1 //===-- ObjectFileMachO.cpp -----------------------------------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 9 #include "llvm/ADT/StringRef.h" 10 11 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 12 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 13 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 14 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 15 #include "lldb/Core/Debugger.h" 16 #include "lldb/Core/FileSpecList.h" 17 #include "lldb/Core/Module.h" 18 #include "lldb/Core/ModuleSpec.h" 19 #include "lldb/Core/PluginManager.h" 20 #include "lldb/Core/Section.h" 21 #include "lldb/Core/StreamFile.h" 22 #include "lldb/Host/Host.h" 23 #include "lldb/Symbol/DWARFCallFrameInfo.h" 24 #include "lldb/Symbol/ObjectFile.h" 25 #include "lldb/Target/DynamicLoader.h" 26 #include "lldb/Target/MemoryRegionInfo.h" 27 #include "lldb/Target/Platform.h" 28 #include "lldb/Target/Process.h" 29 #include "lldb/Target/SectionLoadList.h" 30 #include "lldb/Target/Target.h" 31 #include "lldb/Target/Thread.h" 32 #include "lldb/Target/ThreadList.h" 33 #include "lldb/Utility/ArchSpec.h" 34 #include "lldb/Utility/DataBuffer.h" 35 #include "lldb/Utility/FileSpec.h" 36 #include "lldb/Utility/Log.h" 37 #include "lldb/Utility/RangeMap.h" 38 #include "lldb/Utility/RegisterValue.h" 39 #include "lldb/Utility/Status.h" 40 #include "lldb/Utility/StreamString.h" 41 #include "lldb/Utility/Timer.h" 42 #include "lldb/Utility/UUID.h" 43 44 #include "lldb/Host/SafeMachO.h" 45 46 #include "llvm/Support/MemoryBuffer.h" 47 48 #include "ObjectFileMachO.h" 49 50 #if defined(__APPLE__) && \ 51 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 52 // GetLLDBSharedCacheUUID() needs to call dlsym() 53 #include <dlfcn.h> 54 #endif 55 56 #ifndef __APPLE__ 57 #include "Utility/UuidCompatibility.h" 58 #else 59 #include <uuid/uuid.h> 60 #endif 61 62 #include <memory> 63 64 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 65 using namespace lldb; 66 using namespace lldb_private; 67 using namespace llvm::MachO; 68 69 LLDB_PLUGIN_DEFINE(ObjectFileMachO) 70 71 // Some structure definitions needed for parsing the dyld shared cache files 72 // found on iOS devices. 73 74 struct lldb_copy_dyld_cache_header_v1 { 75 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 76 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 77 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 78 uint32_t imagesOffset; 79 uint32_t imagesCount; 80 uint64_t dyldBaseAddress; 81 uint64_t codeSignatureOffset; 82 uint64_t codeSignatureSize; 83 uint64_t slideInfoOffset; 84 uint64_t slideInfoSize; 85 uint64_t localSymbolsOffset; 86 uint64_t localSymbolsSize; 87 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 88 // and later 89 }; 90 91 struct lldb_copy_dyld_cache_mapping_info { 92 uint64_t address; 93 uint64_t size; 94 uint64_t fileOffset; 95 uint32_t maxProt; 96 uint32_t initProt; 97 }; 98 99 struct lldb_copy_dyld_cache_local_symbols_info { 100 uint32_t nlistOffset; 101 uint32_t nlistCount; 102 uint32_t stringsOffset; 103 uint32_t stringsSize; 104 uint32_t entriesOffset; 105 uint32_t entriesCount; 106 }; 107 struct lldb_copy_dyld_cache_local_symbols_entry { 108 uint32_t dylibOffset; 109 uint32_t nlistStartIndex; 110 uint32_t nlistCount; 111 }; 112 113 static void PrintRegisterValue(RegisterContext *reg_ctx, const char *name, 114 const char *alt_name, size_t reg_byte_size, 115 Stream &data) { 116 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 117 if (reg_info == nullptr) 118 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 119 if (reg_info) { 120 lldb_private::RegisterValue reg_value; 121 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 122 if (reg_info->byte_size >= reg_byte_size) 123 data.Write(reg_value.GetBytes(), reg_byte_size); 124 else { 125 data.Write(reg_value.GetBytes(), reg_info->byte_size); 126 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; ++i) 127 data.PutChar(0); 128 } 129 return; 130 } 131 } 132 // Just write zeros if all else fails 133 for (size_t i = 0; i < reg_byte_size; ++i) 134 data.PutChar(0); 135 } 136 137 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 138 public: 139 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 140 const DataExtractor &data) 141 : RegisterContextDarwin_x86_64(thread, 0) { 142 SetRegisterDataFrom_LC_THREAD(data); 143 } 144 145 void InvalidateAllRegisters() override { 146 // Do nothing... registers are always valid... 147 } 148 149 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 150 lldb::offset_t offset = 0; 151 SetError(GPRRegSet, Read, -1); 152 SetError(FPURegSet, Read, -1); 153 SetError(EXCRegSet, Read, -1); 154 bool done = false; 155 156 while (!done) { 157 int flavor = data.GetU32(&offset); 158 if (flavor == 0) 159 done = true; 160 else { 161 uint32_t i; 162 uint32_t count = data.GetU32(&offset); 163 switch (flavor) { 164 case GPRRegSet: 165 for (i = 0; i < count; ++i) 166 (&gpr.rax)[i] = data.GetU64(&offset); 167 SetError(GPRRegSet, Read, 0); 168 done = true; 169 170 break; 171 case FPURegSet: 172 // TODO: fill in FPU regs.... 173 // SetError (FPURegSet, Read, -1); 174 done = true; 175 176 break; 177 case EXCRegSet: 178 exc.trapno = data.GetU32(&offset); 179 exc.err = data.GetU32(&offset); 180 exc.faultvaddr = data.GetU64(&offset); 181 SetError(EXCRegSet, Read, 0); 182 done = true; 183 break; 184 case 7: 185 case 8: 186 case 9: 187 // fancy flavors that encapsulate of the above flavors... 188 break; 189 190 default: 191 done = true; 192 break; 193 } 194 } 195 } 196 } 197 198 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 199 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 200 if (reg_ctx_sp) { 201 RegisterContext *reg_ctx = reg_ctx_sp.get(); 202 203 data.PutHex32(GPRRegSet); // Flavor 204 data.PutHex32(GPRWordCount); 205 PrintRegisterValue(reg_ctx, "rax", nullptr, 8, data); 206 PrintRegisterValue(reg_ctx, "rbx", nullptr, 8, data); 207 PrintRegisterValue(reg_ctx, "rcx", nullptr, 8, data); 208 PrintRegisterValue(reg_ctx, "rdx", nullptr, 8, data); 209 PrintRegisterValue(reg_ctx, "rdi", nullptr, 8, data); 210 PrintRegisterValue(reg_ctx, "rsi", nullptr, 8, data); 211 PrintRegisterValue(reg_ctx, "rbp", nullptr, 8, data); 212 PrintRegisterValue(reg_ctx, "rsp", nullptr, 8, data); 213 PrintRegisterValue(reg_ctx, "r8", nullptr, 8, data); 214 PrintRegisterValue(reg_ctx, "r9", nullptr, 8, data); 215 PrintRegisterValue(reg_ctx, "r10", nullptr, 8, data); 216 PrintRegisterValue(reg_ctx, "r11", nullptr, 8, data); 217 PrintRegisterValue(reg_ctx, "r12", nullptr, 8, data); 218 PrintRegisterValue(reg_ctx, "r13", nullptr, 8, data); 219 PrintRegisterValue(reg_ctx, "r14", nullptr, 8, data); 220 PrintRegisterValue(reg_ctx, "r15", nullptr, 8, data); 221 PrintRegisterValue(reg_ctx, "rip", nullptr, 8, data); 222 PrintRegisterValue(reg_ctx, "rflags", nullptr, 8, data); 223 PrintRegisterValue(reg_ctx, "cs", nullptr, 8, data); 224 PrintRegisterValue(reg_ctx, "fs", nullptr, 8, data); 225 PrintRegisterValue(reg_ctx, "gs", nullptr, 8, data); 226 227 // // Write out the FPU registers 228 // const size_t fpu_byte_size = sizeof(FPU); 229 // size_t bytes_written = 0; 230 // data.PutHex32 (FPURegSet); 231 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 232 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 233 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 234 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 235 // data); // uint16_t fcw; // "fctrl" 236 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 237 // data); // uint16_t fsw; // "fstat" 238 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 239 // data); // uint8_t ftw; // "ftag" 240 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 241 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 242 // data); // uint16_t fop; // "fop" 243 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 244 // data); // uint32_t ip; // "fioff" 245 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 246 // data); // uint16_t cs; // "fiseg" 247 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 248 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 249 // data); // uint32_t dp; // "fooff" 250 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 251 // data); // uint16_t ds; // "foseg" 252 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 253 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 254 // data); // uint32_t mxcsr; 255 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 256 // 4, data);// uint32_t mxcsrmask; 257 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 258 // sizeof(MMSReg), data); 259 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 260 // sizeof(MMSReg), data); 261 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 262 // sizeof(MMSReg), data); 263 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 264 // sizeof(MMSReg), data); 265 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 266 // sizeof(MMSReg), data); 267 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 268 // sizeof(MMSReg), data); 269 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 270 // sizeof(MMSReg), data); 271 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 272 // sizeof(MMSReg), data); 273 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 274 // sizeof(XMMReg), data); 275 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 276 // sizeof(XMMReg), data); 277 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 278 // sizeof(XMMReg), data); 279 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 280 // sizeof(XMMReg), data); 281 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 282 // sizeof(XMMReg), data); 283 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 284 // sizeof(XMMReg), data); 285 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 286 // sizeof(XMMReg), data); 287 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 288 // sizeof(XMMReg), data); 289 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 290 // sizeof(XMMReg), data); 291 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 292 // sizeof(XMMReg), data); 293 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 294 // sizeof(XMMReg), data); 295 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 296 // sizeof(XMMReg), data); 297 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 298 // sizeof(XMMReg), data); 299 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 300 // sizeof(XMMReg), data); 301 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 302 // sizeof(XMMReg), data); 303 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 304 // sizeof(XMMReg), data); 305 // 306 // // Fill rest with zeros 307 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 308 // i) 309 // data.PutChar(0); 310 311 // Write out the EXC registers 312 data.PutHex32(EXCRegSet); 313 data.PutHex32(EXCWordCount); 314 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 315 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 316 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 8, data); 317 return true; 318 } 319 return false; 320 } 321 322 protected: 323 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 324 325 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 326 327 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 328 329 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 330 return 0; 331 } 332 333 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 334 return 0; 335 } 336 337 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 338 return 0; 339 } 340 }; 341 342 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 343 public: 344 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 345 const DataExtractor &data) 346 : RegisterContextDarwin_i386(thread, 0) { 347 SetRegisterDataFrom_LC_THREAD(data); 348 } 349 350 void InvalidateAllRegisters() override { 351 // Do nothing... registers are always valid... 352 } 353 354 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 355 lldb::offset_t offset = 0; 356 SetError(GPRRegSet, Read, -1); 357 SetError(FPURegSet, Read, -1); 358 SetError(EXCRegSet, Read, -1); 359 bool done = false; 360 361 while (!done) { 362 int flavor = data.GetU32(&offset); 363 if (flavor == 0) 364 done = true; 365 else { 366 uint32_t i; 367 uint32_t count = data.GetU32(&offset); 368 switch (flavor) { 369 case GPRRegSet: 370 for (i = 0; i < count; ++i) 371 (&gpr.eax)[i] = data.GetU32(&offset); 372 SetError(GPRRegSet, Read, 0); 373 done = true; 374 375 break; 376 case FPURegSet: 377 // TODO: fill in FPU regs.... 378 // SetError (FPURegSet, Read, -1); 379 done = true; 380 381 break; 382 case EXCRegSet: 383 exc.trapno = data.GetU32(&offset); 384 exc.err = data.GetU32(&offset); 385 exc.faultvaddr = data.GetU32(&offset); 386 SetError(EXCRegSet, Read, 0); 387 done = true; 388 break; 389 case 7: 390 case 8: 391 case 9: 392 // fancy flavors that encapsulate of the above flavors... 393 break; 394 395 default: 396 done = true; 397 break; 398 } 399 } 400 } 401 } 402 403 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 404 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 405 if (reg_ctx_sp) { 406 RegisterContext *reg_ctx = reg_ctx_sp.get(); 407 408 data.PutHex32(GPRRegSet); // Flavor 409 data.PutHex32(GPRWordCount); 410 PrintRegisterValue(reg_ctx, "eax", nullptr, 4, data); 411 PrintRegisterValue(reg_ctx, "ebx", nullptr, 4, data); 412 PrintRegisterValue(reg_ctx, "ecx", nullptr, 4, data); 413 PrintRegisterValue(reg_ctx, "edx", nullptr, 4, data); 414 PrintRegisterValue(reg_ctx, "edi", nullptr, 4, data); 415 PrintRegisterValue(reg_ctx, "esi", nullptr, 4, data); 416 PrintRegisterValue(reg_ctx, "ebp", nullptr, 4, data); 417 PrintRegisterValue(reg_ctx, "esp", nullptr, 4, data); 418 PrintRegisterValue(reg_ctx, "ss", nullptr, 4, data); 419 PrintRegisterValue(reg_ctx, "eflags", nullptr, 4, data); 420 PrintRegisterValue(reg_ctx, "eip", nullptr, 4, data); 421 PrintRegisterValue(reg_ctx, "cs", nullptr, 4, data); 422 PrintRegisterValue(reg_ctx, "ds", nullptr, 4, data); 423 PrintRegisterValue(reg_ctx, "es", nullptr, 4, data); 424 PrintRegisterValue(reg_ctx, "fs", nullptr, 4, data); 425 PrintRegisterValue(reg_ctx, "gs", nullptr, 4, data); 426 427 // Write out the EXC registers 428 data.PutHex32(EXCRegSet); 429 data.PutHex32(EXCWordCount); 430 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 431 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 432 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 4, data); 433 return true; 434 } 435 return false; 436 } 437 438 protected: 439 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 440 441 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 442 443 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 444 445 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 446 return 0; 447 } 448 449 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 450 return 0; 451 } 452 453 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 454 return 0; 455 } 456 }; 457 458 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 459 public: 460 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 461 const DataExtractor &data) 462 : RegisterContextDarwin_arm(thread, 0) { 463 SetRegisterDataFrom_LC_THREAD(data); 464 } 465 466 void InvalidateAllRegisters() override { 467 // Do nothing... registers are always valid... 468 } 469 470 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 471 lldb::offset_t offset = 0; 472 SetError(GPRRegSet, Read, -1); 473 SetError(FPURegSet, Read, -1); 474 SetError(EXCRegSet, Read, -1); 475 bool done = false; 476 477 while (!done) { 478 int flavor = data.GetU32(&offset); 479 uint32_t count = data.GetU32(&offset); 480 lldb::offset_t next_thread_state = offset + (count * 4); 481 switch (flavor) { 482 case GPRAltRegSet: 483 case GPRRegSet: 484 // On ARM, the CPSR register is also included in the count but it is 485 // not included in gpr.r so loop until (count-1). 486 for (uint32_t i = 0; i < (count - 1); ++i) { 487 gpr.r[i] = data.GetU32(&offset); 488 } 489 // Save cpsr explicitly. 490 gpr.cpsr = data.GetU32(&offset); 491 492 SetError(GPRRegSet, Read, 0); 493 offset = next_thread_state; 494 break; 495 496 case FPURegSet: { 497 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 498 const int fpu_reg_buf_size = sizeof(fpu.floats); 499 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 500 fpu_reg_buf) == fpu_reg_buf_size) { 501 offset += fpu_reg_buf_size; 502 fpu.fpscr = data.GetU32(&offset); 503 SetError(FPURegSet, Read, 0); 504 } else { 505 done = true; 506 } 507 } 508 offset = next_thread_state; 509 break; 510 511 case EXCRegSet: 512 if (count == 3) { 513 exc.exception = data.GetU32(&offset); 514 exc.fsr = data.GetU32(&offset); 515 exc.far = data.GetU32(&offset); 516 SetError(EXCRegSet, Read, 0); 517 } 518 done = true; 519 offset = next_thread_state; 520 break; 521 522 // Unknown register set flavor, stop trying to parse. 523 default: 524 done = true; 525 } 526 } 527 } 528 529 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 530 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 531 if (reg_ctx_sp) { 532 RegisterContext *reg_ctx = reg_ctx_sp.get(); 533 534 data.PutHex32(GPRRegSet); // Flavor 535 data.PutHex32(GPRWordCount); 536 PrintRegisterValue(reg_ctx, "r0", nullptr, 4, data); 537 PrintRegisterValue(reg_ctx, "r1", nullptr, 4, data); 538 PrintRegisterValue(reg_ctx, "r2", nullptr, 4, data); 539 PrintRegisterValue(reg_ctx, "r3", nullptr, 4, data); 540 PrintRegisterValue(reg_ctx, "r4", nullptr, 4, data); 541 PrintRegisterValue(reg_ctx, "r5", nullptr, 4, data); 542 PrintRegisterValue(reg_ctx, "r6", nullptr, 4, data); 543 PrintRegisterValue(reg_ctx, "r7", nullptr, 4, data); 544 PrintRegisterValue(reg_ctx, "r8", nullptr, 4, data); 545 PrintRegisterValue(reg_ctx, "r9", nullptr, 4, data); 546 PrintRegisterValue(reg_ctx, "r10", nullptr, 4, data); 547 PrintRegisterValue(reg_ctx, "r11", nullptr, 4, data); 548 PrintRegisterValue(reg_ctx, "r12", nullptr, 4, data); 549 PrintRegisterValue(reg_ctx, "sp", nullptr, 4, data); 550 PrintRegisterValue(reg_ctx, "lr", nullptr, 4, data); 551 PrintRegisterValue(reg_ctx, "pc", nullptr, 4, data); 552 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 553 554 // Write out the EXC registers 555 // data.PutHex32 (EXCRegSet); 556 // data.PutHex32 (EXCWordCount); 557 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 558 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 559 // WriteRegister (reg_ctx, "far", NULL, 4, data); 560 return true; 561 } 562 return false; 563 } 564 565 protected: 566 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 567 568 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 569 570 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 571 572 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 573 574 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 575 return 0; 576 } 577 578 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 579 return 0; 580 } 581 582 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 583 return 0; 584 } 585 586 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 587 return -1; 588 } 589 }; 590 591 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 592 public: 593 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 594 const DataExtractor &data) 595 : RegisterContextDarwin_arm64(thread, 0) { 596 SetRegisterDataFrom_LC_THREAD(data); 597 } 598 599 void InvalidateAllRegisters() override { 600 // Do nothing... registers are always valid... 601 } 602 603 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 604 lldb::offset_t offset = 0; 605 SetError(GPRRegSet, Read, -1); 606 SetError(FPURegSet, Read, -1); 607 SetError(EXCRegSet, Read, -1); 608 bool done = false; 609 while (!done) { 610 int flavor = data.GetU32(&offset); 611 uint32_t count = data.GetU32(&offset); 612 lldb::offset_t next_thread_state = offset + (count * 4); 613 switch (flavor) { 614 case GPRRegSet: 615 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 616 // 32-bit register) 617 if (count >= (33 * 2) + 1) { 618 for (uint32_t i = 0; i < 29; ++i) 619 gpr.x[i] = data.GetU64(&offset); 620 gpr.fp = data.GetU64(&offset); 621 gpr.lr = data.GetU64(&offset); 622 gpr.sp = data.GetU64(&offset); 623 gpr.pc = data.GetU64(&offset); 624 gpr.cpsr = data.GetU32(&offset); 625 SetError(GPRRegSet, Read, 0); 626 } 627 offset = next_thread_state; 628 break; 629 case FPURegSet: { 630 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 631 const int fpu_reg_buf_size = sizeof(fpu); 632 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 633 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 634 fpu_reg_buf) == fpu_reg_buf_size) { 635 SetError(FPURegSet, Read, 0); 636 } else { 637 done = true; 638 } 639 } 640 offset = next_thread_state; 641 break; 642 case EXCRegSet: 643 if (count == 4) { 644 exc.far = data.GetU64(&offset); 645 exc.esr = data.GetU32(&offset); 646 exc.exception = data.GetU32(&offset); 647 SetError(EXCRegSet, Read, 0); 648 } 649 offset = next_thread_state; 650 break; 651 default: 652 done = true; 653 break; 654 } 655 } 656 } 657 658 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 659 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 660 if (reg_ctx_sp) { 661 RegisterContext *reg_ctx = reg_ctx_sp.get(); 662 663 data.PutHex32(GPRRegSet); // Flavor 664 data.PutHex32(GPRWordCount); 665 PrintRegisterValue(reg_ctx, "x0", nullptr, 8, data); 666 PrintRegisterValue(reg_ctx, "x1", nullptr, 8, data); 667 PrintRegisterValue(reg_ctx, "x2", nullptr, 8, data); 668 PrintRegisterValue(reg_ctx, "x3", nullptr, 8, data); 669 PrintRegisterValue(reg_ctx, "x4", nullptr, 8, data); 670 PrintRegisterValue(reg_ctx, "x5", nullptr, 8, data); 671 PrintRegisterValue(reg_ctx, "x6", nullptr, 8, data); 672 PrintRegisterValue(reg_ctx, "x7", nullptr, 8, data); 673 PrintRegisterValue(reg_ctx, "x8", nullptr, 8, data); 674 PrintRegisterValue(reg_ctx, "x9", nullptr, 8, data); 675 PrintRegisterValue(reg_ctx, "x10", nullptr, 8, data); 676 PrintRegisterValue(reg_ctx, "x11", nullptr, 8, data); 677 PrintRegisterValue(reg_ctx, "x12", nullptr, 8, data); 678 PrintRegisterValue(reg_ctx, "x13", nullptr, 8, data); 679 PrintRegisterValue(reg_ctx, "x14", nullptr, 8, data); 680 PrintRegisterValue(reg_ctx, "x15", nullptr, 8, data); 681 PrintRegisterValue(reg_ctx, "x16", nullptr, 8, data); 682 PrintRegisterValue(reg_ctx, "x17", nullptr, 8, data); 683 PrintRegisterValue(reg_ctx, "x18", nullptr, 8, data); 684 PrintRegisterValue(reg_ctx, "x19", nullptr, 8, data); 685 PrintRegisterValue(reg_ctx, "x20", nullptr, 8, data); 686 PrintRegisterValue(reg_ctx, "x21", nullptr, 8, data); 687 PrintRegisterValue(reg_ctx, "x22", nullptr, 8, data); 688 PrintRegisterValue(reg_ctx, "x23", nullptr, 8, data); 689 PrintRegisterValue(reg_ctx, "x24", nullptr, 8, data); 690 PrintRegisterValue(reg_ctx, "x25", nullptr, 8, data); 691 PrintRegisterValue(reg_ctx, "x26", nullptr, 8, data); 692 PrintRegisterValue(reg_ctx, "x27", nullptr, 8, data); 693 PrintRegisterValue(reg_ctx, "x28", nullptr, 8, data); 694 PrintRegisterValue(reg_ctx, "fp", nullptr, 8, data); 695 PrintRegisterValue(reg_ctx, "lr", nullptr, 8, data); 696 PrintRegisterValue(reg_ctx, "sp", nullptr, 8, data); 697 PrintRegisterValue(reg_ctx, "pc", nullptr, 8, data); 698 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 699 700 // Write out the EXC registers 701 // data.PutHex32 (EXCRegSet); 702 // data.PutHex32 (EXCWordCount); 703 // WriteRegister (reg_ctx, "far", NULL, 8, data); 704 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 705 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 706 return true; 707 } 708 return false; 709 } 710 711 protected: 712 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 713 714 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 715 716 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 717 718 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 719 720 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 721 return 0; 722 } 723 724 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 725 return 0; 726 } 727 728 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 729 return 0; 730 } 731 732 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 733 return -1; 734 } 735 }; 736 737 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 738 switch (magic) { 739 case MH_MAGIC: 740 case MH_CIGAM: 741 return sizeof(struct mach_header); 742 743 case MH_MAGIC_64: 744 case MH_CIGAM_64: 745 return sizeof(struct mach_header_64); 746 break; 747 748 default: 749 break; 750 } 751 return 0; 752 } 753 754 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 755 756 char ObjectFileMachO::ID; 757 758 void ObjectFileMachO::Initialize() { 759 PluginManager::RegisterPlugin( 760 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 761 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 762 } 763 764 void ObjectFileMachO::Terminate() { 765 PluginManager::UnregisterPlugin(CreateInstance); 766 } 767 768 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 769 static ConstString g_name("mach-o"); 770 return g_name; 771 } 772 773 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 774 return "Mach-o object file reader (32 and 64 bit)"; 775 } 776 777 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 778 DataBufferSP &data_sp, 779 lldb::offset_t data_offset, 780 const FileSpec *file, 781 lldb::offset_t file_offset, 782 lldb::offset_t length) { 783 if (!data_sp) { 784 data_sp = MapFileData(*file, length, file_offset); 785 if (!data_sp) 786 return nullptr; 787 data_offset = 0; 788 } 789 790 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 791 return nullptr; 792 793 // Update the data to contain the entire file if it doesn't already 794 if (data_sp->GetByteSize() < length) { 795 data_sp = MapFileData(*file, length, file_offset); 796 if (!data_sp) 797 return nullptr; 798 data_offset = 0; 799 } 800 auto objfile_up = std::make_unique<ObjectFileMachO>( 801 module_sp, data_sp, data_offset, file, file_offset, length); 802 if (!objfile_up || !objfile_up->ParseHeader()) 803 return nullptr; 804 805 return objfile_up.release(); 806 } 807 808 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 809 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 810 const ProcessSP &process_sp, lldb::addr_t header_addr) { 811 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 812 std::unique_ptr<ObjectFile> objfile_up( 813 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 814 if (objfile_up.get() && objfile_up->ParseHeader()) 815 return objfile_up.release(); 816 } 817 return nullptr; 818 } 819 820 size_t ObjectFileMachO::GetModuleSpecifications( 821 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 822 lldb::offset_t data_offset, lldb::offset_t file_offset, 823 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 824 const size_t initial_count = specs.GetSize(); 825 826 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 827 DataExtractor data; 828 data.SetData(data_sp); 829 llvm::MachO::mach_header header; 830 if (ParseHeader(data, &data_offset, header)) { 831 size_t header_and_load_cmds = 832 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 833 if (header_and_load_cmds >= data_sp->GetByteSize()) { 834 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 835 data.SetData(data_sp); 836 data_offset = MachHeaderSizeFromMagic(header.magic); 837 } 838 if (data_sp) { 839 ModuleSpec base_spec; 840 base_spec.GetFileSpec() = file; 841 base_spec.SetObjectOffset(file_offset); 842 base_spec.SetObjectSize(length); 843 GetAllArchSpecs(header, data, data_offset, base_spec, specs); 844 } 845 } 846 } 847 return specs.GetSize() - initial_count; 848 } 849 850 ConstString ObjectFileMachO::GetSegmentNameTEXT() { 851 static ConstString g_segment_name_TEXT("__TEXT"); 852 return g_segment_name_TEXT; 853 } 854 855 ConstString ObjectFileMachO::GetSegmentNameDATA() { 856 static ConstString g_segment_name_DATA("__DATA"); 857 return g_segment_name_DATA; 858 } 859 860 ConstString ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 861 static ConstString g_segment_name("__DATA_DIRTY"); 862 return g_segment_name; 863 } 864 865 ConstString ObjectFileMachO::GetSegmentNameDATA_CONST() { 866 static ConstString g_segment_name("__DATA_CONST"); 867 return g_segment_name; 868 } 869 870 ConstString ObjectFileMachO::GetSegmentNameOBJC() { 871 static ConstString g_segment_name_OBJC("__OBJC"); 872 return g_segment_name_OBJC; 873 } 874 875 ConstString ObjectFileMachO::GetSegmentNameLINKEDIT() { 876 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 877 return g_section_name_LINKEDIT; 878 } 879 880 ConstString ObjectFileMachO::GetSegmentNameDWARF() { 881 static ConstString g_section_name("__DWARF"); 882 return g_section_name; 883 } 884 885 ConstString ObjectFileMachO::GetSectionNameEHFrame() { 886 static ConstString g_section_name_eh_frame("__eh_frame"); 887 return g_section_name_eh_frame; 888 } 889 890 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 891 lldb::addr_t data_offset, 892 lldb::addr_t data_length) { 893 DataExtractor data; 894 data.SetData(data_sp, data_offset, data_length); 895 lldb::offset_t offset = 0; 896 uint32_t magic = data.GetU32(&offset); 897 return MachHeaderSizeFromMagic(magic) != 0; 898 } 899 900 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 901 DataBufferSP &data_sp, 902 lldb::offset_t data_offset, 903 const FileSpec *file, 904 lldb::offset_t file_offset, 905 lldb::offset_t length) 906 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 907 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 908 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 909 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 910 ::memset(&m_header, 0, sizeof(m_header)); 911 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 912 } 913 914 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 915 lldb::DataBufferSP &header_data_sp, 916 const lldb::ProcessSP &process_sp, 917 lldb::addr_t header_addr) 918 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 919 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 920 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 921 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 922 ::memset(&m_header, 0, sizeof(m_header)); 923 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 924 } 925 926 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 927 lldb::offset_t *data_offset_ptr, 928 llvm::MachO::mach_header &header) { 929 data.SetByteOrder(endian::InlHostByteOrder()); 930 // Leave magic in the original byte order 931 header.magic = data.GetU32(data_offset_ptr); 932 bool can_parse = false; 933 bool is_64_bit = false; 934 switch (header.magic) { 935 case MH_MAGIC: 936 data.SetByteOrder(endian::InlHostByteOrder()); 937 data.SetAddressByteSize(4); 938 can_parse = true; 939 break; 940 941 case MH_MAGIC_64: 942 data.SetByteOrder(endian::InlHostByteOrder()); 943 data.SetAddressByteSize(8); 944 can_parse = true; 945 is_64_bit = true; 946 break; 947 948 case MH_CIGAM: 949 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 950 ? eByteOrderLittle 951 : eByteOrderBig); 952 data.SetAddressByteSize(4); 953 can_parse = true; 954 break; 955 956 case MH_CIGAM_64: 957 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 958 ? eByteOrderLittle 959 : eByteOrderBig); 960 data.SetAddressByteSize(8); 961 is_64_bit = true; 962 can_parse = true; 963 break; 964 965 default: 966 break; 967 } 968 969 if (can_parse) { 970 data.GetU32(data_offset_ptr, &header.cputype, 6); 971 if (is_64_bit) 972 *data_offset_ptr += 4; 973 return true; 974 } else { 975 memset(&header, 0, sizeof(header)); 976 } 977 return false; 978 } 979 980 bool ObjectFileMachO::ParseHeader() { 981 ModuleSP module_sp(GetModule()); 982 if (!module_sp) 983 return false; 984 985 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 986 bool can_parse = false; 987 lldb::offset_t offset = 0; 988 m_data.SetByteOrder(endian::InlHostByteOrder()); 989 // Leave magic in the original byte order 990 m_header.magic = m_data.GetU32(&offset); 991 switch (m_header.magic) { 992 case MH_MAGIC: 993 m_data.SetByteOrder(endian::InlHostByteOrder()); 994 m_data.SetAddressByteSize(4); 995 can_parse = true; 996 break; 997 998 case MH_MAGIC_64: 999 m_data.SetByteOrder(endian::InlHostByteOrder()); 1000 m_data.SetAddressByteSize(8); 1001 can_parse = true; 1002 break; 1003 1004 case MH_CIGAM: 1005 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1006 ? eByteOrderLittle 1007 : eByteOrderBig); 1008 m_data.SetAddressByteSize(4); 1009 can_parse = true; 1010 break; 1011 1012 case MH_CIGAM_64: 1013 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1014 ? eByteOrderLittle 1015 : eByteOrderBig); 1016 m_data.SetAddressByteSize(8); 1017 can_parse = true; 1018 break; 1019 1020 default: 1021 break; 1022 } 1023 1024 if (can_parse) { 1025 m_data.GetU32(&offset, &m_header.cputype, 6); 1026 1027 ModuleSpecList all_specs; 1028 ModuleSpec base_spec; 1029 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 1030 base_spec, all_specs); 1031 1032 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 1033 ArchSpec mach_arch = 1034 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 1035 1036 // Check if the module has a required architecture 1037 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1038 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1039 continue; 1040 1041 if (SetModulesArchitecture(mach_arch)) { 1042 const size_t header_and_lc_size = 1043 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1044 if (m_data.GetByteSize() < header_and_lc_size) { 1045 DataBufferSP data_sp; 1046 ProcessSP process_sp(m_process_wp.lock()); 1047 if (process_sp) { 1048 data_sp = ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1049 } else { 1050 // Read in all only the load command data from the file on disk 1051 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset); 1052 if (data_sp->GetByteSize() != header_and_lc_size) 1053 continue; 1054 } 1055 if (data_sp) 1056 m_data.SetData(data_sp); 1057 } 1058 } 1059 return true; 1060 } 1061 // None found. 1062 return false; 1063 } else { 1064 memset(&m_header, 0, sizeof(struct mach_header)); 1065 } 1066 return false; 1067 } 1068 1069 ByteOrder ObjectFileMachO::GetByteOrder() const { 1070 return m_data.GetByteOrder(); 1071 } 1072 1073 bool ObjectFileMachO::IsExecutable() const { 1074 return m_header.filetype == MH_EXECUTE; 1075 } 1076 1077 bool ObjectFileMachO::IsDynamicLoader() const { 1078 return m_header.filetype == MH_DYLINKER; 1079 } 1080 1081 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1082 return m_data.GetAddressByteSize(); 1083 } 1084 1085 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1086 Symtab *symtab = GetSymtab(); 1087 if (!symtab) 1088 return AddressClass::eUnknown; 1089 1090 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1091 if (symbol) { 1092 if (symbol->ValueIsAddress()) { 1093 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1094 if (section_sp) { 1095 const lldb::SectionType section_type = section_sp->GetType(); 1096 switch (section_type) { 1097 case eSectionTypeInvalid: 1098 return AddressClass::eUnknown; 1099 1100 case eSectionTypeCode: 1101 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1102 // For ARM we have a bit in the n_desc field of the symbol that 1103 // tells us ARM/Thumb which is bit 0x0008. 1104 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1105 return AddressClass::eCodeAlternateISA; 1106 } 1107 return AddressClass::eCode; 1108 1109 case eSectionTypeContainer: 1110 return AddressClass::eUnknown; 1111 1112 case eSectionTypeData: 1113 case eSectionTypeDataCString: 1114 case eSectionTypeDataCStringPointers: 1115 case eSectionTypeDataSymbolAddress: 1116 case eSectionTypeData4: 1117 case eSectionTypeData8: 1118 case eSectionTypeData16: 1119 case eSectionTypeDataPointers: 1120 case eSectionTypeZeroFill: 1121 case eSectionTypeDataObjCMessageRefs: 1122 case eSectionTypeDataObjCCFStrings: 1123 case eSectionTypeGoSymtab: 1124 return AddressClass::eData; 1125 1126 case eSectionTypeDebug: 1127 case eSectionTypeDWARFDebugAbbrev: 1128 case eSectionTypeDWARFDebugAbbrevDwo: 1129 case eSectionTypeDWARFDebugAddr: 1130 case eSectionTypeDWARFDebugAranges: 1131 case eSectionTypeDWARFDebugCuIndex: 1132 case eSectionTypeDWARFDebugFrame: 1133 case eSectionTypeDWARFDebugInfo: 1134 case eSectionTypeDWARFDebugInfoDwo: 1135 case eSectionTypeDWARFDebugLine: 1136 case eSectionTypeDWARFDebugLineStr: 1137 case eSectionTypeDWARFDebugLoc: 1138 case eSectionTypeDWARFDebugLocDwo: 1139 case eSectionTypeDWARFDebugLocLists: 1140 case eSectionTypeDWARFDebugLocListsDwo: 1141 case eSectionTypeDWARFDebugMacInfo: 1142 case eSectionTypeDWARFDebugMacro: 1143 case eSectionTypeDWARFDebugNames: 1144 case eSectionTypeDWARFDebugPubNames: 1145 case eSectionTypeDWARFDebugPubTypes: 1146 case eSectionTypeDWARFDebugRanges: 1147 case eSectionTypeDWARFDebugRngLists: 1148 case eSectionTypeDWARFDebugRngListsDwo: 1149 case eSectionTypeDWARFDebugStr: 1150 case eSectionTypeDWARFDebugStrDwo: 1151 case eSectionTypeDWARFDebugStrOffsets: 1152 case eSectionTypeDWARFDebugStrOffsetsDwo: 1153 case eSectionTypeDWARFDebugTuIndex: 1154 case eSectionTypeDWARFDebugTypes: 1155 case eSectionTypeDWARFDebugTypesDwo: 1156 case eSectionTypeDWARFAppleNames: 1157 case eSectionTypeDWARFAppleTypes: 1158 case eSectionTypeDWARFAppleNamespaces: 1159 case eSectionTypeDWARFAppleObjC: 1160 case eSectionTypeDWARFGNUDebugAltLink: 1161 return AddressClass::eDebug; 1162 1163 case eSectionTypeEHFrame: 1164 case eSectionTypeARMexidx: 1165 case eSectionTypeARMextab: 1166 case eSectionTypeCompactUnwind: 1167 return AddressClass::eRuntime; 1168 1169 case eSectionTypeAbsoluteAddress: 1170 case eSectionTypeELFSymbolTable: 1171 case eSectionTypeELFDynamicSymbols: 1172 case eSectionTypeELFRelocationEntries: 1173 case eSectionTypeELFDynamicLinkInfo: 1174 case eSectionTypeOther: 1175 return AddressClass::eUnknown; 1176 } 1177 } 1178 } 1179 1180 const SymbolType symbol_type = symbol->GetType(); 1181 switch (symbol_type) { 1182 case eSymbolTypeAny: 1183 return AddressClass::eUnknown; 1184 case eSymbolTypeAbsolute: 1185 return AddressClass::eUnknown; 1186 1187 case eSymbolTypeCode: 1188 case eSymbolTypeTrampoline: 1189 case eSymbolTypeResolver: 1190 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1191 // For ARM we have a bit in the n_desc field of the symbol that tells 1192 // us ARM/Thumb which is bit 0x0008. 1193 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1194 return AddressClass::eCodeAlternateISA; 1195 } 1196 return AddressClass::eCode; 1197 1198 case eSymbolTypeData: 1199 return AddressClass::eData; 1200 case eSymbolTypeRuntime: 1201 return AddressClass::eRuntime; 1202 case eSymbolTypeException: 1203 return AddressClass::eRuntime; 1204 case eSymbolTypeSourceFile: 1205 return AddressClass::eDebug; 1206 case eSymbolTypeHeaderFile: 1207 return AddressClass::eDebug; 1208 case eSymbolTypeObjectFile: 1209 return AddressClass::eDebug; 1210 case eSymbolTypeCommonBlock: 1211 return AddressClass::eDebug; 1212 case eSymbolTypeBlock: 1213 return AddressClass::eDebug; 1214 case eSymbolTypeLocal: 1215 return AddressClass::eData; 1216 case eSymbolTypeParam: 1217 return AddressClass::eData; 1218 case eSymbolTypeVariable: 1219 return AddressClass::eData; 1220 case eSymbolTypeVariableType: 1221 return AddressClass::eDebug; 1222 case eSymbolTypeLineEntry: 1223 return AddressClass::eDebug; 1224 case eSymbolTypeLineHeader: 1225 return AddressClass::eDebug; 1226 case eSymbolTypeScopeBegin: 1227 return AddressClass::eDebug; 1228 case eSymbolTypeScopeEnd: 1229 return AddressClass::eDebug; 1230 case eSymbolTypeAdditional: 1231 return AddressClass::eUnknown; 1232 case eSymbolTypeCompiler: 1233 return AddressClass::eDebug; 1234 case eSymbolTypeInstrumentation: 1235 return AddressClass::eDebug; 1236 case eSymbolTypeUndefined: 1237 return AddressClass::eUnknown; 1238 case eSymbolTypeObjCClass: 1239 return AddressClass::eRuntime; 1240 case eSymbolTypeObjCMetaClass: 1241 return AddressClass::eRuntime; 1242 case eSymbolTypeObjCIVar: 1243 return AddressClass::eRuntime; 1244 case eSymbolTypeReExported: 1245 return AddressClass::eRuntime; 1246 } 1247 } 1248 return AddressClass::eUnknown; 1249 } 1250 1251 Symtab *ObjectFileMachO::GetSymtab() { 1252 ModuleSP module_sp(GetModule()); 1253 if (module_sp) { 1254 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1255 if (m_symtab_up == nullptr) { 1256 m_symtab_up = 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 (seg_cmd.fileoff > m_length) { 1332 // We have a load command that says it extends past the end of the file. 1333 // This is likely a corrupt file. We don't have any way to return an error 1334 // condition here (this method was likely invoked from something like 1335 // ObjectFile::GetSectionList()), so we just null out the section contents, 1336 // and dump a message to stdout. The most common case here is core file 1337 // debugging with a truncated file. 1338 const char *lc_segment_name = 1339 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1340 GetModule()->ReportWarning( 1341 "load command %u %s has a fileoff (0x%" PRIx64 1342 ") that extends beyond the end of the file (0x%" PRIx64 1343 "), ignoring this section", 1344 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1345 1346 seg_cmd.fileoff = 0; 1347 seg_cmd.filesize = 0; 1348 } 1349 1350 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1351 // We have a load command that says it extends past the end of the file. 1352 // This is likely a corrupt file. We don't have any way to return an error 1353 // condition here (this method was likely invoked from something like 1354 // ObjectFile::GetSectionList()), so we just null out the section contents, 1355 // and dump a message to stdout. The most common case here is core file 1356 // debugging with a truncated file. 1357 const char *lc_segment_name = 1358 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1359 GetModule()->ReportWarning( 1360 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1361 ") that extends beyond the end of the file (0x%" PRIx64 1362 "), the segment will be truncated to match", 1363 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1364 1365 // Truncate the length 1366 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1367 } 1368 } 1369 1370 static uint32_t GetSegmentPermissions(const segment_command_64 &seg_cmd) { 1371 uint32_t result = 0; 1372 if (seg_cmd.initprot & VM_PROT_READ) 1373 result |= ePermissionsReadable; 1374 if (seg_cmd.initprot & VM_PROT_WRITE) 1375 result |= ePermissionsWritable; 1376 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1377 result |= ePermissionsExecutable; 1378 return result; 1379 } 1380 1381 static lldb::SectionType GetSectionType(uint32_t flags, 1382 ConstString section_name) { 1383 1384 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1385 return eSectionTypeCode; 1386 1387 uint32_t mach_sect_type = flags & SECTION_TYPE; 1388 static ConstString g_sect_name_objc_data("__objc_data"); 1389 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1390 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1391 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1392 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1393 static ConstString g_sect_name_objc_const("__objc_const"); 1394 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1395 static ConstString g_sect_name_cfstring("__cfstring"); 1396 1397 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1398 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1399 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1400 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1401 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1402 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1403 static ConstString g_sect_name_dwarf_debug_loclists("__debug_loclists"); 1404 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1405 static ConstString g_sect_name_dwarf_debug_names("__debug_names"); 1406 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1407 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1408 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1409 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1410 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1411 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1412 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1413 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1414 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1415 static ConstString g_sect_name_eh_frame("__eh_frame"); 1416 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1417 static ConstString g_sect_name_text("__text"); 1418 static ConstString g_sect_name_data("__data"); 1419 static ConstString g_sect_name_go_symtab("__gosymtab"); 1420 1421 if (section_name == g_sect_name_dwarf_debug_abbrev) 1422 return eSectionTypeDWARFDebugAbbrev; 1423 if (section_name == g_sect_name_dwarf_debug_aranges) 1424 return eSectionTypeDWARFDebugAranges; 1425 if (section_name == g_sect_name_dwarf_debug_frame) 1426 return eSectionTypeDWARFDebugFrame; 1427 if (section_name == g_sect_name_dwarf_debug_info) 1428 return eSectionTypeDWARFDebugInfo; 1429 if (section_name == g_sect_name_dwarf_debug_line) 1430 return eSectionTypeDWARFDebugLine; 1431 if (section_name == g_sect_name_dwarf_debug_loc) 1432 return eSectionTypeDWARFDebugLoc; 1433 if (section_name == g_sect_name_dwarf_debug_loclists) 1434 return eSectionTypeDWARFDebugLocLists; 1435 if (section_name == g_sect_name_dwarf_debug_macinfo) 1436 return eSectionTypeDWARFDebugMacInfo; 1437 if (section_name == g_sect_name_dwarf_debug_names) 1438 return eSectionTypeDWARFDebugNames; 1439 if (section_name == g_sect_name_dwarf_debug_pubnames) 1440 return eSectionTypeDWARFDebugPubNames; 1441 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1442 return eSectionTypeDWARFDebugPubTypes; 1443 if (section_name == g_sect_name_dwarf_debug_ranges) 1444 return eSectionTypeDWARFDebugRanges; 1445 if (section_name == g_sect_name_dwarf_debug_str) 1446 return eSectionTypeDWARFDebugStr; 1447 if (section_name == g_sect_name_dwarf_debug_types) 1448 return eSectionTypeDWARFDebugTypes; 1449 if (section_name == g_sect_name_dwarf_apple_names) 1450 return eSectionTypeDWARFAppleNames; 1451 if (section_name == g_sect_name_dwarf_apple_types) 1452 return eSectionTypeDWARFAppleTypes; 1453 if (section_name == g_sect_name_dwarf_apple_namespaces) 1454 return eSectionTypeDWARFAppleNamespaces; 1455 if (section_name == g_sect_name_dwarf_apple_objc) 1456 return eSectionTypeDWARFAppleObjC; 1457 if (section_name == g_sect_name_objc_selrefs) 1458 return eSectionTypeDataCStringPointers; 1459 if (section_name == g_sect_name_objc_msgrefs) 1460 return eSectionTypeDataObjCMessageRefs; 1461 if (section_name == g_sect_name_eh_frame) 1462 return eSectionTypeEHFrame; 1463 if (section_name == g_sect_name_compact_unwind) 1464 return eSectionTypeCompactUnwind; 1465 if (section_name == g_sect_name_cfstring) 1466 return eSectionTypeDataObjCCFStrings; 1467 if (section_name == g_sect_name_go_symtab) 1468 return eSectionTypeGoSymtab; 1469 if (section_name == g_sect_name_objc_data || 1470 section_name == g_sect_name_objc_classrefs || 1471 section_name == g_sect_name_objc_superrefs || 1472 section_name == g_sect_name_objc_const || 1473 section_name == g_sect_name_objc_classlist) { 1474 return eSectionTypeDataPointers; 1475 } 1476 1477 switch (mach_sect_type) { 1478 // TODO: categorize sections by other flags for regular sections 1479 case S_REGULAR: 1480 if (section_name == g_sect_name_text) 1481 return eSectionTypeCode; 1482 if (section_name == g_sect_name_data) 1483 return eSectionTypeData; 1484 return eSectionTypeOther; 1485 case S_ZEROFILL: 1486 return eSectionTypeZeroFill; 1487 case S_CSTRING_LITERALS: // section with only literal C strings 1488 return eSectionTypeDataCString; 1489 case S_4BYTE_LITERALS: // section with only 4 byte literals 1490 return eSectionTypeData4; 1491 case S_8BYTE_LITERALS: // section with only 8 byte literals 1492 return eSectionTypeData8; 1493 case S_LITERAL_POINTERS: // section with only pointers to literals 1494 return eSectionTypeDataPointers; 1495 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1496 return eSectionTypeDataPointers; 1497 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1498 return eSectionTypeDataPointers; 1499 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1500 // the reserved2 field 1501 return eSectionTypeCode; 1502 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1503 // initialization 1504 return eSectionTypeDataPointers; 1505 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1506 // termination 1507 return eSectionTypeDataPointers; 1508 case S_COALESCED: 1509 return eSectionTypeOther; 1510 case S_GB_ZEROFILL: 1511 return eSectionTypeZeroFill; 1512 case S_INTERPOSING: // section with only pairs of function pointers for 1513 // interposing 1514 return eSectionTypeCode; 1515 case S_16BYTE_LITERALS: // section with only 16 byte literals 1516 return eSectionTypeData16; 1517 case S_DTRACE_DOF: 1518 return eSectionTypeDebug; 1519 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1520 return eSectionTypeDataPointers; 1521 default: 1522 return eSectionTypeOther; 1523 } 1524 } 1525 1526 struct ObjectFileMachO::SegmentParsingContext { 1527 const EncryptedFileRanges EncryptedRanges; 1528 lldb_private::SectionList &UnifiedList; 1529 uint32_t NextSegmentIdx = 0; 1530 uint32_t NextSectionIdx = 0; 1531 bool FileAddressesChanged = false; 1532 1533 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1534 lldb_private::SectionList &UnifiedList) 1535 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1536 }; 1537 1538 void ObjectFileMachO::ProcessSegmentCommand(const load_command &load_cmd_, 1539 lldb::offset_t offset, 1540 uint32_t cmd_idx, 1541 SegmentParsingContext &context) { 1542 segment_command_64 load_cmd; 1543 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1544 1545 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1546 return; 1547 1548 ModuleSP module_sp = GetModule(); 1549 const bool is_core = GetType() == eTypeCoreFile; 1550 const bool is_dsym = (m_header.filetype == MH_DSYM); 1551 bool add_section = true; 1552 bool add_to_unified = true; 1553 ConstString const_segname( 1554 load_cmd.segname, strnlen(load_cmd.segname, sizeof(load_cmd.segname))); 1555 1556 SectionSP unified_section_sp( 1557 context.UnifiedList.FindSectionByName(const_segname)); 1558 if (is_dsym && unified_section_sp) { 1559 if (const_segname == GetSegmentNameLINKEDIT()) { 1560 // We need to keep the __LINKEDIT segment private to this object file 1561 // only 1562 add_to_unified = false; 1563 } else { 1564 // This is the dSYM file and this section has already been created by the 1565 // object file, no need to create it. 1566 add_section = false; 1567 } 1568 } 1569 load_cmd.vmaddr = m_data.GetAddress(&offset); 1570 load_cmd.vmsize = m_data.GetAddress(&offset); 1571 load_cmd.fileoff = m_data.GetAddress(&offset); 1572 load_cmd.filesize = m_data.GetAddress(&offset); 1573 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1574 return; 1575 1576 SanitizeSegmentCommand(load_cmd, cmd_idx); 1577 1578 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1579 const bool segment_is_encrypted = 1580 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1581 1582 // Keep a list of mach segments around in case we need to get at data that 1583 // isn't stored in the abstracted Sections. 1584 m_mach_segments.push_back(load_cmd); 1585 1586 // Use a segment ID of the segment index shifted left by 8 so they never 1587 // conflict with any of the sections. 1588 SectionSP segment_sp; 1589 if (add_section && (const_segname || is_core)) { 1590 segment_sp = std::make_shared<Section>( 1591 module_sp, // Module to which this section belongs 1592 this, // Object file to which this sections belongs 1593 ++context.NextSegmentIdx 1594 << 8, // Section ID is the 1 based segment index 1595 // shifted right by 8 bits as not to collide with any of the 256 1596 // section IDs that are possible 1597 const_segname, // Name of this section 1598 eSectionTypeContainer, // This section is a container of other 1599 // sections. 1600 load_cmd.vmaddr, // File VM address == addresses as they are 1601 // found in the object file 1602 load_cmd.vmsize, // VM size in bytes of this section 1603 load_cmd.fileoff, // Offset to the data for this section in 1604 // the file 1605 load_cmd.filesize, // Size in bytes of this section as found 1606 // in the file 1607 0, // Segments have no alignment information 1608 load_cmd.flags); // Flags for this section 1609 1610 segment_sp->SetIsEncrypted(segment_is_encrypted); 1611 m_sections_up->AddSection(segment_sp); 1612 segment_sp->SetPermissions(segment_permissions); 1613 if (add_to_unified) 1614 context.UnifiedList.AddSection(segment_sp); 1615 } else if (unified_section_sp) { 1616 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1617 // Check to see if the module was read from memory? 1618 if (module_sp->GetObjectFile()->GetBaseAddress().IsValid()) { 1619 // We have a module that is in memory and needs to have its file 1620 // address adjusted. We need to do this because when we load a file 1621 // from memory, its addresses will be slid already, yet the addresses 1622 // in the new symbol file will still be unslid. Since everything is 1623 // stored as section offset, this shouldn't cause any problems. 1624 1625 // Make sure we've parsed the symbol table from the ObjectFile before 1626 // we go around changing its Sections. 1627 module_sp->GetObjectFile()->GetSymtab(); 1628 // eh_frame would present the same problems but we parse that on a per- 1629 // function basis as-needed so it's more difficult to remove its use of 1630 // the Sections. Realistically, the environments where this code path 1631 // will be taken will not have eh_frame sections. 1632 1633 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1634 1635 // Notify the module that the section addresses have been changed once 1636 // we're done so any file-address caches can be updated. 1637 context.FileAddressesChanged = true; 1638 } 1639 } 1640 m_sections_up->AddSection(unified_section_sp); 1641 } 1642 1643 struct section_64 sect64; 1644 ::memset(§64, 0, sizeof(sect64)); 1645 // Push a section into our mach sections for the section at index zero 1646 // (NO_SECT) if we don't have any mach sections yet... 1647 if (m_mach_sections.empty()) 1648 m_mach_sections.push_back(sect64); 1649 uint32_t segment_sect_idx; 1650 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1651 1652 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1653 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1654 ++segment_sect_idx) { 1655 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1656 sizeof(sect64.sectname)) == nullptr) 1657 break; 1658 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1659 sizeof(sect64.segname)) == nullptr) 1660 break; 1661 sect64.addr = m_data.GetAddress(&offset); 1662 sect64.size = m_data.GetAddress(&offset); 1663 1664 if (m_data.GetU32(&offset, §64.offset, num_u32s) == nullptr) 1665 break; 1666 1667 // Keep a list of mach sections around in case we need to get at data that 1668 // isn't stored in the abstracted Sections. 1669 m_mach_sections.push_back(sect64); 1670 1671 if (add_section) { 1672 ConstString section_name( 1673 sect64.sectname, strnlen(sect64.sectname, sizeof(sect64.sectname))); 1674 if (!const_segname) { 1675 // We have a segment with no name so we need to conjure up segments 1676 // that correspond to the section's segname if there isn't already such 1677 // a section. If there is such a section, we resize the section so that 1678 // it spans all sections. We also mark these sections as fake so 1679 // address matches don't hit if they land in the gaps between the child 1680 // sections. 1681 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1682 sizeof(sect64.segname)); 1683 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1684 if (segment_sp.get()) { 1685 Section *segment = segment_sp.get(); 1686 // Grow the section size as needed. 1687 const lldb::addr_t sect64_min_addr = sect64.addr; 1688 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1689 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1690 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1691 const lldb::addr_t curr_seg_max_addr = 1692 curr_seg_min_addr + curr_seg_byte_size; 1693 if (sect64_min_addr >= curr_seg_min_addr) { 1694 const lldb::addr_t new_seg_byte_size = 1695 sect64_max_addr - curr_seg_min_addr; 1696 // Only grow the section size if needed 1697 if (new_seg_byte_size > curr_seg_byte_size) 1698 segment->SetByteSize(new_seg_byte_size); 1699 } else { 1700 // We need to change the base address of the segment and adjust the 1701 // child section offsets for all existing children. 1702 const lldb::addr_t slide_amount = 1703 sect64_min_addr - curr_seg_min_addr; 1704 segment->Slide(slide_amount, false); 1705 segment->GetChildren().Slide(-slide_amount, false); 1706 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1707 } 1708 1709 // Grow the section size as needed. 1710 if (sect64.offset) { 1711 const lldb::addr_t segment_min_file_offset = 1712 segment->GetFileOffset(); 1713 const lldb::addr_t segment_max_file_offset = 1714 segment_min_file_offset + segment->GetFileSize(); 1715 1716 const lldb::addr_t section_min_file_offset = sect64.offset; 1717 const lldb::addr_t section_max_file_offset = 1718 section_min_file_offset + sect64.size; 1719 const lldb::addr_t new_file_offset = 1720 std::min(section_min_file_offset, segment_min_file_offset); 1721 const lldb::addr_t new_file_size = 1722 std::max(section_max_file_offset, segment_max_file_offset) - 1723 new_file_offset; 1724 segment->SetFileOffset(new_file_offset); 1725 segment->SetFileSize(new_file_size); 1726 } 1727 } else { 1728 // Create a fake section for the section's named segment 1729 segment_sp = std::make_shared<Section>( 1730 segment_sp, // Parent section 1731 module_sp, // Module to which this section belongs 1732 this, // Object file to which this section belongs 1733 ++context.NextSegmentIdx 1734 << 8, // Section ID is the 1 based segment index 1735 // shifted right by 8 bits as not to 1736 // collide with any of the 256 section IDs 1737 // that are possible 1738 const_segname, // Name of this section 1739 eSectionTypeContainer, // This section is a container of 1740 // other sections. 1741 sect64.addr, // File VM address == addresses as they are 1742 // found in the object file 1743 sect64.size, // VM size in bytes of this section 1744 sect64.offset, // Offset to the data for this section in 1745 // the file 1746 sect64.offset ? sect64.size : 0, // Size in bytes of 1747 // this section as 1748 // found in the file 1749 sect64.align, 1750 load_cmd.flags); // Flags for this section 1751 segment_sp->SetIsFake(true); 1752 segment_sp->SetPermissions(segment_permissions); 1753 m_sections_up->AddSection(segment_sp); 1754 if (add_to_unified) 1755 context.UnifiedList.AddSection(segment_sp); 1756 segment_sp->SetIsEncrypted(segment_is_encrypted); 1757 } 1758 } 1759 assert(segment_sp.get()); 1760 1761 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1762 1763 SectionSP section_sp(new Section( 1764 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1765 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1766 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1767 sect64.flags)); 1768 // Set the section to be encrypted to match the segment 1769 1770 bool section_is_encrypted = false; 1771 if (!segment_is_encrypted && load_cmd.filesize != 0) 1772 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1773 sect64.offset) != nullptr; 1774 1775 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1776 section_sp->SetPermissions(segment_permissions); 1777 segment_sp->GetChildren().AddSection(section_sp); 1778 1779 if (segment_sp->IsFake()) { 1780 segment_sp.reset(); 1781 const_segname.Clear(); 1782 } 1783 } 1784 } 1785 if (segment_sp && is_dsym) { 1786 if (first_segment_sectID <= context.NextSectionIdx) { 1787 lldb::user_id_t sect_uid; 1788 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1789 ++sect_uid) { 1790 SectionSP curr_section_sp( 1791 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1792 SectionSP next_section_sp; 1793 if (sect_uid + 1 <= context.NextSectionIdx) 1794 next_section_sp = 1795 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1796 1797 if (curr_section_sp.get()) { 1798 if (curr_section_sp->GetByteSize() == 0) { 1799 if (next_section_sp.get() != nullptr) 1800 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1801 curr_section_sp->GetFileAddress()); 1802 else 1803 curr_section_sp->SetByteSize(load_cmd.vmsize); 1804 } 1805 } 1806 } 1807 } 1808 } 1809 } 1810 1811 void ObjectFileMachO::ProcessDysymtabCommand(const load_command &load_cmd, 1812 lldb::offset_t offset) { 1813 m_dysymtab.cmd = load_cmd.cmd; 1814 m_dysymtab.cmdsize = load_cmd.cmdsize; 1815 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1816 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1817 } 1818 1819 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1820 if (m_sections_up) 1821 return; 1822 1823 m_sections_up = std::make_unique<SectionList>(); 1824 1825 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1826 // bool dump_sections = false; 1827 ModuleSP module_sp(GetModule()); 1828 1829 offset = MachHeaderSizeFromMagic(m_header.magic); 1830 1831 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1832 struct load_command load_cmd; 1833 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1834 const lldb::offset_t load_cmd_offset = offset; 1835 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 1836 break; 1837 1838 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1839 ProcessSegmentCommand(load_cmd, offset, i, context); 1840 else if (load_cmd.cmd == LC_DYSYMTAB) 1841 ProcessDysymtabCommand(load_cmd, offset); 1842 1843 offset = load_cmd_offset + load_cmd.cmdsize; 1844 } 1845 1846 if (context.FileAddressesChanged && module_sp) 1847 module_sp->SectionFileAddressesChanged(); 1848 } 1849 1850 class MachSymtabSectionInfo { 1851 public: 1852 MachSymtabSectionInfo(SectionList *section_list) 1853 : m_section_list(section_list), m_section_infos() { 1854 // Get the number of sections down to a depth of 1 to include all segments 1855 // and their sections, but no other sections that may be added for debug 1856 // map or 1857 m_section_infos.resize(section_list->GetNumSections(1)); 1858 } 1859 1860 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1861 if (n_sect == 0) 1862 return SectionSP(); 1863 if (n_sect < m_section_infos.size()) { 1864 if (!m_section_infos[n_sect].section_sp) { 1865 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1866 m_section_infos[n_sect].section_sp = section_sp; 1867 if (section_sp) { 1868 m_section_infos[n_sect].vm_range.SetBaseAddress( 1869 section_sp->GetFileAddress()); 1870 m_section_infos[n_sect].vm_range.SetByteSize( 1871 section_sp->GetByteSize()); 1872 } else { 1873 const char *filename = "<unknown>"; 1874 SectionSP first_section_sp(m_section_list->GetSectionAtIndex(0)); 1875 if (first_section_sp) 1876 filename = first_section_sp->GetObjectFile()->GetFileSpec().GetPath().c_str(); 1877 1878 Host::SystemLog(Host::eSystemLogError, 1879 "error: unable to find section %d for a symbol in %s, corrupt file?\n", 1880 n_sect, 1881 filename); 1882 } 1883 } 1884 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1885 // Symbol is in section. 1886 return m_section_infos[n_sect].section_sp; 1887 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1888 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1889 file_addr) { 1890 // Symbol is in section with zero size, but has the same start address 1891 // as the section. This can happen with linker symbols (symbols that 1892 // start with the letter 'l' or 'L'. 1893 return m_section_infos[n_sect].section_sp; 1894 } 1895 } 1896 return m_section_list->FindSectionContainingFileAddress(file_addr); 1897 } 1898 1899 protected: 1900 struct SectionInfo { 1901 SectionInfo() : vm_range(), section_sp() {} 1902 1903 VMRange vm_range; 1904 SectionSP section_sp; 1905 }; 1906 SectionList *m_section_list; 1907 std::vector<SectionInfo> m_section_infos; 1908 }; 1909 1910 #define TRIE_SYMBOL_IS_THUMB (1ULL << 63) 1911 struct TrieEntry { 1912 void Dump() const { 1913 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1914 static_cast<unsigned long long>(address), 1915 static_cast<unsigned long long>(flags), 1916 static_cast<unsigned long long>(other), name.GetCString()); 1917 if (import_name) 1918 printf(" -> \"%s\"\n", import_name.GetCString()); 1919 else 1920 printf("\n"); 1921 } 1922 ConstString name; 1923 uint64_t address = LLDB_INVALID_ADDRESS; 1924 uint64_t flags = 1925 0; // EXPORT_SYMBOL_FLAGS_REEXPORT, EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER, 1926 // TRIE_SYMBOL_IS_THUMB 1927 uint64_t other = 0; 1928 ConstString import_name; 1929 }; 1930 1931 struct TrieEntryWithOffset { 1932 lldb::offset_t nodeOffset; 1933 TrieEntry entry; 1934 1935 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1936 1937 void Dump(uint32_t idx) const { 1938 printf("[%3u] 0x%16.16llx: ", idx, 1939 static_cast<unsigned long long>(nodeOffset)); 1940 entry.Dump(); 1941 } 1942 1943 bool operator<(const TrieEntryWithOffset &other) const { 1944 return (nodeOffset < other.nodeOffset); 1945 } 1946 }; 1947 1948 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 1949 const bool is_arm, addr_t text_seg_base_addr, 1950 std::vector<llvm::StringRef> &nameSlices, 1951 std::set<lldb::addr_t> &resolver_addresses, 1952 std::vector<TrieEntryWithOffset> &reexports, 1953 std::vector<TrieEntryWithOffset> &ext_symbols) { 1954 if (!data.ValidOffset(offset)) 1955 return true; 1956 1957 // Terminal node -- end of a branch, possibly add this to 1958 // the symbol table or resolver table. 1959 const uint64_t terminalSize = data.GetULEB128(&offset); 1960 lldb::offset_t children_offset = offset + terminalSize; 1961 if (terminalSize != 0) { 1962 TrieEntryWithOffset e(offset); 1963 e.entry.flags = data.GetULEB128(&offset); 1964 const char *import_name = nullptr; 1965 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 1966 e.entry.address = 0; 1967 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 1968 import_name = data.GetCStr(&offset); 1969 } else { 1970 e.entry.address = data.GetULEB128(&offset); 1971 if (text_seg_base_addr != LLDB_INVALID_ADDRESS) 1972 e.entry.address += text_seg_base_addr; 1973 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 1974 e.entry.other = data.GetULEB128(&offset); 1975 uint64_t resolver_addr = e.entry.other; 1976 if (is_arm) 1977 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 1978 resolver_addresses.insert(resolver_addr); 1979 } else 1980 e.entry.other = 0; 1981 } 1982 bool add_this_entry = false; 1983 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT) && 1984 import_name && import_name[0]) { 1985 // add symbols that are reexport symbols with a valid import name. 1986 add_this_entry = true; 1987 } else if (e.entry.flags == 0 && 1988 (import_name == nullptr || import_name[0] == '\0')) { 1989 // add externally visible symbols, in case the nlist record has 1990 // been stripped/omitted. 1991 add_this_entry = true; 1992 } 1993 if (add_this_entry) { 1994 std::string name; 1995 if (!nameSlices.empty()) { 1996 for (auto name_slice : nameSlices) 1997 name.append(name_slice.data(), name_slice.size()); 1998 } 1999 if (name.size() > 1) { 2000 // Skip the leading '_' 2001 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2002 } 2003 if (import_name) { 2004 // Skip the leading '_' 2005 e.entry.import_name.SetCString(import_name + 1); 2006 } 2007 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT)) { 2008 reexports.push_back(e); 2009 } else { 2010 if (is_arm && (e.entry.address & 1)) { 2011 e.entry.flags |= TRIE_SYMBOL_IS_THUMB; 2012 e.entry.address &= THUMB_ADDRESS_BIT_MASK; 2013 } 2014 ext_symbols.push_back(e); 2015 } 2016 } 2017 } 2018 2019 const uint8_t childrenCount = data.GetU8(&children_offset); 2020 for (uint8_t i = 0; i < childrenCount; ++i) { 2021 const char *cstr = data.GetCStr(&children_offset); 2022 if (cstr) 2023 nameSlices.push_back(llvm::StringRef(cstr)); 2024 else 2025 return false; // Corrupt data 2026 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2027 if (childNodeOffset) { 2028 if (!ParseTrieEntries(data, childNodeOffset, is_arm, text_seg_base_addr, 2029 nameSlices, resolver_addresses, reexports, 2030 ext_symbols)) { 2031 return false; 2032 } 2033 } 2034 nameSlices.pop_back(); 2035 } 2036 return true; 2037 } 2038 2039 static SymbolType GetSymbolType(const char *&symbol_name, 2040 bool &demangled_is_synthesized, 2041 const SectionSP &text_section_sp, 2042 const SectionSP &data_section_sp, 2043 const SectionSP &data_dirty_section_sp, 2044 const SectionSP &data_const_section_sp, 2045 const SectionSP &symbol_section) { 2046 SymbolType type = eSymbolTypeInvalid; 2047 2048 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2049 if (symbol_section->IsDescendant(text_section_sp.get())) { 2050 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 2051 S_ATTR_SELF_MODIFYING_CODE | 2052 S_ATTR_SOME_INSTRUCTIONS)) 2053 type = eSymbolTypeData; 2054 else 2055 type = eSymbolTypeCode; 2056 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 2057 symbol_section->IsDescendant(data_dirty_section_sp.get()) || 2058 symbol_section->IsDescendant(data_const_section_sp.get())) { 2059 if (symbol_sect_name && 2060 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 2061 type = eSymbolTypeRuntime; 2062 2063 if (symbol_name) { 2064 llvm::StringRef symbol_name_ref(symbol_name); 2065 if (symbol_name_ref.startswith("OBJC_")) { 2066 static const llvm::StringRef g_objc_v2_prefix_class("OBJC_CLASS_$_"); 2067 static const llvm::StringRef g_objc_v2_prefix_metaclass( 2068 "OBJC_METACLASS_$_"); 2069 static const llvm::StringRef g_objc_v2_prefix_ivar("OBJC_IVAR_$_"); 2070 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 2071 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2072 type = eSymbolTypeObjCClass; 2073 demangled_is_synthesized = true; 2074 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 2075 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2076 type = eSymbolTypeObjCMetaClass; 2077 demangled_is_synthesized = true; 2078 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 2079 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2080 type = eSymbolTypeObjCIVar; 2081 demangled_is_synthesized = true; 2082 } 2083 } 2084 } 2085 } else if (symbol_sect_name && 2086 ::strstr(symbol_sect_name, "__gcc_except_tab") == 2087 symbol_sect_name) { 2088 type = eSymbolTypeException; 2089 } else { 2090 type = eSymbolTypeData; 2091 } 2092 } else if (symbol_sect_name && 2093 ::strstr(symbol_sect_name, "__IMPORT") == symbol_sect_name) { 2094 type = eSymbolTypeTrampoline; 2095 } 2096 return type; 2097 } 2098 2099 // Read the UUID out of a dyld_shared_cache file on-disk. 2100 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2101 const ByteOrder byte_order, 2102 const uint32_t addr_byte_size) { 2103 UUID dsc_uuid; 2104 DataBufferSP DscData = MapFileData( 2105 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2106 if (!DscData) 2107 return dsc_uuid; 2108 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2109 2110 char version_str[7]; 2111 lldb::offset_t offset = 0; 2112 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2113 version_str[6] = '\0'; 2114 if (strcmp(version_str, "dyld_v") == 0) { 2115 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2116 dsc_uuid = UUID::fromOptionalData( 2117 dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t)); 2118 } 2119 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2120 if (log && dsc_uuid.IsValid()) { 2121 LLDB_LOGF(log, "Shared cache %s has UUID %s", 2122 dyld_shared_cache.GetPath().c_str(), 2123 dsc_uuid.GetAsString().c_str()); 2124 } 2125 return dsc_uuid; 2126 } 2127 2128 static llvm::Optional<struct nlist_64> 2129 ParseNList(DataExtractor &nlist_data, lldb::offset_t &nlist_data_offset, 2130 size_t nlist_byte_size) { 2131 struct nlist_64 nlist; 2132 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2133 return {}; 2134 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 2135 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 2136 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 2137 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 2138 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 2139 return nlist; 2140 } 2141 2142 enum { DebugSymbols = true, NonDebugSymbols = false }; 2143 2144 size_t ObjectFileMachO::ParseSymtab() { 2145 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2146 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2147 m_file.GetFilename().AsCString("")); 2148 ModuleSP module_sp(GetModule()); 2149 if (!module_sp) 2150 return 0; 2151 2152 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2153 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2154 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2155 // The data element of type bool indicates that this entry is thumb 2156 // code. 2157 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2158 2159 // Record the address of every function/data that we add to the symtab. 2160 // We add symbols to the table in the order of most information (nlist 2161 // records) to least (function starts), and avoid duplicating symbols 2162 // via this set. 2163 std::set<addr_t> symbols_added; 2164 FunctionStarts function_starts; 2165 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2166 uint32_t i; 2167 FileSpecList dylib_files; 2168 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2169 llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2170 llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2171 llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2172 2173 for (i = 0; i < m_header.ncmds; ++i) { 2174 const lldb::offset_t cmd_offset = offset; 2175 // Read in the load command and load command size 2176 struct load_command lc; 2177 if (m_data.GetU32(&offset, &lc, 2) == nullptr) 2178 break; 2179 // Watch for the symbol table load command 2180 switch (lc.cmd) { 2181 case LC_SYMTAB: 2182 symtab_load_command.cmd = lc.cmd; 2183 symtab_load_command.cmdsize = lc.cmdsize; 2184 // Read in the rest of the symtab load command 2185 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2186 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2187 return 0; 2188 break; 2189 2190 case LC_DYLD_INFO: 2191 case LC_DYLD_INFO_ONLY: 2192 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2193 dyld_info.cmd = lc.cmd; 2194 dyld_info.cmdsize = lc.cmdsize; 2195 } else { 2196 memset(&dyld_info, 0, sizeof(dyld_info)); 2197 } 2198 break; 2199 2200 case LC_LOAD_DYLIB: 2201 case LC_LOAD_WEAK_DYLIB: 2202 case LC_REEXPORT_DYLIB: 2203 case LC_LOADFVMLIB: 2204 case LC_LOAD_UPWARD_DYLIB: { 2205 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2206 const char *path = m_data.PeekCStr(name_offset); 2207 if (path) { 2208 FileSpec file_spec(path); 2209 // Strip the path if there is @rpath, @executable, etc so we just use 2210 // the basename 2211 if (path[0] == '@') 2212 file_spec.GetDirectory().Clear(); 2213 2214 if (lc.cmd == LC_REEXPORT_DYLIB) { 2215 m_reexported_dylibs.AppendIfUnique(file_spec); 2216 } 2217 2218 dylib_files.Append(file_spec); 2219 } 2220 } break; 2221 2222 case LC_FUNCTION_STARTS: 2223 function_starts_load_command.cmd = lc.cmd; 2224 function_starts_load_command.cmdsize = lc.cmdsize; 2225 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2226 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2227 memset(&function_starts_load_command, 0, 2228 sizeof(function_starts_load_command)); 2229 break; 2230 2231 default: 2232 break; 2233 } 2234 offset = cmd_offset + lc.cmdsize; 2235 } 2236 2237 if (!symtab_load_command.cmd) 2238 return 0; 2239 2240 Symtab *symtab = m_symtab_up.get(); 2241 SectionList *section_list = GetSectionList(); 2242 if (section_list == nullptr) 2243 return 0; 2244 2245 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2246 const ByteOrder byte_order = m_data.GetByteOrder(); 2247 bool bit_width_32 = addr_byte_size == 4; 2248 const size_t nlist_byte_size = 2249 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2250 2251 DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size); 2252 DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size); 2253 DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size); 2254 DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order, 2255 addr_byte_size); 2256 DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size); 2257 2258 const addr_t nlist_data_byte_size = 2259 symtab_load_command.nsyms * nlist_byte_size; 2260 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2261 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2262 2263 ProcessSP process_sp(m_process_wp.lock()); 2264 Process *process = process_sp.get(); 2265 2266 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2267 2268 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2269 Target &target = process->GetTarget(); 2270 2271 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2272 2273 SectionSP linkedit_section_sp( 2274 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2275 // Reading mach file from memory in a process or core file... 2276 2277 if (linkedit_section_sp) { 2278 addr_t linkedit_load_addr = 2279 linkedit_section_sp->GetLoadBaseAddress(&target); 2280 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2281 // We might be trying to access the symbol table before the 2282 // __LINKEDIT's load address has been set in the target. We can't 2283 // fail to read the symbol table, so calculate the right address 2284 // manually 2285 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2286 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2287 } 2288 2289 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2290 const addr_t symoff_addr = linkedit_load_addr + 2291 symtab_load_command.symoff - 2292 linkedit_file_offset; 2293 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2294 linkedit_file_offset; 2295 2296 bool data_was_read = false; 2297 2298 #if defined(__APPLE__) && \ 2299 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2300 if (m_header.flags & MH_DYLIB_IN_CACHE && 2301 process->GetAddressByteSize() == sizeof(void *)) { 2302 // This mach-o memory file is in the dyld shared cache. If this 2303 // program is not remote and this is iOS, then this process will 2304 // share the same shared cache as the process we are debugging and we 2305 // can read the entire __LINKEDIT from the address space in this 2306 // process. This is a needed optimization that is used for local iOS 2307 // debugging only since all shared libraries in the shared cache do 2308 // not have corresponding files that exist in the file system of the 2309 // device. They have been combined into a single file. This means we 2310 // always have to load these files from memory. All of the symbol and 2311 // string tables from all of the __LINKEDIT sections from the shared 2312 // libraries in the shared cache have been merged into a single large 2313 // symbol and string table. Reading all of this symbol and string 2314 // table data across can slow down debug launch times, so we optimize 2315 // this by reading the memory for the __LINKEDIT section from this 2316 // process. 2317 2318 UUID lldb_shared_cache; 2319 addr_t lldb_shared_cache_addr; 2320 GetLLDBSharedCacheUUID(lldb_shared_cache_addr, lldb_shared_cache); 2321 UUID process_shared_cache; 2322 addr_t process_shared_cache_addr; 2323 GetProcessSharedCacheUUID(process, process_shared_cache_addr, 2324 process_shared_cache); 2325 bool use_lldb_cache = true; 2326 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2327 (lldb_shared_cache != process_shared_cache || 2328 process_shared_cache_addr != lldb_shared_cache_addr)) { 2329 use_lldb_cache = false; 2330 } 2331 2332 PlatformSP platform_sp(target.GetPlatform()); 2333 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2334 data_was_read = true; 2335 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2336 eByteOrderLittle); 2337 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2338 eByteOrderLittle); 2339 if (function_starts_load_command.cmd) { 2340 const addr_t func_start_addr = 2341 linkedit_load_addr + function_starts_load_command.dataoff - 2342 linkedit_file_offset; 2343 function_starts_data.SetData((void *)func_start_addr, 2344 function_starts_load_command.datasize, 2345 eByteOrderLittle); 2346 } 2347 } 2348 } 2349 #endif 2350 2351 if (!data_was_read) { 2352 // Always load dyld - the dynamic linker - from memory if we didn't 2353 // find a binary anywhere else. lldb will not register 2354 // dylib/framework/bundle loads/unloads if we don't have the dyld 2355 // symbols, we force dyld to load from memory despite the user's 2356 // target.memory-module-load-level setting. 2357 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2358 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2359 DataBufferSP nlist_data_sp( 2360 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2361 if (nlist_data_sp) 2362 nlist_data.SetData(nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2363 if (m_dysymtab.nindirectsyms != 0) { 2364 const addr_t indirect_syms_addr = linkedit_load_addr + 2365 m_dysymtab.indirectsymoff - 2366 linkedit_file_offset; 2367 DataBufferSP indirect_syms_data_sp(ReadMemory( 2368 process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 2369 if (indirect_syms_data_sp) 2370 indirect_symbol_index_data.SetData( 2371 indirect_syms_data_sp, 0, 2372 indirect_syms_data_sp->GetByteSize()); 2373 // If this binary is outside the shared cache, 2374 // cache the string table. 2375 // Binaries in the shared cache all share a giant string table, 2376 // and we can't share the string tables across multiple 2377 // ObjectFileMachO's, so we'd end up re-reading this mega-strtab 2378 // for every binary in the shared cache - it would be a big perf 2379 // problem. For binaries outside the shared cache, it's faster to 2380 // read the entire strtab at once instead of piece-by-piece as we 2381 // process the nlist records. 2382 if ((m_header.flags & MH_DYLIB_IN_CACHE) == 0) { 2383 DataBufferSP strtab_data_sp( 2384 ReadMemory(process_sp, strtab_addr, strtab_data_byte_size)); 2385 if (strtab_data_sp) { 2386 strtab_data.SetData(strtab_data_sp, 0, 2387 strtab_data_sp->GetByteSize()); 2388 } 2389 } 2390 } 2391 } 2392 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) { 2393 if (function_starts_load_command.cmd) { 2394 const addr_t func_start_addr = 2395 linkedit_load_addr + function_starts_load_command.dataoff - 2396 linkedit_file_offset; 2397 DataBufferSP func_start_data_sp( 2398 ReadMemory(process_sp, func_start_addr, 2399 function_starts_load_command.datasize)); 2400 if (func_start_data_sp) 2401 function_starts_data.SetData(func_start_data_sp, 0, 2402 func_start_data_sp->GetByteSize()); 2403 } 2404 } 2405 } 2406 } 2407 } else { 2408 nlist_data.SetData(m_data, symtab_load_command.symoff, 2409 nlist_data_byte_size); 2410 strtab_data.SetData(m_data, symtab_load_command.stroff, 2411 strtab_data_byte_size); 2412 2413 if (dyld_info.export_size > 0) { 2414 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2415 dyld_info.export_size); 2416 } 2417 2418 if (m_dysymtab.nindirectsyms != 0) { 2419 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2420 m_dysymtab.nindirectsyms * 4); 2421 } 2422 if (function_starts_load_command.cmd) { 2423 function_starts_data.SetData(m_data, function_starts_load_command.dataoff, 2424 function_starts_load_command.datasize); 2425 } 2426 } 2427 2428 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2429 2430 ConstString g_segment_name_TEXT = GetSegmentNameTEXT(); 2431 ConstString g_segment_name_DATA = GetSegmentNameDATA(); 2432 ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2433 ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2434 ConstString g_segment_name_OBJC = GetSegmentNameOBJC(); 2435 ConstString g_section_name_eh_frame = GetSectionNameEHFrame(); 2436 SectionSP text_section_sp( 2437 section_list->FindSectionByName(g_segment_name_TEXT)); 2438 SectionSP data_section_sp( 2439 section_list->FindSectionByName(g_segment_name_DATA)); 2440 SectionSP data_dirty_section_sp( 2441 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2442 SectionSP data_const_section_sp( 2443 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2444 SectionSP objc_section_sp( 2445 section_list->FindSectionByName(g_segment_name_OBJC)); 2446 SectionSP eh_frame_section_sp; 2447 if (text_section_sp.get()) 2448 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2449 g_section_name_eh_frame); 2450 else 2451 eh_frame_section_sp = 2452 section_list->FindSectionByName(g_section_name_eh_frame); 2453 2454 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2455 const bool always_thumb = GetArchitecture().IsAlwaysThumbInstructions(); 2456 2457 // lldb works best if it knows the start address of all functions in a 2458 // module. Linker symbols or debug info are normally the best source of 2459 // information for start addr / size but they may be stripped in a released 2460 // binary. Two additional sources of information exist in Mach-O binaries: 2461 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2462 // function's start address in the 2463 // binary, relative to the text section. 2464 // eh_frame - the eh_frame FDEs have the start addr & size of 2465 // each function 2466 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2467 // all modern binaries. 2468 // Binaries built to run on older releases may need to use eh_frame 2469 // information. 2470 2471 if (text_section_sp && function_starts_data.GetByteSize()) { 2472 FunctionStarts::Entry function_start_entry; 2473 function_start_entry.data = false; 2474 lldb::offset_t function_start_offset = 0; 2475 function_start_entry.addr = text_section_sp->GetFileAddress(); 2476 uint64_t delta; 2477 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2478 0) { 2479 // Now append the current entry 2480 function_start_entry.addr += delta; 2481 if (is_arm) { 2482 if (function_start_entry.addr & 1) { 2483 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2484 function_start_entry.data = true; 2485 } else if (always_thumb) { 2486 function_start_entry.data = true; 2487 } 2488 } 2489 function_starts.Append(function_start_entry); 2490 } 2491 } else { 2492 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2493 // load command claiming an eh_frame but it doesn't actually have the 2494 // eh_frame content. And if we have a dSYM, we don't need to do any of 2495 // this fill-in-the-missing-symbols works anyway - the debug info should 2496 // give us all the functions in the module. 2497 if (text_section_sp.get() && eh_frame_section_sp.get() && 2498 m_type != eTypeDebugInfo) { 2499 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2500 DWARFCallFrameInfo::EH); 2501 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2502 eh_frame.GetFunctionAddressAndSizeVector(functions); 2503 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2504 size_t count = functions.GetSize(); 2505 for (size_t i = 0; i < count; ++i) { 2506 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2507 functions.GetEntryAtIndex(i); 2508 if (func) { 2509 FunctionStarts::Entry function_start_entry; 2510 function_start_entry.addr = func->base - text_base_addr; 2511 if (is_arm) { 2512 if (function_start_entry.addr & 1) { 2513 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; 2514 function_start_entry.data = true; 2515 } else if (always_thumb) { 2516 function_start_entry.data = true; 2517 } 2518 } 2519 function_starts.Append(function_start_entry); 2520 } 2521 } 2522 } 2523 } 2524 2525 const size_t function_starts_count = function_starts.GetSize(); 2526 2527 // For user process binaries (executables, dylibs, frameworks, bundles), if 2528 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2529 // going to assume the binary has been stripped. Don't allow assembly 2530 // language instruction emulation because we don't know proper function 2531 // start boundaries. 2532 // 2533 // For all other types of binaries (kernels, stand-alone bare board 2534 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2535 // sections - we should not make any assumptions about them based on that. 2536 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2537 m_allow_assembly_emulation_unwind_plans = false; 2538 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2539 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2540 2541 if (unwind_or_symbol_log) 2542 module_sp->LogMessage( 2543 unwind_or_symbol_log, 2544 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2545 } 2546 2547 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() 2548 ? eh_frame_section_sp->GetID() 2549 : static_cast<user_id_t>(NO_SECT); 2550 2551 lldb::offset_t nlist_data_offset = 0; 2552 2553 uint32_t N_SO_index = UINT32_MAX; 2554 2555 MachSymtabSectionInfo section_info(section_list); 2556 std::vector<uint32_t> N_FUN_indexes; 2557 std::vector<uint32_t> N_NSYM_indexes; 2558 std::vector<uint32_t> N_INCL_indexes; 2559 std::vector<uint32_t> N_BRAC_indexes; 2560 std::vector<uint32_t> N_COMM_indexes; 2561 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2562 typedef llvm::DenseMap<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2563 typedef llvm::DenseMap<const char *, uint32_t> ConstNameToSymbolIndexMap; 2564 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2565 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2566 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2567 // Any symbols that get merged into another will get an entry in this map 2568 // so we know 2569 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2570 uint32_t nlist_idx = 0; 2571 Symbol *symbol_ptr = nullptr; 2572 2573 uint32_t sym_idx = 0; 2574 Symbol *sym = nullptr; 2575 size_t num_syms = 0; 2576 std::string memory_symbol_name; 2577 uint32_t unmapped_local_symbols_found = 0; 2578 2579 std::vector<TrieEntryWithOffset> reexport_trie_entries; 2580 std::vector<TrieEntryWithOffset> external_sym_trie_entries; 2581 std::set<lldb::addr_t> resolver_addresses; 2582 2583 if (dyld_trie_data.GetByteSize() > 0) { 2584 ConstString text_segment_name("__TEXT"); 2585 SectionSP text_segment_sp = 2586 GetSectionList()->FindSectionByName(text_segment_name); 2587 lldb::addr_t text_segment_file_addr = LLDB_INVALID_ADDRESS; 2588 if (text_segment_sp) 2589 text_segment_file_addr = text_segment_sp->GetFileAddress(); 2590 std::vector<llvm::StringRef> nameSlices; 2591 ParseTrieEntries(dyld_trie_data, 0, is_arm, text_segment_file_addr, 2592 nameSlices, resolver_addresses, reexport_trie_entries, 2593 external_sym_trie_entries); 2594 } 2595 2596 typedef std::set<ConstString> IndirectSymbols; 2597 IndirectSymbols indirect_symbol_names; 2598 2599 #if defined(__APPLE__) && TARGET_OS_EMBEDDED 2600 2601 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2602 // optimized by moving LOCAL symbols out of the memory mapped portion of 2603 // the DSC. The symbol information has all been retained, but it isn't 2604 // available in the normal nlist data. However, there *are* duplicate 2605 // entries of *some* 2606 // LOCAL symbols in the normal nlist data. To handle this situation 2607 // correctly, we must first attempt 2608 // to parse any DSC unmapped symbol information. If we find any, we set a 2609 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2610 2611 if (m_header.flags & MH_DYLIB_IN_CACHE) { 2612 // Before we can start mapping the DSC, we need to make certain the 2613 // target process is actually using the cache we can find. 2614 2615 // Next we need to determine the correct path for the dyld shared cache. 2616 2617 ArchSpec header_arch = GetArchitecture(); 2618 char dsc_path[PATH_MAX]; 2619 char dsc_path_development[PATH_MAX]; 2620 2621 snprintf( 2622 dsc_path, sizeof(dsc_path), "%s%s%s", 2623 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2624 */ 2625 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2626 header_arch.GetArchitectureName()); 2627 2628 snprintf( 2629 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2630 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2631 */ 2632 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2633 header_arch.GetArchitectureName(), ".development"); 2634 2635 FileSpec dsc_nondevelopment_filespec(dsc_path); 2636 FileSpec dsc_development_filespec(dsc_path_development); 2637 FileSpec dsc_filespec; 2638 2639 UUID dsc_uuid; 2640 UUID process_shared_cache_uuid; 2641 addr_t process_shared_cache_base_addr; 2642 2643 if (process) { 2644 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, 2645 process_shared_cache_uuid); 2646 } 2647 2648 // First see if we can find an exact match for the inferior process 2649 // shared cache UUID in the development or non-development shared caches 2650 // on disk. 2651 if (process_shared_cache_uuid.IsValid()) { 2652 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2653 UUID dsc_development_uuid = GetSharedCacheUUID( 2654 dsc_development_filespec, byte_order, addr_byte_size); 2655 if (dsc_development_uuid.IsValid() && 2656 dsc_development_uuid == process_shared_cache_uuid) { 2657 dsc_filespec = dsc_development_filespec; 2658 dsc_uuid = dsc_development_uuid; 2659 } 2660 } 2661 if (!dsc_uuid.IsValid() && 2662 FileSystem::Instance().Exists(dsc_nondevelopment_filespec)) { 2663 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2664 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2665 if (dsc_nondevelopment_uuid.IsValid() && 2666 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2667 dsc_filespec = dsc_nondevelopment_filespec; 2668 dsc_uuid = dsc_nondevelopment_uuid; 2669 } 2670 } 2671 } 2672 2673 // Failing a UUID match, prefer the development dyld_shared cache if both 2674 // are present. 2675 if (!FileSystem::Instance().Exists(dsc_filespec)) { 2676 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2677 dsc_filespec = dsc_development_filespec; 2678 } else { 2679 dsc_filespec = dsc_nondevelopment_filespec; 2680 } 2681 } 2682 2683 /* The dyld_cache_header has a pointer to the 2684 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2685 The dyld_cache_local_symbols_info structure gives us three things: 2686 1. The start and count of the nlist records in the dyld_shared_cache 2687 file 2688 2. The start and size of the strings for these nlist records 2689 3. The start and count of dyld_cache_local_symbols_entry entries 2690 2691 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2692 dyld shared cache. 2693 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2694 the dyld shared cache. 2695 The dyld_cache_local_symbols_entry also lists the start of this 2696 dylib/framework's nlist records 2697 and the count of how many nlist records there are for this 2698 dylib/framework. 2699 */ 2700 2701 // Process the dyld shared cache header to find the unmapped symbols 2702 2703 DataBufferSP dsc_data_sp = MapFileData( 2704 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2705 if (!dsc_uuid.IsValid()) { 2706 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2707 } 2708 if (dsc_data_sp) { 2709 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2710 2711 bool uuid_match = true; 2712 if (dsc_uuid.IsValid() && process) { 2713 if (process_shared_cache_uuid.IsValid() && 2714 dsc_uuid != process_shared_cache_uuid) { 2715 // The on-disk dyld_shared_cache file is not the same as the one in 2716 // this process' memory, don't use it. 2717 uuid_match = false; 2718 ModuleSP module_sp(GetModule()); 2719 if (module_sp) 2720 module_sp->ReportWarning("process shared cache does not match " 2721 "on-disk dyld_shared_cache file, some " 2722 "symbol names will be missing."); 2723 } 2724 } 2725 2726 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2727 2728 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2729 2730 // If the mappingOffset points to a location inside the header, we've 2731 // opened an old dyld shared cache, and should not proceed further. 2732 if (uuid_match && 2733 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2734 2735 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2736 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2737 mappingOffset); 2738 2739 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2740 byte_order, addr_byte_size); 2741 offset = 0; 2742 2743 // The File addresses (from the in-memory Mach-O load commands) for 2744 // the shared libraries in the shared library cache need to be 2745 // adjusted by an offset to match up with the dylibOffset identifying 2746 // field in the dyld_cache_local_symbol_entry's. This offset is 2747 // recorded in mapping_offset_value. 2748 const uint64_t mapping_offset_value = 2749 dsc_mapping_info_data.GetU64(&offset); 2750 2751 offset = 2752 offsetof(struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2753 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2754 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2755 2756 if (localSymbolsOffset && localSymbolsSize) { 2757 // Map the local symbols 2758 DataBufferSP dsc_local_symbols_data_sp = 2759 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2760 2761 if (dsc_local_symbols_data_sp) { 2762 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2763 byte_order, addr_byte_size); 2764 2765 offset = 0; 2766 2767 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 2768 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 2769 UndefinedNameToDescMap undefined_name_to_desc; 2770 SymbolIndexToName reexport_shlib_needs_fixup; 2771 2772 // Read the local_symbols_infos struct in one shot 2773 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2774 dsc_local_symbols_data.GetU32(&offset, 2775 &local_symbols_info.nlistOffset, 6); 2776 2777 SectionSP text_section_sp( 2778 section_list->FindSectionByName(GetSegmentNameTEXT())); 2779 2780 uint32_t header_file_offset = 2781 (text_section_sp->GetFileAddress() - mapping_offset_value); 2782 2783 offset = local_symbols_info.entriesOffset; 2784 for (uint32_t entry_index = 0; 2785 entry_index < local_symbols_info.entriesCount; entry_index++) { 2786 struct lldb_copy_dyld_cache_local_symbols_entry 2787 local_symbols_entry; 2788 local_symbols_entry.dylibOffset = 2789 dsc_local_symbols_data.GetU32(&offset); 2790 local_symbols_entry.nlistStartIndex = 2791 dsc_local_symbols_data.GetU32(&offset); 2792 local_symbols_entry.nlistCount = 2793 dsc_local_symbols_data.GetU32(&offset); 2794 2795 if (header_file_offset == local_symbols_entry.dylibOffset) { 2796 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2797 2798 // The normal nlist code cannot correctly size the Symbols 2799 // array, we need to allocate it here. 2800 sym = symtab->Resize( 2801 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2802 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2803 num_syms = symtab->GetNumSymbols(); 2804 2805 nlist_data_offset = 2806 local_symbols_info.nlistOffset + 2807 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2808 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2809 2810 for (uint32_t nlist_index = 0; 2811 nlist_index < local_symbols_entry.nlistCount; 2812 nlist_index++) { 2813 ///////////////////////////// 2814 { 2815 llvm::Optional<struct nlist_64> nlist_maybe = 2816 ParseNList(dsc_local_symbols_data, nlist_data_offset, 2817 nlist_byte_size); 2818 if (!nlist_maybe) 2819 break; 2820 struct nlist_64 nlist = *nlist_maybe; 2821 2822 SymbolType type = eSymbolTypeInvalid; 2823 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2824 string_table_offset + nlist.n_strx); 2825 2826 if (symbol_name == NULL) { 2827 // No symbol should be NULL, even the symbols with no 2828 // string values should have an offset zero which 2829 // points to an empty C-string 2830 Host::SystemLog( 2831 Host::eSystemLogError, 2832 "error: DSC unmapped local symbol[%u] has invalid " 2833 "string table offset 0x%x in %s, ignoring symbol\n", 2834 entry_index, nlist.n_strx, 2835 module_sp->GetFileSpec().GetPath().c_str()); 2836 continue; 2837 } 2838 if (symbol_name[0] == '\0') 2839 symbol_name = NULL; 2840 2841 const char *symbol_name_non_abi_mangled = NULL; 2842 2843 SectionSP symbol_section; 2844 uint32_t symbol_byte_size = 0; 2845 bool add_nlist = true; 2846 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2847 bool demangled_is_synthesized = false; 2848 bool is_gsym = false; 2849 bool set_value = true; 2850 2851 assert(sym_idx < num_syms); 2852 2853 sym[sym_idx].SetDebug(is_debug); 2854 2855 if (is_debug) { 2856 switch (nlist.n_type) { 2857 case N_GSYM: 2858 // global symbol: name,,NO_SECT,type,0 2859 // Sometimes the N_GSYM value contains the address. 2860 2861 // FIXME: In the .o files, we have a GSYM and a debug 2862 // symbol for all the ObjC data. They 2863 // have the same address, but we want to ensure that 2864 // we always find only the real symbol, 'cause we 2865 // don't currently correctly attribute the 2866 // GSYM one to the ObjCClass/Ivar/MetaClass 2867 // symbol type. This is a temporary hack to make 2868 // sure the ObjectiveC symbols get treated correctly. 2869 // To do this right, we should coalesce all the GSYM 2870 // & global symbols that have the same address. 2871 2872 is_gsym = true; 2873 sym[sym_idx].SetExternal(true); 2874 2875 if (symbol_name && symbol_name[0] == '_' && 2876 symbol_name[1] == 'O') { 2877 llvm::StringRef symbol_name_ref(symbol_name); 2878 if (symbol_name_ref.startswith( 2879 g_objc_v2_prefix_class)) { 2880 symbol_name_non_abi_mangled = symbol_name + 1; 2881 symbol_name = 2882 symbol_name + g_objc_v2_prefix_class.size(); 2883 type = eSymbolTypeObjCClass; 2884 demangled_is_synthesized = true; 2885 2886 } else if (symbol_name_ref.startswith( 2887 g_objc_v2_prefix_metaclass)) { 2888 symbol_name_non_abi_mangled = symbol_name + 1; 2889 symbol_name = 2890 symbol_name + g_objc_v2_prefix_metaclass.size(); 2891 type = eSymbolTypeObjCMetaClass; 2892 demangled_is_synthesized = true; 2893 } else if (symbol_name_ref.startswith( 2894 g_objc_v2_prefix_ivar)) { 2895 symbol_name_non_abi_mangled = symbol_name + 1; 2896 symbol_name = 2897 symbol_name + g_objc_v2_prefix_ivar.size(); 2898 type = eSymbolTypeObjCIVar; 2899 demangled_is_synthesized = true; 2900 } 2901 } else { 2902 if (nlist.n_value != 0) 2903 symbol_section = section_info.GetSection( 2904 nlist.n_sect, nlist.n_value); 2905 type = eSymbolTypeData; 2906 } 2907 break; 2908 2909 case N_FNAME: 2910 // procedure name (f77 kludge): name,,NO_SECT,0,0 2911 type = eSymbolTypeCompiler; 2912 break; 2913 2914 case N_FUN: 2915 // procedure: name,,n_sect,linenumber,address 2916 if (symbol_name) { 2917 type = eSymbolTypeCode; 2918 symbol_section = section_info.GetSection( 2919 nlist.n_sect, nlist.n_value); 2920 2921 N_FUN_addr_to_sym_idx.insert( 2922 std::make_pair(nlist.n_value, sym_idx)); 2923 // We use the current number of symbols in the 2924 // symbol table in lieu of using nlist_idx in case 2925 // we ever start trimming entries out 2926 N_FUN_indexes.push_back(sym_idx); 2927 } else { 2928 type = eSymbolTypeCompiler; 2929 2930 if (!N_FUN_indexes.empty()) { 2931 // Copy the size of the function into the 2932 // original 2933 // STAB entry so we don't have 2934 // to hunt for it later 2935 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2936 ->SetByteSize(nlist.n_value); 2937 N_FUN_indexes.pop_back(); 2938 // We don't really need the end function STAB as 2939 // it contains the size which we already placed 2940 // with the original symbol, so don't add it if 2941 // we want a minimal symbol table 2942 add_nlist = false; 2943 } 2944 } 2945 break; 2946 2947 case N_STSYM: 2948 // static symbol: name,,n_sect,type,address 2949 N_STSYM_addr_to_sym_idx.insert( 2950 std::make_pair(nlist.n_value, sym_idx)); 2951 symbol_section = section_info.GetSection(nlist.n_sect, 2952 nlist.n_value); 2953 if (symbol_name && symbol_name[0]) { 2954 type = ObjectFile::GetSymbolTypeFromName( 2955 symbol_name + 1, eSymbolTypeData); 2956 } 2957 break; 2958 2959 case N_LCSYM: 2960 // .lcomm symbol: name,,n_sect,type,address 2961 symbol_section = section_info.GetSection(nlist.n_sect, 2962 nlist.n_value); 2963 type = eSymbolTypeCommonBlock; 2964 break; 2965 2966 case N_BNSYM: 2967 // We use the current number of symbols in the symbol 2968 // table in lieu of using nlist_idx in case we ever 2969 // start trimming entries out Skip these if we want 2970 // minimal symbol tables 2971 add_nlist = false; 2972 break; 2973 2974 case N_ENSYM: 2975 // Set the size of the N_BNSYM to the terminating 2976 // index of this N_ENSYM so that we can always skip 2977 // the entire symbol if we need to navigate more 2978 // quickly at the source level when parsing STABS 2979 // Skip these if we want minimal symbol tables 2980 add_nlist = false; 2981 break; 2982 2983 case N_OPT: 2984 // emitted with gcc2_compiled and in gcc source 2985 type = eSymbolTypeCompiler; 2986 break; 2987 2988 case N_RSYM: 2989 // register sym: name,,NO_SECT,type,register 2990 type = eSymbolTypeVariable; 2991 break; 2992 2993 case N_SLINE: 2994 // src line: 0,,n_sect,linenumber,address 2995 symbol_section = section_info.GetSection(nlist.n_sect, 2996 nlist.n_value); 2997 type = eSymbolTypeLineEntry; 2998 break; 2999 3000 case N_SSYM: 3001 // structure elt: name,,NO_SECT,type,struct_offset 3002 type = eSymbolTypeVariableType; 3003 break; 3004 3005 case N_SO: 3006 // source file name 3007 type = eSymbolTypeSourceFile; 3008 if (symbol_name == NULL) { 3009 add_nlist = false; 3010 if (N_SO_index != UINT32_MAX) { 3011 // Set the size of the N_SO to the terminating 3012 // index of this N_SO so that we can always skip 3013 // the entire N_SO if we need to navigate more 3014 // quickly at the source level when parsing STABS 3015 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3016 symbol_ptr->SetByteSize(sym_idx); 3017 symbol_ptr->SetSizeIsSibling(true); 3018 } 3019 N_NSYM_indexes.clear(); 3020 N_INCL_indexes.clear(); 3021 N_BRAC_indexes.clear(); 3022 N_COMM_indexes.clear(); 3023 N_FUN_indexes.clear(); 3024 N_SO_index = UINT32_MAX; 3025 } else { 3026 // We use the current number of symbols in the 3027 // symbol table in lieu of using nlist_idx in case 3028 // we ever start trimming entries out 3029 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3030 if (N_SO_has_full_path) { 3031 if ((N_SO_index == sym_idx - 1) && 3032 ((sym_idx - 1) < num_syms)) { 3033 // We have two consecutive N_SO entries where 3034 // the first contains a directory and the 3035 // second contains a full path. 3036 sym[sym_idx - 1].GetMangled().SetValue( 3037 ConstString(symbol_name), false); 3038 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3039 add_nlist = false; 3040 } else { 3041 // This is the first entry in a N_SO that 3042 // contains a directory or 3043 // a full path to the source file 3044 N_SO_index = sym_idx; 3045 } 3046 } else if ((N_SO_index == sym_idx - 1) && 3047 ((sym_idx - 1) < num_syms)) { 3048 // This is usually the second N_SO entry that 3049 // contains just the filename, so here we combine 3050 // it with the first one if we are minimizing the 3051 // symbol table 3052 const char *so_path = 3053 sym[sym_idx - 1] 3054 .GetMangled() 3055 .GetDemangledName( 3056 lldb::eLanguageTypeUnknown) 3057 .AsCString(); 3058 if (so_path && so_path[0]) { 3059 std::string full_so_path(so_path); 3060 const size_t double_slash_pos = 3061 full_so_path.find("//"); 3062 if (double_slash_pos != std::string::npos) { 3063 // The linker has been generating bad N_SO 3064 // entries with doubled up paths 3065 // in the format "%s%s" where the first 3066 // string in the DW_AT_comp_dir, and the 3067 // second is the directory for the source 3068 // file so you end up with a path that looks 3069 // like "/tmp/src//tmp/src/" 3070 FileSpec so_dir(so_path); 3071 if (!FileSystem::Instance().Exists(so_dir)) { 3072 so_dir.SetFile( 3073 &full_so_path[double_slash_pos + 1], 3074 FileSpec::Style::native); 3075 if (FileSystem::Instance().Exists(so_dir)) { 3076 // Trim off the incorrect path 3077 full_so_path.erase(0, double_slash_pos + 1); 3078 } 3079 } 3080 } 3081 if (*full_so_path.rbegin() != '/') 3082 full_so_path += '/'; 3083 full_so_path += symbol_name; 3084 sym[sym_idx - 1].GetMangled().SetValue( 3085 ConstString(full_so_path.c_str()), false); 3086 add_nlist = false; 3087 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3088 } 3089 } else { 3090 // This could be a relative path to a N_SO 3091 N_SO_index = sym_idx; 3092 } 3093 } 3094 break; 3095 3096 case N_OSO: 3097 // object file name: name,,0,0,st_mtime 3098 type = eSymbolTypeObjectFile; 3099 break; 3100 3101 case N_LSYM: 3102 // local sym: name,,NO_SECT,type,offset 3103 type = eSymbolTypeLocal; 3104 break; 3105 3106 // INCL scopes 3107 case N_BINCL: 3108 // include file beginning: name,,NO_SECT,0,sum We use 3109 // the current number of symbols in the symbol table 3110 // in lieu of using nlist_idx in case we ever start 3111 // trimming entries out 3112 N_INCL_indexes.push_back(sym_idx); 3113 type = eSymbolTypeScopeBegin; 3114 break; 3115 3116 case N_EINCL: 3117 // include file end: name,,NO_SECT,0,0 3118 // Set the size of the N_BINCL to the terminating 3119 // index of this N_EINCL so that we can always skip 3120 // the entire symbol if we need to navigate more 3121 // quickly at the source level when parsing STABS 3122 if (!N_INCL_indexes.empty()) { 3123 symbol_ptr = 3124 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3125 symbol_ptr->SetByteSize(sym_idx + 1); 3126 symbol_ptr->SetSizeIsSibling(true); 3127 N_INCL_indexes.pop_back(); 3128 } 3129 type = eSymbolTypeScopeEnd; 3130 break; 3131 3132 case N_SOL: 3133 // #included file name: name,,n_sect,0,address 3134 type = eSymbolTypeHeaderFile; 3135 3136 // We currently don't use the header files on darwin 3137 add_nlist = false; 3138 break; 3139 3140 case N_PARAMS: 3141 // compiler parameters: name,,NO_SECT,0,0 3142 type = eSymbolTypeCompiler; 3143 break; 3144 3145 case N_VERSION: 3146 // compiler version: name,,NO_SECT,0,0 3147 type = eSymbolTypeCompiler; 3148 break; 3149 3150 case N_OLEVEL: 3151 // compiler -O level: name,,NO_SECT,0,0 3152 type = eSymbolTypeCompiler; 3153 break; 3154 3155 case N_PSYM: 3156 // parameter: name,,NO_SECT,type,offset 3157 type = eSymbolTypeVariable; 3158 break; 3159 3160 case N_ENTRY: 3161 // alternate entry: name,,n_sect,linenumber,address 3162 symbol_section = section_info.GetSection(nlist.n_sect, 3163 nlist.n_value); 3164 type = eSymbolTypeLineEntry; 3165 break; 3166 3167 // Left and Right Braces 3168 case N_LBRAC: 3169 // left bracket: 0,,NO_SECT,nesting level,address We 3170 // use the current number of symbols in the symbol 3171 // table in lieu of using nlist_idx in case we ever 3172 // start trimming entries out 3173 symbol_section = section_info.GetSection(nlist.n_sect, 3174 nlist.n_value); 3175 N_BRAC_indexes.push_back(sym_idx); 3176 type = eSymbolTypeScopeBegin; 3177 break; 3178 3179 case N_RBRAC: 3180 // right bracket: 0,,NO_SECT,nesting level,address 3181 // Set the size of the N_LBRAC to the terminating 3182 // index of this N_RBRAC so that we can always skip 3183 // the entire symbol if we need to navigate more 3184 // quickly at the source level when parsing STABS 3185 symbol_section = section_info.GetSection(nlist.n_sect, 3186 nlist.n_value); 3187 if (!N_BRAC_indexes.empty()) { 3188 symbol_ptr = 3189 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3190 symbol_ptr->SetByteSize(sym_idx + 1); 3191 symbol_ptr->SetSizeIsSibling(true); 3192 N_BRAC_indexes.pop_back(); 3193 } 3194 type = eSymbolTypeScopeEnd; 3195 break; 3196 3197 case N_EXCL: 3198 // deleted include file: name,,NO_SECT,0,sum 3199 type = eSymbolTypeHeaderFile; 3200 break; 3201 3202 // COMM scopes 3203 case N_BCOMM: 3204 // begin common: name,,NO_SECT,0,0 3205 // We use the current number of symbols in the symbol 3206 // table in lieu of using nlist_idx in case we ever 3207 // start trimming entries out 3208 type = eSymbolTypeScopeBegin; 3209 N_COMM_indexes.push_back(sym_idx); 3210 break; 3211 3212 case N_ECOML: 3213 // end common (local name): 0,,n_sect,0,address 3214 symbol_section = section_info.GetSection(nlist.n_sect, 3215 nlist.n_value); 3216 // Fall through 3217 3218 case N_ECOMM: 3219 // end common: name,,n_sect,0,0 3220 // Set the size of the N_BCOMM to the terminating 3221 // index of this N_ECOMM/N_ECOML so that we can 3222 // always skip the entire symbol if we need to 3223 // navigate more quickly at the source level when 3224 // parsing STABS 3225 if (!N_COMM_indexes.empty()) { 3226 symbol_ptr = 3227 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3228 symbol_ptr->SetByteSize(sym_idx + 1); 3229 symbol_ptr->SetSizeIsSibling(true); 3230 N_COMM_indexes.pop_back(); 3231 } 3232 type = eSymbolTypeScopeEnd; 3233 break; 3234 3235 case N_LENG: 3236 // second stab entry with length information 3237 type = eSymbolTypeAdditional; 3238 break; 3239 3240 default: 3241 break; 3242 } 3243 } else { 3244 // uint8_t n_pext = N_PEXT & nlist.n_type; 3245 uint8_t n_type = N_TYPE & nlist.n_type; 3246 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3247 3248 switch (n_type) { 3249 case N_INDR: { 3250 const char *reexport_name_cstr = 3251 strtab_data.PeekCStr(nlist.n_value); 3252 if (reexport_name_cstr && reexport_name_cstr[0]) { 3253 type = eSymbolTypeReExported; 3254 ConstString reexport_name( 3255 reexport_name_cstr + 3256 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3257 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3258 set_value = false; 3259 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3260 indirect_symbol_names.insert(ConstString( 3261 symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3262 } else 3263 type = eSymbolTypeUndefined; 3264 } break; 3265 3266 case N_UNDF: 3267 if (symbol_name && symbol_name[0]) { 3268 ConstString undefined_name( 3269 symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3270 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3271 } 3272 // Fall through 3273 case N_PBUD: 3274 type = eSymbolTypeUndefined; 3275 break; 3276 3277 case N_ABS: 3278 type = eSymbolTypeAbsolute; 3279 break; 3280 3281 case N_SECT: { 3282 symbol_section = section_info.GetSection(nlist.n_sect, 3283 nlist.n_value); 3284 3285 if (symbol_section == NULL) { 3286 // TODO: warn about this? 3287 add_nlist = false; 3288 break; 3289 } 3290 3291 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3292 type = eSymbolTypeException; 3293 } else { 3294 uint32_t section_type = 3295 symbol_section->Get() & SECTION_TYPE; 3296 3297 switch (section_type) { 3298 case S_CSTRING_LITERALS: 3299 type = eSymbolTypeData; 3300 break; // section with only literal C strings 3301 case S_4BYTE_LITERALS: 3302 type = eSymbolTypeData; 3303 break; // section with only 4 byte literals 3304 case S_8BYTE_LITERALS: 3305 type = eSymbolTypeData; 3306 break; // section with only 8 byte literals 3307 case S_LITERAL_POINTERS: 3308 type = eSymbolTypeTrampoline; 3309 break; // section with only pointers to literals 3310 case S_NON_LAZY_SYMBOL_POINTERS: 3311 type = eSymbolTypeTrampoline; 3312 break; // section with only non-lazy symbol 3313 // pointers 3314 case S_LAZY_SYMBOL_POINTERS: 3315 type = eSymbolTypeTrampoline; 3316 break; // section with only lazy symbol pointers 3317 case S_SYMBOL_STUBS: 3318 type = eSymbolTypeTrampoline; 3319 break; // section with only symbol stubs, byte 3320 // size of stub in the reserved2 field 3321 case S_MOD_INIT_FUNC_POINTERS: 3322 type = eSymbolTypeCode; 3323 break; // section with only function pointers for 3324 // initialization 3325 case S_MOD_TERM_FUNC_POINTERS: 3326 type = eSymbolTypeCode; 3327 break; // section with only function pointers for 3328 // termination 3329 case S_INTERPOSING: 3330 type = eSymbolTypeTrampoline; 3331 break; // section with only pairs of function 3332 // pointers for interposing 3333 case S_16BYTE_LITERALS: 3334 type = eSymbolTypeData; 3335 break; // section with only 16 byte literals 3336 case S_DTRACE_DOF: 3337 type = eSymbolTypeInstrumentation; 3338 break; 3339 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3340 type = eSymbolTypeTrampoline; 3341 break; 3342 default: 3343 switch (symbol_section->GetType()) { 3344 case lldb::eSectionTypeCode: 3345 type = eSymbolTypeCode; 3346 break; 3347 case eSectionTypeData: 3348 case eSectionTypeDataCString: // Inlined C string 3349 // data 3350 case eSectionTypeDataCStringPointers: // Pointers 3351 // to C 3352 // string 3353 // data 3354 case eSectionTypeDataSymbolAddress: // Address of 3355 // a symbol in 3356 // the symbol 3357 // table 3358 case eSectionTypeData4: 3359 case eSectionTypeData8: 3360 case eSectionTypeData16: 3361 type = eSymbolTypeData; 3362 break; 3363 default: 3364 break; 3365 } 3366 break; 3367 } 3368 3369 if (type == eSymbolTypeInvalid) { 3370 const char *symbol_sect_name = 3371 symbol_section->GetName().AsCString(); 3372 if (symbol_section->IsDescendant( 3373 text_section_sp.get())) { 3374 if (symbol_section->IsClear( 3375 S_ATTR_PURE_INSTRUCTIONS | 3376 S_ATTR_SELF_MODIFYING_CODE | 3377 S_ATTR_SOME_INSTRUCTIONS)) 3378 type = eSymbolTypeData; 3379 else 3380 type = eSymbolTypeCode; 3381 } else if (symbol_section->IsDescendant( 3382 data_section_sp.get()) || 3383 symbol_section->IsDescendant( 3384 data_dirty_section_sp.get()) || 3385 symbol_section->IsDescendant( 3386 data_const_section_sp.get())) { 3387 if (symbol_sect_name && 3388 ::strstr(symbol_sect_name, "__objc") == 3389 symbol_sect_name) { 3390 type = eSymbolTypeRuntime; 3391 3392 if (symbol_name) { 3393 llvm::StringRef symbol_name_ref(symbol_name); 3394 if (symbol_name_ref.startswith("_OBJC_")) { 3395 llvm::StringRef 3396 g_objc_v2_prefix_class( 3397 "_OBJC_CLASS_$_"); 3398 llvm::StringRef 3399 g_objc_v2_prefix_metaclass( 3400 "_OBJC_METACLASS_$_"); 3401 llvm::StringRef 3402 g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 3403 if (symbol_name_ref.startswith( 3404 g_objc_v2_prefix_class)) { 3405 symbol_name_non_abi_mangled = 3406 symbol_name + 1; 3407 symbol_name = 3408 symbol_name + 3409 g_objc_v2_prefix_class.size(); 3410 type = eSymbolTypeObjCClass; 3411 demangled_is_synthesized = true; 3412 } else if ( 3413 symbol_name_ref.startswith( 3414 g_objc_v2_prefix_metaclass)) { 3415 symbol_name_non_abi_mangled = 3416 symbol_name + 1; 3417 symbol_name = 3418 symbol_name + 3419 g_objc_v2_prefix_metaclass.size(); 3420 type = eSymbolTypeObjCMetaClass; 3421 demangled_is_synthesized = true; 3422 } else if (symbol_name_ref.startswith( 3423 g_objc_v2_prefix_ivar)) { 3424 symbol_name_non_abi_mangled = 3425 symbol_name + 1; 3426 symbol_name = 3427 symbol_name + 3428 g_objc_v2_prefix_ivar.size(); 3429 type = eSymbolTypeObjCIVar; 3430 demangled_is_synthesized = true; 3431 } 3432 } 3433 } 3434 } else if (symbol_sect_name && 3435 ::strstr(symbol_sect_name, 3436 "__gcc_except_tab") == 3437 symbol_sect_name) { 3438 type = eSymbolTypeException; 3439 } else { 3440 type = eSymbolTypeData; 3441 } 3442 } else if (symbol_sect_name && 3443 ::strstr(symbol_sect_name, "__IMPORT") == 3444 symbol_sect_name) { 3445 type = eSymbolTypeTrampoline; 3446 } else if (symbol_section->IsDescendant( 3447 objc_section_sp.get())) { 3448 type = eSymbolTypeRuntime; 3449 if (symbol_name && symbol_name[0] == '.') { 3450 llvm::StringRef symbol_name_ref(symbol_name); 3451 llvm::StringRef 3452 g_objc_v1_prefix_class(".objc_class_name_"); 3453 if (symbol_name_ref.startswith( 3454 g_objc_v1_prefix_class)) { 3455 symbol_name_non_abi_mangled = symbol_name; 3456 symbol_name = symbol_name + 3457 g_objc_v1_prefix_class.size(); 3458 type = eSymbolTypeObjCClass; 3459 demangled_is_synthesized = true; 3460 } 3461 } 3462 } 3463 } 3464 } 3465 } break; 3466 } 3467 } 3468 3469 if (add_nlist) { 3470 uint64_t symbol_value = nlist.n_value; 3471 if (symbol_name_non_abi_mangled) { 3472 sym[sym_idx].GetMangled().SetMangledName( 3473 ConstString(symbol_name_non_abi_mangled)); 3474 sym[sym_idx].GetMangled().SetDemangledName( 3475 ConstString(symbol_name)); 3476 } else { 3477 bool symbol_name_is_mangled = false; 3478 3479 if (symbol_name && symbol_name[0] == '_') { 3480 symbol_name_is_mangled = symbol_name[1] == '_'; 3481 symbol_name++; // Skip the leading underscore 3482 } 3483 3484 if (symbol_name) { 3485 ConstString const_symbol_name(symbol_name); 3486 sym[sym_idx].GetMangled().SetValue( 3487 const_symbol_name, symbol_name_is_mangled); 3488 if (is_gsym && is_debug) { 3489 const char *gsym_name = 3490 sym[sym_idx] 3491 .GetMangled() 3492 .GetName(lldb::eLanguageTypeUnknown, 3493 Mangled::ePreferMangled) 3494 .GetCString(); 3495 if (gsym_name) 3496 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3497 } 3498 } 3499 } 3500 if (symbol_section) { 3501 const addr_t section_file_addr = 3502 symbol_section->GetFileAddress(); 3503 if (symbol_byte_size == 0 && 3504 function_starts_count > 0) { 3505 addr_t symbol_lookup_file_addr = nlist.n_value; 3506 // Do an exact address match for non-ARM addresses, 3507 // else get the closest since the symbol might be a 3508 // thumb symbol which has an address with bit zero 3509 // set 3510 FunctionStarts::Entry *func_start_entry = 3511 function_starts.FindEntry(symbol_lookup_file_addr, 3512 !is_arm); 3513 if (is_arm && func_start_entry) { 3514 // Verify that the function start address is the 3515 // symbol address (ARM) or the symbol address + 1 3516 // (thumb) 3517 if (func_start_entry->addr != 3518 symbol_lookup_file_addr && 3519 func_start_entry->addr != 3520 (symbol_lookup_file_addr + 1)) { 3521 // Not the right entry, NULL it out... 3522 func_start_entry = NULL; 3523 } 3524 } 3525 if (func_start_entry) { 3526 func_start_entry->data = true; 3527 3528 addr_t symbol_file_addr = func_start_entry->addr; 3529 uint32_t symbol_flags = 0; 3530 if (is_arm) { 3531 if (symbol_file_addr & 1) 3532 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3533 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3534 } 3535 3536 const FunctionStarts::Entry *next_func_start_entry = 3537 function_starts.FindNextEntry(func_start_entry); 3538 const addr_t section_end_file_addr = 3539 section_file_addr + 3540 symbol_section->GetByteSize(); 3541 if (next_func_start_entry) { 3542 addr_t next_symbol_file_addr = 3543 next_func_start_entry->addr; 3544 // Be sure the clear the Thumb address bit when 3545 // we calculate the size from the current and 3546 // next address 3547 if (is_arm) 3548 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3549 symbol_byte_size = std::min<lldb::addr_t>( 3550 next_symbol_file_addr - symbol_file_addr, 3551 section_end_file_addr - symbol_file_addr); 3552 } else { 3553 symbol_byte_size = 3554 section_end_file_addr - symbol_file_addr; 3555 } 3556 } 3557 } 3558 symbol_value -= section_file_addr; 3559 } 3560 3561 if (is_debug == false) { 3562 if (type == eSymbolTypeCode) { 3563 // See if we can find a N_FUN entry for any code 3564 // symbols. If we do find a match, and the name 3565 // matches, then we can merge the two into just the 3566 // function symbol to avoid duplicate entries in 3567 // the symbol table 3568 auto range = 3569 N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3570 if (range.first != range.second) { 3571 bool found_it = false; 3572 for (auto pos = range.first; pos != range.second; 3573 ++pos) { 3574 if (sym[sym_idx].GetMangled().GetName( 3575 lldb::eLanguageTypeUnknown, 3576 Mangled::ePreferMangled) == 3577 sym[pos->second].GetMangled().GetName( 3578 lldb::eLanguageTypeUnknown, 3579 Mangled::ePreferMangled)) { 3580 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3581 // We just need the flags from the linker 3582 // symbol, so put these flags 3583 // into the N_FUN flags to avoid duplicate 3584 // symbols in the symbol table 3585 sym[pos->second].SetExternal( 3586 sym[sym_idx].IsExternal()); 3587 sym[pos->second].SetFlags(nlist.n_type << 16 | 3588 nlist.n_desc); 3589 if (resolver_addresses.find(nlist.n_value) != 3590 resolver_addresses.end()) 3591 sym[pos->second].SetType(eSymbolTypeResolver); 3592 sym[sym_idx].Clear(); 3593 found_it = true; 3594 break; 3595 } 3596 } 3597 if (found_it) 3598 continue; 3599 } else { 3600 if (resolver_addresses.find(nlist.n_value) != 3601 resolver_addresses.end()) 3602 type = eSymbolTypeResolver; 3603 } 3604 } else if (type == eSymbolTypeData || 3605 type == eSymbolTypeObjCClass || 3606 type == eSymbolTypeObjCMetaClass || 3607 type == eSymbolTypeObjCIVar) { 3608 // See if we can find a N_STSYM entry for any data 3609 // symbols. If we do find a match, and the name 3610 // matches, then we can merge the two into just the 3611 // Static symbol to avoid duplicate entries in the 3612 // symbol table 3613 auto range = N_STSYM_addr_to_sym_idx.equal_range( 3614 nlist.n_value); 3615 if (range.first != range.second) { 3616 bool found_it = false; 3617 for (auto pos = range.first; pos != range.second; 3618 ++pos) { 3619 if (sym[sym_idx].GetMangled().GetName( 3620 lldb::eLanguageTypeUnknown, 3621 Mangled::ePreferMangled) == 3622 sym[pos->second].GetMangled().GetName( 3623 lldb::eLanguageTypeUnknown, 3624 Mangled::ePreferMangled)) { 3625 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3626 // We just need the flags from the linker 3627 // symbol, so put these flags 3628 // into the N_STSYM flags to avoid duplicate 3629 // symbols in the symbol table 3630 sym[pos->second].SetExternal( 3631 sym[sym_idx].IsExternal()); 3632 sym[pos->second].SetFlags(nlist.n_type << 16 | 3633 nlist.n_desc); 3634 sym[sym_idx].Clear(); 3635 found_it = true; 3636 break; 3637 } 3638 } 3639 if (found_it) 3640 continue; 3641 } else { 3642 const char *gsym_name = 3643 sym[sym_idx] 3644 .GetMangled() 3645 .GetName(lldb::eLanguageTypeUnknown, 3646 Mangled::ePreferMangled) 3647 .GetCString(); 3648 if (gsym_name) { 3649 // Combine N_GSYM stab entries with the non 3650 // stab symbol 3651 ConstNameToSymbolIndexMap::const_iterator pos = 3652 N_GSYM_name_to_sym_idx.find(gsym_name); 3653 if (pos != N_GSYM_name_to_sym_idx.end()) { 3654 const uint32_t GSYM_sym_idx = pos->second; 3655 m_nlist_idx_to_sym_idx[nlist_idx] = 3656 GSYM_sym_idx; 3657 // Copy the address, because often the N_GSYM 3658 // address has an invalid address of zero 3659 // when the global is a common symbol 3660 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3661 symbol_section); 3662 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3663 symbol_value); 3664 symbols_added.insert(sym[GSYM_sym_idx] 3665 .GetAddress() 3666 .GetFileAddress()); 3667 // We just need the flags from the linker 3668 // symbol, so put these flags 3669 // into the N_GSYM flags to avoid duplicate 3670 // symbols in the symbol table 3671 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 3672 nlist.n_desc); 3673 sym[sym_idx].Clear(); 3674 continue; 3675 } 3676 } 3677 } 3678 } 3679 } 3680 3681 sym[sym_idx].SetID(nlist_idx); 3682 sym[sym_idx].SetType(type); 3683 if (set_value) { 3684 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 3685 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3686 symbols_added.insert( 3687 sym[sym_idx].GetAddress().GetFileAddress()); 3688 } 3689 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 3690 3691 if (symbol_byte_size > 0) 3692 sym[sym_idx].SetByteSize(symbol_byte_size); 3693 3694 if (demangled_is_synthesized) 3695 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3696 ++sym_idx; 3697 } else { 3698 sym[sym_idx].Clear(); 3699 } 3700 } 3701 ///////////////////////////// 3702 } 3703 break; // No more entries to consider 3704 } 3705 } 3706 3707 for (const auto &pos : reexport_shlib_needs_fixup) { 3708 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3709 if (undef_pos != undefined_name_to_desc.end()) { 3710 const uint8_t dylib_ordinal = 3711 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3712 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3713 sym[pos.first].SetReExportedSymbolSharedLibrary( 3714 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3715 } 3716 } 3717 } 3718 } 3719 } 3720 } 3721 } 3722 3723 // Must reset this in case it was mutated above! 3724 nlist_data_offset = 0; 3725 #endif 3726 3727 if (nlist_data.GetByteSize() > 0) { 3728 3729 // If the sym array was not created while parsing the DSC unmapped 3730 // symbols, create it now. 3731 if (sym == nullptr) { 3732 sym = 3733 symtab->Resize(symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3734 num_syms = symtab->GetNumSymbols(); 3735 } 3736 3737 if (unmapped_local_symbols_found) { 3738 assert(m_dysymtab.ilocalsym == 0); 3739 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3740 nlist_idx = m_dysymtab.nlocalsym; 3741 } else { 3742 nlist_idx = 0; 3743 } 3744 3745 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 3746 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 3747 UndefinedNameToDescMap undefined_name_to_desc; 3748 SymbolIndexToName reexport_shlib_needs_fixup; 3749 3750 // Symtab parsing is a huge mess. Everything is entangled and the code 3751 // requires access to a ridiculous amount of variables. LLDB depends 3752 // heavily on the proper merging of symbols and to get that right we need 3753 // to make sure we have parsed all the debug symbols first. Therefore we 3754 // invoke the lambda twice, once to parse only the debug symbols and then 3755 // once more to parse the remaining symbols. 3756 auto ParseSymbolLambda = [&](struct nlist_64 &nlist, uint32_t nlist_idx, 3757 bool debug_only) { 3758 const bool is_debug = ((nlist.n_type & N_STAB) != 0); 3759 if (is_debug != debug_only) 3760 return true; 3761 3762 const char *symbol_name_non_abi_mangled = nullptr; 3763 const char *symbol_name = nullptr; 3764 3765 if (have_strtab_data) { 3766 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3767 3768 if (symbol_name == nullptr) { 3769 // No symbol should be NULL, even the symbols with no string values 3770 // should have an offset zero which points to an empty C-string 3771 Host::SystemLog(Host::eSystemLogError, 3772 "error: symbol[%u] has invalid string table offset " 3773 "0x%x in %s, ignoring symbol\n", 3774 nlist_idx, nlist.n_strx, 3775 module_sp->GetFileSpec().GetPath().c_str()); 3776 return true; 3777 } 3778 if (symbol_name[0] == '\0') 3779 symbol_name = nullptr; 3780 } else { 3781 const addr_t str_addr = strtab_addr + nlist.n_strx; 3782 Status str_error; 3783 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3784 str_error)) 3785 symbol_name = memory_symbol_name.c_str(); 3786 } 3787 3788 SymbolType type = eSymbolTypeInvalid; 3789 SectionSP symbol_section; 3790 lldb::addr_t symbol_byte_size = 0; 3791 bool add_nlist = true; 3792 bool is_gsym = false; 3793 bool demangled_is_synthesized = false; 3794 bool set_value = true; 3795 3796 assert(sym_idx < num_syms); 3797 sym[sym_idx].SetDebug(is_debug); 3798 3799 if (is_debug) { 3800 switch (nlist.n_type) { 3801 case N_GSYM: 3802 // global symbol: name,,NO_SECT,type,0 3803 // Sometimes the N_GSYM value contains the address. 3804 3805 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3806 // the ObjC data. They 3807 // have the same address, but we want to ensure that we always find 3808 // only the real symbol, 'cause we don't currently correctly 3809 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3810 // type. This is a temporary hack to make sure the ObjectiveC 3811 // symbols get treated correctly. To do this right, we should 3812 // coalesce all the GSYM & global symbols that have the same 3813 // address. 3814 is_gsym = true; 3815 sym[sym_idx].SetExternal(true); 3816 3817 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3818 llvm::StringRef symbol_name_ref(symbol_name); 3819 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3820 symbol_name_non_abi_mangled = symbol_name + 1; 3821 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3822 type = eSymbolTypeObjCClass; 3823 demangled_is_synthesized = true; 3824 3825 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 3826 symbol_name_non_abi_mangled = symbol_name + 1; 3827 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3828 type = eSymbolTypeObjCMetaClass; 3829 demangled_is_synthesized = true; 3830 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3831 symbol_name_non_abi_mangled = symbol_name + 1; 3832 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3833 type = eSymbolTypeObjCIVar; 3834 demangled_is_synthesized = true; 3835 } 3836 } else { 3837 if (nlist.n_value != 0) 3838 symbol_section = 3839 section_info.GetSection(nlist.n_sect, nlist.n_value); 3840 type = eSymbolTypeData; 3841 } 3842 break; 3843 3844 case N_FNAME: 3845 // procedure name (f77 kludge): name,,NO_SECT,0,0 3846 type = eSymbolTypeCompiler; 3847 break; 3848 3849 case N_FUN: 3850 // procedure: name,,n_sect,linenumber,address 3851 if (symbol_name) { 3852 type = eSymbolTypeCode; 3853 symbol_section = 3854 section_info.GetSection(nlist.n_sect, nlist.n_value); 3855 3856 N_FUN_addr_to_sym_idx.insert( 3857 std::make_pair(nlist.n_value, sym_idx)); 3858 // We use the current number of symbols in the symbol table in 3859 // lieu of using nlist_idx in case we ever start trimming entries 3860 // out 3861 N_FUN_indexes.push_back(sym_idx); 3862 } else { 3863 type = eSymbolTypeCompiler; 3864 3865 if (!N_FUN_indexes.empty()) { 3866 // Copy the size of the function into the original STAB entry 3867 // so we don't have to hunt for it later 3868 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3869 ->SetByteSize(nlist.n_value); 3870 N_FUN_indexes.pop_back(); 3871 // We don't really need the end function STAB as it contains 3872 // the size which we already placed with the original symbol, 3873 // so don't add it if we want a minimal symbol table 3874 add_nlist = false; 3875 } 3876 } 3877 break; 3878 3879 case N_STSYM: 3880 // static symbol: name,,n_sect,type,address 3881 N_STSYM_addr_to_sym_idx.insert( 3882 std::make_pair(nlist.n_value, sym_idx)); 3883 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3884 if (symbol_name && symbol_name[0]) { 3885 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3886 eSymbolTypeData); 3887 } 3888 break; 3889 3890 case N_LCSYM: 3891 // .lcomm symbol: name,,n_sect,type,address 3892 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3893 type = eSymbolTypeCommonBlock; 3894 break; 3895 3896 case N_BNSYM: 3897 // We use the current number of symbols in the symbol table in lieu 3898 // of using nlist_idx in case we ever start trimming entries out 3899 // Skip these if we want minimal symbol tables 3900 add_nlist = false; 3901 break; 3902 3903 case N_ENSYM: 3904 // Set the size of the N_BNSYM to the terminating index of this 3905 // N_ENSYM so that we can always skip the entire symbol if we need 3906 // to navigate more quickly at the source level when parsing STABS 3907 // Skip these if we want minimal symbol tables 3908 add_nlist = false; 3909 break; 3910 3911 case N_OPT: 3912 // emitted with gcc2_compiled and in gcc source 3913 type = eSymbolTypeCompiler; 3914 break; 3915 3916 case N_RSYM: 3917 // register sym: name,,NO_SECT,type,register 3918 type = eSymbolTypeVariable; 3919 break; 3920 3921 case N_SLINE: 3922 // src line: 0,,n_sect,linenumber,address 3923 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3924 type = eSymbolTypeLineEntry; 3925 break; 3926 3927 case N_SSYM: 3928 // structure elt: name,,NO_SECT,type,struct_offset 3929 type = eSymbolTypeVariableType; 3930 break; 3931 3932 case N_SO: 3933 // source file name 3934 type = eSymbolTypeSourceFile; 3935 if (symbol_name == nullptr) { 3936 add_nlist = false; 3937 if (N_SO_index != UINT32_MAX) { 3938 // Set the size of the N_SO to the terminating index of this 3939 // N_SO so that we can always skip the entire N_SO if we need 3940 // to navigate more quickly at the source level when parsing 3941 // STABS 3942 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3943 symbol_ptr->SetByteSize(sym_idx); 3944 symbol_ptr->SetSizeIsSibling(true); 3945 } 3946 N_NSYM_indexes.clear(); 3947 N_INCL_indexes.clear(); 3948 N_BRAC_indexes.clear(); 3949 N_COMM_indexes.clear(); 3950 N_FUN_indexes.clear(); 3951 N_SO_index = UINT32_MAX; 3952 } else { 3953 // We use the current number of symbols in the symbol table in 3954 // lieu of using nlist_idx in case we ever start trimming entries 3955 // out 3956 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3957 if (N_SO_has_full_path) { 3958 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3959 // We have two consecutive N_SO entries where the first 3960 // contains a directory and the second contains a full path. 3961 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), 3962 false); 3963 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3964 add_nlist = false; 3965 } else { 3966 // This is the first entry in a N_SO that contains a 3967 // directory or a full path to the source file 3968 N_SO_index = sym_idx; 3969 } 3970 } else if ((N_SO_index == sym_idx - 1) && 3971 ((sym_idx - 1) < num_syms)) { 3972 // This is usually the second N_SO entry that contains just the 3973 // filename, so here we combine it with the first one if we are 3974 // minimizing the symbol table 3975 const char *so_path = 3976 sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3977 if (so_path && so_path[0]) { 3978 std::string full_so_path(so_path); 3979 const size_t double_slash_pos = full_so_path.find("//"); 3980 if (double_slash_pos != std::string::npos) { 3981 // The linker has been generating bad N_SO entries with 3982 // doubled up paths in the format "%s%s" where the first 3983 // string in the DW_AT_comp_dir, and the second is the 3984 // directory for the source file so you end up with a path 3985 // that looks like "/tmp/src//tmp/src/" 3986 FileSpec so_dir(so_path); 3987 if (!FileSystem::Instance().Exists(so_dir)) { 3988 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 3989 FileSpec::Style::native); 3990 if (FileSystem::Instance().Exists(so_dir)) { 3991 // Trim off the incorrect path 3992 full_so_path.erase(0, double_slash_pos + 1); 3993 } 3994 } 3995 } 3996 if (*full_so_path.rbegin() != '/') 3997 full_so_path += '/'; 3998 full_so_path += symbol_name; 3999 sym[sym_idx - 1].GetMangled().SetValue( 4000 ConstString(full_so_path.c_str()), false); 4001 add_nlist = false; 4002 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4003 } 4004 } else { 4005 // This could be a relative path to a N_SO 4006 N_SO_index = sym_idx; 4007 } 4008 } 4009 break; 4010 4011 case N_OSO: 4012 // object file name: name,,0,0,st_mtime 4013 type = eSymbolTypeObjectFile; 4014 break; 4015 4016 case N_LSYM: 4017 // local sym: name,,NO_SECT,type,offset 4018 type = eSymbolTypeLocal; 4019 break; 4020 4021 // INCL scopes 4022 case N_BINCL: 4023 // include file beginning: name,,NO_SECT,0,sum We use the current 4024 // number of symbols in the symbol table in lieu of using nlist_idx 4025 // in case we ever start trimming entries out 4026 N_INCL_indexes.push_back(sym_idx); 4027 type = eSymbolTypeScopeBegin; 4028 break; 4029 4030 case N_EINCL: 4031 // include file end: name,,NO_SECT,0,0 4032 // Set the size of the N_BINCL to the terminating index of this 4033 // N_EINCL so that we can always skip the entire symbol if we need 4034 // to navigate more quickly at the source level when parsing STABS 4035 if (!N_INCL_indexes.empty()) { 4036 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4037 symbol_ptr->SetByteSize(sym_idx + 1); 4038 symbol_ptr->SetSizeIsSibling(true); 4039 N_INCL_indexes.pop_back(); 4040 } 4041 type = eSymbolTypeScopeEnd; 4042 break; 4043 4044 case N_SOL: 4045 // #included file name: name,,n_sect,0,address 4046 type = eSymbolTypeHeaderFile; 4047 4048 // We currently don't use the header files on darwin 4049 add_nlist = false; 4050 break; 4051 4052 case N_PARAMS: 4053 // compiler parameters: name,,NO_SECT,0,0 4054 type = eSymbolTypeCompiler; 4055 break; 4056 4057 case N_VERSION: 4058 // compiler version: name,,NO_SECT,0,0 4059 type = eSymbolTypeCompiler; 4060 break; 4061 4062 case N_OLEVEL: 4063 // compiler -O level: name,,NO_SECT,0,0 4064 type = eSymbolTypeCompiler; 4065 break; 4066 4067 case N_PSYM: 4068 // parameter: name,,NO_SECT,type,offset 4069 type = eSymbolTypeVariable; 4070 break; 4071 4072 case N_ENTRY: 4073 // alternate entry: name,,n_sect,linenumber,address 4074 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4075 type = eSymbolTypeLineEntry; 4076 break; 4077 4078 // Left and Right Braces 4079 case N_LBRAC: 4080 // left bracket: 0,,NO_SECT,nesting level,address We use the 4081 // current number of symbols in the symbol table in lieu of using 4082 // nlist_idx in case we ever start trimming entries out 4083 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4084 N_BRAC_indexes.push_back(sym_idx); 4085 type = eSymbolTypeScopeBegin; 4086 break; 4087 4088 case N_RBRAC: 4089 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4090 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4091 // can always skip the entire symbol if we need to navigate more 4092 // quickly at the source level when parsing STABS 4093 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4094 if (!N_BRAC_indexes.empty()) { 4095 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4096 symbol_ptr->SetByteSize(sym_idx + 1); 4097 symbol_ptr->SetSizeIsSibling(true); 4098 N_BRAC_indexes.pop_back(); 4099 } 4100 type = eSymbolTypeScopeEnd; 4101 break; 4102 4103 case N_EXCL: 4104 // deleted include file: name,,NO_SECT,0,sum 4105 type = eSymbolTypeHeaderFile; 4106 break; 4107 4108 // COMM scopes 4109 case N_BCOMM: 4110 // begin common: name,,NO_SECT,0,0 4111 // We use the current number of symbols in the symbol table in lieu 4112 // of using nlist_idx in case we ever start trimming entries out 4113 type = eSymbolTypeScopeBegin; 4114 N_COMM_indexes.push_back(sym_idx); 4115 break; 4116 4117 case N_ECOML: 4118 // end common (local name): 0,,n_sect,0,address 4119 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4120 LLVM_FALLTHROUGH; 4121 4122 case N_ECOMM: 4123 // end common: name,,n_sect,0,0 4124 // Set the size of the N_BCOMM to the terminating index of this 4125 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4126 // we need to navigate more quickly at the source level when 4127 // parsing STABS 4128 if (!N_COMM_indexes.empty()) { 4129 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4130 symbol_ptr->SetByteSize(sym_idx + 1); 4131 symbol_ptr->SetSizeIsSibling(true); 4132 N_COMM_indexes.pop_back(); 4133 } 4134 type = eSymbolTypeScopeEnd; 4135 break; 4136 4137 case N_LENG: 4138 // second stab entry with length information 4139 type = eSymbolTypeAdditional; 4140 break; 4141 4142 default: 4143 break; 4144 } 4145 } else { 4146 uint8_t n_type = N_TYPE & nlist.n_type; 4147 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4148 4149 switch (n_type) { 4150 case N_INDR: { 4151 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4152 if (reexport_name_cstr && reexport_name_cstr[0]) { 4153 type = eSymbolTypeReExported; 4154 ConstString reexport_name(reexport_name_cstr + 4155 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4156 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4157 set_value = false; 4158 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4159 indirect_symbol_names.insert( 4160 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4161 } else 4162 type = eSymbolTypeUndefined; 4163 } break; 4164 4165 case N_UNDF: 4166 if (symbol_name && symbol_name[0]) { 4167 ConstString undefined_name(symbol_name + 4168 ((symbol_name[0] == '_') ? 1 : 0)); 4169 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4170 } 4171 LLVM_FALLTHROUGH; 4172 4173 case N_PBUD: 4174 type = eSymbolTypeUndefined; 4175 break; 4176 4177 case N_ABS: 4178 type = eSymbolTypeAbsolute; 4179 break; 4180 4181 case N_SECT: { 4182 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4183 4184 if (!symbol_section) { 4185 // TODO: warn about this? 4186 add_nlist = false; 4187 break; 4188 } 4189 4190 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4191 type = eSymbolTypeException; 4192 } else { 4193 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4194 4195 switch (section_type) { 4196 case S_CSTRING_LITERALS: 4197 type = eSymbolTypeData; 4198 break; // section with only literal C strings 4199 case S_4BYTE_LITERALS: 4200 type = eSymbolTypeData; 4201 break; // section with only 4 byte literals 4202 case S_8BYTE_LITERALS: 4203 type = eSymbolTypeData; 4204 break; // section with only 8 byte literals 4205 case S_LITERAL_POINTERS: 4206 type = eSymbolTypeTrampoline; 4207 break; // section with only pointers to literals 4208 case S_NON_LAZY_SYMBOL_POINTERS: 4209 type = eSymbolTypeTrampoline; 4210 break; // section with only non-lazy symbol pointers 4211 case S_LAZY_SYMBOL_POINTERS: 4212 type = eSymbolTypeTrampoline; 4213 break; // section with only lazy symbol pointers 4214 case S_SYMBOL_STUBS: 4215 type = eSymbolTypeTrampoline; 4216 break; // section with only symbol stubs, byte size of stub in 4217 // the reserved2 field 4218 case S_MOD_INIT_FUNC_POINTERS: 4219 type = eSymbolTypeCode; 4220 break; // section with only function pointers for initialization 4221 case S_MOD_TERM_FUNC_POINTERS: 4222 type = eSymbolTypeCode; 4223 break; // section with only function pointers for termination 4224 case S_INTERPOSING: 4225 type = eSymbolTypeTrampoline; 4226 break; // section with only pairs of function pointers for 4227 // interposing 4228 case S_16BYTE_LITERALS: 4229 type = eSymbolTypeData; 4230 break; // section with only 16 byte literals 4231 case S_DTRACE_DOF: 4232 type = eSymbolTypeInstrumentation; 4233 break; 4234 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4235 type = eSymbolTypeTrampoline; 4236 break; 4237 default: 4238 switch (symbol_section->GetType()) { 4239 case lldb::eSectionTypeCode: 4240 type = eSymbolTypeCode; 4241 break; 4242 case eSectionTypeData: 4243 case eSectionTypeDataCString: // Inlined C string data 4244 case eSectionTypeDataCStringPointers: // Pointers to C string 4245 // data 4246 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4247 // the symbol table 4248 case eSectionTypeData4: 4249 case eSectionTypeData8: 4250 case eSectionTypeData16: 4251 type = eSymbolTypeData; 4252 break; 4253 default: 4254 break; 4255 } 4256 break; 4257 } 4258 4259 if (type == eSymbolTypeInvalid) { 4260 const char *symbol_sect_name = 4261 symbol_section->GetName().AsCString(); 4262 if (symbol_section->IsDescendant(text_section_sp.get())) { 4263 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4264 S_ATTR_SELF_MODIFYING_CODE | 4265 S_ATTR_SOME_INSTRUCTIONS)) 4266 type = eSymbolTypeData; 4267 else 4268 type = eSymbolTypeCode; 4269 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 4270 symbol_section->IsDescendant( 4271 data_dirty_section_sp.get()) || 4272 symbol_section->IsDescendant( 4273 data_const_section_sp.get())) { 4274 if (symbol_sect_name && 4275 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 4276 type = eSymbolTypeRuntime; 4277 4278 if (symbol_name) { 4279 llvm::StringRef symbol_name_ref(symbol_name); 4280 if (symbol_name_ref.startswith("_OBJC_")) { 4281 llvm::StringRef g_objc_v2_prefix_class( 4282 "_OBJC_CLASS_$_"); 4283 llvm::StringRef g_objc_v2_prefix_metaclass( 4284 "_OBJC_METACLASS_$_"); 4285 llvm::StringRef g_objc_v2_prefix_ivar( 4286 "_OBJC_IVAR_$_"); 4287 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 4288 symbol_name_non_abi_mangled = symbol_name + 1; 4289 symbol_name = 4290 symbol_name + g_objc_v2_prefix_class.size(); 4291 type = eSymbolTypeObjCClass; 4292 demangled_is_synthesized = true; 4293 } else if (symbol_name_ref.startswith( 4294 g_objc_v2_prefix_metaclass)) { 4295 symbol_name_non_abi_mangled = symbol_name + 1; 4296 symbol_name = 4297 symbol_name + g_objc_v2_prefix_metaclass.size(); 4298 type = eSymbolTypeObjCMetaClass; 4299 demangled_is_synthesized = true; 4300 } else if (symbol_name_ref.startswith( 4301 g_objc_v2_prefix_ivar)) { 4302 symbol_name_non_abi_mangled = symbol_name + 1; 4303 symbol_name = 4304 symbol_name + g_objc_v2_prefix_ivar.size(); 4305 type = eSymbolTypeObjCIVar; 4306 demangled_is_synthesized = true; 4307 } 4308 } 4309 } 4310 } else if (symbol_sect_name && 4311 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4312 symbol_sect_name) { 4313 type = eSymbolTypeException; 4314 } else { 4315 type = eSymbolTypeData; 4316 } 4317 } else if (symbol_sect_name && 4318 ::strstr(symbol_sect_name, "__IMPORT") == 4319 symbol_sect_name) { 4320 type = eSymbolTypeTrampoline; 4321 } else if (symbol_section->IsDescendant(objc_section_sp.get())) { 4322 type = eSymbolTypeRuntime; 4323 if (symbol_name && symbol_name[0] == '.') { 4324 llvm::StringRef symbol_name_ref(symbol_name); 4325 llvm::StringRef g_objc_v1_prefix_class( 4326 ".objc_class_name_"); 4327 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4328 symbol_name_non_abi_mangled = symbol_name; 4329 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4330 type = eSymbolTypeObjCClass; 4331 demangled_is_synthesized = true; 4332 } 4333 } 4334 } 4335 } 4336 } 4337 } break; 4338 } 4339 } 4340 4341 if (!add_nlist) { 4342 sym[sym_idx].Clear(); 4343 return true; 4344 } 4345 4346 uint64_t symbol_value = nlist.n_value; 4347 4348 if (symbol_name_non_abi_mangled) { 4349 sym[sym_idx].GetMangled().SetMangledName( 4350 ConstString(symbol_name_non_abi_mangled)); 4351 sym[sym_idx].GetMangled().SetDemangledName(ConstString(symbol_name)); 4352 } else { 4353 bool symbol_name_is_mangled = false; 4354 4355 if (symbol_name && symbol_name[0] == '_') { 4356 symbol_name_is_mangled = symbol_name[1] == '_'; 4357 symbol_name++; // Skip the leading underscore 4358 } 4359 4360 if (symbol_name) { 4361 ConstString const_symbol_name(symbol_name); 4362 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4363 symbol_name_is_mangled); 4364 } 4365 } 4366 4367 if (is_gsym) { 4368 const char *gsym_name = sym[sym_idx] 4369 .GetMangled() 4370 .GetName(Mangled::ePreferMangled) 4371 .GetCString(); 4372 if (gsym_name) 4373 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4374 } 4375 4376 if (symbol_section) { 4377 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4378 if (symbol_byte_size == 0 && function_starts_count > 0) { 4379 addr_t symbol_lookup_file_addr = nlist.n_value; 4380 // Do an exact address match for non-ARM addresses, else get the 4381 // closest since the symbol might be a thumb symbol which has an 4382 // address with bit zero set. 4383 FunctionStarts::Entry *func_start_entry = 4384 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4385 if (is_arm && func_start_entry) { 4386 // Verify that the function start address is the symbol address 4387 // (ARM) or the symbol address + 1 (thumb). 4388 if (func_start_entry->addr != symbol_lookup_file_addr && 4389 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4390 // Not the right entry, NULL it out... 4391 func_start_entry = nullptr; 4392 } 4393 } 4394 if (func_start_entry) { 4395 func_start_entry->data = true; 4396 4397 addr_t symbol_file_addr = func_start_entry->addr; 4398 if (is_arm) 4399 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4400 4401 const FunctionStarts::Entry *next_func_start_entry = 4402 function_starts.FindNextEntry(func_start_entry); 4403 const addr_t section_end_file_addr = 4404 section_file_addr + symbol_section->GetByteSize(); 4405 if (next_func_start_entry) { 4406 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4407 // Be sure the clear the Thumb address bit when we calculate the 4408 // size from the current and next address 4409 if (is_arm) 4410 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4411 symbol_byte_size = std::min<lldb::addr_t>( 4412 next_symbol_file_addr - symbol_file_addr, 4413 section_end_file_addr - symbol_file_addr); 4414 } else { 4415 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4416 } 4417 } 4418 } 4419 symbol_value -= section_file_addr; 4420 } 4421 4422 if (!is_debug) { 4423 if (type == eSymbolTypeCode) { 4424 // See if we can find a N_FUN entry for any code symbols. If we do 4425 // find a match, and the name matches, then we can merge the two into 4426 // just the function symbol to avoid duplicate entries in the symbol 4427 // table. 4428 std::pair<ValueToSymbolIndexMap::const_iterator, 4429 ValueToSymbolIndexMap::const_iterator> 4430 range; 4431 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4432 if (range.first != range.second) { 4433 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4434 pos != range.second; ++pos) { 4435 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4436 sym[pos->second].GetMangled().GetName( 4437 Mangled::ePreferMangled)) { 4438 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4439 // We just need the flags from the linker symbol, so put these 4440 // flags into the N_FUN flags to avoid duplicate symbols in the 4441 // symbol table. 4442 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4443 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4444 if (resolver_addresses.find(nlist.n_value) != 4445 resolver_addresses.end()) 4446 sym[pos->second].SetType(eSymbolTypeResolver); 4447 sym[sym_idx].Clear(); 4448 return true; 4449 } 4450 } 4451 } else { 4452 if (resolver_addresses.find(nlist.n_value) != 4453 resolver_addresses.end()) 4454 type = eSymbolTypeResolver; 4455 } 4456 } else if (type == eSymbolTypeData || type == eSymbolTypeObjCClass || 4457 type == eSymbolTypeObjCMetaClass || 4458 type == eSymbolTypeObjCIVar) { 4459 // See if we can find a N_STSYM entry for any data symbols. If we do 4460 // find a match, and the name matches, then we can merge the two into 4461 // just the Static symbol to avoid duplicate entries in the symbol 4462 // table. 4463 std::pair<ValueToSymbolIndexMap::const_iterator, 4464 ValueToSymbolIndexMap::const_iterator> 4465 range; 4466 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4467 if (range.first != range.second) { 4468 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4469 pos != range.second; ++pos) { 4470 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == 4471 sym[pos->second].GetMangled().GetName( 4472 Mangled::ePreferMangled)) { 4473 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4474 // We just need the flags from the linker symbol, so put these 4475 // flags into the N_STSYM flags to avoid duplicate symbols in 4476 // the symbol table. 4477 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4478 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4479 sym[sym_idx].Clear(); 4480 return true; 4481 } 4482 } 4483 } else { 4484 // Combine N_GSYM stab entries with the non stab symbol. 4485 const char *gsym_name = sym[sym_idx] 4486 .GetMangled() 4487 .GetName(Mangled::ePreferMangled) 4488 .GetCString(); 4489 if (gsym_name) { 4490 ConstNameToSymbolIndexMap::const_iterator pos = 4491 N_GSYM_name_to_sym_idx.find(gsym_name); 4492 if (pos != N_GSYM_name_to_sym_idx.end()) { 4493 const uint32_t GSYM_sym_idx = pos->second; 4494 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4495 // Copy the address, because often the N_GSYM address has an 4496 // invalid address of zero when the global is a common symbol. 4497 sym[GSYM_sym_idx].GetAddressRef().SetSection(symbol_section); 4498 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4499 symbols_added.insert( 4500 sym[GSYM_sym_idx].GetAddress().GetFileAddress()); 4501 // We just need the flags from the linker symbol, so put these 4502 // flags into the N_GSYM flags to avoid duplicate symbols in 4503 // the symbol table. 4504 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4505 sym[sym_idx].Clear(); 4506 return true; 4507 } 4508 } 4509 } 4510 } 4511 } 4512 4513 sym[sym_idx].SetID(nlist_idx); 4514 sym[sym_idx].SetType(type); 4515 if (set_value) { 4516 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4517 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4518 symbols_added.insert(sym[sym_idx].GetAddress().GetFileAddress()); 4519 } 4520 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4521 if (nlist.n_desc & N_WEAK_REF) 4522 sym[sym_idx].SetIsWeak(true); 4523 4524 if (symbol_byte_size > 0) 4525 sym[sym_idx].SetByteSize(symbol_byte_size); 4526 4527 if (demangled_is_synthesized) 4528 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4529 4530 ++sym_idx; 4531 return true; 4532 }; 4533 4534 // First parse all the nlists but don't process them yet. See the next 4535 // comment for an explanation why. 4536 std::vector<struct nlist_64> nlists; 4537 nlists.reserve(symtab_load_command.nsyms); 4538 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 4539 if (auto nlist = 4540 ParseNList(nlist_data, nlist_data_offset, nlist_byte_size)) 4541 nlists.push_back(*nlist); 4542 else 4543 break; 4544 } 4545 4546 // Now parse all the debug symbols. This is needed to merge non-debug 4547 // symbols in the next step. Non-debug symbols are always coalesced into 4548 // the debug symbol. Doing this in one step would mean that some symbols 4549 // won't be merged. 4550 nlist_idx = 0; 4551 for (auto &nlist : nlists) { 4552 if (!ParseSymbolLambda(nlist, nlist_idx++, DebugSymbols)) 4553 break; 4554 } 4555 4556 // Finally parse all the non debug symbols. 4557 nlist_idx = 0; 4558 for (auto &nlist : nlists) { 4559 if (!ParseSymbolLambda(nlist, nlist_idx++, NonDebugSymbols)) 4560 break; 4561 } 4562 4563 for (const auto &pos : reexport_shlib_needs_fixup) { 4564 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4565 if (undef_pos != undefined_name_to_desc.end()) { 4566 const uint8_t dylib_ordinal = 4567 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4568 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4569 sym[pos.first].SetReExportedSymbolSharedLibrary( 4570 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4571 } 4572 } 4573 } 4574 4575 // Count how many trie symbols we'll add to the symbol table 4576 int trie_symbol_table_augment_count = 0; 4577 for (auto &e : external_sym_trie_entries) { 4578 if (symbols_added.find(e.entry.address) == symbols_added.end()) 4579 trie_symbol_table_augment_count++; 4580 } 4581 4582 if (num_syms < sym_idx + trie_symbol_table_augment_count) { 4583 num_syms = sym_idx + trie_symbol_table_augment_count; 4584 sym = symtab->Resize(num_syms); 4585 } 4586 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4587 4588 // Add symbols from the trie to the symbol table. 4589 for (auto &e : external_sym_trie_entries) { 4590 if (symbols_added.find(e.entry.address) != symbols_added.end()) 4591 continue; 4592 4593 // Find the section that this trie address is in, use that to annotate 4594 // symbol type as we add the trie address and name to the symbol table. 4595 Address symbol_addr; 4596 if (module_sp->ResolveFileAddress(e.entry.address, symbol_addr)) { 4597 SectionSP symbol_section(symbol_addr.GetSection()); 4598 const char *symbol_name = e.entry.name.GetCString(); 4599 bool demangled_is_synthesized = false; 4600 SymbolType type = 4601 GetSymbolType(symbol_name, demangled_is_synthesized, text_section_sp, 4602 data_section_sp, data_dirty_section_sp, 4603 data_const_section_sp, symbol_section); 4604 4605 sym[sym_idx].SetType(type); 4606 if (symbol_section) { 4607 sym[sym_idx].SetID(synthetic_sym_id++); 4608 sym[sym_idx].GetMangled().SetMangledName(ConstString(symbol_name)); 4609 if (demangled_is_synthesized) 4610 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4611 sym[sym_idx].SetIsSynthetic(true); 4612 sym[sym_idx].SetExternal(true); 4613 sym[sym_idx].GetAddressRef() = symbol_addr; 4614 symbols_added.insert(symbol_addr.GetFileAddress()); 4615 if (e.entry.flags & TRIE_SYMBOL_IS_THUMB) 4616 sym[sym_idx].SetFlags(MACHO_NLIST_ARM_SYMBOL_IS_THUMB); 4617 ++sym_idx; 4618 } 4619 } 4620 } 4621 4622 if (function_starts_count > 0) { 4623 uint32_t num_synthetic_function_symbols = 0; 4624 for (i = 0; i < function_starts_count; ++i) { 4625 if (symbols_added.find(function_starts.GetEntryRef(i).addr) == 4626 symbols_added.end()) 4627 ++num_synthetic_function_symbols; 4628 } 4629 4630 if (num_synthetic_function_symbols > 0) { 4631 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4632 num_syms = sym_idx + num_synthetic_function_symbols; 4633 sym = symtab->Resize(num_syms); 4634 } 4635 for (i = 0; i < function_starts_count; ++i) { 4636 const FunctionStarts::Entry *func_start_entry = 4637 function_starts.GetEntryAtIndex(i); 4638 if (symbols_added.find(func_start_entry->addr) == symbols_added.end()) { 4639 addr_t symbol_file_addr = func_start_entry->addr; 4640 uint32_t symbol_flags = 0; 4641 if (func_start_entry->data) 4642 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4643 Address symbol_addr; 4644 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4645 SectionSP symbol_section(symbol_addr.GetSection()); 4646 uint32_t symbol_byte_size = 0; 4647 if (symbol_section) { 4648 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4649 const FunctionStarts::Entry *next_func_start_entry = 4650 function_starts.FindNextEntry(func_start_entry); 4651 const addr_t section_end_file_addr = 4652 section_file_addr + symbol_section->GetByteSize(); 4653 if (next_func_start_entry) { 4654 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4655 if (is_arm) 4656 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4657 symbol_byte_size = std::min<lldb::addr_t>( 4658 next_symbol_file_addr - symbol_file_addr, 4659 section_end_file_addr - symbol_file_addr); 4660 } else { 4661 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4662 } 4663 sym[sym_idx].SetID(synthetic_sym_id++); 4664 sym[sym_idx].GetMangled().SetDemangledName( 4665 GetNextSyntheticSymbolName()); 4666 sym[sym_idx].SetType(eSymbolTypeCode); 4667 sym[sym_idx].SetIsSynthetic(true); 4668 sym[sym_idx].GetAddressRef() = symbol_addr; 4669 symbols_added.insert(symbol_addr.GetFileAddress()); 4670 if (symbol_flags) 4671 sym[sym_idx].SetFlags(symbol_flags); 4672 if (symbol_byte_size) 4673 sym[sym_idx].SetByteSize(symbol_byte_size); 4674 ++sym_idx; 4675 } 4676 } 4677 } 4678 } 4679 } 4680 } 4681 4682 // Trim our symbols down to just what we ended up with after removing any 4683 // symbols. 4684 if (sym_idx < num_syms) { 4685 num_syms = sym_idx; 4686 sym = symtab->Resize(num_syms); 4687 } 4688 4689 // Now synthesize indirect symbols 4690 if (m_dysymtab.nindirectsyms != 0) { 4691 if (indirect_symbol_index_data.GetByteSize()) { 4692 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4693 m_nlist_idx_to_sym_idx.end(); 4694 4695 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4696 ++sect_idx) { 4697 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4698 S_SYMBOL_STUBS) { 4699 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4700 if (symbol_stub_byte_size == 0) 4701 continue; 4702 4703 const uint32_t num_symbol_stubs = 4704 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4705 4706 if (num_symbol_stubs == 0) 4707 continue; 4708 4709 const uint32_t symbol_stub_index_offset = 4710 m_mach_sections[sect_idx].reserved1; 4711 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) { 4712 const uint32_t symbol_stub_index = 4713 symbol_stub_index_offset + stub_idx; 4714 const lldb::addr_t symbol_stub_addr = 4715 m_mach_sections[sect_idx].addr + 4716 (stub_idx * symbol_stub_byte_size); 4717 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4718 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4719 symbol_stub_offset, 4)) { 4720 const uint32_t stub_sym_id = 4721 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4722 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4723 continue; 4724 4725 NListIndexToSymbolIndexMap::const_iterator index_pos = 4726 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4727 Symbol *stub_symbol = nullptr; 4728 if (index_pos != end_index_pos) { 4729 // We have a remapping from the original nlist index to a 4730 // current symbol index, so just look this up by index 4731 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4732 } else { 4733 // We need to lookup a symbol using the original nlist symbol 4734 // index since this index is coming from the S_SYMBOL_STUBS 4735 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4736 } 4737 4738 if (stub_symbol) { 4739 Address so_addr(symbol_stub_addr, section_list); 4740 4741 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4742 // Change the external symbol into a trampoline that makes 4743 // sense These symbols were N_UNDF N_EXT, and are useless 4744 // to us, so we can re-use them so we don't have to make up 4745 // a synthetic symbol for no good reason. 4746 if (resolver_addresses.find(symbol_stub_addr) == 4747 resolver_addresses.end()) 4748 stub_symbol->SetType(eSymbolTypeTrampoline); 4749 else 4750 stub_symbol->SetType(eSymbolTypeResolver); 4751 stub_symbol->SetExternal(false); 4752 stub_symbol->GetAddressRef() = so_addr; 4753 stub_symbol->SetByteSize(symbol_stub_byte_size); 4754 } else { 4755 // Make a synthetic symbol to describe the trampoline stub 4756 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4757 if (sym_idx >= num_syms) { 4758 sym = symtab->Resize(++num_syms); 4759 stub_symbol = nullptr; // this pointer no longer valid 4760 } 4761 sym[sym_idx].SetID(synthetic_sym_id++); 4762 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4763 if (resolver_addresses.find(symbol_stub_addr) == 4764 resolver_addresses.end()) 4765 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4766 else 4767 sym[sym_idx].SetType(eSymbolTypeResolver); 4768 sym[sym_idx].SetIsSynthetic(true); 4769 sym[sym_idx].GetAddressRef() = so_addr; 4770 symbols_added.insert(so_addr.GetFileAddress()); 4771 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4772 ++sym_idx; 4773 } 4774 } else { 4775 if (log) 4776 log->Warning("symbol stub referencing symbol table symbol " 4777 "%u that isn't in our minimal symbol table, " 4778 "fix this!!!", 4779 stub_sym_id); 4780 } 4781 } 4782 } 4783 } 4784 } 4785 } 4786 } 4787 4788 if (!reexport_trie_entries.empty()) { 4789 for (const auto &e : reexport_trie_entries) { 4790 if (e.entry.import_name) { 4791 // Only add indirect symbols from the Trie entries if we didn't have 4792 // a N_INDR nlist entry for this already 4793 if (indirect_symbol_names.find(e.entry.name) == 4794 indirect_symbol_names.end()) { 4795 // Make a synthetic symbol to describe re-exported symbol. 4796 if (sym_idx >= num_syms) 4797 sym = symtab->Resize(++num_syms); 4798 sym[sym_idx].SetID(synthetic_sym_id++); 4799 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4800 sym[sym_idx].SetType(eSymbolTypeReExported); 4801 sym[sym_idx].SetIsSynthetic(true); 4802 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4803 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4804 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4805 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4806 } 4807 ++sym_idx; 4808 } 4809 } 4810 } 4811 } 4812 4813 // StreamFile s(stdout, false); 4814 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4815 // symtab->Dump(&s, NULL, eSortOrderNone); 4816 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4817 // sizes 4818 symtab->CalculateSymbolSizes(); 4819 4820 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4821 // symtab->Dump(&s, NULL, eSortOrderNone); 4822 4823 return symtab->GetNumSymbols(); 4824 } 4825 4826 void ObjectFileMachO::Dump(Stream *s) { 4827 ModuleSP module_sp(GetModule()); 4828 if (module_sp) { 4829 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4830 s->Printf("%p: ", static_cast<void *>(this)); 4831 s->Indent(); 4832 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4833 s->PutCString("ObjectFileMachO64"); 4834 else 4835 s->PutCString("ObjectFileMachO32"); 4836 4837 *s << ", file = '" << m_file; 4838 ModuleSpecList all_specs; 4839 ModuleSpec base_spec; 4840 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 4841 base_spec, all_specs); 4842 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 4843 *s << "', triple"; 4844 if (e) 4845 s->Printf("[%d]", i); 4846 *s << " = "; 4847 *s << all_specs.GetModuleSpecRefAtIndex(i) 4848 .GetArchitecture() 4849 .GetTriple() 4850 .getTriple(); 4851 } 4852 *s << "\n"; 4853 SectionList *sections = GetSectionList(); 4854 if (sections) 4855 sections->Dump(s->AsRawOstream(), s->GetIndentLevel(), nullptr, true, 4856 UINT32_MAX); 4857 4858 if (m_symtab_up) 4859 m_symtab_up->Dump(s, nullptr, eSortOrderNone); 4860 } 4861 } 4862 4863 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4864 const lldb_private::DataExtractor &data, 4865 lldb::offset_t lc_offset) { 4866 uint32_t i; 4867 struct uuid_command load_cmd; 4868 4869 lldb::offset_t offset = lc_offset; 4870 for (i = 0; i < header.ncmds; ++i) { 4871 const lldb::offset_t cmd_offset = offset; 4872 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4873 break; 4874 4875 if (load_cmd.cmd == LC_UUID) { 4876 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4877 4878 if (uuid_bytes) { 4879 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4880 // We pretend these object files have no UUID to prevent crashing. 4881 4882 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4883 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4884 0xbb, 0x14, 0xf0, 0x0d}; 4885 4886 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4887 return UUID(); 4888 4889 return UUID::fromOptionalData(uuid_bytes, 16); 4890 } 4891 return UUID(); 4892 } 4893 offset = cmd_offset + load_cmd.cmdsize; 4894 } 4895 return UUID(); 4896 } 4897 4898 static llvm::StringRef GetOSName(uint32_t cmd) { 4899 switch (cmd) { 4900 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4901 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4902 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4903 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4904 case llvm::MachO::LC_VERSION_MIN_TVOS: 4905 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4906 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4907 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4908 default: 4909 llvm_unreachable("unexpected LC_VERSION load command"); 4910 } 4911 } 4912 4913 namespace { 4914 struct OSEnv { 4915 llvm::StringRef os_type; 4916 llvm::StringRef environment; 4917 OSEnv(uint32_t cmd) { 4918 switch (cmd) { 4919 case llvm::MachO::PLATFORM_MACOS: 4920 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4921 return; 4922 case llvm::MachO::PLATFORM_IOS: 4923 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4924 return; 4925 case llvm::MachO::PLATFORM_TVOS: 4926 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4927 return; 4928 case llvm::MachO::PLATFORM_WATCHOS: 4929 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4930 return; 4931 // NEED_BRIDGEOS_TRIPLE case llvm::MachO::PLATFORM_BRIDGEOS: 4932 // NEED_BRIDGEOS_TRIPLE os_type = 4933 // llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4934 // NEED_BRIDGEOS_TRIPLE return; 4935 case llvm::MachO::PLATFORM_MACCATALYST: 4936 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4937 environment = llvm::Triple::getEnvironmentTypeName(llvm::Triple::MacABI); 4938 return; 4939 case llvm::MachO::PLATFORM_IOSSIMULATOR: 4940 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4941 environment = 4942 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4943 return; 4944 case llvm::MachO::PLATFORM_TVOSSIMULATOR: 4945 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4946 environment = 4947 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4948 return; 4949 case llvm::MachO::PLATFORM_WATCHOSSIMULATOR: 4950 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4951 environment = 4952 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4953 return; 4954 default: { 4955 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 4956 LIBLLDB_LOG_PROCESS)); 4957 LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION"); 4958 } 4959 } 4960 } 4961 }; 4962 4963 struct MinOS { 4964 uint32_t major_version, minor_version, patch_version; 4965 MinOS(uint32_t version) 4966 : major_version(version >> 16), minor_version((version >> 8) & 0xffu), 4967 patch_version(version & 0xffu) {} 4968 }; 4969 } // namespace 4970 4971 void ObjectFileMachO::GetAllArchSpecs(const llvm::MachO::mach_header &header, 4972 const lldb_private::DataExtractor &data, 4973 lldb::offset_t lc_offset, 4974 ModuleSpec &base_spec, 4975 lldb_private::ModuleSpecList &all_specs) { 4976 auto &base_arch = base_spec.GetArchitecture(); 4977 base_arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4978 if (!base_arch.IsValid()) 4979 return; 4980 4981 bool found_any = false; 4982 auto add_triple = [&](const llvm::Triple &triple) { 4983 auto spec = base_spec; 4984 spec.GetArchitecture().GetTriple() = triple; 4985 if (spec.GetArchitecture().IsValid()) { 4986 spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset); 4987 all_specs.Append(spec); 4988 found_any = true; 4989 } 4990 }; 4991 4992 // Set OS to an unspecified unknown or a "*" so it can match any OS 4993 llvm::Triple base_triple = base_arch.GetTriple(); 4994 base_triple.setOS(llvm::Triple::UnknownOS); 4995 base_triple.setOSName(llvm::StringRef()); 4996 4997 if (header.filetype == MH_PRELOAD) { 4998 if (header.cputype == CPU_TYPE_ARM) { 4999 // If this is a 32-bit arm binary, and it's a standalone binary, force 5000 // the Vendor to Apple so we don't accidentally pick up the generic 5001 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 5002 // frame pointer register; most other armv7 ABIs use a combination of 5003 // r7 and r11. 5004 base_triple.setVendor(llvm::Triple::Apple); 5005 } else { 5006 // Set vendor to an unspecified unknown or a "*" so it can match any 5007 // vendor This is required for correct behavior of EFI debugging on 5008 // x86_64 5009 base_triple.setVendor(llvm::Triple::UnknownVendor); 5010 base_triple.setVendorName(llvm::StringRef()); 5011 } 5012 return add_triple(base_triple); 5013 } 5014 5015 struct load_command load_cmd; 5016 5017 // See if there is an LC_VERSION_MIN_* load command that can give 5018 // us the OS type. 5019 lldb::offset_t offset = lc_offset; 5020 for (uint32_t i = 0; i < header.ncmds; ++i) { 5021 const lldb::offset_t cmd_offset = offset; 5022 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 5023 break; 5024 5025 struct version_min_command version_min; 5026 switch (load_cmd.cmd) { 5027 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 5028 case llvm::MachO::LC_VERSION_MIN_MACOSX: 5029 case llvm::MachO::LC_VERSION_MIN_TVOS: 5030 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 5031 if (load_cmd.cmdsize != sizeof(version_min)) 5032 break; 5033 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 5034 data.GetByteOrder(), &version_min) == 0) 5035 break; 5036 MinOS min_os(version_min.version); 5037 llvm::SmallString<32> os_name; 5038 llvm::raw_svector_ostream os(os_name); 5039 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 5040 << min_os.minor_version << '.' << min_os.patch_version; 5041 5042 auto triple = base_triple; 5043 triple.setOSName(os.str()); 5044 os_name.clear(); 5045 add_triple(triple); 5046 break; 5047 } 5048 default: 5049 break; 5050 } 5051 5052 offset = cmd_offset + load_cmd.cmdsize; 5053 } 5054 5055 // See if there are LC_BUILD_VERSION load commands that can give 5056 // us the OS type. 5057 offset = lc_offset; 5058 for (uint32_t i = 0; i < header.ncmds; ++i) { 5059 const lldb::offset_t cmd_offset = offset; 5060 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 5061 break; 5062 5063 do { 5064 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 5065 struct build_version_command build_version; 5066 if (load_cmd.cmdsize < sizeof(build_version)) { 5067 // Malformed load command. 5068 break; 5069 } 5070 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 5071 data.GetByteOrder(), &build_version) == 0) 5072 break; 5073 MinOS min_os(build_version.minos); 5074 OSEnv os_env(build_version.platform); 5075 llvm::SmallString<16> os_name; 5076 llvm::raw_svector_ostream os(os_name); 5077 os << os_env.os_type << min_os.major_version << '.' 5078 << min_os.minor_version << '.' << min_os.patch_version; 5079 auto triple = base_triple; 5080 triple.setOSName(os.str()); 5081 os_name.clear(); 5082 if (!os_env.environment.empty()) 5083 triple.setEnvironmentName(os_env.environment); 5084 add_triple(triple); 5085 } 5086 } while (0); 5087 offset = cmd_offset + load_cmd.cmdsize; 5088 } 5089 5090 if (!found_any) { 5091 if (header.filetype == MH_KEXT_BUNDLE) { 5092 base_triple.setVendor(llvm::Triple::Apple); 5093 add_triple(base_triple); 5094 } else { 5095 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 5096 // so lets not say our Vendor is Apple, leave it as an unspecified 5097 // unknown. 5098 base_triple.setVendor(llvm::Triple::UnknownVendor); 5099 base_triple.setVendorName(llvm::StringRef()); 5100 add_triple(base_triple); 5101 } 5102 } 5103 } 5104 5105 ArchSpec ObjectFileMachO::GetArchitecture( 5106 ModuleSP module_sp, const llvm::MachO::mach_header &header, 5107 const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) { 5108 ModuleSpecList all_specs; 5109 ModuleSpec base_spec; 5110 GetAllArchSpecs(header, data, MachHeaderSizeFromMagic(header.magic), 5111 base_spec, all_specs); 5112 5113 // If the object file offers multiple alternative load commands, 5114 // pick the one that matches the module. 5115 if (module_sp) { 5116 const ArchSpec &module_arch = module_sp->GetArchitecture(); 5117 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 5118 ArchSpec mach_arch = 5119 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 5120 if (module_arch.IsCompatibleMatch(mach_arch)) 5121 return mach_arch; 5122 } 5123 } 5124 5125 // Return the first arch we found. 5126 if (all_specs.GetSize() == 0) 5127 return {}; 5128 return all_specs.GetModuleSpecRefAtIndex(0).GetArchitecture(); 5129 } 5130 5131 UUID ObjectFileMachO::GetUUID() { 5132 ModuleSP module_sp(GetModule()); 5133 if (module_sp) { 5134 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5135 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5136 return GetUUID(m_header, m_data, offset); 5137 } 5138 return UUID(); 5139 } 5140 5141 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5142 uint32_t count = 0; 5143 ModuleSP module_sp(GetModule()); 5144 if (module_sp) { 5145 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5146 struct load_command load_cmd; 5147 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5148 std::vector<std::string> rpath_paths; 5149 std::vector<std::string> rpath_relative_paths; 5150 std::vector<std::string> at_exec_relative_paths; 5151 uint32_t i; 5152 for (i = 0; i < m_header.ncmds; ++i) { 5153 const uint32_t cmd_offset = offset; 5154 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5155 break; 5156 5157 switch (load_cmd.cmd) { 5158 case LC_RPATH: 5159 case LC_LOAD_DYLIB: 5160 case LC_LOAD_WEAK_DYLIB: 5161 case LC_REEXPORT_DYLIB: 5162 case LC_LOAD_DYLINKER: 5163 case LC_LOADFVMLIB: 5164 case LC_LOAD_UPWARD_DYLIB: { 5165 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5166 const char *path = m_data.PeekCStr(name_offset); 5167 if (path) { 5168 if (load_cmd.cmd == LC_RPATH) 5169 rpath_paths.push_back(path); 5170 else { 5171 if (path[0] == '@') { 5172 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5173 rpath_relative_paths.push_back(path + strlen("@rpath")); 5174 else if (strncmp(path, "@executable_path", 5175 strlen("@executable_path")) == 0) 5176 at_exec_relative_paths.push_back(path + 5177 strlen("@executable_path")); 5178 } else { 5179 FileSpec file_spec(path); 5180 if (files.AppendIfUnique(file_spec)) 5181 count++; 5182 } 5183 } 5184 } 5185 } break; 5186 5187 default: 5188 break; 5189 } 5190 offset = cmd_offset + load_cmd.cmdsize; 5191 } 5192 5193 FileSpec this_file_spec(m_file); 5194 FileSystem::Instance().Resolve(this_file_spec); 5195 5196 if (!rpath_paths.empty()) { 5197 // Fixup all LC_RPATH values to be absolute paths 5198 std::string loader_path("@loader_path"); 5199 std::string executable_path("@executable_path"); 5200 for (auto &rpath : rpath_paths) { 5201 if (llvm::StringRef(rpath).startswith(loader_path)) { 5202 rpath.erase(0, loader_path.size()); 5203 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5204 } else if (llvm::StringRef(rpath).startswith(executable_path)) { 5205 rpath.erase(0, executable_path.size()); 5206 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5207 } 5208 } 5209 5210 for (const auto &rpath_relative_path : rpath_relative_paths) { 5211 for (const auto &rpath : rpath_paths) { 5212 std::string path = rpath; 5213 path += rpath_relative_path; 5214 // It is OK to resolve this path because we must find a file on disk 5215 // for us to accept it anyway if it is rpath relative. 5216 FileSpec file_spec(path); 5217 FileSystem::Instance().Resolve(file_spec); 5218 if (FileSystem::Instance().Exists(file_spec) && 5219 files.AppendIfUnique(file_spec)) { 5220 count++; 5221 break; 5222 } 5223 } 5224 } 5225 } 5226 5227 // We may have @executable_paths but no RPATHS. Figure those out here. 5228 // Only do this if this object file is the executable. We have no way to 5229 // get back to the actual executable otherwise, so we won't get the right 5230 // path. 5231 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5232 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5233 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5234 FileSpec file_spec = 5235 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5236 if (FileSystem::Instance().Exists(file_spec) && 5237 files.AppendIfUnique(file_spec)) 5238 count++; 5239 } 5240 } 5241 } 5242 return count; 5243 } 5244 5245 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5246 // If the object file is not an executable it can't hold the entry point. 5247 // m_entry_point_address is initialized to an invalid address, so we can just 5248 // return that. If m_entry_point_address is valid it means we've found it 5249 // already, so return the cached value. 5250 5251 if ((!IsExecutable() && !IsDynamicLoader()) || 5252 m_entry_point_address.IsValid()) { 5253 return m_entry_point_address; 5254 } 5255 5256 // Otherwise, look for the UnixThread or Thread command. The data for the 5257 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5258 // 5259 // uint32_t flavor - this is the flavor argument you would pass to 5260 // thread_get_state 5261 // uint32_t count - this is the count of longs in the thread state data 5262 // struct XXX_thread_state state - this is the structure from 5263 // <machine/thread_status.h> corresponding to the flavor. 5264 // <repeat this trio> 5265 // 5266 // So we just keep reading the various register flavors till we find the GPR 5267 // one, then read the PC out of there. 5268 // FIXME: We will need to have a "RegisterContext data provider" class at some 5269 // point that can get all the registers 5270 // out of data in this form & attach them to a given thread. That should 5271 // underlie the MacOS X User process plugin, and we'll also need it for the 5272 // MacOS X Core File process plugin. When we have that we can also use it 5273 // here. 5274 // 5275 // For now we hard-code the offsets and flavors we need: 5276 // 5277 // 5278 5279 ModuleSP module_sp(GetModule()); 5280 if (module_sp) { 5281 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5282 struct load_command load_cmd; 5283 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5284 uint32_t i; 5285 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5286 bool done = false; 5287 5288 for (i = 0; i < m_header.ncmds; ++i) { 5289 const lldb::offset_t cmd_offset = offset; 5290 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5291 break; 5292 5293 switch (load_cmd.cmd) { 5294 case LC_UNIXTHREAD: 5295 case LC_THREAD: { 5296 while (offset < cmd_offset + load_cmd.cmdsize) { 5297 uint32_t flavor = m_data.GetU32(&offset); 5298 uint32_t count = m_data.GetU32(&offset); 5299 if (count == 0) { 5300 // We've gotten off somehow, log and exit; 5301 return m_entry_point_address; 5302 } 5303 5304 switch (m_header.cputype) { 5305 case llvm::MachO::CPU_TYPE_ARM: 5306 if (flavor == 1 || 5307 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 5308 // from mach/arm/thread_status.h 5309 { 5310 offset += 60; // This is the offset of pc in the GPR thread state 5311 // data structure. 5312 start_address = m_data.GetU32(&offset); 5313 done = true; 5314 } 5315 break; 5316 case llvm::MachO::CPU_TYPE_ARM64: 5317 case llvm::MachO::CPU_TYPE_ARM64_32: 5318 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5319 { 5320 offset += 256; // This is the offset of pc in the GPR thread state 5321 // data structure. 5322 start_address = m_data.GetU64(&offset); 5323 done = true; 5324 } 5325 break; 5326 case llvm::MachO::CPU_TYPE_I386: 5327 if (flavor == 5328 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5329 { 5330 offset += 40; // This is the offset of eip in the GPR thread state 5331 // data structure. 5332 start_address = m_data.GetU32(&offset); 5333 done = true; 5334 } 5335 break; 5336 case llvm::MachO::CPU_TYPE_X86_64: 5337 if (flavor == 5338 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5339 { 5340 offset += 16 * 8; // This is the offset of rip in the GPR thread 5341 // state data structure. 5342 start_address = m_data.GetU64(&offset); 5343 done = true; 5344 } 5345 break; 5346 default: 5347 return m_entry_point_address; 5348 } 5349 // Haven't found the GPR flavor yet, skip over the data for this 5350 // flavor: 5351 if (done) 5352 break; 5353 offset += count * 4; 5354 } 5355 } break; 5356 case LC_MAIN: { 5357 ConstString text_segment_name("__TEXT"); 5358 uint64_t entryoffset = m_data.GetU64(&offset); 5359 SectionSP text_segment_sp = 5360 GetSectionList()->FindSectionByName(text_segment_name); 5361 if (text_segment_sp) { 5362 done = true; 5363 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5364 } 5365 } break; 5366 5367 default: 5368 break; 5369 } 5370 if (done) 5371 break; 5372 5373 // Go to the next load command: 5374 offset = cmd_offset + load_cmd.cmdsize; 5375 } 5376 5377 if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) { 5378 if (GetSymtab()) { 5379 Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType( 5380 ConstString("_dyld_start"), SymbolType::eSymbolTypeCode, 5381 Symtab::eDebugAny, Symtab::eVisibilityAny); 5382 if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) { 5383 start_address = dyld_start_sym->GetAddress().GetFileAddress(); 5384 } 5385 } 5386 } 5387 5388 if (start_address != LLDB_INVALID_ADDRESS) { 5389 // We got the start address from the load commands, so now resolve that 5390 // address in the sections of this ObjectFile: 5391 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5392 start_address, GetSectionList())) { 5393 m_entry_point_address.Clear(); 5394 } 5395 } else { 5396 // We couldn't read the UnixThread load command - maybe it wasn't there. 5397 // As a fallback look for the "start" symbol in the main executable. 5398 5399 ModuleSP module_sp(GetModule()); 5400 5401 if (module_sp) { 5402 SymbolContextList contexts; 5403 SymbolContext context; 5404 module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5405 eSymbolTypeCode, contexts); 5406 if (contexts.GetSize()) { 5407 if (contexts.GetContextAtIndex(0, context)) 5408 m_entry_point_address = context.symbol->GetAddress(); 5409 } 5410 } 5411 } 5412 } 5413 5414 return m_entry_point_address; 5415 } 5416 5417 lldb_private::Address ObjectFileMachO::GetBaseAddress() { 5418 lldb_private::Address header_addr; 5419 SectionList *section_list = GetSectionList(); 5420 if (section_list) { 5421 SectionSP text_segment_sp( 5422 section_list->FindSectionByName(GetSegmentNameTEXT())); 5423 if (text_segment_sp) { 5424 header_addr.SetSection(text_segment_sp); 5425 header_addr.SetOffset(0); 5426 } 5427 } 5428 return header_addr; 5429 } 5430 5431 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5432 ModuleSP module_sp(GetModule()); 5433 if (module_sp) { 5434 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5435 if (!m_thread_context_offsets_valid) { 5436 m_thread_context_offsets_valid = true; 5437 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5438 FileRangeArray::Entry file_range; 5439 thread_command thread_cmd; 5440 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5441 const uint32_t cmd_offset = offset; 5442 if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr) 5443 break; 5444 5445 if (thread_cmd.cmd == LC_THREAD) { 5446 file_range.SetRangeBase(offset); 5447 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5448 m_thread_context_offsets.Append(file_range); 5449 } 5450 offset = cmd_offset + thread_cmd.cmdsize; 5451 } 5452 } 5453 } 5454 return m_thread_context_offsets.GetSize(); 5455 } 5456 5457 std::string ObjectFileMachO::GetIdentifierString() { 5458 std::string result; 5459 ModuleSP module_sp(GetModule()); 5460 if (module_sp) { 5461 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5462 5463 // First, look over the load commands for an LC_NOTE load command with 5464 // data_owner string "kern ver str" & use that if found. 5465 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5466 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5467 const uint32_t cmd_offset = offset; 5468 load_command lc; 5469 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5470 break; 5471 if (lc.cmd == LC_NOTE) { 5472 char data_owner[17]; 5473 m_data.CopyData(offset, 16, data_owner); 5474 data_owner[16] = '\0'; 5475 offset += 16; 5476 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5477 uint64_t size = m_data.GetU64_unchecked(&offset); 5478 5479 // "kern ver str" has a uint32_t version and then a nul terminated 5480 // c-string. 5481 if (strcmp("kern ver str", data_owner) == 0) { 5482 offset = fileoff; 5483 uint32_t version; 5484 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5485 if (version == 1) { 5486 uint32_t strsize = size - sizeof(uint32_t); 5487 char *buf = (char *)malloc(strsize); 5488 if (buf) { 5489 m_data.CopyData(offset, strsize, buf); 5490 buf[strsize - 1] = '\0'; 5491 result = buf; 5492 if (buf) 5493 free(buf); 5494 return result; 5495 } 5496 } 5497 } 5498 } 5499 } 5500 offset = cmd_offset + lc.cmdsize; 5501 } 5502 5503 // Second, make a pass over the load commands looking for an obsolete 5504 // LC_IDENT load command. 5505 offset = MachHeaderSizeFromMagic(m_header.magic); 5506 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5507 const uint32_t cmd_offset = offset; 5508 struct ident_command ident_command; 5509 if (m_data.GetU32(&offset, &ident_command, 2) == nullptr) 5510 break; 5511 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5512 char *buf = (char *)malloc(ident_command.cmdsize); 5513 if (buf != nullptr && m_data.CopyData(offset, ident_command.cmdsize, 5514 buf) == ident_command.cmdsize) { 5515 buf[ident_command.cmdsize - 1] = '\0'; 5516 result = buf; 5517 } 5518 if (buf) 5519 free(buf); 5520 } 5521 offset = cmd_offset + ident_command.cmdsize; 5522 } 5523 } 5524 return result; 5525 } 5526 5527 bool ObjectFileMachO::GetCorefileMainBinaryInfo(addr_t &address, UUID &uuid) { 5528 address = LLDB_INVALID_ADDRESS; 5529 uuid.Clear(); 5530 ModuleSP module_sp(GetModule()); 5531 if (module_sp) { 5532 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5533 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5534 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5535 const uint32_t cmd_offset = offset; 5536 load_command lc; 5537 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5538 break; 5539 if (lc.cmd == LC_NOTE) { 5540 char data_owner[17]; 5541 memset(data_owner, 0, sizeof(data_owner)); 5542 m_data.CopyData(offset, 16, data_owner); 5543 offset += 16; 5544 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5545 uint64_t size = m_data.GetU64_unchecked(&offset); 5546 5547 // "main bin spec" (main binary specification) data payload is 5548 // formatted: 5549 // uint32_t version [currently 1] 5550 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user 5551 // process] uint64_t address [ UINT64_MAX if address not 5552 // specified ] uuid_t uuid [ all zero's if uuid not 5553 // specified ] uint32_t log2_pagesize [ process page size in log base 5554 // 2, e.g. 4k pages are 12. 0 for unspecified ] 5555 5556 if (strcmp("main bin spec", data_owner) == 0 && size >= 32) { 5557 offset = fileoff; 5558 uint32_t version; 5559 if (m_data.GetU32(&offset, &version, 1) != nullptr && version == 1) { 5560 uint32_t type = 0; 5561 uuid_t raw_uuid; 5562 memset(raw_uuid, 0, sizeof(uuid_t)); 5563 5564 if (m_data.GetU32(&offset, &type, 1) && 5565 m_data.GetU64(&offset, &address, 1) && 5566 m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5567 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5568 return true; 5569 } 5570 } 5571 } 5572 } 5573 offset = cmd_offset + lc.cmdsize; 5574 } 5575 } 5576 return false; 5577 } 5578 5579 lldb::RegisterContextSP 5580 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5581 lldb_private::Thread &thread) { 5582 lldb::RegisterContextSP reg_ctx_sp; 5583 5584 ModuleSP module_sp(GetModule()); 5585 if (module_sp) { 5586 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5587 if (!m_thread_context_offsets_valid) 5588 GetNumThreadContexts(); 5589 5590 const FileRangeArray::Entry *thread_context_file_range = 5591 m_thread_context_offsets.GetEntryAtIndex(idx); 5592 if (thread_context_file_range) { 5593 5594 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5595 thread_context_file_range->GetByteSize()); 5596 5597 switch (m_header.cputype) { 5598 case llvm::MachO::CPU_TYPE_ARM64: 5599 case llvm::MachO::CPU_TYPE_ARM64_32: 5600 reg_ctx_sp = 5601 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data); 5602 break; 5603 5604 case llvm::MachO::CPU_TYPE_ARM: 5605 reg_ctx_sp = 5606 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data); 5607 break; 5608 5609 case llvm::MachO::CPU_TYPE_I386: 5610 reg_ctx_sp = 5611 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data); 5612 break; 5613 5614 case llvm::MachO::CPU_TYPE_X86_64: 5615 reg_ctx_sp = 5616 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data); 5617 break; 5618 } 5619 } 5620 } 5621 return reg_ctx_sp; 5622 } 5623 5624 ObjectFile::Type ObjectFileMachO::CalculateType() { 5625 switch (m_header.filetype) { 5626 case MH_OBJECT: // 0x1u 5627 if (GetAddressByteSize() == 4) { 5628 // 32 bit kexts are just object files, but they do have a valid 5629 // UUID load command. 5630 if (GetUUID()) { 5631 // this checking for the UUID load command is not enough we could 5632 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5633 // this is required of kexts 5634 if (m_strata == eStrataInvalid) 5635 m_strata = eStrataKernel; 5636 return eTypeSharedLibrary; 5637 } 5638 } 5639 return eTypeObjectFile; 5640 5641 case MH_EXECUTE: 5642 return eTypeExecutable; // 0x2u 5643 case MH_FVMLIB: 5644 return eTypeSharedLibrary; // 0x3u 5645 case MH_CORE: 5646 return eTypeCoreFile; // 0x4u 5647 case MH_PRELOAD: 5648 return eTypeSharedLibrary; // 0x5u 5649 case MH_DYLIB: 5650 return eTypeSharedLibrary; // 0x6u 5651 case MH_DYLINKER: 5652 return eTypeDynamicLinker; // 0x7u 5653 case MH_BUNDLE: 5654 return eTypeSharedLibrary; // 0x8u 5655 case MH_DYLIB_STUB: 5656 return eTypeStubLibrary; // 0x9u 5657 case MH_DSYM: 5658 return eTypeDebugInfo; // 0xAu 5659 case MH_KEXT_BUNDLE: 5660 return eTypeSharedLibrary; // 0xBu 5661 default: 5662 break; 5663 } 5664 return eTypeUnknown; 5665 } 5666 5667 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5668 switch (m_header.filetype) { 5669 case MH_OBJECT: // 0x1u 5670 { 5671 // 32 bit kexts are just object files, but they do have a valid 5672 // UUID load command. 5673 if (GetUUID()) { 5674 // this checking for the UUID load command is not enough we could 5675 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5676 // this is required of kexts 5677 if (m_type == eTypeInvalid) 5678 m_type = eTypeSharedLibrary; 5679 5680 return eStrataKernel; 5681 } 5682 } 5683 return eStrataUnknown; 5684 5685 case MH_EXECUTE: // 0x2u 5686 // Check for the MH_DYLDLINK bit in the flags 5687 if (m_header.flags & MH_DYLDLINK) { 5688 return eStrataUser; 5689 } else { 5690 SectionList *section_list = GetSectionList(); 5691 if (section_list) { 5692 static ConstString g_kld_section_name("__KLD"); 5693 if (section_list->FindSectionByName(g_kld_section_name)) 5694 return eStrataKernel; 5695 } 5696 } 5697 return eStrataRawImage; 5698 5699 case MH_FVMLIB: 5700 return eStrataUser; // 0x3u 5701 case MH_CORE: 5702 return eStrataUnknown; // 0x4u 5703 case MH_PRELOAD: 5704 return eStrataRawImage; // 0x5u 5705 case MH_DYLIB: 5706 return eStrataUser; // 0x6u 5707 case MH_DYLINKER: 5708 return eStrataUser; // 0x7u 5709 case MH_BUNDLE: 5710 return eStrataUser; // 0x8u 5711 case MH_DYLIB_STUB: 5712 return eStrataUser; // 0x9u 5713 case MH_DSYM: 5714 return eStrataUnknown; // 0xAu 5715 case MH_KEXT_BUNDLE: 5716 return eStrataKernel; // 0xBu 5717 default: 5718 break; 5719 } 5720 return eStrataUnknown; 5721 } 5722 5723 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5724 ModuleSP module_sp(GetModule()); 5725 if (module_sp) { 5726 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5727 struct dylib_command load_cmd; 5728 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5729 uint32_t version_cmd = 0; 5730 uint64_t version = 0; 5731 uint32_t i; 5732 for (i = 0; i < m_header.ncmds; ++i) { 5733 const lldb::offset_t cmd_offset = offset; 5734 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5735 break; 5736 5737 if (load_cmd.cmd == LC_ID_DYLIB) { 5738 if (version_cmd == 0) { 5739 version_cmd = load_cmd.cmd; 5740 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr) 5741 break; 5742 version = load_cmd.dylib.current_version; 5743 } 5744 break; // Break for now unless there is another more complete version 5745 // number load command in the future. 5746 } 5747 offset = cmd_offset + load_cmd.cmdsize; 5748 } 5749 5750 if (version_cmd == LC_ID_DYLIB) { 5751 unsigned major = (version & 0xFFFF0000ull) >> 16; 5752 unsigned minor = (version & 0x0000FF00ull) >> 8; 5753 unsigned subminor = (version & 0x000000FFull); 5754 return llvm::VersionTuple(major, minor, subminor); 5755 } 5756 } 5757 return llvm::VersionTuple(); 5758 } 5759 5760 ArchSpec ObjectFileMachO::GetArchitecture() { 5761 ModuleSP module_sp(GetModule()); 5762 ArchSpec arch; 5763 if (module_sp) { 5764 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5765 5766 return GetArchitecture(module_sp, m_header, m_data, 5767 MachHeaderSizeFromMagic(m_header.magic)); 5768 } 5769 return arch; 5770 } 5771 5772 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, 5773 addr_t &base_addr, UUID &uuid) { 5774 uuid.Clear(); 5775 base_addr = LLDB_INVALID_ADDRESS; 5776 if (process && process->GetDynamicLoader()) { 5777 DynamicLoader *dl = process->GetDynamicLoader(); 5778 LazyBool using_shared_cache; 5779 LazyBool private_shared_cache; 5780 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5781 private_shared_cache); 5782 } 5783 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5784 LIBLLDB_LOG_PROCESS)); 5785 LLDB_LOGF( 5786 log, 5787 "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64, 5788 uuid.GetAsString().c_str(), base_addr); 5789 } 5790 5791 // From dyld SPI header dyld_process_info.h 5792 typedef void *dyld_process_info; 5793 struct lldb_copy__dyld_process_cache_info { 5794 uuid_t cacheUUID; // UUID of cache used by process 5795 uint64_t cacheBaseAddress; // load address of dyld shared cache 5796 bool noCache; // process is running without a dyld cache 5797 bool privateCache; // process is using a private copy of its dyld cache 5798 }; 5799 5800 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with 5801 // llvm enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile 5802 // errors. So we need to use the actual underlying types of task_t and 5803 // kern_return_t below. 5804 extern "C" unsigned int /*task_t*/ mach_task_self(); 5805 5806 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5807 uuid.Clear(); 5808 base_addr = LLDB_INVALID_ADDRESS; 5809 5810 #if defined(__APPLE__) && \ 5811 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5812 uint8_t *(*dyld_get_all_image_infos)(void); 5813 dyld_get_all_image_infos = 5814 (uint8_t * (*)()) dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5815 if (dyld_get_all_image_infos) { 5816 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5817 if (dyld_all_image_infos_address) { 5818 uint32_t *version = (uint32_t *) 5819 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5820 if (*version >= 13) { 5821 uuid_t *sharedCacheUUID_address = 0; 5822 int wordsize = sizeof(uint8_t *); 5823 if (wordsize == 8) { 5824 sharedCacheUUID_address = 5825 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5826 160); // sharedCacheUUID <mach-o/dyld_images.h> 5827 if (*version >= 15) 5828 base_addr = 5829 *(uint64_t 5830 *)((uint8_t *)dyld_all_image_infos_address + 5831 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5832 } else { 5833 sharedCacheUUID_address = 5834 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5835 84); // sharedCacheUUID <mach-o/dyld_images.h> 5836 if (*version >= 15) { 5837 base_addr = 0; 5838 base_addr = 5839 *(uint32_t 5840 *)((uint8_t *)dyld_all_image_infos_address + 5841 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5842 } 5843 } 5844 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5845 } 5846 } 5847 } else { 5848 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5849 dyld_process_info (*dyld_process_info_create)( 5850 unsigned int /* task_t */ task, uint64_t timestamp, 5851 unsigned int /*kern_return_t*/ *kernelError); 5852 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5853 void (*dyld_process_info_release)(dyld_process_info info); 5854 5855 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, 5856 unsigned int /*kern_return_t*/ *)) 5857 dlsym(RTLD_DEFAULT, "_dyld_process_info_create"); 5858 dyld_process_info_get_cache = (void (*)(void *, void *))dlsym( 5859 RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5860 dyld_process_info_release = 5861 (void (*)(void *))dlsym(RTLD_DEFAULT, "_dyld_process_info_release"); 5862 5863 if (dyld_process_info_create && dyld_process_info_get_cache) { 5864 unsigned int /*kern_return_t */ kern_ret; 5865 dyld_process_info process_info = 5866 dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5867 if (process_info) { 5868 struct lldb_copy__dyld_process_cache_info sc_info; 5869 memset(&sc_info, 0, sizeof(struct lldb_copy__dyld_process_cache_info)); 5870 dyld_process_info_get_cache(process_info, &sc_info); 5871 if (sc_info.cacheBaseAddress != 0) { 5872 base_addr = sc_info.cacheBaseAddress; 5873 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 5874 } 5875 dyld_process_info_release(process_info); 5876 } 5877 } 5878 } 5879 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5880 LIBLLDB_LOG_PROCESS)); 5881 if (log && uuid.IsValid()) 5882 LLDB_LOGF(log, 5883 "lldb's in-memory shared cache has a UUID of %s base address of " 5884 "0x%" PRIx64, 5885 uuid.GetAsString().c_str(), base_addr); 5886 #endif 5887 } 5888 5889 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5890 if (!m_min_os_version) { 5891 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5892 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5893 const lldb::offset_t load_cmd_offset = offset; 5894 5895 version_min_command lc; 5896 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5897 break; 5898 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5899 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5900 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5901 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5902 if (m_data.GetU32(&offset, &lc.version, 5903 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5904 const uint32_t xxxx = lc.version >> 16; 5905 const uint32_t yy = (lc.version >> 8) & 0xffu; 5906 const uint32_t zz = lc.version & 0xffu; 5907 if (xxxx) { 5908 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5909 break; 5910 } 5911 } 5912 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5913 // struct build_version_command { 5914 // uint32_t cmd; /* LC_BUILD_VERSION */ 5915 // uint32_t cmdsize; /* sizeof(struct 5916 // build_version_command) plus */ 5917 // /* ntools * sizeof(struct 5918 // build_tool_version) */ 5919 // uint32_t platform; /* platform */ 5920 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5921 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded in 5922 // nibbles xxxx.yy.zz */ uint32_t ntools; /* number of 5923 // tool entries following this */ 5924 // }; 5925 5926 offset += 4; // skip platform 5927 uint32_t minos = m_data.GetU32(&offset); 5928 5929 const uint32_t xxxx = minos >> 16; 5930 const uint32_t yy = (minos >> 8) & 0xffu; 5931 const uint32_t zz = minos & 0xffu; 5932 if (xxxx) { 5933 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5934 break; 5935 } 5936 } 5937 5938 offset = load_cmd_offset + lc.cmdsize; 5939 } 5940 5941 if (!m_min_os_version) { 5942 // Set version to an empty value so we don't keep trying to 5943 m_min_os_version = llvm::VersionTuple(); 5944 } 5945 } 5946 5947 return *m_min_os_version; 5948 } 5949 5950 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { 5951 if (!m_sdk_versions.hasValue()) { 5952 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5953 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5954 const lldb::offset_t load_cmd_offset = offset; 5955 5956 version_min_command lc; 5957 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5958 break; 5959 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5960 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5961 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5962 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5963 if (m_data.GetU32(&offset, &lc.version, 5964 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5965 const uint32_t xxxx = lc.sdk >> 16; 5966 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5967 const uint32_t zz = lc.sdk & 0xffu; 5968 if (xxxx) { 5969 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 5970 break; 5971 } else { 5972 GetModule()->ReportWarning("minimum OS version load command with " 5973 "invalid (0) version found."); 5974 } 5975 } 5976 } 5977 offset = load_cmd_offset + lc.cmdsize; 5978 } 5979 5980 if (!m_sdk_versions.hasValue()) { 5981 offset = MachHeaderSizeFromMagic(m_header.magic); 5982 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5983 const lldb::offset_t load_cmd_offset = offset; 5984 5985 version_min_command lc; 5986 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5987 break; 5988 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5989 // struct build_version_command { 5990 // uint32_t cmd; /* LC_BUILD_VERSION */ 5991 // uint32_t cmdsize; /* sizeof(struct 5992 // build_version_command) plus */ 5993 // /* ntools * sizeof(struct 5994 // build_tool_version) */ 5995 // uint32_t platform; /* platform */ 5996 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5997 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded 5998 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number 5999 // of tool entries following this */ 6000 // }; 6001 6002 offset += 4; // skip platform 6003 uint32_t minos = m_data.GetU32(&offset); 6004 6005 const uint32_t xxxx = minos >> 16; 6006 const uint32_t yy = (minos >> 8) & 0xffu; 6007 const uint32_t zz = minos & 0xffu; 6008 if (xxxx) { 6009 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 6010 break; 6011 } 6012 } 6013 offset = load_cmd_offset + lc.cmdsize; 6014 } 6015 } 6016 6017 if (!m_sdk_versions.hasValue()) 6018 m_sdk_versions = llvm::VersionTuple(); 6019 } 6020 6021 return m_sdk_versions.getValue(); 6022 } 6023 6024 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 6025 return m_header.filetype == llvm::MachO::MH_DYLINKER; 6026 } 6027 6028 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 6029 return m_allow_assembly_emulation_unwind_plans; 6030 } 6031 6032 // PluginInterface protocol 6033 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 6034 return GetPluginNameStatic(); 6035 } 6036 6037 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 6038 6039 Section *ObjectFileMachO::GetMachHeaderSection() { 6040 // Find the first address of the mach header which is the first non-zero file 6041 // sized section whose file offset is zero. This is the base file address of 6042 // the mach-o file which can be subtracted from the vmaddr of the other 6043 // segments found in memory and added to the load address 6044 ModuleSP module_sp = GetModule(); 6045 if (!module_sp) 6046 return nullptr; 6047 SectionList *section_list = GetSectionList(); 6048 if (!section_list) 6049 return nullptr; 6050 const size_t num_sections = section_list->GetSize(); 6051 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6052 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 6053 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 6054 return section; 6055 } 6056 return nullptr; 6057 } 6058 6059 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 6060 if (!section) 6061 return false; 6062 const bool is_dsym = (m_header.filetype == MH_DSYM); 6063 if (section->GetFileSize() == 0 && !is_dsym) 6064 return false; 6065 if (section->IsThreadSpecific()) 6066 return false; 6067 if (GetModule().get() != section->GetModule().get()) 6068 return false; 6069 // Be careful with __LINKEDIT and __DWARF segments 6070 if (section->GetName() == GetSegmentNameLINKEDIT() || 6071 section->GetName() == GetSegmentNameDWARF()) { 6072 // Only map __LINKEDIT and __DWARF if we have an in memory image and 6073 // this isn't a kernel binary like a kext or mach_kernel. 6074 const bool is_memory_image = (bool)m_process_wp.lock(); 6075 const Strata strata = GetStrata(); 6076 if (is_memory_image == false || strata == eStrataKernel) 6077 return false; 6078 } 6079 return true; 6080 } 6081 6082 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 6083 lldb::addr_t header_load_address, const Section *header_section, 6084 const Section *section) { 6085 ModuleSP module_sp = GetModule(); 6086 if (module_sp && header_section && section && 6087 header_load_address != LLDB_INVALID_ADDRESS) { 6088 lldb::addr_t file_addr = header_section->GetFileAddress(); 6089 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 6090 return section->GetFileAddress() - file_addr + header_load_address; 6091 } 6092 return LLDB_INVALID_ADDRESS; 6093 } 6094 6095 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 6096 bool value_is_offset) { 6097 ModuleSP module_sp = GetModule(); 6098 if (!module_sp) 6099 return false; 6100 6101 SectionList *section_list = GetSectionList(); 6102 if (!section_list) 6103 return false; 6104 6105 size_t num_loaded_sections = 0; 6106 const size_t num_sections = section_list->GetSize(); 6107 6108 if (value_is_offset) { 6109 // "value" is an offset to apply to each top level segment 6110 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6111 // Iterate through the object file sections to find all of the 6112 // sections that size on disk (to avoid __PAGEZERO) and load them 6113 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6114 if (SectionIsLoadable(section_sp.get())) 6115 if (target.GetSectionLoadList().SetSectionLoadAddress( 6116 section_sp, section_sp->GetFileAddress() + value)) 6117 ++num_loaded_sections; 6118 } 6119 } else { 6120 // "value" is the new base address of the mach_header, adjust each 6121 // section accordingly 6122 6123 Section *mach_header_section = GetMachHeaderSection(); 6124 if (mach_header_section) { 6125 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6126 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6127 6128 lldb::addr_t section_load_addr = 6129 CalculateSectionLoadAddressForMemoryImage( 6130 value, mach_header_section, section_sp.get()); 6131 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6132 if (target.GetSectionLoadList().SetSectionLoadAddress( 6133 section_sp, section_load_addr)) 6134 ++num_loaded_sections; 6135 } 6136 } 6137 } 6138 } 6139 return num_loaded_sections > 0; 6140 } 6141 6142 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6143 const FileSpec &outfile, Status &error) { 6144 if (!process_sp) 6145 return false; 6146 6147 Target &target = process_sp->GetTarget(); 6148 const ArchSpec target_arch = target.GetArchitecture(); 6149 const llvm::Triple &target_triple = target_arch.GetTriple(); 6150 if (target_triple.getVendor() == llvm::Triple::Apple && 6151 (target_triple.getOS() == llvm::Triple::MacOSX || 6152 target_triple.getOS() == llvm::Triple::IOS || 6153 target_triple.getOS() == llvm::Triple::WatchOS || 6154 target_triple.getOS() == llvm::Triple::TvOS)) { 6155 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) 6156 // { 6157 bool make_core = false; 6158 switch (target_arch.GetMachine()) { 6159 case llvm::Triple::aarch64: 6160 case llvm::Triple::aarch64_32: 6161 case llvm::Triple::arm: 6162 case llvm::Triple::thumb: 6163 case llvm::Triple::x86: 6164 case llvm::Triple::x86_64: 6165 make_core = true; 6166 break; 6167 default: 6168 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6169 target_triple.str().c_str()); 6170 break; 6171 } 6172 6173 if (make_core) { 6174 std::vector<segment_command_64> segment_load_commands; 6175 // uint32_t range_info_idx = 0; 6176 MemoryRegionInfo range_info; 6177 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6178 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6179 const ByteOrder byte_order = target_arch.GetByteOrder(); 6180 if (range_error.Success()) { 6181 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6182 const addr_t addr = range_info.GetRange().GetRangeBase(); 6183 const addr_t size = range_info.GetRange().GetByteSize(); 6184 6185 if (size == 0) 6186 break; 6187 6188 // Calculate correct protections 6189 uint32_t prot = 0; 6190 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6191 prot |= VM_PROT_READ; 6192 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6193 prot |= VM_PROT_WRITE; 6194 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6195 prot |= VM_PROT_EXECUTE; 6196 6197 if (prot != 0) { 6198 uint32_t cmd_type = LC_SEGMENT_64; 6199 uint32_t segment_size = sizeof(segment_command_64); 6200 if (addr_byte_size == 4) { 6201 cmd_type = LC_SEGMENT; 6202 segment_size = sizeof(segment_command); 6203 } 6204 segment_command_64 segment = { 6205 cmd_type, // uint32_t cmd; 6206 segment_size, // uint32_t cmdsize; 6207 {0}, // char segname[16]; 6208 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6209 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6210 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6211 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6212 prot, // uint32_t maxprot; 6213 prot, // uint32_t initprot; 6214 0, // uint32_t nsects; 6215 0}; // uint32_t flags; 6216 segment_load_commands.push_back(segment); 6217 } else { 6218 // No protections and a size of 1 used to be returned from old 6219 // debugservers when we asked about a region that was past the 6220 // last memory region and it indicates the end... 6221 if (size == 1) 6222 break; 6223 } 6224 6225 range_error = process_sp->GetMemoryRegionInfo( 6226 range_info.GetRange().GetRangeEnd(), range_info); 6227 if (range_error.Fail()) 6228 break; 6229 } 6230 6231 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6232 6233 mach_header_64 mach_header; 6234 if (addr_byte_size == 8) { 6235 mach_header.magic = MH_MAGIC_64; 6236 } else { 6237 mach_header.magic = MH_MAGIC; 6238 } 6239 mach_header.cputype = target_arch.GetMachOCPUType(); 6240 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6241 mach_header.filetype = MH_CORE; 6242 mach_header.ncmds = segment_load_commands.size(); 6243 mach_header.flags = 0; 6244 mach_header.reserved = 0; 6245 ThreadList &thread_list = process_sp->GetThreadList(); 6246 const uint32_t num_threads = thread_list.GetSize(); 6247 6248 // Make an array of LC_THREAD data items. Each one contains the 6249 // contents of the LC_THREAD load command. The data doesn't contain 6250 // the load command + load command size, we will add the load command 6251 // and load command size as we emit the data. 6252 std::vector<StreamString> LC_THREAD_datas(num_threads); 6253 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6254 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6255 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6256 LC_THREAD_data.SetByteOrder(byte_order); 6257 } 6258 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) { 6259 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6260 if (thread_sp) { 6261 switch (mach_header.cputype) { 6262 case llvm::MachO::CPU_TYPE_ARM64: 6263 case llvm::MachO::CPU_TYPE_ARM64_32: 6264 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6265 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6266 break; 6267 6268 case llvm::MachO::CPU_TYPE_ARM: 6269 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6270 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6271 break; 6272 6273 case llvm::MachO::CPU_TYPE_I386: 6274 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6275 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6276 break; 6277 6278 case llvm::MachO::CPU_TYPE_X86_64: 6279 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6280 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6281 break; 6282 } 6283 } 6284 } 6285 6286 // The size of the load command is the size of the segments... 6287 if (addr_byte_size == 8) { 6288 mach_header.sizeofcmds = 6289 segment_load_commands.size() * sizeof(struct segment_command_64); 6290 } else { 6291 mach_header.sizeofcmds = 6292 segment_load_commands.size() * sizeof(struct segment_command); 6293 } 6294 6295 // and the size of all LC_THREAD load command 6296 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6297 ++mach_header.ncmds; 6298 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6299 } 6300 6301 // Write the mach header 6302 buffer.PutHex32(mach_header.magic); 6303 buffer.PutHex32(mach_header.cputype); 6304 buffer.PutHex32(mach_header.cpusubtype); 6305 buffer.PutHex32(mach_header.filetype); 6306 buffer.PutHex32(mach_header.ncmds); 6307 buffer.PutHex32(mach_header.sizeofcmds); 6308 buffer.PutHex32(mach_header.flags); 6309 if (addr_byte_size == 8) { 6310 buffer.PutHex32(mach_header.reserved); 6311 } 6312 6313 // Skip the mach header and all load commands and align to the next 6314 // 0x1000 byte boundary 6315 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6316 if (file_offset & 0x00000fff) { 6317 file_offset += 0x00001000ull; 6318 file_offset &= (~0x00001000ull + 1); 6319 } 6320 6321 for (auto &segment : segment_load_commands) { 6322 segment.fileoff = file_offset; 6323 file_offset += segment.filesize; 6324 } 6325 6326 // Write out all of the LC_THREAD load commands 6327 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6328 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6329 buffer.PutHex32(LC_THREAD); 6330 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6331 buffer.Write(LC_THREAD_data.GetString().data(), LC_THREAD_data_size); 6332 } 6333 6334 // Write out all of the segment load commands 6335 for (const auto &segment : segment_load_commands) { 6336 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6337 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6338 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6339 segment.cmd, segment.cmdsize, segment.vmaddr, 6340 segment.vmaddr + segment.vmsize, segment.fileoff, 6341 segment.filesize, segment.maxprot, segment.initprot, 6342 segment.nsects, segment.flags); 6343 6344 buffer.PutHex32(segment.cmd); 6345 buffer.PutHex32(segment.cmdsize); 6346 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6347 if (addr_byte_size == 8) { 6348 buffer.PutHex64(segment.vmaddr); 6349 buffer.PutHex64(segment.vmsize); 6350 buffer.PutHex64(segment.fileoff); 6351 buffer.PutHex64(segment.filesize); 6352 } else { 6353 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6354 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6355 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6356 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6357 } 6358 buffer.PutHex32(segment.maxprot); 6359 buffer.PutHex32(segment.initprot); 6360 buffer.PutHex32(segment.nsects); 6361 buffer.PutHex32(segment.flags); 6362 } 6363 6364 std::string core_file_path(outfile.GetPath()); 6365 auto core_file = FileSystem::Instance().Open( 6366 outfile, File::eOpenOptionWrite | File::eOpenOptionTruncate | 6367 File::eOpenOptionCanCreate); 6368 if (!core_file) { 6369 error = core_file.takeError(); 6370 } else { 6371 // Read 1 page at a time 6372 uint8_t bytes[0x1000]; 6373 // Write the mach header and load commands out to the core file 6374 size_t bytes_written = buffer.GetString().size(); 6375 error = 6376 core_file.get()->Write(buffer.GetString().data(), bytes_written); 6377 if (error.Success()) { 6378 // Now write the file data for all memory segments in the process 6379 for (const auto &segment : segment_load_commands) { 6380 if (core_file.get()->SeekFromStart(segment.fileoff) == -1) { 6381 error.SetErrorStringWithFormat( 6382 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6383 segment.fileoff, core_file_path.c_str()); 6384 break; 6385 } 6386 6387 printf("Saving %" PRId64 6388 " bytes of data for memory region at 0x%" PRIx64 "\n", 6389 segment.vmsize, segment.vmaddr); 6390 addr_t bytes_left = segment.vmsize; 6391 addr_t addr = segment.vmaddr; 6392 Status memory_read_error; 6393 while (bytes_left > 0 && error.Success()) { 6394 const size_t bytes_to_read = 6395 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6396 6397 // In a savecore setting, we don't really care about caching, 6398 // as the data is dumped and very likely never read again, 6399 // so we call ReadMemoryFromInferior to bypass it. 6400 const size_t bytes_read = process_sp->ReadMemoryFromInferior( 6401 addr, bytes, bytes_to_read, memory_read_error); 6402 6403 if (bytes_read == bytes_to_read) { 6404 size_t bytes_written = bytes_read; 6405 error = core_file.get()->Write(bytes, bytes_written); 6406 bytes_left -= bytes_read; 6407 addr += bytes_read; 6408 } else { 6409 // Some pages within regions are not readable, those should 6410 // be zero filled 6411 memset(bytes, 0, bytes_to_read); 6412 size_t bytes_written = bytes_to_read; 6413 error = core_file.get()->Write(bytes, bytes_written); 6414 bytes_left -= bytes_to_read; 6415 addr += bytes_to_read; 6416 } 6417 } 6418 } 6419 } 6420 } 6421 } else { 6422 error.SetErrorString( 6423 "process doesn't support getting memory region info"); 6424 } 6425 } 6426 return true; // This is the right plug to handle saving core files for 6427 // this process 6428 } 6429 return false; 6430 } 6431