need to get at data that 1583 // isn't stored in the abstracted Sections. 1584 m_mach_segments.push_back(load_cmd); 1585 1586 // Use a segment ID of the segment index shifted left by 8 so they never 1587 // conflict with any of the sections. 1588 SectionSP segment_sp; 1589 if (add_section && (const_segname || is_core)) { 1590 segment_sp = std::make_shared<Section>( 1591 module_sp, // Module to which this section belongs 1592 this, // Object file to which this sections belongs 1593 ++context.NextSegmentIdx 1594 << 8, // Section ID is the 1 based segment index 1595 // shifted right by 8 bits as not to collide with any of the 256 1596 // section IDs that are possible 1597 const_segname, // Name of this section 1598 eSectionTypeContainer, // This section is a container of other 1599 // sections. 1600 load_cmd.vmaddr, // File VM address == addresses as they are 1601 // found in the object file 1602 load_cmd.vmsize, // VM size in bytes of this section 1603 load_cmd.fileoff, // Offset to the data for this section in 1604 // the file 1605 load_cmd.filesize, // Size in bytes of this section as found 1606 // in the file 1607 0, // Segments have no alignment information 1608 load_cmd.flags); // Flags for this section 1609 1610 segment_sp->SetIsEncrypted(segment_is_encrypted); 1611 m_sections_up->AddSection(segment_sp); 1612 segment_sp->SetPermissions(segment_permissions); 1613 if (add_to_unified) 1614 context.UnifiedList.AddSection(segment_sp); 1615 } else if (unified_section_sp) { 1616 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1617 // Check to see if the module was read from memory? 1618 if (module_sp->GetObjectFile()->GetBaseAddress().IsValid()) { 1619 // We have a module that is in memory and needs to have its file 1620 // address adjusted. We need to do this because when we load a file 1621 // from memory, its addresses will be slid already, yet the addresses 1622 // in the new symbol file will still be unslid. Since everything is 1623 // stored as section offset, this shouldn't cause any problems. 1624 1625 // Make sure we've parsed the symbol table from the ObjectFile before 1626 // we go around changing its Sections. 1627 module_sp->GetObjectFile()->GetSymtab(); 1628 // eh_frame would present the same problems but we parse that on a per- 1629 // function basis as-needed so it's more difficult to remove its use of 1630 // the Sections. Realistically, the environments where this code path 1631 // will be taken will not have eh_frame sections. 1632 1633 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1634 1635 // Notify the module that the section addresses have been changed once 1636 // we're done so any file-address caches can be updated. 1637 context.FileAddressesChanged = true; 1638 } 1639 } 1640 m_sections_up->AddSection(unified_section_sp); 1641 } 1642 1643 struct section_64 sect64; 1644 ::memset(§64, 0, sizeof(sect64)); 1645 // Push a section into our mach sections for the section at index zero 1646 // (NO_SECT) if we don't have any mach sections yet... 1647 if (m_mach_sections.empty()) 1648 m_mach_sections.push_back(sect64); 1649 uint32_t segment_sect_idx; 1650 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1651 1652 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1653 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1654 ++segment_sect_idx) { 1655 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1656 sizeof(sect64.sectname)) == nullptr) 1657 break; 1658 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1659 sizeof(sect64.segname)) == nullptr) 1660 break; 1661 sect64.addr = m_data.GetAddress(&offset); 1662 sect64.size = m_data.GetAddress(&offset); 1663 1664 if (m_data.GetU32(&offset, §64.offset, num_u32s) == nullptr) 1665 break; 1666 1667 // Keep a list of mach sections around in case we need to get at data that 1668 // isn't stored in the abstracted Sections. 1669 m_mach_sections.push_back(sect64); 1670 1671 if (add_section) { 1672 ConstString section_name( 1673 sect64.sectname, strnlen(sect64.sectname, sizeof(sect64.sectname))); 1674 if (!const_segname) { 1675 // We have a segment with no name so we need to conjure up segments 1676 // that correspond to the section's segname if there isn't already such 1677 // a section. If there is such a section, we resize the section so that 1678 // it spans all sections. We also mark these sections as fake so 1679 // address matches don't hit if they land in the gaps between the child 1680 // sections. 1681 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1682 sizeof(sect64.segname)); 1683 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1684 if (segment_sp.get()) { 1685 Section *segment = segment_sp.get(); 1686 // Grow the section size as needed. 1687 const lldb::addr_t sect64_min_addr = sect64.addr; 1688 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1689 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1690 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1691 const lldb::addr_t curr_seg_max_addr = 1692 curr_seg_min_addr + curr_seg_byte_size; 1693 if (sect64_min_addr >= curr_seg_min_addr) { 1694 const lldb::addr_t new_seg_byte_size = 1695 sect64_max_addr - curr_seg_min_addr; 1696 // Only grow the section size if needed 1697 if (new_seg_byte_size > curr_seg_byte_size) 1698 segment->SetByteSize(new_seg_byte_size); 1699 } else { 1700 // We need to change the base address of the segment and adjust the 1701 // child section offsets for all existing children. 1702 const lldb::addr_t slide_amount = 1703 sect64_min_addr - curr_seg_min_addr; 1704 segment->Slide(slide_amount, false); 1705 segment->GetChildren().Slide(-slide_amount, false); 1706 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1707 } 1708 1709 // Grow the section size as needed. 1710 if (sect64.offset) { 1711 const lldb::addr_t segment_min_file_offset = 1712 segment->GetFileOffset(); 1713 const lldb::addr_t segment_max_file_offset = 1714 segment_min_file_offset + segment->GetFileSize(); 1715 1716 const lldb::addr_t section_min_file_offset = sect64.offset; 1717 const lldb::addr_t section_max_file_offset = 1718 section_min_file_offset + sect64.size; 1719 const lldb::addr_t new_file_offset = 1720 std::min(section_min_file_offset, segment_min_file_offset); 1721 const lldb::addr_t new_file_size = 1722 std::max(section_max_file_offset, segment_max_file_offset) - 1723 new_file_offset; 1724 segment->SetFileOffset(new_file_offset); 1725 segment->SetFileSize(new_file_size); 1726 } 1727 } else { 1728 // Create a fake section for the section's named segment 1729 segment_sp = std::make_shared<Section>( 1730 segment_sp, // Parent section 1731 module_sp, // Module to which this section belongs 1732 this, // Object file to which this section belongs 1733 ++context.NextSegmentIdx 1734 << 8, // Section ID is the 1 based segment index 1735 // shifted right by 8 bits as not to 1736 // collide with any of the 256 section IDs 1737 // that are possible 1738 const_segname, // Name of this section 1739 eSectionTypeContainer, // This section is a container of 1740 // other sections. 1741 sect64.addr, // File VM address == addresses as they are 1742 // found in the object file 1743 sect64.size, // VM size in bytes of this section 1744 sect64.offset, // Offset to the data for this section in 1745 // the file 1746 sect64.offset ? sect64.size : 0, // Size in bytes of 1747 // this section as 1748 // found in the file 1749 sect64.align, 1750 load_cmd.flags); // Flags for this section 1751 segment_sp->SetIsFake(true); 1752 segment_sp->SetPermissions(segment_permissions); 1753 m_sections_up->AddSection(segment_sp); 1754 if (add_to_unified) 1755 context.UnifiedList.AddSection(segment_sp); 1756 segment_sp->SetIsEncrypted(segment_is_encrypted); 1757 } 1758 } 1759 assert(segment_sp.get()); 1760 1761 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1762 1763 SectionSP section_sp(new Section( 1764 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1765 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1766 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1767 sect64.flags)); 1768 // Set the section to be encrypted to match the segment 1769 1770 bool section_is_encrypted = false; 1771 if (!segment_is_encrypted && load_cmd.filesize != 0) 1772 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1773 sect64.offset) != nullptr; 1774 1775 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1776 section_sp->SetPermissions(segment_permissions); 1777 segment_sp->GetChildren().AddSection(section_sp); 1778 1779 if (segment_sp->IsFake()) { 1780 segment_sp.reset(); 1781 const_segname.Clear(); 1782 } 1783 } 1784 } 1785 if (segment_sp && is_dsym) { 1786 if (first_segment_sectID <= context.NextSectionIdx) { 1787 lldb::user_id_t sect_uid; 1788 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1789 ++sect_uid) { 1790 SectionSP curr_section_sp( 1791 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1792 SectionSP next_section_sp; 1793 if (sect_uid + 1 <= context.NextSectionIdx) 1794 next_section_sp = 1795 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1796 1797 if (curr_section_sp.get()) { 1798 if (curr_section_sp->GetByteSize() == 0) { 1799 if (next_section_sp.get() != nullptr) 1800 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1801 curr_section_sp->GetFileAddress()); 1802 else 1803 curr_section_sp->SetByteSize(load_cmd.vmsize); 1804 } 1805 } 1806 } 1807 } 1808 } 1809 } 1810 1811 void ObjectFileMachO::ProcessDysymtabCommand(const load_command &load_cmd, 1812 lldb::offset_t offset) { 1813 m_dysymtab.cmd = load_cmd.cmd; 1814 m_dysymtab.cmdsize = load_cmd.cmdsize; 1815 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1816 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1817 } 1818 1819 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1820 if (m_sections_up) 1821 return; 1822 1823 m_sections_up.reset(new SectionList()); 1824 1825 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1826 // bool dump_sections = false; 1827 ModuleSP module_sp(GetModule()); 1828 1829 offset = MachHeaderSizeFromMagic(m_header.magic); 1830 1831 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1832 struct load_command load_cmd; 1833 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1834 const lldb::offset_t load_cmd_offset = offset; 1835 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 1836 break; 1837 1838 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1839 ProcessSegmentCommand(load_cmd, offset, i, context); 1840 else if (load_cmd.cmd == LC_DYSYMTAB) 1841 ProcessDysymtabCommand(load_cmd, offset); 1842 1843 offset = load_cmd_offset + load_cmd.cmdsize; 1844 } 1845 1846 if (context.FileAddressesChanged && module_sp) 1847 module_sp->SectionFileAddressesChanged(); 1848 } 1849 1850 class MachSymtabSectionInfo { 1851 public: 1852 MachSymtabSectionInfo(SectionList *section_list) 1853 : m_section_list(section_list), m_section_infos() { 1854 // Get the number of sections down to a depth of 1 to include all segments 1855 // and their sections, but no other sections that may be added for debug 1856 // map or 1857 m_section_infos.resize(section_list->GetNumSections(1)); 1858 } 1859 1860 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1861 if (n_sect == 0) 1862 return SectionSP(); 1863 if (n_sect < m_section_infos.size()) { 1864 if (!m_section_infos[n_sect].section_sp) { 1865 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1866 m_section_infos[n_sect].section_sp = section_sp; 1867 if (section_sp) { 1868 m_section_infos[n_sect].vm_range.SetBaseAddress( 1869 section_sp->GetFileAddress()); 1870 m_section_infos[n_sect].vm_range.SetByteSize( 1871 section_sp->GetByteSize()); 1872 } else { 1873 const char *filename = "<unknown>"; 1874 SectionSP first_section_sp(m_section_list->GetSectionAtIndex(0)); 1875 if (first_section_sp) 1876 filename = first_section_sp->GetObjectFile()->GetFileSpec().GetPath().c_str(); 1877 1878 Host::SystemLog(Host::eSystemLogError, 1879 "error: unable to find section %d for a symbol in %s, corrupt file?\n", 1880 n_sect, 1881 filename); 1882 } 1883 } 1884 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1885 // Symbol is in section. 1886 return m_section_infos[n_sect].section_sp; 1887 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1888 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1889 file_addr) { 1890 // Symbol is in section with zero size, but has the same start address 1891 // as the section. This can happen with linker symbols (symbols that 1892 // start with the letter 'l' or 'L'. 1893 return m_section_infos[n_sect].section_sp; 1894 } 1895 } 1896 return m_section_list->FindSectionContainingFileAddress(file_addr); 1897 } 1898 1899 protected: 1900 struct SectionInfo { 1901 SectionInfo() : vm_range(), section_sp() {} 1902 1903 VMRange vm_range; 1904 SectionSP section_sp; 1905 }; 1906 SectionList *m_section_list; 1907 std::vector<SectionInfo> m_section_infos; 1908 }; 1909 1910 #define TRIE_SYMBOL_IS_THUMB (1ULL << 63) 1911 struct TrieEntry { 1912 void Dump() const { 1913 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1914 static_cast<unsigned long long>(address), 1915 static_cast<unsigned long long>(flags), 1916 static_cast<unsigned long long>(other), name.GetCString()); 1917 if (import_name) 1918 printf(" -> \"%s\"\n", import_name.GetCString()); 1919 else 1920 printf("\n"); 1921 } 1922 ConstString name; 1923 uint64_t address = LLDB_INVALID_ADDRESS; 1924 uint64_t flags = 1925 0; // EXPORT_SYMBOL_FLAGS_REEXPORT, EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER, 1926 // TRIE_SYMBOL_IS_THUMB 1927 uint64_t other = 0; 1928 ConstString import_name; 1929 }; 1930 1931 struct TrieEntryWithOffset { 1932 lldb::offset_t nodeOffset; 1933 TrieEntry entry; 1934 1935 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1936 1937 void Dump(uint32_t idx) const { 1938 printf("[%3u] 0x%16.16llx: ", idx, 1939 static_cast<unsigned long long>(nodeOffset)); 1940 entry.Dump(); 1941 } 1942 1943 bool operator<(const TrieEntryWithOffset &other) const { 1944 return (nodeOffset < other.nodeOffset); 1945 } 1946 }; 1947 1948 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 1949 const bool is_arm, addr_t text_seg_base_addr, 1950 std::vector<llvm::StringRef> &nameSlices, 1951 std::set<lldb::addr_t> &resolver_addresses, 1952 std::vector<TrieEntryWithOffset> &reexports, 1953 std::vector<TrieEntryWithOffset> &ext_symbols) { 1954 if (!data.ValidOffset(offset)) 1955 return true; 1956 1957 // Terminal node -- end of a branch, possibly add this to 1958 // the symbol table or resolver table. 1959 const uint64_t terminalSize = data.GetULEB128(&offset); 1960 lldb::offset_t children_offset = offset + terminalSize; 1961 if (terminalSize != 0) { 1962 TrieEntryWithOffset e(offset); 1963 e.entry.flags = data.GetULEB128(&offset); 1964 const char *import_name = nullptr; 1965 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 1966 e.entry.address = 0; 1967 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 1968 import_name = data.GetCStr(&offset); 1969 } else { 1970 e.entry.address = data.GetULEB128(&offset); 1971 if (text_seg_base_addr != LLDB_INVALID_ADDRESS) 1972 e.entry.address += text_seg_base_addr; 1973 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 1974 e.entry.other = data.GetULEB128(&offset); 1975 uint64_t resolver_addr = e.entry.other; 1976 if (is_arm) 1977 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 1978 resolver_addresses.insert(resolver_addr); 1979 } else 1980 e.entry.other = 0; 1981 } 1982 bool add_this_entry = false; 1983 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT) && 1984 import_name && import_name[0]) { 1985 // add symbols that are reexport symbols with a valid import name. 1986 add_this_entry = true; 1987 } else if (e.entry.flags == 0 && 1988 (import_name == nullptr || import_name[0] == '\0')) { 1989 // add externally visible symbols, in case the nlist record has 1990 // been stripped/omitted. 1991 add_this_entry = true; 1992 } 1993 if (add_this_entry) { 1994 std::string name; 1995 if (!nameSlices.empty()) { 1996 for (auto name_slice : nameSlices) 1997 name.append(name_slice.data(), name_slice.size()); 1998 } 1999 if (name.size() > 1) { 2000 // Skip the leading '_' 2001 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2002 } 2003 if (import_name) { 2004 // Skip the leading '_' 2005 e.entry.import_name.SetCString(import_name + 1); 2006 } 2007 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT)) { 2008 reexports.push_back(e); 2009 } else { 2010 if (is_arm && (e.entry.address & 1)) { 2011 e.entry.flags |= TRIE_SYMBOL_IS_THUMB; 2012 e.entry.address &= THUMB_ADDRESS_BIT_MASK; 2013 } 2014 ext_symbols.push_back(e); 2015 } 2016 } 2017 } 2018 2019 const uint8_t childrenCount = data.GetU8(&children_offset); 2020 for (uint8_t i = 0; i < childrenCount; ++i) { 2021 const char *cstr = data.GetCStr(&children_offset); 2022 if (cstr) 2023 nameSlices.push_back(llvm::StringRef(cstr)); 2024 else 2025 return false; // Corrupt data 2026 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2027 if (childNodeOffset) { 2028 if (!ParseTrieEntries(data, childNodeOffset, is_arm, text_seg_base_addr, 2029 nameSlices, resolver_addresses, reexports, 2030 ext_symbols)) { 2031 return false; 2032 } 2033 } 2034 nameSlices.pop_back(); 2035 } 2036 return true; 2037 } 2038 2039 static SymbolType GetSymbolType(const char *&symbol_name, 2040 bool &demangled_is_synthesized, 2041 const SectionSP &text_section_sp, 2042 const SectionSP &data_section_sp, 2043 const SectionSP &data_dirty_section_sp, 2044 const SectionSP &data_const_section_sp, 2045 const SectionSP &symbol_section) { 2046 SymbolType type = eSymbolTypeInvalid; 2047 2048 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2049 if (symbol_section->IsDescendant(text_section_sp.get())) { 2050 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 2051 S_ATTR_SELF_MODIFYING_CODE | 2052 S_ATTR_SOME_INSTRUCTIONS)) 2053 type = eSymbolTypeData; 2054 else 2055 type = eSymbolTypeCode; 2056 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 2057 symbol_section->IsDescendant(data_dirty_section_sp.get()) || 2058 symbol_section->IsDescendant(data_const_section_sp.get())) { 2059 if (symbol_sect_name && 2060 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 2061 type = eSymbolTypeRuntime; 2062 2063 if (symbol_name) { 2064 llvm::StringRef symbol_name_ref(symbol_name); 2065 if (symbol_name_ref.startswith("OBJC_")) { 2066 static const llvm::StringRef g_objc_v2_prefix_class("OBJC_CLASS_$_"); 2067 static const llvm::StringRef g_objc_v2_prefix_metaclass( 2068 "OBJC_METACLASS_$_"); 2069 static const llvm::StringRef g_objc_v2_prefix_ivar("OBJC_IVAR_$_"); 2070 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 2071 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2072 type = eSymbolTypeObjCClass; 2073 demangled_is_synthesized = true; 2074 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 2075 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2076 type = eSymbolTypeObjCMetaClass; 2077 demangled_is_synthesized = true; 2078 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 2079 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2080 type = eSymbolTypeObjCIVar; 2081 demangled_is_synthesized = true; 2082 } 2083 } 2084 } 2085 } else if (symbol_sect_name && 2086 ::strstr(symbol_sect_name, "__gcc_except_tab") == 2087 symbol_sect_name) { 2088 type = eSymbolTypeException; 2089 } else { 2090 type = eSymbolTypeData; 2091 } 2092 } else if (symbol_sect_name && 2093 ::strstr(symbol_sect_name, "__IMPORT") == symbol_sect_name) { 2094 type = eSymbolTypeTrampoline; 2095 } 2096 return type; 2097 } 2098 2099 // Read the UUID out of a dyld_shared_cache file on-disk. 2100 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2101 const ByteOrder byte_order, 2102 const uint32_t addr_byte_size) { 2103 UUID dsc_uuid; 2104 DataBufferSP DscData = MapFileData( 2105 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2106 if (!DscData) 2107 return dsc_uuid; 2108 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2109 2110 char version_str[7]; 2111 lldb::offset_t offset = 0; 2112 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2113 version_str[6] = '\0'; 2114 if (strcmp(version_str, "dyld_v") == 0) { 2115 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2116 dsc_uuid = UUID::fromOptionalData( 2117 dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t)); 2118 } 2119 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2120 if (log && dsc_uuid.IsValid()) { 2121 LLDB_LOGF(log, "Shared cache %s has UUID %s", 2122 dyld_shared_cache.GetPath().c_str(), 2123 dsc_uuid.GetAsString().c_str()); 2124 } 2125 return dsc_uuid; 2126 } 2127 2128 static llvm::Optional<struct nlist_64> 2129 ParseNList(DataExtractor &nlist_data, lldb::offset_t &nlist_data_offset, 2130 size_t nlist_byte_size) { 2131 struct nlist_64 nlist; 2132 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2133 return {}; 2134 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 2135 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 2136 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 2137 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 2138 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 2139 return nlist; 2140 } 2141 2142 enum { DebugSymbols = true, NonDebugSymbols = false }; 2143 2144 size_t ObjectFileMachO::ParseSymtab() { 2145 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2146 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2147 m_file.GetFilename().AsCString("")); 2148 ModuleSP module_sp(GetModule()); 2149 if (!module_sp) 2150 return 0; 2151 2152 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2153 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2154 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2155 // The data element of type bool indicates that this entry is thumb 2156 // code. 2157 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2158 2159 // Record the address of every function/data that we add to the symtab. 2160 // We add symbols to the table in the order of most information (nlist 2161 // records) to least (function starts), and avoid duplicating symbols 2162 // via this set. 2163 std::set<addr_t> symbols_added; 2164 FunctionStarts function_starts; 2165 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2166 uint32_t i; 2167 FileSpecList dylib_files; 2168 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2169 llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2170 llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2171 llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2172 2173 for (i = 0; i < m_header.ncmds; ++i) { 2174 const lldb::offset_t cmd_offset = offset; 2175 // Read in the load command and load command size 2176 struct load_command lc; 2177 if (m_data.GetU32(&offset, &lc, 2) == nullptr) 2178 break; 2179 // Watch for the symbol table load command 2180 switch (lc.cmd) { 2181 case LC_SYMTAB: 2182 symtab_load_command.cmd = lc.cmd; 2183 symtab_load_command.cmdsize = lc.cmdsize; 2184 // Read in the rest of the symtab load command 2185 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2186 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2187 return 0; 2188 break; 2189 2190 case LC_DYLD_INFO: 2191 case LC_DYLD_INFO_ONLY: 2192 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2193 dyld_info.cmd = lc.cmd; 2194 dyld_info.cmdsize = lc.cmdsize; 2195 } else { 2196 memset(&dyld_info, 0, sizeof(dyld_info)); 2197 } 2198 break; 2199 2200 case LC_LOAD_DYLIB: 2201 case LC_LOAD_WEAK_DYLIB: 2202 case LC_REEXPORT_DYLIB: 2203 case LC_LOADFVMLIB: 2204 case LC_LOAD_UPWARD_DYLIB: { 2205 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2206 const char *path = m_data.PeekCStr(name_offset); 2207 if (path) { 2208 FileSpec file_spec(path); 2209 // Strip the path if there is @rpath, @executable, etc so we just use 2210 // the basename 2211 if (path[0] == '@') 2212 file_spec.GetDirectory().Clear(); 2213 2214 if (lc.cmd == LC_REEXPORT_DYLIB) { 2215 m_reexported_dylibs.AppendIfUnique(file_spec); 2216 } 2217 2218 dylib_files.Append(file_spec); 2219 } 2220 } break; 2221 2222 case LC_FUNCTION_STARTS: 2223 function_starts_load_command.cmd = lc.cmd; 2224 function_starts_load_command.cmdsize = lc.cmdsize; 2225 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2226 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2227 memset(&function_starts_load_command, 0, 2228 sizeof(function_starts_load_command)); 2229 break; 2230 2231 default: 2232 break; 2233 } 2234 offset = cmd_offset + lc.cmdsize; 2235 } 2236 2237 if (!symtab_load_command.cmd) 2238 return 0; 2239 2240 Symtab *symtab = m_symtab_up.get(); 2241 SectionList *section_list = GetSectionList(); 2242 if (section_list == nullptr) 2243 return 0; 2244 2245 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2246 const ByteOrder byte_order = m_data.GetByteOrder(); 2247 bool bit_width_32 = addr_byte_size == 4; 2248 const size_t nlist_byte_size = 2249 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2250 2251 DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size); 2252 DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size); 2253 DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size); 2254 DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order, 2255 addr_byte_size); 2256 DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size); 2257 2258 const addr_t nlist_data_byte_size = 2259 symtab_load_command.nsyms * nlist_byte_size; 2260 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2261 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2262 2263 ProcessSP process_sp(m_process_wp.lock()); 2264 Process *process = process_sp.get(); 2265 2266 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2267 2268 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2269 Target &target = process->GetTarget(); 2270 2271 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2272 2273 SectionSP linkedit_section_sp( 2274 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2275 // Reading mach file from memory in a process or core file... 2276 2277 if (linkedit_section_sp) { 2278 addr_t linkedit_load_addr = 2279 linkedit_section_sp->GetLoadBaseAddress(&target); 2280 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2281 // We might be trying to access the symbol table before the 2282 // __LINKEDIT's load address has been set in the target. We can't 2283 // fail to read the symbol table, so calculate the right address 2284 // manually 2285 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2286 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2287 } 2288 2289 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2290 const addr_t symoff_addr = linkedit_load_addr + 2291 symtab_load_command.symoff - 2292 linkedit_file_offset; 2293 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2294 linkedit_file_offset; 2295 2296 bool data_was_read = false; 2297 2298 #if defined(__APPLE__) && \ 2299 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2300 if (m_header.flags & 0x80000000u && 2301 process->GetAddressByteSize() == sizeof(void *)) { 2302 // This mach-o memory file is in the dyld shared cache. If this 2303 // program is not remote and this is iOS, then this process will 2304 // share the same shared cache as the process we are debugging and we 2305 // can read the entire __LINKEDIT from the address space in this 2306 // process. This is a needed optimization that is used for local iOS 2307 // debugging only since all shared libraries in the shared cache do 2308 // not have corresponding files that exist in the file system of the 2309 // device. They have been combined into a single file. This means we 2310 // always have to load these files from memory. All of the symbol and 2311 // string tables from all of the __LINKEDIT sections from the shared 2312 // libraries in the shared cache have been merged into a single large 2313 // symbol and string table. Reading all of this symbol and string 2314 // table data across can slow down debug launch times, so we optimize 2315 // this by reading the memory for the __LINKEDIT section from this 2316 // process. 2317 2318 UUID lldb_shared_cache; 2319 addr_t lldb_shared_cache_addr; 2320 GetLLDBSharedCacheUUID(lldb_shared_cache_addr, lldb_shared_cache); 2321 UUID process_shared_cache; 2322 addr_t process_shared_cache_addr; 2323 GetProcessSharedCacheUUID(process, process_shared_cache_addr, 2324 process_shared_cache); 2325 bool use_lldb_cache = true; 2326 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2327 (lldb_shared_cache != process_shared_cache || 2328 process_shared_cache_addr != lldb_shared_cache_addr)) { 2329 use_lldb_cache = false; 2330 } 2331 2332 PlatformSP platform_sp(target.GetPlatform()); 2333 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2334 data_was_read = true; 2335 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2336 eByteOrderLittle); 2337 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2338 eByteOrderLittle); 2339 if (function_starts_load_command.cmd) { 2340 const addr_t func_start_addr = 2341 linkedit_load_addr + function_starts_load_command.dataoff - 2342 linkedit_file_offset; 2343 function_starts_data.SetData((void *)func_start_addr, 2344 function_starts_load_command.datasize, 2345 eByteOrderLittle); 2346 } 2347 } 2348 } 2349 #endif 2350 2351 if (!data_was_read) { 2352 // Always load dyld - the dynamic linker - from memory if we didn't 2353 // find a binary anywhere else. lldb will not register 2354 // dylib/framework/bundle loads/unloads if we don't have the dyld 2355 // symbols, we force dyld to load from memory despite the user's 2356 // target.memory-module-load-level setting. 2357 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2358 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2359 DataBufferSP nlist_data_sp( 2360 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2361 if (nlist_data_sp) 2362 nlist_data.SetData(nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2363 if (m_dysymtab.nindirectsyms != 0) { 2364 const addr_t indirect_syms_addr = linkedit_load_addr + 2365 m_dysymtab.indirectsymoff - 2366 linkedit_file_offset; 2367 DataBufferSP indirect_syms_data_sp(ReadMemory( 2368 process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 2369 if (indirect_syms_data_sp) 2370 indirect_symbol_index_data.SetData( 2371 indirect_syms_data_sp, 0, 2372 indirect_syms_data_sp->GetByteSize()); 2373 // If this binary is outside the shared cache, 2374 // cache the string table. 2375 // Binaries in the shared cache all share a giant string table, 2376 // and we can't share the string tables across multiple 2377 // ObjectFileMachO's, so we'd end up re-reading this mega-strtab 2378 // for every binary in the shared cache - it would be a big perf 2379 // problem. For binaries outside the shared cache, it's faster to 2380 // read the entire strtab at once instead of piece-by-piece as we 2381 // process the nlist records. 2382 if ((m_header.flags & 0x80000000u) == 0) { 2383 DataBufferSP strtab_data_sp( 2384 ReadMemory(process_sp, strtab_addr, strtab_data_byte_size)); 2385 if (strtab_data_sp) { 2386 strtab_data.SetData(strtab_data_sp, 0, 2387 strtab_data_sp->GetByteSize()); 2388 } 2389 } 2390 } 2391 } 2392 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) { 2393 if (function_starts_load_command.cmd) { 2394 const addr_t func_start_addr = 2395 linkedit_load_addr + function_starts_load_command.dataoff - 2396 linkedit_file_offset; 2397 DataBufferSP func_start_data_sp( 2398 ReadMemory(process_sp, func_start_addr, 2399 function_starts_load_command.datasize)); 2400 if (func_start_data_sp) 2401 function_starts_data.SetData(func_start_data_sp, 0, 2402 func_start_data_sp->GetByteSize()); 2403 } 2404 } 2405 } 2406 } 2407 } else { 2408 nlist_data.SetData(m_data, symtab_load_command.symoff, 2409 nlist_data_byte_size); 2410 strtab_data.SetData(m_data, symtab_load_command.stroff, 2411 strtab_data_byte_size); 2412 2413 if (dyld_info.export_size > 0) { 2414 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2415 dyld_info.export_size); 2416 } 2417 2418 if (m_dysymtab.nindirectsyms != 0) { 2419 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2420 m_dysymtab.nindirectsyms * 4); 2421 } 2422 if (function_starts_load_command.cmd) { 2423 function_starts_data.SetData(m_data, function_starts_load_command.dataoff, 2424 function_starts_load_command.datasize); 2425 } 2426 } 2427 2428 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2429 2430 ConstString g_segment_name_TEXT = GetSegmentNameTEXT(); 2431 ConstString g_segment_name_DATA = GetSegmentNameDATA(); 2432 ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2433 ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2434 ConstString g_segment_name_OBJC = GetSegmentNameOBJC(); 2435 ConstString g_section_name_eh_frame = GetSectionNameEHFrame(); 2436 SectionSP text_section_sp( 2437 section_list->FindSectionByName(g_segment_name_TEXT)); 2438 SectionSP data_section_sp( 2439 section_list->FindSectionByName(g_segment_name_DATA)); 2440 SectionSP data_dirty_section_sp( 2441 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2442 SectionSP data_const_section_sp( 2443 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2444 SectionSP objc_section_sp( 2445 section_list->FindSectionByName(g_segment_name_OBJC)); 2446 SectionSP eh_frame_section_sp; 2447 if (text_section_sp.get()) 2448 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2449 g_section_name_eh_frame); 2450 else 2451 eh_frame_section_sp = 2452 section_list->FindSectionByName(g_section_name_eh_frame); 2453 2454 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2455 const bool always_thumb = GetArchitecture().IsAlwaysThumbInstructions(); 2456 2457 // lldb works best if it knows the start address of all functions in a 2458 // module. Linker symbols or debug info are normally the best source of 2459 // information for start addr / size but they may be stripped in a released 2460 // binary. Two additional sources of information exist in Mach-O binaries: 2461 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2462 // function's start address in the 2463 // binary, relative to the text section. 2464 // eh_frame - the eh_frame FDEs have the start addr & size of 2465 // each function 2466 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2467 // all modern binaries. 2468 // Binaries built to run on older releases may need to use eh_frame 2469 // information. 2470 2471 if (text_section_sp && function_starts_data.GetByteSize()) { 2472 FunctionStarts::Entry function_start_entry; 2473 function_start_entry.data = false; 2474 lldb::offset_t function_start_offset = 0; 2475 function_start_entry.addr = text_section_sp->GetFileAddress(); 2476 uint64_t delta; 2477 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2478 0) { 2479 // Now append the current entry 2480 function_start_entry.addr += delta; 2481 if (is_arm) { 2482 if (function_start_entry.addr & 1) { 2483 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2484 function_start_entry.data = true; 2485 } else if (always_thumb) { 2486 function_start_entry.data = true; 2487 } 2488 } 2489 function_starts.Append(function_start_entry); 2490 } 2491 } else { 2492 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2493 // load command claiming an eh_frame but it doesn't actually have the 2494 // eh_frame content. And if we have a dSYM, we don't need to do any of 2495 // this fill-in-the-missing-symbols works anyway - the debug info should 2496 // give us all the functions in the module. 2497 if (text_section_sp.get() && eh_frame_section_sp.get() && 2498 m_type != eTypeDebugInfo) { 2499 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2500 DWARFCallFrameInfo::EH); 2501 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2502 eh_frame.GetFunctionAddressAndSizeVector(functions); 2503 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2504 size_t count = functions.GetSize(); 2505 for (size_t i = 0; i < count; ++i) { 2506 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2507 functions.GetEntryAtIndex(i); 2508 if (func) { 2509 FunctionStarts::Entry function_start_entry; 2510 function_start_entry.addr = func->base - text_base_addr; 2511 if (is_arm) { 2512 if (function_start_entry.addr & 1) { 2513 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2514 function_start_entry.data = true; 2515 } else if (always_thumb) { 2516 function_start_entry.data = true; 2517 } 2518 } 2519 function_starts.Append(function_start_entry); 2520 } 2521 } 2522 } 2523 } 2524 2525 const size_t function_starts_count = function_starts.GetSize(); 2526 2527 // For user process binaries (executables, dylibs, frameworks, bundles), if 2528 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2529 // going to assume the binary has been stripped. Don't allow assembly 2530 // language instruction emulation because we don't know proper function 2531 // start boundaries. 2532 // 2533 // For all other types of binaries (kernels, stand-alone bare board 2534 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2535 // sections - we should not make any assumptions about them based on that. 2536 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2537 m_allow_assembly_emulation_unwind_plans = false; 2538 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2539 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2540 2541 if (unwind_or_symbol_log) 2542 module_sp->LogMessage( 2543 unwind_or_symbol_log, 2544 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2545 } 2546 2547 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() 2548 ? eh_frame_section_sp->GetID() 2549 : static_cast<user_id_t>(NO_SECT); 2550 2551 lldb::offset_t nlist_data_offset = 0; 2552 2553 uint32_t N_SO_index = UINT32_MAX; 2554 2555 MachSymtabSectionInfo section_info(section_list); 2556 std::vector<uint32_t> N_FUN_indexes; 2557 std::vector<uint32_t> N_NSYM_indexes; 2558 std::vector<uint32_t> N_INCL_indexes; 2559 std::vector<uint32_t> N_BRAC_indexes; 2560 std::vector<uint32_t> N_COMM_indexes; 2561 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2562 typedef llvm::DenseMap<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2563 typedef llvm::DenseMap<const char *, uint32_t> ConstNameToSymbolIndexMap; 2564 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2565 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2566 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2567 // Any symbols that get merged into another will get an entry in this map 2568 // so we know 2569 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2570 uint32_t nlist_idx = 0; 2571 Symbol *symbol_ptr = nullptr; 2572 2573 uint32_t sym_idx = 0; 2574 Symbol *sym = nullptr; 2575 size_t num_syms = 0; 2576 std::string memory_symbol_name; 2577 uint32_t unmapped_local_symbols_found = 0; 2578 2579 std::vector<TrieEntryWithOffset> reexport_trie_entries; 2580 std::vector<TrieEntryWithOffset> external_sym_trie_entries; 2581 std::set<lldb::addr_t> resolver_addresses; 2582 2583 if (dyld_trie_data.GetByteSize() > 0) { 2584 ConstString text_segment_name("__TEXT"); 2585 SectionSP text_segment_sp = 2586 GetSectionList()->FindSectionByName(text_segment_name); 2587 lldb::addr_t text_segment_file_addr = LLDB_INVALID_ADDRESS; 2588 if (text_segment_sp) 2589 text_segment_file_addr = text_segment_sp->GetFileAddress(); 2590 std::vector<llvm::StringRef> nameSlices; 2591 ParseTrieEntries(dyld_trie_data, 0, is_arm, text_segment_file_addr, 2592 nameSlices, resolver_addresses, reexport_trie_entries, 2593 external_sym_trie_entries); 2594 } 2595 2596 typedef std::set<ConstString> IndirectSymbols; 2597 IndirectSymbols indirect_symbol_names; 2598 2599 #if defined(__APPLE__) && \ 2600 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2601 2602 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2603 // optimized by moving LOCAL symbols out of the memory mapped portion of 2604 // the DSC. The symbol information has all been retained, but it isn't 2605 // available in the normal nlist data. However, there *are* duplicate 2606 // entries of *some* 2607 // LOCAL symbols in the normal nlist data. To handle this situation 2608 // correctly, we must first attempt 2609 // to parse any DSC unmapped symbol information. If we find any, we set a 2610 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2611 2612 if (m_header.flags & 0x80000000u) { 2613 // Before we can start mapping the DSC, we need to make certain the 2614 // target process is actually using the cache we can find. 2615 2616 // Next we need to determine the correct path for the dyld shared cache. 2617 2618 ArchSpec header_arch = GetArchitecture(); 2619 char dsc_path[PATH_MAX]; 2620 char dsc_path_development[PATH_MAX]; 2621 2622 snprintf( 2623 dsc_path, sizeof(dsc_path), "%s%s%s", 2624 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2625 */ 2626 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2627 header_arch.GetArchitectureName()); 2628 2629 snprintf( 2630 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2631 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2632 */ 2633 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2634 header_arch.GetArchitectureName(), ".development"); 2635 2636 FileSpec dsc_nondevelopment_filespec(dsc_path); 2637 FileSpec dsc_development_filespec(dsc_path_development); 2638 FileSpec dsc_filespec; 2639 2640 UUID dsc_uuid; 2641 UUID process_shared_cache_uuid; 2642 addr_t process_shared_cache_base_addr; 2643 2644 if (process) { 2645 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, 2646 process_shared_cache_uuid); 2647 } 2648 2649 // First see if we can find an exact match for the inferior process 2650 // shared cache UUID in the development or non-development shared caches 2651 // on disk. 2652 if (process_shared_cache_uuid.IsValid()) { 2653 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2654 UUID dsc_development_uuid = GetSharedCacheUUID( 2655 dsc_development_filespec, byte_order, addr_byte_size); 2656 if (dsc_development_uuid.IsValid() && 2657 dsc_development_uuid == process_shared_cache_uuid) { 2658 dsc_filespec = dsc_development_filespec; 2659 dsc_uuid = dsc_development_uuid; 2660 } 2661 } 2662 if (!dsc_uuid.IsValid() && 2663 FileSystem::Instance().Exists(dsc_nondevelopment_filespec)) { 2664 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2665 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2666 if (dsc_nondevelopment_uuid.IsValid() && 2667 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2668 dsc_filespec = dsc_nondevelopment_filespec; 2669 dsc_uuid = dsc_nondevelopment_uuid; 2670 } 2671 } 2672 } 2673 2674 // Failing a UUID match, prefer the development dyld_shared cache if both 2675 // are present. 2676 if (!FileSystem::Instance().Exists(dsc_filespec)) { 2677 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2678 dsc_filespec = dsc_development_filespec; 2679 } else { 2680 dsc_filespec = dsc_nondevelopment_filespec; 2681 } 2682 } 2683 2684 /* The dyld_cache_header has a pointer to the 2685 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2686 The dyld_cache_local_symbols_info structure gives us three things: 2687 1. The start and count of the nlist records in the dyld_shared_cache 2688 file 2689 2. The start and size of the strings for these nlist records 2690 3. The start and count of dyld_cache_local_symbols_entry entries 2691 2692 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2693 dyld shared cache. 2694 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2695 the dyld shared cache. 2696 The dyld_cache_local_symbols_entry also lists the start of this 2697 dylib/framework's nlist records 2698 and the count of how many nlist records there are for this 2699 dylib/framework. 2700 */ 2701 2702 // Process the dyld shared cache header to find the unmapped symbols 2703 2704 DataBufferSP dsc_data_sp = MapFileData( 2705 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2706 if (!dsc_uuid.IsValid()) { 2707 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2708 } 2709 if (dsc_data_sp) { 2710 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2711 2712 bool uuid_match = true; 2713 if (dsc_uuid.IsValid() && process) { 2714 if (process_shared_cache_uuid.IsValid() && 2715 dsc_uuid != process_shared_cache_uuid) { 2716 // The on-disk dyld_shared_cache file is not the same as the one in 2717 // this process' memory, don't use it. 2718 uuid_match = false; 2719 ModuleSP module_sp(GetModule()); 2720 if (module_sp) 2721 module_sp->ReportWarning("process shared cache does not match " 2722 "on-disk dyld_shared_cache file, some " 2723 "symbol names will be missing."); 2724 } 2725 } 2726 2727 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2728 2729 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2730 2731 // If the mappingOffset points to a location inside the header, we've 2732 // opened an old dyld shared cache, and should not proceed further. 2733 if (uuid_match && 2734 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2735 2736 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2737 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2738 mappingOffset); 2739 2740 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2741 byte_order, addr_byte_size); 2742 offset = 0; 2743 2744 // The File addresses (from the in-memory Mach-O load commands) for 2745 // the shared libraries in the shared library cache need to be 2746 // adjusted by an offset to match up with the dylibOffset identifying 2747 // field in the dyld_cache_local_symbol_entry's. This offset is 2748 // recorded in mapping_offset_value. 2749 const uint64_t mapping_offset_value = 2750 dsc_mapping_info_data.GetU64(&offset); 2751 2752 offset = 2753 offsetof(struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2754 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2755 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2756 2757 if (localSymbolsOffset && localSymbolsSize) { 2758 // Map the local symbols 2759 DataBufferSP dsc_local_symbols_data_sp = 2760 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2761 2762 if (dsc_local_symbols_data_sp) { 2763 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2764 byte_order, addr_byte_size); 2765 2766 offset = 0; 2767 2768 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 2769 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 2770 UndefinedNameToDescMap undefined_name_to_desc; 2771 SymbolIndexToName reexport_shlib_needs_fixup; 2772 2773 // Read the local_symbols_infos struct in one shot 2774 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2775 dsc_local_symbols_data.GetU32(&offset, 2776 &local_symbols_info.nlistOffset, 6); 2777 2778 SectionSP text_section_sp( 2779 section_list->FindSectionByName(GetSegmentNameTEXT())); 2780 2781 uint32_t header_file_offset = 2782 (text_section_sp->GetFileAddress() - mapping_offset_value); 2783 2784 offset = local_symbols_info.entriesOffset; 2785 for (uint32_t entry_index = 0; 2786 entry_index < local_symbols_info.entriesCount; entry_index++) { 2787 struct lldb_copy_dyld_cache_local_symbols_entry 2788 local_symbols_entry; 2789 local_symbols_entry.dylibOffset = 2790 dsc_local_symbols_data.GetU32(&offset); 2791 local_symbols_entry.nlistStartIndex = 2792 dsc_local_symbols_data.GetU32(&offset); 2793 local_symbols_entry.nlistCount = 2794 dsc_local_symbols_data.GetU32(&offset); 2795 2796 if (header_file_offset == local_symbols_entry.dylibOffset) { 2797 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2798 2799 // The normal nlist code cannot correctly size the Symbols 2800 // array, we need to allocate it here. 2801 sym = symtab->Resize( 2802 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2803 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2804 num_syms = symtab->GetNumSymbols(); 2805 2806 nlist_data_offset = 2807 local_symbols_info.nlistOffset + 2808 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2809 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2810 2811 for (uint32_t nlist_index = 0; 2812 nlist_index < local_symbols_entry.nlistCount; 2813 nlist_index++) { 2814 ///////////////////////////// 2815 { 2816 llvm::Optional<struct nlist_64> nlist_maybe = 2817 ParseNList(dsc_local_symbols_data, nlist_data_offset, 2818 nlist_byte_size); 2819 if (!nlist_maybe) 2820 break; 2821 struct nlist_64 nlist = *nlist_maybe; 2822 2823 SymbolType type = eSymbolTypeInvalid; 2824 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2825 string_table_offset + nlist.n_strx); 2826 2827 if (symbol_name == NULL) { 2828 // No symbol should be NULL, even the symbols with no 2829 // string values should have an offset zero which 2830 // points to an empty C-string 2831 Host::SystemLog( 2832 Host::eSystemLogError, 2833 "error: DSC unmapped local symbol[%u] has invalid " 2834 "string table offset 0x%x in %s, ignoring symbol\n", 2835 entry_index, nlist.n_strx, 2836 module_sp->GetFileSpec().GetPath().c_str()); 2837 continue; 2838 } 2839 if (symbol_name[0] == '\0') 2840 symbol_name = NULL; 2841 2842 const char *symbol_name_non_abi_mangled = NULL; 2843 2844 SectionSP symbol_section; 2845 uint32_t symbol_byte_size = 0; 2846 bool add_nlist = true; 2847 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2848 bool demangled_is_synthesized = false; 2849 bool is_gsym = false; 2850 bool set_value = true; 2851 2852 assert(sym_idx < num_syms); 2853 2854 sym[sym_idx].SetDebug(is_debug); 2855 2856 if (is_debug) { 2857 switch (nlist.n_type) { 2858 case N_GSYM: 2859 // global symbol: name,,NO_SECT,type,0 2860 // Sometimes the N_GSYM value contains the address. 2861 2862 // FIXME: In the .o files, we have a GSYM and a debug 2863 // symbol for all the ObjC data. They 2864 // have the same address, but we want to ensure that 2865 // we always find only the real symbol, 'cause we 2866 // don't currently correctly attribute the 2867 // GSYM one to the ObjCClass/Ivar/MetaClass 2868 // symbol type. This is a temporary hack to make 2869 // sure the ObjectiveC symbols get treated correctly. 2870 // To do this right, we should coalesce all the GSYM 2871 // & global symbols that have the same address. 2872 2873 is_gsym = true; 2874 sym[sym_idx].SetExternal(true); 2875 2876 if (symbol_name && symbol_name[0] == '_' && 2877 symbol_name[1] == 'O') { 2878 llvm::StringRef symbol_name_ref(symbol_name); 2879 if (symbol_name_ref.startswith( 2880 g_objc_v2_prefix_class)) { 2881 symbol_name_non_abi_mangled = symbol_name + 1; 2882 symbol_name = 2883 symbol_name + g_objc_v2_prefix_class.size(); 2884 type = eSymbolTypeObjCClass; 2885 demangled_is_synthesized = true; 2886 2887 } else if (symbol_name_ref.startswith( 2888 g_objc_v2_prefix_metaclass)) { 2889 symbol_name_non_abi_mangled = symbol_name + 1; 2890 symbol_name = 2891 symbol_name + g_objc_v2_prefix_metaclass.size(); 2892 type = eSymbolTypeObjCMetaClass; 2893 demangled_is_synthesized = true; 2894 } else if (symbol_name_ref.startswith( 2895 g_objc_v2_prefix_ivar)) { 2896 symbol_name_non_abi_mangled = symbol_name + 1; 2897 symbol_name = 2898 symbol_name + g_objc_v2_prefix_ivar.size(); 2899 type = eSymbolTypeObjCIVar; 2900 demangled_is_synthesized = true; 2901 } 2902 } else { 2903 if (nlist.n_value != 0) 2904 symbol_section = section_info.GetSection( 2905 nlist.n_sect, nlist.n_value); 2906 type = eSymbolTypeData; 2907 } 2908 break; 2909 2910 case N_FNAME: 2911 // procedure name (f77 kludge): name,,NO_SECT,0,0 2912 type = eSymbolTypeCompiler; 2913 break; 2914 2915 case N_FUN: 2916 // procedure: name,,n_sect,linenumber,address 2917 if (symbol_name) { 2918 type = eSymbolTypeCode; 2919 symbol_section = section_info.GetSection( 2920 nlist.n_sect, nlist.n_value); 2921 2922 N_FUN_addr_to_sym_idx.insert( 2923 std::make_pair(nlist.n_value, sym_idx)); 2924 // We use the current number of symbols in the 2925 // symbol table in lieu of using nlist_idx in case 2926 // we ever start trimming entries out 2927 N_FUN_indexes.push_back(sym_idx); 2928 } else { 2929 type = eSymbolTypeCompiler; 2930 2931 if (!N_FUN_indexes.empty()) { 2932 // Copy the size of the function into the 2933 // original 2934 // STAB entry so we don't have 2935 // to hunt for it later 2936 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2937 ->SetByteSize(nlist.n_value); 2938 N_FUN_indexes.pop_back(); 2939 // We don't really need the end function STAB as 2940 // it contains the size which we already placed 2941 // with the original symbol, so don't add it if 2942 // we want a minimal symbol table 2943 add_nlist = false; 2944 } 2945 } 2946 break; 2947 2948 case N_STSYM: 2949 // static symbol: name,,n_sect,type,address 2950 N_STSYM_addr_to_sym_idx.insert( 2951 std::make_pair(nlist.n_value, sym_idx)); 2952 symbol_section = section_info.GetSection(nlist.n_sect, 2953 nlist.n_value); 2954 if (symbol_name && symbol_name[0]) { 2955 type = ObjectFile::GetSymbolTypeFromName( 2956 symbol_name + 1, eSymbolTypeData); 2957 } 2958 break; 2959 2960 case N_LCSYM: 2961 // .lcomm symbol: name,,n_sect,type,address 2962 symbol_section = section_info.GetSection(nlist.n_sect, 2963 nlist.n_value); 2964 type = eSymbolTypeCommonBlock; 2965 break; 2966 2967 case N_BNSYM: 2968 // We use the current number of symbols in the symbol 2969 // table in lieu of using nlist_idx in case we ever 2970 // start trimming entries out Skip these if we want 2971 // minimal symbol tables 2972 add_nlist = false; 2973 break; 2974 2975 case N_ENSYM: 2976 // Set the size of the N_BNSYM to the terminating 2977 // index of this N_ENSYM so that we can always skip 2978 // the entire symbol if we need to navigate more 2979 // quickly at the source level when parsing STABS 2980 // Skip these if we want minimal symbol tables 2981 add_nlist = false; 2982 break; 2983 2984 case N_OPT: 2985 // emitted with gcc2_compiled and in gcc source 2986 type = eSymbolTypeCompiler; 2987 break; 2988 2989 case N_RSYM: 2990 // register sym: name,,NO_SECT,type,register 2991 type = eSymbolTypeVariable; 2992 break; 2993 2994 case N_SLINE: 2995 // src line: 0,,n_sect,linenumber,address 2996 symbol_section = section_info.GetSection(nlist.n_sect, 2997 nlist.n_value); 2998 type = eSymbolTypeLineEntry; 2999 break; 3000 3001 case N_SSYM: 3002 // structure elt: name,,NO_SECT,type,struct_offset 3003 type = eSymbolTypeVariableType; 3004 break; 3005 3006 case N_SO: 3007 // source file name 3008 type = eSymbolTypeSourceFile; 3009 if (symbol_name == NULL) { 3010 add_nlist = false; 3011 if (N_SO_index != UINT32_MAX) { 3012 // Set the size of the N_SO to the terminating 3013 // index of this N_SO so that we can always skip 3014 // the entire N_SO if we need to navigate more 3015 // quickly at the source level when parsing STABS 3016 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3017 symbol_ptr->SetByteSize(sym_idx); 3018 symbol_ptr->SetSizeIsSibling(true); 3019 } 3020 N_NSYM_indexes.clear(); 3021 N_INCL_indexes.clear(); 3022 N_BRAC_indexes.clear(); 3023 N_COMM_indexes.clear(); 3024 N_FUN_indexes.clear(); 3025 N_SO_index = UINT32_MAX; 3026 } else { 3027 // We use the current number of symbols in the 3028 // symbol table in lieu of using nlist_idx in case 3029 // we ever start trimming entries out 3030 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3031 if (N_SO_has_full_path) { 3032 if ((N_SO_index == sym_idx - 1) && 3033 ((sym_idx - 1) < num_syms)) { 3034 // We have two consecutive N_SO entries where 3035 // the first contains a directory and the 3036 // second contains a full path. 3037 sym[sym_idx - 1].GetMangled().SetValue( 3038 ConstString(symbol_name), false); 3039 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3040 add_nlist = false; 3041 } else { 3042 // This is the first entry in a N_SO that 3043 // contains a directory or 3044 // a full path to the source file 3045 N_SO_index = sym_idx; 3046 } 3047 } else if ((N_SO_index == sym_idx - 1) && 3048 ((sym_idx - 1) < num_syms)) { 3049 // This is usually the second N_SO entry that 3050 // contains just the filename, so here we combine 3051 // it with the first one if we are minimizing the 3052 // symbol table 3053 const char *so_path = 3054 sym[sym_idx - 1] 3055 .GetMangled() 3056 .GetDemangledName( 3057 lldb::eLanguageTypeUnknown) 3058 .AsCString(); 3059 if (so_path && so_path[0]) { 3060 std::string full_so_path(so_path); 3061 const size_t double_slash_pos = 3062 full_so_path.find("//"); 3063 if (double_slash_pos != std::string::npos) { 3064 // The linker has been generating bad N_SO 3065 // entries with doubled up paths 3066 // in the format "%s%s" where the first 3067 // string in the DW_AT_comp_dir, and the 3068 // second is the directory for the source 3069 // file so you end up with a path that looks 3070 // like "/tmp/src//tmp/src/" 3071 FileSpec so_dir(so_path); 3072 if (!FileSystem::Instance().Exists(so_dir)) { 3073 so_dir.SetFile( 3074 &full_so_path[double_slash_pos + 1], 3075 FileSpec::Style::native); 3076 if (FileSystem::Instance().Exists(so_dir)) { 3077 // Trim off the incorrect path 3078 full_so_path.erase(0, double_slash_pos + 1); 3079 } 3080 } 3081 } 3082 if (*full_so_path.rbegin() != '/') 3083 full_so_path += '/'; 3084 full_so_path += symbol_name; 3085 sym[sym_idx - 1].GetMangled().SetValue( 3086 ConstString(full_so_path.c_str()), false); 3087 add_nlist = false; 3088 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3089 } 3090 } else { 3091 // This could be a relative path to a N_SO 3092 N_SO_index = sym_idx; 3093 } 3094 } 3095 break; 3096 3097 case N_OSO: 3098 // object file name: name,,0,0,st_mtime 3099 type = eSymbolTypeObjectFile; 3100 break; 3101 3102 case N_LSYM: 3103 // local sym: name,,NO_SECT,type,offset 3104 type = eSymbolTypeLocal; 3105 break; 3106 3107 // INCL scopes 3108 case N_BINCL: 3109 // include file beginning: name,,NO_SECT,0,sum We use 3110 // the current number of symbols in the symbol table 3111 // in lieu of using nlist_idx in case we ever start 3112 // trimming entries out 3113 N_INCL_indexes.push_back(sym_idx); 3114 type = eSymbolTypeScopeBegin; 3115 break; 3116 3117 case N_EINCL: 3118 // include file end: name,,NO_SECT,0,0 3119 // Set the size of the N_BINCL to the terminating 3120 // index of this N_EINCL so that we can always skip 3121 // the entire symbol if we need to navigate more 3122 // quickly at the source level when parsing STABS 3123 if (!N_INCL_indexes.empty()) { 3124 symbol_ptr = 3125 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3126 symbol_ptr->SetByteSize(sym_idx + 1); 3127 symbol_ptr->SetSizeIsSibling(true); 3128 N_INCL_indexes.pop_back(); 3129 } 3130 type = eSymbolTypeScopeEnd; 3131 break; 3132 3133 case N_SOL: 3134 // #included file name: name,,n_sect,0,address 3135 type = eSymbolTypeHeaderFile; 3136 3137 // We currently don't use the header files on darwin 3138 add_nlist = false; 3139 break; 3140 3141 case N_PARAMS: 3142 // compiler parameters: name,,NO_SECT,0,0 3143 type = eSymbolTypeCompiler; 3144 break; 3145 3146 case N_VERSION: 3147 // compiler version: name,,NO_SECT,0,0 3148 type = eSymbolTypeCompiler; 3149 break; 3150 3151 case N_OLEVEL: 3152 // compiler -O level: name,,NO_SECT,0,0 3153 type = eSymbolTypeCompiler; 3154 break; 3155 3156 case N_PSYM: 3157 // parameter: name,,NO_SECT,type,offset 3158 type = eSymbolTypeVariable; 3159 break; 3160 3161 case N_ENTRY: 3162 // alternate entry: name,,n_sect,linenumber,address 3163 symbol_section = section_info.GetSection(nlist.n_sect, 3164 nlist.n_value); 3165 type = eSymbolTypeLineEntry; 3166 break; 3167 3168 // Left and Right Braces 3169 case N_LBRAC: 3170 // left bracket: 0,,NO_SECT,nesting level,address We 3171 // use the current number of symbols in the symbol 3172 // table in lieu of using nlist_idx in case we ever 3173 // start trimming entries out 3174 symbol_section = section_info.GetSection(nlist.n_sect, 3175 nlist.n_value); 3176 N_BRAC_indexes.push_back(sym_idx); 3177 type = eSymbolTypeScopeBegin; 3178 break; 3179 3180 case N_RBRAC: 3181 // right bracket: 0,,NO_SECT,nesting level,address 3182 // Set the size of the N_LBRAC to the terminating 3183 // index of this N_RBRAC so that we can always skip 3184 // the entire symbol if we need to navigate more 3185 // quickly at the source level when parsing STABS 3186 symbol_section = section_info.GetSection(nlist.n_sect, 3187 nlist.n_value); 3188 if (!N_BRAC_indexes.empty()) { 3189 symbol_ptr = 3190 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3191 symbol_ptr->SetByteSize(sym_idx + 1); 3192 symbol_ptr->SetSizeIsSibling(true); 3193 N_BRAC_indexes.pop_back(); 3194 } 3195 type = eSymbolTypeScopeEnd; 3196 break; 3197 3198 case N_EXCL: 3199 // deleted include file: name,,NO_SECT,0,sum 3200 type = eSymbolTypeHeaderFile; 3201 break; 3202 3203 // COMM scopes 3204 case N_BCOMM: 3205 // begin common: name,,NO_SECT,0,0 3206 // We use the current number of symbols in the symbol 3207 // table in lieu of using nlist_idx in case we ever 3208 // start trimming entries out 3209 type = eSymbolTypeScopeBegin; 3210 N_COMM_indexes.push_back(sym_idx); 3211 break; 3212 3213 case N_ECOML: 3214 // end common (local name): 0,,n_sect,0,address 3215 symbol_section = section_info.GetSection(nlist.n_sect, 3216 nlist.n_value); 3217 // Fall through 3218 3219 case N_ECOMM: 3220 // end common: name,,n_sect,0,0 3221 // Set the size of the N_BCOMM to the terminating 3222 // index of this N_ECOMM/N_ECOML so that we can 3223 // always skip the entire symbol if we need to 3224 // navigate more quickly at the source level when 3225 // parsing STABS 3226 if (!N_COMM_indexes.empty()) { 3227 symbol_ptr = 3228 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3229 symbol_ptr->SetByteSize(sym_idx + 1); 3230 symbol_ptr->SetSizeIsSibling(true); 3231 N_COMM_indexes.pop_back(); 3232 } 3233 type = eSymbolTypeScopeEnd; 3234 break; 3235 3236 case N_LENG: 3237 // second stab entry with length information 3238 type = eSymbolTypeAdditional; 3239 break; 3240 3241 default: 3242 break; 3243 } 3244 } else { 3245 // uint8_t n_pext = N_PEXT & nlist.n_type; 3246 uint8_t n_type = N_TYPE & nlist.n_type; 3247 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3248 3249 switch (n_type) { 3250 case N_INDR: { 3251 const char *reexport_name_cstr = 3252 strtab_data.PeekCStr(nlist.n_value); 3253 if (reexport_name_cstr && reexport_name_cstr[0]) { 3254 type = eSymbolTypeReExported; 3255 ConstString reexport_name( 3256 reexport_name_cstr + 3257 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3258 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3259 set_value = false; 3260 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3261 indirect_symbol_names.insert(ConstString( 3262 symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3263 } else 3264 type = eSymbolTypeUndefined; 3265 } break; 3266 3267 case N_UNDF: 3268 if (symbol_name && symbol_name[0]) { 3269 ConstString undefined_name( 3270 symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3271 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3272 } 3273 // Fall through 3274 case N_PBUD: 3275 type = eSymbolTypeUndefined; 3276 break; 3277 3278 case N_ABS: 3279 type = eSymbolTypeAbsolute; 3280 break; 3281 3282 case N_SECT: { 3283 symbol_section = section_info.GetSection(nlist.n_sect, 3284 nlist.n_value); 3285 3286 if (symbol_section == NULL) { 3287 // TODO: warn about this? 3288 add_nlist = false; 3289 break; 3290 } 3291 3292 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3293 type = eSymbolTypeException; 3294 } else { 3295 uint32_t section_type = 3296 symbol_section->Get() & SECTION_TYPE; 3297 3298 switch (section_type) { 3299 case S_CSTRING_LITERALS: 3300 type = eSymbolTypeData; 3301 break; // section with only literal C strings 3302 case S_4BYTE_LITERALS: 3303 type = eSymbolTypeData; 3304 break; // section with only 4 byte literals 3305 case S_8BYTE_LITERALS: 3306 type = eSymbolTypeData; 3307 break; // section with only 8 byte literals 3308 case S_LITERAL_POINTERS: 3309 type = eSymbolTypeTrampoline; 3310 break; // section with only pointers to literals 3311 case S_NON_LAZY_SYMBOL_POINTERS: 3312 type = eSymbolTypeTrampoline; 3313 break; // section with only non-lazy symbol 3314 // pointers 3315 case S_LAZY_SYMBOL_POINTERS: 3316 type = eSymbolTypeTrampoline; 3317 break; // section with only lazy symbol pointers 3318 case S_SYMBOL_STUBS: 3319 type = eSymbolTypeTrampoline; 3320 break; // section with only symbol stubs, byte 3321 // size of stub in the reserved2 field 3322 case S_MOD_INIT_FUNC_POINTERS: 3323 type = eSymbolTypeCode; 3324 break; // section with only function pointers for 3325 // initialization 3326 case S_MOD_TERM_FUNC_POINTERS: 3327 type = eSymbolTypeCode; 3328 break; // section with only function pointers for 3329 // termination 3330 case S_INTERPOSING: 3331 type = eSymbolTypeTrampoline; 3332 break; // section with only pairs of function 3333 // pointers for interposing 3334 case S_16BYTE_LITERALS: 3335 type = eSymbolTypeData; 3336 break; // section with only 16 byte literals 3337 case S_DTRACE_DOF: 3338 type = eSymbolTypeInstrumentation; 3339 break; 3340 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3341 type = eSymbolTypeTrampoline; 3342 break; 3343 default: 3344 switch (symbol_section->GetType()) { 3345 case lldb::eSectionTypeCode: 3346 type = eSymbolTypeCode; 3347 break; 3348 case eSectionTypeData: 3349 case eSectionTypeDataCString: // Inlined C string 3350 // data 3351 case eSectionTypeDataCStringPointers: // Pointers 3352 // to C 3353 // string 3354 // data 3355 case eSectionTypeDataSymbolAddress: // Address of 3356 // a symbol in 3357 // the symbol 3358 // table 3359 case eSectionTypeData4: 3360 case eSectionTypeData8: 3361 case eSectionTypeData16: 3362 type = eSymbolTypeData; 3363 break; 3364 default: 3365 break; 3366 } 3367 break; 3368 } 3369 3370 if (type == eSymbolTypeInvalid) { 3371 const char *symbol_sect_name = 3372 symbol_section->GetName().AsCString(); 3373 if (symbol_section->IsDescendant( 3374 text_section_sp.get())) { 3375 if (symbol_section->IsClear( 3376 S_ATTR_PURE_INSTRUCTIONS | 3377 S_ATTR_SELF_MODIFYING_CODE | 3378 S_ATTR_SOME_INSTRUCTIONS)) 3379 type = eSymbolTypeData; 3380 else 3381 type = eSymbolTypeCode; 3382 } else if (symbol_section->IsDescendant( 3383 data_section_sp.get()) || 3384 symbol_section->IsDescendant( 3385 data_dirty_section_sp.get()) || 3386 symbol_section->IsDescendant( 3387 data_const_section_sp.get())) { 3388 if (symbol_sect_name && 3389 ::strstr(symbol_sect_name, "__objc") == 3390 symbol_sect_name) { 3391 type = eSymbolTypeRuntime; 3392 3393 if (symbol_name) { 3394 llvm::StringRef symbol_name_ref(symbol_name); 3395 if (symbol_name_ref.startswith("_OBJC_")) { 3396 llvm::StringRef 3397 g_objc_v2_prefix_class( 3398 "_OBJC_CLASS_$_"); 3399 llvm::StringRef 3400 g_objc_v2_prefix_metaclass( 3401 "_OBJC_METACLASS_$_"); 3402 llvm::StringRef 3403 g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 3404 if (symbol_name_ref.startswith( 3405 g_objc_v2_prefix_class)) { 3406 symbol_name_non_abi_mangled = 3407 symbol_name + 1; 3408 symbol_name = 3409 symbol_name + 3410 g_objc_v2_prefix_class.size(); 3411 type = eSymbolTypeObjCClass; 3412 demangled_is_synthesized = true; 3413 } else if ( 3414 symbol_name_ref.startswith( 3415 g_objc_v2_prefix_metaclass)) { 3416 symbol_name_non_abi_mangled = 3417 symbol_name + 1; 3418 symbol_name = 3419 symbol_name + 3420 g_objc_v2_prefix_metaclass.size(); 3421 type = eSymbolTypeObjCMetaClass; 3422 demangled_is_synthesized = true; 3423 } else if (symbol_name_ref.startswith( 3424 g_objc_v2_prefix_ivar)) { 3425 symbol_name_non_abi_mangled = 3426 symbol_name + 1; 3427 symbol_name = 3428 symbol_name + 3429 g_objc_v2_prefix_ivar.size(); 3430 type = eSymbolTypeObjCIVar; 3431 demangled_is_synthesized = true; 3432 } 3433 } 3434 } 3435 } else if (symbol_sect_name && 3436 ::strstr(symbol_sect_name, 3437 "__gcc_except_tab") == 3438 symbol_sect_name) { 3439 type = eSymbolTypeException; 3440 } else { 3441 type = eSymbolTypeData; 3442 } 3443 } else if (symbol_sect_name && 3444 ::strstr(symbol_sect_name, "__IMPORT") == 3445 symbol_sect_name) { 3446 type = eSymbolTypeTrampoline; 3447 } else if (symbol_section->IsDescendant( 3448 objc_section_sp.get())) { 3449 type = eSymbolTypeRuntime; 3450 if (symbol_name && symbol_name[0] == '.') { 3451 llvm::StringRef symbol_name_ref(symbol_name); 3452 llvm::StringRef 3453 g_objc_v1_prefix_class(".objc_class_name_"); 3454 if (symbol_name_ref.startswith( 3455 g_objc_v1_prefix_class)) { 3456 symbol_name_non_abi_mangled = symbol_name; 3457 symbol_name = symbol_name + 3458 g_objc_v1_prefix_class.size(); 3459 type = eSymbolTypeObjCClass; 3460 demangled_is_synthesized = true; 3461 } 3462 } 3463 } 3464 } 3465 } 3466 } break; 3467 } 3468 } 3469 3470 if (add_nlist) { 3471 uint64_t symbol_value = nlist.n_value; 3472 if (symbol_name_non_abi_mangled) { 3473 sym[sym_idx].GetMangled().SetMangledName( 3474 ConstString(symbol_name_non_abi_mangled)); 3475 sym[sym_idx].GetMangled().SetDemangledName( 3476 ConstString(symbol_name)); 3477 } else { 3478 bool symbol_name_is_mangled = false; 3479 3480 if (symbol_name && symbol_name[0] == '_') { 3481 symbol_name_is_mangled = symbol_name[1] == '_'; 3482 symbol_name++; // Skip the leading underscore 3483 } 3484 3485 if (symbol_name) { 3486 ConstString const_symbol_name(symbol_name); 3487 sym[sym_idx].GetMangled().SetValue( 3488 const_symbol_name, symbol_name_is_mangled); 3489 if (is_gsym && is_debug) { 3490 const char *gsym_name = 3491 sym[sym_idx] 3492 .GetMangled() 3493 .GetName(lldb::eLanguageTypeUnknown, 3494 Mangled::ePreferMangled) 3495 .GetCString(); 3496 if (gsym_name) 3497 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3498 } 3499 } 3500 } 3501 if (symbol_section) { 3502 const addr_t section_file_addr = 3503 symbol_section->GetFileAddress(); 3504 if (symbol_byte_size == 0 && 3505 function_starts_count > 0) { 3506 addr_t symbol_lookup_file_addr = nlist.n_value; 3507 // Do an exact address match for non-ARM addresses, 3508 // else get the closest since the symbol might be a 3509 // thumb symbol which has an address with bit zero 3510 // set 3511 FunctionStarts::Entry *func_start_entry = 3512 function_starts.FindEntry(symbol_lookup_file_addr, 3513 !is_arm); 3514 if (is_arm && func_start_entry) { 3515 // Verify that the function start address is the 3516 // symbol address (ARM) or the symbol address + 1 3517 // (thumb) 3518 if (func_start_entry->addr != 3519 symbol_lookup_file_addr && 3520 func_start_entry->addr != 3521 (symbol_lookup_file_addr + 1)) { 3522 // Not the right entry, NULL it out... 3523 func_start_entry = NULL; 3524 } 3525 } 3526 if (func_start_entry) { 3527 func_start_entry->data = true; 3528 3529 addr_t symbol_file_addr = func_start_entry->addr; 3530 uint32_t symbol_flags = 0; 3531 if (is_arm) { 3532 if (symbol_file_addr & 1) 3533 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3534 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3535 } 3536 3537 const FunctionStarts::Entry *next_func_start_entry = 3538 function_starts.FindNextEntry(func_start_entry); 3539 const addr_t section_end_file_addr = 3540 section_file_addr + 3541 symbol_section->GetByteSize(); 3542 if (next_func_start_entry) { 3543 addr_t next_symbol_file_addr = 3544 next_func_start_entry->addr; 3545 // Be sure the clear the Thumb address bit when 3546 // we calculate the size from the current and 3547 // next address 3548 if (is_arm) 3549 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3550 symbol_byte_size = std::min<lldb::addr_t>( 3551 next_symbol_file_addr - symbol_file_addr, 3552 section_end_file_addr - symbol_file_addr); 3553 } else { 3554 symbol_byte_size = 3555 section_end_file_addr - symbol_file_addr; 3556 } 3557 } 3558 } 3559 symbol_value -= section_file_addr; 3560 } 3561 3562 if (is_debug == false) { 3563 if (type == eSymbolTypeCode) { 3564 // See if we can find a N_FUN entry for any code 3565 // symbols. If we do find a match, and the name 3566 // matches, then we can merge the two into just the 3567 // function symbol to avoid duplicate entries in 3568 // the symbol table 3569 auto range = 3570 N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3571 if (range.first != range.second) { 3572 bool found_it = false; 3573 for (auto pos = range.first; pos != range.second; 3574 ++pos) { 3575 if (sym[sym_idx].GetMangled().GetName( 3576 lldb::eLanguageTypeUnknown, 3577 Mangled::ePreferMangled) == 3578 sym[pos->second].GetMangled().GetName( 3579 lldb::eLanguageTypeUnknown, 3580 Mangled::ePreferMangled)) { 3581 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3582 // We just need the flags from the linker 3583 // symbol, so put these flags 3584 // into the N_FUN flags to avoid duplicate 3585 // symbols in the symbol table 3586 sym[pos->second].SetExternal( 3587 sym[sym_idx].IsExternal()); 3588 sym[pos->second].SetFlags(nlist.n_type << 16 | 3589 nlist.n_desc); 3590 if (resolver_addresses.find(nlist.n_value) != 3591 resolver_addresses.end()) 3592 sym[pos->second].SetType(eSymbolTypeResolver); 3593 sym[sym_idx].Clear(); 3594 found_it = true; 3595 break; 3596 } 3597 } 3598 if (found_it) 3599 continue; 3600 } else { 3601 if (resolver_addresses.find(nlist.n_value) != 3602 resolver_addresses.end()) 3603 type = eSymbolTypeResolver; 3604 } 3605 } else if (type == eSymbolTypeData || 3606 type == eSymbolTypeObjCClass || 3607 type == eSymbolTypeObjCMetaClass || 3608 type == eSymbolTypeObjCIVar) { 3609 // See if we can find a N_STSYM entry for any data 3610 // symbols. If we do find a match, and the name 3611 // matches, then we can merge the two into just the 3612 // Static symbol to avoid duplicate entries in the 3613 // symbol table 3614 auto range = N_STSYM_addr_to_sym_idx.equal_range( 3615 nlist.n_value); 3616 if (range.first != range.second) { 3617 bool found_it = false; 3618 for (auto pos = range.first; pos != range.second; 3619 ++pos) { 3620 if (sym[sym_idx].GetMangled().GetName( 3621 lldb::eLanguageTypeUnknown, 3622 Mangled::ePreferMangled) == 3623 sym[pos->second].GetMangled().GetName( 3624 lldb::eLanguageTypeUnknown, 3625 Mangled::ePreferMangled)) { 3626 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3627 // We just need the flags from the linker 3628 // symbol, so put these flags 3629 // into the N_STSYM flags to avoid duplicate 3630 // symbols in the symbol table 3631 sym[pos->second].SetExternal( 3632 sym[sym_idx].IsExternal()); 3633 sym[pos->second].SetFlags(nlist.n_type << 16 | 3634 nlist.n_desc); 3635 sym[sym_idx].Clear(); 3636 found_it = true; 3637 break; 3638 } 3639 } 3640 if (found_it) 3641 continue; 3642 } else { 3643 const char *gsym_name = 3644 sym[sym_idx] 3645 .GetMangled() 3646 .GetName(lldb::eLanguageTypeUnknown, 3647 Mangled::ePreferMangled) 3648 .GetCString(); 3649 if (gsym_name) { 3650 // Combine N_GSYM stab entries with the non 3651 // stab symbol 3652 ConstNameToSymbolIndexMap::const_iterator pos = 3653 N_GSYM_name_to_sym_idx.find(gsym_name); 3654 if (pos != N_GSYM_name_to_sym_idx.end()) { 3655 const uint32_t GSYM_sym_idx = pos->second; 3656 m_nlist_idx_to_sym_idx[nlist_idx] = 3657 GSYM_sym_idx; 3658 // Copy the address, because often the N_GSYM 3659 // address has an invalid address of zero 3660 // when the global is a common symbol 3661 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3662 symbol_section); 3663 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3664 symbol_value); 3665 symbols_added.insert(sym[GSYM_sym_idx] 3666 .GetAddress() 3667 .GetFileAddress()); 3668 // We just need the flags from the linker 3669 // symbol, so put these flags 3670 // into the N_GSYM flags to avoid duplicate 3671 // symbols in the symbol table 3672 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 3673 nlist.n_desc); 3674 sym[sym_idx].Clear(); 3675 continue; 3676 } 3677 } 3678 } 3679 } 3680 } 3681 3682 sym[sym_idx].SetID(nlist_idx); 3683 sym[sym_idx].SetType(type); 3684 if (set_value) { 3685 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 3686 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3687 symbols_added.insert( 3688 sym[sym_idx].GetAddress().GetFileAddress()); 3689 } 3690 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 3691 3692 if (symbol_byte_size > 0) 3693 sym[sym_idx].SetByteSize(symbol_byte_size); 3694 3695 if (demangled_is_synthesized) 3696 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3697 ++sym_idx; 3698 } else { 3699 sym[sym_idx].Clear(); 3700 } 3701 } 3702 ///////////////////////////// 3703 } 3704 break; // No more entries to consider 3705 } 3706 } 3707 3708 for (const auto &pos : reexport_shlib_needs_fixup) { 3709 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3710 if (undef_pos != undefined_name_to_desc.end()) { 3711 const uint8_t dylib_ordinal = 3712 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3713 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3714 sym[pos.first].SetReExportedSymbolSharedLibrary( 3715 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3716 } 3717 } 3718 } 3719 } 3720 } 3721 } 3722 } 3723 3724 // Must reset this in case it was mutated above! 3725 nlist_data_offset = 0; 3726 #endif 3727 3728 if (nlist_data.GetByteSize() > 0) { 3729 3730 // If the sym array was not created while parsing the DSC unmapped 3731 // symbols, create it now. 3732 if (sym == nullptr) { 3733 sym = 3734 symtab->Resize(symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3735 num_syms = symtab->GetNumSymbols(); 3736 } 3737 3738 if (unmapped_local_symbols_found) { 3739 assert(m_dysymtab.ilocalsym == 0); 3740 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3741 nlist_idx = m_dysymtab.nlocalsym; 3742 } else { 3743 nlist_idx = 0; 3744 } 3745 3746 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 3747 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 3748 UndefinedNameToDescMap undefined_name_to_desc; 3749 SymbolIndexToName reexport_shlib_needs_fixup; 3750 3751 // Symtab parsing is a huge mess. Everything is entangled and the code 3752 // requires access to a ridiculous amount of variables. LLDB depends 3753 // heavily on the proper merging of symbols and to get that right we need 3754 // to make sure we have parsed all the debug symbols first. Therefore we 3755 // invoke the lambda twice, once to parse only the debug symbols and then 3756 // once more to parse the remaining symbols. 3757 auto ParseSymbolLambda = [&](struct nlist_64 &nlist, uint32_t nlist_idx, 3758 bool debug_only) { 3759 const bool is_debug = ((nlist.n_type & N_STAB) != 0); 3760 if (is_debug != debug_only) 3761 return true; 3762 3763 const char *symbol_name_non_abi_mangled = nullptr; 3764 const char *symbol_name = nullptr; 3765 3766 if (have_strtab_data) { 3767 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3768 3769 if (symbol_name == nullptr) { 3770 // No symbol should be NULL, even the symbols with no string values 3771 // should have an offset zero which points to an empty C-string 3772 Host::SystemLog(Host::eSystemLogError, 3773 "error: symbol[%u] has invalid string table offset " 3774 "0x%x in %s, ignoring symbol\n", 3775 nlist_idx, nlist.n_strx, 3776 module_sp->GetFileSpec().GetPath().c_str()); 3777 return true; 3778 } 3779 if (symbol_name[0] == '\0') 3780 symbol_name = nullptr; 3781 } else { 3782 const addr_t str_addr = strtab_addr + nlist.n_strx; 3783 Status str_error; 3784 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3785 str_error)) 3786 symbol_name = memory_symbol_name.c_str(); 3787 } 3788 3789 SymbolType type = eSymbolTypeInvalid; 3790 SectionSP symbol_section; 3791 lldb::addr_t symbol_byte_size = 0; 3792 bool add_nlist = true; 3793 bool is_gsym = false; 3794 bool demangled_is_synthesized = false; 3795 bool set_value = true; 3796 3797 assert(sym_idx < num_syms); 3798 sym[sym_idx].SetDebug(is_debug); 3799 3800 if (is_debug) { 3801 switch (nlist.n_type) { 3802 case N_GSYM: 3803 // global symbol: name,,NO_SECT,type,0 3804 // Sometimes the N_GSYM value contains the address. 3805 3806 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3807 // the ObjC data. They 3808 // have the same address, but we want to ensure that we always find 3809 // only the real symbol, 'cause we don't currently correctly 3810 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3811 // type. This is a temporary hack to make sure the ObjectiveC 3812 // symbols get treated correctly. To do this right, we should 3813 // coalesce all the GSYM & global symbols that have the same 3814 // address. 3815 is_gsym = true; 3816 sym[sym_idx].SetExternal(true); 3817 3818 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3819 llvm::StringRef symbol_name_ref(symbol_name); 3820 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3821 symbol_name_non_abi_mangled = symbol_name + 1; 3822 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3823 type = eSymbolTypeObjCClass; 3824 demangled_is_synthesized = true; 3825 3826 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 3827 symbol_name_non_abi_mangled = symbol_name + 1; 3828 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3829 type = eSymbolTypeObjCMetaClass; 3830 demangled_is_synthesized = true; 3831 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3832 symbol_name_non_abi_mangled = symbol_name + 1; 3833 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3834 type = eSymbolTypeObjCIVar; 3835 demangled_is_synthesized = true; 3836 } 3837 } else { 3838 if (nlist.n_value != 0) 3839 symbol_section = 3840 section_info.GetSection(nlist.n_sect, nlist.n_value); 3841 type = eSymbolTypeData; 3842 } 3843 break; 3844 3845 case N_FNAME: 3846 // procedure name (f77 kludge): name,,NO_SECT,0,0 3847 type = eSymbolTypeCompiler; 3848 break; 3849 3850 case N_FUN: 3851 // procedure: name,,n_sect,linenumber,address 3852 if (symbol_name) { 3853 type = eSymbolTypeCode; 3854 symbol_section = 3855 section_info.GetSection(nlist.n_sect, nlist.n_value); 3856 3857 N_FUN_addr_to_sym_idx.insert( 3858 std::make_pair(nlist.n_value, sym_idx)); 3859 // We use the current number of symbols in the symbol table in 3860 // lieu of using nlist_idx in case we ever start trimming entries 3861 // out 3862 N_FUN_indexes.push_back(sym_idx); 3863 } else { 3864 type = eSymbolTypeCompiler; 3865 3866 if (!N_FUN_indexes.empty()) { 3867 // Copy the size of the function into the original STAB entry 3868 // so we don't have to hunt for it later 3869 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3870 ->SetByteSize(nlist.n_value); 3871 N_FUN_indexes.pop_back(); 3872 // We don't really need the end function STAB as it contains 3873 // the size which we already placed with the original symbol, 3874 // so don't add it if we want a minimal symbol table 3875 add_nlist = false; 3876 } 3877 } 3878 break; 3879 3880 case N_STSYM: 3881 // static symbol: name,,n_sect,type,address 3882 N_STSYM_addr_to_sym_idx.insert( 3883 std::make_pair(nlist.n_value, sym_idx)); 3884 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3885 if (symbol_name && symbol_name[0]) { 3886 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3887 eSymbolTypeData); 3888 } 3889 break; 3890 3891 case N_LCSYM: 3892 // .lcomm symbol: name,,n_sect,type,address 3893 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3894 type = eSymbolTypeCommonBlock; 3895 break; 3896 3897 case N_BNSYM: 3898 // We use the current number of symbols in the symbol table in lieu 3899 // of using nlist_idx in case we ever start trimming entries out 3900 // Skip these if we want minimal symbol tables 3901 add_nlist = false; 3902 break; 3903 3904 case N_ENSYM: 3905 // Set the size of the N_BNSYM to the terminating index of this 3906 // N_ENSYM so that we can always skip the entire symbol if we need 3907 // to navigate more quickly at the source level when parsing STABS 3908 // Skip these if we want minimal symbol tables 3909 add_nlist = false; 3910 break; 3911 3912 case N_OPT: 3913 // emitted with gcc2_compiled and in gcc source 3914 type = eSymbolTypeCompiler; 3915 break; 3916 3917 case N_RSYM: 3918 // register sym: name,,NO_SECT,type,register 3919 type = eSymbolTypeVariable; 3920 break; 3921 3922 case N_SLINE: 3923 // src line: 0,,n_sect,linenumber,address 3924 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3925 type = eSymbolTypeLineEntry; 3926 break; 3927 3928 case N_SSYM: 3929 // structure elt: name,,NO_SECT,type,struct_offset 3930 type = eSymbolTypeVariableType; 3931 break; 3932 3933 case N_SO: 3934 // source file name 3935 type = eSymbolTypeSourceFile; 3936 if (symbol_name == nullptr) { 3937 add_nlist = false; 3938 if (N_SO_index != UINT32_MAX) { 3939 // Set the size of the N_SO to the terminating index of this 3940 // N_SO so that we can always skip the entire N_SO if we need 3941 // to navigate more quickly at the source level when parsing 3942 // STABS 3943 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3944 symbol_ptr->SetByteSize(sym_idx); 3945 symbol_ptr->SetSizeIsSibling(true); 3946 } 3947 N_NSYM_indexes.clear(); 3948 N_INCL_indexes.clear(); 3949 N_BRAC_indexes.clear(); 3950 N_COMM_indexes.clear(); 3951 N_FUN_indexes.clear(); 3952 N_SO_index = UINT32_MAX; 3953 } else { 3954 // We use the current number of symbols in the symbol table in 3955 // lieu of using nlist_idx in case we ever start trimming entries 3956 // out 3957 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3958 if (N_SO_has_full_path) { 3959 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3960 // We have two consecutive N_SO entries where the first 3961 // contains a directory and the second contains a full path. 3962 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), 3963 false); 3964 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3965 add_nlist = false; 3966 } else { 3967 // This is the first entry in a N_SO that contains a 3968 // directory or a full path to the source file 3969 N_SO_index = sym_idx; 3970 } 3971 } else if ((N_SO_index == sym_idx - 1) && 3972 ((sym_idx - 1) < num_syms)) { 3973 // This is usually the second N_SO entry that contains just the 3974 // filename, so here we combine it with the first one if we are 3975 // minimizing the symbol table 3976 const char *so_path = 3977 sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3978 if (so_path && so_path[0]) { 3979 std::string full_so_path(so_path); 3980 const size_t double_slash_pos = full_so_path.find("//"); 3981 if (double_slash_pos != std::string::npos) { 3982 // The linker has been generating bad N_SO entries with 3983 // doubled up paths in the format "%s%s" where the first 3984 // string in the DW_AT_comp_dir, and the second is the 3985 // directory for the source file so you end up with a path 3986 // that looks like "/tmp/src//tmp/src/" 3987 FileSpec so_dir(so_path); 3988 if (!FileSystem::Instance().Exists(so_dir)) { 3989 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 3990 FileSpec::Style::native); 3991 if (FileSystem::Instance().Exists(so_dir)) { 3992 // Trim off the incorrect path 3993 full_so_path.erase(0, double_slash_pos + 1); 3994 } 3995 } 3996 } 3997 if (*full_so_path.rbegin() != '/') 3998 full_so_path += '/'; 3999 full_so_path += symbol_name; 4000 sym[sym_idx - 1].GetMangled().SetValue( 4001 ConstString(full_so_path.c_str()), false); 4002 add_nlist = false; 4003 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4004 } 4005 } else { 4006 // This could be a relative path to a N_SO 4007 N_SO_index = sym_idx; 4008 } 4009 } 4010 break; 4011 4012 case N_OSO: 4013 // object file name: name,,0,0,st_mtime 4014 type = eSymbolTypeObjectFile; 4015 break; 4016 4017 case N_LSYM: 4018 // local sym: name,,NO_SECT,type,offset 4019 type = eSymbolTypeLocal; 4020 break; 4021 4022 // INCL scopes 4023 case N_BINCL: 4024 // include file beginning: name,,NO_SECT,0,sum We use the current 4025 // number of symbols in the symbol table in lieu of using nlist_idx 4026 // in case we ever start trimming entries out 4027 N_INCL_indexes.push_back(sym_idx); 4028 type = eSymbolTypeScopeBegin; 4029 break; 4030 4031 case N_EINCL: 4032 // include file end: name,,NO_SECT,0,0 4033 // Set the size of the N_BINCL to the terminating index of this 4034 // N_EINCL so that we can always skip the entire symbol if we need 4035 // to navigate more quickly at the source level when parsing STABS 4036 if (!N_INCL_indexes.empty()) { 4037 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4038 symbol_ptr->SetByteSize(sym_idx + 1); 4039 symbol_ptr->SetSizeIsSibling(true); 4040 N_INCL_indexes.pop_back(); 4041 } 4042 type = eSymbolTypeScopeEnd; 4043 break; 4044 4045 case N_SOL: 4046 // #included file name: name,,n_sect,0,address 4047 type = eSymbolTypeHeaderFile; 4048 4049 // We currently don't use the header files on darwin 4050 add_nlist = false; 4051 break; 4052 4053 case N_PARAMS: 4054 // compiler parameters: name,,NO_SECT,0,0 4055 type = eSymbolTypeCompiler; 4056 break; 4057 4058 case N_VERSION: 4059 // compiler version: name,,NO_SECT,0,0 4060 type = eSymbolTypeCompiler; 4061 break; 4062 4063 case N_OLEVEL: 4064 // compiler -O level: name,,NO_SECT,0,0 4065 type = eSymbolTypeCompiler; 4066 break; 4067 4068 case N_PSYM: 4069 // parameter: name,,NO_SECT,type,offset 4070 type = eSymbolTypeVariable; 4071 break; 4072 4073 case N_ENTRY: 4074 // alternate entry: name,,n_sect,linenumber,address 4075 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4076 type = eSymbolTypeLineEntry; 4077 break; 4078 4079 // Left and Right Braces 4080 case N_LBRAC: 4081 // left bracket: 0,,NO_SECT,nesting level,address We use the 4082 // current number of symbols in the symbol table in lieu of using 4083 // nlist_idx in case we ever start trimming entries out 4084 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4085 N_BRAC_indexes.push_back(sym_idx); 4086 type = eSymbolTypeScopeBegin; 4087 break; 4088 4089 case N_RBRAC: 4090 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4091 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4092 // can always skip the entire symbol if we need to navigate more 4093 // quickly at the source level when parsing STABS 4094 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4095 if (!N_BRAC_indexes.empty()) { 4096 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4097 symbol_ptr->SetByteSize(sym_idx + 1); 4098 symbol_ptr->SetSizeIsSibling(true); 4099 N_BRAC_indexes.pop_back(); 4100 } 4101 type = eSymbolTypeScopeEnd; 4102 break; 4103 4104 case N_EXCL: 4105 // deleted include file: name,,NO_SECT,0,sum 4106 type = eSymbolTypeHeaderFile; 4107 break; 4108 4109 // COMM scopes 4110 case N_BCOMM: 4111 // begin common: name,,NO_SECT,0,0 4112 // We use the current number of symbols in the symbol table in lieu 4113 // of using nlist_idx in case we ever start trimming entries out 4114 type = eSymbolTypeScopeBegin; 4115 N_COMM_indexes.push_back(sym_idx); 4116 break; 4117 4118 case N_ECOML: 4119 // end common (local name): 0,,n_sect,0,address 4120 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4121 LLVM_FALLTHROUGH; 4122 4123 case N_ECOMM: 4124 // end common: name,,n_sect,0,0 4125 // Set the size of the N_BCOMM to the terminating index of this 4126 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4127 // we need to navigate more quickly at the source level when 4128 // parsing STABS 4129 if (!N_COMM_indexes.empty()) { 4130 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4131 symbol_ptr->SetByteSize(sym_idx + 1); 4132 symbol_ptr->SetSizeIsSibling(true); 4133 N_COMM_indexes.pop_back(); 4134 } 4135 type = eSymbolTypeScopeEnd; 4136 break; 4137 4138 case N_LENG: 4139 // second stab entry with length information 4140 type = eSymbolTypeAdditional; 4141 break; 4142 4143 default: 4144 break; 4145 } 4146 } else { 4147 uint8_t n_type = N_TYPE & nlist.n_type; 4148 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4149 4150 switch (n_type) { 4151 case N_INDR: { 4152 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4153 if (reexport_name_cstr && reexport_name_cstr[0]) { 4154 type = eSymbolTypeReExported; 4155 ConstString reexport_name(reexport_name_cstr + 4156 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4157 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4158 set_value = false; 4159 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4160 indirect_symbol_names.insert( 4161 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4162 } else 4163 type = eSymbolTypeUndefined; 4164 } break; 4165 4166 case N_UNDF: 4167 if (symbol_name && symbol_name[0]) { 4168 ConstString undefined_name(symbol_name + 4169 ((symbol_name[0] == '_') ? 1 : 0)); 4170 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4171 } 4172 LLVM_FALLTHROUGH; 4173 4174 case N_PBUD: 4175 type = eSymbolTypeUndefined; 4176 break; 4177 4178 case N_ABS: 4179 type = eSymbolTypeAbsolute; 4180 break; 4181 4182 case N_SECT: { 4183 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4184 4185 if (!symbol_section) { 4186 // TODO: warn about this? 4187 add_nlist = false; 4188 break; 4189 } 4190 4191 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4192 type = eSymbolTypeException; 4193 } else { 4194 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4195 4196 switch (section_type) { 4197 case S_CSTRING_LITERALS: 4198 type = eSymbolTypeData; 4199 break; // section with only literal C strings 4200 case S_4BYTE_LITERALS: 4201 type = eSymbolTypeData; 4202 break; // section with only 4 byte literals 4203 case S_8BYTE_LITERALS: 4204 type = eSymbolTypeData; 4205 break; // section with only 8 byte literals 4206 case S_LITERAL_POINTERS: 4207 type = eSymbolTypeTrampoline; 4208 break; // section with only pointers to literals 4209 case S_NON_LAZY_SYMBOL_POINTERS: 4210 type = eSymbolTypeTrampoline; 4211 break; // section with only non-lazy symbol pointers 4212 case S_LAZY_SYMBOL_POINTERS: 4213 type = eSymbolTypeTrampoline; 4214 break; // section with only lazy symbol pointers 4215 case S_SYMBOL_STUBS: 4216 type = eSymbolTypeTrampoline; 4217 break; // section with only symbol stubs, byte size of stub in 4218 // the reserved2 field 4219 case S_MOD_INIT_FUNC_POINTERS: 4220 type = eSymbolTypeCode; 4221 break; // section with only function pointers for initialization 4222 case S_MOD_TERM_FUNC_POINTERS: 4223 type = eSymbolTypeCode; 4224 break; // section with only function pointers for termination 4225 case S_INTERPOSING: 4226 type = eSymbolTypeTrampoline; 4227 break; // section with only pairs of function pointers for 4228 // interposing 4229 case S_16BYTE_LITERALS: 4230 type = eSymbolTypeData; 4231 break; // section with only 16 byte literals 4232 case S_DTRACE_DOF: 4233 type = eSymbolTypeInstrumentation; 4234 break; 4235 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4236 type = eSymbolTypeTrampoline; 4237 break; 4238 default: 4239 switch (symbol_section->GetType()) { 4240 case lldb::eSectionTypeCode: 4241 type = eSymbolTypeCode; 4242 break; 4243 case eSectionTypeData: 4244 case eSectionTypeDataCString: // Inlined C string data 4245 case eSectionTypeDataCStringPointers: // Pointers to C string 4246 // data 4247 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4248 // the symbol table 4249 case eSectionTypeData4: 4250 case eSectionTypeData8: 4251 case eSectionTypeData16: 4252 type = eSymbolTypeData; 4253 break; 4254 default: 4255 break; 4256 } 4257 break; 4258 } 4259 4260 if (type == eSymbolTypeInvalid) { 4261 const char *symbol_sect_name = 4262 symbol_section->GetName().AsCString(); 4263 if (symbol_section->IsDescendant(text_section_sp.get())) { 4264 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4265 S_ATTR_SELF_MODIFYING_CODE | 4266 S_ATTR_SOME_INSTRUCTIONS)) 4267 type = eSymbolTypeData; 4268 else 4269 type = eSymbolTypeCode; 4270 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 4271 symbol_section->IsDescendant( 4272 data_dirty_section_sp.get()) || 4273 symbol_section->IsDescendant( 4274 data_const_section_sp.get())) { 4275 if (symbol_sect_name && 4276 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 4277 type = eSymbolTypeRuntime; 4278 4279 if (symbol_name) { 4280 llvm::StringRef symbol_name_ref(symbol_name); 4281 if (symbol_name_ref.startswith("_OBJC_")) { 4282 llvm::StringRef g_objc_v2_prefix_class( 4283 "_OBJC_CLASS_$_"); 4284 llvm::StringRef g_objc_v2_prefix_metaclass( 4285 "_OBJC_METACLASS_$_"); 4286 llvm::StringRef g_objc_v2_prefix_ivar( 4287 "_OBJC_IVAR_$_"); 4288 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 4289 symbol_name_non_abi_mangled = symbol_name + 1; 4290 symbol_name = 4291 symbol_name + g_objc_v2_prefix_class.size(); 4292 type = eSymbolTypeObjCClass; 4293 demangled_is_synthesized = true; 4294 } else if (symbol_name_ref.startswith( 4295 g_objc_v2_prefix_metaclass)) { 4296 symbol_name_non_abi_mangled = symbol_name + 1; 4297 symbol_name = 4298 symbol_name + g_objc_v2_prefix_metaclass.size(); 4299 type = eSymbolTypeObjCMetaClass; 4300 demangled_is_synthesized = true; 4301 } else if (symbol_name_ref.startswith( 4302 g_objc_v2_prefix_ivar)) { 4303 symbol_name_non_abi_mangled = symbol_name + 1; 4304 symbol_name = 4305 symbol_name + g_objc_v2_prefix_ivar.size(); 4306 type = eSymbolTypeObjCIVar; 4307 demangled_is_synthesized = true; 4308 } 4309 } 4310 } 4311 } else if (symbol_sect_name && 4312 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4313 symbol_sect_name) { 4314 type = eSymbolTypeException; 4315 } else { 4316 type = eSymbolTypeData; 4317 } 4318 } else if (symbol_sect_name && 4319 ::strstr(symbol_sect_name, "__IMPORT") == 4320 symbol_sect_name) { 4321 type = eSymbolTypeTrampoline; 4322 } else if (symbol_section->IsDescendant(objc_section_sp.get())) { 4323 type = eSymbolTypeRuntime; 4324 if (symbol_name && symbol_name[0] == '.') { 4325 llvm::StringRef symbol_name_ref(symbol_name); 4326 llvm::StringRef g_objc_v1_prefix_class( 4327 ".objc_class_name_"); 4328 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4329 symbol_name_non_abi_mangled = symbol_name; 4330 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4331 type = eSymbolTypeObjCClass; 4332 demangled_is_synthesized = true; 4333 } 4334 } 4335 } 4336 } 4337 } 4338 } break; 4339 } 4340 } 4341 4342 if (!add_nlist) { 4343 sym[sym_idx].Clear(); 4344 return true; 4345 } 4346 4347 uint64_t symbol_value = nlist.n_value; 4348 4349 if (symbol_name_non_abi_mangled) { 4350 sym[sym_idx].GetMangled().SetMangledName( 4351 ConstString(symbol_name_non_abi_mangled)); 4352 sym[sym_idx].GetMangled().SetDemangledName(ConstString(symbol_name)); 4353 } else { 4354 bool symbol_name_is_mangled = false; 4355 4356 if (symbol_name && symbol_name[0] == '_') { 4357 symbol_name_is_mangled = symbol_name[1] == '_'; 4358 symbol_name++; // Skip the leading underscore 4359 } 4360 4361 if (symbol_name) { 4362 ConstString const_symbol_name(symbol_name); 4363 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4364 symbol_name_is_mangled); 4365 } 4366 } 4367 4368 if (is_gsym) { 4369 const char *gsym_name = sym[sym_idx] 4370 .GetMangled() 4371 .GetName(Mangled::ePreferMangled) 4372 .GetCString(); 4373 if (gsym_name) 4374 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4375 } 4376 4377 if (symbol_section) { 4378 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4379 if (symbol_byte_size == 0 && function_starts_count > 0) { 4380 addr_t symbol_lookup_file_addr = nlist.n_value; 4381 // Do an exact address match for non-ARM addresses, else get the 4382 // closest since the symbol might be a thumb symbol which has an 4383 // address with bit zero set. 4384 FunctionStarts::Entry *func_start_entry = 4385 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4386 if (is_arm && func_start_entry) { 4387 // Verify that the function start address is the symbol address 4388 // (ARM) or the symbol address + 1 (thumb). 4389 if (func_start_entry->addr != symbol_lookup_file_addr && 4390 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4391 // Not the right entry, NULL it out... 4392 func_start_entry = nullptr; 4393 } 4394 } 4395 if (func_start_entry) { 4396 func_start_entry->data = true; 4397 4398 addr_t symbol_file_addr = func_start_entry->addr; 4399 if (is_arm) 4400 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4401 4402 const FunctionStarts::Entry *next_func_start_entry = 4403 function_starts.FindNextEntry(func_start_entry); 4404 const addr_t section_end_file_addr = 4405 section_file_addr + symbol_section->GetByteSize(); 4406 if (next_func_start_entry) { 4407 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4408 // Be sure the clear the Thumb address bit when we calculate the 4409 // size from the current and next address 4410 if (is_arm) 4411 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4412 symbol_byte_size = std::min<lldb::addr_t>( 4413 next_symbol_file_addr - symbol_file_addr, 4414 section_end_file_addr - symbol_file_addr); 4415 } else { 4416 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4417 } 4418 } 4419 } 4420 symbol_value -= section_file_addr; 4421 } 4422 4423 if (!is_debug) { 4424 if (type == eSymbolTypeCode) { 4425 // See if we can find a N_FUN entry for any code symbols. If we do 4426 // find a match, and the name matches, then we can merge the two into 4427 // just the function symbol to avoid duplicate entries in the symbol 4428 // table. 4429 std::pair<ValueToSymbolIndexMap::const_iterator, 4430 ValueToSymbolIndexMap::const_iterator> 4431 range; 4432 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4433 if (range.first != range.second) { 4434 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4435 pos != range.second; ++pos) { 4436 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4437 sym[pos->second].GetMangled().GetName( 4438 Mangled::ePreferMangled)) { 4439 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4440 // We just need the flags from the linker symbol, so put these 4441 // flags into the N_FUN flags to avoid duplicate symbols in the 4442 // symbol table. 4443 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4444 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4445 if (resolver_addresses.find(nlist.n_value) != 4446 resolver_addresses.end()) 4447 sym[pos->second].SetType(eSymbolTypeResolver); 4448 sym[sym_idx].Clear(); 4449 return true; 4450 } 4451 } 4452 } else { 4453 if (resolver_addresses.find(nlist.n_value) != 4454 resolver_addresses.end()) 4455 type = eSymbolTypeResolver; 4456 } 4457 } else if (type == eSymbolTypeData || type == eSymbolTypeObjCClass || 4458 type == eSymbolTypeObjCMetaClass || 4459 type == eSymbolTypeObjCIVar) { 4460 // See if we can find a N_STSYM entry for any data symbols. If we do 4461 // find a match, and the name matches, then we can merge the two into 4462 // just the Static symbol to avoid duplicate entries in the symbol 4463 // table. 4464 std::pair<ValueToSymbolIndexMap::const_iterator, 4465 ValueToSymbolIndexMap::const_iterator> 4466 range; 4467 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4468 if (range.first != range.second) { 4469 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4470 pos != range.second; ++pos) { 4471 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4472 sym[pos->second].GetMangled().GetName( 4473 Mangled::ePreferMangled)) { 4474 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4475 // We just need the flags from the linker symbol, so put these 4476 // flags into the N_STSYM flags to avoid duplicate symbols in 4477 // the symbol table. 4478 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4479 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4480 sym[sym_idx].Clear(); 4481 return true; 4482 } 4483 } 4484 } else { 4485 // Combine N_GSYM stab entries with the non stab symbol. 4486 const char *gsym_name = sym[sym_idx] 4487 .GetMangled() 4488 .GetName(Mangled::ePreferMangled) 4489 .GetCString(); 4490 if (gsym_name) { 4491 ConstNameToSymbolIndexMap::const_iterator pos = 4492 N_GSYM_name_to_sym_idx.find(gsym_name); 4493 if (pos != N_GSYM_name_to_sym_idx.end()) { 4494 const uint32_t GSYM_sym_idx = pos->second; 4495 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4496 // Copy the address, because often the N_GSYM address has an 4497 // invalid address of zero when the global is a common symbol. 4498 sym[GSYM_sym_idx].GetAddressRef().SetSection(symbol_section); 4499 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4500 symbols_added.insert( 4501 sym[GSYM_sym_idx].GetAddress().GetFileAddress()); 4502 // We just need the flags from the linker symbol, so put these 4503 // flags into the N_GSYM flags to avoid duplicate symbols in 4504 // the symbol table. 4505 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4506 sym[sym_idx].Clear(); 4507 return true; 4508 } 4509 } 4510 } 4511 } 4512 } 4513 4514 sym[sym_idx].SetID(nlist_idx); 4515 sym[sym_idx].SetType(type); 4516 if (set_value) { 4517 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4518 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4519 symbols_added.insert(sym[sym_idx].GetAddress().GetFileAddress()); 4520 } 4521 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4522 if (nlist.n_desc & N_WEAK_REF) 4523 sym[sym_idx].SetIsWeak(true); 4524 4525 if (symbol_byte_size > 0) 4526 sym[sym_idx].SetByteSize(symbol_byte_size); 4527 4528 if (demangled_is_synthesized) 4529 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4530 4531 ++sym_idx; 4532 return true; 4533 }; 4534 4535 // First parse all the nlists but don't process them yet. See the next 4536 // comment for an explanation why. 4537 std::vector<struct nlist_64> nlists; 4538 nlists.reserve(symtab_load_command.nsyms); 4539 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 4540 if (auto nlist = 4541 ParseNList(nlist_data, nlist_data_offset, nlist_byte_size)) 4542 nlists.push_back(*nlist); 4543 else 4544 break; 4545 } 4546 4547 // Now parse all the debug symbols. This is needed to merge non-debug 4548 // symbols in the next step. Non-debug symbols are always coalesced into 4549 // the debug symbol. Doing this in one step would mean that some symbols 4550 // won't be merged. 4551 nlist_idx = 0; 4552 for (auto &nlist : nlists) { 4553 if (!ParseSymbolLambda(nlist, nlist_idx++, DebugSymbols)) 4554 break; 4555 } 4556 4557 // Finally parse all the non debug symbols. 4558 nlist_idx = 0; 4559 for (auto &nlist : nlists) { 4560 if (!ParseSymbolLambda(nlist, nlist_idx++, NonDebugSymbols)) 4561 break; 4562 } 4563 4564 for (const auto &pos : reexport_shlib_needs_fixup) { 4565 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4566 if (undef_pos != undefined_name_to_desc.end()) { 4567 const uint8_t dylib_ordinal = 4568 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4569 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4570 sym[pos.first].SetReExportedSymbolSharedLibrary( 4571 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4572 } 4573 } 4574 } 4575 4576 // Count how many trie symbols we'll add to the symbol table 4577 int trie_symbol_table_augment_count = 0; 4578 for (auto &e : external_sym_trie_entries) { 4579 if (symbols_added.find(e.entry.address) == symbols_added.end()) 4580 trie_symbol_table_augment_count++; 4581 } 4582 4583 if (num_syms < sym_idx + trie_symbol_table_augment_count) { 4584 num_syms = sym_idx + trie_symbol_table_augment_count; 4585 sym = symtab->Resize(num_syms); 4586 } 4587 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4588 4589 // Add symbols from the trie to the symbol table. 4590 for (auto &e : external_sym_trie_entries) { 4591 if (symbols_added.find(e.entry.address) != symbols_added.end()) 4592 continue; 4593 4594 // Find the section that this trie address is in, use that to annotate 4595 // symbol type as we add the trie address and name to the symbol table. 4596 Address symbol_addr; 4597 if (module_sp->ResolveFileAddress(e.entry.address, symbol_addr)) { 4598 SectionSP symbol_section(symbol_addr.GetSection()); 4599 const char *symbol_name = e.entry.name.GetCString(); 4600 bool demangled_is_synthesized = false; 4601 SymbolType type = 4602 GetSymbolType(symbol_name, demangled_is_synthesized, text_section_sp, 4603 data_section_sp, data_dirty_section_sp, 4604 data_const_section_sp, symbol_section); 4605 4606 sym[sym_idx].SetType(type); 4607 if (symbol_section) { 4608 sym[sym_idx].SetID(synthetic_sym_id++); 4609 sym[sym_idx].GetMangled().SetMangledName(ConstString(symbol_name)); 4610 if (demangled_is_synthesized) 4611 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4612 sym[sym_idx].SetIsSynthetic(true); 4613 sym[sym_idx].SetExternal(true); 4614 sym[sym_idx].GetAddressRef() = symbol_addr; 4615 symbols_added.insert(symbol_addr.GetFileAddress()); 4616 if (e.entry.flags & TRIE_SYMBOL_IS_THUMB) 4617 sym[sym_idx].SetFlags(MACHO_NLIST_ARM_SYMBOL_IS_THUMB); 4618 ++sym_idx; 4619 } 4620 } 4621 } 4622 4623 if (function_starts_count > 0) { 4624 uint32_t num_synthetic_function_symbols = 0; 4625 for (i = 0; i < function_starts_count; ++i) { 4626 if (symbols_added.find(function_starts.GetEntryRef(i).addr) == 4627 symbols_added.end()) 4628 ++num_synthetic_function_symbols; 4629 } 4630 4631 if (num_synthetic_function_symbols > 0) { 4632 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4633 num_syms = sym_idx + num_synthetic_function_symbols; 4634 sym = symtab->Resize(num_syms); 4635 } 4636 for (i = 0; i < function_starts_count; ++i) { 4637 const FunctionStarts::Entry *func_start_entry = 4638 function_starts.GetEntryAtIndex(i); 4639 if (symbols_added.find(func_start_entry->addr) == symbols_added.end()) { 4640 addr_t symbol_file_addr = func_start_entry->addr; 4641 uint32_t symbol_flags = 0; 4642 if (func_start_entry->data) 4643 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4644 Address symbol_addr; 4645 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4646 SectionSP symbol_section(symbol_addr.GetSection()); 4647 uint32_t symbol_byte_size = 0; 4648 if (symbol_section) { 4649 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4650 const FunctionStarts::Entry *next_func_start_entry = 4651 function_starts.FindNextEntry(func_start_entry); 4652 const addr_t section_end_file_addr = 4653 section_file_addr + symbol_section->GetByteSize(); 4654 if (next_func_start_entry) { 4655 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4656 if (is_arm) 4657 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4658 symbol_byte_size = std::min<lldb::addr_t>( 4659 next_symbol_file_addr - symbol_file_addr, 4660 section_end_file_addr - symbol_file_addr); 4661 } else { 4662 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4663 } 4664 sym[sym_idx].SetID(synthetic_sym_id++); 4665 sym[sym_idx].GetMangled().SetDemangledName( 4666 GetNextSyntheticSymbolName()); 4667 sym[sym_idx].SetType(eSymbolTypeCode); 4668 sym[sym_idx].SetIsSynthetic(true); 4669 sym[sym_idx].GetAddressRef() = symbol_addr; 4670 symbols_added.insert(symbol_addr.GetFileAddress()); 4671 if (symbol_flags) 4672 sym[sym_idx].SetFlags(symbol_flags); 4673 if (symbol_byte_size) 4674 sym[sym_idx].SetByteSize(symbol_byte_size); 4675 ++sym_idx; 4676 } 4677 } 4678 } 4679 } 4680 } 4681 } 4682 4683 // Trim our symbols down to just what we ended up with after removing any 4684 // symbols. 4685 if (sym_idx < num_syms) { 4686 num_syms = sym_idx; 4687 sym = symtab->Resize(num_syms); 4688 } 4689 4690 // Now synthesize indirect symbols 4691 if (m_dysymtab.nindirectsyms != 0) { 4692 if (indirect_symbol_index_data.GetByteSize()) { 4693 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4694 m_nlist_idx_to_sym_idx.end(); 4695 4696 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4697 ++sect_idx) { 4698 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4699 S_SYMBOL_STUBS) { 4700 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4701 if (symbol_stub_byte_size == 0) 4702 continue; 4703 4704 const uint32_t num_symbol_stubs = 4705 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4706 4707 if (num_symbol_stubs == 0) 4708 continue; 4709 4710 const uint32_t symbol_stub_index_offset = 4711 m_mach_sections[sect_idx].reserved1; 4712 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) { 4713 const uint32_t symbol_stub_index = 4714 symbol_stub_index_offset + stub_idx; 4715 const lldb::addr_t symbol_stub_addr = 4716 m_mach_sections[sect_idx].addr + 4717 (stub_idx * symbol_stub_byte_size); 4718 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4719 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4720 symbol_stub_offset, 4)) { 4721 const uint32_t stub_sym_id = 4722 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4723 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4724 continue; 4725 4726 NListIndexToSymbolIndexMap::const_iterator index_pos = 4727 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4728 Symbol *stub_symbol = nullptr; 4729 if (index_pos != end_index_pos) { 4730 // We have a remapping from the original nlist index to a 4731 // current symbol index, so just look this up by index 4732 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4733 } else { 4734 // We need to lookup a symbol using the original nlist symbol 4735 // index since this index is coming from the S_SYMBOL_STUBS 4736 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4737 } 4738 4739 if (stub_symbol) { 4740 Address so_addr(symbol_stub_addr, section_list); 4741 4742 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4743 // Change the external symbol into a trampoline that makes 4744 // sense These symbols were N_UNDF N_EXT, and are useless 4745 // to us, so we can re-use them so we don't have to make up 4746 // a synthetic symbol for no good reason. 4747 if (resolver_addresses.find(symbol_stub_addr) == 4748 resolver_addresses.end()) 4749 stub_symbol->SetType(eSymbolTypeTrampoline); 4750 else 4751 stub_symbol->SetType(eSymbolTypeResolver); 4752 stub_symbol->SetExternal(false); 4753 stub_symbol->GetAddressRef() = so_addr; 4754 stub_symbol->SetByteSize(symbol_stub_byte_size); 4755 } else { 4756 // Make a synthetic symbol to describe the trampoline stub 4757 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4758 if (sym_idx >= num_syms) { 4759 sym = symtab->Resize(++num_syms); 4760 stub_symbol = nullptr; // this pointer no longer valid 4761 } 4762 sym[sym_idx].SetID(synthetic_sym_id++); 4763 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4764 if (resolver_addresses.find(symbol_stub_addr) == 4765 resolver_addresses.end()) 4766 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4767 else 4768 sym[sym_idx].SetType(eSymbolTypeResolver); 4769 sym[sym_idx].SetIsSynthetic(true); 4770 sym[sym_idx].GetAddressRef() = so_addr; 4771 symbols_added.insert(so_addr.GetFileAddress()); 4772 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4773 ++sym_idx; 4774 } 4775 } else { 4776 if (log) 4777 log->Warning("symbol stub referencing symbol table symbol " 4778 "%u that isn't in our minimal symbol table, " 4779 "fix this!!!", 4780 stub_sym_id); 4781 } 4782 } 4783 } 4784 } 4785 } 4786 } 4787 } 4788 4789 if (!reexport_trie_entries.empty()) { 4790 for (const auto &e : reexport_trie_entries) { 4791 if (e.entry.import_name) { 4792 // Only add indirect symbols from the Trie entries if we didn't have 4793 // a N_INDR nlist entry for this already 4794 if (indirect_symbol_names.find(e.entry.name) == 4795 indirect_symbol_names.end()) { 4796 // Make a synthetic symbol to describe re-exported symbol. 4797 if (sym_idx >= num_syms) 4798 sym = symtab->Resize(++num_syms); 4799 sym[sym_idx].SetID(synthetic_sym_id++); 4800 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4801 sym[sym_idx].SetType(eSymbolTypeReExported); 4802 sym[sym_idx].SetIsSynthetic(true); 4803 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4804 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4805 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4806 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4807 } 4808 ++sym_idx; 4809 } 4810 } 4811 } 4812 } 4813 4814 // StreamFile s(stdout, false); 4815 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4816 // symtab->Dump(&s, NULL, eSortOrderNone); 4817 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4818 // sizes 4819 symtab->CalculateSymbolSizes(); 4820 4821 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4822 // symtab->Dump(&s, NULL, eSortOrderNone); 4823 4824 return symtab->GetNumSymbols(); 4825 } 4826 4827 void ObjectFileMachO::Dump(Stream *s) { 4828 ModuleSP module_sp(GetModule()); 4829 if (module_sp) { 4830 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4831 s->Printf("%p: ", static_cast<void *>(this)); 4832 s->Indent(); 4833 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4834 s->PutCString("ObjectFileMachO64"); 4835 else 4836 s->PutCString("ObjectFileMachO32"); 4837 4838 *s << ", file = '" << m_file; 4839 ModuleSpecList all_specs; 4840 ModuleSpec base_spec; 4841 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 4842 base_spec, all_specs); 4843 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 4844 *s << "', triple"; 4845 if (e) 4846 s->Printf("[%d]", i); 4847 *s << " = "; 4848 *s << all_specs.GetModuleSpecRefAtIndex(i) 4849 .GetArchitecture() 4850 .GetTriple() 4851 .getTriple(); 4852 } 4853 *s << "\n"; 4854 SectionList *sections = GetSectionList(); 4855 if (sections) 4856 sections->Dump(s->AsRawOstream(), s->GetIndentLevel(), nullptr, true, 4857 UINT32_MAX); 4858 4859 if (m_symtab_up) 4860 m_symtab_up->Dump(s, nullptr, eSortOrderNone); 4861 } 4862 } 4863 4864 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4865 const lldb_private::DataExtractor &data, 4866 lldb::offset_t lc_offset) { 4867 uint32_t i; 4868 struct uuid_command load_cmd; 4869 4870 lldb::offset_t offset = lc_offset; 4871 for (i = 0; i < header.ncmds; ++i) { 4872 const lldb::offset_t cmd_offset = offset; 4873 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4874 break; 4875 4876 if (load_cmd.cmd == LC_UUID) { 4877 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4878 4879 if (uuid_bytes) { 4880 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4881 // We pretend these object files have no UUID to prevent crashing. 4882 4883 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4884 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4885 0xbb, 0x14, 0xf0, 0x0d}; 4886 4887 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4888 return UUID(); 4889 4890 return UUID::fromOptionalData(uuid_bytes, 16); 4891 } 4892 return UUID(); 4893 } 4894 offset = cmd_offset + load_cmd.cmdsize; 4895 } 4896 return UUID(); 4897 } 4898 4899 static llvm::StringRef GetOSName(uint32_t cmd) { 4900 switch (cmd) { 4901 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4902 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4903 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4904 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4905 case llvm::MachO::LC_VERSION_MIN_TVOS: 4906 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4907 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4908 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4909 default: 4910 llvm_unreachable("unexpected LC_VERSION load command"); 4911 } 4912 } 4913 4914 namespace { 4915 struct OSEnv { 4916 llvm::StringRef os_type; 4917 llvm::StringRef environment; 4918 OSEnv(uint32_t cmd) { 4919 switch (cmd) { 4920 case llvm::MachO::PLATFORM_MACOS: 4921 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4922 return; 4923 case llvm::MachO::PLATFORM_IOS: 4924 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4925 return; 4926 case llvm::MachO::PLATFORM_TVOS: 4927 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4928 return; 4929 case llvm::MachO::PLATFORM_WATCHOS: 4930 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4931 return; 4932 // NEED_BRIDGEOS_TRIPLE case llvm::MachO::PLATFORM_BRIDGEOS: 4933 // NEED_BRIDGEOS_TRIPLE os_type = 4934 // llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4935 // NEED_BRIDGEOS_TRIPLE return; 4936 case llvm::MachO::PLATFORM_MACCATALYST: 4937 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4938 environment = llvm::Triple::getEnvironmentTypeName(llvm::Triple::MacABI); 4939 return; 4940 case llvm::MachO::PLATFORM_IOSSIMULATOR: 4941 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4942 environment = 4943 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4944 return; 4945 case llvm::MachO::PLATFORM_TVOSSIMULATOR: 4946 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4947 environment = 4948 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4949 return; 4950 case llvm::MachO::PLATFORM_WATCHOSSIMULATOR: 4951 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4952 environment = 4953 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4954 return; 4955 default: { 4956 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 4957 LIBLLDB_LOG_PROCESS)); 4958 LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION"); 4959 } 4960 } 4961 } 4962 }; 4963 4964 struct MinOS { 4965 uint32_t major_version, minor_version, patch_version; 4966 MinOS(uint32_t version) 4967 : major_version(version >> 16), minor_version((version >> 8) & 0xffu), 4968 patch_version(version & 0xffu) {} 4969 }; 4970 } // namespace 4971 4972 void ObjectFileMachO::GetAllArchSpecs(const llvm::MachO::mach_header &header, 4973 const lldb_private::DataExtractor &data, 4974 lldb::offset_t lc_offset, 4975 ModuleSpec &base_spec, 4976 lldb_private::ModuleSpecList &all_specs) { 4977 auto &base_arch = base_spec.GetArchitecture(); 4978 base_arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4979 if (!base_arch.IsValid()) 4980 return; 4981 4982 bool found_any = false; 4983 auto add_triple = [&](const llvm::Triple &triple) { 4984 auto spec = base_spec; 4985 spec.GetArchitecture().GetTriple() = triple; 4986 if (spec.GetArchitecture().IsValid()) { 4987 spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset); 4988 all_specs.Append(spec); 4989 found_any = true; 4990 } 4991 }; 4992 4993 // Set OS to an unspecified unknown or a "*" so it can match any OS 4994 llvm::Triple base_triple = base_arch.GetTriple(); 4995 base_triple.setOS(llvm::Triple::UnknownOS); 4996 base_triple.setOSName(llvm::StringRef()); 4997 4998 if (header.filetype == MH_PRELOAD) { 4999 if (header.cputype == CPU_TYPE_ARM) { 5000 // If this is a 32-bit arm binary, and it's a standalone binary, force 5001 // the Vendor to Apple so we don't accidentally pick up the generic 5002 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 5003 // frame pointer register; most other armv7 ABIs use a combination of 5004 // r7 and r11. 5005 base_triple.setVendor(llvm::Triple::Apple); 5006 } else { 5007 // Set vendor to an unspecified unknown or a "*" so it can match any 5008 // vendor This is required for correct behavior of EFI debugging on 5009 // x86_64 5010 base_triple.setVendor(llvm::Triple::UnknownVendor); 5011 base_triple.setVendorName(llvm::StringRef()); 5012 } 5013 return add_triple(base_triple); 5014 } 5015 5016 struct load_command load_cmd; 5017 5018 // See if there is an LC_VERSION_MIN_* load command that can give 5019 // us the OS type. 5020 lldb::offset_t offset = lc_offset; 5021 for (uint32_t i = 0; i < header.ncmds; ++i) { 5022 const lldb::offset_t cmd_offset = offset; 5023 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 5024 break; 5025 5026 struct version_min_command version_min; 5027 switch (load_cmd.cmd) { 5028 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 5029 case llvm::MachO::LC_VERSION_MIN_MACOSX: 5030 case llvm::MachO::LC_VERSION_MIN_TVOS: 5031 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 5032 if (load_cmd.cmdsize != sizeof(version_min)) 5033 break; 5034 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 5035 data.GetByteOrder(), &version_min) == 0) 5036 break; 5037 MinOS min_os(version_min.version); 5038 llvm::SmallString<32> os_name; 5039 llvm::raw_svector_ostream os(os_name); 5040 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 5041 << min_os.minor_version << '.' << min_os.patch_version; 5042 5043 auto triple = base_triple; 5044 triple.setOSName(os.str()); 5045 os_name.clear(); 5046 add_triple(triple); 5047 break; 5048 } 5049 default: 5050 break; 5051 } 5052 5053 offset = cmd_offset + load_cmd.cmdsize; 5054 } 5055 5056 // See if there are LC_BUILD_VERSION load commands that can give 5057 // us the OS type. 5058 offset = lc_offset; 5059 for (uint32_t i = 0; i < header.ncmds; ++i) { 5060 const lldb::offset_t cmd_offset = offset; 5061 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 5062 break; 5063 5064 do { 5065 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 5066 struct build_version_command build_version; 5067 if (load_cmd.cmdsize < sizeof(build_version)) { 5068 // Malformed load command. 5069 break; 5070 } 5071 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 5072 data.GetByteOrder(), &build_version) == 0) 5073 break; 5074 MinOS min_os(build_version.minos); 5075 OSEnv os_env(build_version.platform); 5076 llvm::SmallString<16> os_name; 5077 llvm::raw_svector_ostream os(os_name); 5078 os << os_env.os_type << min_os.major_version << '.' 5079 << min_os.minor_version << '.' << min_os.patch_version; 5080 auto triple = base_triple; 5081 triple.setOSName(os.str()); 5082 os_name.clear(); 5083 if (!os_env.environment.empty()) 5084 triple.setEnvironmentName(os_env.environment); 5085 add_triple(triple); 5086 } 5087 } while (0); 5088 offset = cmd_offset + load_cmd.cmdsize; 5089 } 5090 5091 if (!found_any) { 5092 if (header.filetype == MH_KEXT_BUNDLE) { 5093 base_triple.setVendor(llvm::Triple::Apple); 5094 add_triple(base_triple); 5095 } else { 5096 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 5097 // so lets not say our Vendor is Apple, leave it as an unspecified 5098 // unknown. 5099 base_triple.setVendor(llvm::Triple::UnknownVendor); 5100 base_triple.setVendorName(llvm::StringRef()); 5101 add_triple(base_triple); 5102 } 5103 } 5104 } 5105 5106 ArchSpec ObjectFileMachO::GetArchitecture( 5107 ModuleSP module_sp, const llvm::MachO::mach_header &header, 5108 const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) { 5109 ModuleSpecList all_specs; 5110 ModuleSpec base_spec; 5111 GetAllArchSpecs(header, data, MachHeaderSizeFromMagic(header.magic), 5112 base_spec, all_specs); 5113 5114 // If the object file offers multiple alternative load commands, 5115 // pick the one that matches the module. 5116 if (module_sp) { 5117 const ArchSpec &module_arch = module_sp->GetArchitecture(); 5118 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 5119 ArchSpec mach_arch = 5120 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 5121 if (module_arch.IsCompatibleMatch(mach_arch)) 5122 return mach_arch; 5123 } 5124 } 5125 5126 // Return the first arch we found. 5127 if (all_specs.GetSize() == 0) 5128 return {}; 5129 return all_specs.GetModuleSpecRefAtIndex(0).GetArchitecture(); 5130 } 5131 5132 UUID ObjectFileMachO::GetUUID() { 5133 ModuleSP module_sp(GetModule()); 5134 if (module_sp) { 5135 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5136 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5137 return GetUUID(m_header, m_data, offset); 5138 } 5139 return UUID(); 5140 } 5141 5142 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5143 uint32_t count = 0; 5144 ModuleSP module_sp(GetModule()); 5145 if (module_sp) { 5146 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5147 struct load_command load_cmd; 5148 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5149 std::vector<std::string> rpath_paths; 5150 std::vector<std::string> rpath_relative_paths; 5151 std::vector<std::string> at_exec_relative_paths; 5152 uint32_t i; 5153 for (i = 0; i < m_header.ncmds; ++i) { 5154 const uint32_t cmd_offset = offset; 5155 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5156 break; 5157 5158 switch (load_cmd.cmd) { 5159 case LC_RPATH: 5160 case LC_LOAD_DYLIB: 5161 case LC_LOAD_WEAK_DYLIB: 5162 case LC_REEXPORT_DYLIB: 5163 case LC_LOAD_DYLINKER: 5164 case LC_LOADFVMLIB: 5165 case LC_LOAD_UPWARD_DYLIB: { 5166 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5167 const char *path = m_data.PeekCStr(name_offset); 5168 if (path) { 5169 if (load_cmd.cmd == LC_RPATH) 5170 rpath_paths.push_back(path); 5171 else { 5172 if (path[0] == '@') { 5173 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5174 rpath_relative_paths.push_back(path + strlen("@rpath")); 5175 else if (strncmp(path, "@executable_path", 5176 strlen("@executable_path")) == 0) 5177 at_exec_relative_paths.push_back(path + 5178 strlen("@executable_path")); 5179 } else { 5180 FileSpec file_spec(path); 5181 if (files.AppendIfUnique(file_spec)) 5182 count++; 5183 } 5184 } 5185 } 5186 } break; 5187 5188 default: 5189 break; 5190 } 5191 offset = cmd_offset + load_cmd.cmdsize; 5192 } 5193 5194 FileSpec this_file_spec(m_file); 5195 FileSystem::Instance().Resolve(this_file_spec); 5196 5197 if (!rpath_paths.empty()) { 5198 // Fixup all LC_RPATH values to be absolute paths 5199 std::string loader_path("@loader_path"); 5200 std::string executable_path("@executable_path"); 5201 for (auto &rpath : rpath_paths) { 5202 if (llvm::StringRef(rpath).startswith(loader_path)) { 5203 rpath.erase(0, loader_path.size()); 5204 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5205 } else if (llvm::StringRef(rpath).startswith(executable_path)) { 5206 rpath.erase(0, executable_path.size()); 5207 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5208 } 5209 } 5210 5211 for (const auto &rpath_relative_path : rpath_relative_paths) { 5212 for (const auto &rpath : rpath_paths) { 5213 std::string path = rpath; 5214 path += rpath_relative_path; 5215 // It is OK to resolve this path because we must find a file on disk 5216 // for us to accept it anyway if it is rpath relative. 5217 FileSpec file_spec(path); 5218 FileSystem::Instance().Resolve(file_spec); 5219 if (FileSystem::Instance().Exists(file_spec) && 5220 files.AppendIfUnique(file_spec)) { 5221 count++; 5222 break; 5223 } 5224 } 5225 } 5226 } 5227 5228 // We may have @executable_paths but no RPATHS. Figure those out here. 5229 // Only do this if this object file is the executable. We have no way to 5230 // get back to the actual executable otherwise, so we won't get the right 5231 // path. 5232 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5233 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5234 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5235 FileSpec file_spec = 5236 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5237 if (FileSystem::Instance().Exists(file_spec) && 5238 files.AppendIfUnique(file_spec)) 5239 count++; 5240 } 5241 } 5242 } 5243 return count; 5244 } 5245 5246 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5247 // If the object file is not an executable it can't hold the entry point. 5248 // m_entry_point_address is initialized to an invalid address, so we can just 5249 // return that. If m_entry_point_address is valid it means we've found it 5250 // already, so return the cached value. 5251 5252 if ((!IsExecutable() && !IsDynamicLoader()) || 5253 m_entry_point_address.IsValid()) { 5254 return m_entry_point_address; 5255 } 5256 5257 // Otherwise, look for the UnixThread or Thread command. The data for the 5258 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5259 // 5260 // uint32_t flavor - this is the flavor argument you would pass to 5261 // thread_get_state 5262 // uint32_t count - this is the count of longs in the thread state data 5263 // struct XXX_thread_state state - this is the structure from 5264 // <machine/thread_status.h> corresponding to the flavor. 5265 // <repeat this trio> 5266 // 5267 // So we just keep reading the various register flavors till we find the GPR 5268 // one, then read the PC out of there. 5269 // FIXME: We will need to have a "RegisterContext data provider" class at some 5270 // point that can get all the registers 5271 // out of data in this form & attach them to a given thread. That should 5272 // underlie the MacOS X User process plugin, and we'll also need it for the 5273 // MacOS X Core File process plugin. When we have that we can also use it 5274 // here. 5275 // 5276 // For now we hard-code the offsets and flavors we need: 5277 // 5278 // 5279 5280 ModuleSP module_sp(GetModule()); 5281 if (module_sp) { 5282 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5283 struct load_command load_cmd; 5284 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5285 uint32_t i; 5286 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5287 bool done = false; 5288 5289 for (i = 0; i < m_header.ncmds; ++i) { 5290 const lldb::offset_t cmd_offset = offset; 5291 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5292 break; 5293 5294 switch (load_cmd.cmd) { 5295 case LC_UNIXTHREAD: 5296 case LC_THREAD: { 5297 while (offset < cmd_offset + load_cmd.cmdsize) { 5298 uint32_t flavor = m_data.GetU32(&offset); 5299 uint32_t count = m_data.GetU32(&offset); 5300 if (count == 0) { 5301 // We've gotten off somehow, log and exit; 5302 return m_entry_point_address; 5303 } 5304 5305 switch (m_header.cputype) { 5306 case llvm::MachO::CPU_TYPE_ARM: 5307 if (flavor == 1 || 5308 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 5309 // from mach/arm/thread_status.h 5310 { 5311 offset += 60; // This is the offset of pc in the GPR thread state 5312 // data structure. 5313 start_address = m_data.GetU32(&offset); 5314 done = true; 5315 } 5316 break; 5317 case llvm::MachO::CPU_TYPE_ARM64: 5318 case llvm::MachO::CPU_TYPE_ARM64_32: 5319 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5320 { 5321 offset += 256; // This is the offset of pc in the GPR thread state 5322 // data structure. 5323 start_address = m_data.GetU64(&offset); 5324 done = true; 5325 } 5326 break; 5327 case llvm::MachO::CPU_TYPE_I386: 5328 if (flavor == 5329 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5330 { 5331 offset += 40; // This is the offset of eip in the GPR thread state 5332 // data structure. 5333 start_address = m_data.GetU32(&offset); 5334 done = true; 5335 } 5336 break; 5337 case llvm::MachO::CPU_TYPE_X86_64: 5338 if (flavor == 5339 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5340 { 5341 offset += 16 * 8; // This is the offset of rip in the GPR thread 5342 // state data structure. 5343 start_address = m_data.GetU64(&offset); 5344 done = true; 5345 } 5346 break; 5347 default: 5348 return m_entry_point_address; 5349 } 5350 // Haven't found the GPR flavor yet, skip over the data for this 5351 // flavor: 5352 if (done) 5353 break; 5354 offset += count * 4; 5355 } 5356 } break; 5357 case LC_MAIN: { 5358 ConstString text_segment_name("__TEXT"); 5359 uint64_t entryoffset = m_data.GetU64(&offset); 5360 SectionSP text_segment_sp = 5361 GetSectionList()->FindSectionByName(text_segment_name); 5362 if (text_segment_sp) { 5363 done = true; 5364 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5365 } 5366 } break; 5367 5368 default: 5369 break; 5370 } 5371 if (done) 5372 break; 5373 5374 // Go to the next load command: 5375 offset = cmd_offset + load_cmd.cmdsize; 5376 } 5377 5378 if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) { 5379 if (GetSymtab()) { 5380 Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType( 5381 ConstString("_dyld_start"), SymbolType::eSymbolTypeCode, 5382 Symtab::eDebugAny, Symtab::eVisibilityAny); 5383 if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) { 5384 start_address = dyld_start_sym->GetAddress().GetFileAddress(); 5385 } 5386 } 5387 } 5388 5389 if (start_address != LLDB_INVALID_ADDRESS) { 5390 // We got the start address from the load commands, so now resolve that 5391 // address in the sections of this ObjectFile: 5392 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5393 start_address, GetSectionList())) { 5394 m_entry_point_address.Clear(); 5395 } 5396 } else { 5397 // We couldn't read the UnixThread load command - maybe it wasn't there. 5398 // As a fallback look for the "start" symbol in the main executable. 5399 5400 ModuleSP module_sp(GetModule()); 5401 5402 if (module_sp) { 5403 SymbolContextList contexts; 5404 SymbolContext context; 5405 module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5406 eSymbolTypeCode, contexts); 5407 if (contexts.GetSize()) { 5408 if (contexts.GetContextAtIndex(0, context)) 5409 m_entry_point_address = context.symbol->GetAddress(); 5410 } 5411 } 5412 } 5413 } 5414 5415 return m_entry_point_address; 5416 } 5417 5418 lldb_private::Address ObjectFileMachO::GetBaseAddress() { 5419 lldb_private::Address header_addr; 5420 SectionList *section_list = GetSectionList(); 5421 if (section_list) { 5422 SectionSP text_segment_sp( 5423 section_list->FindSectionByName(GetSegmentNameTEXT())); 5424 if (text_segment_sp) { 5425 header_addr.SetSection(text_segment_sp); 5426 header_addr.SetOffset(0); 5427 } 5428 } 5429 return header_addr; 5430 } 5431 5432 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5433 ModuleSP module_sp(GetModule()); 5434 if (module_sp) { 5435 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5436 if (!m_thread_context_offsets_valid) { 5437 m_thread_context_offsets_valid = true; 5438 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5439 FileRangeArray::Entry file_range; 5440 thread_command thread_cmd; 5441 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5442 const uint32_t cmd_offset = offset; 5443 if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr) 5444 break; 5445 5446 if (thread_cmd.cmd == LC_THREAD) { 5447 file_range.SetRangeBase(offset); 5448 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5449 m_thread_context_offsets.Append(file_range); 5450 } 5451 offset = cmd_offset + thread_cmd.cmdsize; 5452 } 5453 } 5454 } 5455 return m_thread_context_offsets.GetSize(); 5456 } 5457 5458 std::string ObjectFileMachO::GetIdentifierString() { 5459 std::string result; 5460 ModuleSP module_sp(GetModule()); 5461 if (module_sp) { 5462 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5463 5464 // First, look over the load commands for an LC_NOTE load command with 5465 // data_owner string "kern ver str" & use that if found. 5466 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5467 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5468 const uint32_t cmd_offset = offset; 5469 load_command lc; 5470 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5471 break; 5472 if (lc.cmd == LC_NOTE) { 5473 char data_owner[17]; 5474 m_data.CopyData(offset, 16, data_owner); 5475 data_owner[16] = '\0'; 5476 offset += 16; 5477 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5478 uint64_t size = m_data.GetU64_unchecked(&offset); 5479 5480 // "kern ver str" has a uint32_t version and then a nul terminated 5481 // c-string. 5482 if (strcmp("kern ver str", data_owner) == 0) { 5483 offset = fileoff; 5484 uint32_t version; 5485 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5486 if (version == 1) { 5487 uint32_t strsize = size - sizeof(uint32_t); 5488 char *buf = (char *)malloc(strsize); 5489 if (buf) { 5490 m_data.CopyData(offset, strsize, buf); 5491 buf[strsize - 1] = '\0'; 5492 result = buf; 5493 if (buf) 5494 free(buf); 5495 return result; 5496 } 5497 } 5498 } 5499 } 5500 } 5501 offset = cmd_offset + lc.cmdsize; 5502 } 5503 5504 // Second, make a pass over the load commands looking for an obsolete 5505 // LC_IDENT load command. 5506 offset = MachHeaderSizeFromMagic(m_header.magic); 5507 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5508 const uint32_t cmd_offset = offset; 5509 struct ident_command ident_command; 5510 if (m_data.GetU32(&offset, &ident_command, 2) == nullptr) 5511 break; 5512 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5513 char *buf = (char *)malloc(ident_command.cmdsize); 5514 if (buf != nullptr && m_data.CopyData(offset, ident_command.cmdsize, 5515 buf) == ident_command.cmdsize) { 5516 buf[ident_command.cmdsize - 1] = '\0'; 5517 result = buf; 5518 } 5519 if (buf) 5520 free(buf); 5521 } 5522 offset = cmd_offset + ident_command.cmdsize; 5523 } 5524 } 5525 return result; 5526 } 5527 5528 bool ObjectFileMachO::GetCorefileMainBinaryInfo(addr_t &address, UUID &uuid) { 5529 address = LLDB_INVALID_ADDRESS; 5530 uuid.Clear(); 5531 ModuleSP module_sp(GetModule()); 5532 if (module_sp) { 5533 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5534 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5535 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5536 const uint32_t cmd_offset = offset; 5537 load_command lc; 5538 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5539 break; 5540 if (lc.cmd == LC_NOTE) { 5541 char data_owner[17]; 5542 memset(data_owner, 0, sizeof(data_owner)); 5543 m_data.CopyData(offset, 16, data_owner); 5544 offset += 16; 5545 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5546 uint64_t size = m_data.GetU64_unchecked(&offset); 5547 5548 // "main bin spec" (main binary specification) data payload is 5549 // formatted: 5550 // uint32_t version [currently 1] 5551 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user 5552 // process] uint64_t address [ UINT64_MAX if address not 5553 // specified ] uuid_t uuid [ all zero's if uuid not 5554 // specified ] uint32_t log2_pagesize [ process page size in log base 5555 // 2, e.g. 4k pages are 12. 0 for unspecified ] 5556 5557 if (strcmp("main bin spec", data_owner) == 0 && size >= 32) { 5558 offset = fileoff; 5559 uint32_t version; 5560 if (m_data.GetU32(&offset, &version, 1) != nullptr && version == 1) { 5561 uint32_t type = 0; 5562 uuid_t raw_uuid; 5563 memset(raw_uuid, 0, sizeof(uuid_t)); 5564 5565 if (m_data.GetU32(&offset, &type, 1) && 5566 m_data.GetU64(&offset, &address, 1) && 5567 m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5568 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5569 return true; 5570 } 5571 } 5572 } 5573 } 5574 offset = cmd_offset + lc.cmdsize; 5575 } 5576 } 5577 return false; 5578 } 5579 5580 lldb::RegisterContextSP 5581 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5582 lldb_private::Thread &thread) { 5583 lldb::RegisterContextSP reg_ctx_sp; 5584 5585 ModuleSP module_sp(GetModule()); 5586 if (module_sp) { 5587 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5588 if (!m_thread_context_offsets_valid) 5589 GetNumThreadContexts(); 5590 5591 const FileRangeArray::Entry *thread_context_file_range = 5592 m_thread_context_offsets.GetEntryAtIndex(idx); 5593 if (thread_context_file_range) { 5594 5595 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5596 thread_context_file_range->GetByteSize()); 5597 5598 switch (m_header.cputype) { 5599 case llvm::MachO::CPU_TYPE_ARM64: 5600 case llvm::MachO::CPU_TYPE_ARM64_32: 5601 reg_ctx_sp = 5602 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data); 5603 break; 5604 5605 case llvm::MachO::CPU_TYPE_ARM: 5606 reg_ctx_sp = 5607 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data); 5608 break; 5609 5610 case llvm::MachO::CPU_TYPE_I386: 5611 reg_ctx_sp = 5612 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data); 5613 break; 5614 5615 case llvm::MachO::CPU_TYPE_X86_64: 5616 reg_ctx_sp = 5617 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data); 5618 break; 5619 } 5620 } 5621 } 5622 return reg_ctx_sp; 5623 } 5624 5625 ObjectFile::Type ObjectFileMachO::CalculateType() { 5626 switch (m_header.filetype) { 5627 case MH_OBJECT: // 0x1u 5628 if (GetAddressByteSize() == 4) { 5629 // 32 bit kexts are just object files, but they do have a valid 5630 // UUID load command. 5631 if (GetUUID()) { 5632 // this checking for the UUID load command is not enough we could 5633 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5634 // this is required of kexts 5635 if (m_strata == eStrataInvalid) 5636 m_strata = eStrataKernel; 5637 return eTypeSharedLibrary; 5638 } 5639 } 5640 return eTypeObjectFile; 5641 5642 case MH_EXECUTE: 5643 return eTypeExecutable; // 0x2u 5644 case MH_FVMLIB: 5645 return eTypeSharedLibrary; // 0x3u 5646 case MH_CORE: 5647 return eTypeCoreFile; // 0x4u 5648 case MH_PRELOAD: 5649 return eTypeSharedLibrary; // 0x5u 5650 case MH_DYLIB: 5651 return eTypeSharedLibrary; // 0x6u 5652 case MH_DYLINKER: 5653 return eTypeDynamicLinker; // 0x7u 5654 case MH_BUNDLE: 5655 return eTypeSharedLibrary; // 0x8u 5656 case MH_DYLIB_STUB: 5657 return eTypeStubLibrary; // 0x9u 5658 case MH_DSYM: 5659 return eTypeDebugInfo; // 0xAu 5660 case MH_KEXT_BUNDLE: 5661 return eTypeSharedLibrary; // 0xBu 5662 default: 5663 break; 5664 } 5665 return eTypeUnknown; 5666 } 5667 5668 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5669 switch (m_header.filetype) { 5670 case MH_OBJECT: // 0x1u 5671 { 5672 // 32 bit kexts are just object files, but they do have a valid 5673 // UUID load command. 5674 if (GetUUID()) { 5675 // this checking for the UUID load command is not enough we could 5676 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5677 // this is required of kexts 5678 if (m_type == eTypeInvalid) 5679 m_type = eTypeSharedLibrary; 5680 5681 return eStrataKernel; 5682 } 5683 } 5684 return eStrataUnknown; 5685 5686 case MH_EXECUTE: // 0x2u 5687 // Check for the MH_DYLDLINK bit in the flags 5688 if (m_header.flags & MH_DYLDLINK) { 5689 return eStrataUser; 5690 } else { 5691 SectionList *section_list = GetSectionList(); 5692 if (section_list) { 5693 static ConstString g_kld_section_name("__KLD"); 5694 if (section_list->FindSectionByName(g_kld_section_name)) 5695 return eStrataKernel; 5696 } 5697 } 5698 return eStrataRawImage; 5699 5700 case MH_FVMLIB: 5701 return eStrataUser; // 0x3u 5702 case MH_CORE: 5703 return eStrataUnknown; // 0x4u 5704 case MH_PRELOAD: 5705 return eStrataRawImage; // 0x5u 5706 case MH_DYLIB: 5707 return eStrataUser; // 0x6u 5708 case MH_DYLINKER: 5709 return eStrataUser; // 0x7u 5710 case MH_BUNDLE: 5711 return eStrataUser; // 0x8u 5712 case MH_DYLIB_STUB: 5713 return eStrataUser; // 0x9u 5714 case MH_DSYM: 5715 return eStrataUnknown; // 0xAu 5716 case MH_KEXT_BUNDLE: 5717 return eStrataKernel; // 0xBu 5718 default: 5719 break; 5720 } 5721 return eStrataUnknown; 5722 } 5723 5724 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5725 ModuleSP module_sp(GetModule()); 5726 if (module_sp) { 5727 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5728 struct dylib_command load_cmd; 5729 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5730 uint32_t version_cmd = 0; 5731 uint64_t version = 0; 5732 uint32_t i; 5733 for (i = 0; i < m_header.ncmds; ++i) { 5734 const lldb::offset_t cmd_offset = offset; 5735 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5736 break; 5737 5738 if (load_cmd.cmd == LC_ID_DYLIB) { 5739 if (version_cmd == 0) { 5740 version_cmd = load_cmd.cmd; 5741 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr) 5742 break; 5743 version = load_cmd.dylib.current_version; 5744 } 5745 break; // Break for now unless there is another more complete version 5746 // number load command in the future. 5747 } 5748 offset = cmd_offset + load_cmd.cmdsize; 5749 } 5750 5751 if (version_cmd == LC_ID_DYLIB) { 5752 unsigned major = (version & 0xFFFF0000ull) >> 16; 5753 unsigned minor = (version & 0x0000FF00ull) >> 8; 5754 unsigned subminor = (version & 0x000000FFull); 5755 return llvm::VersionTuple(major, minor, subminor); 5756 } 5757 } 5758 return llvm::VersionTuple(); 5759 } 5760 5761 ArchSpec ObjectFileMachO::GetArchitecture() { 5762 ModuleSP module_sp(GetModule()); 5763 ArchSpec arch; 5764 if (module_sp) { 5765 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5766 5767 return GetArchitecture(module_sp, m_header, m_data, 5768 MachHeaderSizeFromMagic(m_header.magic)); 5769 } 5770 return arch; 5771 } 5772 5773 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, 5774 addr_t &base_addr, UUID &uuid) { 5775 uuid.Clear(); 5776 base_addr = LLDB_INVALID_ADDRESS; 5777 if (process && process->GetDynamicLoader()) { 5778 DynamicLoader *dl = process->GetDynamicLoader(); 5779 LazyBool using_shared_cache; 5780 LazyBool private_shared_cache; 5781 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5782 private_shared_cache); 5783 } 5784 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5785 LIBLLDB_LOG_PROCESS)); 5786 LLDB_LOGF( 5787 log, 5788 "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64, 5789 uuid.GetAsString().c_str(), base_addr); 5790 } 5791 5792 // From dyld SPI header dyld_process_info.h 5793 typedef void *dyld_process_info; 5794 struct lldb_copy__dyld_process_cache_info { 5795 uuid_t cacheUUID; // UUID of cache used by process 5796 uint64_t cacheBaseAddress; // load address of dyld shared cache 5797 bool noCache; // process is running without a dyld cache 5798 bool privateCache; // process is using a private copy of its dyld cache 5799 }; 5800 5801 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with 5802 // llvm enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile 5803 // errors. So we need to use the actual underlying types of task_t and 5804 // kern_return_t below. 5805 extern "C" unsigned int /*task_t*/ mach_task_self(); 5806 5807 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5808 uuid.Clear(); 5809 base_addr = LLDB_INVALID_ADDRESS; 5810 5811 #if defined(__APPLE__) && \ 5812 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5813 uint8_t *(*dyld_get_all_image_infos)(void); 5814 dyld_get_all_image_infos = 5815 (uint8_t * (*)()) dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5816 if (dyld_get_all_image_infos) { 5817 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5818 if (dyld_all_image_infos_address) { 5819 uint32_t *version = (uint32_t *) 5820 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5821 if (*version >= 13) { 5822 uuid_t *sharedCacheUUID_address = 0; 5823 int wordsize = sizeof(uint8_t *); 5824 if (wordsize == 8) { 5825 sharedCacheUUID_address = 5826 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5827 160); // sharedCacheUUID <mach-o/dyld_images.h> 5828 if (*version >= 15) 5829 base_addr = 5830 *(uint64_t 5831 *)((uint8_t *)dyld_all_image_infos_address + 5832 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5833 } else { 5834 sharedCacheUUID_address = 5835 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5836 84); // sharedCacheUUID <mach-o/dyld_images.h> 5837 if (*version >= 15) { 5838 base_addr = 0; 5839 base_addr = 5840 *(uint32_t 5841 *)((uint8_t *)dyld_all_image_infos_address + 5842 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5843 } 5844 } 5845 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5846 } 5847 } 5848 } else { 5849 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5850 dyld_process_info (*dyld_process_info_create)( 5851 unsigned int /* task_t */ task, uint64_t timestamp, 5852 unsigned int /*kern_return_t*/ *kernelError); 5853 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5854 void (*dyld_process_info_release)(dyld_process_info info); 5855 5856 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, 5857 unsigned int /*kern_return_t*/ *)) 5858 dlsym(RTLD_DEFAULT, "_dyld_process_info_create"); 5859 dyld_process_info_get_cache = (void (*)(void *, void *))dlsym( 5860 RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5861 dyld_process_info_release = 5862 (void (*)(void *))dlsym(RTLD_DEFAULT, "_dyld_process_info_release"); 5863 5864 if (dyld_process_info_create && dyld_process_info_get_cache) { 5865 unsigned int /*kern_return_t */ kern_ret; 5866 dyld_process_info process_info = 5867 dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5868 if (process_info) { 5869 struct lldb_copy__dyld_process_cache_info sc_info; 5870 memset(&sc_info, 0, sizeof(struct lldb_copy__dyld_process_cache_info)); 5871 dyld_process_info_get_cache(process_info, &sc_info); 5872 if (sc_info.cacheBaseAddress != 0) { 5873 base_addr = sc_info.cacheBaseAddress; 5874 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 5875 } 5876 dyld_process_info_release(process_info); 5877 } 5878 } 5879 } 5880 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5881 LIBLLDB_LOG_PROCESS)); 5882 if (log && uuid.IsValid()) 5883 LLDB_LOGF(log, 5884 "lldb's in-memory shared cache has a UUID of %s base address of " 5885 "0x%" PRIx64, 5886 uuid.GetAsString().c_str(), base_addr); 5887 #endif 5888 } 5889 5890 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5891 if (!m_min_os_version) { 5892 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5893 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5894 const lldb::offset_t load_cmd_offset = offset; 5895 5896 version_min_command lc; 5897 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5898 break; 5899 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5900 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5901 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5902 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5903 if (m_data.GetU32(&offset, &lc.version, 5904 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5905 const uint32_t xxxx = lc.version >> 16; 5906 const uint32_t yy = (lc.version >> 8) & 0xffu; 5907 const uint32_t zz = lc.version & 0xffu; 5908 if (xxxx) { 5909 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5910 break; 5911 } 5912 } 5913 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5914 // struct build_version_command { 5915 // uint32_t cmd; /* LC_BUILD_VERSION */ 5916 // uint32_t cmdsize; /* sizeof(struct 5917 // build_version_command) plus */ 5918 // /* ntools * sizeof(struct 5919 // build_tool_version) */ 5920 // uint32_t platform; /* platform */ 5921 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5922 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded in 5923 // nibbles xxxx.yy.zz */ uint32_t ntools; /* number of 5924 // tool entries following this */ 5925 // }; 5926 5927 offset += 4; // skip platform 5928 uint32_t minos = m_data.GetU32(&offset); 5929 5930 const uint32_t xxxx = minos >> 16; 5931 const uint32_t yy = (minos >> 8) & 0xffu; 5932 const uint32_t zz = minos & 0xffu; 5933 if (xxxx) { 5934 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5935 break; 5936 } 5937 } 5938 5939 offset = load_cmd_offset + lc.cmdsize; 5940 } 5941 5942 if (!m_min_os_version) { 5943 // Set version to an empty value so we don't keep trying to 5944 m_min_os_version = llvm::VersionTuple(); 5945 } 5946 } 5947 5948 return *m_min_os_version; 5949 } 5950 5951 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { 5952 if (!m_sdk_versions.hasValue()) { 5953 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5954 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5955 const lldb::offset_t load_cmd_offset = offset; 5956 5957 version_min_command lc; 5958 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5959 break; 5960 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5961 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5962 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5963 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5964 if (m_data.GetU32(&offset, &lc.version, 5965 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5966 const uint32_t xxxx = lc.sdk >> 16; 5967 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5968 const uint32_t zz = lc.sdk & 0xffu; 5969 if (xxxx) { 5970 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 5971 break; 5972 } else { 5973 GetModule()->ReportWarning("minimum OS version load command with " 5974 "invalid (0) version found."); 5975 } 5976 } 5977 } 5978 offset = load_cmd_offset + lc.cmdsize; 5979 } 5980 5981 if (!m_sdk_versions.hasValue()) { 5982 offset = MachHeaderSizeFromMagic(m_header.magic); 5983 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5984 const lldb::offset_t load_cmd_offset = offset; 5985 5986 version_min_command lc; 5987 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5988 break; 5989 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5990 // struct build_version_command { 5991 // uint32_t cmd; /* LC_BUILD_VERSION */ 5992 // uint32_t cmdsize; /* sizeof(struct 5993 // build_version_command) plus */ 5994 // /* ntools * sizeof(struct 5995 // build_tool_version) */ 5996 // uint32_t platform; /* platform */ 5997 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5998 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded 5999 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number 6000 // of tool entries following this */ 6001 // }; 6002 6003 offset += 4; // skip platform 6004 uint32_t minos = m_data.GetU32(&offset); 6005 6006 const uint32_t xxxx = minos >> 16; 6007 const uint32_t yy = (minos >> 8) & 0xffu; 6008 const uint32_t zz = minos & 0xffu; 6009 if (xxxx) { 6010 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 6011 break; 6012 } 6013 } 6014 offset = load_cmd_offset + lc.cmdsize; 6015 } 6016 } 6017 6018 if (!m_sdk_versions.hasValue()) 6019 m_sdk_versions = llvm::VersionTuple(); 6020 } 6021 6022 return m_sdk_versions.getValue(); 6023 } 6024 6025 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 6026 return m_header.filetype == llvm::MachO::MH_DYLINKER; 6027 } 6028 6029 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 6030 return m_allow_assembly_emulation_unwind_plans; 6031 } 6032 6033 // PluginInterface protocol 6034 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 6035 return GetPluginNameStatic(); 6036 } 6037 6038 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 6039 6040 Section *ObjectFileMachO::GetMachHeaderSection() { 6041 // Find the first address of the mach header which is the first non-zero file 6042 // sized section whose file offset is zero. This is the base file address of 6043 // the mach-o file which can be subtracted from the vmaddr of the other 6044 // segments found in memory and added to the load address 6045 ModuleSP module_sp = GetModule(); 6046 if (!module_sp) 6047 return nullptr; 6048 SectionList *section_list = GetSectionList(); 6049 if (!section_list) 6050 return nullptr; 6051 const size_t num_sections = section_list->GetSize(); 6052 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6053 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 6054 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 6055 return section; 6056 } 6057 return nullptr; 6058 } 6059 6060 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 6061 if (!section) 6062 return false; 6063 const bool is_dsym = (m_header.filetype == MH_DSYM); 6064 if (section->GetFileSize() == 0 && !is_dsym) 6065 return false; 6066 if (section->IsThreadSpecific()) 6067 return false; 6068 if (GetModule().get() != section->GetModule().get()) 6069 return false; 6070 // Be careful with __LINKEDIT and __DWARF segments 6071 if (section->GetName() == GetSegmentNameLINKEDIT() || 6072 section->GetName() == GetSegmentNameDWARF()) { 6073 // Only map __LINKEDIT and __DWARF if we have an in memory image and 6074 // this isn't a kernel binary like a kext or mach_kernel. 6075 const bool is_memory_image = (bool)m_process_wp.lock(); 6076 const Strata strata = GetStrata(); 6077 if (is_memory_image == false || strata == eStrataKernel) 6078 return false; 6079 } 6080 return true; 6081 } 6082 6083 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 6084 lldb::addr_t header_load_address, const Section *header_section, 6085 const Section *section) { 6086 ModuleSP module_sp = GetModule(); 6087 if (module_sp && header_section && section && 6088 header_load_address != LLDB_INVALID_ADDRESS) { 6089 lldb::addr_t file_addr = header_section->GetFileAddress(); 6090 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 6091 return section->GetFileAddress() - file_addr + header_load_address; 6092 } 6093 return LLDB_INVALID_ADDRESS; 6094 } 6095 6096 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 6097 bool value_is_offset) { 6098 ModuleSP module_sp = GetModule(); 6099 if (!module_sp) 6100 return false; 6101 6102 SectionList *section_list = GetSectionList(); 6103 if (!section_list) 6104 return false; 6105 6106 size_t num_loaded_sections = 0; 6107 const size_t num_sections = section_list->GetSize(); 6108 6109 if (value_is_offset) { 6110 // "value" is an offset to apply to each top level segment 6111 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6112 // Iterate through the object file sections to find all of the 6113 // sections that size on disk (to avoid __PAGEZERO) and load them 6114 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6115 if (SectionIsLoadable(section_sp.get())) 6116 if (target.GetSectionLoadList().SetSectionLoadAddress( 6117 section_sp, section_sp->GetFileAddress() + value)) 6118 ++num_loaded_sections; 6119 } 6120 } else { 6121 // "value" is the new base address of the mach_header, adjust each 6122 // section accordingly 6123 6124 Section *mach_header_section = GetMachHeaderSection(); 6125 if (mach_header_section) { 6126 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6127 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6128 6129 lldb::addr_t section_load_addr = 6130 CalculateSectionLoadAddressForMemoryImage( 6131 value, mach_header_section, section_sp.get()); 6132 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6133 if (target.GetSectionLoadList().SetSectionLoadAddress( 6134 section_sp, section_load_addr)) 6135 ++num_loaded_sections; 6136 } 6137 } 6138 } 6139 } 6140 return num_loaded_sections > 0; 6141 } 6142 6143 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6144 const FileSpec &outfile, Status &error) { 6145 if (!process_sp) 6146 return false; 6147 6148 Target &target = process_sp->GetTarget(); 6149 const ArchSpec target_arch = target.GetArchitecture(); 6150 const llvm::Triple &target_triple = target_arch.GetTriple(); 6151 if (target_triple.getVendor() == llvm::Triple::Apple && 6152 (target_triple.getOS() == llvm::Triple::MacOSX || 6153 target_triple.getOS() == llvm::Triple::IOS || 6154 target_triple.getOS() == llvm::Triple::WatchOS || 6155 target_triple.getOS() == llvm::Triple::TvOS)) { 6156 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) 6157 // { 6158 bool make_core = false; 6159 switch (target_arch.GetMachine()) { 6160 case llvm::Triple::aarch64: 6161 case llvm::Triple::aarch64_32: 6162 case llvm::Triple::arm: 6163 case llvm::Triple::thumb: 6164 case llvm::Triple::x86: 6165 case llvm::Triple::x86_64: 6166 make_core = true; 6167 break; 6168 default: 6169 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6170 target_triple.str().c_str()); 6171 break; 6172 } 6173 6174 if (make_core) { 6175 std::vector<segment_command_64> segment_load_commands; 6176 // uint32_t range_info_idx = 0; 6177 MemoryRegionInfo range_info; 6178 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6179 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6180 const ByteOrder byte_order = target_arch.GetByteOrder(); 6181 if (range_error.Success()) { 6182 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6183 const addr_t addr = range_info.GetRange().GetRangeBase(); 6184 const addr_t size = range_info.GetRange().GetByteSize(); 6185 6186 if (size == 0) 6187 break; 6188 6189 // Calculate correct protections 6190 uint32_t prot = 0; 6191 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6192 prot |= VM_PROT_READ; 6193 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6194 prot |= VM_PROT_WRITE; 6195 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6196 prot |= VM_PROT_EXECUTE; 6197 6198 if (prot != 0) { 6199 uint32_t cmd_type = LC_SEGMENT_64; 6200 uint32_t segment_size = sizeof(segment_command_64); 6201 if (addr_byte_size == 4) { 6202 cmd_type = LC_SEGMENT; 6203 segment_size = sizeof(segment_command); 6204 } 6205 segment_command_64 segment = { 6206 cmd_type, // uint32_t cmd; 6207 segment_size, // uint32_t cmdsize; 6208 {0}, // char segname[16]; 6209 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6210 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6211 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6212 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6213 prot, // uint32_t maxprot; 6214 prot, // uint32_t initprot; 6215 0, // uint32_t nsects; 6216 0}; // uint32_t flags; 6217 segment_load_commands.push_back(segment); 6218 } else { 6219 // No protections and a size of 1 used to be returned from old 6220 // debugservers when we asked about a region that was past the 6221 // last memory region and it indicates the end... 6222 if (size == 1) 6223 break; 6224 } 6225 6226 range_error = process_sp->GetMemoryRegionInfo( 6227 range_info.GetRange().GetRangeEnd(), range_info); 6228 if (range_error.Fail()) 6229 break; 6230 } 6231 6232 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6233 6234 mach_header_64 mach_header; 6235 if (addr_byte_size == 8) { 6236 mach_header.magic = MH_MAGIC_64; 6237 } else { 6238 mach_header.magic = MH_MAGIC; 6239 } 6240 mach_header.cputype = target_arch.GetMachOCPUType(); 6241 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6242 mach_header.filetype = MH_CORE; 6243 mach_header.ncmds = segment_load_commands.size(); 6244 mach_header.flags = 0; 6245 mach_header.reserved = 0; 6246 ThreadList &thread_list = process_sp->GetThreadList(); 6247 const uint32_t num_threads = thread_list.GetSize(); 6248 6249 // Make an array of LC_THREAD data items. Each one contains the 6250 // contents of the LC_THREAD load command. The data doesn't contain 6251 // the load command + load command size, we will add the load command 6252 // and load command size as we emit the data. 6253 std::vector<StreamString> LC_THREAD_datas(num_threads); 6254 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6255 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6256 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6257 LC_THREAD_data.SetByteOrder(byte_order); 6258 } 6259 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) { 6260 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6261 if (thread_sp) { 6262 switch (mach_header.cputype) { 6263 case llvm::MachO::CPU_TYPE_ARM64: 6264 case llvm::MachO::CPU_TYPE_ARM64_32: 6265 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6266 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6267 break; 6268 6269 case llvm::MachO::CPU_TYPE_ARM: 6270 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6271 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6272 break; 6273 6274 case llvm::MachO::CPU_TYPE_I386: 6275 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6276 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6277 break; 6278 6279 case llvm::MachO::CPU_TYPE_X86_64: 6280 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6281 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6282 break; 6283 } 6284 } 6285 } 6286 6287 // The size of the load command is the size of the segments... 6288 if (addr_byte_size == 8) { 6289 mach_header.sizeofcmds = 6290 segment_load_commands.size() * sizeof(struct segment_command_64); 6291 } else { 6292 mach_header.sizeofcmds = 6293 segment_load_commands.size() * sizeof(struct segment_command); 6294 } 6295 6296 // and the size of all LC_THREAD load command 6297 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6298 ++mach_header.ncmds; 6299 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6300 } 6301 6302 // Write the mach header 6303 buffer.PutHex32(mach_header.magic); 6304 buffer.PutHex32(mach_header.cputype); 6305 buffer.PutHex32(mach_header.cpusubtype); 6306 buffer.PutHex32(mach_header.filetype); 6307 buffer.PutHex32(mach_header.ncmds); 6308 buffer.PutHex32(mach_header.sizeofcmds); 6309 buffer.PutHex32(mach_header.flags); 6310 if (addr_byte_size == 8) { 6311 buffer.PutHex32(mach_header.reserved); 6312 } 6313 6314 // Skip the mach header and all load commands and align to the next 6315 // 0x1000 byte boundary 6316 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6317 if (file_offset & 0x00000fff) { 6318 file_offset += 0x00001000ull; 6319 file_offset &= (~0x00001000ull + 1); 6320 } 6321 6322 for (auto &segment : segment_load_commands) { 6323 segment.fileoff = file_offset; 6324 file_offset += segment.filesize; 6325 } 6326 6327 // Write out all of the LC_THREAD load commands 6328 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6329 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6330 buffer.PutHex32(LC_THREAD); 6331 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6332 buffer.Write(LC_THREAD_data.GetString().data(), LC_THREAD_data_size); 6333 } 6334 6335 // Write out all of the segment load commands 6336 for (const auto &segment : segment_load_commands) { 6337 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6338 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6339 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6340 segment.cmd, segment.cmdsize, segment.vmaddr, 6341 segment.vmaddr + segment.vmsize, segment.fileoff, 6342 segment.filesize, segment.maxprot, segment.initprot, 6343 segment.nsects, segment.flags); 6344 6345 buffer.PutHex32(segment.cmd); 6346 buffer.PutHex32(segment.cmdsize); 6347 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6348 if (addr_byte_size == 8) { 6349 buffer.PutHex64(segment.vmaddr); 6350 buffer.PutHex64(segment.vmsize); 6351 buffer.PutHex64(segment.fileoff); 6352 buffer.PutHex64(segment.filesize); 6353 } else { 6354 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6355 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6356 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6357 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6358 } 6359 buffer.PutHex32(segment.maxprot); 6360 buffer.PutHex32(segment.initprot); 6361 buffer.PutHex32(segment.nsects); 6362 buffer.PutHex32(segment.flags); 6363 } 6364 6365 std::string core_file_path(outfile.GetPath()); 6366 auto core_file = FileSystem::Instance().Open( 6367 outfile, File::eOpenOptionWrite | File::eOpenOptionTruncate | 6368 File::eOpenOptionCanCreate); 6369 if (!core_file) { 6370 error = core_file.takeError(); 6371 } else { 6372 // Read 1 page at a time 6373 uint8_t bytes[0x1000]; 6374 // Write the mach header and load commands out to the core file 6375 size_t bytes_written = buffer.GetString().size(); 6376 error = 6377 core_file.get()->Write(buffer.GetString().data(), bytes_written); 6378 if (error.Success()) { 6379 // Now write the file data for all memory segments in the process 6380 for (const auto &segment : segment_load_commands) { 6381 if (core_file.get()->SeekFromStart(segment.fileoff) == -1) { 6382 error.SetErrorStringWithFormat( 6383 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6384 segment.fileoff, core_file_path.c_str()); 6385 break; 6386 } 6387 6388 printf("Saving %" PRId64 6389 " bytes of data for memory region at 0x%" PRIx64 "\n", 6390 segment.vmsize, segment.vmaddr); 6391 addr_t bytes_left = segment.vmsize; 6392 addr_t addr = segment.vmaddr; 6393 Status memory_read_error; 6394 while (bytes_left > 0 && error.Success()) { 6395 const size_t bytes_to_read = 6396 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6397 6398 // In a savecore setting, we don't really care about caching, 6399 // as the data is dumped and very likely never read again, 6400 // so we call ReadMemoryFromInferior to bypass it. 6401 const size_t bytes_read = process_sp->ReadMemoryFromInferior( 6402 addr, bytes, bytes_to_read, memory_read_error); 6403 6404 if (bytes_read == bytes_to_read) { 6405 size_t bytes_written = bytes_read; 6406 error = core_file.get()->Write(bytes, bytes_written); 6407 bytes_left -= bytes_read; 6408 addr += bytes_read; 6409 } else { 6410 // Some pages within regions are not readable, those should 6411 // be zero filled 6412 memset(bytes, 0, bytes_to_read); 6413 size_t bytes_written = bytes_to_read; 6414 error = core_file.get()->Write(bytes, bytes_written); 6415 bytes_left -= bytes_to_read; 6416 addr += bytes_to_read; 6417 } 6418 } 6419 } 6420 } 6421 } 6422 } else { 6423 error.SetErrorString( 6424 "process doesn't support getting memory region info"); 6425 } 6426 } 6427 return true; // This is the right plug to handle saving core files for 6428 // this process 6429 } 6430 return false; 6431 } 6432