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 #include <TargetConditionals.h> 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 = std::make_unique<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 ((m_header.flags & MH_DYLIB_IN_CACHE) && !IsInMemory()) { 1332 // In shared cache images, the load commands are relative to the 1333 // shared cache file, and not the the specific image we are 1334 // examining. Let's fix this up so that it looks like a normal 1335 // image. 1336 if (strncmp(seg_cmd.segname, "__TEXT", sizeof(seg_cmd.segname)) == 0) 1337 m_text_address = seg_cmd.vmaddr; 1338 if (strncmp(seg_cmd.segname, "__LINKEDIT", sizeof(seg_cmd.segname)) == 0) 1339 m_linkedit_original_offset = seg_cmd.fileoff; 1340 1341 seg_cmd.fileoff = seg_cmd.vmaddr - m_text_address; 1342 } 1343 1344 if (seg_cmd.fileoff > m_length) { 1345 // We have a load command that says it extends past the end of the file. 1346 // This is likely a corrupt file. We don't have any way to return an error 1347 // condition here (this method was likely invoked from something like 1348 // ObjectFile::GetSectionList()), so we just null out the section contents, 1349 // and dump a message to stdout. The most common case here is core file 1350 // debugging with a truncated file. 1351 const char *lc_segment_name = 1352 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1353 GetModule()->ReportWarning( 1354 "load command %u %s has a fileoff (0x%" PRIx64 1355 ") that extends beyond the end of the file (0x%" PRIx64 1356 "), ignoring this section", 1357 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1358 1359 seg_cmd.fileoff = 0; 1360 seg_cmd.filesize = 0; 1361 } 1362 1363 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1364 // We have a load command that says it extends past the end of the file. 1365 // This is likely a corrupt file. We don't have any way to return an error 1366 // condition here (this method was likely invoked from something like 1367 // ObjectFile::GetSectionList()), so we just null out the section contents, 1368 // and dump a message to stdout. The most common case here is core file 1369 // debugging with a truncated file. 1370 const char *lc_segment_name = 1371 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1372 GetModule()->ReportWarning( 1373 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1374 ") that extends beyond the end of the file (0x%" PRIx64 1375 "), the segment will be truncated to match", 1376 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1377 1378 // Truncate the length 1379 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1380 } 1381 } 1382 1383 static uint32_t GetSegmentPermissions(const segment_command_64 &seg_cmd) { 1384 uint32_t result = 0; 1385 if (seg_cmd.initprot & VM_PROT_READ) 1386 result |= ePermissionsReadable; 1387 if (seg_cmd.initprot & VM_PROT_WRITE) 1388 result |= ePermissionsWritable; 1389 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1390 result |= ePermissionsExecutable; 1391 return result; 1392 } 1393 1394 static lldb::SectionType GetSectionType(uint32_t flags, 1395 ConstString section_name) { 1396 1397 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1398 return eSectionTypeCode; 1399 1400 uint32_t mach_sect_type = flags & SECTION_TYPE; 1401 static ConstString g_sect_name_objc_data("__objc_data"); 1402 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1403 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1404 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1405 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1406 static ConstString g_sect_name_objc_const("__objc_const"); 1407 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1408 static ConstString g_sect_name_cfstring("__cfstring"); 1409 1410 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1411 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1412 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1413 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1414 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1415 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1416 static ConstString g_sect_name_dwarf_debug_loclists("__debug_loclists"); 1417 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1418 static ConstString g_sect_name_dwarf_debug_names("__debug_names"); 1419 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1420 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1421 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1422 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1423 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1424 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1425 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1426 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1427 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1428 static ConstString g_sect_name_eh_frame("__eh_frame"); 1429 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1430 static ConstString g_sect_name_text("__text"); 1431 static ConstString g_sect_name_data("__data"); 1432 static ConstString g_sect_name_go_symtab("__gosymtab"); 1433 1434 if (section_name == g_sect_name_dwarf_debug_abbrev) 1435 return eSectionTypeDWARFDebugAbbrev; 1436 if (section_name == g_sect_name_dwarf_debug_aranges) 1437 return eSectionTypeDWARFDebugAranges; 1438 if (section_name == g_sect_name_dwarf_debug_frame) 1439 return eSectionTypeDWARFDebugFrame; 1440 if (section_name == g_sect_name_dwarf_debug_info) 1441 return eSectionTypeDWARFDebugInfo; 1442 if (section_name == g_sect_name_dwarf_debug_line) 1443 return eSectionTypeDWARFDebugLine; 1444 if (section_name == g_sect_name_dwarf_debug_loc) 1445 return eSectionTypeDWARFDebugLoc; 1446 if (section_name == g_sect_name_dwarf_debug_loclists) 1447 return eSectionTypeDWARFDebugLocLists; 1448 if (section_name == g_sect_name_dwarf_debug_macinfo) 1449 return eSectionTypeDWARFDebugMacInfo; 1450 if (section_name == g_sect_name_dwarf_debug_names) 1451 return eSectionTypeDWARFDebugNames; 1452 if (section_name == g_sect_name_dwarf_debug_pubnames) 1453 return eSectionTypeDWARFDebugPubNames; 1454 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1455 return eSectionTypeDWARFDebugPubTypes; 1456 if (section_name == g_sect_name_dwarf_debug_ranges) 1457 return eSectionTypeDWARFDebugRanges; 1458 if (section_name == g_sect_name_dwarf_debug_str) 1459 return eSectionTypeDWARFDebugStr; 1460 if (section_name == g_sect_name_dwarf_debug_types) 1461 return eSectionTypeDWARFDebugTypes; 1462 if (section_name == g_sect_name_dwarf_apple_names) 1463 return eSectionTypeDWARFAppleNames; 1464 if (section_name == g_sect_name_dwarf_apple_types) 1465 return eSectionTypeDWARFAppleTypes; 1466 if (section_name == g_sect_name_dwarf_apple_namespaces) 1467 return eSectionTypeDWARFAppleNamespaces; 1468 if (section_name == g_sect_name_dwarf_apple_objc) 1469 return eSectionTypeDWARFAppleObjC; 1470 if (section_name == g_sect_name_objc_selrefs) 1471 return eSectionTypeDataCStringPointers; 1472 if (section_name == g_sect_name_objc_msgrefs) 1473 return eSectionTypeDataObjCMessageRefs; 1474 if (section_name == g_sect_name_eh_frame) 1475 return eSectionTypeEHFrame; 1476 if (section_name == g_sect_name_compact_unwind) 1477 return eSectionTypeCompactUnwind; 1478 if (section_name == g_sect_name_cfstring) 1479 return eSectionTypeDataObjCCFStrings; 1480 if (section_name == g_sect_name_go_symtab) 1481 return eSectionTypeGoSymtab; 1482 if (section_name == g_sect_name_objc_data || 1483 section_name == g_sect_name_objc_classrefs || 1484 section_name == g_sect_name_objc_superrefs || 1485 section_name == g_sect_name_objc_const || 1486 section_name == g_sect_name_objc_classlist) { 1487 return eSectionTypeDataPointers; 1488 } 1489 1490 switch (mach_sect_type) { 1491 // TODO: categorize sections by other flags for regular sections 1492 case S_REGULAR: 1493 if (section_name == g_sect_name_text) 1494 return eSectionTypeCode; 1495 if (section_name == g_sect_name_data) 1496 return eSectionTypeData; 1497 return eSectionTypeOther; 1498 case S_ZEROFILL: 1499 return eSectionTypeZeroFill; 1500 case S_CSTRING_LITERALS: // section with only literal C strings 1501 return eSectionTypeDataCString; 1502 case S_4BYTE_LITERALS: // section with only 4 byte literals 1503 return eSectionTypeData4; 1504 case S_8BYTE_LITERALS: // section with only 8 byte literals 1505 return eSectionTypeData8; 1506 case S_LITERAL_POINTERS: // section with only pointers to literals 1507 return eSectionTypeDataPointers; 1508 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1509 return eSectionTypeDataPointers; 1510 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1511 return eSectionTypeDataPointers; 1512 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1513 // the reserved2 field 1514 return eSectionTypeCode; 1515 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1516 // initialization 1517 return eSectionTypeDataPointers; 1518 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1519 // termination 1520 return eSectionTypeDataPointers; 1521 case S_COALESCED: 1522 return eSectionTypeOther; 1523 case S_GB_ZEROFILL: 1524 return eSectionTypeZeroFill; 1525 case S_INTERPOSING: // section with only pairs of function pointers for 1526 // interposing 1527 return eSectionTypeCode; 1528 case S_16BYTE_LITERALS: // section with only 16 byte literals 1529 return eSectionTypeData16; 1530 case S_DTRACE_DOF: 1531 return eSectionTypeDebug; 1532 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1533 return eSectionTypeDataPointers; 1534 default: 1535 return eSectionTypeOther; 1536 } 1537 } 1538 1539 struct ObjectFileMachO::SegmentParsingContext { 1540 const EncryptedFileRanges EncryptedRanges; 1541 lldb_private::SectionList &UnifiedList; 1542 uint32_t NextSegmentIdx = 0; 1543 uint32_t NextSectionIdx = 0; 1544 bool FileAddressesChanged = false; 1545 1546 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1547 lldb_private::SectionList &UnifiedList) 1548 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1549 }; 1550 1551 void ObjectFileMachO::ProcessSegmentCommand(const load_command &load_cmd_, 1552 lldb::offset_t offset, 1553 uint32_t cmd_idx, 1554 SegmentParsingContext &context) { 1555 segment_command_64 load_cmd; 1556 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1557 1558 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1559 return; 1560 1561 ModuleSP module_sp = GetModule(); 1562 const bool is_core = GetType() == eTypeCoreFile; 1563 const bool is_dsym = (m_header.filetype == MH_DSYM); 1564 bool add_section = true; 1565 bool add_to_unified = true; 1566 ConstString const_segname( 1567 load_cmd.segname, strnlen(load_cmd.segname, sizeof(load_cmd.segname))); 1568 1569 SectionSP unified_section_sp( 1570 context.UnifiedList.FindSectionByName(const_segname)); 1571 if (is_dsym && unified_section_sp) { 1572 if (const_segname == GetSegmentNameLINKEDIT()) { 1573 // We need to keep the __LINKEDIT segment private to this object file 1574 // only 1575 add_to_unified = false; 1576 } else { 1577 // This is the dSYM file and this section has already been created by the 1578 // object file, no need to create it. 1579 add_section = false; 1580 } 1581 } 1582 load_cmd.vmaddr = m_data.GetAddress(&offset); 1583 load_cmd.vmsize = m_data.GetAddress(&offset); 1584 load_cmd.fileoff = m_data.GetAddress(&offset); 1585 load_cmd.filesize = m_data.GetAddress(&offset); 1586 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1587 return; 1588 1589 SanitizeSegmentCommand(load_cmd, cmd_idx); 1590 1591 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1592 const bool segment_is_encrypted = 1593 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1594 1595 // Keep a list of mach segments around in case we need to get at data that 1596 // isn't stored in the abstracted Sections. 1597 m_mach_segments.push_back(load_cmd); 1598 1599 // Use a segment ID of the segment index shifted left by 8 so they never 1600 // conflict with any of the sections. 1601 SectionSP segment_sp; 1602 if (add_section && (const_segname || is_core)) { 1603 segment_sp = std::make_shared<Section>( 1604 module_sp, // Module to which this section belongs 1605 this, // Object file to which this sections belongs 1606 ++context.NextSegmentIdx 1607 << 8, // Section ID is the 1 based segment index 1608 // shifted right by 8 bits as not to collide with any of the 256 1609 // section IDs that are possible 1610 const_segname, // Name of this section 1611 eSectionTypeContainer, // This section is a container of other 1612 // sections. 1613 load_cmd.vmaddr, // File VM address == addresses as they are 1614 // found in the object file 1615 load_cmd.vmsize, // VM size in bytes of this section 1616 load_cmd.fileoff, // Offset to the data for this section in 1617 // the file 1618 load_cmd.filesize, // Size in bytes of this section as found 1619 // in the file 1620 0, // Segments have no alignment information 1621 load_cmd.flags); // Flags for this section 1622 1623 segment_sp->SetIsEncrypted(segment_is_encrypted); 1624 m_sections_up->AddSection(segment_sp); 1625 segment_sp->SetPermissions(segment_permissions); 1626 if (add_to_unified) 1627 context.UnifiedList.AddSection(segment_sp); 1628 } else if (unified_section_sp) { 1629 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1630 // Check to see if the module was read from memory? 1631 if (module_sp->GetObjectFile()->GetBaseAddress().IsValid()) { 1632 // We have a module that is in memory and needs to have its file 1633 // address adjusted. We need to do this because when we load a file 1634 // from memory, its addresses will be slid already, yet the addresses 1635 // in the new symbol file will still be unslid. Since everything is 1636 // stored as section offset, this shouldn't cause any problems. 1637 1638 // Make sure we've parsed the symbol table from the ObjectFile before 1639 // we go around changing its Sections. 1640 module_sp->GetObjectFile()->GetSymtab(); 1641 // eh_frame would present the same problems but we parse that on a per- 1642 // function basis as-needed so it's more difficult to remove its use of 1643 // the Sections. Realistically, the environments where this code path 1644 // will be taken will not have eh_frame sections. 1645 1646 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1647 1648 // Notify the module that the section addresses have been changed once 1649 // we're done so any file-address caches can be updated. 1650 context.FileAddressesChanged = true; 1651 } 1652 } 1653 m_sections_up->AddSection(unified_section_sp); 1654 } 1655 1656 struct section_64 sect64; 1657 ::memset(§64, 0, sizeof(sect64)); 1658 // Push a section into our mach sections for the section at index zero 1659 // (NO_SECT) if we don't have any mach sections yet... 1660 if (m_mach_sections.empty()) 1661 m_mach_sections.push_back(sect64); 1662 uint32_t segment_sect_idx; 1663 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1664 1665 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1666 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1667 ++segment_sect_idx) { 1668 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1669 sizeof(sect64.sectname)) == nullptr) 1670 break; 1671 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1672 sizeof(sect64.segname)) == nullptr) 1673 break; 1674 sect64.addr = m_data.GetAddress(&offset); 1675 sect64.size = m_data.GetAddress(&offset); 1676 1677 if (m_data.GetU32(&offset, §64.offset, num_u32s) == nullptr) 1678 break; 1679 1680 if ((m_header.flags & MH_DYLIB_IN_CACHE) && !IsInMemory()) { 1681 sect64.offset = sect64.addr - m_text_address; 1682 } 1683 1684 // Keep a list of mach sections around in case we need to get at data that 1685 // isn't stored in the abstracted Sections. 1686 m_mach_sections.push_back(sect64); 1687 1688 if (add_section) { 1689 ConstString section_name( 1690 sect64.sectname, strnlen(sect64.sectname, sizeof(sect64.sectname))); 1691 if (!const_segname) { 1692 // We have a segment with no name so we need to conjure up segments 1693 // that correspond to the section's segname if there isn't already such 1694 // a section. If there is such a section, we resize the section so that 1695 // it spans all sections. We also mark these sections as fake so 1696 // address matches don't hit if they land in the gaps between the child 1697 // sections. 1698 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1699 sizeof(sect64.segname)); 1700 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1701 if (segment_sp.get()) { 1702 Section *segment = segment_sp.get(); 1703 // Grow the section size as needed. 1704 const lldb::addr_t sect64_min_addr = sect64.addr; 1705 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1706 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1707 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1708 const lldb::addr_t curr_seg_max_addr = 1709 curr_seg_min_addr + curr_seg_byte_size; 1710 if (sect64_min_addr >= curr_seg_min_addr) { 1711 const lldb::addr_t new_seg_byte_size = 1712 sect64_max_addr - curr_seg_min_addr; 1713 // Only grow the section size if needed 1714 if (new_seg_byte_size > curr_seg_byte_size) 1715 segment->SetByteSize(new_seg_byte_size); 1716 } else { 1717 // We need to change the base address of the segment and adjust the 1718 // child section offsets for all existing children. 1719 const lldb::addr_t slide_amount = 1720 sect64_min_addr - curr_seg_min_addr; 1721 segment->Slide(slide_amount, false); 1722 segment->GetChildren().Slide(-slide_amount, false); 1723 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1724 } 1725 1726 // Grow the section size as needed. 1727 if (sect64.offset) { 1728 const lldb::addr_t segment_min_file_offset = 1729 segment->GetFileOffset(); 1730 const lldb::addr_t segment_max_file_offset = 1731 segment_min_file_offset + segment->GetFileSize(); 1732 1733 const lldb::addr_t section_min_file_offset = sect64.offset; 1734 const lldb::addr_t section_max_file_offset = 1735 section_min_file_offset + sect64.size; 1736 const lldb::addr_t new_file_offset = 1737 std::min(section_min_file_offset, segment_min_file_offset); 1738 const lldb::addr_t new_file_size = 1739 std::max(section_max_file_offset, segment_max_file_offset) - 1740 new_file_offset; 1741 segment->SetFileOffset(new_file_offset); 1742 segment->SetFileSize(new_file_size); 1743 } 1744 } else { 1745 // Create a fake section for the section's named segment 1746 segment_sp = std::make_shared<Section>( 1747 segment_sp, // Parent section 1748 module_sp, // Module to which this section belongs 1749 this, // Object file to which this section belongs 1750 ++context.NextSegmentIdx 1751 << 8, // Section ID is the 1 based segment index 1752 // shifted right by 8 bits as not to 1753 // collide with any of the 256 section IDs 1754 // that are possible 1755 const_segname, // Name of this section 1756 eSectionTypeContainer, // This section is a container of 1757 // other sections. 1758 sect64.addr, // File VM address == addresses as they are 1759 // found in the object file 1760 sect64.size, // VM size in bytes of this section 1761 sect64.offset, // Offset to the data for this section in 1762 // the file 1763 sect64.offset ? sect64.size : 0, // Size in bytes of 1764 // this section as 1765 // found in the file 1766 sect64.align, 1767 load_cmd.flags); // Flags for this section 1768 segment_sp->SetIsFake(true); 1769 segment_sp->SetPermissions(segment_permissions); 1770 m_sections_up->AddSection(segment_sp); 1771 if (add_to_unified) 1772 context.UnifiedList.AddSection(segment_sp); 1773 segment_sp->SetIsEncrypted(segment_is_encrypted); 1774 } 1775 } 1776 assert(segment_sp.get()); 1777 1778 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1779 1780 SectionSP section_sp(new Section( 1781 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1782 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1783 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1784 sect64.flags)); 1785 // Set the section to be encrypted to match the segment 1786 1787 bool section_is_encrypted = false; 1788 if (!segment_is_encrypted && load_cmd.filesize != 0) 1789 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1790 sect64.offset) != nullptr; 1791 1792 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1793 section_sp->SetPermissions(segment_permissions); 1794 segment_sp->GetChildren().AddSection(section_sp); 1795 1796 if (segment_sp->IsFake()) { 1797 segment_sp.reset(); 1798 const_segname.Clear(); 1799 } 1800 } 1801 } 1802 if (segment_sp && is_dsym) { 1803 if (first_segment_sectID <= context.NextSectionIdx) { 1804 lldb::user_id_t sect_uid; 1805 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1806 ++sect_uid) { 1807 SectionSP curr_section_sp( 1808 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1809 SectionSP next_section_sp; 1810 if (sect_uid + 1 <= context.NextSectionIdx) 1811 next_section_sp = 1812 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1813 1814 if (curr_section_sp.get()) { 1815 if (curr_section_sp->GetByteSize() == 0) { 1816 if (next_section_sp.get() != nullptr) 1817 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1818 curr_section_sp->GetFileAddress()); 1819 else 1820 curr_section_sp->SetByteSize(load_cmd.vmsize); 1821 } 1822 } 1823 } 1824 } 1825 } 1826 } 1827 1828 void ObjectFileMachO::ProcessDysymtabCommand(const load_command &load_cmd, 1829 lldb::offset_t offset) { 1830 m_dysymtab.cmd = load_cmd.cmd; 1831 m_dysymtab.cmdsize = load_cmd.cmdsize; 1832 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1833 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1834 } 1835 1836 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1837 if (m_sections_up) 1838 return; 1839 1840 m_sections_up = std::make_unique<SectionList>(); 1841 1842 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1843 // bool dump_sections = false; 1844 ModuleSP module_sp(GetModule()); 1845 1846 offset = MachHeaderSizeFromMagic(m_header.magic); 1847 1848 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1849 struct load_command load_cmd; 1850 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1851 const lldb::offset_t load_cmd_offset = offset; 1852 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 1853 break; 1854 1855 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1856 ProcessSegmentCommand(load_cmd, offset, i, context); 1857 else if (load_cmd.cmd == LC_DYSYMTAB) 1858 ProcessDysymtabCommand(load_cmd, offset); 1859 1860 offset = load_cmd_offset + load_cmd.cmdsize; 1861 } 1862 1863 if (context.FileAddressesChanged && module_sp) 1864 module_sp->SectionFileAddressesChanged(); 1865 } 1866 1867 class MachSymtabSectionInfo { 1868 public: 1869 MachSymtabSectionInfo(SectionList *section_list) 1870 : m_section_list(section_list), m_section_infos() { 1871 // Get the number of sections down to a depth of 1 to include all segments 1872 // and their sections, but no other sections that may be added for debug 1873 // map or 1874 m_section_infos.resize(section_list->GetNumSections(1)); 1875 } 1876 1877 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1878 if (n_sect == 0) 1879 return SectionSP(); 1880 if (n_sect < m_section_infos.size()) { 1881 if (!m_section_infos[n_sect].section_sp) { 1882 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1883 m_section_infos[n_sect].section_sp = section_sp; 1884 if (section_sp) { 1885 m_section_infos[n_sect].vm_range.SetBaseAddress( 1886 section_sp->GetFileAddress()); 1887 m_section_infos[n_sect].vm_range.SetByteSize( 1888 section_sp->GetByteSize()); 1889 } else { 1890 const char *filename = "<unknown>"; 1891 SectionSP first_section_sp(m_section_list->GetSectionAtIndex(0)); 1892 if (first_section_sp) 1893 filename = first_section_sp->GetObjectFile()->GetFileSpec().GetPath().c_str(); 1894 1895 Host::SystemLog(Host::eSystemLogError, 1896 "error: unable to find section %d for a symbol in %s, corrupt file?\n", 1897 n_sect, 1898 filename); 1899 } 1900 } 1901 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1902 // Symbol is in section. 1903 return m_section_infos[n_sect].section_sp; 1904 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1905 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1906 file_addr) { 1907 // Symbol is in section with zero size, but has the same start address 1908 // as the section. This can happen with linker symbols (symbols that 1909 // start with the letter 'l' or 'L'. 1910 return m_section_infos[n_sect].section_sp; 1911 } 1912 } 1913 return m_section_list->FindSectionContainingFileAddress(file_addr); 1914 } 1915 1916 protected: 1917 struct SectionInfo { 1918 SectionInfo() : vm_range(), section_sp() {} 1919 1920 VMRange vm_range; 1921 SectionSP section_sp; 1922 }; 1923 SectionList *m_section_list; 1924 std::vector<SectionInfo> m_section_infos; 1925 }; 1926 1927 #define TRIE_SYMBOL_IS_THUMB (1ULL << 63) 1928 struct TrieEntry { 1929 void Dump() const { 1930 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1931 static_cast<unsigned long long>(address), 1932 static_cast<unsigned long long>(flags), 1933 static_cast<unsigned long long>(other), name.GetCString()); 1934 if (import_name) 1935 printf(" -> \"%s\"\n", import_name.GetCString()); 1936 else 1937 printf("\n"); 1938 } 1939 ConstString name; 1940 uint64_t address = LLDB_INVALID_ADDRESS; 1941 uint64_t flags = 1942 0; // EXPORT_SYMBOL_FLAGS_REEXPORT, EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER, 1943 // TRIE_SYMBOL_IS_THUMB 1944 uint64_t other = 0; 1945 ConstString import_name; 1946 }; 1947 1948 struct TrieEntryWithOffset { 1949 lldb::offset_t nodeOffset; 1950 TrieEntry entry; 1951 1952 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1953 1954 void Dump(uint32_t idx) const { 1955 printf("[%3u] 0x%16.16llx: ", idx, 1956 static_cast<unsigned long long>(nodeOffset)); 1957 entry.Dump(); 1958 } 1959 1960 bool operator<(const TrieEntryWithOffset &other) const { 1961 return (nodeOffset < other.nodeOffset); 1962 } 1963 }; 1964 1965 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 1966 const bool is_arm, addr_t text_seg_base_addr, 1967 std::vector<llvm::StringRef> &nameSlices, 1968 std::set<lldb::addr_t> &resolver_addresses, 1969 std::vector<TrieEntryWithOffset> &reexports, 1970 std::vector<TrieEntryWithOffset> &ext_symbols) { 1971 if (!data.ValidOffset(offset)) 1972 return true; 1973 1974 // Terminal node -- end of a branch, possibly add this to 1975 // the symbol table or resolver table. 1976 const uint64_t terminalSize = data.GetULEB128(&offset); 1977 lldb::offset_t children_offset = offset + terminalSize; 1978 if (terminalSize != 0) { 1979 TrieEntryWithOffset e(offset); 1980 e.entry.flags = data.GetULEB128(&offset); 1981 const char *import_name = nullptr; 1982 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 1983 e.entry.address = 0; 1984 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 1985 import_name = data.GetCStr(&offset); 1986 } else { 1987 e.entry.address = data.GetULEB128(&offset); 1988 if (text_seg_base_addr != LLDB_INVALID_ADDRESS) 1989 e.entry.address += text_seg_base_addr; 1990 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 1991 e.entry.other = data.GetULEB128(&offset); 1992 uint64_t resolver_addr = e.entry.other; 1993 if (is_arm) 1994 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 1995 resolver_addresses.insert(resolver_addr); 1996 } else 1997 e.entry.other = 0; 1998 } 1999 bool add_this_entry = false; 2000 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT) && 2001 import_name && import_name[0]) { 2002 // add symbols that are reexport symbols with a valid import name. 2003 add_this_entry = true; 2004 } else if (e.entry.flags == 0 && 2005 (import_name == nullptr || import_name[0] == '\0')) { 2006 // add externally visible symbols, in case the nlist record has 2007 // been stripped/omitted. 2008 add_this_entry = true; 2009 } 2010 if (add_this_entry) { 2011 std::string name; 2012 if (!nameSlices.empty()) { 2013 for (auto name_slice : nameSlices) 2014 name.append(name_slice.data(), name_slice.size()); 2015 } 2016 if (name.size() > 1) { 2017 // Skip the leading '_' 2018 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2019 } 2020 if (import_name) { 2021 // Skip the leading '_' 2022 e.entry.import_name.SetCString(import_name + 1); 2023 } 2024 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT)) { 2025 reexports.push_back(e); 2026 } else { 2027 if (is_arm && (e.entry.address & 1)) { 2028 e.entry.flags |= TRIE_SYMBOL_IS_THUMB; 2029 e.entry.address &= THUMB_ADDRESS_BIT_MASK; 2030 } 2031 ext_symbols.push_back(e); 2032 } 2033 } 2034 } 2035 2036 const uint8_t childrenCount = data.GetU8(&children_offset); 2037 for (uint8_t i = 0; i < childrenCount; ++i) { 2038 const char *cstr = data.GetCStr(&children_offset); 2039 if (cstr) 2040 nameSlices.push_back(llvm::StringRef(cstr)); 2041 else 2042 return false; // Corrupt data 2043 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2044 if (childNodeOffset) { 2045 if (!ParseTrieEntries(data, childNodeOffset, is_arm, text_seg_base_addr, 2046 nameSlices, resolver_addresses, reexports, 2047 ext_symbols)) { 2048 return false; 2049 } 2050 } 2051 nameSlices.pop_back(); 2052 } 2053 return true; 2054 } 2055 2056 static SymbolType GetSymbolType(const char *&symbol_name, 2057 bool &demangled_is_synthesized, 2058 const SectionSP &text_section_sp, 2059 const SectionSP &data_section_sp, 2060 const SectionSP &data_dirty_section_sp, 2061 const SectionSP &data_const_section_sp, 2062 const SectionSP &symbol_section) { 2063 SymbolType type = eSymbolTypeInvalid; 2064 2065 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2066 if (symbol_section->IsDescendant(text_section_sp.get())) { 2067 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 2068 S_ATTR_SELF_MODIFYING_CODE | 2069 S_ATTR_SOME_INSTRUCTIONS)) 2070 type = eSymbolTypeData; 2071 else 2072 type = eSymbolTypeCode; 2073 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 2074 symbol_section->IsDescendant(data_dirty_section_sp.get()) || 2075 symbol_section->IsDescendant(data_const_section_sp.get())) { 2076 if (symbol_sect_name && 2077 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 2078 type = eSymbolTypeRuntime; 2079 2080 if (symbol_name) { 2081 llvm::StringRef symbol_name_ref(symbol_name); 2082 if (symbol_name_ref.startswith("OBJC_")) { 2083 static const llvm::StringRef g_objc_v2_prefix_class("OBJC_CLASS_$_"); 2084 static const llvm::StringRef g_objc_v2_prefix_metaclass( 2085 "OBJC_METACLASS_$_"); 2086 static const llvm::StringRef g_objc_v2_prefix_ivar("OBJC_IVAR_$_"); 2087 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 2088 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2089 type = eSymbolTypeObjCClass; 2090 demangled_is_synthesized = true; 2091 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 2092 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2093 type = eSymbolTypeObjCMetaClass; 2094 demangled_is_synthesized = true; 2095 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 2096 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2097 type = eSymbolTypeObjCIVar; 2098 demangled_is_synthesized = true; 2099 } 2100 } 2101 } 2102 } else if (symbol_sect_name && 2103 ::strstr(symbol_sect_name, "__gcc_except_tab") == 2104 symbol_sect_name) { 2105 type = eSymbolTypeException; 2106 } else { 2107 type = eSymbolTypeData; 2108 } 2109 } else if (symbol_sect_name && 2110 ::strstr(symbol_sect_name, "__IMPORT") == symbol_sect_name) { 2111 type = eSymbolTypeTrampoline; 2112 } 2113 return type; 2114 } 2115 2116 // Read the UUID out of a dyld_shared_cache file on-disk. 2117 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2118 const ByteOrder byte_order, 2119 const uint32_t addr_byte_size) { 2120 UUID dsc_uuid; 2121 DataBufferSP DscData = MapFileData( 2122 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2123 if (!DscData) 2124 return dsc_uuid; 2125 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2126 2127 char version_str[7]; 2128 lldb::offset_t offset = 0; 2129 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2130 version_str[6] = '\0'; 2131 if (strcmp(version_str, "dyld_v") == 0) { 2132 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2133 dsc_uuid = UUID::fromOptionalData( 2134 dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t)); 2135 } 2136 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2137 if (log && dsc_uuid.IsValid()) { 2138 LLDB_LOGF(log, "Shared cache %s has UUID %s", 2139 dyld_shared_cache.GetPath().c_str(), 2140 dsc_uuid.GetAsString().c_str()); 2141 } 2142 return dsc_uuid; 2143 } 2144 2145 static llvm::Optional<struct nlist_64> 2146 ParseNList(DataExtractor &nlist_data, lldb::offset_t &nlist_data_offset, 2147 size_t nlist_byte_size) { 2148 struct nlist_64 nlist; 2149 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2150 return {}; 2151 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 2152 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 2153 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 2154 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 2155 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 2156 return nlist; 2157 } 2158 2159 enum { DebugSymbols = true, NonDebugSymbols = false }; 2160 2161 size_t ObjectFileMachO::ParseSymtab() { 2162 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2163 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2164 m_file.GetFilename().AsCString("")); 2165 ModuleSP module_sp(GetModule()); 2166 if (!module_sp) 2167 return 0; 2168 2169 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2170 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2171 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2172 // The data element of type bool indicates that this entry is thumb 2173 // code. 2174 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2175 2176 // Record the address of every function/data that we add to the symtab. 2177 // We add symbols to the table in the order of most information (nlist 2178 // records) to least (function starts), and avoid duplicating symbols 2179 // via this set. 2180 std::set<addr_t> symbols_added; 2181 FunctionStarts function_starts; 2182 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2183 uint32_t i; 2184 FileSpecList dylib_files; 2185 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2186 llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2187 llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2188 llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2189 2190 for (i = 0; i < m_header.ncmds; ++i) { 2191 const lldb::offset_t cmd_offset = offset; 2192 // Read in the load command and load command size 2193 struct load_command lc; 2194 if (m_data.GetU32(&offset, &lc, 2) == nullptr) 2195 break; 2196 // Watch for the symbol table load command 2197 switch (lc.cmd) { 2198 case LC_SYMTAB: 2199 symtab_load_command.cmd = lc.cmd; 2200 symtab_load_command.cmdsize = lc.cmdsize; 2201 // Read in the rest of the symtab load command 2202 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2203 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2204 return 0; 2205 break; 2206 2207 case LC_DYLD_INFO: 2208 case LC_DYLD_INFO_ONLY: 2209 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2210 dyld_info.cmd = lc.cmd; 2211 dyld_info.cmdsize = lc.cmdsize; 2212 } else { 2213 memset(&dyld_info, 0, sizeof(dyld_info)); 2214 } 2215 break; 2216 2217 case LC_LOAD_DYLIB: 2218 case LC_LOAD_WEAK_DYLIB: 2219 case LC_REEXPORT_DYLIB: 2220 case LC_LOADFVMLIB: 2221 case LC_LOAD_UPWARD_DYLIB: { 2222 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2223 const char *path = m_data.PeekCStr(name_offset); 2224 if (path) { 2225 FileSpec file_spec(path); 2226 // Strip the path if there is @rpath, @executable, etc so we just use 2227 // the basename 2228 if (path[0] == '@') 2229 file_spec.GetDirectory().Clear(); 2230 2231 if (lc.cmd == LC_REEXPORT_DYLIB) { 2232 m_reexported_dylibs.AppendIfUnique(file_spec); 2233 } 2234 2235 dylib_files.Append(file_spec); 2236 } 2237 } break; 2238 2239 case LC_FUNCTION_STARTS: 2240 function_starts_load_command.cmd = lc.cmd; 2241 function_starts_load_command.cmdsize = lc.cmdsize; 2242 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2243 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2244 memset(&function_starts_load_command, 0, 2245 sizeof(function_starts_load_command)); 2246 break; 2247 2248 default: 2249 break; 2250 } 2251 offset = cmd_offset + lc.cmdsize; 2252 } 2253 2254 if (!symtab_load_command.cmd) 2255 return 0; 2256 2257 Symtab *symtab = m_symtab_up.get(); 2258 SectionList *section_list = GetSectionList(); 2259 if (section_list == nullptr) 2260 return 0; 2261 2262 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2263 const ByteOrder byte_order = m_data.GetByteOrder(); 2264 bool bit_width_32 = addr_byte_size == 4; 2265 const size_t nlist_byte_size = 2266 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2267 2268 DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size); 2269 DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size); 2270 DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size); 2271 DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order, 2272 addr_byte_size); 2273 DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size); 2274 2275 const addr_t nlist_data_byte_size = 2276 symtab_load_command.nsyms * nlist_byte_size; 2277 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2278 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2279 2280 ProcessSP process_sp(m_process_wp.lock()); 2281 Process *process = process_sp.get(); 2282 2283 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2284 bool is_shared_cache_image = m_header.flags & MH_DYLIB_IN_CACHE; 2285 bool is_local_shared_cache_image = is_shared_cache_image && !IsInMemory(); 2286 SectionSP linkedit_section_sp( 2287 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2288 2289 if (process && m_header.filetype != llvm::MachO::MH_OBJECT && 2290 !is_local_shared_cache_image) { 2291 Target &target = process->GetTarget(); 2292 2293 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2294 2295 // Reading mach file from memory in a process or core file... 2296 2297 if (linkedit_section_sp) { 2298 addr_t linkedit_load_addr = 2299 linkedit_section_sp->GetLoadBaseAddress(&target); 2300 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2301 // We might be trying to access the symbol table before the 2302 // __LINKEDIT's load address has been set in the target. We can't 2303 // fail to read the symbol table, so calculate the right address 2304 // manually 2305 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2306 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2307 } 2308 2309 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2310 const addr_t symoff_addr = linkedit_load_addr + 2311 symtab_load_command.symoff - 2312 linkedit_file_offset; 2313 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2314 linkedit_file_offset; 2315 2316 // Always load dyld - the dynamic linker - from memory if we didn't 2317 // find a binary anywhere else. lldb will not register 2318 // dylib/framework/bundle loads/unloads if we don't have the dyld 2319 // symbols, we force dyld to load from memory despite the user's 2320 // target.memory-module-load-level setting. 2321 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2322 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2323 DataBufferSP nlist_data_sp( 2324 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2325 if (nlist_data_sp) 2326 nlist_data.SetData(nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2327 if (m_dysymtab.nindirectsyms != 0) { 2328 const addr_t indirect_syms_addr = linkedit_load_addr + 2329 m_dysymtab.indirectsymoff - 2330 linkedit_file_offset; 2331 DataBufferSP indirect_syms_data_sp(ReadMemory( 2332 process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 2333 if (indirect_syms_data_sp) 2334 indirect_symbol_index_data.SetData( 2335 indirect_syms_data_sp, 0, 2336 indirect_syms_data_sp->GetByteSize()); 2337 // If this binary is outside the shared cache, 2338 // cache the string table. 2339 // Binaries in the shared cache all share a giant string table, 2340 // and we can't share the string tables across multiple 2341 // ObjectFileMachO's, so we'd end up re-reading this mega-strtab 2342 // for every binary in the shared cache - it would be a big perf 2343 // problem. For binaries outside the shared cache, it's faster to 2344 // read the entire strtab at once instead of piece-by-piece as we 2345 // process the nlist records. 2346 if (!is_shared_cache_image) { 2347 DataBufferSP strtab_data_sp( 2348 ReadMemory(process_sp, strtab_addr, strtab_data_byte_size)); 2349 if (strtab_data_sp) { 2350 strtab_data.SetData(strtab_data_sp, 0, 2351 strtab_data_sp->GetByteSize()); 2352 } 2353 } 2354 } 2355 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) { 2356 if (function_starts_load_command.cmd) { 2357 const addr_t func_start_addr = 2358 linkedit_load_addr + function_starts_load_command.dataoff - 2359 linkedit_file_offset; 2360 DataBufferSP func_start_data_sp( 2361 ReadMemory(process_sp, func_start_addr, 2362 function_starts_load_command.datasize)); 2363 if (func_start_data_sp) 2364 function_starts_data.SetData(func_start_data_sp, 0, 2365 func_start_data_sp->GetByteSize()); 2366 } 2367 } 2368 } 2369 } 2370 } else { 2371 if (is_local_shared_cache_image) { 2372 // The load commands in shared cache images are relative to the 2373 // beginning of the shared cache, not the library image. The 2374 // data we get handed when creating the ObjectFileMachO starts 2375 // at the beginning of a specific library and spans to the end 2376 // of the cache to be able to reach the shared LINKEDIT 2377 // segments. We need to convert the load command offsets to be 2378 // relative to the beginning of our specific image. 2379 lldb::addr_t linkedit_offset = linkedit_section_sp->GetFileOffset(); 2380 lldb::offset_t linkedit_slide = 2381 linkedit_offset - m_linkedit_original_offset; 2382 symtab_load_command.symoff += linkedit_slide; 2383 symtab_load_command.stroff += linkedit_slide; 2384 dyld_info.export_off += linkedit_slide; 2385 m_dysymtab.indirectsymoff += linkedit_slide; 2386 function_starts_load_command.dataoff += linkedit_slide; 2387 } 2388 2389 nlist_data.SetData(m_data, symtab_load_command.symoff, 2390 nlist_data_byte_size); 2391 strtab_data.SetData(m_data, symtab_load_command.stroff, 2392 strtab_data_byte_size); 2393 2394 if (dyld_info.export_size > 0) { 2395 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2396 dyld_info.export_size); 2397 } 2398 2399 if (m_dysymtab.nindirectsyms != 0) { 2400 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2401 m_dysymtab.nindirectsyms * 4); 2402 } 2403 if (function_starts_load_command.cmd) { 2404 function_starts_data.SetData(m_data, function_starts_load_command.dataoff, 2405 function_starts_load_command.datasize); 2406 } 2407 } 2408 2409 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2410 2411 ConstString g_segment_name_TEXT = GetSegmentNameTEXT(); 2412 ConstString g_segment_name_DATA = GetSegmentNameDATA(); 2413 ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2414 ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2415 ConstString g_segment_name_OBJC = GetSegmentNameOBJC(); 2416 ConstString g_section_name_eh_frame = GetSectionNameEHFrame(); 2417 SectionSP text_section_sp( 2418 section_list->FindSectionByName(g_segment_name_TEXT)); 2419 SectionSP data_section_sp( 2420 section_list->FindSectionByName(g_segment_name_DATA)); 2421 SectionSP data_dirty_section_sp( 2422 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2423 SectionSP data_const_section_sp( 2424 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2425 SectionSP objc_section_sp( 2426 section_list->FindSectionByName(g_segment_name_OBJC)); 2427 SectionSP eh_frame_section_sp; 2428 if (text_section_sp.get()) 2429 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2430 g_section_name_eh_frame); 2431 else 2432 eh_frame_section_sp = 2433 section_list->FindSectionByName(g_section_name_eh_frame); 2434 2435 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2436 const bool always_thumb = GetArchitecture().IsAlwaysThumbInstructions(); 2437 2438 // lldb works best if it knows the start address of all functions in a 2439 // module. Linker symbols or debug info are normally the best source of 2440 // information for start addr / size but they may be stripped in a released 2441 // binary. Two additional sources of information exist in Mach-O binaries: 2442 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2443 // function's start address in the 2444 // binary, relative to the text section. 2445 // eh_frame - the eh_frame FDEs have the start addr & size of 2446 // each function 2447 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2448 // all modern binaries. 2449 // Binaries built to run on older releases may need to use eh_frame 2450 // information. 2451 2452 if (text_section_sp && function_starts_data.GetByteSize()) { 2453 FunctionStarts::Entry function_start_entry; 2454 function_start_entry.data = false; 2455 lldb::offset_t function_start_offset = 0; 2456 function_start_entry.addr = text_section_sp->GetFileAddress(); 2457 uint64_t delta; 2458 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2459 0) { 2460 // Now append the current entry 2461 function_start_entry.addr += delta; 2462 if (is_arm) { 2463 if (function_start_entry.addr & 1) { 2464 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2465 function_start_entry.data = true; 2466 } else if (always_thumb) { 2467 function_start_entry.data = true; 2468 } 2469 } 2470 function_starts.Append(function_start_entry); 2471 } 2472 } else { 2473 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2474 // load command claiming an eh_frame but it doesn't actually have the 2475 // eh_frame content. And if we have a dSYM, we don't need to do any of 2476 // this fill-in-the-missing-symbols works anyway - the debug info should 2477 // give us all the functions in the module. 2478 if (text_section_sp.get() && eh_frame_section_sp.get() && 2479 m_type != eTypeDebugInfo) { 2480 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2481 DWARFCallFrameInfo::EH); 2482 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2483 eh_frame.GetFunctionAddressAndSizeVector(functions); 2484 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2485 size_t count = functions.GetSize(); 2486 for (size_t i = 0; i < count; ++i) { 2487 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2488 functions.GetEntryAtIndex(i); 2489 if (func) { 2490 FunctionStarts::Entry function_start_entry; 2491 function_start_entry.addr = func->base - text_base_addr; 2492 if (is_arm) { 2493 if (function_start_entry.addr & 1) { 2494 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2495 function_start_entry.data = true; 2496 } else if (always_thumb) { 2497 function_start_entry.data = true; 2498 } 2499 } 2500 function_starts.Append(function_start_entry); 2501 } 2502 } 2503 } 2504 } 2505 2506 const size_t function_starts_count = function_starts.GetSize(); 2507 2508 // For user process binaries (executables, dylibs, frameworks, bundles), if 2509 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2510 // going to assume the binary has been stripped. Don't allow assembly 2511 // language instruction emulation because we don't know proper function 2512 // start boundaries. 2513 // 2514 // For all other types of binaries (kernels, stand-alone bare board 2515 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2516 // sections - we should not make any assumptions about them based on that. 2517 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2518 m_allow_assembly_emulation_unwind_plans = false; 2519 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2520 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2521 2522 if (unwind_or_symbol_log) 2523 module_sp->LogMessage( 2524 unwind_or_symbol_log, 2525 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2526 } 2527 2528 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() 2529 ? eh_frame_section_sp->GetID() 2530 : static_cast<user_id_t>(NO_SECT); 2531 2532 lldb::offset_t nlist_data_offset = 0; 2533 2534 uint32_t N_SO_index = UINT32_MAX; 2535 2536 MachSymtabSectionInfo section_info(section_list); 2537 std::vector<uint32_t> N_FUN_indexes; 2538 std::vector<uint32_t> N_NSYM_indexes; 2539 std::vector<uint32_t> N_INCL_indexes; 2540 std::vector<uint32_t> N_BRAC_indexes; 2541 std::vector<uint32_t> N_COMM_indexes; 2542 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2543 typedef llvm::DenseMap<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2544 typedef llvm::DenseMap<const char *, uint32_t> ConstNameToSymbolIndexMap; 2545 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2546 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2547 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2548 // Any symbols that get merged into another will get an entry in this map 2549 // so we know 2550 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2551 uint32_t nlist_idx = 0; 2552 Symbol *symbol_ptr = nullptr; 2553 2554 uint32_t sym_idx = 0; 2555 Symbol *sym = nullptr; 2556 size_t num_syms = 0; 2557 std::string memory_symbol_name; 2558 uint32_t unmapped_local_symbols_found = 0; 2559 2560 std::vector<TrieEntryWithOffset> reexport_trie_entries; 2561 std::vector<TrieEntryWithOffset> external_sym_trie_entries; 2562 std::set<lldb::addr_t> resolver_addresses; 2563 2564 if (dyld_trie_data.GetByteSize() > 0) { 2565 ConstString text_segment_name("__TEXT"); 2566 SectionSP text_segment_sp = 2567 GetSectionList()->FindSectionByName(text_segment_name); 2568 lldb::addr_t text_segment_file_addr = LLDB_INVALID_ADDRESS; 2569 if (text_segment_sp) 2570 text_segment_file_addr = text_segment_sp->GetFileAddress(); 2571 std::vector<llvm::StringRef> nameSlices; 2572 ParseTrieEntries(dyld_trie_data, 0, is_arm, text_segment_file_addr, 2573 nameSlices, resolver_addresses, reexport_trie_entries, 2574 external_sym_trie_entries); 2575 } 2576 2577 typedef std::set<ConstString> IndirectSymbols; 2578 IndirectSymbols indirect_symbol_names; 2579 2580 #if defined(__APPLE__) && TARGET_OS_EMBEDDED 2581 2582 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2583 // optimized by moving LOCAL symbols out of the memory mapped portion of 2584 // the DSC. The symbol information has all been retained, but it isn't 2585 // available in the normal nlist data. However, there *are* duplicate 2586 // entries of *some* 2587 // LOCAL symbols in the normal nlist data. To handle this situation 2588 // correctly, we must first attempt 2589 // to parse any DSC unmapped symbol information. If we find any, we set a 2590 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2591 2592 if (m_header.flags & MH_DYLIB_IN_CACHE) { 2593 // Before we can start mapping the DSC, we need to make certain the 2594 // target process is actually using the cache we can find. 2595 2596 // Next we need to determine the correct path for the dyld shared cache. 2597 2598 ArchSpec header_arch = GetArchitecture(); 2599 char dsc_path[PATH_MAX]; 2600 char dsc_path_development[PATH_MAX]; 2601 2602 snprintf( 2603 dsc_path, sizeof(dsc_path), "%s%s%s", 2604 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2605 */ 2606 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2607 header_arch.GetArchitectureName()); 2608 2609 snprintf( 2610 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2611 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2612 */ 2613 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2614 header_arch.GetArchitectureName(), ".development"); 2615 2616 FileSpec dsc_nondevelopment_filespec(dsc_path); 2617 FileSpec dsc_development_filespec(dsc_path_development); 2618 FileSpec dsc_filespec; 2619 2620 UUID dsc_uuid; 2621 UUID process_shared_cache_uuid; 2622 addr_t process_shared_cache_base_addr; 2623 2624 if (process) { 2625 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, 2626 process_shared_cache_uuid); 2627 } 2628 2629 // First see if we can find an exact match for the inferior process 2630 // shared cache UUID in the development or non-development shared caches 2631 // on disk. 2632 if (process_shared_cache_uuid.IsValid()) { 2633 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2634 UUID dsc_development_uuid = GetSharedCacheUUID( 2635 dsc_development_filespec, byte_order, addr_byte_size); 2636 if (dsc_development_uuid.IsValid() && 2637 dsc_development_uuid == process_shared_cache_uuid) { 2638 dsc_filespec = dsc_development_filespec; 2639 dsc_uuid = dsc_development_uuid; 2640 } 2641 } 2642 if (!dsc_uuid.IsValid() && 2643 FileSystem::Instance().Exists(dsc_nondevelopment_filespec)) { 2644 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2645 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2646 if (dsc_nondevelopment_uuid.IsValid() && 2647 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2648 dsc_filespec = dsc_nondevelopment_filespec; 2649 dsc_uuid = dsc_nondevelopment_uuid; 2650 } 2651 } 2652 } 2653 2654 // Failing a UUID match, prefer the development dyld_shared cache if both 2655 // are present. 2656 if (!FileSystem::Instance().Exists(dsc_filespec)) { 2657 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2658 dsc_filespec = dsc_development_filespec; 2659 } else { 2660 dsc_filespec = dsc_nondevelopment_filespec; 2661 } 2662 } 2663 2664 /* The dyld_cache_header has a pointer to the 2665 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2666 The dyld_cache_local_symbols_info structure gives us three things: 2667 1. The start and count of the nlist records in the dyld_shared_cache 2668 file 2669 2. The start and size of the strings for these nlist records 2670 3. The start and count of dyld_cache_local_symbols_entry entries 2671 2672 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2673 dyld shared cache. 2674 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2675 the dyld shared cache. 2676 The dyld_cache_local_symbols_entry also lists the start of this 2677 dylib/framework's nlist records 2678 and the count of how many nlist records there are for this 2679 dylib/framework. 2680 */ 2681 2682 // Process the dyld shared cache header to find the unmapped symbols 2683 2684 DataBufferSP dsc_data_sp = MapFileData( 2685 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2686 if (!dsc_uuid.IsValid()) { 2687 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2688 } 2689 if (dsc_data_sp) { 2690 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2691 2692 bool uuid_match = true; 2693 if (dsc_uuid.IsValid() && process) { 2694 if (process_shared_cache_uuid.IsValid() && 2695 dsc_uuid != process_shared_cache_uuid) { 2696 // The on-disk dyld_shared_cache file is not the same as the one in 2697 // this process' memory, don't use it. 2698 uuid_match = false; 2699 ModuleSP module_sp(GetModule()); 2700 if (module_sp) 2701 module_sp->ReportWarning("process shared cache does not match " 2702 "on-disk dyld_shared_cache file, some " 2703 "symbol names will be missing."); 2704 } 2705 } 2706 2707 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2708 2709 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2710 2711 // If the mappingOffset points to a location inside the header, we've 2712 // opened an old dyld shared cache, and should not proceed further. 2713 if (uuid_match && 2714 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2715 2716 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2717 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2718 mappingOffset); 2719 2720 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2721 byte_order, addr_byte_size); 2722 offset = 0; 2723 2724 // The File addresses (from the in-memory Mach-O load commands) for 2725 // the shared libraries in the shared library cache need to be 2726 // adjusted by an offset to match up with the dylibOffset identifying 2727 // field in the dyld_cache_local_symbol_entry's. This offset is 2728 // recorded in mapping_offset_value. 2729 const uint64_t mapping_offset_value = 2730 dsc_mapping_info_data.GetU64(&offset); 2731 2732 offset = 2733 offsetof(struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2734 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2735 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2736 2737 if (localSymbolsOffset && localSymbolsSize) { 2738 // Map the local symbols 2739 DataBufferSP dsc_local_symbols_data_sp = 2740 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2741 2742 if (dsc_local_symbols_data_sp) { 2743 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2744 byte_order, addr_byte_size); 2745 2746 offset = 0; 2747 2748 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 2749 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 2750 UndefinedNameToDescMap undefined_name_to_desc; 2751 SymbolIndexToName reexport_shlib_needs_fixup; 2752 2753 // Read the local_symbols_infos struct in one shot 2754 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2755 dsc_local_symbols_data.GetU32(&offset, 2756 &local_symbols_info.nlistOffset, 6); 2757 2758 SectionSP text_section_sp( 2759 section_list->FindSectionByName(GetSegmentNameTEXT())); 2760 2761 uint32_t header_file_offset = 2762 (text_section_sp->GetFileAddress() - mapping_offset_value); 2763 2764 offset = local_symbols_info.entriesOffset; 2765 for (uint32_t entry_index = 0; 2766 entry_index < local_symbols_info.entriesCount; entry_index++) { 2767 struct lldb_copy_dyld_cache_local_symbols_entry 2768 local_symbols_entry; 2769 local_symbols_entry.dylibOffset = 2770 dsc_local_symbols_data.GetU32(&offset); 2771 local_symbols_entry.nlistStartIndex = 2772 dsc_local_symbols_data.GetU32(&offset); 2773 local_symbols_entry.nlistCount = 2774 dsc_local_symbols_data.GetU32(&offset); 2775 2776 if (header_file_offset == local_symbols_entry.dylibOffset) { 2777 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2778 2779 // The normal nlist code cannot correctly size the Symbols 2780 // array, we need to allocate it here. 2781 sym = symtab->Resize( 2782 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2783 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2784 num_syms = symtab->GetNumSymbols(); 2785 2786 nlist_data_offset = 2787 local_symbols_info.nlistOffset + 2788 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2789 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2790 2791 for (uint32_t nlist_index = 0; 2792 nlist_index < local_symbols_entry.nlistCount; 2793 nlist_index++) { 2794 ///////////////////////////// 2795 { 2796 llvm::Optional<struct nlist_64> nlist_maybe = 2797 ParseNList(dsc_local_symbols_data, nlist_data_offset, 2798 nlist_byte_size); 2799 if (!nlist_maybe) 2800 break; 2801 struct nlist_64 nlist = *nlist_maybe; 2802 2803 SymbolType type = eSymbolTypeInvalid; 2804 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2805 string_table_offset + nlist.n_strx); 2806 2807 if (symbol_name == NULL) { 2808 // No symbol should be NULL, even the symbols with no 2809 // string values should have an offset zero which 2810 // points to an empty C-string 2811 Host::SystemLog( 2812 Host::eSystemLogError, 2813 "error: DSC unmapped local symbol[%u] has invalid " 2814 "string table offset 0x%x in %s, ignoring symbol\n", 2815 entry_index, nlist.n_strx, 2816 module_sp->GetFileSpec().GetPath().c_str()); 2817 continue; 2818 } 2819 if (symbol_name[0] == '\0') 2820 symbol_name = NULL; 2821 2822 const char *symbol_name_non_abi_mangled = NULL; 2823 2824 SectionSP symbol_section; 2825 uint32_t symbol_byte_size = 0; 2826 bool add_nlist = true; 2827 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2828 bool demangled_is_synthesized = false; 2829 bool is_gsym = false; 2830 bool set_value = true; 2831 2832 assert(sym_idx < num_syms); 2833 2834 sym[sym_idx].SetDebug(is_debug); 2835 2836 if (is_debug) { 2837 switch (nlist.n_type) { 2838 case N_GSYM: 2839 // global symbol: name,,NO_SECT,type,0 2840 // Sometimes the N_GSYM value contains the address. 2841 2842 // FIXME: In the .o files, we have a GSYM and a debug 2843 // symbol for all the ObjC data. They 2844 // have the same address, but we want to ensure that 2845 // we always find only the real symbol, 'cause we 2846 // don't currently correctly attribute the 2847 // GSYM one to the ObjCClass/Ivar/MetaClass 2848 // symbol type. This is a temporary hack to make 2849 // sure the ObjectiveC symbols get treated correctly. 2850 // To do this right, we should coalesce all the GSYM 2851 // & global symbols that have the same address. 2852 2853 is_gsym = true; 2854 sym[sym_idx].SetExternal(true); 2855 2856 if (symbol_name && symbol_name[0] == '_' && 2857 symbol_name[1] == 'O') { 2858 llvm::StringRef symbol_name_ref(symbol_name); 2859 if (symbol_name_ref.startswith( 2860 g_objc_v2_prefix_class)) { 2861 symbol_name_non_abi_mangled = symbol_name + 1; 2862 symbol_name = 2863 symbol_name + g_objc_v2_prefix_class.size(); 2864 type = eSymbolTypeObjCClass; 2865 demangled_is_synthesized = true; 2866 2867 } else if (symbol_name_ref.startswith( 2868 g_objc_v2_prefix_metaclass)) { 2869 symbol_name_non_abi_mangled = symbol_name + 1; 2870 symbol_name = 2871 symbol_name + g_objc_v2_prefix_metaclass.size(); 2872 type = eSymbolTypeObjCMetaClass; 2873 demangled_is_synthesized = true; 2874 } else if (symbol_name_ref.startswith( 2875 g_objc_v2_prefix_ivar)) { 2876 symbol_name_non_abi_mangled = symbol_name + 1; 2877 symbol_name = 2878 symbol_name + g_objc_v2_prefix_ivar.size(); 2879 type = eSymbolTypeObjCIVar; 2880 demangled_is_synthesized = true; 2881 } 2882 } else { 2883 if (nlist.n_value != 0) 2884 symbol_section = section_info.GetSection( 2885 nlist.n_sect, nlist.n_value); 2886 type = eSymbolTypeData; 2887 } 2888 break; 2889 2890 case N_FNAME: 2891 // procedure name (f77 kludge): name,,NO_SECT,0,0 2892 type = eSymbolTypeCompiler; 2893 break; 2894 2895 case N_FUN: 2896 // procedure: name,,n_sect,linenumber,address 2897 if (symbol_name) { 2898 type = eSymbolTypeCode; 2899 symbol_section = section_info.GetSection( 2900 nlist.n_sect, nlist.n_value); 2901 2902 N_FUN_addr_to_sym_idx.insert( 2903 std::make_pair(nlist.n_value, sym_idx)); 2904 // We use the current number of symbols in the 2905 // symbol table in lieu of using nlist_idx in case 2906 // we ever start trimming entries out 2907 N_FUN_indexes.push_back(sym_idx); 2908 } else { 2909 type = eSymbolTypeCompiler; 2910 2911 if (!N_FUN_indexes.empty()) { 2912 // Copy the size of the function into the 2913 // original 2914 // STAB entry so we don't have 2915 // to hunt for it later 2916 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2917 ->SetByteSize(nlist.n_value); 2918 N_FUN_indexes.pop_back(); 2919 // We don't really need the end function STAB as 2920 // it contains the size which we already placed 2921 // with the original symbol, so don't add it if 2922 // we want a minimal symbol table 2923 add_nlist = false; 2924 } 2925 } 2926 break; 2927 2928 case N_STSYM: 2929 // static symbol: name,,n_sect,type,address 2930 N_STSYM_addr_to_sym_idx.insert( 2931 std::make_pair(nlist.n_value, sym_idx)); 2932 symbol_section = section_info.GetSection(nlist.n_sect, 2933 nlist.n_value); 2934 if (symbol_name && symbol_name[0]) { 2935 type = ObjectFile::GetSymbolTypeFromName( 2936 symbol_name + 1, eSymbolTypeData); 2937 } 2938 break; 2939 2940 case N_LCSYM: 2941 // .lcomm symbol: name,,n_sect,type,address 2942 symbol_section = section_info.GetSection(nlist.n_sect, 2943 nlist.n_value); 2944 type = eSymbolTypeCommonBlock; 2945 break; 2946 2947 case N_BNSYM: 2948 // We use the current number of symbols in the symbol 2949 // table in lieu of using nlist_idx in case we ever 2950 // start trimming entries out Skip these if we want 2951 // minimal symbol tables 2952 add_nlist = false; 2953 break; 2954 2955 case N_ENSYM: 2956 // Set the size of the N_BNSYM to the terminating 2957 // index of this N_ENSYM so that we can always skip 2958 // the entire symbol if we need to navigate more 2959 // quickly at the source level when parsing STABS 2960 // Skip these if we want minimal symbol tables 2961 add_nlist = false; 2962 break; 2963 2964 case N_OPT: 2965 // emitted with gcc2_compiled and in gcc source 2966 type = eSymbolTypeCompiler; 2967 break; 2968 2969 case N_RSYM: 2970 // register sym: name,,NO_SECT,type,register 2971 type = eSymbolTypeVariable; 2972 break; 2973 2974 case N_SLINE: 2975 // src line: 0,,n_sect,linenumber,address 2976 symbol_section = section_info.GetSection(nlist.n_sect, 2977 nlist.n_value); 2978 type = eSymbolTypeLineEntry; 2979 break; 2980 2981 case N_SSYM: 2982 // structure elt: name,,NO_SECT,type,struct_offset 2983 type = eSymbolTypeVariableType; 2984 break; 2985 2986 case N_SO: 2987 // source file name 2988 type = eSymbolTypeSourceFile; 2989 if (symbol_name == NULL) { 2990 add_nlist = false; 2991 if (N_SO_index != UINT32_MAX) { 2992 // Set the size of the N_SO to the terminating 2993 // index of this N_SO so that we can always skip 2994 // the entire N_SO if we need to navigate more 2995 // quickly at the source level when parsing STABS 2996 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2997 symbol_ptr->SetByteSize(sym_idx); 2998 symbol_ptr->SetSizeIsSibling(true); 2999 } 3000 N_NSYM_indexes.clear(); 3001 N_INCL_indexes.clear(); 3002 N_BRAC_indexes.clear(); 3003 N_COMM_indexes.clear(); 3004 N_FUN_indexes.clear(); 3005 N_SO_index = UINT32_MAX; 3006 } else { 3007 // We use the current number of symbols in the 3008 // symbol table in lieu of using nlist_idx in case 3009 // we ever start trimming entries out 3010 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3011 if (N_SO_has_full_path) { 3012 if ((N_SO_index == sym_idx - 1) && 3013 ((sym_idx - 1) < num_syms)) { 3014 // We have two consecutive N_SO entries where 3015 // the first contains a directory and the 3016 // second contains a full path. 3017 sym[sym_idx - 1].GetMangled().SetValue( 3018 ConstString(symbol_name), false); 3019 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3020 add_nlist = false; 3021 } else { 3022 // This is the first entry in a N_SO that 3023 // contains a directory or 3024 // a full path to the source file 3025 N_SO_index = sym_idx; 3026 } 3027 } else if ((N_SO_index == sym_idx - 1) && 3028 ((sym_idx - 1) < num_syms)) { 3029 // This is usually the second N_SO entry that 3030 // contains just the filename, so here we combine 3031 // it with the first one if we are minimizing the 3032 // symbol table 3033 const char *so_path = 3034 sym[sym_idx - 1] 3035 .GetMangled() 3036 .GetDemangledName( 3037 lldb::eLanguageTypeUnknown) 3038 .AsCString(); 3039 if (so_path && so_path[0]) { 3040 std::string full_so_path(so_path); 3041 const size_t double_slash_pos = 3042 full_so_path.find("//"); 3043 if (double_slash_pos != std::string::npos) { 3044 // The linker has been generating bad N_SO 3045 // entries with doubled up paths 3046 // in the format "%s%s" where the first 3047 // string in the DW_AT_comp_dir, and the 3048 // second is the directory for the source 3049 // file so you end up with a path that looks 3050 // like "/tmp/src//tmp/src/" 3051 FileSpec so_dir(so_path); 3052 if (!FileSystem::Instance().Exists(so_dir)) { 3053 so_dir.SetFile( 3054 &full_so_path[double_slash_pos + 1], 3055 FileSpec::Style::native); 3056 if (FileSystem::Instance().Exists(so_dir)) { 3057 // Trim off the incorrect path 3058 full_so_path.erase(0, double_slash_pos + 1); 3059 } 3060 } 3061 } 3062 if (*full_so_path.rbegin() != '/') 3063 full_so_path += '/'; 3064 full_so_path += symbol_name; 3065 sym[sym_idx - 1].GetMangled().SetValue( 3066 ConstString(full_so_path.c_str()), false); 3067 add_nlist = false; 3068 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3069 } 3070 } else { 3071 // This could be a relative path to a N_SO 3072 N_SO_index = sym_idx; 3073 } 3074 } 3075 break; 3076 3077 case N_OSO: 3078 // object file name: name,,0,0,st_mtime 3079 type = eSymbolTypeObjectFile; 3080 break; 3081 3082 case N_LSYM: 3083 // local sym: name,,NO_SECT,type,offset 3084 type = eSymbolTypeLocal; 3085 break; 3086 3087 // INCL scopes 3088 case N_BINCL: 3089 // include file beginning: name,,NO_SECT,0,sum We use 3090 // the current number of symbols in the symbol table 3091 // in lieu of using nlist_idx in case we ever start 3092 // trimming entries out 3093 N_INCL_indexes.push_back(sym_idx); 3094 type = eSymbolTypeScopeBegin; 3095 break; 3096 3097 case N_EINCL: 3098 // include file end: name,,NO_SECT,0,0 3099 // Set the size of the N_BINCL to the terminating 3100 // index of this N_EINCL so that we can always skip 3101 // the entire symbol if we need to navigate more 3102 // quickly at the source level when parsing STABS 3103 if (!N_INCL_indexes.empty()) { 3104 symbol_ptr = 3105 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3106 symbol_ptr->SetByteSize(sym_idx + 1); 3107 symbol_ptr->SetSizeIsSibling(true); 3108 N_INCL_indexes.pop_back(); 3109 } 3110 type = eSymbolTypeScopeEnd; 3111 break; 3112 3113 case N_SOL: 3114 // #included file name: name,,n_sect,0,address 3115 type = eSymbolTypeHeaderFile; 3116 3117 // We currently don't use the header files on darwin 3118 add_nlist = false; 3119 break; 3120 3121 case N_PARAMS: 3122 // compiler parameters: name,,NO_SECT,0,0 3123 type = eSymbolTypeCompiler; 3124 break; 3125 3126 case N_VERSION: 3127 // compiler version: name,,NO_SECT,0,0 3128 type = eSymbolTypeCompiler; 3129 break; 3130 3131 case N_OLEVEL: 3132 // compiler -O level: name,,NO_SECT,0,0 3133 type = eSymbolTypeCompiler; 3134 break; 3135 3136 case N_PSYM: 3137 // parameter: name,,NO_SECT,type,offset 3138 type = eSymbolTypeVariable; 3139 break; 3140 3141 case N_ENTRY: 3142 // alternate entry: name,,n_sect,linenumber,address 3143 symbol_section = section_info.GetSection(nlist.n_sect, 3144 nlist.n_value); 3145 type = eSymbolTypeLineEntry; 3146 break; 3147 3148 // Left and Right Braces 3149 case N_LBRAC: 3150 // left bracket: 0,,NO_SECT,nesting level,address We 3151 // use the current number of symbols in the symbol 3152 // table in lieu of using nlist_idx in case we ever 3153 // start trimming entries out 3154 symbol_section = section_info.GetSection(nlist.n_sect, 3155 nlist.n_value); 3156 N_BRAC_indexes.push_back(sym_idx); 3157 type = eSymbolTypeScopeBegin; 3158 break; 3159 3160 case N_RBRAC: 3161 // right bracket: 0,,NO_SECT,nesting level,address 3162 // Set the size of the N_LBRAC to the terminating 3163 // index of this N_RBRAC so that we can always skip 3164 // the entire symbol if we need to navigate more 3165 // quickly at the source level when parsing STABS 3166 symbol_section = section_info.GetSection(nlist.n_sect, 3167 nlist.n_value); 3168 if (!N_BRAC_indexes.empty()) { 3169 symbol_ptr = 3170 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3171 symbol_ptr->SetByteSize(sym_idx + 1); 3172 symbol_ptr->SetSizeIsSibling(true); 3173 N_BRAC_indexes.pop_back(); 3174 } 3175 type = eSymbolTypeScopeEnd; 3176 break; 3177 3178 case N_EXCL: 3179 // deleted include file: name,,NO_SECT,0,sum 3180 type = eSymbolTypeHeaderFile; 3181 break; 3182 3183 // COMM scopes 3184 case N_BCOMM: 3185 // begin common: name,,NO_SECT,0,0 3186 // We use the current number of symbols in the symbol 3187 // table in lieu of using nlist_idx in case we ever 3188 // start trimming entries out 3189 type = eSymbolTypeScopeBegin; 3190 N_COMM_indexes.push_back(sym_idx); 3191 break; 3192 3193 case N_ECOML: 3194 // end common (local name): 0,,n_sect,0,address 3195 symbol_section = section_info.GetSection(nlist.n_sect, 3196 nlist.n_value); 3197 // Fall through 3198 3199 case N_ECOMM: 3200 // end common: name,,n_sect,0,0 3201 // Set the size of the N_BCOMM to the terminating 3202 // index of this N_ECOMM/N_ECOML so that we can 3203 // always skip the entire symbol if we need to 3204 // navigate more quickly at the source level when 3205 // parsing STABS 3206 if (!N_COMM_indexes.empty()) { 3207 symbol_ptr = 3208 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3209 symbol_ptr->SetByteSize(sym_idx + 1); 3210 symbol_ptr->SetSizeIsSibling(true); 3211 N_COMM_indexes.pop_back(); 3212 } 3213 type = eSymbolTypeScopeEnd; 3214 break; 3215 3216 case N_LENG: 3217 // second stab entry with length information 3218 type = eSymbolTypeAdditional; 3219 break; 3220 3221 default: 3222 break; 3223 } 3224 } else { 3225 // uint8_t n_pext = N_PEXT & nlist.n_type; 3226 uint8_t n_type = N_TYPE & nlist.n_type; 3227 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3228 3229 switch (n_type) { 3230 case N_INDR: { 3231 const char *reexport_name_cstr = 3232 strtab_data.PeekCStr(nlist.n_value); 3233 if (reexport_name_cstr && reexport_name_cstr[0]) { 3234 type = eSymbolTypeReExported; 3235 ConstString reexport_name( 3236 reexport_name_cstr + 3237 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3238 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3239 set_value = false; 3240 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3241 indirect_symbol_names.insert(ConstString( 3242 symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3243 } else 3244 type = eSymbolTypeUndefined; 3245 } break; 3246 3247 case N_UNDF: 3248 if (symbol_name && symbol_name[0]) { 3249 ConstString undefined_name( 3250 symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3251 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3252 } 3253 // Fall through 3254 case N_PBUD: 3255 type = eSymbolTypeUndefined; 3256 break; 3257 3258 case N_ABS: 3259 type = eSymbolTypeAbsolute; 3260 break; 3261 3262 case N_SECT: { 3263 symbol_section = section_info.GetSection(nlist.n_sect, 3264 nlist.n_value); 3265 3266 if (symbol_section == NULL) { 3267 // TODO: warn about this? 3268 add_nlist = false; 3269 break; 3270 } 3271 3272 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3273 type = eSymbolTypeException; 3274 } else { 3275 uint32_t section_type = 3276 symbol_section->Get() & SECTION_TYPE; 3277 3278 switch (section_type) { 3279 case S_CSTRING_LITERALS: 3280 type = eSymbolTypeData; 3281 break; // section with only literal C strings 3282 case S_4BYTE_LITERALS: 3283 type = eSymbolTypeData; 3284 break; // section with only 4 byte literals 3285 case S_8BYTE_LITERALS: 3286 type = eSymbolTypeData; 3287 break; // section with only 8 byte literals 3288 case S_LITERAL_POINTERS: 3289 type = eSymbolTypeTrampoline; 3290 break; // section with only pointers to literals 3291 case S_NON_LAZY_SYMBOL_POINTERS: 3292 type = eSymbolTypeTrampoline; 3293 break; // section with only non-lazy symbol 3294 // pointers 3295 case S_LAZY_SYMBOL_POINTERS: 3296 type = eSymbolTypeTrampoline; 3297 break; // section with only lazy symbol pointers 3298 case S_SYMBOL_STUBS: 3299 type = eSymbolTypeTrampoline; 3300 break; // section with only symbol stubs, byte 3301 // size of stub in the reserved2 field 3302 case S_MOD_INIT_FUNC_POINTERS: 3303 type = eSymbolTypeCode; 3304 break; // section with only function pointers for 3305 // initialization 3306 case S_MOD_TERM_FUNC_POINTERS: 3307 type = eSymbolTypeCode; 3308 break; // section with only function pointers for 3309 // termination 3310 case S_INTERPOSING: 3311 type = eSymbolTypeTrampoline; 3312 break; // section with only pairs of function 3313 // pointers for interposing 3314 case S_16BYTE_LITERALS: 3315 type = eSymbolTypeData; 3316 break; // section with only 16 byte literals 3317 case S_DTRACE_DOF: 3318 type = eSymbolTypeInstrumentation; 3319 break; 3320 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3321 type = eSymbolTypeTrampoline; 3322 break; 3323 default: 3324 switch (symbol_section->GetType()) { 3325 case lldb::eSectionTypeCode: 3326 type = eSymbolTypeCode; 3327 break; 3328 case eSectionTypeData: 3329 case eSectionTypeDataCString: // Inlined C string 3330 // data 3331 case eSectionTypeDataCStringPointers: // Pointers 3332 // to C 3333 // string 3334 // data 3335 case eSectionTypeDataSymbolAddress: // Address of 3336 // a symbol in 3337 // the symbol 3338 // table 3339 case eSectionTypeData4: 3340 case eSectionTypeData8: 3341 case eSectionTypeData16: 3342 type = eSymbolTypeData; 3343 break; 3344 default: 3345 break; 3346 } 3347 break; 3348 } 3349 3350 if (type == eSymbolTypeInvalid) { 3351 const char *symbol_sect_name = 3352 symbol_section->GetName().AsCString(); 3353 if (symbol_section->IsDescendant( 3354 text_section_sp.get())) { 3355 if (symbol_section->IsClear( 3356 S_ATTR_PURE_INSTRUCTIONS | 3357 S_ATTR_SELF_MODIFYING_CODE | 3358 S_ATTR_SOME_INSTRUCTIONS)) 3359 type = eSymbolTypeData; 3360 else 3361 type = eSymbolTypeCode; 3362 } else if (symbol_section->IsDescendant( 3363 data_section_sp.get()) || 3364 symbol_section->IsDescendant( 3365 data_dirty_section_sp.get()) || 3366 symbol_section->IsDescendant( 3367 data_const_section_sp.get())) { 3368 if (symbol_sect_name && 3369 ::strstr(symbol_sect_name, "__objc") == 3370 symbol_sect_name) { 3371 type = eSymbolTypeRuntime; 3372 3373 if (symbol_name) { 3374 llvm::StringRef symbol_name_ref(symbol_name); 3375 if (symbol_name_ref.startswith("_OBJC_")) { 3376 llvm::StringRef 3377 g_objc_v2_prefix_class( 3378 "_OBJC_CLASS_$_"); 3379 llvm::StringRef 3380 g_objc_v2_prefix_metaclass( 3381 "_OBJC_METACLASS_$_"); 3382 llvm::StringRef 3383 g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 3384 if (symbol_name_ref.startswith( 3385 g_objc_v2_prefix_class)) { 3386 symbol_name_non_abi_mangled = 3387 symbol_name + 1; 3388 symbol_name = 3389 symbol_name + 3390 g_objc_v2_prefix_class.size(); 3391 type = eSymbolTypeObjCClass; 3392 demangled_is_synthesized = true; 3393 } else if ( 3394 symbol_name_ref.startswith( 3395 g_objc_v2_prefix_metaclass)) { 3396 symbol_name_non_abi_mangled = 3397 symbol_name + 1; 3398 symbol_name = 3399 symbol_name + 3400 g_objc_v2_prefix_metaclass.size(); 3401 type = eSymbolTypeObjCMetaClass; 3402 demangled_is_synthesized = true; 3403 } else if (symbol_name_ref.startswith( 3404 g_objc_v2_prefix_ivar)) { 3405 symbol_name_non_abi_mangled = 3406 symbol_name + 1; 3407 symbol_name = 3408 symbol_name + 3409 g_objc_v2_prefix_ivar.size(); 3410 type = eSymbolTypeObjCIVar; 3411 demangled_is_synthesized = true; 3412 } 3413 } 3414 } 3415 } else if (symbol_sect_name && 3416 ::strstr(symbol_sect_name, 3417 "__gcc_except_tab") == 3418 symbol_sect_name) { 3419 type = eSymbolTypeException; 3420 } else { 3421 type = eSymbolTypeData; 3422 } 3423 } else if (symbol_sect_name && 3424 ::strstr(symbol_sect_name, "__IMPORT") == 3425 symbol_sect_name) { 3426 type = eSymbolTypeTrampoline; 3427 } else if (symbol_section->IsDescendant( 3428 objc_section_sp.get())) { 3429 type = eSymbolTypeRuntime; 3430 if (symbol_name && symbol_name[0] == '.') { 3431 llvm::StringRef symbol_name_ref(symbol_name); 3432 llvm::StringRef 3433 g_objc_v1_prefix_class(".objc_class_name_"); 3434 if (symbol_name_ref.startswith( 3435 g_objc_v1_prefix_class)) { 3436 symbol_name_non_abi_mangled = symbol_name; 3437 symbol_name = symbol_name + 3438 g_objc_v1_prefix_class.size(); 3439 type = eSymbolTypeObjCClass; 3440 demangled_is_synthesized = true; 3441 } 3442 } 3443 } 3444 } 3445 } 3446 } break; 3447 } 3448 } 3449 3450 if (add_nlist) { 3451 uint64_t symbol_value = nlist.n_value; 3452 if (symbol_name_non_abi_mangled) { 3453 sym[sym_idx].GetMangled().SetMangledName( 3454 ConstString(symbol_name_non_abi_mangled)); 3455 sym[sym_idx].GetMangled().SetDemangledName( 3456 ConstString(symbol_name)); 3457 } else { 3458 bool symbol_name_is_mangled = false; 3459 3460 if (symbol_name && symbol_name[0] == '_') { 3461 symbol_name_is_mangled = symbol_name[1] == '_'; 3462 symbol_name++; // Skip the leading underscore 3463 } 3464 3465 if (symbol_name) { 3466 ConstString const_symbol_name(symbol_name); 3467 sym[sym_idx].GetMangled().SetValue( 3468 const_symbol_name, symbol_name_is_mangled); 3469 if (is_gsym && is_debug) { 3470 const char *gsym_name = 3471 sym[sym_idx] 3472 .GetMangled() 3473 .GetName(lldb::eLanguageTypeUnknown, 3474 Mangled::ePreferMangled) 3475 .GetCString(); 3476 if (gsym_name) 3477 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3478 } 3479 } 3480 } 3481 if (symbol_section) { 3482 const addr_t section_file_addr = 3483 symbol_section->GetFileAddress(); 3484 if (symbol_byte_size == 0 && 3485 function_starts_count > 0) { 3486 addr_t symbol_lookup_file_addr = nlist.n_value; 3487 // Do an exact address match for non-ARM addresses, 3488 // else get the closest since the symbol might be a 3489 // thumb symbol which has an address with bit zero 3490 // set 3491 FunctionStarts::Entry *func_start_entry = 3492 function_starts.FindEntry(symbol_lookup_file_addr, 3493 !is_arm); 3494 if (is_arm && func_start_entry) { 3495 // Verify that the function start address is the 3496 // symbol address (ARM) or the symbol address + 1 3497 // (thumb) 3498 if (func_start_entry->addr != 3499 symbol_lookup_file_addr && 3500 func_start_entry->addr != 3501 (symbol_lookup_file_addr + 1)) { 3502 // Not the right entry, NULL it out... 3503 func_start_entry = NULL; 3504 } 3505 } 3506 if (func_start_entry) { 3507 func_start_entry->data = true; 3508 3509 addr_t symbol_file_addr = func_start_entry->addr; 3510 uint32_t symbol_flags = 0; 3511 if (is_arm) { 3512 if (symbol_file_addr & 1) 3513 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3514 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3515 } 3516 3517 const FunctionStarts::Entry *next_func_start_entry = 3518 function_starts.FindNextEntry(func_start_entry); 3519 const addr_t section_end_file_addr = 3520 section_file_addr + 3521 symbol_section->GetByteSize(); 3522 if (next_func_start_entry) { 3523 addr_t next_symbol_file_addr = 3524 next_func_start_entry->addr; 3525 // Be sure the clear the Thumb address bit when 3526 // we calculate the size from the current and 3527 // next address 3528 if (is_arm) 3529 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3530 symbol_byte_size = std::min<lldb::addr_t>( 3531 next_symbol_file_addr - symbol_file_addr, 3532 section_end_file_addr - symbol_file_addr); 3533 } else { 3534 symbol_byte_size = 3535 section_end_file_addr - symbol_file_addr; 3536 } 3537 } 3538 } 3539 symbol_value -= section_file_addr; 3540 } 3541 3542 if (is_debug == false) { 3543 if (type == eSymbolTypeCode) { 3544 // See if we can find a N_FUN entry for any code 3545 // symbols. If we do find a match, and the name 3546 // matches, then we can merge the two into just the 3547 // function symbol to avoid duplicate entries in 3548 // the symbol table 3549 auto range = 3550 N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3551 if (range.first != range.second) { 3552 bool found_it = false; 3553 for (auto pos = range.first; pos != range.second; 3554 ++pos) { 3555 if (sym[sym_idx].GetMangled().GetName( 3556 lldb::eLanguageTypeUnknown, 3557 Mangled::ePreferMangled) == 3558 sym[pos->second].GetMangled().GetName( 3559 lldb::eLanguageTypeUnknown, 3560 Mangled::ePreferMangled)) { 3561 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3562 // We just need the flags from the linker 3563 // symbol, so put these flags 3564 // into the N_FUN flags to avoid duplicate 3565 // symbols in the symbol table 3566 sym[pos->second].SetExternal( 3567 sym[sym_idx].IsExternal()); 3568 sym[pos->second].SetFlags(nlist.n_type << 16 | 3569 nlist.n_desc); 3570 if (resolver_addresses.find(nlist.n_value) != 3571 resolver_addresses.end()) 3572 sym[pos->second].SetType(eSymbolTypeResolver); 3573 sym[sym_idx].Clear(); 3574 found_it = true; 3575 break; 3576 } 3577 } 3578 if (found_it) 3579 continue; 3580 } else { 3581 if (resolver_addresses.find(nlist.n_value) != 3582 resolver_addresses.end()) 3583 type = eSymbolTypeResolver; 3584 } 3585 } else if (type == eSymbolTypeData || 3586 type == eSymbolTypeObjCClass || 3587 type == eSymbolTypeObjCMetaClass || 3588 type == eSymbolTypeObjCIVar) { 3589 // See if we can find a N_STSYM entry for any data 3590 // symbols. If we do find a match, and the name 3591 // matches, then we can merge the two into just the 3592 // Static symbol to avoid duplicate entries in the 3593 // symbol table 3594 auto range = N_STSYM_addr_to_sym_idx.equal_range( 3595 nlist.n_value); 3596 if (range.first != range.second) { 3597 bool found_it = false; 3598 for (auto pos = range.first; pos != range.second; 3599 ++pos) { 3600 if (sym[sym_idx].GetMangled().GetName( 3601 lldb::eLanguageTypeUnknown, 3602 Mangled::ePreferMangled) == 3603 sym[pos->second].GetMangled().GetName( 3604 lldb::eLanguageTypeUnknown, 3605 Mangled::ePreferMangled)) { 3606 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3607 // We just need the flags from the linker 3608 // symbol, so put these flags 3609 // into the N_STSYM flags to avoid duplicate 3610 // symbols in the symbol table 3611 sym[pos->second].SetExternal( 3612 sym[sym_idx].IsExternal()); 3613 sym[pos->second].SetFlags(nlist.n_type << 16 | 3614 nlist.n_desc); 3615 sym[sym_idx].Clear(); 3616 found_it = true; 3617 break; 3618 } 3619 } 3620 if (found_it) 3621 continue; 3622 } else { 3623 const char *gsym_name = 3624 sym[sym_idx] 3625 .GetMangled() 3626 .GetName(lldb::eLanguageTypeUnknown, 3627 Mangled::ePreferMangled) 3628 .GetCString(); 3629 if (gsym_name) { 3630 // Combine N_GSYM stab entries with the non 3631 // stab symbol 3632 ConstNameToSymbolIndexMap::const_iterator pos = 3633 N_GSYM_name_to_sym_idx.find(gsym_name); 3634 if (pos != N_GSYM_name_to_sym_idx.end()) { 3635 const uint32_t GSYM_sym_idx = pos->second; 3636 m_nlist_idx_to_sym_idx[nlist_idx] = 3637 GSYM_sym_idx; 3638 // Copy the address, because often the N_GSYM 3639 // address has an invalid address of zero 3640 // when the global is a common symbol 3641 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3642 symbol_section); 3643 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3644 symbol_value); 3645 symbols_added.insert(sym[GSYM_sym_idx] 3646 .GetAddress() 3647 .GetFileAddress()); 3648 // We just need the flags from the linker 3649 // symbol, so put these flags 3650 // into the N_GSYM flags to avoid duplicate 3651 // symbols in the symbol table 3652 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 3653 nlist.n_desc); 3654 sym[sym_idx].Clear(); 3655 continue; 3656 } 3657 } 3658 } 3659 } 3660 } 3661 3662 sym[sym_idx].SetID(nlist_idx); 3663 sym[sym_idx].SetType(type); 3664 if (set_value) { 3665 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 3666 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3667 symbols_added.insert( 3668 sym[sym_idx].GetAddress().GetFileAddress()); 3669 } 3670 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 3671 3672 if (symbol_byte_size > 0) 3673 sym[sym_idx].SetByteSize(symbol_byte_size); 3674 3675 if (demangled_is_synthesized) 3676 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3677 ++sym_idx; 3678 } else { 3679 sym[sym_idx].Clear(); 3680 } 3681 } 3682 ///////////////////////////// 3683 } 3684 break; // No more entries to consider 3685 } 3686 } 3687 3688 for (const auto &pos : reexport_shlib_needs_fixup) { 3689 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3690 if (undef_pos != undefined_name_to_desc.end()) { 3691 const uint8_t dylib_ordinal = 3692 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3693 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3694 sym[pos.first].SetReExportedSymbolSharedLibrary( 3695 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3696 } 3697 } 3698 } 3699 } 3700 } 3701 } 3702 } 3703 3704 // Must reset this in case it was mutated above! 3705 nlist_data_offset = 0; 3706 #endif 3707 3708 if (nlist_data.GetByteSize() > 0) { 3709 3710 // If the sym array was not created while parsing the DSC unmapped 3711 // symbols, create it now. 3712 if (sym == nullptr) { 3713 sym = 3714 symtab->Resize(symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3715 num_syms = symtab->GetNumSymbols(); 3716 } 3717 3718 if (unmapped_local_symbols_found) { 3719 assert(m_dysymtab.ilocalsym == 0); 3720 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3721 nlist_idx = m_dysymtab.nlocalsym; 3722 } else { 3723 nlist_idx = 0; 3724 } 3725 3726 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 3727 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 3728 UndefinedNameToDescMap undefined_name_to_desc; 3729 SymbolIndexToName reexport_shlib_needs_fixup; 3730 3731 // Symtab parsing is a huge mess. Everything is entangled and the code 3732 // requires access to a ridiculous amount of variables. LLDB depends 3733 // heavily on the proper merging of symbols and to get that right we need 3734 // to make sure we have parsed all the debug symbols first. Therefore we 3735 // invoke the lambda twice, once to parse only the debug symbols and then 3736 // once more to parse the remaining symbols. 3737 auto ParseSymbolLambda = [&](struct nlist_64 &nlist, uint32_t nlist_idx, 3738 bool debug_only) { 3739 const bool is_debug = ((nlist.n_type & N_STAB) != 0); 3740 if (is_debug != debug_only) 3741 return true; 3742 3743 const char *symbol_name_non_abi_mangled = nullptr; 3744 const char *symbol_name = nullptr; 3745 3746 if (have_strtab_data) { 3747 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3748 3749 if (symbol_name == nullptr) { 3750 // No symbol should be NULL, even the symbols with no string values 3751 // should have an offset zero which points to an empty C-string 3752 Host::SystemLog(Host::eSystemLogError, 3753 "error: symbol[%u] has invalid string table offset " 3754 "0x%x in %s, ignoring symbol\n", 3755 nlist_idx, nlist.n_strx, 3756 module_sp->GetFileSpec().GetPath().c_str()); 3757 return true; 3758 } 3759 if (symbol_name[0] == '\0') 3760 symbol_name = nullptr; 3761 } else { 3762 const addr_t str_addr = strtab_addr + nlist.n_strx; 3763 Status str_error; 3764 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3765 str_error)) 3766 symbol_name = memory_symbol_name.c_str(); 3767 } 3768 3769 SymbolType type = eSymbolTypeInvalid; 3770 SectionSP symbol_section; 3771 lldb::addr_t symbol_byte_size = 0; 3772 bool add_nlist = true; 3773 bool is_gsym = false; 3774 bool demangled_is_synthesized = false; 3775 bool set_value = true; 3776 3777 assert(sym_idx < num_syms); 3778 sym[sym_idx].SetDebug(is_debug); 3779 3780 if (is_debug) { 3781 switch (nlist.n_type) { 3782 case N_GSYM: 3783 // global symbol: name,,NO_SECT,type,0 3784 // Sometimes the N_GSYM value contains the address. 3785 3786 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3787 // the ObjC data. They 3788 // have the same address, but we want to ensure that we always find 3789 // only the real symbol, 'cause we don't currently correctly 3790 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3791 // type. This is a temporary hack to make sure the ObjectiveC 3792 // symbols get treated correctly. To do this right, we should 3793 // coalesce all the GSYM & global symbols that have the same 3794 // address. 3795 is_gsym = true; 3796 sym[sym_idx].SetExternal(true); 3797 3798 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3799 llvm::StringRef symbol_name_ref(symbol_name); 3800 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3801 symbol_name_non_abi_mangled = symbol_name + 1; 3802 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3803 type = eSymbolTypeObjCClass; 3804 demangled_is_synthesized = true; 3805 3806 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 3807 symbol_name_non_abi_mangled = symbol_name + 1; 3808 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3809 type = eSymbolTypeObjCMetaClass; 3810 demangled_is_synthesized = true; 3811 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3812 symbol_name_non_abi_mangled = symbol_name + 1; 3813 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3814 type = eSymbolTypeObjCIVar; 3815 demangled_is_synthesized = true; 3816 } 3817 } else { 3818 if (nlist.n_value != 0) 3819 symbol_section = 3820 section_info.GetSection(nlist.n_sect, nlist.n_value); 3821 type = eSymbolTypeData; 3822 } 3823 break; 3824 3825 case N_FNAME: 3826 // procedure name (f77 kludge): name,,NO_SECT,0,0 3827 type = eSymbolTypeCompiler; 3828 break; 3829 3830 case N_FUN: 3831 // procedure: name,,n_sect,linenumber,address 3832 if (symbol_name) { 3833 type = eSymbolTypeCode; 3834 symbol_section = 3835 section_info.GetSection(nlist.n_sect, nlist.n_value); 3836 3837 N_FUN_addr_to_sym_idx.insert( 3838 std::make_pair(nlist.n_value, sym_idx)); 3839 // We use the current number of symbols in the symbol table in 3840 // lieu of using nlist_idx in case we ever start trimming entries 3841 // out 3842 N_FUN_indexes.push_back(sym_idx); 3843 } else { 3844 type = eSymbolTypeCompiler; 3845 3846 if (!N_FUN_indexes.empty()) { 3847 // Copy the size of the function into the original STAB entry 3848 // so we don't have to hunt for it later 3849 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3850 ->SetByteSize(nlist.n_value); 3851 N_FUN_indexes.pop_back(); 3852 // We don't really need the end function STAB as it contains 3853 // the size which we already placed with the original symbol, 3854 // so don't add it if we want a minimal symbol table 3855 add_nlist = false; 3856 } 3857 } 3858 break; 3859 3860 case N_STSYM: 3861 // static symbol: name,,n_sect,type,address 3862 N_STSYM_addr_to_sym_idx.insert( 3863 std::make_pair(nlist.n_value, sym_idx)); 3864 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3865 if (symbol_name && symbol_name[0]) { 3866 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3867 eSymbolTypeData); 3868 } 3869 break; 3870 3871 case N_LCSYM: 3872 // .lcomm symbol: name,,n_sect,type,address 3873 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3874 type = eSymbolTypeCommonBlock; 3875 break; 3876 3877 case N_BNSYM: 3878 // We use the current number of symbols in the symbol table in lieu 3879 // of using nlist_idx in case we ever start trimming entries out 3880 // Skip these if we want minimal symbol tables 3881 add_nlist = false; 3882 break; 3883 3884 case N_ENSYM: 3885 // Set the size of the N_BNSYM to the terminating index of this 3886 // N_ENSYM so that we can always skip the entire symbol if we need 3887 // to navigate more quickly at the source level when parsing STABS 3888 // Skip these if we want minimal symbol tables 3889 add_nlist = false; 3890 break; 3891 3892 case N_OPT: 3893 // emitted with gcc2_compiled and in gcc source 3894 type = eSymbolTypeCompiler; 3895 break; 3896 3897 case N_RSYM: 3898 // register sym: name,,NO_SECT,type,register 3899 type = eSymbolTypeVariable; 3900 break; 3901 3902 case N_SLINE: 3903 // src line: 0,,n_sect,linenumber,address 3904 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3905 type = eSymbolTypeLineEntry; 3906 break; 3907 3908 case N_SSYM: 3909 // structure elt: name,,NO_SECT,type,struct_offset 3910 type = eSymbolTypeVariableType; 3911 break; 3912 3913 case N_SO: 3914 // source file name 3915 type = eSymbolTypeSourceFile; 3916 if (symbol_name == nullptr) { 3917 add_nlist = false; 3918 if (N_SO_index != UINT32_MAX) { 3919 // Set the size of the N_SO to the terminating index of this 3920 // N_SO so that we can always skip the entire N_SO if we need 3921 // to navigate more quickly at the source level when parsing 3922 // STABS 3923 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3924 symbol_ptr->SetByteSize(sym_idx); 3925 symbol_ptr->SetSizeIsSibling(true); 3926 } 3927 N_NSYM_indexes.clear(); 3928 N_INCL_indexes.clear(); 3929 N_BRAC_indexes.clear(); 3930 N_COMM_indexes.clear(); 3931 N_FUN_indexes.clear(); 3932 N_SO_index = UINT32_MAX; 3933 } else { 3934 // We use the current number of symbols in the symbol table in 3935 // lieu of using nlist_idx in case we ever start trimming entries 3936 // out 3937 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3938 if (N_SO_has_full_path) { 3939 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3940 // We have two consecutive N_SO entries where the first 3941 // contains a directory and the second contains a full path. 3942 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), 3943 false); 3944 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3945 add_nlist = false; 3946 } else { 3947 // This is the first entry in a N_SO that contains a 3948 // directory or a full path to the source file 3949 N_SO_index = sym_idx; 3950 } 3951 } else if ((N_SO_index == sym_idx - 1) && 3952 ((sym_idx - 1) < num_syms)) { 3953 // This is usually the second N_SO entry that contains just the 3954 // filename, so here we combine it with the first one if we are 3955 // minimizing the symbol table 3956 const char *so_path = 3957 sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3958 if (so_path && so_path[0]) { 3959 std::string full_so_path(so_path); 3960 const size_t double_slash_pos = full_so_path.find("//"); 3961 if (double_slash_pos != std::string::npos) { 3962 // The linker has been generating bad N_SO entries with 3963 // doubled up paths in the format "%s%s" where the first 3964 // string in the DW_AT_comp_dir, and the second is the 3965 // directory for the source file so you end up with a path 3966 // that looks like "/tmp/src//tmp/src/" 3967 FileSpec so_dir(so_path); 3968 if (!FileSystem::Instance().Exists(so_dir)) { 3969 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 3970 FileSpec::Style::native); 3971 if (FileSystem::Instance().Exists(so_dir)) { 3972 // Trim off the incorrect path 3973 full_so_path.erase(0, double_slash_pos + 1); 3974 } 3975 } 3976 } 3977 if (*full_so_path.rbegin() != '/') 3978 full_so_path += '/'; 3979 full_so_path += symbol_name; 3980 sym[sym_idx - 1].GetMangled().SetValue( 3981 ConstString(full_so_path.c_str()), false); 3982 add_nlist = false; 3983 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3984 } 3985 } else { 3986 // This could be a relative path to a N_SO 3987 N_SO_index = sym_idx; 3988 } 3989 } 3990 break; 3991 3992 case N_OSO: 3993 // object file name: name,,0,0,st_mtime 3994 type = eSymbolTypeObjectFile; 3995 break; 3996 3997 case N_LSYM: 3998 // local sym: name,,NO_SECT,type,offset 3999 type = eSymbolTypeLocal; 4000 break; 4001 4002 // INCL scopes 4003 case N_BINCL: 4004 // include file beginning: name,,NO_SECT,0,sum We use the current 4005 // number of symbols in the symbol table in lieu of using nlist_idx 4006 // in case we ever start trimming entries out 4007 N_INCL_indexes.push_back(sym_idx); 4008 type = eSymbolTypeScopeBegin; 4009 break; 4010 4011 case N_EINCL: 4012 // include file end: name,,NO_SECT,0,0 4013 // Set the size of the N_BINCL to the terminating index of this 4014 // N_EINCL so that we can always skip the entire symbol if we need 4015 // to navigate more quickly at the source level when parsing STABS 4016 if (!N_INCL_indexes.empty()) { 4017 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4018 symbol_ptr->SetByteSize(sym_idx + 1); 4019 symbol_ptr->SetSizeIsSibling(true); 4020 N_INCL_indexes.pop_back(); 4021 } 4022 type = eSymbolTypeScopeEnd; 4023 break; 4024 4025 case N_SOL: 4026 // #included file name: name,,n_sect,0,address 4027 type = eSymbolTypeHeaderFile; 4028 4029 // We currently don't use the header files on darwin 4030 add_nlist = false; 4031 break; 4032 4033 case N_PARAMS: 4034 // compiler parameters: name,,NO_SECT,0,0 4035 type = eSymbolTypeCompiler; 4036 break; 4037 4038 case N_VERSION: 4039 // compiler version: name,,NO_SECT,0,0 4040 type = eSymbolTypeCompiler; 4041 break; 4042 4043 case N_OLEVEL: 4044 // compiler -O level: name,,NO_SECT,0,0 4045 type = eSymbolTypeCompiler; 4046 break; 4047 4048 case N_PSYM: 4049 // parameter: name,,NO_SECT,type,offset 4050 type = eSymbolTypeVariable; 4051 break; 4052 4053 case N_ENTRY: 4054 // alternate entry: name,,n_sect,linenumber,address 4055 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4056 type = eSymbolTypeLineEntry; 4057 break; 4058 4059 // Left and Right Braces 4060 case N_LBRAC: 4061 // left bracket: 0,,NO_SECT,nesting level,address We use the 4062 // current number of symbols in the symbol table in lieu of using 4063 // nlist_idx in case we ever start trimming entries out 4064 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4065 N_BRAC_indexes.push_back(sym_idx); 4066 type = eSymbolTypeScopeBegin; 4067 break; 4068 4069 case N_RBRAC: 4070 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4071 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4072 // can always skip the entire symbol if we need to navigate more 4073 // quickly at the source level when parsing STABS 4074 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4075 if (!N_BRAC_indexes.empty()) { 4076 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4077 symbol_ptr->SetByteSize(sym_idx + 1); 4078 symbol_ptr->SetSizeIsSibling(true); 4079 N_BRAC_indexes.pop_back(); 4080 } 4081 type = eSymbolTypeScopeEnd; 4082 break; 4083 4084 case N_EXCL: 4085 // deleted include file: name,,NO_SECT,0,sum 4086 type = eSymbolTypeHeaderFile; 4087 break; 4088 4089 // COMM scopes 4090 case N_BCOMM: 4091 // begin common: name,,NO_SECT,0,0 4092 // We use the current number of symbols in the symbol table in lieu 4093 // of using nlist_idx in case we ever start trimming entries out 4094 type = eSymbolTypeScopeBegin; 4095 N_COMM_indexes.push_back(sym_idx); 4096 break; 4097 4098 case N_ECOML: 4099 // end common (local name): 0,,n_sect,0,address 4100 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4101 LLVM_FALLTHROUGH; 4102 4103 case N_ECOMM: 4104 // end common: name,,n_sect,0,0 4105 // Set the size of the N_BCOMM to the terminating index of this 4106 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4107 // we need to navigate more quickly at the source level when 4108 // parsing STABS 4109 if (!N_COMM_indexes.empty()) { 4110 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4111 symbol_ptr->SetByteSize(sym_idx + 1); 4112 symbol_ptr->SetSizeIsSibling(true); 4113 N_COMM_indexes.pop_back(); 4114 } 4115 type = eSymbolTypeScopeEnd; 4116 break; 4117 4118 case N_LENG: 4119 // second stab entry with length information 4120 type = eSymbolTypeAdditional; 4121 break; 4122 4123 default: 4124 break; 4125 } 4126 } else { 4127 uint8_t n_type = N_TYPE & nlist.n_type; 4128 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4129 4130 switch (n_type) { 4131 case N_INDR: { 4132 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4133 if (reexport_name_cstr && reexport_name_cstr[0]) { 4134 type = eSymbolTypeReExported; 4135 ConstString reexport_name(reexport_name_cstr + 4136 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4137 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4138 set_value = false; 4139 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4140 indirect_symbol_names.insert( 4141 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4142 } else 4143 type = eSymbolTypeUndefined; 4144 } break; 4145 4146 case N_UNDF: 4147 if (symbol_name && symbol_name[0]) { 4148 ConstString undefined_name(symbol_name + 4149 ((symbol_name[0] == '_') ? 1 : 0)); 4150 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4151 } 4152 LLVM_FALLTHROUGH; 4153 4154 case N_PBUD: 4155 type = eSymbolTypeUndefined; 4156 break; 4157 4158 case N_ABS: 4159 type = eSymbolTypeAbsolute; 4160 break; 4161 4162 case N_SECT: { 4163 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4164 4165 if (!symbol_section) { 4166 // TODO: warn about this? 4167 add_nlist = false; 4168 break; 4169 } 4170 4171 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4172 type = eSymbolTypeException; 4173 } else { 4174 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4175 4176 switch (section_type) { 4177 case S_CSTRING_LITERALS: 4178 type = eSymbolTypeData; 4179 break; // section with only literal C strings 4180 case S_4BYTE_LITERALS: 4181 type = eSymbolTypeData; 4182 break; // section with only 4 byte literals 4183 case S_8BYTE_LITERALS: 4184 type = eSymbolTypeData; 4185 break; // section with only 8 byte literals 4186 case S_LITERAL_POINTERS: 4187 type = eSymbolTypeTrampoline; 4188 break; // section with only pointers to literals 4189 case S_NON_LAZY_SYMBOL_POINTERS: 4190 type = eSymbolTypeTrampoline; 4191 break; // section with only non-lazy symbol pointers 4192 case S_LAZY_SYMBOL_POINTERS: 4193 type = eSymbolTypeTrampoline; 4194 break; // section with only lazy symbol pointers 4195 case S_SYMBOL_STUBS: 4196 type = eSymbolTypeTrampoline; 4197 break; // section with only symbol stubs, byte size of stub in 4198 // the reserved2 field 4199 case S_MOD_INIT_FUNC_POINTERS: 4200 type = eSymbolTypeCode; 4201 break; // section with only function pointers for initialization 4202 case S_MOD_TERM_FUNC_POINTERS: 4203 type = eSymbolTypeCode; 4204 break; // section with only function pointers for termination 4205 case S_INTERPOSING: 4206 type = eSymbolTypeTrampoline; 4207 break; // section with only pairs of function pointers for 4208 // interposing 4209 case S_16BYTE_LITERALS: 4210 type = eSymbolTypeData; 4211 break; // section with only 16 byte literals 4212 case S_DTRACE_DOF: 4213 type = eSymbolTypeInstrumentation; 4214 break; 4215 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4216 type = eSymbolTypeTrampoline; 4217 break; 4218 default: 4219 switch (symbol_section->GetType()) { 4220 case lldb::eSectionTypeCode: 4221 type = eSymbolTypeCode; 4222 break; 4223 case eSectionTypeData: 4224 case eSectionTypeDataCString: // Inlined C string data 4225 case eSectionTypeDataCStringPointers: // Pointers to C string 4226 // data 4227 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4228 // the symbol table 4229 case eSectionTypeData4: 4230 case eSectionTypeData8: 4231 case eSectionTypeData16: 4232 type = eSymbolTypeData; 4233 break; 4234 default: 4235 break; 4236 } 4237 break; 4238 } 4239 4240 if (type == eSymbolTypeInvalid) { 4241 const char *symbol_sect_name = 4242 symbol_section->GetName().AsCString(); 4243 if (symbol_section->IsDescendant(text_section_sp.get())) { 4244 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4245 S_ATTR_SELF_MODIFYING_CODE | 4246 S_ATTR_SOME_INSTRUCTIONS)) 4247 type = eSymbolTypeData; 4248 else 4249 type = eSymbolTypeCode; 4250 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 4251 symbol_section->IsDescendant( 4252 data_dirty_section_sp.get()) || 4253 symbol_section->IsDescendant( 4254 data_const_section_sp.get())) { 4255 if (symbol_sect_name && 4256 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 4257 type = eSymbolTypeRuntime; 4258 4259 if (symbol_name) { 4260 llvm::StringRef symbol_name_ref(symbol_name); 4261 if (symbol_name_ref.startswith("_OBJC_")) { 4262 llvm::StringRef g_objc_v2_prefix_class( 4263 "_OBJC_CLASS_$_"); 4264 llvm::StringRef g_objc_v2_prefix_metaclass( 4265 "_OBJC_METACLASS_$_"); 4266 llvm::StringRef g_objc_v2_prefix_ivar( 4267 "_OBJC_IVAR_$_"); 4268 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 4269 symbol_name_non_abi_mangled = symbol_name + 1; 4270 symbol_name = 4271 symbol_name + g_objc_v2_prefix_class.size(); 4272 type = eSymbolTypeObjCClass; 4273 demangled_is_synthesized = true; 4274 } else if (symbol_name_ref.startswith( 4275 g_objc_v2_prefix_metaclass)) { 4276 symbol_name_non_abi_mangled = symbol_name + 1; 4277 symbol_name = 4278 symbol_name + g_objc_v2_prefix_metaclass.size(); 4279 type = eSymbolTypeObjCMetaClass; 4280 demangled_is_synthesized = true; 4281 } else if (symbol_name_ref.startswith( 4282 g_objc_v2_prefix_ivar)) { 4283 symbol_name_non_abi_mangled = symbol_name + 1; 4284 symbol_name = 4285 symbol_name + g_objc_v2_prefix_ivar.size(); 4286 type = eSymbolTypeObjCIVar; 4287 demangled_is_synthesized = true; 4288 } 4289 } 4290 } 4291 } else if (symbol_sect_name && 4292 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4293 symbol_sect_name) { 4294 type = eSymbolTypeException; 4295 } else { 4296 type = eSymbolTypeData; 4297 } 4298 } else if (symbol_sect_name && 4299 ::strstr(symbol_sect_name, "__IMPORT") == 4300 symbol_sect_name) { 4301 type = eSymbolTypeTrampoline; 4302 } else if (symbol_section->IsDescendant(objc_section_sp.get())) { 4303 type = eSymbolTypeRuntime; 4304 if (symbol_name && symbol_name[0] == '.') { 4305 llvm::StringRef symbol_name_ref(symbol_name); 4306 llvm::StringRef g_objc_v1_prefix_class( 4307 ".objc_class_name_"); 4308 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4309 symbol_name_non_abi_mangled = symbol_name; 4310 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4311 type = eSymbolTypeObjCClass; 4312 demangled_is_synthesized = true; 4313 } 4314 } 4315 } 4316 } 4317 } 4318 } break; 4319 } 4320 } 4321 4322 if (!add_nlist) { 4323 sym[sym_idx].Clear(); 4324 return true; 4325 } 4326 4327 uint64_t symbol_value = nlist.n_value; 4328 4329 if (symbol_name_non_abi_mangled) { 4330 sym[sym_idx].GetMangled().SetMangledName( 4331 ConstString(symbol_name_non_abi_mangled)); 4332 sym[sym_idx].GetMangled().SetDemangledName(ConstString(symbol_name)); 4333 } else { 4334 bool symbol_name_is_mangled = false; 4335 4336 if (symbol_name && symbol_name[0] == '_') { 4337 symbol_name_is_mangled = symbol_name[1] == '_'; 4338 symbol_name++; // Skip the leading underscore 4339 } 4340 4341 if (symbol_name) { 4342 ConstString const_symbol_name(symbol_name); 4343 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4344 symbol_name_is_mangled); 4345 } 4346 } 4347 4348 if (is_gsym) { 4349 const char *gsym_name = sym[sym_idx] 4350 .GetMangled() 4351 .GetName(Mangled::ePreferMangled) 4352 .GetCString(); 4353 if (gsym_name) 4354 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4355 } 4356 4357 if (symbol_section) { 4358 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4359 if (symbol_byte_size == 0 && function_starts_count > 0) { 4360 addr_t symbol_lookup_file_addr = nlist.n_value; 4361 // Do an exact address match for non-ARM addresses, else get the 4362 // closest since the symbol might be a thumb symbol which has an 4363 // address with bit zero set. 4364 FunctionStarts::Entry *func_start_entry = 4365 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4366 if (is_arm && func_start_entry) { 4367 // Verify that the function start address is the symbol address 4368 // (ARM) or the symbol address + 1 (thumb). 4369 if (func_start_entry->addr != symbol_lookup_file_addr && 4370 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4371 // Not the right entry, NULL it out... 4372 func_start_entry = nullptr; 4373 } 4374 } 4375 if (func_start_entry) { 4376 func_start_entry->data = true; 4377 4378 addr_t symbol_file_addr = func_start_entry->addr; 4379 if (is_arm) 4380 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4381 4382 const FunctionStarts::Entry *next_func_start_entry = 4383 function_starts.FindNextEntry(func_start_entry); 4384 const addr_t section_end_file_addr = 4385 section_file_addr + symbol_section->GetByteSize(); 4386 if (next_func_start_entry) { 4387 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4388 // Be sure the clear the Thumb address bit when we calculate the 4389 // size from the current and next address 4390 if (is_arm) 4391 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4392 symbol_byte_size = std::min<lldb::addr_t>( 4393 next_symbol_file_addr - symbol_file_addr, 4394 section_end_file_addr - symbol_file_addr); 4395 } else { 4396 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4397 } 4398 } 4399 } 4400 symbol_value -= section_file_addr; 4401 } 4402 4403 if (!is_debug) { 4404 if (type == eSymbolTypeCode) { 4405 // See if we can find a N_FUN entry for any code symbols. If we do 4406 // find a match, and the name matches, then we can merge the two into 4407 // just the function symbol to avoid duplicate entries in the symbol 4408 // table. 4409 std::pair<ValueToSymbolIndexMap::const_iterator, 4410 ValueToSymbolIndexMap::const_iterator> 4411 range; 4412 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4413 if (range.first != range.second) { 4414 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4415 pos != range.second; ++pos) { 4416 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4417 sym[pos->second].GetMangled().GetName( 4418 Mangled::ePreferMangled)) { 4419 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4420 // We just need the flags from the linker symbol, so put these 4421 // flags into the N_FUN flags to avoid duplicate symbols in the 4422 // symbol table. 4423 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4424 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4425 if (resolver_addresses.find(nlist.n_value) != 4426 resolver_addresses.end()) 4427 sym[pos->second].SetType(eSymbolTypeResolver); 4428 sym[sym_idx].Clear(); 4429 return true; 4430 } 4431 } 4432 } else { 4433 if (resolver_addresses.find(nlist.n_value) != 4434 resolver_addresses.end()) 4435 type = eSymbolTypeResolver; 4436 } 4437 } else if (type == eSymbolTypeData || type == eSymbolTypeObjCClass || 4438 type == eSymbolTypeObjCMetaClass || 4439 type == eSymbolTypeObjCIVar) { 4440 // See if we can find a N_STSYM entry for any data symbols. If we do 4441 // find a match, and the name matches, then we can merge the two into 4442 // just the Static symbol to avoid duplicate entries in the symbol 4443 // table. 4444 std::pair<ValueToSymbolIndexMap::const_iterator, 4445 ValueToSymbolIndexMap::const_iterator> 4446 range; 4447 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4448 if (range.first != range.second) { 4449 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4450 pos != range.second; ++pos) { 4451 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4452 sym[pos->second].GetMangled().GetName( 4453 Mangled::ePreferMangled)) { 4454 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4455 // We just need the flags from the linker symbol, so put these 4456 // flags into the N_STSYM flags to avoid duplicate symbols in 4457 // the symbol table. 4458 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4459 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4460 sym[sym_idx].Clear(); 4461 return true; 4462 } 4463 } 4464 } else { 4465 // Combine N_GSYM stab entries with the non stab symbol. 4466 const char *gsym_name = sym[sym_idx] 4467 .GetMangled() 4468 .GetName(Mangled::ePreferMangled) 4469 .GetCString(); 4470 if (gsym_name) { 4471 ConstNameToSymbolIndexMap::const_iterator pos = 4472 N_GSYM_name_to_sym_idx.find(gsym_name); 4473 if (pos != N_GSYM_name_to_sym_idx.end()) { 4474 const uint32_t GSYM_sym_idx = pos->second; 4475 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4476 // Copy the address, because often the N_GSYM address has an 4477 // invalid address of zero when the global is a common symbol. 4478 sym[GSYM_sym_idx].GetAddressRef().SetSection(symbol_section); 4479 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4480 symbols_added.insert( 4481 sym[GSYM_sym_idx].GetAddress().GetFileAddress()); 4482 // We just need the flags from the linker symbol, so put these 4483 // flags into the N_GSYM flags to avoid duplicate symbols in 4484 // the symbol table. 4485 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4486 sym[sym_idx].Clear(); 4487 return true; 4488 } 4489 } 4490 } 4491 } 4492 } 4493 4494 sym[sym_idx].SetID(nlist_idx); 4495 sym[sym_idx].SetType(type); 4496 if (set_value) { 4497 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4498 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4499 symbols_added.insert(sym[sym_idx].GetAddress().GetFileAddress()); 4500 } 4501 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4502 if (nlist.n_desc & N_WEAK_REF) 4503 sym[sym_idx].SetIsWeak(true); 4504 4505 if (symbol_byte_size > 0) 4506 sym[sym_idx].SetByteSize(symbol_byte_size); 4507 4508 if (demangled_is_synthesized) 4509 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4510 4511 ++sym_idx; 4512 return true; 4513 }; 4514 4515 // First parse all the nlists but don't process them yet. See the next 4516 // comment for an explanation why. 4517 std::vector<struct nlist_64> nlists; 4518 nlists.reserve(symtab_load_command.nsyms); 4519 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 4520 if (auto nlist = 4521 ParseNList(nlist_data, nlist_data_offset, nlist_byte_size)) 4522 nlists.push_back(*nlist); 4523 else 4524 break; 4525 } 4526 4527 // Now parse all the debug symbols. This is needed to merge non-debug 4528 // symbols in the next step. Non-debug symbols are always coalesced into 4529 // the debug symbol. Doing this in one step would mean that some symbols 4530 // won't be merged. 4531 nlist_idx = 0; 4532 for (auto &nlist : nlists) { 4533 if (!ParseSymbolLambda(nlist, nlist_idx++, DebugSymbols)) 4534 break; 4535 } 4536 4537 // Finally parse all the non debug symbols. 4538 nlist_idx = 0; 4539 for (auto &nlist : nlists) { 4540 if (!ParseSymbolLambda(nlist, nlist_idx++, NonDebugSymbols)) 4541 break; 4542 } 4543 4544 for (const auto &pos : reexport_shlib_needs_fixup) { 4545 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4546 if (undef_pos != undefined_name_to_desc.end()) { 4547 const uint8_t dylib_ordinal = 4548 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4549 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4550 sym[pos.first].SetReExportedSymbolSharedLibrary( 4551 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4552 } 4553 } 4554 } 4555 4556 // Count how many trie symbols we'll add to the symbol table 4557 int trie_symbol_table_augment_count = 0; 4558 for (auto &e : external_sym_trie_entries) { 4559 if (symbols_added.find(e.entry.address) == symbols_added.end()) 4560 trie_symbol_table_augment_count++; 4561 } 4562 4563 if (num_syms < sym_idx + trie_symbol_table_augment_count) { 4564 num_syms = sym_idx + trie_symbol_table_augment_count; 4565 sym = symtab->Resize(num_syms); 4566 } 4567 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4568 4569 // Add symbols from the trie to the symbol table. 4570 for (auto &e : external_sym_trie_entries) { 4571 if (symbols_added.find(e.entry.address) != symbols_added.end()) 4572 continue; 4573 4574 // Find the section that this trie address is in, use that to annotate 4575 // symbol type as we add the trie address and name to the symbol table. 4576 Address symbol_addr; 4577 if (module_sp->ResolveFileAddress(e.entry.address, symbol_addr)) { 4578 SectionSP symbol_section(symbol_addr.GetSection()); 4579 const char *symbol_name = e.entry.name.GetCString(); 4580 bool demangled_is_synthesized = false; 4581 SymbolType type = 4582 GetSymbolType(symbol_name, demangled_is_synthesized, text_section_sp, 4583 data_section_sp, data_dirty_section_sp, 4584 data_const_section_sp, symbol_section); 4585 4586 sym[sym_idx].SetType(type); 4587 if (symbol_section) { 4588 sym[sym_idx].SetID(synthetic_sym_id++); 4589 sym[sym_idx].GetMangled().SetMangledName(ConstString(symbol_name)); 4590 if (demangled_is_synthesized) 4591 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4592 sym[sym_idx].SetIsSynthetic(true); 4593 sym[sym_idx].SetExternal(true); 4594 sym[sym_idx].GetAddressRef() = symbol_addr; 4595 symbols_added.insert(symbol_addr.GetFileAddress()); 4596 if (e.entry.flags & TRIE_SYMBOL_IS_THUMB) 4597 sym[sym_idx].SetFlags(MACHO_NLIST_ARM_SYMBOL_IS_THUMB); 4598 ++sym_idx; 4599 } 4600 } 4601 } 4602 4603 if (function_starts_count > 0) { 4604 uint32_t num_synthetic_function_symbols = 0; 4605 for (i = 0; i < function_starts_count; ++i) { 4606 if (symbols_added.find(function_starts.GetEntryRef(i).addr) == 4607 symbols_added.end()) 4608 ++num_synthetic_function_symbols; 4609 } 4610 4611 if (num_synthetic_function_symbols > 0) { 4612 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4613 num_syms = sym_idx + num_synthetic_function_symbols; 4614 sym = symtab->Resize(num_syms); 4615 } 4616 for (i = 0; i < function_starts_count; ++i) { 4617 const FunctionStarts::Entry *func_start_entry = 4618 function_starts.GetEntryAtIndex(i); 4619 if (symbols_added.find(func_start_entry->addr) == symbols_added.end()) { 4620 addr_t symbol_file_addr = func_start_entry->addr; 4621 uint32_t symbol_flags = 0; 4622 if (func_start_entry->data) 4623 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4624 Address symbol_addr; 4625 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4626 SectionSP symbol_section(symbol_addr.GetSection()); 4627 uint32_t symbol_byte_size = 0; 4628 if (symbol_section) { 4629 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4630 const FunctionStarts::Entry *next_func_start_entry = 4631 function_starts.FindNextEntry(func_start_entry); 4632 const addr_t section_end_file_addr = 4633 section_file_addr + symbol_section->GetByteSize(); 4634 if (next_func_start_entry) { 4635 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4636 if (is_arm) 4637 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4638 symbol_byte_size = std::min<lldb::addr_t>( 4639 next_symbol_file_addr - symbol_file_addr, 4640 section_end_file_addr - symbol_file_addr); 4641 } else { 4642 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4643 } 4644 sym[sym_idx].SetID(synthetic_sym_id++); 4645 sym[sym_idx].GetMangled().SetDemangledName( 4646 GetNextSyntheticSymbolName()); 4647 sym[sym_idx].SetType(eSymbolTypeCode); 4648 sym[sym_idx].SetIsSynthetic(true); 4649 sym[sym_idx].GetAddressRef() = symbol_addr; 4650 symbols_added.insert(symbol_addr.GetFileAddress()); 4651 if (symbol_flags) 4652 sym[sym_idx].SetFlags(symbol_flags); 4653 if (symbol_byte_size) 4654 sym[sym_idx].SetByteSize(symbol_byte_size); 4655 ++sym_idx; 4656 } 4657 } 4658 } 4659 } 4660 } 4661 } 4662 4663 // Trim our symbols down to just what we ended up with after removing any 4664 // symbols. 4665 if (sym_idx < num_syms) { 4666 num_syms = sym_idx; 4667 sym = symtab->Resize(num_syms); 4668 } 4669 4670 // Now synthesize indirect symbols 4671 if (m_dysymtab.nindirectsyms != 0) { 4672 if (indirect_symbol_index_data.GetByteSize()) { 4673 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4674 m_nlist_idx_to_sym_idx.end(); 4675 4676 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4677 ++sect_idx) { 4678 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4679 S_SYMBOL_STUBS) { 4680 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4681 if (symbol_stub_byte_size == 0) 4682 continue; 4683 4684 const uint32_t num_symbol_stubs = 4685 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4686 4687 if (num_symbol_stubs == 0) 4688 continue; 4689 4690 const uint32_t symbol_stub_index_offset = 4691 m_mach_sections[sect_idx].reserved1; 4692 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) { 4693 const uint32_t symbol_stub_index = 4694 symbol_stub_index_offset + stub_idx; 4695 const lldb::addr_t symbol_stub_addr = 4696 m_mach_sections[sect_idx].addr + 4697 (stub_idx * symbol_stub_byte_size); 4698 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4699 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4700 symbol_stub_offset, 4)) { 4701 const uint32_t stub_sym_id = 4702 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4703 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4704 continue; 4705 4706 NListIndexToSymbolIndexMap::const_iterator index_pos = 4707 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4708 Symbol *stub_symbol = nullptr; 4709 if (index_pos != end_index_pos) { 4710 // We have a remapping from the original nlist index to a 4711 // current symbol index, so just look this up by index 4712 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4713 } else { 4714 // We need to lookup a symbol using the original nlist symbol 4715 // index since this index is coming from the S_SYMBOL_STUBS 4716 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4717 } 4718 4719 if (stub_symbol) { 4720 Address so_addr(symbol_stub_addr, section_list); 4721 4722 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4723 // Change the external symbol into a trampoline that makes 4724 // sense These symbols were N_UNDF N_EXT, and are useless 4725 // to us, so we can re-use them so we don't have to make up 4726 // a synthetic symbol for no good reason. 4727 if (resolver_addresses.find(symbol_stub_addr) == 4728 resolver_addresses.end()) 4729 stub_symbol->SetType(eSymbolTypeTrampoline); 4730 else 4731 stub_symbol->SetType(eSymbolTypeResolver); 4732 stub_symbol->SetExternal(false); 4733 stub_symbol->GetAddressRef() = so_addr; 4734 stub_symbol->SetByteSize(symbol_stub_byte_size); 4735 } else { 4736 // Make a synthetic symbol to describe the trampoline stub 4737 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4738 if (sym_idx >= num_syms) { 4739 sym = symtab->Resize(++num_syms); 4740 stub_symbol = nullptr; // this pointer no longer valid 4741 } 4742 sym[sym_idx].SetID(synthetic_sym_id++); 4743 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4744 if (resolver_addresses.find(symbol_stub_addr) == 4745 resolver_addresses.end()) 4746 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4747 else 4748 sym[sym_idx].SetType(eSymbolTypeResolver); 4749 sym[sym_idx].SetIsSynthetic(true); 4750 sym[sym_idx].GetAddressRef() = so_addr; 4751 symbols_added.insert(so_addr.GetFileAddress()); 4752 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4753 ++sym_idx; 4754 } 4755 } else { 4756 if (log) 4757 log->Warning("symbol stub referencing symbol table symbol " 4758 "%u that isn't in our minimal symbol table, " 4759 "fix this!!!", 4760 stub_sym_id); 4761 } 4762 } 4763 } 4764 } 4765 } 4766 } 4767 } 4768 4769 if (!reexport_trie_entries.empty()) { 4770 for (const auto &e : reexport_trie_entries) { 4771 if (e.entry.import_name) { 4772 // Only add indirect symbols from the Trie entries if we didn't have 4773 // a N_INDR nlist entry for this already 4774 if (indirect_symbol_names.find(e.entry.name) == 4775 indirect_symbol_names.end()) { 4776 // Make a synthetic symbol to describe re-exported symbol. 4777 if (sym_idx >= num_syms) 4778 sym = symtab->Resize(++num_syms); 4779 sym[sym_idx].SetID(synthetic_sym_id++); 4780 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4781 sym[sym_idx].SetType(eSymbolTypeReExported); 4782 sym[sym_idx].SetIsSynthetic(true); 4783 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4784 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4785 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4786 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4787 } 4788 ++sym_idx; 4789 } 4790 } 4791 } 4792 } 4793 4794 // StreamFile s(stdout, false); 4795 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4796 // symtab->Dump(&s, NULL, eSortOrderNone); 4797 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4798 // sizes 4799 symtab->CalculateSymbolSizes(); 4800 4801 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4802 // symtab->Dump(&s, NULL, eSortOrderNone); 4803 4804 return symtab->GetNumSymbols(); 4805 } 4806 4807 void ObjectFileMachO::Dump(Stream *s) { 4808 ModuleSP module_sp(GetModule()); 4809 if (module_sp) { 4810 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4811 s->Printf("%p: ", static_cast<void *>(this)); 4812 s->Indent(); 4813 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4814 s->PutCString("ObjectFileMachO64"); 4815 else 4816 s->PutCString("ObjectFileMachO32"); 4817 4818 *s << ", file = '" << m_file; 4819 ModuleSpecList all_specs; 4820 ModuleSpec base_spec; 4821 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 4822 base_spec, all_specs); 4823 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 4824 *s << "', triple"; 4825 if (e) 4826 s->Printf("[%d]", i); 4827 *s << " = "; 4828 *s << all_specs.GetModuleSpecRefAtIndex(i) 4829 .GetArchitecture() 4830 .GetTriple() 4831 .getTriple(); 4832 } 4833 *s << "\n"; 4834 SectionList *sections = GetSectionList(); 4835 if (sections) 4836 sections->Dump(s->AsRawOstream(), s->GetIndentLevel(), nullptr, true, 4837 UINT32_MAX); 4838 4839 if (m_symtab_up) 4840 m_symtab_up->Dump(s, nullptr, eSortOrderNone); 4841 } 4842 } 4843 4844 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4845 const lldb_private::DataExtractor &data, 4846 lldb::offset_t lc_offset) { 4847 uint32_t i; 4848 struct uuid_command load_cmd; 4849 4850 lldb::offset_t offset = lc_offset; 4851 for (i = 0; i < header.ncmds; ++i) { 4852 const lldb::offset_t cmd_offset = offset; 4853 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4854 break; 4855 4856 if (load_cmd.cmd == LC_UUID) { 4857 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4858 4859 if (uuid_bytes) { 4860 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4861 // We pretend these object files have no UUID to prevent crashing. 4862 4863 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4864 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4865 0xbb, 0x14, 0xf0, 0x0d}; 4866 4867 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4868 return UUID(); 4869 4870 return UUID::fromOptionalData(uuid_bytes, 16); 4871 } 4872 return UUID(); 4873 } 4874 offset = cmd_offset + load_cmd.cmdsize; 4875 } 4876 return UUID(); 4877 } 4878 4879 static llvm::StringRef GetOSName(uint32_t cmd) { 4880 switch (cmd) { 4881 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4882 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4883 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4884 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4885 case llvm::MachO::LC_VERSION_MIN_TVOS: 4886 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4887 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4888 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4889 default: 4890 llvm_unreachable("unexpected LC_VERSION load command"); 4891 } 4892 } 4893 4894 namespace { 4895 struct OSEnv { 4896 llvm::StringRef os_type; 4897 llvm::StringRef environment; 4898 OSEnv(uint32_t cmd) { 4899 switch (cmd) { 4900 case llvm::MachO::PLATFORM_MACOS: 4901 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4902 return; 4903 case llvm::MachO::PLATFORM_IOS: 4904 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4905 return; 4906 case llvm::MachO::PLATFORM_TVOS: 4907 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4908 return; 4909 case llvm::MachO::PLATFORM_WATCHOS: 4910 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4911 return; 4912 // NEED_BRIDGEOS_TRIPLE case llvm::MachO::PLATFORM_BRIDGEOS: 4913 // NEED_BRIDGEOS_TRIPLE os_type = 4914 // llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4915 // NEED_BRIDGEOS_TRIPLE return; 4916 case llvm::MachO::PLATFORM_MACCATALYST: 4917 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4918 environment = llvm::Triple::getEnvironmentTypeName(llvm::Triple::MacABI); 4919 return; 4920 case llvm::MachO::PLATFORM_IOSSIMULATOR: 4921 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4922 environment = 4923 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4924 return; 4925 case llvm::MachO::PLATFORM_TVOSSIMULATOR: 4926 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4927 environment = 4928 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4929 return; 4930 case llvm::MachO::PLATFORM_WATCHOSSIMULATOR: 4931 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4932 environment = 4933 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4934 return; 4935 default: { 4936 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 4937 LIBLLDB_LOG_PROCESS)); 4938 LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION"); 4939 } 4940 } 4941 } 4942 }; 4943 4944 struct MinOS { 4945 uint32_t major_version, minor_version, patch_version; 4946 MinOS(uint32_t version) 4947 : major_version(version >> 16), minor_version((version >> 8) & 0xffu), 4948 patch_version(version & 0xffu) {} 4949 }; 4950 } // namespace 4951 4952 void ObjectFileMachO::GetAllArchSpecs(const llvm::MachO::mach_header &header, 4953 const lldb_private::DataExtractor &data, 4954 lldb::offset_t lc_offset, 4955 ModuleSpec &base_spec, 4956 lldb_private::ModuleSpecList &all_specs) { 4957 auto &base_arch = base_spec.GetArchitecture(); 4958 base_arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4959 if (!base_arch.IsValid()) 4960 return; 4961 4962 bool found_any = false; 4963 auto add_triple = [&](const llvm::Triple &triple) { 4964 auto spec = base_spec; 4965 spec.GetArchitecture().GetTriple() = triple; 4966 if (spec.GetArchitecture().IsValid()) { 4967 spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset); 4968 all_specs.Append(spec); 4969 found_any = true; 4970 } 4971 }; 4972 4973 // Set OS to an unspecified unknown or a "*" so it can match any OS 4974 llvm::Triple base_triple = base_arch.GetTriple(); 4975 base_triple.setOS(llvm::Triple::UnknownOS); 4976 base_triple.setOSName(llvm::StringRef()); 4977 4978 if (header.filetype == MH_PRELOAD) { 4979 if (header.cputype == CPU_TYPE_ARM) { 4980 // If this is a 32-bit arm binary, and it's a standalone binary, force 4981 // the Vendor to Apple so we don't accidentally pick up the generic 4982 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4983 // frame pointer register; most other armv7 ABIs use a combination of 4984 // r7 and r11. 4985 base_triple.setVendor(llvm::Triple::Apple); 4986 } else { 4987 // Set vendor to an unspecified unknown or a "*" so it can match any 4988 // vendor This is required for correct behavior of EFI debugging on 4989 // x86_64 4990 base_triple.setVendor(llvm::Triple::UnknownVendor); 4991 base_triple.setVendorName(llvm::StringRef()); 4992 } 4993 return add_triple(base_triple); 4994 } 4995 4996 struct load_command load_cmd; 4997 4998 // See if there is an LC_VERSION_MIN_* load command that can give 4999 // us the OS type. 5000 lldb::offset_t offset = lc_offset; 5001 for (uint32_t i = 0; i < header.ncmds; ++i) { 5002 const lldb::offset_t cmd_offset = offset; 5003 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 5004 break; 5005 5006 struct version_min_command version_min; 5007 switch (load_cmd.cmd) { 5008 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 5009 case llvm::MachO::LC_VERSION_MIN_MACOSX: 5010 case llvm::MachO::LC_VERSION_MIN_TVOS: 5011 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 5012 if (load_cmd.cmdsize != sizeof(version_min)) 5013 break; 5014 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 5015 data.GetByteOrder(), &version_min) == 0) 5016 break; 5017 MinOS min_os(version_min.version); 5018 llvm::SmallString<32> os_name; 5019 llvm::raw_svector_ostream os(os_name); 5020 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 5021 << min_os.minor_version << '.' << min_os.patch_version; 5022 5023 auto triple = base_triple; 5024 triple.setOSName(os.str()); 5025 os_name.clear(); 5026 add_triple(triple); 5027 break; 5028 } 5029 default: 5030 break; 5031 } 5032 5033 offset = cmd_offset + load_cmd.cmdsize; 5034 } 5035 5036 // See if there are LC_BUILD_VERSION load commands that can give 5037 // us the OS type. 5038 offset = lc_offset; 5039 for (uint32_t i = 0; i < header.ncmds; ++i) { 5040 const lldb::offset_t cmd_offset = offset; 5041 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 5042 break; 5043 5044 do { 5045 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 5046 struct build_version_command build_version; 5047 if (load_cmd.cmdsize < sizeof(build_version)) { 5048 // Malformed load command. 5049 break; 5050 } 5051 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 5052 data.GetByteOrder(), &build_version) == 0) 5053 break; 5054 MinOS min_os(build_version.minos); 5055 OSEnv os_env(build_version.platform); 5056 llvm::SmallString<16> os_name; 5057 llvm::raw_svector_ostream os(os_name); 5058 os << os_env.os_type << min_os.major_version << '.' 5059 << min_os.minor_version << '.' << min_os.patch_version; 5060 auto triple = base_triple; 5061 triple.setOSName(os.str()); 5062 os_name.clear(); 5063 if (!os_env.environment.empty()) 5064 triple.setEnvironmentName(os_env.environment); 5065 add_triple(triple); 5066 } 5067 } while (0); 5068 offset = cmd_offset + load_cmd.cmdsize; 5069 } 5070 5071 if (!found_any) { 5072 if (header.filetype == MH_KEXT_BUNDLE) { 5073 base_triple.setVendor(llvm::Triple::Apple); 5074 add_triple(base_triple); 5075 } else { 5076 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 5077 // so lets not say our Vendor is Apple, leave it as an unspecified 5078 // unknown. 5079 base_triple.setVendor(llvm::Triple::UnknownVendor); 5080 base_triple.setVendorName(llvm::StringRef()); 5081 add_triple(base_triple); 5082 } 5083 } 5084 } 5085 5086 ArchSpec ObjectFileMachO::GetArchitecture( 5087 ModuleSP module_sp, const llvm::MachO::mach_header &header, 5088 const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) { 5089 ModuleSpecList all_specs; 5090 ModuleSpec base_spec; 5091 GetAllArchSpecs(header, data, MachHeaderSizeFromMagic(header.magic), 5092 base_spec, all_specs); 5093 5094 // If the object file offers multiple alternative load commands, 5095 // pick the one that matches the module. 5096 if (module_sp) { 5097 const ArchSpec &module_arch = module_sp->GetArchitecture(); 5098 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 5099 ArchSpec mach_arch = 5100 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 5101 if (module_arch.IsCompatibleMatch(mach_arch)) 5102 return mach_arch; 5103 } 5104 } 5105 5106 // Return the first arch we found. 5107 if (all_specs.GetSize() == 0) 5108 return {}; 5109 return all_specs.GetModuleSpecRefAtIndex(0).GetArchitecture(); 5110 } 5111 5112 UUID ObjectFileMachO::GetUUID() { 5113 ModuleSP module_sp(GetModule()); 5114 if (module_sp) { 5115 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5116 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5117 return GetUUID(m_header, m_data, offset); 5118 } 5119 return UUID(); 5120 } 5121 5122 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5123 uint32_t count = 0; 5124 ModuleSP module_sp(GetModule()); 5125 if (module_sp) { 5126 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5127 struct load_command load_cmd; 5128 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5129 std::vector<std::string> rpath_paths; 5130 std::vector<std::string> rpath_relative_paths; 5131 std::vector<std::string> at_exec_relative_paths; 5132 uint32_t i; 5133 for (i = 0; i < m_header.ncmds; ++i) { 5134 const uint32_t cmd_offset = offset; 5135 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5136 break; 5137 5138 switch (load_cmd.cmd) { 5139 case LC_RPATH: 5140 case LC_LOAD_DYLIB: 5141 case LC_LOAD_WEAK_DYLIB: 5142 case LC_REEXPORT_DYLIB: 5143 case LC_LOAD_DYLINKER: 5144 case LC_LOADFVMLIB: 5145 case LC_LOAD_UPWARD_DYLIB: { 5146 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5147 const char *path = m_data.PeekCStr(name_offset); 5148 if (path) { 5149 if (load_cmd.cmd == LC_RPATH) 5150 rpath_paths.push_back(path); 5151 else { 5152 if (path[0] == '@') { 5153 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5154 rpath_relative_paths.push_back(path + strlen("@rpath")); 5155 else if (strncmp(path, "@executable_path", 5156 strlen("@executable_path")) == 0) 5157 at_exec_relative_paths.push_back(path + 5158 strlen("@executable_path")); 5159 } else { 5160 FileSpec file_spec(path); 5161 if (files.AppendIfUnique(file_spec)) 5162 count++; 5163 } 5164 } 5165 } 5166 } break; 5167 5168 default: 5169 break; 5170 } 5171 offset = cmd_offset + load_cmd.cmdsize; 5172 } 5173 5174 FileSpec this_file_spec(m_file); 5175 FileSystem::Instance().Resolve(this_file_spec); 5176 5177 if (!rpath_paths.empty()) { 5178 // Fixup all LC_RPATH values to be absolute paths 5179 std::string loader_path("@loader_path"); 5180 std::string executable_path("@executable_path"); 5181 for (auto &rpath : rpath_paths) { 5182 if (llvm::StringRef(rpath).startswith(loader_path)) { 5183 rpath.erase(0, loader_path.size()); 5184 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5185 } else if (llvm::StringRef(rpath).startswith(executable_path)) { 5186 rpath.erase(0, executable_path.size()); 5187 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5188 } 5189 } 5190 5191 for (const auto &rpath_relative_path : rpath_relative_paths) { 5192 for (const auto &rpath : rpath_paths) { 5193 std::string path = rpath; 5194 path += rpath_relative_path; 5195 // It is OK to resolve this path because we must find a file on disk 5196 // for us to accept it anyway if it is rpath relative. 5197 FileSpec file_spec(path); 5198 FileSystem::Instance().Resolve(file_spec); 5199 if (FileSystem::Instance().Exists(file_spec) && 5200 files.AppendIfUnique(file_spec)) { 5201 count++; 5202 break; 5203 } 5204 } 5205 } 5206 } 5207 5208 // We may have @executable_paths but no RPATHS. Figure those out here. 5209 // Only do this if this object file is the executable. We have no way to 5210 // get back to the actual executable otherwise, so we won't get the right 5211 // path. 5212 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5213 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5214 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5215 FileSpec file_spec = 5216 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5217 if (FileSystem::Instance().Exists(file_spec) && 5218 files.AppendIfUnique(file_spec)) 5219 count++; 5220 } 5221 } 5222 } 5223 return count; 5224 } 5225 5226 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5227 // If the object file is not an executable it can't hold the entry point. 5228 // m_entry_point_address is initialized to an invalid address, so we can just 5229 // return that. If m_entry_point_address is valid it means we've found it 5230 // already, so return the cached value. 5231 5232 if ((!IsExecutable() && !IsDynamicLoader()) || 5233 m_entry_point_address.IsValid()) { 5234 return m_entry_point_address; 5235 } 5236 5237 // Otherwise, look for the UnixThread or Thread command. The data for the 5238 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5239 // 5240 // uint32_t flavor - this is the flavor argument you would pass to 5241 // thread_get_state 5242 // uint32_t count - this is the count of longs in the thread state data 5243 // struct XXX_thread_state state - this is the structure from 5244 // <machine/thread_status.h> corresponding to the flavor. 5245 // <repeat this trio> 5246 // 5247 // So we just keep reading the various register flavors till we find the GPR 5248 // one, then read the PC out of there. 5249 // FIXME: We will need to have a "RegisterContext data provider" class at some 5250 // point that can get all the registers 5251 // out of data in this form & attach them to a given thread. That should 5252 // underlie the MacOS X User process plugin, and we'll also need it for the 5253 // MacOS X Core File process plugin. When we have that we can also use it 5254 // here. 5255 // 5256 // For now we hard-code the offsets and flavors we need: 5257 // 5258 // 5259 5260 ModuleSP module_sp(GetModule()); 5261 if (module_sp) { 5262 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5263 struct load_command load_cmd; 5264 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5265 uint32_t i; 5266 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5267 bool done = false; 5268 5269 for (i = 0; i < m_header.ncmds; ++i) { 5270 const lldb::offset_t cmd_offset = offset; 5271 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5272 break; 5273 5274 switch (load_cmd.cmd) { 5275 case LC_UNIXTHREAD: 5276 case LC_THREAD: { 5277 while (offset < cmd_offset + load_cmd.cmdsize) { 5278 uint32_t flavor = m_data.GetU32(&offset); 5279 uint32_t count = m_data.GetU32(&offset); 5280 if (count == 0) { 5281 // We've gotten off somehow, log and exit; 5282 return m_entry_point_address; 5283 } 5284 5285 switch (m_header.cputype) { 5286 case llvm::MachO::CPU_TYPE_ARM: 5287 if (flavor == 1 || 5288 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 5289 // from mach/arm/thread_status.h 5290 { 5291 offset += 60; // This is the offset of pc in the GPR thread state 5292 // data structure. 5293 start_address = m_data.GetU32(&offset); 5294 done = true; 5295 } 5296 break; 5297 case llvm::MachO::CPU_TYPE_ARM64: 5298 case llvm::MachO::CPU_TYPE_ARM64_32: 5299 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5300 { 5301 offset += 256; // This is the offset of pc in the GPR thread state 5302 // data structure. 5303 start_address = m_data.GetU64(&offset); 5304 done = true; 5305 } 5306 break; 5307 case llvm::MachO::CPU_TYPE_I386: 5308 if (flavor == 5309 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5310 { 5311 offset += 40; // This is the offset of eip 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_X86_64: 5318 if (flavor == 5319 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5320 { 5321 offset += 16 * 8; // This is the offset of rip in the GPR thread 5322 // state data structure. 5323 start_address = m_data.GetU64(&offset); 5324 done = true; 5325 } 5326 break; 5327 default: 5328 return m_entry_point_address; 5329 } 5330 // Haven't found the GPR flavor yet, skip over the data for this 5331 // flavor: 5332 if (done) 5333 break; 5334 offset += count * 4; 5335 } 5336 } break; 5337 case LC_MAIN: { 5338 ConstString text_segment_name("__TEXT"); 5339 uint64_t entryoffset = m_data.GetU64(&offset); 5340 SectionSP text_segment_sp = 5341 GetSectionList()->FindSectionByName(text_segment_name); 5342 if (text_segment_sp) { 5343 done = true; 5344 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5345 } 5346 } break; 5347 5348 default: 5349 break; 5350 } 5351 if (done) 5352 break; 5353 5354 // Go to the next load command: 5355 offset = cmd_offset + load_cmd.cmdsize; 5356 } 5357 5358 if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) { 5359 if (GetSymtab()) { 5360 Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType( 5361 ConstString("_dyld_start"), SymbolType::eSymbolTypeCode, 5362 Symtab::eDebugAny, Symtab::eVisibilityAny); 5363 if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) { 5364 start_address = dyld_start_sym->GetAddress().GetFileAddress(); 5365 } 5366 } 5367 } 5368 5369 if (start_address != LLDB_INVALID_ADDRESS) { 5370 // We got the start address from the load commands, so now resolve that 5371 // address in the sections of this ObjectFile: 5372 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5373 start_address, GetSectionList())) { 5374 m_entry_point_address.Clear(); 5375 } 5376 } else { 5377 // We couldn't read the UnixThread load command - maybe it wasn't there. 5378 // As a fallback look for the "start" symbol in the main executable. 5379 5380 ModuleSP module_sp(GetModule()); 5381 5382 if (module_sp) { 5383 SymbolContextList contexts; 5384 SymbolContext context; 5385 module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5386 eSymbolTypeCode, contexts); 5387 if (contexts.GetSize()) { 5388 if (contexts.GetContextAtIndex(0, context)) 5389 m_entry_point_address = context.symbol->GetAddress(); 5390 } 5391 } 5392 } 5393 } 5394 5395 return m_entry_point_address; 5396 } 5397 5398 lldb_private::Address ObjectFileMachO::GetBaseAddress() { 5399 lldb_private::Address header_addr; 5400 SectionList *section_list = GetSectionList(); 5401 if (section_list) { 5402 SectionSP text_segment_sp( 5403 section_list->FindSectionByName(GetSegmentNameTEXT())); 5404 if (text_segment_sp) { 5405 header_addr.SetSection(text_segment_sp); 5406 header_addr.SetOffset(0); 5407 } 5408 } 5409 return header_addr; 5410 } 5411 5412 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5413 ModuleSP module_sp(GetModule()); 5414 if (module_sp) { 5415 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5416 if (!m_thread_context_offsets_valid) { 5417 m_thread_context_offsets_valid = true; 5418 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5419 FileRangeArray::Entry file_range; 5420 thread_command thread_cmd; 5421 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5422 const uint32_t cmd_offset = offset; 5423 if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr) 5424 break; 5425 5426 if (thread_cmd.cmd == LC_THREAD) { 5427 file_range.SetRangeBase(offset); 5428 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5429 m_thread_context_offsets.Append(file_range); 5430 } 5431 offset = cmd_offset + thread_cmd.cmdsize; 5432 } 5433 } 5434 } 5435 return m_thread_context_offsets.GetSize(); 5436 } 5437 5438 std::string ObjectFileMachO::GetIdentifierString() { 5439 std::string result; 5440 ModuleSP module_sp(GetModule()); 5441 if (module_sp) { 5442 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5443 5444 // First, look over the load commands for an LC_NOTE load command with 5445 // data_owner string "kern ver str" & use that if found. 5446 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5447 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5448 const uint32_t cmd_offset = offset; 5449 load_command lc; 5450 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5451 break; 5452 if (lc.cmd == LC_NOTE) { 5453 char data_owner[17]; 5454 m_data.CopyData(offset, 16, data_owner); 5455 data_owner[16] = '\0'; 5456 offset += 16; 5457 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5458 uint64_t size = m_data.GetU64_unchecked(&offset); 5459 5460 // "kern ver str" has a uint32_t version and then a nul terminated 5461 // c-string. 5462 if (strcmp("kern ver str", data_owner) == 0) { 5463 offset = fileoff; 5464 uint32_t version; 5465 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5466 if (version == 1) { 5467 uint32_t strsize = size - sizeof(uint32_t); 5468 char *buf = (char *)malloc(strsize); 5469 if (buf) { 5470 m_data.CopyData(offset, strsize, buf); 5471 buf[strsize - 1] = '\0'; 5472 result = buf; 5473 if (buf) 5474 free(buf); 5475 return result; 5476 } 5477 } 5478 } 5479 } 5480 } 5481 offset = cmd_offset + lc.cmdsize; 5482 } 5483 5484 // Second, make a pass over the load commands looking for an obsolete 5485 // LC_IDENT load command. 5486 offset = MachHeaderSizeFromMagic(m_header.magic); 5487 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5488 const uint32_t cmd_offset = offset; 5489 struct ident_command ident_command; 5490 if (m_data.GetU32(&offset, &ident_command, 2) == nullptr) 5491 break; 5492 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5493 char *buf = (char *)malloc(ident_command.cmdsize); 5494 if (buf != nullptr && m_data.CopyData(offset, ident_command.cmdsize, 5495 buf) == ident_command.cmdsize) { 5496 buf[ident_command.cmdsize - 1] = '\0'; 5497 result = buf; 5498 } 5499 if (buf) 5500 free(buf); 5501 } 5502 offset = cmd_offset + ident_command.cmdsize; 5503 } 5504 } 5505 return result; 5506 } 5507 5508 bool ObjectFileMachO::GetCorefileMainBinaryInfo(addr_t &address, UUID &uuid) { 5509 address = LLDB_INVALID_ADDRESS; 5510 uuid.Clear(); 5511 ModuleSP module_sp(GetModule()); 5512 if (module_sp) { 5513 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5514 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5515 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5516 const uint32_t cmd_offset = offset; 5517 load_command lc; 5518 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5519 break; 5520 if (lc.cmd == LC_NOTE) { 5521 char data_owner[17]; 5522 memset(data_owner, 0, sizeof(data_owner)); 5523 m_data.CopyData(offset, 16, data_owner); 5524 offset += 16; 5525 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5526 uint64_t size = m_data.GetU64_unchecked(&offset); 5527 5528 // "main bin spec" (main binary specification) data payload is 5529 // formatted: 5530 // uint32_t version [currently 1] 5531 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user 5532 // process] uint64_t address [ UINT64_MAX if address not 5533 // specified ] uuid_t uuid [ all zero's if uuid not 5534 // specified ] uint32_t log2_pagesize [ process page size in log base 5535 // 2, e.g. 4k pages are 12. 0 for unspecified ] 5536 5537 if (strcmp("main bin spec", data_owner) == 0 && size >= 32) { 5538 offset = fileoff; 5539 uint32_t version; 5540 if (m_data.GetU32(&offset, &version, 1) != nullptr && version == 1) { 5541 uint32_t type = 0; 5542 uuid_t raw_uuid; 5543 memset(raw_uuid, 0, sizeof(uuid_t)); 5544 5545 if (m_data.GetU32(&offset, &type, 1) && 5546 m_data.GetU64(&offset, &address, 1) && 5547 m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5548 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5549 return true; 5550 } 5551 } 5552 } 5553 } 5554 offset = cmd_offset + lc.cmdsize; 5555 } 5556 } 5557 return false; 5558 } 5559 5560 lldb::RegisterContextSP 5561 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5562 lldb_private::Thread &thread) { 5563 lldb::RegisterContextSP reg_ctx_sp; 5564 5565 ModuleSP module_sp(GetModule()); 5566 if (module_sp) { 5567 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5568 if (!m_thread_context_offsets_valid) 5569 GetNumThreadContexts(); 5570 5571 const FileRangeArray::Entry *thread_context_file_range = 5572 m_thread_context_offsets.GetEntryAtIndex(idx); 5573 if (thread_context_file_range) { 5574 5575 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5576 thread_context_file_range->GetByteSize()); 5577 5578 switch (m_header.cputype) { 5579 case llvm::MachO::CPU_TYPE_ARM64: 5580 case llvm::MachO::CPU_TYPE_ARM64_32: 5581 reg_ctx_sp = 5582 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data); 5583 break; 5584 5585 case llvm::MachO::CPU_TYPE_ARM: 5586 reg_ctx_sp = 5587 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data); 5588 break; 5589 5590 case llvm::MachO::CPU_TYPE_I386: 5591 reg_ctx_sp = 5592 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data); 5593 break; 5594 5595 case llvm::MachO::CPU_TYPE_X86_64: 5596 reg_ctx_sp = 5597 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data); 5598 break; 5599 } 5600 } 5601 } 5602 return reg_ctx_sp; 5603 } 5604 5605 ObjectFile::Type ObjectFileMachO::CalculateType() { 5606 switch (m_header.filetype) { 5607 case MH_OBJECT: // 0x1u 5608 if (GetAddressByteSize() == 4) { 5609 // 32 bit kexts are just object files, but they do have a valid 5610 // UUID load command. 5611 if (GetUUID()) { 5612 // this checking for the UUID load command is not enough we could 5613 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5614 // this is required of kexts 5615 if (m_strata == eStrataInvalid) 5616 m_strata = eStrataKernel; 5617 return eTypeSharedLibrary; 5618 } 5619 } 5620 return eTypeObjectFile; 5621 5622 case MH_EXECUTE: 5623 return eTypeExecutable; // 0x2u 5624 case MH_FVMLIB: 5625 return eTypeSharedLibrary; // 0x3u 5626 case MH_CORE: 5627 return eTypeCoreFile; // 0x4u 5628 case MH_PRELOAD: 5629 return eTypeSharedLibrary; // 0x5u 5630 case MH_DYLIB: 5631 return eTypeSharedLibrary; // 0x6u 5632 case MH_DYLINKER: 5633 return eTypeDynamicLinker; // 0x7u 5634 case MH_BUNDLE: 5635 return eTypeSharedLibrary; // 0x8u 5636 case MH_DYLIB_STUB: 5637 return eTypeStubLibrary; // 0x9u 5638 case MH_DSYM: 5639 return eTypeDebugInfo; // 0xAu 5640 case MH_KEXT_BUNDLE: 5641 return eTypeSharedLibrary; // 0xBu 5642 default: 5643 break; 5644 } 5645 return eTypeUnknown; 5646 } 5647 5648 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5649 switch (m_header.filetype) { 5650 case MH_OBJECT: // 0x1u 5651 { 5652 // 32 bit kexts are just object files, but they do have a valid 5653 // UUID load command. 5654 if (GetUUID()) { 5655 // this checking for the UUID load command is not enough we could 5656 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5657 // this is required of kexts 5658 if (m_type == eTypeInvalid) 5659 m_type = eTypeSharedLibrary; 5660 5661 return eStrataKernel; 5662 } 5663 } 5664 return eStrataUnknown; 5665 5666 case MH_EXECUTE: // 0x2u 5667 // Check for the MH_DYLDLINK bit in the flags 5668 if (m_header.flags & MH_DYLDLINK) { 5669 return eStrataUser; 5670 } else { 5671 SectionList *section_list = GetSectionList(); 5672 if (section_list) { 5673 static ConstString g_kld_section_name("__KLD"); 5674 if (section_list->FindSectionByName(g_kld_section_name)) 5675 return eStrataKernel; 5676 } 5677 } 5678 return eStrataRawImage; 5679 5680 case MH_FVMLIB: 5681 return eStrataUser; // 0x3u 5682 case MH_CORE: 5683 return eStrataUnknown; // 0x4u 5684 case MH_PRELOAD: 5685 return eStrataRawImage; // 0x5u 5686 case MH_DYLIB: 5687 return eStrataUser; // 0x6u 5688 case MH_DYLINKER: 5689 return eStrataUser; // 0x7u 5690 case MH_BUNDLE: 5691 return eStrataUser; // 0x8u 5692 case MH_DYLIB_STUB: 5693 return eStrataUser; // 0x9u 5694 case MH_DSYM: 5695 return eStrataUnknown; // 0xAu 5696 case MH_KEXT_BUNDLE: 5697 return eStrataKernel; // 0xBu 5698 default: 5699 break; 5700 } 5701 return eStrataUnknown; 5702 } 5703 5704 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5705 ModuleSP module_sp(GetModule()); 5706 if (module_sp) { 5707 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5708 struct dylib_command load_cmd; 5709 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5710 uint32_t version_cmd = 0; 5711 uint64_t version = 0; 5712 uint32_t i; 5713 for (i = 0; i < m_header.ncmds; ++i) { 5714 const lldb::offset_t cmd_offset = offset; 5715 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5716 break; 5717 5718 if (load_cmd.cmd == LC_ID_DYLIB) { 5719 if (version_cmd == 0) { 5720 version_cmd = load_cmd.cmd; 5721 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr) 5722 break; 5723 version = load_cmd.dylib.current_version; 5724 } 5725 break; // Break for now unless there is another more complete version 5726 // number load command in the future. 5727 } 5728 offset = cmd_offset + load_cmd.cmdsize; 5729 } 5730 5731 if (version_cmd == LC_ID_DYLIB) { 5732 unsigned major = (version & 0xFFFF0000ull) >> 16; 5733 unsigned minor = (version & 0x0000FF00ull) >> 8; 5734 unsigned subminor = (version & 0x000000FFull); 5735 return llvm::VersionTuple(major, minor, subminor); 5736 } 5737 } 5738 return llvm::VersionTuple(); 5739 } 5740 5741 ArchSpec ObjectFileMachO::GetArchitecture() { 5742 ModuleSP module_sp(GetModule()); 5743 ArchSpec arch; 5744 if (module_sp) { 5745 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5746 5747 return GetArchitecture(module_sp, m_header, m_data, 5748 MachHeaderSizeFromMagic(m_header.magic)); 5749 } 5750 return arch; 5751 } 5752 5753 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, 5754 addr_t &base_addr, UUID &uuid) { 5755 uuid.Clear(); 5756 base_addr = LLDB_INVALID_ADDRESS; 5757 if (process && process->GetDynamicLoader()) { 5758 DynamicLoader *dl = process->GetDynamicLoader(); 5759 LazyBool using_shared_cache; 5760 LazyBool private_shared_cache; 5761 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5762 private_shared_cache); 5763 } 5764 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5765 LIBLLDB_LOG_PROCESS)); 5766 LLDB_LOGF( 5767 log, 5768 "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64, 5769 uuid.GetAsString().c_str(), base_addr); 5770 } 5771 5772 // From dyld SPI header dyld_process_info.h 5773 typedef void *dyld_process_info; 5774 struct lldb_copy__dyld_process_cache_info { 5775 uuid_t cacheUUID; // UUID of cache used by process 5776 uint64_t cacheBaseAddress; // load address of dyld shared cache 5777 bool noCache; // process is running without a dyld cache 5778 bool privateCache; // process is using a private copy of its dyld cache 5779 }; 5780 5781 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with 5782 // llvm enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile 5783 // errors. So we need to use the actual underlying types of task_t and 5784 // kern_return_t below. 5785 extern "C" unsigned int /*task_t*/ mach_task_self(); 5786 5787 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5788 uuid.Clear(); 5789 base_addr = LLDB_INVALID_ADDRESS; 5790 5791 #if defined(__APPLE__) 5792 uint8_t *(*dyld_get_all_image_infos)(void); 5793 dyld_get_all_image_infos = 5794 (uint8_t * (*)()) dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5795 if (dyld_get_all_image_infos) { 5796 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5797 if (dyld_all_image_infos_address) { 5798 uint32_t *version = (uint32_t *) 5799 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5800 if (*version >= 13) { 5801 uuid_t *sharedCacheUUID_address = 0; 5802 int wordsize = sizeof(uint8_t *); 5803 if (wordsize == 8) { 5804 sharedCacheUUID_address = 5805 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5806 160); // sharedCacheUUID <mach-o/dyld_images.h> 5807 if (*version >= 15) 5808 base_addr = 5809 *(uint64_t 5810 *)((uint8_t *)dyld_all_image_infos_address + 5811 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5812 } else { 5813 sharedCacheUUID_address = 5814 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5815 84); // sharedCacheUUID <mach-o/dyld_images.h> 5816 if (*version >= 15) { 5817 base_addr = 0; 5818 base_addr = 5819 *(uint32_t 5820 *)((uint8_t *)dyld_all_image_infos_address + 5821 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5822 } 5823 } 5824 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5825 } 5826 } 5827 } else { 5828 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5829 dyld_process_info (*dyld_process_info_create)( 5830 unsigned int /* task_t */ task, uint64_t timestamp, 5831 unsigned int /*kern_return_t*/ *kernelError); 5832 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5833 void (*dyld_process_info_release)(dyld_process_info info); 5834 5835 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, 5836 unsigned int /*kern_return_t*/ *)) 5837 dlsym(RTLD_DEFAULT, "_dyld_process_info_create"); 5838 dyld_process_info_get_cache = (void (*)(void *, void *))dlsym( 5839 RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5840 dyld_process_info_release = 5841 (void (*)(void *))dlsym(RTLD_DEFAULT, "_dyld_process_info_release"); 5842 5843 if (dyld_process_info_create && dyld_process_info_get_cache) { 5844 unsigned int /*kern_return_t */ kern_ret; 5845 dyld_process_info process_info = 5846 dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5847 if (process_info) { 5848 struct lldb_copy__dyld_process_cache_info sc_info; 5849 memset(&sc_info, 0, sizeof(struct lldb_copy__dyld_process_cache_info)); 5850 dyld_process_info_get_cache(process_info, &sc_info); 5851 if (sc_info.cacheBaseAddress != 0) { 5852 base_addr = sc_info.cacheBaseAddress; 5853 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 5854 } 5855 dyld_process_info_release(process_info); 5856 } 5857 } 5858 } 5859 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5860 LIBLLDB_LOG_PROCESS)); 5861 if (log && uuid.IsValid()) 5862 LLDB_LOGF(log, 5863 "lldb's in-memory shared cache has a UUID of %s base address of " 5864 "0x%" PRIx64, 5865 uuid.GetAsString().c_str(), base_addr); 5866 #endif 5867 } 5868 5869 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5870 if (!m_min_os_version) { 5871 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5872 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5873 const lldb::offset_t load_cmd_offset = offset; 5874 5875 version_min_command lc; 5876 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5877 break; 5878 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5879 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5880 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5881 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5882 if (m_data.GetU32(&offset, &lc.version, 5883 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5884 const uint32_t xxxx = lc.version >> 16; 5885 const uint32_t yy = (lc.version >> 8) & 0xffu; 5886 const uint32_t zz = lc.version & 0xffu; 5887 if (xxxx) { 5888 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5889 break; 5890 } 5891 } 5892 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5893 // struct build_version_command { 5894 // uint32_t cmd; /* LC_BUILD_VERSION */ 5895 // uint32_t cmdsize; /* sizeof(struct 5896 // build_version_command) plus */ 5897 // /* ntools * sizeof(struct 5898 // build_tool_version) */ 5899 // uint32_t platform; /* platform */ 5900 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5901 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded in 5902 // nibbles xxxx.yy.zz */ uint32_t ntools; /* number of 5903 // tool entries following this */ 5904 // }; 5905 5906 offset += 4; // skip platform 5907 uint32_t minos = m_data.GetU32(&offset); 5908 5909 const uint32_t xxxx = minos >> 16; 5910 const uint32_t yy = (minos >> 8) & 0xffu; 5911 const uint32_t zz = minos & 0xffu; 5912 if (xxxx) { 5913 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5914 break; 5915 } 5916 } 5917 5918 offset = load_cmd_offset + lc.cmdsize; 5919 } 5920 5921 if (!m_min_os_version) { 5922 // Set version to an empty value so we don't keep trying to 5923 m_min_os_version = llvm::VersionTuple(); 5924 } 5925 } 5926 5927 return *m_min_os_version; 5928 } 5929 5930 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { 5931 if (!m_sdk_versions.hasValue()) { 5932 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5933 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5934 const lldb::offset_t load_cmd_offset = offset; 5935 5936 version_min_command lc; 5937 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5938 break; 5939 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5940 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5941 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5942 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5943 if (m_data.GetU32(&offset, &lc.version, 5944 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5945 const uint32_t xxxx = lc.sdk >> 16; 5946 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5947 const uint32_t zz = lc.sdk & 0xffu; 5948 if (xxxx) { 5949 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 5950 break; 5951 } else { 5952 GetModule()->ReportWarning("minimum OS version load command with " 5953 "invalid (0) version found."); 5954 } 5955 } 5956 } 5957 offset = load_cmd_offset + lc.cmdsize; 5958 } 5959 5960 if (!m_sdk_versions.hasValue()) { 5961 offset = MachHeaderSizeFromMagic(m_header.magic); 5962 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5963 const lldb::offset_t load_cmd_offset = offset; 5964 5965 version_min_command lc; 5966 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5967 break; 5968 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5969 // struct build_version_command { 5970 // uint32_t cmd; /* LC_BUILD_VERSION */ 5971 // uint32_t cmdsize; /* sizeof(struct 5972 // build_version_command) plus */ 5973 // /* ntools * sizeof(struct 5974 // build_tool_version) */ 5975 // uint32_t platform; /* platform */ 5976 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5977 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded 5978 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number 5979 // of tool entries following this */ 5980 // }; 5981 5982 offset += 4; // skip platform 5983 uint32_t minos = m_data.GetU32(&offset); 5984 5985 const uint32_t xxxx = minos >> 16; 5986 const uint32_t yy = (minos >> 8) & 0xffu; 5987 const uint32_t zz = minos & 0xffu; 5988 if (xxxx) { 5989 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 5990 break; 5991 } 5992 } 5993 offset = load_cmd_offset + lc.cmdsize; 5994 } 5995 } 5996 5997 if (!m_sdk_versions.hasValue()) 5998 m_sdk_versions = llvm::VersionTuple(); 5999 } 6000 6001 return m_sdk_versions.getValue(); 6002 } 6003 6004 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 6005 return m_header.filetype == llvm::MachO::MH_DYLINKER; 6006 } 6007 6008 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 6009 return m_allow_assembly_emulation_unwind_plans; 6010 } 6011 6012 // PluginInterface protocol 6013 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 6014 return GetPluginNameStatic(); 6015 } 6016 6017 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 6018 6019 Section *ObjectFileMachO::GetMachHeaderSection() { 6020 // Find the first address of the mach header which is the first non-zero file 6021 // sized section whose file offset is zero. This is the base file address of 6022 // the mach-o file which can be subtracted from the vmaddr of the other 6023 // segments found in memory and added to the load address 6024 ModuleSP module_sp = GetModule(); 6025 if (!module_sp) 6026 return nullptr; 6027 SectionList *section_list = GetSectionList(); 6028 if (!section_list) 6029 return nullptr; 6030 const size_t num_sections = section_list->GetSize(); 6031 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6032 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 6033 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 6034 return section; 6035 } 6036 return nullptr; 6037 } 6038 6039 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 6040 if (!section) 6041 return false; 6042 const bool is_dsym = (m_header.filetype == MH_DSYM); 6043 if (section->GetFileSize() == 0 && !is_dsym) 6044 return false; 6045 if (section->IsThreadSpecific()) 6046 return false; 6047 if (GetModule().get() != section->GetModule().get()) 6048 return false; 6049 // Be careful with __LINKEDIT and __DWARF segments 6050 if (section->GetName() == GetSegmentNameLINKEDIT() || 6051 section->GetName() == GetSegmentNameDWARF()) { 6052 // Only map __LINKEDIT and __DWARF if we have an in memory image and 6053 // this isn't a kernel binary like a kext or mach_kernel. 6054 const bool is_memory_image = (bool)m_process_wp.lock(); 6055 const Strata strata = GetStrata(); 6056 if (is_memory_image == false || strata == eStrataKernel) 6057 return false; 6058 } 6059 return true; 6060 } 6061 6062 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 6063 lldb::addr_t header_load_address, const Section *header_section, 6064 const Section *section) { 6065 ModuleSP module_sp = GetModule(); 6066 if (module_sp && header_section && section && 6067 header_load_address != LLDB_INVALID_ADDRESS) { 6068 lldb::addr_t file_addr = header_section->GetFileAddress(); 6069 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 6070 return section->GetFileAddress() - file_addr + header_load_address; 6071 } 6072 return LLDB_INVALID_ADDRESS; 6073 } 6074 6075 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 6076 bool value_is_offset) { 6077 ModuleSP module_sp = GetModule(); 6078 if (!module_sp) 6079 return false; 6080 6081 SectionList *section_list = GetSectionList(); 6082 if (!section_list) 6083 return false; 6084 6085 size_t num_loaded_sections = 0; 6086 const size_t num_sections = section_list->GetSize(); 6087 6088 if (value_is_offset) { 6089 // "value" is an offset to apply to each top level segment 6090 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6091 // Iterate through the object file sections to find all of the 6092 // sections that size on disk (to avoid __PAGEZERO) and load them 6093 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6094 if (SectionIsLoadable(section_sp.get())) 6095 if (target.GetSectionLoadList().SetSectionLoadAddress( 6096 section_sp, section_sp->GetFileAddress() + value)) 6097 ++num_loaded_sections; 6098 } 6099 } else { 6100 // "value" is the new base address of the mach_header, adjust each 6101 // section accordingly 6102 6103 Section *mach_header_section = GetMachHeaderSection(); 6104 if (mach_header_section) { 6105 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6106 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6107 6108 lldb::addr_t section_load_addr = 6109 CalculateSectionLoadAddressForMemoryImage( 6110 value, mach_header_section, section_sp.get()); 6111 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6112 if (target.GetSectionLoadList().SetSectionLoadAddress( 6113 section_sp, section_load_addr)) 6114 ++num_loaded_sections; 6115 } 6116 } 6117 } 6118 } 6119 return num_loaded_sections > 0; 6120 } 6121 6122 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6123 const FileSpec &outfile, Status &error) { 6124 if (!process_sp) 6125 return false; 6126 6127 Target &target = process_sp->GetTarget(); 6128 const ArchSpec target_arch = target.GetArchitecture(); 6129 const llvm::Triple &target_triple = target_arch.GetTriple(); 6130 if (target_triple.getVendor() == llvm::Triple::Apple && 6131 (target_triple.getOS() == llvm::Triple::MacOSX || 6132 target_triple.getOS() == llvm::Triple::IOS || 6133 target_triple.getOS() == llvm::Triple::WatchOS || 6134 target_triple.getOS() == llvm::Triple::TvOS)) { 6135 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) 6136 // { 6137 bool make_core = false; 6138 switch (target_arch.GetMachine()) { 6139 case llvm::Triple::aarch64: 6140 case llvm::Triple::aarch64_32: 6141 case llvm::Triple::arm: 6142 case llvm::Triple::thumb: 6143 case llvm::Triple::x86: 6144 case llvm::Triple::x86_64: 6145 make_core = true; 6146 break; 6147 default: 6148 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6149 target_triple.str().c_str()); 6150 break; 6151 } 6152 6153 if (make_core) { 6154 std::vector<segment_command_64> segment_load_commands; 6155 // uint32_t range_info_idx = 0; 6156 MemoryRegionInfo range_info; 6157 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6158 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6159 const ByteOrder byte_order = target_arch.GetByteOrder(); 6160 if (range_error.Success()) { 6161 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6162 const addr_t addr = range_info.GetRange().GetRangeBase(); 6163 const addr_t size = range_info.GetRange().GetByteSize(); 6164 6165 if (size == 0) 6166 break; 6167 6168 // Calculate correct protections 6169 uint32_t prot = 0; 6170 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6171 prot |= VM_PROT_READ; 6172 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6173 prot |= VM_PROT_WRITE; 6174 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6175 prot |= VM_PROT_EXECUTE; 6176 6177 if (prot != 0) { 6178 uint32_t cmd_type = LC_SEGMENT_64; 6179 uint32_t segment_size = sizeof(segment_command_64); 6180 if (addr_byte_size == 4) { 6181 cmd_type = LC_SEGMENT; 6182 segment_size = sizeof(segment_command); 6183 } 6184 segment_command_64 segment = { 6185 cmd_type, // uint32_t cmd; 6186 segment_size, // uint32_t cmdsize; 6187 {0}, // char segname[16]; 6188 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6189 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6190 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6191 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6192 prot, // uint32_t maxprot; 6193 prot, // uint32_t initprot; 6194 0, // uint32_t nsects; 6195 0}; // uint32_t flags; 6196 segment_load_commands.push_back(segment); 6197 } else { 6198 // No protections and a size of 1 used to be returned from old 6199 // debugservers when we asked about a region that was past the 6200 // last memory region and it indicates the end... 6201 if (size == 1) 6202 break; 6203 } 6204 6205 range_error = process_sp->GetMemoryRegionInfo( 6206 range_info.GetRange().GetRangeEnd(), range_info); 6207 if (range_error.Fail()) 6208 break; 6209 } 6210 6211 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6212 6213 mach_header_64 mach_header; 6214 if (addr_byte_size == 8) { 6215 mach_header.magic = MH_MAGIC_64; 6216 } else { 6217 mach_header.magic = MH_MAGIC; 6218 } 6219 mach_header.cputype = target_arch.GetMachOCPUType(); 6220 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6221 mach_header.filetype = MH_CORE; 6222 mach_header.ncmds = segment_load_commands.size(); 6223 mach_header.flags = 0; 6224 mach_header.reserved = 0; 6225 ThreadList &thread_list = process_sp->GetThreadList(); 6226 const uint32_t num_threads = thread_list.GetSize(); 6227 6228 // Make an array of LC_THREAD data items. Each one contains the 6229 // contents of the LC_THREAD load command. The data doesn't contain 6230 // the load command + load command size, we will add the load command 6231 // and load command size as we emit the data. 6232 std::vector<StreamString> LC_THREAD_datas(num_threads); 6233 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6234 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6235 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6236 LC_THREAD_data.SetByteOrder(byte_order); 6237 } 6238 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) { 6239 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6240 if (thread_sp) { 6241 switch (mach_header.cputype) { 6242 case llvm::MachO::CPU_TYPE_ARM64: 6243 case llvm::MachO::CPU_TYPE_ARM64_32: 6244 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6245 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6246 break; 6247 6248 case llvm::MachO::CPU_TYPE_ARM: 6249 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6250 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6251 break; 6252 6253 case llvm::MachO::CPU_TYPE_I386: 6254 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6255 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6256 break; 6257 6258 case llvm::MachO::CPU_TYPE_X86_64: 6259 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6260 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6261 break; 6262 } 6263 } 6264 } 6265 6266 // The size of the load command is the size of the segments... 6267 if (addr_byte_size == 8) { 6268 mach_header.sizeofcmds = 6269 segment_load_commands.size() * sizeof(struct segment_command_64); 6270 } else { 6271 mach_header.sizeofcmds = 6272 segment_load_commands.size() * sizeof(struct segment_command); 6273 } 6274 6275 // and the size of all LC_THREAD load command 6276 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6277 ++mach_header.ncmds; 6278 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6279 } 6280 6281 // Write the mach header 6282 buffer.PutHex32(mach_header.magic); 6283 buffer.PutHex32(mach_header.cputype); 6284 buffer.PutHex32(mach_header.cpusubtype); 6285 buffer.PutHex32(mach_header.filetype); 6286 buffer.PutHex32(mach_header.ncmds); 6287 buffer.PutHex32(mach_header.sizeofcmds); 6288 buffer.PutHex32(mach_header.flags); 6289 if (addr_byte_size == 8) { 6290 buffer.PutHex32(mach_header.reserved); 6291 } 6292 6293 // Skip the mach header and all load commands and align to the next 6294 // 0x1000 byte boundary 6295 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6296 if (file_offset & 0x00000fff) { 6297 file_offset += 0x00001000ull; 6298 file_offset &= (~0x00001000ull + 1); 6299 } 6300 6301 for (auto &segment : segment_load_commands) { 6302 segment.fileoff = file_offset; 6303 file_offset += segment.filesize; 6304 } 6305 6306 // Write out all of the LC_THREAD load commands 6307 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6308 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6309 buffer.PutHex32(LC_THREAD); 6310 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6311 buffer.Write(LC_THREAD_data.GetString().data(), LC_THREAD_data_size); 6312 } 6313 6314 // Write out all of the segment load commands 6315 for (const auto &segment : segment_load_commands) { 6316 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6317 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6318 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6319 segment.cmd, segment.cmdsize, segment.vmaddr, 6320 segment.vmaddr + segment.vmsize, segment.fileoff, 6321 segment.filesize, segment.maxprot, segment.initprot, 6322 segment.nsects, segment.flags); 6323 6324 buffer.PutHex32(segment.cmd); 6325 buffer.PutHex32(segment.cmdsize); 6326 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6327 if (addr_byte_size == 8) { 6328 buffer.PutHex64(segment.vmaddr); 6329 buffer.PutHex64(segment.vmsize); 6330 buffer.PutHex64(segment.fileoff); 6331 buffer.PutHex64(segment.filesize); 6332 } else { 6333 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6334 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6335 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6336 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6337 } 6338 buffer.PutHex32(segment.maxprot); 6339 buffer.PutHex32(segment.initprot); 6340 buffer.PutHex32(segment.nsects); 6341 buffer.PutHex32(segment.flags); 6342 } 6343 6344 std::string core_file_path(outfile.GetPath()); 6345 auto core_file = FileSystem::Instance().Open( 6346 outfile, File::eOpenOptionWrite | File::eOpenOptionTruncate | 6347 File::eOpenOptionCanCreate); 6348 if (!core_file) { 6349 error = core_file.takeError(); 6350 } else { 6351 // Read 1 page at a time 6352 uint8_t bytes[0x1000]; 6353 // Write the mach header and load commands out to the core file 6354 size_t bytes_written = buffer.GetString().size(); 6355 error = 6356 core_file.get()->Write(buffer.GetString().data(), bytes_written); 6357 if (error.Success()) { 6358 // Now write the file data for all memory segments in the process 6359 for (const auto &segment : segment_load_commands) { 6360 if (core_file.get()->SeekFromStart(segment.fileoff) == -1) { 6361 error.SetErrorStringWithFormat( 6362 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6363 segment.fileoff, core_file_path.c_str()); 6364 break; 6365 } 6366 6367 printf("Saving %" PRId64 6368 " bytes of data for memory region at 0x%" PRIx64 "\n", 6369 segment.vmsize, segment.vmaddr); 6370 addr_t bytes_left = segment.vmsize; 6371 addr_t addr = segment.vmaddr; 6372 Status memory_read_error; 6373 while (bytes_left > 0 && error.Success()) { 6374 const size_t bytes_to_read = 6375 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6376 6377 // In a savecore setting, we don't really care about caching, 6378 // as the data is dumped and very likely never read again, 6379 // so we call ReadMemoryFromInferior to bypass it. 6380 const size_t bytes_read = process_sp->ReadMemoryFromInferior( 6381 addr, bytes, bytes_to_read, memory_read_error); 6382 6383 if (bytes_read == bytes_to_read) { 6384 size_t bytes_written = bytes_read; 6385 error = core_file.get()->Write(bytes, bytes_written); 6386 bytes_left -= bytes_read; 6387 addr += bytes_read; 6388 } else { 6389 // Some pages within regions are not readable, those should 6390 // be zero filled 6391 memset(bytes, 0, bytes_to_read); 6392 size_t bytes_written = bytes_to_read; 6393 error = core_file.get()->Write(bytes, bytes_written); 6394 bytes_left -= bytes_to_read; 6395 addr += bytes_to_read; 6396 } 6397 } 6398 } 6399 } 6400 } 6401 } else { 6402 error.SetErrorString( 6403 "process doesn't support getting memory region info"); 6404 } 6405 } 6406 return true; // This is the right plug to handle saving core files for 6407 // this process 6408 } 6409 return false; 6410 } 6411