1 //===-- ObjectFileMachO.cpp -------------------------------------*- C++ -*-===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 9 #include "llvm/ADT/StringRef.h" 10 11 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 12 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 13 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 14 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 15 #include "lldb/Core/Debugger.h" 16 #include "lldb/Core/FileSpecList.h" 17 #include "lldb/Core/Module.h" 18 #include "lldb/Core/ModuleSpec.h" 19 #include "lldb/Core/PluginManager.h" 20 #include "lldb/Core/Section.h" 21 #include "lldb/Core/StreamFile.h" 22 #include "lldb/Host/Host.h" 23 #include "lldb/Symbol/DWARFCallFrameInfo.h" 24 #include "lldb/Symbol/ObjectFile.h" 25 #include "lldb/Target/DynamicLoader.h" 26 #include "lldb/Target/MemoryRegionInfo.h" 27 #include "lldb/Target/Platform.h" 28 #include "lldb/Target/Process.h" 29 #include "lldb/Target/SectionLoadList.h" 30 #include "lldb/Target/Target.h" 31 #include "lldb/Target/Thread.h" 32 #include "lldb/Target/ThreadList.h" 33 #include "lldb/Utility/ArchSpec.h" 34 #include "lldb/Utility/DataBuffer.h" 35 #include "lldb/Utility/FileSpec.h" 36 #include "lldb/Utility/Log.h" 37 #include "lldb/Utility/RangeMap.h" 38 #include "lldb/Utility/RegisterValue.h" 39 #include "lldb/Utility/Status.h" 40 #include "lldb/Utility/StreamString.h" 41 #include "lldb/Utility/Timer.h" 42 #include "lldb/Utility/UUID.h" 43 44 #include "lldb/Host/SafeMachO.h" 45 46 #include "llvm/Support/MemoryBuffer.h" 47 48 #include "ObjectFileMachO.h" 49 50 #if defined(__APPLE__) && \ 51 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 52 // GetLLDBSharedCacheUUID() needs to call dlsym() 53 #include <dlfcn.h> 54 #endif 55 56 #ifndef __APPLE__ 57 #include "Utility/UuidCompatibility.h" 58 #else 59 #include <uuid/uuid.h> 60 #endif 61 62 #include <memory> 63 64 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 65 using namespace lldb; 66 using namespace lldb_private; 67 using namespace llvm::MachO; 68 69 // Some structure definitions needed for parsing the dyld shared cache files 70 // found on iOS devices. 71 72 struct lldb_copy_dyld_cache_header_v1 { 73 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 74 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 75 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 76 uint32_t imagesOffset; 77 uint32_t imagesCount; 78 uint64_t dyldBaseAddress; 79 uint64_t codeSignatureOffset; 80 uint64_t codeSignatureSize; 81 uint64_t slideInfoOffset; 82 uint64_t slideInfoSize; 83 uint64_t localSymbolsOffset; 84 uint64_t localSymbolsSize; 85 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 86 // and later 87 }; 88 89 struct lldb_copy_dyld_cache_mapping_info { 90 uint64_t address; 91 uint64_t size; 92 uint64_t fileOffset; 93 uint32_t maxProt; 94 uint32_t initProt; 95 }; 96 97 struct lldb_copy_dyld_cache_local_symbols_info { 98 uint32_t nlistOffset; 99 uint32_t nlistCount; 100 uint32_t stringsOffset; 101 uint32_t stringsSize; 102 uint32_t entriesOffset; 103 uint32_t entriesCount; 104 }; 105 struct lldb_copy_dyld_cache_local_symbols_entry { 106 uint32_t dylibOffset; 107 uint32_t nlistStartIndex; 108 uint32_t nlistCount; 109 }; 110 111 static void PrintRegisterValue(RegisterContext *reg_ctx, const char *name, 112 const char *alt_name, size_t reg_byte_size, 113 Stream &data) { 114 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 115 if (reg_info == nullptr) 116 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 117 if (reg_info) { 118 lldb_private::RegisterValue reg_value; 119 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 120 if (reg_info->byte_size >= reg_byte_size) 121 data.Write(reg_value.GetBytes(), reg_byte_size); 122 else { 123 data.Write(reg_value.GetBytes(), reg_info->byte_size); 124 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; ++i) 125 data.PutChar(0); 126 } 127 return; 128 } 129 } 130 // Just write zeros if all else fails 131 for (size_t i = 0; i < reg_byte_size; ++i) 132 data.PutChar(0); 133 } 134 135 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 136 public: 137 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 138 const DataExtractor &data) 139 : RegisterContextDarwin_x86_64(thread, 0) { 140 SetRegisterDataFrom_LC_THREAD(data); 141 } 142 143 void InvalidateAllRegisters() override { 144 // Do nothing... registers are always valid... 145 } 146 147 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 148 lldb::offset_t offset = 0; 149 SetError(GPRRegSet, Read, -1); 150 SetError(FPURegSet, Read, -1); 151 SetError(EXCRegSet, Read, -1); 152 bool done = false; 153 154 while (!done) { 155 int flavor = data.GetU32(&offset); 156 if (flavor == 0) 157 done = true; 158 else { 159 uint32_t i; 160 uint32_t count = data.GetU32(&offset); 161 switch (flavor) { 162 case GPRRegSet: 163 for (i = 0; i < count; ++i) 164 (&gpr.rax)[i] = data.GetU64(&offset); 165 SetError(GPRRegSet, Read, 0); 166 done = true; 167 168 break; 169 case FPURegSet: 170 // TODO: fill in FPU regs.... 171 // SetError (FPURegSet, Read, -1); 172 done = true; 173 174 break; 175 case EXCRegSet: 176 exc.trapno = data.GetU32(&offset); 177 exc.err = data.GetU32(&offset); 178 exc.faultvaddr = data.GetU64(&offset); 179 SetError(EXCRegSet, Read, 0); 180 done = true; 181 break; 182 case 7: 183 case 8: 184 case 9: 185 // fancy flavors that encapsulate of the above flavors... 186 break; 187 188 default: 189 done = true; 190 break; 191 } 192 } 193 } 194 } 195 196 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 197 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 198 if (reg_ctx_sp) { 199 RegisterContext *reg_ctx = reg_ctx_sp.get(); 200 201 data.PutHex32(GPRRegSet); // Flavor 202 data.PutHex32(GPRWordCount); 203 PrintRegisterValue(reg_ctx, "rax", nullptr, 8, data); 204 PrintRegisterValue(reg_ctx, "rbx", nullptr, 8, data); 205 PrintRegisterValue(reg_ctx, "rcx", nullptr, 8, data); 206 PrintRegisterValue(reg_ctx, "rdx", nullptr, 8, data); 207 PrintRegisterValue(reg_ctx, "rdi", nullptr, 8, data); 208 PrintRegisterValue(reg_ctx, "rsi", nullptr, 8, data); 209 PrintRegisterValue(reg_ctx, "rbp", nullptr, 8, data); 210 PrintRegisterValue(reg_ctx, "rsp", nullptr, 8, data); 211 PrintRegisterValue(reg_ctx, "r8", nullptr, 8, data); 212 PrintRegisterValue(reg_ctx, "r9", nullptr, 8, data); 213 PrintRegisterValue(reg_ctx, "r10", nullptr, 8, data); 214 PrintRegisterValue(reg_ctx, "r11", nullptr, 8, data); 215 PrintRegisterValue(reg_ctx, "r12", nullptr, 8, data); 216 PrintRegisterValue(reg_ctx, "r13", nullptr, 8, data); 217 PrintRegisterValue(reg_ctx, "r14", nullptr, 8, data); 218 PrintRegisterValue(reg_ctx, "r15", nullptr, 8, data); 219 PrintRegisterValue(reg_ctx, "rip", nullptr, 8, data); 220 PrintRegisterValue(reg_ctx, "rflags", nullptr, 8, data); 221 PrintRegisterValue(reg_ctx, "cs", nullptr, 8, data); 222 PrintRegisterValue(reg_ctx, "fs", nullptr, 8, data); 223 PrintRegisterValue(reg_ctx, "gs", nullptr, 8, data); 224 225 // // Write out the FPU registers 226 // const size_t fpu_byte_size = sizeof(FPU); 227 // size_t bytes_written = 0; 228 // data.PutHex32 (FPURegSet); 229 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 230 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 231 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 232 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 233 // data); // uint16_t fcw; // "fctrl" 234 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 235 // data); // uint16_t fsw; // "fstat" 236 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 237 // data); // uint8_t ftw; // "ftag" 238 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 239 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 240 // data); // uint16_t fop; // "fop" 241 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 242 // data); // uint32_t ip; // "fioff" 243 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 244 // data); // uint16_t cs; // "fiseg" 245 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 246 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 247 // data); // uint32_t dp; // "fooff" 248 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 249 // data); // uint16_t ds; // "foseg" 250 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 251 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 252 // data); // uint32_t mxcsr; 253 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 254 // 4, data);// uint32_t mxcsrmask; 255 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 256 // sizeof(MMSReg), data); 257 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 258 // sizeof(MMSReg), data); 259 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 260 // sizeof(MMSReg), data); 261 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 262 // sizeof(MMSReg), data); 263 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 264 // sizeof(MMSReg), data); 265 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 266 // sizeof(MMSReg), data); 267 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 268 // sizeof(MMSReg), data); 269 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 270 // sizeof(MMSReg), data); 271 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 272 // sizeof(XMMReg), data); 273 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 274 // sizeof(XMMReg), data); 275 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 276 // sizeof(XMMReg), data); 277 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 278 // sizeof(XMMReg), data); 279 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 280 // sizeof(XMMReg), data); 281 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 282 // sizeof(XMMReg), data); 283 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 284 // sizeof(XMMReg), data); 285 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 286 // sizeof(XMMReg), data); 287 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 288 // sizeof(XMMReg), data); 289 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 290 // sizeof(XMMReg), data); 291 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 292 // sizeof(XMMReg), data); 293 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 294 // sizeof(XMMReg), data); 295 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 296 // sizeof(XMMReg), data); 297 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 298 // sizeof(XMMReg), data); 299 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 300 // sizeof(XMMReg), data); 301 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 302 // sizeof(XMMReg), data); 303 // 304 // // Fill rest with zeros 305 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 306 // i) 307 // data.PutChar(0); 308 309 // Write out the EXC registers 310 data.PutHex32(EXCRegSet); 311 data.PutHex32(EXCWordCount); 312 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 313 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 314 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 8, data); 315 return true; 316 } 317 return false; 318 } 319 320 protected: 321 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 322 323 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 324 325 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 326 327 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 328 return 0; 329 } 330 331 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 332 return 0; 333 } 334 335 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 336 return 0; 337 } 338 }; 339 340 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 341 public: 342 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 343 const DataExtractor &data) 344 : RegisterContextDarwin_i386(thread, 0) { 345 SetRegisterDataFrom_LC_THREAD(data); 346 } 347 348 void InvalidateAllRegisters() override { 349 // Do nothing... registers are always valid... 350 } 351 352 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 353 lldb::offset_t offset = 0; 354 SetError(GPRRegSet, Read, -1); 355 SetError(FPURegSet, Read, -1); 356 SetError(EXCRegSet, Read, -1); 357 bool done = false; 358 359 while (!done) { 360 int flavor = data.GetU32(&offset); 361 if (flavor == 0) 362 done = true; 363 else { 364 uint32_t i; 365 uint32_t count = data.GetU32(&offset); 366 switch (flavor) { 367 case GPRRegSet: 368 for (i = 0; i < count; ++i) 369 (&gpr.eax)[i] = data.GetU32(&offset); 370 SetError(GPRRegSet, Read, 0); 371 done = true; 372 373 break; 374 case FPURegSet: 375 // TODO: fill in FPU regs.... 376 // SetError (FPURegSet, Read, -1); 377 done = true; 378 379 break; 380 case EXCRegSet: 381 exc.trapno = data.GetU32(&offset); 382 exc.err = data.GetU32(&offset); 383 exc.faultvaddr = data.GetU32(&offset); 384 SetError(EXCRegSet, Read, 0); 385 done = true; 386 break; 387 case 7: 388 case 8: 389 case 9: 390 // fancy flavors that encapsulate of the above flavors... 391 break; 392 393 default: 394 done = true; 395 break; 396 } 397 } 398 } 399 } 400 401 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 402 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 403 if (reg_ctx_sp) { 404 RegisterContext *reg_ctx = reg_ctx_sp.get(); 405 406 data.PutHex32(GPRRegSet); // Flavor 407 data.PutHex32(GPRWordCount); 408 PrintRegisterValue(reg_ctx, "eax", nullptr, 4, data); 409 PrintRegisterValue(reg_ctx, "ebx", nullptr, 4, data); 410 PrintRegisterValue(reg_ctx, "ecx", nullptr, 4, data); 411 PrintRegisterValue(reg_ctx, "edx", nullptr, 4, data); 412 PrintRegisterValue(reg_ctx, "edi", nullptr, 4, data); 413 PrintRegisterValue(reg_ctx, "esi", nullptr, 4, data); 414 PrintRegisterValue(reg_ctx, "ebp", nullptr, 4, data); 415 PrintRegisterValue(reg_ctx, "esp", nullptr, 4, data); 416 PrintRegisterValue(reg_ctx, "ss", nullptr, 4, data); 417 PrintRegisterValue(reg_ctx, "eflags", nullptr, 4, data); 418 PrintRegisterValue(reg_ctx, "eip", nullptr, 4, data); 419 PrintRegisterValue(reg_ctx, "cs", nullptr, 4, data); 420 PrintRegisterValue(reg_ctx, "ds", nullptr, 4, data); 421 PrintRegisterValue(reg_ctx, "es", nullptr, 4, data); 422 PrintRegisterValue(reg_ctx, "fs", nullptr, 4, data); 423 PrintRegisterValue(reg_ctx, "gs", nullptr, 4, data); 424 425 // Write out the EXC registers 426 data.PutHex32(EXCRegSet); 427 data.PutHex32(EXCWordCount); 428 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data); 429 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data); 430 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 4, data); 431 return true; 432 } 433 return false; 434 } 435 436 protected: 437 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 438 439 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 440 441 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 442 443 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 444 return 0; 445 } 446 447 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 448 return 0; 449 } 450 451 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 452 return 0; 453 } 454 }; 455 456 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 457 public: 458 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 459 const DataExtractor &data) 460 : RegisterContextDarwin_arm(thread, 0) { 461 SetRegisterDataFrom_LC_THREAD(data); 462 } 463 464 void InvalidateAllRegisters() override { 465 // Do nothing... registers are always valid... 466 } 467 468 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 469 lldb::offset_t offset = 0; 470 SetError(GPRRegSet, Read, -1); 471 SetError(FPURegSet, Read, -1); 472 SetError(EXCRegSet, Read, -1); 473 bool done = false; 474 475 while (!done) { 476 int flavor = data.GetU32(&offset); 477 uint32_t count = data.GetU32(&offset); 478 lldb::offset_t next_thread_state = offset + (count * 4); 479 switch (flavor) { 480 case GPRAltRegSet: 481 case GPRRegSet: 482 // On ARM, the CPSR register is also included in the count but it is 483 // not included in gpr.r so loop until (count-1). 484 for (uint32_t i = 0; i < (count - 1); ++i) { 485 gpr.r[i] = data.GetU32(&offset); 486 } 487 // Save cpsr explicitly. 488 gpr.cpsr = data.GetU32(&offset); 489 490 SetError(GPRRegSet, Read, 0); 491 offset = next_thread_state; 492 break; 493 494 case FPURegSet: { 495 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 496 const int fpu_reg_buf_size = sizeof(fpu.floats); 497 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 498 fpu_reg_buf) == fpu_reg_buf_size) { 499 offset += fpu_reg_buf_size; 500 fpu.fpscr = data.GetU32(&offset); 501 SetError(FPURegSet, Read, 0); 502 } else { 503 done = true; 504 } 505 } 506 offset = next_thread_state; 507 break; 508 509 case EXCRegSet: 510 if (count == 3) { 511 exc.exception = data.GetU32(&offset); 512 exc.fsr = data.GetU32(&offset); 513 exc.far = data.GetU32(&offset); 514 SetError(EXCRegSet, Read, 0); 515 } 516 done = true; 517 offset = next_thread_state; 518 break; 519 520 // Unknown register set flavor, stop trying to parse. 521 default: 522 done = true; 523 } 524 } 525 } 526 527 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 528 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 529 if (reg_ctx_sp) { 530 RegisterContext *reg_ctx = reg_ctx_sp.get(); 531 532 data.PutHex32(GPRRegSet); // Flavor 533 data.PutHex32(GPRWordCount); 534 PrintRegisterValue(reg_ctx, "r0", nullptr, 4, data); 535 PrintRegisterValue(reg_ctx, "r1", nullptr, 4, data); 536 PrintRegisterValue(reg_ctx, "r2", nullptr, 4, data); 537 PrintRegisterValue(reg_ctx, "r3", nullptr, 4, data); 538 PrintRegisterValue(reg_ctx, "r4", nullptr, 4, data); 539 PrintRegisterValue(reg_ctx, "r5", nullptr, 4, data); 540 PrintRegisterValue(reg_ctx, "r6", nullptr, 4, data); 541 PrintRegisterValue(reg_ctx, "r7", nullptr, 4, data); 542 PrintRegisterValue(reg_ctx, "r8", nullptr, 4, data); 543 PrintRegisterValue(reg_ctx, "r9", nullptr, 4, data); 544 PrintRegisterValue(reg_ctx, "r10", nullptr, 4, data); 545 PrintRegisterValue(reg_ctx, "r11", nullptr, 4, data); 546 PrintRegisterValue(reg_ctx, "r12", nullptr, 4, data); 547 PrintRegisterValue(reg_ctx, "sp", nullptr, 4, data); 548 PrintRegisterValue(reg_ctx, "lr", nullptr, 4, data); 549 PrintRegisterValue(reg_ctx, "pc", nullptr, 4, data); 550 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 551 552 // Write out the EXC registers 553 // data.PutHex32 (EXCRegSet); 554 // data.PutHex32 (EXCWordCount); 555 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 556 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 557 // WriteRegister (reg_ctx, "far", NULL, 4, data); 558 return true; 559 } 560 return false; 561 } 562 563 protected: 564 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 565 566 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 567 568 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 569 570 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 571 572 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 573 return 0; 574 } 575 576 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 577 return 0; 578 } 579 580 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 581 return 0; 582 } 583 584 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 585 return -1; 586 } 587 }; 588 589 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 590 public: 591 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 592 const DataExtractor &data) 593 : RegisterContextDarwin_arm64(thread, 0) { 594 SetRegisterDataFrom_LC_THREAD(data); 595 } 596 597 void InvalidateAllRegisters() override { 598 // Do nothing... registers are always valid... 599 } 600 601 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 602 lldb::offset_t offset = 0; 603 SetError(GPRRegSet, Read, -1); 604 SetError(FPURegSet, Read, -1); 605 SetError(EXCRegSet, Read, -1); 606 bool done = false; 607 while (!done) { 608 int flavor = data.GetU32(&offset); 609 uint32_t count = data.GetU32(&offset); 610 lldb::offset_t next_thread_state = offset + (count * 4); 611 switch (flavor) { 612 case GPRRegSet: 613 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 614 // 32-bit register) 615 if (count >= (33 * 2) + 1) { 616 for (uint32_t i = 0; i < 29; ++i) 617 gpr.x[i] = data.GetU64(&offset); 618 gpr.fp = data.GetU64(&offset); 619 gpr.lr = data.GetU64(&offset); 620 gpr.sp = data.GetU64(&offset); 621 gpr.pc = data.GetU64(&offset); 622 gpr.cpsr = data.GetU32(&offset); 623 SetError(GPRRegSet, Read, 0); 624 } 625 offset = next_thread_state; 626 break; 627 case FPURegSet: { 628 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 629 const int fpu_reg_buf_size = sizeof(fpu); 630 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 631 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 632 fpu_reg_buf) == fpu_reg_buf_size) { 633 SetError(FPURegSet, Read, 0); 634 } else { 635 done = true; 636 } 637 } 638 offset = next_thread_state; 639 break; 640 case EXCRegSet: 641 if (count == 4) { 642 exc.far = data.GetU64(&offset); 643 exc.esr = data.GetU32(&offset); 644 exc.exception = data.GetU32(&offset); 645 SetError(EXCRegSet, Read, 0); 646 } 647 offset = next_thread_state; 648 break; 649 default: 650 done = true; 651 break; 652 } 653 } 654 } 655 656 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 657 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 658 if (reg_ctx_sp) { 659 RegisterContext *reg_ctx = reg_ctx_sp.get(); 660 661 data.PutHex32(GPRRegSet); // Flavor 662 data.PutHex32(GPRWordCount); 663 PrintRegisterValue(reg_ctx, "x0", nullptr, 8, data); 664 PrintRegisterValue(reg_ctx, "x1", nullptr, 8, data); 665 PrintRegisterValue(reg_ctx, "x2", nullptr, 8, data); 666 PrintRegisterValue(reg_ctx, "x3", nullptr, 8, data); 667 PrintRegisterValue(reg_ctx, "x4", nullptr, 8, data); 668 PrintRegisterValue(reg_ctx, "x5", nullptr, 8, data); 669 PrintRegisterValue(reg_ctx, "x6", nullptr, 8, data); 670 PrintRegisterValue(reg_ctx, "x7", nullptr, 8, data); 671 PrintRegisterValue(reg_ctx, "x8", nullptr, 8, data); 672 PrintRegisterValue(reg_ctx, "x9", nullptr, 8, data); 673 PrintRegisterValue(reg_ctx, "x10", nullptr, 8, data); 674 PrintRegisterValue(reg_ctx, "x11", nullptr, 8, data); 675 PrintRegisterValue(reg_ctx, "x12", nullptr, 8, data); 676 PrintRegisterValue(reg_ctx, "x13", nullptr, 8, data); 677 PrintRegisterValue(reg_ctx, "x14", nullptr, 8, data); 678 PrintRegisterValue(reg_ctx, "x15", nullptr, 8, data); 679 PrintRegisterValue(reg_ctx, "x16", nullptr, 8, data); 680 PrintRegisterValue(reg_ctx, "x17", nullptr, 8, data); 681 PrintRegisterValue(reg_ctx, "x18", nullptr, 8, data); 682 PrintRegisterValue(reg_ctx, "x19", nullptr, 8, data); 683 PrintRegisterValue(reg_ctx, "x20", nullptr, 8, data); 684 PrintRegisterValue(reg_ctx, "x21", nullptr, 8, data); 685 PrintRegisterValue(reg_ctx, "x22", nullptr, 8, data); 686 PrintRegisterValue(reg_ctx, "x23", nullptr, 8, data); 687 PrintRegisterValue(reg_ctx, "x24", nullptr, 8, data); 688 PrintRegisterValue(reg_ctx, "x25", nullptr, 8, data); 689 PrintRegisterValue(reg_ctx, "x26", nullptr, 8, data); 690 PrintRegisterValue(reg_ctx, "x27", nullptr, 8, data); 691 PrintRegisterValue(reg_ctx, "x28", nullptr, 8, data); 692 PrintRegisterValue(reg_ctx, "fp", nullptr, 8, data); 693 PrintRegisterValue(reg_ctx, "lr", nullptr, 8, data); 694 PrintRegisterValue(reg_ctx, "sp", nullptr, 8, data); 695 PrintRegisterValue(reg_ctx, "pc", nullptr, 8, data); 696 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data); 697 698 // Write out the EXC registers 699 // data.PutHex32 (EXCRegSet); 700 // data.PutHex32 (EXCWordCount); 701 // WriteRegister (reg_ctx, "far", NULL, 8, data); 702 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 703 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 704 return true; 705 } 706 return false; 707 } 708 709 protected: 710 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 711 712 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 713 714 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 715 716 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 717 718 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 719 return 0; 720 } 721 722 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 723 return 0; 724 } 725 726 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 727 return 0; 728 } 729 730 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 731 return -1; 732 } 733 }; 734 735 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 736 switch (magic) { 737 case MH_MAGIC: 738 case MH_CIGAM: 739 return sizeof(struct mach_header); 740 741 case MH_MAGIC_64: 742 case MH_CIGAM_64: 743 return sizeof(struct mach_header_64); 744 break; 745 746 default: 747 break; 748 } 749 return 0; 750 } 751 752 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 753 754 char ObjectFileMachO::ID; 755 756 void ObjectFileMachO::Initialize() { 757 PluginManager::RegisterPlugin( 758 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 759 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 760 } 761 762 void ObjectFileMachO::Terminate() { 763 PluginManager::UnregisterPlugin(CreateInstance); 764 } 765 766 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 767 static ConstString g_name("mach-o"); 768 return g_name; 769 } 770 771 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 772 return "Mach-o object file reader (32 and 64 bit)"; 773 } 774 775 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 776 DataBufferSP &data_sp, 777 lldb::offset_t data_offset, 778 const FileSpec *file, 779 lldb::offset_t file_offset, 780 lldb::offset_t length) { 781 if (!data_sp) { 782 data_sp = MapFileData(*file, length, file_offset); 783 if (!data_sp) 784 return nullptr; 785 data_offset = 0; 786 } 787 788 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 789 return nullptr; 790 791 // Update the data to contain the entire file if it doesn't already 792 if (data_sp->GetByteSize() < length) { 793 data_sp = MapFileData(*file, length, file_offset); 794 if (!data_sp) 795 return nullptr; 796 data_offset = 0; 797 } 798 auto objfile_up = std::make_unique<ObjectFileMachO>( 799 module_sp, data_sp, data_offset, file, file_offset, length); 800 if (!objfile_up || !objfile_up->ParseHeader()) 801 return nullptr; 802 803 return objfile_up.release(); 804 } 805 806 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 807 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 808 const ProcessSP &process_sp, lldb::addr_t header_addr) { 809 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 810 std::unique_ptr<ObjectFile> objfile_up( 811 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 812 if (objfile_up.get() && objfile_up->ParseHeader()) 813 return objfile_up.release(); 814 } 815 return nullptr; 816 } 817 818 size_t ObjectFileMachO::GetModuleSpecifications( 819 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 820 lldb::offset_t data_offset, lldb::offset_t file_offset, 821 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 822 const size_t initial_count = specs.GetSize(); 823 824 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 825 DataExtractor data; 826 data.SetData(data_sp); 827 llvm::MachO::mach_header header; 828 if (ParseHeader(data, &data_offset, header)) { 829 size_t header_and_load_cmds = 830 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 831 if (header_and_load_cmds >= data_sp->GetByteSize()) { 832 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 833 data.SetData(data_sp); 834 data_offset = MachHeaderSizeFromMagic(header.magic); 835 } 836 if (data_sp) { 837 ModuleSpec base_spec; 838 base_spec.GetFileSpec() = file; 839 base_spec.SetObjectOffset(file_offset); 840 base_spec.SetObjectSize(length); 841 GetAllArchSpecs(header, data, data_offset, base_spec, specs); 842 } 843 } 844 } 845 return specs.GetSize() - initial_count; 846 } 847 848 ConstString ObjectFileMachO::GetSegmentNameTEXT() { 849 static ConstString g_segment_name_TEXT("__TEXT"); 850 return g_segment_name_TEXT; 851 } 852 853 ConstString ObjectFileMachO::GetSegmentNameDATA() { 854 static ConstString g_segment_name_DATA("__DATA"); 855 return g_segment_name_DATA; 856 } 857 858 ConstString ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 859 static ConstString g_segment_name("__DATA_DIRTY"); 860 return g_segment_name; 861 } 862 863 ConstString ObjectFileMachO::GetSegmentNameDATA_CONST() { 864 static ConstString g_segment_name("__DATA_CONST"); 865 return g_segment_name; 866 } 867 868 ConstString ObjectFileMachO::GetSegmentNameOBJC() { 869 static ConstString g_segment_name_OBJC("__OBJC"); 870 return g_segment_name_OBJC; 871 } 872 873 ConstString ObjectFileMachO::GetSegmentNameLINKEDIT() { 874 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 875 return g_section_name_LINKEDIT; 876 } 877 878 ConstString ObjectFileMachO::GetSegmentNameDWARF() { 879 static ConstString g_section_name("__DWARF"); 880 return g_section_name; 881 } 882 883 ConstString ObjectFileMachO::GetSectionNameEHFrame() { 884 static ConstString g_section_name_eh_frame("__eh_frame"); 885 return g_section_name_eh_frame; 886 } 887 888 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 889 lldb::addr_t data_offset, 890 lldb::addr_t data_length) { 891 DataExtractor data; 892 data.SetData(data_sp, data_offset, data_length); 893 lldb::offset_t offset = 0; 894 uint32_t magic = data.GetU32(&offset); 895 return MachHeaderSizeFromMagic(magic) != 0; 896 } 897 898 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 899 DataBufferSP &data_sp, 900 lldb::offset_t data_offset, 901 const FileSpec *file, 902 lldb::offset_t file_offset, 903 lldb::offset_t length) 904 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 905 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 906 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 907 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 908 ::memset(&m_header, 0, sizeof(m_header)); 909 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 910 } 911 912 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 913 lldb::DataBufferSP &header_data_sp, 914 const lldb::ProcessSP &process_sp, 915 lldb::addr_t header_addr) 916 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 917 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 918 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 919 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 920 ::memset(&m_header, 0, sizeof(m_header)); 921 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 922 } 923 924 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 925 lldb::offset_t *data_offset_ptr, 926 llvm::MachO::mach_header &header) { 927 data.SetByteOrder(endian::InlHostByteOrder()); 928 // Leave magic in the original byte order 929 header.magic = data.GetU32(data_offset_ptr); 930 bool can_parse = false; 931 bool is_64_bit = false; 932 switch (header.magic) { 933 case MH_MAGIC: 934 data.SetByteOrder(endian::InlHostByteOrder()); 935 data.SetAddressByteSize(4); 936 can_parse = true; 937 break; 938 939 case MH_MAGIC_64: 940 data.SetByteOrder(endian::InlHostByteOrder()); 941 data.SetAddressByteSize(8); 942 can_parse = true; 943 is_64_bit = true; 944 break; 945 946 case MH_CIGAM: 947 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 948 ? eByteOrderLittle 949 : eByteOrderBig); 950 data.SetAddressByteSize(4); 951 can_parse = true; 952 break; 953 954 case MH_CIGAM_64: 955 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 956 ? eByteOrderLittle 957 : eByteOrderBig); 958 data.SetAddressByteSize(8); 959 is_64_bit = true; 960 can_parse = true; 961 break; 962 963 default: 964 break; 965 } 966 967 if (can_parse) { 968 data.GetU32(data_offset_ptr, &header.cputype, 6); 969 if (is_64_bit) 970 *data_offset_ptr += 4; 971 return true; 972 } else { 973 memset(&header, 0, sizeof(header)); 974 } 975 return false; 976 } 977 978 bool ObjectFileMachO::ParseHeader() { 979 ModuleSP module_sp(GetModule()); 980 if (!module_sp) 981 return false; 982 983 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 984 bool can_parse = false; 985 lldb::offset_t offset = 0; 986 m_data.SetByteOrder(endian::InlHostByteOrder()); 987 // Leave magic in the original byte order 988 m_header.magic = m_data.GetU32(&offset); 989 switch (m_header.magic) { 990 case MH_MAGIC: 991 m_data.SetByteOrder(endian::InlHostByteOrder()); 992 m_data.SetAddressByteSize(4); 993 can_parse = true; 994 break; 995 996 case MH_MAGIC_64: 997 m_data.SetByteOrder(endian::InlHostByteOrder()); 998 m_data.SetAddressByteSize(8); 999 can_parse = true; 1000 break; 1001 1002 case MH_CIGAM: 1003 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1004 ? eByteOrderLittle 1005 : eByteOrderBig); 1006 m_data.SetAddressByteSize(4); 1007 can_parse = true; 1008 break; 1009 1010 case MH_CIGAM_64: 1011 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1012 ? eByteOrderLittle 1013 : eByteOrderBig); 1014 m_data.SetAddressByteSize(8); 1015 can_parse = true; 1016 break; 1017 1018 default: 1019 break; 1020 } 1021 1022 if (can_parse) { 1023 m_data.GetU32(&offset, &m_header.cputype, 6); 1024 1025 ModuleSpecList all_specs; 1026 ModuleSpec base_spec; 1027 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 1028 base_spec, all_specs); 1029 1030 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 1031 ArchSpec mach_arch = 1032 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 1033 1034 // Check if the module has a required architecture 1035 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1036 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1037 continue; 1038 1039 if (SetModulesArchitecture(mach_arch)) { 1040 const size_t header_and_lc_size = 1041 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1042 if (m_data.GetByteSize() < header_and_lc_size) { 1043 DataBufferSP data_sp; 1044 ProcessSP process_sp(m_process_wp.lock()); 1045 if (process_sp) { 1046 data_sp = ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1047 } else { 1048 // Read in all only the load command data from the file on disk 1049 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset); 1050 if (data_sp->GetByteSize() != header_and_lc_size) 1051 continue; 1052 } 1053 if (data_sp) 1054 m_data.SetData(data_sp); 1055 } 1056 } 1057 return true; 1058 } 1059 // None found. 1060 return false; 1061 } else { 1062 memset(&m_header, 0, sizeof(struct mach_header)); 1063 } 1064 return false; 1065 } 1066 1067 ByteOrder ObjectFileMachO::GetByteOrder() const { 1068 return m_data.GetByteOrder(); 1069 } 1070 1071 bool ObjectFileMachO::IsExecutable() const { 1072 return m_header.filetype == MH_EXECUTE; 1073 } 1074 1075 bool ObjectFileMachO::IsDynamicLoader() const { 1076 return m_header.filetype == MH_DYLINKER; 1077 } 1078 1079 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1080 return m_data.GetAddressByteSize(); 1081 } 1082 1083 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1084 Symtab *symtab = GetSymtab(); 1085 if (!symtab) 1086 return AddressClass::eUnknown; 1087 1088 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1089 if (symbol) { 1090 if (symbol->ValueIsAddress()) { 1091 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1092 if (section_sp) { 1093 const lldb::SectionType section_type = section_sp->GetType(); 1094 switch (section_type) { 1095 case eSectionTypeInvalid: 1096 return AddressClass::eUnknown; 1097 1098 case eSectionTypeCode: 1099 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1100 // For ARM we have a bit in the n_desc field of the symbol that 1101 // tells us ARM/Thumb which is bit 0x0008. 1102 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1103 return AddressClass::eCodeAlternateISA; 1104 } 1105 return AddressClass::eCode; 1106 1107 case eSectionTypeContainer: 1108 return AddressClass::eUnknown; 1109 1110 case eSectionTypeData: 1111 case eSectionTypeDataCString: 1112 case eSectionTypeDataCStringPointers: 1113 case eSectionTypeDataSymbolAddress: 1114 case eSectionTypeData4: 1115 case eSectionTypeData8: 1116 case eSectionTypeData16: 1117 case eSectionTypeDataPointers: 1118 case eSectionTypeZeroFill: 1119 case eSectionTypeDataObjCMessageRefs: 1120 case eSectionTypeDataObjCCFStrings: 1121 case eSectionTypeGoSymtab: 1122 return AddressClass::eData; 1123 1124 case eSectionTypeDebug: 1125 case eSectionTypeDWARFDebugAbbrev: 1126 case eSectionTypeDWARFDebugAbbrevDwo: 1127 case eSectionTypeDWARFDebugAddr: 1128 case eSectionTypeDWARFDebugAranges: 1129 case eSectionTypeDWARFDebugCuIndex: 1130 case eSectionTypeDWARFDebugFrame: 1131 case eSectionTypeDWARFDebugInfo: 1132 case eSectionTypeDWARFDebugInfoDwo: 1133 case eSectionTypeDWARFDebugLine: 1134 case eSectionTypeDWARFDebugLineStr: 1135 case eSectionTypeDWARFDebugLoc: 1136 case eSectionTypeDWARFDebugLocDwo: 1137 case eSectionTypeDWARFDebugLocLists: 1138 case eSectionTypeDWARFDebugLocListsDwo: 1139 case eSectionTypeDWARFDebugMacInfo: 1140 case eSectionTypeDWARFDebugMacro: 1141 case eSectionTypeDWARFDebugNames: 1142 case eSectionTypeDWARFDebugPubNames: 1143 case eSectionTypeDWARFDebugPubTypes: 1144 case eSectionTypeDWARFDebugRanges: 1145 case eSectionTypeDWARFDebugRngLists: 1146 case eSectionTypeDWARFDebugRngListsDwo: 1147 case eSectionTypeDWARFDebugStr: 1148 case eSectionTypeDWARFDebugStrDwo: 1149 case eSectionTypeDWARFDebugStrOffsets: 1150 case eSectionTypeDWARFDebugStrOffsetsDwo: 1151 case eSectionTypeDWARFDebugTypes: 1152 case eSectionTypeDWARFDebugTypesDwo: 1153 case eSectionTypeDWARFAppleNames: 1154 case eSectionTypeDWARFAppleTypes: 1155 case eSectionTypeDWARFAppleNamespaces: 1156 case eSectionTypeDWARFAppleObjC: 1157 case eSectionTypeDWARFGNUDebugAltLink: 1158 return AddressClass::eDebug; 1159 1160 case eSectionTypeEHFrame: 1161 case eSectionTypeARMexidx: 1162 case eSectionTypeARMextab: 1163 case eSectionTypeCompactUnwind: 1164 return AddressClass::eRuntime; 1165 1166 case eSectionTypeAbsoluteAddress: 1167 case eSectionTypeELFSymbolTable: 1168 case eSectionTypeELFDynamicSymbols: 1169 case eSectionTypeELFRelocationEntries: 1170 case eSectionTypeELFDynamicLinkInfo: 1171 case eSectionTypeOther: 1172 return AddressClass::eUnknown; 1173 } 1174 } 1175 } 1176 1177 const SymbolType symbol_type = symbol->GetType(); 1178 switch (symbol_type) { 1179 case eSymbolTypeAny: 1180 return AddressClass::eUnknown; 1181 case eSymbolTypeAbsolute: 1182 return AddressClass::eUnknown; 1183 1184 case eSymbolTypeCode: 1185 case eSymbolTypeTrampoline: 1186 case eSymbolTypeResolver: 1187 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1188 // For ARM we have a bit in the n_desc field of the symbol that tells 1189 // us ARM/Thumb which is bit 0x0008. 1190 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1191 return AddressClass::eCodeAlternateISA; 1192 } 1193 return AddressClass::eCode; 1194 1195 case eSymbolTypeData: 1196 return AddressClass::eData; 1197 case eSymbolTypeRuntime: 1198 return AddressClass::eRuntime; 1199 case eSymbolTypeException: 1200 return AddressClass::eRuntime; 1201 case eSymbolTypeSourceFile: 1202 return AddressClass::eDebug; 1203 case eSymbolTypeHeaderFile: 1204 return AddressClass::eDebug; 1205 case eSymbolTypeObjectFile: 1206 return AddressClass::eDebug; 1207 case eSymbolTypeCommonBlock: 1208 return AddressClass::eDebug; 1209 case eSymbolTypeBlock: 1210 return AddressClass::eDebug; 1211 case eSymbolTypeLocal: 1212 return AddressClass::eData; 1213 case eSymbolTypeParam: 1214 return AddressClass::eData; 1215 case eSymbolTypeVariable: 1216 return AddressClass::eData; 1217 case eSymbolTypeVariableType: 1218 return AddressClass::eDebug; 1219 case eSymbolTypeLineEntry: 1220 return AddressClass::eDebug; 1221 case eSymbolTypeLineHeader: 1222 return AddressClass::eDebug; 1223 case eSymbolTypeScopeBegin: 1224 return AddressClass::eDebug; 1225 case eSymbolTypeScopeEnd: 1226 return AddressClass::eDebug; 1227 case eSymbolTypeAdditional: 1228 return AddressClass::eUnknown; 1229 case eSymbolTypeCompiler: 1230 return AddressClass::eDebug; 1231 case eSymbolTypeInstrumentation: 1232 return AddressClass::eDebug; 1233 case eSymbolTypeUndefined: 1234 return AddressClass::eUnknown; 1235 case eSymbolTypeObjCClass: 1236 return AddressClass::eRuntime; 1237 case eSymbolTypeObjCMetaClass: 1238 return AddressClass::eRuntime; 1239 case eSymbolTypeObjCIVar: 1240 return AddressClass::eRuntime; 1241 case eSymbolTypeReExported: 1242 return AddressClass::eRuntime; 1243 } 1244 } 1245 return AddressClass::eUnknown; 1246 } 1247 1248 Symtab *ObjectFileMachO::GetSymtab() { 1249 ModuleSP module_sp(GetModule()); 1250 if (module_sp) { 1251 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1252 if (m_symtab_up == nullptr) { 1253 m_symtab_up.reset(new Symtab(this)); 1254 std::lock_guard<std::recursive_mutex> symtab_guard( 1255 m_symtab_up->GetMutex()); 1256 ParseSymtab(); 1257 m_symtab_up->Finalize(); 1258 } 1259 } 1260 return m_symtab_up.get(); 1261 } 1262 1263 bool ObjectFileMachO::IsStripped() { 1264 if (m_dysymtab.cmd == 0) { 1265 ModuleSP module_sp(GetModule()); 1266 if (module_sp) { 1267 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1268 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1269 const lldb::offset_t load_cmd_offset = offset; 1270 1271 load_command lc; 1272 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 1273 break; 1274 if (lc.cmd == LC_DYSYMTAB) { 1275 m_dysymtab.cmd = lc.cmd; 1276 m_dysymtab.cmdsize = lc.cmdsize; 1277 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1278 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1279 nullptr) { 1280 // Clear m_dysymtab if we were unable to read all items from the 1281 // load command 1282 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1283 } 1284 } 1285 offset = load_cmd_offset + lc.cmdsize; 1286 } 1287 } 1288 } 1289 if (m_dysymtab.cmd) 1290 return m_dysymtab.nlocalsym <= 1; 1291 return false; 1292 } 1293 1294 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1295 EncryptedFileRanges result; 1296 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1297 1298 encryption_info_command encryption_cmd; 1299 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1300 const lldb::offset_t load_cmd_offset = offset; 1301 if (m_data.GetU32(&offset, &encryption_cmd, 2) == nullptr) 1302 break; 1303 1304 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the 1305 // 3 fields we care about, so treat them the same. 1306 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1307 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1308 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1309 if (encryption_cmd.cryptid != 0) { 1310 EncryptedFileRanges::Entry entry; 1311 entry.SetRangeBase(encryption_cmd.cryptoff); 1312 entry.SetByteSize(encryption_cmd.cryptsize); 1313 result.Append(entry); 1314 } 1315 } 1316 } 1317 offset = load_cmd_offset + encryption_cmd.cmdsize; 1318 } 1319 1320 return result; 1321 } 1322 1323 void ObjectFileMachO::SanitizeSegmentCommand(segment_command_64 &seg_cmd, 1324 uint32_t cmd_idx) { 1325 if (m_length == 0 || seg_cmd.filesize == 0) 1326 return; 1327 1328 if (seg_cmd.fileoff > m_length) { 1329 // We have a load command that says it extends past the end of the file. 1330 // This is likely a corrupt file. We don't have any way to return an error 1331 // condition here (this method was likely invoked from something like 1332 // ObjectFile::GetSectionList()), so we just null out the section contents, 1333 // and dump a message to stdout. The most common case here is core file 1334 // debugging with a truncated file. 1335 const char *lc_segment_name = 1336 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1337 GetModule()->ReportWarning( 1338 "load command %u %s has a fileoff (0x%" PRIx64 1339 ") that extends beyond the end of the file (0x%" PRIx64 1340 "), ignoring this section", 1341 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1342 1343 seg_cmd.fileoff = 0; 1344 seg_cmd.filesize = 0; 1345 } 1346 1347 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1348 // We have a load command that says it extends past the end of the file. 1349 // This is likely a corrupt file. We don't have any way to return an error 1350 // condition here (this method was likely invoked from something like 1351 // ObjectFile::GetSectionList()), so we just null out the section contents, 1352 // and dump a message to stdout. The most common case here is core file 1353 // debugging with a truncated file. 1354 const char *lc_segment_name = 1355 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1356 GetModule()->ReportWarning( 1357 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1358 ") that extends beyond the end of the file (0x%" PRIx64 1359 "), the segment will be truncated to match", 1360 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1361 1362 // Truncate the length 1363 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1364 } 1365 } 1366 1367 static uint32_t GetSegmentPermissions(const segment_command_64 &seg_cmd) { 1368 uint32_t result = 0; 1369 if (seg_cmd.initprot & VM_PROT_READ) 1370 result |= ePermissionsReadable; 1371 if (seg_cmd.initprot & VM_PROT_WRITE) 1372 result |= ePermissionsWritable; 1373 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1374 result |= ePermissionsExecutable; 1375 return result; 1376 } 1377 1378 static lldb::SectionType GetSectionType(uint32_t flags, 1379 ConstString section_name) { 1380 1381 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1382 return eSectionTypeCode; 1383 1384 uint32_t mach_sect_type = flags & SECTION_TYPE; 1385 static ConstString g_sect_name_objc_data("__objc_data"); 1386 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1387 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1388 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1389 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1390 static ConstString g_sect_name_objc_const("__objc_const"); 1391 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1392 static ConstString g_sect_name_cfstring("__cfstring"); 1393 1394 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1395 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1396 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1397 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1398 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1399 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1400 static ConstString g_sect_name_dwarf_debug_loclists("__debug_loclists"); 1401 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1402 static ConstString g_sect_name_dwarf_debug_names("__debug_names"); 1403 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1404 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1405 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1406 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1407 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1408 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1409 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1410 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1411 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1412 static ConstString g_sect_name_eh_frame("__eh_frame"); 1413 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1414 static ConstString g_sect_name_text("__text"); 1415 static ConstString g_sect_name_data("__data"); 1416 static ConstString g_sect_name_go_symtab("__gosymtab"); 1417 1418 if (section_name == g_sect_name_dwarf_debug_abbrev) 1419 return eSectionTypeDWARFDebugAbbrev; 1420 if (section_name == g_sect_name_dwarf_debug_aranges) 1421 return eSectionTypeDWARFDebugAranges; 1422 if (section_name == g_sect_name_dwarf_debug_frame) 1423 return eSectionTypeDWARFDebugFrame; 1424 if (section_name == g_sect_name_dwarf_debug_info) 1425 return eSectionTypeDWARFDebugInfo; 1426 if (section_name == g_sect_name_dwarf_debug_line) 1427 return eSectionTypeDWARFDebugLine; 1428 if (section_name == g_sect_name_dwarf_debug_loc) 1429 return eSectionTypeDWARFDebugLoc; 1430 if (section_name == g_sect_name_dwarf_debug_loclists) 1431 return eSectionTypeDWARFDebugLocLists; 1432 if (section_name == g_sect_name_dwarf_debug_macinfo) 1433 return eSectionTypeDWARFDebugMacInfo; 1434 if (section_name == g_sect_name_dwarf_debug_names) 1435 return eSectionTypeDWARFDebugNames; 1436 if (section_name == g_sect_name_dwarf_debug_pubnames) 1437 return eSectionTypeDWARFDebugPubNames; 1438 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1439 return eSectionTypeDWARFDebugPubTypes; 1440 if (section_name == g_sect_name_dwarf_debug_ranges) 1441 return eSectionTypeDWARFDebugRanges; 1442 if (section_name == g_sect_name_dwarf_debug_str) 1443 return eSectionTypeDWARFDebugStr; 1444 if (section_name == g_sect_name_dwarf_debug_types) 1445 return eSectionTypeDWARFDebugTypes; 1446 if (section_name == g_sect_name_dwarf_apple_names) 1447 return eSectionTypeDWARFAppleNames; 1448 if (section_name == g_sect_name_dwarf_apple_types) 1449 return eSectionTypeDWARFAppleTypes; 1450 if (section_name == g_sect_name_dwarf_apple_namespaces) 1451 return eSectionTypeDWARFAppleNamespaces; 1452 if (section_name == g_sect_name_dwarf_apple_objc) 1453 return eSectionTypeDWARFAppleObjC; 1454 if (section_name == g_sect_name_objc_selrefs) 1455 return eSectionTypeDataCStringPointers; 1456 if (section_name == g_sect_name_objc_msgrefs) 1457 return eSectionTypeDataObjCMessageRefs; 1458 if (section_name == g_sect_name_eh_frame) 1459 return eSectionTypeEHFrame; 1460 if (section_name == g_sect_name_compact_unwind) 1461 return eSectionTypeCompactUnwind; 1462 if (section_name == g_sect_name_cfstring) 1463 return eSectionTypeDataObjCCFStrings; 1464 if (section_name == g_sect_name_go_symtab) 1465 return eSectionTypeGoSymtab; 1466 if (section_name == g_sect_name_objc_data || 1467 section_name == g_sect_name_objc_classrefs || 1468 section_name == g_sect_name_objc_superrefs || 1469 section_name == g_sect_name_objc_const || 1470 section_name == g_sect_name_objc_classlist) { 1471 return eSectionTypeDataPointers; 1472 } 1473 1474 switch (mach_sect_type) { 1475 // TODO: categorize sections by other flags for regular sections 1476 case S_REGULAR: 1477 if (section_name == g_sect_name_text) 1478 return eSectionTypeCode; 1479 if (section_name == g_sect_name_data) 1480 return eSectionTypeData; 1481 return eSectionTypeOther; 1482 case S_ZEROFILL: 1483 return eSectionTypeZeroFill; 1484 case S_CSTRING_LITERALS: // section with only literal C strings 1485 return eSectionTypeDataCString; 1486 case S_4BYTE_LITERALS: // section with only 4 byte literals 1487 return eSectionTypeData4; 1488 case S_8BYTE_LITERALS: // section with only 8 byte literals 1489 return eSectionTypeData8; 1490 case S_LITERAL_POINTERS: // section with only pointers to literals 1491 return eSectionTypeDataPointers; 1492 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1493 return eSectionTypeDataPointers; 1494 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1495 return eSectionTypeDataPointers; 1496 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1497 // the reserved2 field 1498 return eSectionTypeCode; 1499 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1500 // initialization 1501 return eSectionTypeDataPointers; 1502 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1503 // termination 1504 return eSectionTypeDataPointers; 1505 case S_COALESCED: 1506 return eSectionTypeOther; 1507 case S_GB_ZEROFILL: 1508 return eSectionTypeZeroFill; 1509 case S_INTERPOSING: // section with only pairs of function pointers for 1510 // interposing 1511 return eSectionTypeCode; 1512 case S_16BYTE_LITERALS: // section with only 16 byte literals 1513 return eSectionTypeData16; 1514 case S_DTRACE_DOF: 1515 return eSectionTypeDebug; 1516 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1517 return eSectionTypeDataPointers; 1518 default: 1519 return eSectionTypeOther; 1520 } 1521 } 1522 1523 struct ObjectFileMachO::SegmentParsingContext { 1524 const EncryptedFileRanges EncryptedRanges; 1525 lldb_private::SectionList &UnifiedList; 1526 uint32_t NextSegmentIdx = 0; 1527 uint32_t NextSectionIdx = 0; 1528 bool FileAddressesChanged = false; 1529 1530 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1531 lldb_private::SectionList &UnifiedList) 1532 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1533 }; 1534 1535 void ObjectFileMachO::ProcessSegmentCommand(const load_command &load_cmd_, 1536 lldb::offset_t offset, 1537 uint32_t cmd_idx, 1538 SegmentParsingContext &context) { 1539 segment_command_64 load_cmd; 1540 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1541 1542 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1543 return; 1544 1545 ModuleSP module_sp = GetModule(); 1546 const bool is_core = GetType() == eTypeCoreFile; 1547 const bool is_dsym = (m_header.filetype == MH_DSYM); 1548 bool add_section = true; 1549 bool add_to_unified = true; 1550 ConstString const_segname( 1551 load_cmd.segname, strnlen(load_cmd.segname, sizeof(load_cmd.segname))); 1552 1553 SectionSP unified_section_sp( 1554 context.UnifiedList.FindSectionByName(const_segname)); 1555 if (is_dsym && unified_section_sp) { 1556 if (const_segname == GetSegmentNameLINKEDIT()) { 1557 // We need to keep the __LINKEDIT segment private to this object file 1558 // only 1559 add_to_unified = false; 1560 } else { 1561 // This is the dSYM file and this section has already been created by the 1562 // object file, no need to create it. 1563 add_section = false; 1564 } 1565 } 1566 load_cmd.vmaddr = m_data.GetAddress(&offset); 1567 load_cmd.vmsize = m_data.GetAddress(&offset); 1568 load_cmd.fileoff = m_data.GetAddress(&offset); 1569 load_cmd.filesize = m_data.GetAddress(&offset); 1570 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1571 return; 1572 1573 SanitizeSegmentCommand(load_cmd, cmd_idx); 1574 1575 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1576 const bool segment_is_encrypted = 1577 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1578 1579 // Keep a list of mach segments around in case we need to get at data that 1580 // isn't stored in the abstracted Sections. 1581 m_mach_segments.push_back(load_cmd); 1582 1583 // Use a segment ID of the segment index shifted left by 8 so they never 1584 // conflict with any of the sections. 1585 SectionSP segment_sp; 1586 if (add_section && (const_segname || is_core)) { 1587 segment_sp = std::make_shared<Section>( 1588 module_sp, // Module to which this section belongs 1589 this, // Object file to which this sections belongs 1590 ++context.NextSegmentIdx 1591 << 8, // Section ID is the 1 based segment index 1592 // shifted right by 8 bits as not to collide with any of the 256 1593 // section IDs that are possible 1594 const_segname, // Name of this section 1595 eSectionTypeContainer, // This section is a container of other 1596 // sections. 1597 load_cmd.vmaddr, // File VM address == addresses as they are 1598 // found in the object file 1599 load_cmd.vmsize, // VM size in bytes of this section 1600 load_cmd.fileoff, // Offset to the data for this section in 1601 // the file 1602 load_cmd.filesize, // Size in bytes of this section as found 1603 // in the file 1604 0, // Segments have no alignment information 1605 load_cmd.flags); // Flags for this section 1606 1607 segment_sp->SetIsEncrypted(segment_is_encrypted); 1608 m_sections_up->AddSection(segment_sp); 1609 segment_sp->SetPermissions(segment_permissions); 1610 if (add_to_unified) 1611 context.UnifiedList.AddSection(segment_sp); 1612 } else if (unified_section_sp) { 1613 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1614 // Check to see if the module was read from memory? 1615 if (module_sp->GetObjectFile()->GetBaseAddress().IsValid()) { 1616 // We have a module that is in memory and needs to have its file 1617 // address adjusted. We need to do this because when we load a file 1618 // from memory, its addresses will be slid already, yet the addresses 1619 // in the new symbol file will still be unslid. Since everything is 1620 // stored as section offset, this shouldn't cause any problems. 1621 1622 // Make sure we've parsed the symbol table from the ObjectFile before 1623 // we go around changing its Sections. 1624 module_sp->GetObjectFile()->GetSymtab(); 1625 // eh_frame would present the same problems but we parse that on a per- 1626 // function basis as-needed so it's more difficult to remove its use of 1627 // the Sections. Realistically, the environments where this code path 1628 // will be taken will not have eh_frame sections. 1629 1630 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1631 1632 // Notify the module that the section addresses have been changed once 1633 // we're done so any file-address caches can be updated. 1634 context.FileAddressesChanged = true; 1635 } 1636 } 1637 m_sections_up->AddSection(unified_section_sp); 1638 } 1639 1640 struct section_64 sect64; 1641 ::memset(§64, 0, sizeof(sect64)); 1642 // Push a section into our mach sections for the section at index zero 1643 // (NO_SECT) if we don't have any mach sections yet... 1644 if (m_mach_sections.empty()) 1645 m_mach_sections.push_back(sect64); 1646 uint32_t segment_sect_idx; 1647 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1648 1649 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1650 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1651 ++segment_sect_idx) { 1652 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1653 sizeof(sect64.sectname)) == nullptr) 1654 break; 1655 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1656 sizeof(sect64.segname)) == nullptr) 1657 break; 1658 sect64.addr = m_data.GetAddress(&offset); 1659 sect64.size = m_data.GetAddress(&offset); 1660 1661 if (m_data.GetU32(&offset, §64.offset, num_u32s) == nullptr) 1662 break; 1663 1664 // Keep a list of mach sections around in case we need to get at data that 1665 // isn't stored in the abstracted Sections. 1666 m_mach_sections.push_back(sect64); 1667 1668 if (add_section) { 1669 ConstString section_name( 1670 sect64.sectname, strnlen(sect64.sectname, sizeof(sect64.sectname))); 1671 if (!const_segname) { 1672 // We have a segment with no name so we need to conjure up segments 1673 // that correspond to the section's segname if there isn't already such 1674 // a section. If there is such a section, we resize the section so that 1675 // it spans all sections. We also mark these sections as fake so 1676 // address matches don't hit if they land in the gaps between the child 1677 // sections. 1678 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1679 sizeof(sect64.segname)); 1680 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1681 if (segment_sp.get()) { 1682 Section *segment = segment_sp.get(); 1683 // Grow the section size as needed. 1684 const lldb::addr_t sect64_min_addr = sect64.addr; 1685 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1686 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1687 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1688 const lldb::addr_t curr_seg_max_addr = 1689 curr_seg_min_addr + curr_seg_byte_size; 1690 if (sect64_min_addr >= curr_seg_min_addr) { 1691 const lldb::addr_t new_seg_byte_size = 1692 sect64_max_addr - curr_seg_min_addr; 1693 // Only grow the section size if needed 1694 if (new_seg_byte_size > curr_seg_byte_size) 1695 segment->SetByteSize(new_seg_byte_size); 1696 } else { 1697 // We need to change the base address of the segment and adjust the 1698 // child section offsets for all existing children. 1699 const lldb::addr_t slide_amount = 1700 sect64_min_addr - curr_seg_min_addr; 1701 segment->Slide(slide_amount, false); 1702 segment->GetChildren().Slide(-slide_amount, false); 1703 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1704 } 1705 1706 // Grow the section size as needed. 1707 if (sect64.offset) { 1708 const lldb::addr_t segment_min_file_offset = 1709 segment->GetFileOffset(); 1710 const lldb::addr_t segment_max_file_offset = 1711 segment_min_file_offset + segment->GetFileSize(); 1712 1713 const lldb::addr_t section_min_file_offset = sect64.offset; 1714 const lldb::addr_t section_max_file_offset = 1715 section_min_file_offset + sect64.size; 1716 const lldb::addr_t new_file_offset = 1717 std::min(section_min_file_offset, segment_min_file_offset); 1718 const lldb::addr_t new_file_size = 1719 std::max(section_max_file_offset, segment_max_file_offset) - 1720 new_file_offset; 1721 segment->SetFileOffset(new_file_offset); 1722 segment->SetFileSize(new_file_size); 1723 } 1724 } else { 1725 // Create a fake section for the section's named segment 1726 segment_sp = std::make_shared<Section>( 1727 segment_sp, // Parent section 1728 module_sp, // Module to which this section belongs 1729 this, // Object file to which this section belongs 1730 ++context.NextSegmentIdx 1731 << 8, // Section ID is the 1 based segment index 1732 // shifted right by 8 bits as not to 1733 // collide with any of the 256 section IDs 1734 // that are possible 1735 const_segname, // Name of this section 1736 eSectionTypeContainer, // This section is a container of 1737 // other sections. 1738 sect64.addr, // File VM address == addresses as they are 1739 // found in the object file 1740 sect64.size, // VM size in bytes of this section 1741 sect64.offset, // Offset to the data for this section in 1742 // the file 1743 sect64.offset ? sect64.size : 0, // Size in bytes of 1744 // this section as 1745 // found in the file 1746 sect64.align, 1747 load_cmd.flags); // Flags for this section 1748 segment_sp->SetIsFake(true); 1749 segment_sp->SetPermissions(segment_permissions); 1750 m_sections_up->AddSection(segment_sp); 1751 if (add_to_unified) 1752 context.UnifiedList.AddSection(segment_sp); 1753 segment_sp->SetIsEncrypted(segment_is_encrypted); 1754 } 1755 } 1756 assert(segment_sp.get()); 1757 1758 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1759 1760 SectionSP section_sp(new Section( 1761 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1762 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1763 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1764 sect64.flags)); 1765 // Set the section to be encrypted to match the segment 1766 1767 bool section_is_encrypted = false; 1768 if (!segment_is_encrypted && load_cmd.filesize != 0) 1769 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1770 sect64.offset) != nullptr; 1771 1772 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1773 section_sp->SetPermissions(segment_permissions); 1774 segment_sp->GetChildren().AddSection(section_sp); 1775 1776 if (segment_sp->IsFake()) { 1777 segment_sp.reset(); 1778 const_segname.Clear(); 1779 } 1780 } 1781 } 1782 if (segment_sp && is_dsym) { 1783 if (first_segment_sectID <= context.NextSectionIdx) { 1784 lldb::user_id_t sect_uid; 1785 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1786 ++sect_uid) { 1787 SectionSP curr_section_sp( 1788 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1789 SectionSP next_section_sp; 1790 if (sect_uid + 1 <= context.NextSectionIdx) 1791 next_section_sp = 1792 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1793 1794 if (curr_section_sp.get()) { 1795 if (curr_section_sp->GetByteSize() == 0) { 1796 if (next_section_sp.get() != nullptr) 1797 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1798 curr_section_sp->GetFileAddress()); 1799 else 1800 curr_section_sp->SetByteSize(load_cmd.vmsize); 1801 } 1802 } 1803 } 1804 } 1805 } 1806 } 1807 1808 void ObjectFileMachO::ProcessDysymtabCommand(const load_command &load_cmd, 1809 lldb::offset_t offset) { 1810 m_dysymtab.cmd = load_cmd.cmd; 1811 m_dysymtab.cmdsize = load_cmd.cmdsize; 1812 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1813 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1814 } 1815 1816 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1817 if (m_sections_up) 1818 return; 1819 1820 m_sections_up.reset(new SectionList()); 1821 1822 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1823 // bool dump_sections = false; 1824 ModuleSP module_sp(GetModule()); 1825 1826 offset = MachHeaderSizeFromMagic(m_header.magic); 1827 1828 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1829 struct load_command load_cmd; 1830 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1831 const lldb::offset_t load_cmd_offset = offset; 1832 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 1833 break; 1834 1835 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1836 ProcessSegmentCommand(load_cmd, offset, i, context); 1837 else if (load_cmd.cmd == LC_DYSYMTAB) 1838 ProcessDysymtabCommand(load_cmd, offset); 1839 1840 offset = load_cmd_offset + load_cmd.cmdsize; 1841 } 1842 1843 if (context.FileAddressesChanged && module_sp) 1844 module_sp->SectionFileAddressesChanged(); 1845 } 1846 1847 class MachSymtabSectionInfo { 1848 public: 1849 MachSymtabSectionInfo(SectionList *section_list) 1850 : m_section_list(section_list), m_section_infos() { 1851 // Get the number of sections down to a depth of 1 to include all segments 1852 // and their sections, but no other sections that may be added for debug 1853 // map or 1854 m_section_infos.resize(section_list->GetNumSections(1)); 1855 } 1856 1857 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1858 if (n_sect == 0) 1859 return SectionSP(); 1860 if (n_sect < m_section_infos.size()) { 1861 if (!m_section_infos[n_sect].section_sp) { 1862 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1863 m_section_infos[n_sect].section_sp = section_sp; 1864 if (section_sp) { 1865 m_section_infos[n_sect].vm_range.SetBaseAddress( 1866 section_sp->GetFileAddress()); 1867 m_section_infos[n_sect].vm_range.SetByteSize( 1868 section_sp->GetByteSize()); 1869 } else { 1870 const char *filename = "<unknown>"; 1871 SectionSP first_section_sp(m_section_list->GetSectionAtIndex(0)); 1872 if (first_section_sp) 1873 filename = first_section_sp->GetObjectFile()->GetFileSpec().GetPath().c_str(); 1874 1875 Host::SystemLog(Host::eSystemLogError, 1876 "error: unable to find section %d for a symbol in %s, corrupt file?\n", 1877 n_sect, 1878 filename); 1879 } 1880 } 1881 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1882 // Symbol is in section. 1883 return m_section_infos[n_sect].section_sp; 1884 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1885 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1886 file_addr) { 1887 // Symbol is in section with zero size, but has the same start address 1888 // as the section. This can happen with linker symbols (symbols that 1889 // start with the letter 'l' or 'L'. 1890 return m_section_infos[n_sect].section_sp; 1891 } 1892 } 1893 return m_section_list->FindSectionContainingFileAddress(file_addr); 1894 } 1895 1896 protected: 1897 struct SectionInfo { 1898 SectionInfo() : vm_range(), section_sp() {} 1899 1900 VMRange vm_range; 1901 SectionSP section_sp; 1902 }; 1903 SectionList *m_section_list; 1904 std::vector<SectionInfo> m_section_infos; 1905 }; 1906 1907 struct TrieEntry { 1908 void Dump() const { 1909 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1910 static_cast<unsigned long long>(address), 1911 static_cast<unsigned long long>(flags), 1912 static_cast<unsigned long long>(other), name.GetCString()); 1913 if (import_name) 1914 printf(" -> \"%s\"\n", import_name.GetCString()); 1915 else 1916 printf("\n"); 1917 } 1918 ConstString name; 1919 uint64_t address = LLDB_INVALID_ADDRESS; 1920 uint64_t flags = 0; 1921 uint64_t other = 0; 1922 ConstString import_name; 1923 }; 1924 1925 struct TrieEntryWithOffset { 1926 lldb::offset_t nodeOffset; 1927 TrieEntry entry; 1928 1929 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1930 1931 void Dump(uint32_t idx) const { 1932 printf("[%3u] 0x%16.16llx: ", idx, 1933 static_cast<unsigned long long>(nodeOffset)); 1934 entry.Dump(); 1935 } 1936 1937 bool operator<(const TrieEntryWithOffset &other) const { 1938 return (nodeOffset < other.nodeOffset); 1939 } 1940 }; 1941 1942 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 1943 const bool is_arm, 1944 std::vector<llvm::StringRef> &nameSlices, 1945 std::set<lldb::addr_t> &resolver_addresses, 1946 std::vector<TrieEntryWithOffset> &output) { 1947 if (!data.ValidOffset(offset)) 1948 return true; 1949 1950 const uint64_t terminalSize = data.GetULEB128(&offset); 1951 lldb::offset_t children_offset = offset + terminalSize; 1952 if (terminalSize != 0) { 1953 TrieEntryWithOffset e(offset); 1954 e.entry.flags = data.GetULEB128(&offset); 1955 const char *import_name = nullptr; 1956 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 1957 e.entry.address = 0; 1958 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 1959 import_name = data.GetCStr(&offset); 1960 } else { 1961 e.entry.address = data.GetULEB128(&offset); 1962 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 1963 e.entry.other = data.GetULEB128(&offset); 1964 uint64_t resolver_addr = e.entry.other; 1965 if (is_arm) 1966 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 1967 resolver_addresses.insert(resolver_addr); 1968 } else 1969 e.entry.other = 0; 1970 } 1971 // Only add symbols that are reexport symbols with a valid import name 1972 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && 1973 import_name[0]) { 1974 std::string name; 1975 if (!nameSlices.empty()) { 1976 for (auto name_slice : nameSlices) 1977 name.append(name_slice.data(), name_slice.size()); 1978 } 1979 if (name.size() > 1) { 1980 // Skip the leading '_' 1981 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 1982 } 1983 if (import_name) { 1984 // Skip the leading '_' 1985 e.entry.import_name.SetCString(import_name + 1); 1986 } 1987 output.push_back(e); 1988 } 1989 } 1990 1991 const uint8_t childrenCount = data.GetU8(&children_offset); 1992 for (uint8_t i = 0; i < childrenCount; ++i) { 1993 const char *cstr = data.GetCStr(&children_offset); 1994 if (cstr) 1995 nameSlices.push_back(llvm::StringRef(cstr)); 1996 else 1997 return false; // Corrupt data 1998 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 1999 if (childNodeOffset) { 2000 if (!ParseTrieEntries(data, childNodeOffset, is_arm, nameSlices, 2001 resolver_addresses, output)) { 2002 return false; 2003 } 2004 } 2005 nameSlices.pop_back(); 2006 } 2007 return true; 2008 } 2009 2010 // Read the UUID out of a dyld_shared_cache file on-disk. 2011 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2012 const ByteOrder byte_order, 2013 const uint32_t addr_byte_size) { 2014 UUID dsc_uuid; 2015 DataBufferSP DscData = MapFileData( 2016 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2017 if (!DscData) 2018 return dsc_uuid; 2019 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2020 2021 char version_str[7]; 2022 lldb::offset_t offset = 0; 2023 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2024 version_str[6] = '\0'; 2025 if (strcmp(version_str, "dyld_v") == 0) { 2026 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2027 dsc_uuid = UUID::fromOptionalData( 2028 dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t)); 2029 } 2030 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2031 if (log && dsc_uuid.IsValid()) { 2032 LLDB_LOGF(log, "Shared cache %s has UUID %s", 2033 dyld_shared_cache.GetPath().c_str(), 2034 dsc_uuid.GetAsString().c_str()); 2035 } 2036 return dsc_uuid; 2037 } 2038 2039 static llvm::Optional<struct nlist_64> 2040 ParseNList(DataExtractor &nlist_data, lldb::offset_t &nlist_data_offset, 2041 size_t nlist_byte_size) { 2042 struct nlist_64 nlist; 2043 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2044 return {}; 2045 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 2046 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 2047 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 2048 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 2049 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 2050 return nlist; 2051 } 2052 2053 enum { DebugSymbols = true, NonDebugSymbols = false }; 2054 2055 size_t ObjectFileMachO::ParseSymtab() { 2056 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2057 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2058 m_file.GetFilename().AsCString("")); 2059 ModuleSP module_sp(GetModule()); 2060 if (!module_sp) 2061 return 0; 2062 2063 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2064 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2065 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2066 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2067 FunctionStarts function_starts; 2068 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2069 uint32_t i; 2070 FileSpecList dylib_files; 2071 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2072 llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2073 llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2074 llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2075 2076 for (i = 0; i < m_header.ncmds; ++i) { 2077 const lldb::offset_t cmd_offset = offset; 2078 // Read in the load command and load command size 2079 struct load_command lc; 2080 if (m_data.GetU32(&offset, &lc, 2) == nullptr) 2081 break; 2082 // Watch for the symbol table load command 2083 switch (lc.cmd) { 2084 case LC_SYMTAB: 2085 symtab_load_command.cmd = lc.cmd; 2086 symtab_load_command.cmdsize = lc.cmdsize; 2087 // Read in the rest of the symtab load command 2088 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2089 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2090 return 0; 2091 if (symtab_load_command.symoff == 0) { 2092 if (log) 2093 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2094 return 0; 2095 } 2096 2097 if (symtab_load_command.stroff == 0) { 2098 if (log) 2099 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2100 return 0; 2101 } 2102 2103 if (symtab_load_command.nsyms == 0) { 2104 if (log) 2105 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2106 return 0; 2107 } 2108 2109 if (symtab_load_command.strsize == 0) { 2110 if (log) 2111 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2112 return 0; 2113 } 2114 break; 2115 2116 case LC_DYLD_INFO: 2117 case LC_DYLD_INFO_ONLY: 2118 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2119 dyld_info.cmd = lc.cmd; 2120 dyld_info.cmdsize = lc.cmdsize; 2121 } else { 2122 memset(&dyld_info, 0, sizeof(dyld_info)); 2123 } 2124 break; 2125 2126 case LC_LOAD_DYLIB: 2127 case LC_LOAD_WEAK_DYLIB: 2128 case LC_REEXPORT_DYLIB: 2129 case LC_LOADFVMLIB: 2130 case LC_LOAD_UPWARD_DYLIB: { 2131 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2132 const char *path = m_data.PeekCStr(name_offset); 2133 if (path) { 2134 FileSpec file_spec(path); 2135 // Strip the path if there is @rpath, @executable, etc so we just use 2136 // the basename 2137 if (path[0] == '@') 2138 file_spec.GetDirectory().Clear(); 2139 2140 if (lc.cmd == LC_REEXPORT_DYLIB) { 2141 m_reexported_dylibs.AppendIfUnique(file_spec); 2142 } 2143 2144 dylib_files.Append(file_spec); 2145 } 2146 } break; 2147 2148 case LC_FUNCTION_STARTS: 2149 function_starts_load_command.cmd = lc.cmd; 2150 function_starts_load_command.cmdsize = lc.cmdsize; 2151 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2152 nullptr) // fill in symoff, nsyms, stroff, strsize fields 2153 memset(&function_starts_load_command, 0, 2154 sizeof(function_starts_load_command)); 2155 break; 2156 2157 default: 2158 break; 2159 } 2160 offset = cmd_offset + lc.cmdsize; 2161 } 2162 2163 if (!symtab_load_command.cmd) 2164 return 0; 2165 2166 Symtab *symtab = m_symtab_up.get(); 2167 SectionList *section_list = GetSectionList(); 2168 if (section_list == nullptr) 2169 return 0; 2170 2171 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2172 const ByteOrder byte_order = m_data.GetByteOrder(); 2173 bool bit_width_32 = addr_byte_size == 4; 2174 const size_t nlist_byte_size = 2175 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2176 2177 DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size); 2178 DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size); 2179 DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size); 2180 DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order, 2181 addr_byte_size); 2182 DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size); 2183 2184 const addr_t nlist_data_byte_size = 2185 symtab_load_command.nsyms * nlist_byte_size; 2186 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2187 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2188 2189 ProcessSP process_sp(m_process_wp.lock()); 2190 Process *process = process_sp.get(); 2191 2192 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2193 2194 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2195 Target &target = process->GetTarget(); 2196 2197 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2198 2199 SectionSP linkedit_section_sp( 2200 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2201 // Reading mach file from memory in a process or core file... 2202 2203 if (linkedit_section_sp) { 2204 addr_t linkedit_load_addr = 2205 linkedit_section_sp->GetLoadBaseAddress(&target); 2206 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2207 // We might be trying to access the symbol table before the 2208 // __LINKEDIT's load address has been set in the target. We can't 2209 // fail to read the symbol table, so calculate the right address 2210 // manually 2211 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2212 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2213 } 2214 2215 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2216 const addr_t symoff_addr = linkedit_load_addr + 2217 symtab_load_command.symoff - 2218 linkedit_file_offset; 2219 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2220 linkedit_file_offset; 2221 2222 bool data_was_read = false; 2223 2224 #if defined(__APPLE__) && \ 2225 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2226 if (m_header.flags & 0x80000000u && 2227 process->GetAddressByteSize() == sizeof(void *)) { 2228 // This mach-o memory file is in the dyld shared cache. If this 2229 // program is not remote and this is iOS, then this process will 2230 // share the same shared cache as the process we are debugging and we 2231 // can read the entire __LINKEDIT from the address space in this 2232 // process. This is a needed optimization that is used for local iOS 2233 // debugging only since all shared libraries in the shared cache do 2234 // not have corresponding files that exist in the file system of the 2235 // device. They have been combined into a single file. This means we 2236 // always have to load these files from memory. All of the symbol and 2237 // string tables from all of the __LINKEDIT sections from the shared 2238 // libraries in the shared cache have been merged into a single large 2239 // symbol and string table. Reading all of this symbol and string 2240 // table data across can slow down debug launch times, so we optimize 2241 // this by reading the memory for the __LINKEDIT section from this 2242 // process. 2243 2244 UUID lldb_shared_cache; 2245 addr_t lldb_shared_cache_addr; 2246 GetLLDBSharedCacheUUID(lldb_shared_cache_addr, lldb_shared_cache); 2247 UUID process_shared_cache; 2248 addr_t process_shared_cache_addr; 2249 GetProcessSharedCacheUUID(process, process_shared_cache_addr, 2250 process_shared_cache); 2251 bool use_lldb_cache = true; 2252 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2253 (lldb_shared_cache != process_shared_cache || 2254 process_shared_cache_addr != lldb_shared_cache_addr)) { 2255 use_lldb_cache = false; 2256 } 2257 2258 PlatformSP platform_sp(target.GetPlatform()); 2259 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2260 data_was_read = true; 2261 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2262 eByteOrderLittle); 2263 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2264 eByteOrderLittle); 2265 if (function_starts_load_command.cmd) { 2266 const addr_t func_start_addr = 2267 linkedit_load_addr + function_starts_load_command.dataoff - 2268 linkedit_file_offset; 2269 function_starts_data.SetData((void *)func_start_addr, 2270 function_starts_load_command.datasize, 2271 eByteOrderLittle); 2272 } 2273 } 2274 } 2275 #endif 2276 2277 if (!data_was_read) { 2278 // Always load dyld - the dynamic linker - from memory if we didn't 2279 // find a binary anywhere else. lldb will not register 2280 // dylib/framework/bundle loads/unloads if we don't have the dyld 2281 // symbols, we force dyld to load from memory despite the user's 2282 // target.memory-module-load-level setting. 2283 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2284 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2285 DataBufferSP nlist_data_sp( 2286 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2287 if (nlist_data_sp) 2288 nlist_data.SetData(nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2289 if (m_dysymtab.nindirectsyms != 0) { 2290 const addr_t indirect_syms_addr = linkedit_load_addr + 2291 m_dysymtab.indirectsymoff - 2292 linkedit_file_offset; 2293 DataBufferSP indirect_syms_data_sp(ReadMemory( 2294 process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 2295 if (indirect_syms_data_sp) 2296 indirect_symbol_index_data.SetData( 2297 indirect_syms_data_sp, 0, 2298 indirect_syms_data_sp->GetByteSize()); 2299 // If this binary is outside the shared cache, 2300 // cache the string table. 2301 // Binaries in the shared cache all share a giant string table, 2302 // and we can't share the string tables across multiple 2303 // ObjectFileMachO's, so we'd end up re-reading this mega-strtab 2304 // for every binary in the shared cache - it would be a big perf 2305 // problem. For binaries outside the shared cache, it's faster to 2306 // read the entire strtab at once instead of piece-by-piece as we 2307 // process the nlist records. 2308 if ((m_header.flags & 0x80000000u) == 0) { 2309 DataBufferSP strtab_data_sp( 2310 ReadMemory(process_sp, strtab_addr, strtab_data_byte_size)); 2311 if (strtab_data_sp) { 2312 strtab_data.SetData(strtab_data_sp, 0, 2313 strtab_data_sp->GetByteSize()); 2314 } 2315 } 2316 } 2317 } 2318 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) { 2319 if (function_starts_load_command.cmd) { 2320 const addr_t func_start_addr = 2321 linkedit_load_addr + function_starts_load_command.dataoff - 2322 linkedit_file_offset; 2323 DataBufferSP func_start_data_sp( 2324 ReadMemory(process_sp, func_start_addr, 2325 function_starts_load_command.datasize)); 2326 if (func_start_data_sp) 2327 function_starts_data.SetData(func_start_data_sp, 0, 2328 func_start_data_sp->GetByteSize()); 2329 } 2330 } 2331 } 2332 } 2333 } else { 2334 nlist_data.SetData(m_data, symtab_load_command.symoff, 2335 nlist_data_byte_size); 2336 strtab_data.SetData(m_data, symtab_load_command.stroff, 2337 strtab_data_byte_size); 2338 2339 if (dyld_info.export_size > 0) { 2340 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2341 dyld_info.export_size); 2342 } 2343 2344 if (m_dysymtab.nindirectsyms != 0) { 2345 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2346 m_dysymtab.nindirectsyms * 4); 2347 } 2348 if (function_starts_load_command.cmd) { 2349 function_starts_data.SetData(m_data, function_starts_load_command.dataoff, 2350 function_starts_load_command.datasize); 2351 } 2352 } 2353 2354 if (nlist_data.GetByteSize() == 0 && 2355 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2356 if (log) 2357 module_sp->LogMessage(log, "failed to read nlist data"); 2358 return 0; 2359 } 2360 2361 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2362 if (!have_strtab_data) { 2363 if (process) { 2364 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2365 if (log) 2366 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2367 return 0; 2368 } 2369 } else { 2370 if (log) 2371 module_sp->LogMessage(log, "failed to read strtab data"); 2372 return 0; 2373 } 2374 } 2375 2376 ConstString g_segment_name_TEXT = GetSegmentNameTEXT(); 2377 ConstString g_segment_name_DATA = GetSegmentNameDATA(); 2378 ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2379 ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2380 ConstString g_segment_name_OBJC = GetSegmentNameOBJC(); 2381 ConstString g_section_name_eh_frame = GetSectionNameEHFrame(); 2382 SectionSP text_section_sp( 2383 section_list->FindSectionByName(g_segment_name_TEXT)); 2384 SectionSP data_section_sp( 2385 section_list->FindSectionByName(g_segment_name_DATA)); 2386 SectionSP data_dirty_section_sp( 2387 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2388 SectionSP data_const_section_sp( 2389 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2390 SectionSP objc_section_sp( 2391 section_list->FindSectionByName(g_segment_name_OBJC)); 2392 SectionSP eh_frame_section_sp; 2393 if (text_section_sp.get()) 2394 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2395 g_section_name_eh_frame); 2396 else 2397 eh_frame_section_sp = 2398 section_list->FindSectionByName(g_section_name_eh_frame); 2399 2400 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2401 2402 // lldb works best if it knows the start address of all functions in a 2403 // module. Linker symbols or debug info are normally the best source of 2404 // information for start addr / size but they may be stripped in a released 2405 // binary. Two additional sources of information exist in Mach-O binaries: 2406 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2407 // function's start address in the 2408 // binary, relative to the text section. 2409 // eh_frame - the eh_frame FDEs have the start addr & size of 2410 // each function 2411 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2412 // all modern binaries. 2413 // Binaries built to run on older releases may need to use eh_frame 2414 // information. 2415 2416 if (text_section_sp && function_starts_data.GetByteSize()) { 2417 FunctionStarts::Entry function_start_entry; 2418 function_start_entry.data = false; 2419 lldb::offset_t function_start_offset = 0; 2420 function_start_entry.addr = text_section_sp->GetFileAddress(); 2421 uint64_t delta; 2422 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2423 0) { 2424 // Now append the current entry 2425 function_start_entry.addr += delta; 2426 function_starts.Append(function_start_entry); 2427 } 2428 } else { 2429 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2430 // load command claiming an eh_frame but it doesn't actually have the 2431 // eh_frame content. And if we have a dSYM, we don't need to do any of 2432 // this fill-in-the-missing-symbols works anyway - the debug info should 2433 // give us all the functions in the module. 2434 if (text_section_sp.get() && eh_frame_section_sp.get() && 2435 m_type != eTypeDebugInfo) { 2436 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2437 DWARFCallFrameInfo::EH); 2438 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2439 eh_frame.GetFunctionAddressAndSizeVector(functions); 2440 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2441 size_t count = functions.GetSize(); 2442 for (size_t i = 0; i < count; ++i) { 2443 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2444 functions.GetEntryAtIndex(i); 2445 if (func) { 2446 FunctionStarts::Entry function_start_entry; 2447 function_start_entry.addr = func->base - text_base_addr; 2448 function_starts.Append(function_start_entry); 2449 } 2450 } 2451 } 2452 } 2453 2454 const size_t function_starts_count = function_starts.GetSize(); 2455 2456 // For user process binaries (executables, dylibs, frameworks, bundles), if 2457 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2458 // going to assume the binary has been stripped. Don't allow assembly 2459 // language instruction emulation because we don't know proper function 2460 // start boundaries. 2461 // 2462 // For all other types of binaries (kernels, stand-alone bare board 2463 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2464 // sections - we should not make any assumptions about them based on that. 2465 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2466 m_allow_assembly_emulation_unwind_plans = false; 2467 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2468 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2469 2470 if (unwind_or_symbol_log) 2471 module_sp->LogMessage( 2472 unwind_or_symbol_log, 2473 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2474 } 2475 2476 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() 2477 ? eh_frame_section_sp->GetID() 2478 : static_cast<user_id_t>(NO_SECT); 2479 2480 lldb::offset_t nlist_data_offset = 0; 2481 2482 uint32_t N_SO_index = UINT32_MAX; 2483 2484 MachSymtabSectionInfo section_info(section_list); 2485 std::vector<uint32_t> N_FUN_indexes; 2486 std::vector<uint32_t> N_NSYM_indexes; 2487 std::vector<uint32_t> N_INCL_indexes; 2488 std::vector<uint32_t> N_BRAC_indexes; 2489 std::vector<uint32_t> N_COMM_indexes; 2490 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2491 typedef llvm::DenseMap<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2492 typedef llvm::DenseMap<const char *, uint32_t> ConstNameToSymbolIndexMap; 2493 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2494 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2495 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2496 // Any symbols that get merged into another will get an entry in this map 2497 // so we know 2498 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2499 uint32_t nlist_idx = 0; 2500 Symbol *symbol_ptr = nullptr; 2501 2502 uint32_t sym_idx = 0; 2503 Symbol *sym = nullptr; 2504 size_t num_syms = 0; 2505 std::string memory_symbol_name; 2506 uint32_t unmapped_local_symbols_found = 0; 2507 2508 std::vector<TrieEntryWithOffset> trie_entries; 2509 std::set<lldb::addr_t> resolver_addresses; 2510 2511 if (dyld_trie_data.GetByteSize() > 0) { 2512 std::vector<llvm::StringRef> nameSlices; 2513 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, resolver_addresses, 2514 trie_entries); 2515 2516 ConstString text_segment_name("__TEXT"); 2517 SectionSP text_segment_sp = 2518 GetSectionList()->FindSectionByName(text_segment_name); 2519 if (text_segment_sp) { 2520 const lldb::addr_t text_segment_file_addr = 2521 text_segment_sp->GetFileAddress(); 2522 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2523 for (auto &e : trie_entries) 2524 e.entry.address += text_segment_file_addr; 2525 } 2526 } 2527 } 2528 2529 typedef std::set<ConstString> IndirectSymbols; 2530 IndirectSymbols indirect_symbol_names; 2531 2532 #if defined(__APPLE__) && \ 2533 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2534 2535 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2536 // optimized by moving LOCAL symbols out of the memory mapped portion of 2537 // the DSC. The symbol information has all been retained, but it isn't 2538 // available in the normal nlist data. However, there *are* duplicate 2539 // entries of *some* 2540 // LOCAL symbols in the normal nlist data. To handle this situation 2541 // correctly, we must first attempt 2542 // to parse any DSC unmapped symbol information. If we find any, we set a 2543 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2544 2545 if (m_header.flags & 0x80000000u) { 2546 // Before we can start mapping the DSC, we need to make certain the 2547 // target process is actually using the cache we can find. 2548 2549 // Next we need to determine the correct path for the dyld shared cache. 2550 2551 ArchSpec header_arch = GetArchitecture(); 2552 char dsc_path[PATH_MAX]; 2553 char dsc_path_development[PATH_MAX]; 2554 2555 snprintf( 2556 dsc_path, sizeof(dsc_path), "%s%s%s", 2557 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2558 */ 2559 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2560 header_arch.GetArchitectureName()); 2561 2562 snprintf( 2563 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2564 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2565 */ 2566 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2567 header_arch.GetArchitectureName(), ".development"); 2568 2569 FileSpec dsc_nondevelopment_filespec(dsc_path); 2570 FileSpec dsc_development_filespec(dsc_path_development); 2571 FileSpec dsc_filespec; 2572 2573 UUID dsc_uuid; 2574 UUID process_shared_cache_uuid; 2575 addr_t process_shared_cache_base_addr; 2576 2577 if (process) { 2578 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, 2579 process_shared_cache_uuid); 2580 } 2581 2582 // First see if we can find an exact match for the inferior process 2583 // shared cache UUID in the development or non-development shared caches 2584 // on disk. 2585 if (process_shared_cache_uuid.IsValid()) { 2586 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2587 UUID dsc_development_uuid = GetSharedCacheUUID( 2588 dsc_development_filespec, byte_order, addr_byte_size); 2589 if (dsc_development_uuid.IsValid() && 2590 dsc_development_uuid == process_shared_cache_uuid) { 2591 dsc_filespec = dsc_development_filespec; 2592 dsc_uuid = dsc_development_uuid; 2593 } 2594 } 2595 if (!dsc_uuid.IsValid() && 2596 FileSystem::Instance().Exists(dsc_nondevelopment_filespec)) { 2597 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2598 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2599 if (dsc_nondevelopment_uuid.IsValid() && 2600 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2601 dsc_filespec = dsc_nondevelopment_filespec; 2602 dsc_uuid = dsc_nondevelopment_uuid; 2603 } 2604 } 2605 } 2606 2607 // Failing a UUID match, prefer the development dyld_shared cache if both 2608 // are present. 2609 if (!FileSystem::Instance().Exists(dsc_filespec)) { 2610 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2611 dsc_filespec = dsc_development_filespec; 2612 } else { 2613 dsc_filespec = dsc_nondevelopment_filespec; 2614 } 2615 } 2616 2617 /* The dyld_cache_header has a pointer to the 2618 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2619 The dyld_cache_local_symbols_info structure gives us three things: 2620 1. The start and count of the nlist records in the dyld_shared_cache 2621 file 2622 2. The start and size of the strings for these nlist records 2623 3. The start and count of dyld_cache_local_symbols_entry entries 2624 2625 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2626 dyld shared cache. 2627 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2628 the dyld shared cache. 2629 The dyld_cache_local_symbols_entry also lists the start of this 2630 dylib/framework's nlist records 2631 and the count of how many nlist records there are for this 2632 dylib/framework. 2633 */ 2634 2635 // Process the dyld shared cache header to find the unmapped symbols 2636 2637 DataBufferSP dsc_data_sp = MapFileData( 2638 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2639 if (!dsc_uuid.IsValid()) { 2640 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2641 } 2642 if (dsc_data_sp) { 2643 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2644 2645 bool uuid_match = true; 2646 if (dsc_uuid.IsValid() && process) { 2647 if (process_shared_cache_uuid.IsValid() && 2648 dsc_uuid != process_shared_cache_uuid) { 2649 // The on-disk dyld_shared_cache file is not the same as the one in 2650 // this process' memory, don't use it. 2651 uuid_match = false; 2652 ModuleSP module_sp(GetModule()); 2653 if (module_sp) 2654 module_sp->ReportWarning("process shared cache does not match " 2655 "on-disk dyld_shared_cache file, some " 2656 "symbol names will be missing."); 2657 } 2658 } 2659 2660 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2661 2662 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2663 2664 // If the mappingOffset points to a location inside the header, we've 2665 // opened an old dyld shared cache, and should not proceed further. 2666 if (uuid_match && 2667 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2668 2669 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2670 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2671 mappingOffset); 2672 2673 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2674 byte_order, addr_byte_size); 2675 offset = 0; 2676 2677 // The File addresses (from the in-memory Mach-O load commands) for 2678 // the shared libraries in the shared library cache need to be 2679 // adjusted by an offset to match up with the dylibOffset identifying 2680 // field in the dyld_cache_local_symbol_entry's. This offset is 2681 // recorded in mapping_offset_value. 2682 const uint64_t mapping_offset_value = 2683 dsc_mapping_info_data.GetU64(&offset); 2684 2685 offset = 2686 offsetof(struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2687 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2688 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2689 2690 if (localSymbolsOffset && localSymbolsSize) { 2691 // Map the local symbols 2692 DataBufferSP dsc_local_symbols_data_sp = 2693 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2694 2695 if (dsc_local_symbols_data_sp) { 2696 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2697 byte_order, addr_byte_size); 2698 2699 offset = 0; 2700 2701 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 2702 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 2703 UndefinedNameToDescMap undefined_name_to_desc; 2704 SymbolIndexToName reexport_shlib_needs_fixup; 2705 2706 // Read the local_symbols_infos struct in one shot 2707 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2708 dsc_local_symbols_data.GetU32(&offset, 2709 &local_symbols_info.nlistOffset, 6); 2710 2711 SectionSP text_section_sp( 2712 section_list->FindSectionByName(GetSegmentNameTEXT())); 2713 2714 uint32_t header_file_offset = 2715 (text_section_sp->GetFileAddress() - mapping_offset_value); 2716 2717 offset = local_symbols_info.entriesOffset; 2718 for (uint32_t entry_index = 0; 2719 entry_index < local_symbols_info.entriesCount; entry_index++) { 2720 struct lldb_copy_dyld_cache_local_symbols_entry 2721 local_symbols_entry; 2722 local_symbols_entry.dylibOffset = 2723 dsc_local_symbols_data.GetU32(&offset); 2724 local_symbols_entry.nlistStartIndex = 2725 dsc_local_symbols_data.GetU32(&offset); 2726 local_symbols_entry.nlistCount = 2727 dsc_local_symbols_data.GetU32(&offset); 2728 2729 if (header_file_offset == local_symbols_entry.dylibOffset) { 2730 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2731 2732 // The normal nlist code cannot correctly size the Symbols 2733 // array, we need to allocate it here. 2734 sym = symtab->Resize( 2735 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2736 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2737 num_syms = symtab->GetNumSymbols(); 2738 2739 nlist_data_offset = 2740 local_symbols_info.nlistOffset + 2741 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2742 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2743 2744 for (uint32_t nlist_index = 0; 2745 nlist_index < local_symbols_entry.nlistCount; 2746 nlist_index++) { 2747 ///////////////////////////// 2748 { 2749 llvm::Optional<struct nlist_64> nlist_maybe = 2750 ParseNList(dsc_local_symbols_data, nlist_data_offset, 2751 nlist_byte_size); 2752 if (!nlist_maybe) 2753 break; 2754 struct nlist_64 nlist = *nlist_maybe; 2755 2756 SymbolType type = eSymbolTypeInvalid; 2757 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2758 string_table_offset + nlist.n_strx); 2759 2760 if (symbol_name == NULL) { 2761 // No symbol should be NULL, even the symbols with no 2762 // string values should have an offset zero which 2763 // points to an empty C-string 2764 Host::SystemLog( 2765 Host::eSystemLogError, 2766 "error: DSC unmapped local symbol[%u] has invalid " 2767 "string table offset 0x%x in %s, ignoring symbol\n", 2768 entry_index, nlist.n_strx, 2769 module_sp->GetFileSpec().GetPath().c_str()); 2770 continue; 2771 } 2772 if (symbol_name[0] == '\0') 2773 symbol_name = NULL; 2774 2775 const char *symbol_name_non_abi_mangled = NULL; 2776 2777 SectionSP symbol_section; 2778 uint32_t symbol_byte_size = 0; 2779 bool add_nlist = true; 2780 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2781 bool demangled_is_synthesized = false; 2782 bool is_gsym = false; 2783 bool set_value = true; 2784 2785 assert(sym_idx < num_syms); 2786 2787 sym[sym_idx].SetDebug(is_debug); 2788 2789 if (is_debug) { 2790 switch (nlist.n_type) { 2791 case N_GSYM: 2792 // global symbol: name,,NO_SECT,type,0 2793 // Sometimes the N_GSYM value contains the address. 2794 2795 // FIXME: In the .o files, we have a GSYM and a debug 2796 // symbol for all the ObjC data. They 2797 // have the same address, but we want to ensure that 2798 // we always find only the real symbol, 'cause we 2799 // don't currently correctly attribute the 2800 // GSYM one to the ObjCClass/Ivar/MetaClass 2801 // symbol type. This is a temporary hack to make 2802 // sure the ObjectiveC symbols get treated correctly. 2803 // To do this right, we should coalesce all the GSYM 2804 // & global symbols that have the same address. 2805 2806 is_gsym = true; 2807 sym[sym_idx].SetExternal(true); 2808 2809 if (symbol_name && symbol_name[0] == '_' && 2810 symbol_name[1] == 'O') { 2811 llvm::StringRef symbol_name_ref(symbol_name); 2812 if (symbol_name_ref.startswith( 2813 g_objc_v2_prefix_class)) { 2814 symbol_name_non_abi_mangled = symbol_name + 1; 2815 symbol_name = 2816 symbol_name + g_objc_v2_prefix_class.size(); 2817 type = eSymbolTypeObjCClass; 2818 demangled_is_synthesized = true; 2819 2820 } else if (symbol_name_ref.startswith( 2821 g_objc_v2_prefix_metaclass)) { 2822 symbol_name_non_abi_mangled = symbol_name + 1; 2823 symbol_name = 2824 symbol_name + g_objc_v2_prefix_metaclass.size(); 2825 type = eSymbolTypeObjCMetaClass; 2826 demangled_is_synthesized = true; 2827 } else if (symbol_name_ref.startswith( 2828 g_objc_v2_prefix_ivar)) { 2829 symbol_name_non_abi_mangled = symbol_name + 1; 2830 symbol_name = 2831 symbol_name + g_objc_v2_prefix_ivar.size(); 2832 type = eSymbolTypeObjCIVar; 2833 demangled_is_synthesized = true; 2834 } 2835 } else { 2836 if (nlist.n_value != 0) 2837 symbol_section = section_info.GetSection( 2838 nlist.n_sect, nlist.n_value); 2839 type = eSymbolTypeData; 2840 } 2841 break; 2842 2843 case N_FNAME: 2844 // procedure name (f77 kludge): name,,NO_SECT,0,0 2845 type = eSymbolTypeCompiler; 2846 break; 2847 2848 case N_FUN: 2849 // procedure: name,,n_sect,linenumber,address 2850 if (symbol_name) { 2851 type = eSymbolTypeCode; 2852 symbol_section = section_info.GetSection( 2853 nlist.n_sect, nlist.n_value); 2854 2855 N_FUN_addr_to_sym_idx.insert( 2856 std::make_pair(nlist.n_value, sym_idx)); 2857 // We use the current number of symbols in the 2858 // symbol table in lieu of using nlist_idx in case 2859 // we ever start trimming entries out 2860 N_FUN_indexes.push_back(sym_idx); 2861 } else { 2862 type = eSymbolTypeCompiler; 2863 2864 if (!N_FUN_indexes.empty()) { 2865 // Copy the size of the function into the 2866 // original 2867 // STAB entry so we don't have 2868 // to hunt for it later 2869 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2870 ->SetByteSize(nlist.n_value); 2871 N_FUN_indexes.pop_back(); 2872 // We don't really need the end function STAB as 2873 // it contains the size which we already placed 2874 // with the original symbol, so don't add it if 2875 // we want a minimal symbol table 2876 add_nlist = false; 2877 } 2878 } 2879 break; 2880 2881 case N_STSYM: 2882 // static symbol: name,,n_sect,type,address 2883 N_STSYM_addr_to_sym_idx.insert( 2884 std::make_pair(nlist.n_value, sym_idx)); 2885 symbol_section = section_info.GetSection(nlist.n_sect, 2886 nlist.n_value); 2887 if (symbol_name && symbol_name[0]) { 2888 type = ObjectFile::GetSymbolTypeFromName( 2889 symbol_name + 1, eSymbolTypeData); 2890 } 2891 break; 2892 2893 case N_LCSYM: 2894 // .lcomm symbol: name,,n_sect,type,address 2895 symbol_section = section_info.GetSection(nlist.n_sect, 2896 nlist.n_value); 2897 type = eSymbolTypeCommonBlock; 2898 break; 2899 2900 case N_BNSYM: 2901 // We use the current number of symbols in the symbol 2902 // table in lieu of using nlist_idx in case we ever 2903 // start trimming entries out Skip these if we want 2904 // minimal symbol tables 2905 add_nlist = false; 2906 break; 2907 2908 case N_ENSYM: 2909 // Set the size of the N_BNSYM to the terminating 2910 // index of this N_ENSYM so that we can always skip 2911 // the entire symbol if we need to navigate more 2912 // quickly at the source level when parsing STABS 2913 // Skip these if we want minimal symbol tables 2914 add_nlist = false; 2915 break; 2916 2917 case N_OPT: 2918 // emitted with gcc2_compiled and in gcc source 2919 type = eSymbolTypeCompiler; 2920 break; 2921 2922 case N_RSYM: 2923 // register sym: name,,NO_SECT,type,register 2924 type = eSymbolTypeVariable; 2925 break; 2926 2927 case N_SLINE: 2928 // src line: 0,,n_sect,linenumber,address 2929 symbol_section = section_info.GetSection(nlist.n_sect, 2930 nlist.n_value); 2931 type = eSymbolTypeLineEntry; 2932 break; 2933 2934 case N_SSYM: 2935 // structure elt: name,,NO_SECT,type,struct_offset 2936 type = eSymbolTypeVariableType; 2937 break; 2938 2939 case N_SO: 2940 // source file name 2941 type = eSymbolTypeSourceFile; 2942 if (symbol_name == NULL) { 2943 add_nlist = false; 2944 if (N_SO_index != UINT32_MAX) { 2945 // Set the size of the N_SO to the terminating 2946 // index of this N_SO so that we can always skip 2947 // the entire N_SO if we need to navigate more 2948 // quickly at the source level when parsing STABS 2949 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2950 symbol_ptr->SetByteSize(sym_idx); 2951 symbol_ptr->SetSizeIsSibling(true); 2952 } 2953 N_NSYM_indexes.clear(); 2954 N_INCL_indexes.clear(); 2955 N_BRAC_indexes.clear(); 2956 N_COMM_indexes.clear(); 2957 N_FUN_indexes.clear(); 2958 N_SO_index = UINT32_MAX; 2959 } else { 2960 // We use the current number of symbols in the 2961 // symbol table in lieu of using nlist_idx in case 2962 // we ever start trimming entries out 2963 const bool N_SO_has_full_path = symbol_name[0] == '/'; 2964 if (N_SO_has_full_path) { 2965 if ((N_SO_index == sym_idx - 1) && 2966 ((sym_idx - 1) < num_syms)) { 2967 // We have two consecutive N_SO entries where 2968 // the first contains a directory and the 2969 // second contains a full path. 2970 sym[sym_idx - 1].GetMangled().SetValue( 2971 ConstString(symbol_name), false); 2972 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2973 add_nlist = false; 2974 } else { 2975 // This is the first entry in a N_SO that 2976 // contains a directory or 2977 // a full path to the source file 2978 N_SO_index = sym_idx; 2979 } 2980 } else if ((N_SO_index == sym_idx - 1) && 2981 ((sym_idx - 1) < num_syms)) { 2982 // This is usually the second N_SO entry that 2983 // contains just the filename, so here we combine 2984 // it with the first one if we are minimizing the 2985 // symbol table 2986 const char *so_path = 2987 sym[sym_idx - 1] 2988 .GetMangled() 2989 .GetDemangledName( 2990 lldb::eLanguageTypeUnknown) 2991 .AsCString(); 2992 if (so_path && so_path[0]) { 2993 std::string full_so_path(so_path); 2994 const size_t double_slash_pos = 2995 full_so_path.find("//"); 2996 if (double_slash_pos != std::string::npos) { 2997 // The linker has been generating bad N_SO 2998 // entries with doubled up paths 2999 // in the format "%s%s" where the first 3000 // string in the DW_AT_comp_dir, and the 3001 // second is the directory for the source 3002 // file so you end up with a path that looks 3003 // like "/tmp/src//tmp/src/" 3004 FileSpec so_dir(so_path); 3005 if (!FileSystem::Instance().Exists(so_dir)) { 3006 so_dir.SetFile( 3007 &full_so_path[double_slash_pos + 1], 3008 FileSpec::Style::native); 3009 if (FileSystem::Instance().Exists(so_dir)) { 3010 // Trim off the incorrect path 3011 full_so_path.erase(0, double_slash_pos + 1); 3012 } 3013 } 3014 } 3015 if (*full_so_path.rbegin() != '/') 3016 full_so_path += '/'; 3017 full_so_path += symbol_name; 3018 sym[sym_idx - 1].GetMangled().SetValue( 3019 ConstString(full_so_path.c_str()), false); 3020 add_nlist = false; 3021 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3022 } 3023 } else { 3024 // This could be a relative path to a N_SO 3025 N_SO_index = sym_idx; 3026 } 3027 } 3028 break; 3029 3030 case N_OSO: 3031 // object file name: name,,0,0,st_mtime 3032 type = eSymbolTypeObjectFile; 3033 break; 3034 3035 case N_LSYM: 3036 // local sym: name,,NO_SECT,type,offset 3037 type = eSymbolTypeLocal; 3038 break; 3039 3040 // INCL scopes 3041 case N_BINCL: 3042 // include file beginning: name,,NO_SECT,0,sum We use 3043 // the current number of symbols in the symbol table 3044 // in lieu of using nlist_idx in case we ever start 3045 // trimming entries out 3046 N_INCL_indexes.push_back(sym_idx); 3047 type = eSymbolTypeScopeBegin; 3048 break; 3049 3050 case N_EINCL: 3051 // include file end: name,,NO_SECT,0,0 3052 // Set the size of the N_BINCL to the terminating 3053 // index of this N_EINCL so that we can always skip 3054 // the entire symbol if we need to navigate more 3055 // quickly at the source level when parsing STABS 3056 if (!N_INCL_indexes.empty()) { 3057 symbol_ptr = 3058 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3059 symbol_ptr->SetByteSize(sym_idx + 1); 3060 symbol_ptr->SetSizeIsSibling(true); 3061 N_INCL_indexes.pop_back(); 3062 } 3063 type = eSymbolTypeScopeEnd; 3064 break; 3065 3066 case N_SOL: 3067 // #included file name: name,,n_sect,0,address 3068 type = eSymbolTypeHeaderFile; 3069 3070 // We currently don't use the header files on darwin 3071 add_nlist = false; 3072 break; 3073 3074 case N_PARAMS: 3075 // compiler parameters: name,,NO_SECT,0,0 3076 type = eSymbolTypeCompiler; 3077 break; 3078 3079 case N_VERSION: 3080 // compiler version: name,,NO_SECT,0,0 3081 type = eSymbolTypeCompiler; 3082 break; 3083 3084 case N_OLEVEL: 3085 // compiler -O level: name,,NO_SECT,0,0 3086 type = eSymbolTypeCompiler; 3087 break; 3088 3089 case N_PSYM: 3090 // parameter: name,,NO_SECT,type,offset 3091 type = eSymbolTypeVariable; 3092 break; 3093 3094 case N_ENTRY: 3095 // alternate entry: name,,n_sect,linenumber,address 3096 symbol_section = section_info.GetSection(nlist.n_sect, 3097 nlist.n_value); 3098 type = eSymbolTypeLineEntry; 3099 break; 3100 3101 // Left and Right Braces 3102 case N_LBRAC: 3103 // left bracket: 0,,NO_SECT,nesting level,address We 3104 // use the current number of symbols in the symbol 3105 // table in lieu of using nlist_idx in case we ever 3106 // start trimming entries out 3107 symbol_section = section_info.GetSection(nlist.n_sect, 3108 nlist.n_value); 3109 N_BRAC_indexes.push_back(sym_idx); 3110 type = eSymbolTypeScopeBegin; 3111 break; 3112 3113 case N_RBRAC: 3114 // right bracket: 0,,NO_SECT,nesting level,address 3115 // Set the size of the N_LBRAC to the terminating 3116 // index of this N_RBRAC so that we can always skip 3117 // the entire symbol if we need to navigate more 3118 // quickly at the source level when parsing STABS 3119 symbol_section = section_info.GetSection(nlist.n_sect, 3120 nlist.n_value); 3121 if (!N_BRAC_indexes.empty()) { 3122 symbol_ptr = 3123 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3124 symbol_ptr->SetByteSize(sym_idx + 1); 3125 symbol_ptr->SetSizeIsSibling(true); 3126 N_BRAC_indexes.pop_back(); 3127 } 3128 type = eSymbolTypeScopeEnd; 3129 break; 3130 3131 case N_EXCL: 3132 // deleted include file: name,,NO_SECT,0,sum 3133 type = eSymbolTypeHeaderFile; 3134 break; 3135 3136 // COMM scopes 3137 case N_BCOMM: 3138 // begin common: name,,NO_SECT,0,0 3139 // We use the current number of symbols in the symbol 3140 // table in lieu of using nlist_idx in case we ever 3141 // start trimming entries out 3142 type = eSymbolTypeScopeBegin; 3143 N_COMM_indexes.push_back(sym_idx); 3144 break; 3145 3146 case N_ECOML: 3147 // end common (local name): 0,,n_sect,0,address 3148 symbol_section = section_info.GetSection(nlist.n_sect, 3149 nlist.n_value); 3150 // Fall through 3151 3152 case N_ECOMM: 3153 // end common: name,,n_sect,0,0 3154 // Set the size of the N_BCOMM to the terminating 3155 // index of this N_ECOMM/N_ECOML so that we can 3156 // always skip the entire symbol if we need to 3157 // navigate more quickly at the source level when 3158 // parsing STABS 3159 if (!N_COMM_indexes.empty()) { 3160 symbol_ptr = 3161 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3162 symbol_ptr->SetByteSize(sym_idx + 1); 3163 symbol_ptr->SetSizeIsSibling(true); 3164 N_COMM_indexes.pop_back(); 3165 } 3166 type = eSymbolTypeScopeEnd; 3167 break; 3168 3169 case N_LENG: 3170 // second stab entry with length information 3171 type = eSymbolTypeAdditional; 3172 break; 3173 3174 default: 3175 break; 3176 } 3177 } else { 3178 // uint8_t n_pext = N_PEXT & nlist.n_type; 3179 uint8_t n_type = N_TYPE & nlist.n_type; 3180 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3181 3182 switch (n_type) { 3183 case N_INDR: { 3184 const char *reexport_name_cstr = 3185 strtab_data.PeekCStr(nlist.n_value); 3186 if (reexport_name_cstr && reexport_name_cstr[0]) { 3187 type = eSymbolTypeReExported; 3188 ConstString reexport_name( 3189 reexport_name_cstr + 3190 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3191 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3192 set_value = false; 3193 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3194 indirect_symbol_names.insert(ConstString( 3195 symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 3196 } else 3197 type = eSymbolTypeUndefined; 3198 } break; 3199 3200 case N_UNDF: 3201 if (symbol_name && symbol_name[0]) { 3202 ConstString undefined_name( 3203 symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); 3204 undefined_name_to_desc[undefined_name] = nlist.n_desc; 3205 } 3206 // Fall through 3207 case N_PBUD: 3208 type = eSymbolTypeUndefined; 3209 break; 3210 3211 case N_ABS: 3212 type = eSymbolTypeAbsolute; 3213 break; 3214 3215 case N_SECT: { 3216 symbol_section = section_info.GetSection(nlist.n_sect, 3217 nlist.n_value); 3218 3219 if (symbol_section == NULL) { 3220 // TODO: warn about this? 3221 add_nlist = false; 3222 break; 3223 } 3224 3225 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3226 type = eSymbolTypeException; 3227 } else { 3228 uint32_t section_type = 3229 symbol_section->Get() & SECTION_TYPE; 3230 3231 switch (section_type) { 3232 case S_CSTRING_LITERALS: 3233 type = eSymbolTypeData; 3234 break; // section with only literal C strings 3235 case S_4BYTE_LITERALS: 3236 type = eSymbolTypeData; 3237 break; // section with only 4 byte literals 3238 case S_8BYTE_LITERALS: 3239 type = eSymbolTypeData; 3240 break; // section with only 8 byte literals 3241 case S_LITERAL_POINTERS: 3242 type = eSymbolTypeTrampoline; 3243 break; // section with only pointers to literals 3244 case S_NON_LAZY_SYMBOL_POINTERS: 3245 type = eSymbolTypeTrampoline; 3246 break; // section with only non-lazy symbol 3247 // pointers 3248 case S_LAZY_SYMBOL_POINTERS: 3249 type = eSymbolTypeTrampoline; 3250 break; // section with only lazy symbol pointers 3251 case S_SYMBOL_STUBS: 3252 type = eSymbolTypeTrampoline; 3253 break; // section with only symbol stubs, byte 3254 // size of stub in the reserved2 field 3255 case S_MOD_INIT_FUNC_POINTERS: 3256 type = eSymbolTypeCode; 3257 break; // section with only function pointers for 3258 // initialization 3259 case S_MOD_TERM_FUNC_POINTERS: 3260 type = eSymbolTypeCode; 3261 break; // section with only function pointers for 3262 // termination 3263 case S_INTERPOSING: 3264 type = eSymbolTypeTrampoline; 3265 break; // section with only pairs of function 3266 // pointers for interposing 3267 case S_16BYTE_LITERALS: 3268 type = eSymbolTypeData; 3269 break; // section with only 16 byte literals 3270 case S_DTRACE_DOF: 3271 type = eSymbolTypeInstrumentation; 3272 break; 3273 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3274 type = eSymbolTypeTrampoline; 3275 break; 3276 default: 3277 switch (symbol_section->GetType()) { 3278 case lldb::eSectionTypeCode: 3279 type = eSymbolTypeCode; 3280 break; 3281 case eSectionTypeData: 3282 case eSectionTypeDataCString: // Inlined C string 3283 // data 3284 case eSectionTypeDataCStringPointers: // Pointers 3285 // to C 3286 // string 3287 // data 3288 case eSectionTypeDataSymbolAddress: // Address of 3289 // a symbol in 3290 // the symbol 3291 // table 3292 case eSectionTypeData4: 3293 case eSectionTypeData8: 3294 case eSectionTypeData16: 3295 type = eSymbolTypeData; 3296 break; 3297 default: 3298 break; 3299 } 3300 break; 3301 } 3302 3303 if (type == eSymbolTypeInvalid) { 3304 const char *symbol_sect_name = 3305 symbol_section->GetName().AsCString(); 3306 if (symbol_section->IsDescendant( 3307 text_section_sp.get())) { 3308 if (symbol_section->IsClear( 3309 S_ATTR_PURE_INSTRUCTIONS | 3310 S_ATTR_SELF_MODIFYING_CODE | 3311 S_ATTR_SOME_INSTRUCTIONS)) 3312 type = eSymbolTypeData; 3313 else 3314 type = eSymbolTypeCode; 3315 } else if (symbol_section->IsDescendant( 3316 data_section_sp.get()) || 3317 symbol_section->IsDescendant( 3318 data_dirty_section_sp.get()) || 3319 symbol_section->IsDescendant( 3320 data_const_section_sp.get())) { 3321 if (symbol_sect_name && 3322 ::strstr(symbol_sect_name, "__objc") == 3323 symbol_sect_name) { 3324 type = eSymbolTypeRuntime; 3325 3326 if (symbol_name) { 3327 llvm::StringRef symbol_name_ref(symbol_name); 3328 if (symbol_name_ref.startswith("_OBJC_")) { 3329 llvm::StringRef 3330 g_objc_v2_prefix_class( 3331 "_OBJC_CLASS_$_"); 3332 llvm::StringRef 3333 g_objc_v2_prefix_metaclass( 3334 "_OBJC_METACLASS_$_"); 3335 llvm::StringRef 3336 g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 3337 if (symbol_name_ref.startswith( 3338 g_objc_v2_prefix_class)) { 3339 symbol_name_non_abi_mangled = 3340 symbol_name + 1; 3341 symbol_name = 3342 symbol_name + 3343 g_objc_v2_prefix_class.size(); 3344 type = eSymbolTypeObjCClass; 3345 demangled_is_synthesized = true; 3346 } else if ( 3347 symbol_name_ref.startswith( 3348 g_objc_v2_prefix_metaclass)) { 3349 symbol_name_non_abi_mangled = 3350 symbol_name + 1; 3351 symbol_name = 3352 symbol_name + 3353 g_objc_v2_prefix_metaclass.size(); 3354 type = eSymbolTypeObjCMetaClass; 3355 demangled_is_synthesized = true; 3356 } else if (symbol_name_ref.startswith( 3357 g_objc_v2_prefix_ivar)) { 3358 symbol_name_non_abi_mangled = 3359 symbol_name + 1; 3360 symbol_name = 3361 symbol_name + 3362 g_objc_v2_prefix_ivar.size(); 3363 type = eSymbolTypeObjCIVar; 3364 demangled_is_synthesized = true; 3365 } 3366 } 3367 } 3368 } else if (symbol_sect_name && 3369 ::strstr(symbol_sect_name, 3370 "__gcc_except_tab") == 3371 symbol_sect_name) { 3372 type = eSymbolTypeException; 3373 } else { 3374 type = eSymbolTypeData; 3375 } 3376 } else if (symbol_sect_name && 3377 ::strstr(symbol_sect_name, "__IMPORT") == 3378 symbol_sect_name) { 3379 type = eSymbolTypeTrampoline; 3380 } else if (symbol_section->IsDescendant( 3381 objc_section_sp.get())) { 3382 type = eSymbolTypeRuntime; 3383 if (symbol_name && symbol_name[0] == '.') { 3384 llvm::StringRef symbol_name_ref(symbol_name); 3385 llvm::StringRef 3386 g_objc_v1_prefix_class(".objc_class_name_"); 3387 if (symbol_name_ref.startswith( 3388 g_objc_v1_prefix_class)) { 3389 symbol_name_non_abi_mangled = symbol_name; 3390 symbol_name = symbol_name + 3391 g_objc_v1_prefix_class.size(); 3392 type = eSymbolTypeObjCClass; 3393 demangled_is_synthesized = true; 3394 } 3395 } 3396 } 3397 } 3398 } 3399 } break; 3400 } 3401 } 3402 3403 if (add_nlist) { 3404 uint64_t symbol_value = nlist.n_value; 3405 if (symbol_name_non_abi_mangled) { 3406 sym[sym_idx].GetMangled().SetMangledName( 3407 ConstString(symbol_name_non_abi_mangled)); 3408 sym[sym_idx].GetMangled().SetDemangledName( 3409 ConstString(symbol_name)); 3410 } else { 3411 bool symbol_name_is_mangled = false; 3412 3413 if (symbol_name && symbol_name[0] == '_') { 3414 symbol_name_is_mangled = symbol_name[1] == '_'; 3415 symbol_name++; // Skip the leading underscore 3416 } 3417 3418 if (symbol_name) { 3419 ConstString const_symbol_name(symbol_name); 3420 sym[sym_idx].GetMangled().SetValue( 3421 const_symbol_name, symbol_name_is_mangled); 3422 if (is_gsym && is_debug) { 3423 const char *gsym_name = 3424 sym[sym_idx] 3425 .GetMangled() 3426 .GetName(lldb::eLanguageTypeUnknown, 3427 Mangled::ePreferMangled) 3428 .GetCString(); 3429 if (gsym_name) 3430 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3431 } 3432 } 3433 } 3434 if (symbol_section) { 3435 const addr_t section_file_addr = 3436 symbol_section->GetFileAddress(); 3437 if (symbol_byte_size == 0 && 3438 function_starts_count > 0) { 3439 addr_t symbol_lookup_file_addr = nlist.n_value; 3440 // Do an exact address match for non-ARM addresses, 3441 // else get the closest since the symbol might be a 3442 // thumb symbol which has an address with bit zero 3443 // set 3444 FunctionStarts::Entry *func_start_entry = 3445 function_starts.FindEntry(symbol_lookup_file_addr, 3446 !is_arm); 3447 if (is_arm && func_start_entry) { 3448 // Verify that the function start address is the 3449 // symbol address (ARM) or the symbol address + 1 3450 // (thumb) 3451 if (func_start_entry->addr != 3452 symbol_lookup_file_addr && 3453 func_start_entry->addr != 3454 (symbol_lookup_file_addr + 1)) { 3455 // Not the right entry, NULL it out... 3456 func_start_entry = NULL; 3457 } 3458 } 3459 if (func_start_entry) { 3460 func_start_entry->data = true; 3461 3462 addr_t symbol_file_addr = func_start_entry->addr; 3463 uint32_t symbol_flags = 0; 3464 if (is_arm) { 3465 if (symbol_file_addr & 1) 3466 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3467 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3468 } 3469 3470 const FunctionStarts::Entry *next_func_start_entry = 3471 function_starts.FindNextEntry(func_start_entry); 3472 const addr_t section_end_file_addr = 3473 section_file_addr + 3474 symbol_section->GetByteSize(); 3475 if (next_func_start_entry) { 3476 addr_t next_symbol_file_addr = 3477 next_func_start_entry->addr; 3478 // Be sure the clear the Thumb address bit when 3479 // we calculate the size from the current and 3480 // next address 3481 if (is_arm) 3482 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3483 symbol_byte_size = std::min<lldb::addr_t>( 3484 next_symbol_file_addr - symbol_file_addr, 3485 section_end_file_addr - symbol_file_addr); 3486 } else { 3487 symbol_byte_size = 3488 section_end_file_addr - symbol_file_addr; 3489 } 3490 } 3491 } 3492 symbol_value -= section_file_addr; 3493 } 3494 3495 if (is_debug == false) { 3496 if (type == eSymbolTypeCode) { 3497 // See if we can find a N_FUN entry for any code 3498 // symbols. If we do find a match, and the name 3499 // matches, then we can merge the two into just the 3500 // function symbol to avoid duplicate entries in 3501 // the symbol table 3502 auto range = 3503 N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3504 if (range.first != range.second) { 3505 bool found_it = false; 3506 for (const auto pos = range.first; 3507 pos != range.second; ++pos) { 3508 if (sym[sym_idx].GetMangled().GetName( 3509 lldb::eLanguageTypeUnknown, 3510 Mangled::ePreferMangled) == 3511 sym[pos->second].GetMangled().GetName( 3512 lldb::eLanguageTypeUnknown, 3513 Mangled::ePreferMangled)) { 3514 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3515 // We just need the flags from the linker 3516 // symbol, so put these flags 3517 // into the N_FUN flags to avoid duplicate 3518 // symbols in the symbol table 3519 sym[pos->second].SetExternal( 3520 sym[sym_idx].IsExternal()); 3521 sym[pos->second].SetFlags(nlist.n_type << 16 | 3522 nlist.n_desc); 3523 if (resolver_addresses.find(nlist.n_value) != 3524 resolver_addresses.end()) 3525 sym[pos->second].SetType(eSymbolTypeResolver); 3526 sym[sym_idx].Clear(); 3527 found_it = true; 3528 break; 3529 } 3530 } 3531 if (found_it) 3532 continue; 3533 } else { 3534 if (resolver_addresses.find(nlist.n_value) != 3535 resolver_addresses.end()) 3536 type = eSymbolTypeResolver; 3537 } 3538 } else if (type == eSymbolTypeData || 3539 type == eSymbolTypeObjCClass || 3540 type == eSymbolTypeObjCMetaClass || 3541 type == eSymbolTypeObjCIVar) { 3542 // See if we can find a N_STSYM entry for any data 3543 // symbols. If we do find a match, and the name 3544 // matches, then we can merge the two into just the 3545 // Static symbol to avoid duplicate entries in the 3546 // symbol table 3547 auto range = N_STSYM_addr_to_sym_idx.equal_range( 3548 nlist.n_value); 3549 if (range.first != range.second) { 3550 bool found_it = false; 3551 for (const auto pos = range.first; 3552 pos != range.second; ++pos) { 3553 if (sym[sym_idx].GetMangled().GetName( 3554 lldb::eLanguageTypeUnknown, 3555 Mangled::ePreferMangled) == 3556 sym[pos->second].GetMangled().GetName( 3557 lldb::eLanguageTypeUnknown, 3558 Mangled::ePreferMangled)) { 3559 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3560 // We just need the flags from the linker 3561 // symbol, so put these flags 3562 // into the N_STSYM flags to avoid duplicate 3563 // symbols in the symbol table 3564 sym[pos->second].SetExternal( 3565 sym[sym_idx].IsExternal()); 3566 sym[pos->second].SetFlags(nlist.n_type << 16 | 3567 nlist.n_desc); 3568 sym[sym_idx].Clear(); 3569 found_it = true; 3570 break; 3571 } 3572 } 3573 if (found_it) 3574 continue; 3575 } else { 3576 const char *gsym_name = 3577 sym[sym_idx] 3578 .GetMangled() 3579 .GetName(lldb::eLanguageTypeUnknown, 3580 Mangled::ePreferMangled) 3581 .GetCString(); 3582 if (gsym_name) { 3583 // Combine N_GSYM stab entries with the non 3584 // stab symbol 3585 ConstNameToSymbolIndexMap::const_iterator pos = 3586 N_GSYM_name_to_sym_idx.find(gsym_name); 3587 if (pos != N_GSYM_name_to_sym_idx.end()) { 3588 const uint32_t GSYM_sym_idx = pos->second; 3589 m_nlist_idx_to_sym_idx[nlist_idx] = 3590 GSYM_sym_idx; 3591 // Copy the address, because often the N_GSYM 3592 // address has an invalid address of zero 3593 // when the global is a common symbol 3594 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3595 symbol_section); 3596 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3597 symbol_value); 3598 // We just need the flags from the linker 3599 // symbol, so put these flags 3600 // into the N_GSYM flags to avoid duplicate 3601 // symbols in the symbol table 3602 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 3603 nlist.n_desc); 3604 sym[sym_idx].Clear(); 3605 continue; 3606 } 3607 } 3608 } 3609 } 3610 } 3611 3612 sym[sym_idx].SetID(nlist_idx); 3613 sym[sym_idx].SetType(type); 3614 if (set_value) { 3615 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 3616 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3617 } 3618 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 3619 3620 if (symbol_byte_size > 0) 3621 sym[sym_idx].SetByteSize(symbol_byte_size); 3622 3623 if (demangled_is_synthesized) 3624 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3625 ++sym_idx; 3626 } else { 3627 sym[sym_idx].Clear(); 3628 } 3629 } 3630 ///////////////////////////// 3631 } 3632 break; // No more entries to consider 3633 } 3634 } 3635 3636 for (const auto &pos : reexport_shlib_needs_fixup) { 3637 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3638 if (undef_pos != undefined_name_to_desc.end()) { 3639 const uint8_t dylib_ordinal = 3640 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3641 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 3642 sym[pos.first].SetReExportedSymbolSharedLibrary( 3643 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3644 } 3645 } 3646 } 3647 } 3648 } 3649 } 3650 } 3651 3652 // Must reset this in case it was mutated above! 3653 nlist_data_offset = 0; 3654 #endif 3655 3656 if (nlist_data.GetByteSize() > 0) { 3657 3658 // If the sym array was not created while parsing the DSC unmapped 3659 // symbols, create it now. 3660 if (sym == nullptr) { 3661 sym = 3662 symtab->Resize(symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3663 num_syms = symtab->GetNumSymbols(); 3664 } 3665 3666 if (unmapped_local_symbols_found) { 3667 assert(m_dysymtab.ilocalsym == 0); 3668 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3669 nlist_idx = m_dysymtab.nlocalsym; 3670 } else { 3671 nlist_idx = 0; 3672 } 3673 3674 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; 3675 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; 3676 UndefinedNameToDescMap undefined_name_to_desc; 3677 SymbolIndexToName reexport_shlib_needs_fixup; 3678 3679 // Symtab parsing is a huge mess. Everything is entangled and the code 3680 // requires access to a ridiculous amount of variables. LLDB depends 3681 // heavily on the proper merging of symbols and to get that right we need 3682 // to make sure we have parsed all the debug symbols first. Therefore we 3683 // invoke the lambda twice, once to parse only the debug symbols and then 3684 // once more to parse the remaining symbols. 3685 auto ParseSymbolLambda = [&](struct nlist_64 &nlist, uint32_t nlist_idx, 3686 bool debug_only) { 3687 const bool is_debug = ((nlist.n_type & N_STAB) != 0); 3688 if (is_debug != debug_only) 3689 return true; 3690 3691 const char *symbol_name_non_abi_mangled = nullptr; 3692 const char *symbol_name = nullptr; 3693 3694 if (have_strtab_data) { 3695 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3696 3697 if (symbol_name == nullptr) { 3698 // No symbol should be NULL, even the symbols with no string values 3699 // should have an offset zero which points to an empty C-string 3700 Host::SystemLog(Host::eSystemLogError, 3701 "error: symbol[%u] has invalid string table offset " 3702 "0x%x in %s, ignoring symbol\n", 3703 nlist_idx, nlist.n_strx, 3704 module_sp->GetFileSpec().GetPath().c_str()); 3705 return true; 3706 } 3707 if (symbol_name[0] == '\0') 3708 symbol_name = nullptr; 3709 } else { 3710 const addr_t str_addr = strtab_addr + nlist.n_strx; 3711 Status str_error; 3712 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3713 str_error)) 3714 symbol_name = memory_symbol_name.c_str(); 3715 } 3716 3717 SymbolType type = eSymbolTypeInvalid; 3718 SectionSP symbol_section; 3719 lldb::addr_t symbol_byte_size = 0; 3720 bool add_nlist = true; 3721 bool is_gsym = false; 3722 bool demangled_is_synthesized = false; 3723 bool set_value = true; 3724 3725 assert(sym_idx < num_syms); 3726 sym[sym_idx].SetDebug(is_debug); 3727 3728 if (is_debug) { 3729 switch (nlist.n_type) { 3730 case N_GSYM: 3731 // global symbol: name,,NO_SECT,type,0 3732 // Sometimes the N_GSYM value contains the address. 3733 3734 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3735 // the ObjC data. They 3736 // have the same address, but we want to ensure that we always find 3737 // only the real symbol, 'cause we don't currently correctly 3738 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3739 // type. This is a temporary hack to make sure the ObjectiveC 3740 // symbols get treated correctly. To do this right, we should 3741 // coalesce all the GSYM & global symbols that have the same 3742 // address. 3743 is_gsym = true; 3744 sym[sym_idx].SetExternal(true); 3745 3746 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3747 llvm::StringRef symbol_name_ref(symbol_name); 3748 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3749 symbol_name_non_abi_mangled = symbol_name + 1; 3750 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3751 type = eSymbolTypeObjCClass; 3752 demangled_is_synthesized = true; 3753 3754 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) { 3755 symbol_name_non_abi_mangled = symbol_name + 1; 3756 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3757 type = eSymbolTypeObjCMetaClass; 3758 demangled_is_synthesized = true; 3759 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3760 symbol_name_non_abi_mangled = symbol_name + 1; 3761 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3762 type = eSymbolTypeObjCIVar; 3763 demangled_is_synthesized = true; 3764 } 3765 } else { 3766 if (nlist.n_value != 0) 3767 symbol_section = 3768 section_info.GetSection(nlist.n_sect, nlist.n_value); 3769 type = eSymbolTypeData; 3770 } 3771 break; 3772 3773 case N_FNAME: 3774 // procedure name (f77 kludge): name,,NO_SECT,0,0 3775 type = eSymbolTypeCompiler; 3776 break; 3777 3778 case N_FUN: 3779 // procedure: name,,n_sect,linenumber,address 3780 if (symbol_name) { 3781 type = eSymbolTypeCode; 3782 symbol_section = 3783 section_info.GetSection(nlist.n_sect, nlist.n_value); 3784 3785 N_FUN_addr_to_sym_idx.insert( 3786 std::make_pair(nlist.n_value, sym_idx)); 3787 // We use the current number of symbols in the symbol table in 3788 // lieu of using nlist_idx in case we ever start trimming entries 3789 // out 3790 N_FUN_indexes.push_back(sym_idx); 3791 } else { 3792 type = eSymbolTypeCompiler; 3793 3794 if (!N_FUN_indexes.empty()) { 3795 // Copy the size of the function into the original STAB entry 3796 // so we don't have to hunt for it later 3797 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3798 ->SetByteSize(nlist.n_value); 3799 N_FUN_indexes.pop_back(); 3800 // We don't really need the end function STAB as it contains 3801 // the size which we already placed with the original symbol, 3802 // so don't add it if we want a minimal symbol table 3803 add_nlist = false; 3804 } 3805 } 3806 break; 3807 3808 case N_STSYM: 3809 // static symbol: name,,n_sect,type,address 3810 N_STSYM_addr_to_sym_idx.insert( 3811 std::make_pair(nlist.n_value, sym_idx)); 3812 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3813 if (symbol_name && symbol_name[0]) { 3814 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3815 eSymbolTypeData); 3816 } 3817 break; 3818 3819 case N_LCSYM: 3820 // .lcomm symbol: name,,n_sect,type,address 3821 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3822 type = eSymbolTypeCommonBlock; 3823 break; 3824 3825 case N_BNSYM: 3826 // We use the current number of symbols in the symbol table in lieu 3827 // of using nlist_idx in case we ever start trimming entries out 3828 // Skip these if we want minimal symbol tables 3829 add_nlist = false; 3830 break; 3831 3832 case N_ENSYM: 3833 // Set the size of the N_BNSYM to the terminating index of this 3834 // N_ENSYM so that we can always skip the entire symbol if we need 3835 // to navigate more quickly at the source level when parsing STABS 3836 // Skip these if we want minimal symbol tables 3837 add_nlist = false; 3838 break; 3839 3840 case N_OPT: 3841 // emitted with gcc2_compiled and in gcc source 3842 type = eSymbolTypeCompiler; 3843 break; 3844 3845 case N_RSYM: 3846 // register sym: name,,NO_SECT,type,register 3847 type = eSymbolTypeVariable; 3848 break; 3849 3850 case N_SLINE: 3851 // src line: 0,,n_sect,linenumber,address 3852 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 3853 type = eSymbolTypeLineEntry; 3854 break; 3855 3856 case N_SSYM: 3857 // structure elt: name,,NO_SECT,type,struct_offset 3858 type = eSymbolTypeVariableType; 3859 break; 3860 3861 case N_SO: 3862 // source file name 3863 type = eSymbolTypeSourceFile; 3864 if (symbol_name == nullptr) { 3865 add_nlist = false; 3866 if (N_SO_index != UINT32_MAX) { 3867 // Set the size of the N_SO to the terminating index of this 3868 // N_SO so that we can always skip the entire N_SO if we need 3869 // to navigate more quickly at the source level when parsing 3870 // STABS 3871 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3872 symbol_ptr->SetByteSize(sym_idx); 3873 symbol_ptr->SetSizeIsSibling(true); 3874 } 3875 N_NSYM_indexes.clear(); 3876 N_INCL_indexes.clear(); 3877 N_BRAC_indexes.clear(); 3878 N_COMM_indexes.clear(); 3879 N_FUN_indexes.clear(); 3880 N_SO_index = UINT32_MAX; 3881 } else { 3882 // We use the current number of symbols in the symbol table in 3883 // lieu of using nlist_idx in case we ever start trimming entries 3884 // out 3885 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3886 if (N_SO_has_full_path) { 3887 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3888 // We have two consecutive N_SO entries where the first 3889 // contains a directory and the second contains a full path. 3890 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), 3891 false); 3892 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3893 add_nlist = false; 3894 } else { 3895 // This is the first entry in a N_SO that contains a 3896 // directory or a full path to the source file 3897 N_SO_index = sym_idx; 3898 } 3899 } else if ((N_SO_index == sym_idx - 1) && 3900 ((sym_idx - 1) < num_syms)) { 3901 // This is usually the second N_SO entry that contains just the 3902 // filename, so here we combine it with the first one if we are 3903 // minimizing the symbol table 3904 const char *so_path = 3905 sym[sym_idx - 1] 3906 .GetMangled() 3907 .GetDemangledName(lldb::eLanguageTypeUnknown) 3908 .AsCString(); 3909 if (so_path && so_path[0]) { 3910 std::string full_so_path(so_path); 3911 const size_t double_slash_pos = full_so_path.find("//"); 3912 if (double_slash_pos != std::string::npos) { 3913 // The linker has been generating bad N_SO entries with 3914 // doubled up paths in the format "%s%s" where the first 3915 // string in the DW_AT_comp_dir, and the second is the 3916 // directory for the source file so you end up with a path 3917 // that looks like "/tmp/src//tmp/src/" 3918 FileSpec so_dir(so_path); 3919 if (!FileSystem::Instance().Exists(so_dir)) { 3920 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 3921 FileSpec::Style::native); 3922 if (FileSystem::Instance().Exists(so_dir)) { 3923 // Trim off the incorrect path 3924 full_so_path.erase(0, double_slash_pos + 1); 3925 } 3926 } 3927 } 3928 if (*full_so_path.rbegin() != '/') 3929 full_so_path += '/'; 3930 full_so_path += symbol_name; 3931 sym[sym_idx - 1].GetMangled().SetValue( 3932 ConstString(full_so_path.c_str()), false); 3933 add_nlist = false; 3934 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3935 } 3936 } else { 3937 // This could be a relative path to a N_SO 3938 N_SO_index = sym_idx; 3939 } 3940 } 3941 break; 3942 3943 case N_OSO: 3944 // object file name: name,,0,0,st_mtime 3945 type = eSymbolTypeObjectFile; 3946 break; 3947 3948 case N_LSYM: 3949 // local sym: name,,NO_SECT,type,offset 3950 type = eSymbolTypeLocal; 3951 break; 3952 3953 // INCL scopes 3954 case N_BINCL: 3955 // include file beginning: name,,NO_SECT,0,sum We use the current 3956 // number of symbols in the symbol table in lieu of using nlist_idx 3957 // in case we ever start trimming entries out 3958 N_INCL_indexes.push_back(sym_idx); 3959 type = eSymbolTypeScopeBegin; 3960 break; 3961 3962 case N_EINCL: 3963 // include file end: name,,NO_SECT,0,0 3964 // Set the size of the N_BINCL to the terminating index of this 3965 // N_EINCL so that we can always skip the entire symbol if we need 3966 // to navigate more quickly at the source level when parsing STABS 3967 if (!N_INCL_indexes.empty()) { 3968 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3969 symbol_ptr->SetByteSize(sym_idx + 1); 3970 symbol_ptr->SetSizeIsSibling(true); 3971 N_INCL_indexes.pop_back(); 3972 } 3973 type = eSymbolTypeScopeEnd; 3974 break; 3975 3976 case N_SOL: 3977 // #included file name: name,,n_sect,0,address 3978 type = eSymbolTypeHeaderFile; 3979 3980 // We currently don't use the header files on darwin 3981 add_nlist = false; 3982 break; 3983 3984 case N_PARAMS: 3985 // compiler parameters: name,,NO_SECT,0,0 3986 type = eSymbolTypeCompiler; 3987 break; 3988 3989 case N_VERSION: 3990 // compiler version: name,,NO_SECT,0,0 3991 type = eSymbolTypeCompiler; 3992 break; 3993 3994 case N_OLEVEL: 3995 // compiler -O level: name,,NO_SECT,0,0 3996 type = eSymbolTypeCompiler; 3997 break; 3998 3999 case N_PSYM: 4000 // parameter: name,,NO_SECT,type,offset 4001 type = eSymbolTypeVariable; 4002 break; 4003 4004 case N_ENTRY: 4005 // alternate entry: name,,n_sect,linenumber,address 4006 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4007 type = eSymbolTypeLineEntry; 4008 break; 4009 4010 // Left and Right Braces 4011 case N_LBRAC: 4012 // left bracket: 0,,NO_SECT,nesting level,address We use the 4013 // current number of symbols in the symbol table in lieu of using 4014 // nlist_idx in case we ever start trimming entries out 4015 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4016 N_BRAC_indexes.push_back(sym_idx); 4017 type = eSymbolTypeScopeBegin; 4018 break; 4019 4020 case N_RBRAC: 4021 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4022 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4023 // can always skip the entire symbol if we need to navigate more 4024 // quickly at the source level when parsing STABS 4025 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4026 if (!N_BRAC_indexes.empty()) { 4027 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4028 symbol_ptr->SetByteSize(sym_idx + 1); 4029 symbol_ptr->SetSizeIsSibling(true); 4030 N_BRAC_indexes.pop_back(); 4031 } 4032 type = eSymbolTypeScopeEnd; 4033 break; 4034 4035 case N_EXCL: 4036 // deleted include file: name,,NO_SECT,0,sum 4037 type = eSymbolTypeHeaderFile; 4038 break; 4039 4040 // COMM scopes 4041 case N_BCOMM: 4042 // begin common: name,,NO_SECT,0,0 4043 // We use the current number of symbols in the symbol table in lieu 4044 // of using nlist_idx in case we ever start trimming entries out 4045 type = eSymbolTypeScopeBegin; 4046 N_COMM_indexes.push_back(sym_idx); 4047 break; 4048 4049 case N_ECOML: 4050 // end common (local name): 0,,n_sect,0,address 4051 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4052 LLVM_FALLTHROUGH; 4053 4054 case N_ECOMM: 4055 // end common: name,,n_sect,0,0 4056 // Set the size of the N_BCOMM to the terminating index of this 4057 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4058 // we need to navigate more quickly at the source level when 4059 // parsing STABS 4060 if (!N_COMM_indexes.empty()) { 4061 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4062 symbol_ptr->SetByteSize(sym_idx + 1); 4063 symbol_ptr->SetSizeIsSibling(true); 4064 N_COMM_indexes.pop_back(); 4065 } 4066 type = eSymbolTypeScopeEnd; 4067 break; 4068 4069 case N_LENG: 4070 // second stab entry with length information 4071 type = eSymbolTypeAdditional; 4072 break; 4073 4074 default: 4075 break; 4076 } 4077 } else { 4078 uint8_t n_type = N_TYPE & nlist.n_type; 4079 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4080 4081 switch (n_type) { 4082 case N_INDR: { 4083 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value); 4084 if (reexport_name_cstr && reexport_name_cstr[0]) { 4085 type = eSymbolTypeReExported; 4086 ConstString reexport_name(reexport_name_cstr + 4087 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4088 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4089 set_value = false; 4090 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4091 indirect_symbol_names.insert( 4092 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4093 } else 4094 type = eSymbolTypeUndefined; 4095 } break; 4096 4097 case N_UNDF: 4098 if (symbol_name && symbol_name[0]) { 4099 ConstString undefined_name(symbol_name + 4100 ((symbol_name[0] == '_') ? 1 : 0)); 4101 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4102 } 4103 LLVM_FALLTHROUGH; 4104 4105 case N_PBUD: 4106 type = eSymbolTypeUndefined; 4107 break; 4108 4109 case N_ABS: 4110 type = eSymbolTypeAbsolute; 4111 break; 4112 4113 case N_SECT: { 4114 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value); 4115 4116 if (!symbol_section) { 4117 // TODO: warn about this? 4118 add_nlist = false; 4119 break; 4120 } 4121 4122 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4123 type = eSymbolTypeException; 4124 } else { 4125 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4126 4127 switch (section_type) { 4128 case S_CSTRING_LITERALS: 4129 type = eSymbolTypeData; 4130 break; // section with only literal C strings 4131 case S_4BYTE_LITERALS: 4132 type = eSymbolTypeData; 4133 break; // section with only 4 byte literals 4134 case S_8BYTE_LITERALS: 4135 type = eSymbolTypeData; 4136 break; // section with only 8 byte literals 4137 case S_LITERAL_POINTERS: 4138 type = eSymbolTypeTrampoline; 4139 break; // section with only pointers to literals 4140 case S_NON_LAZY_SYMBOL_POINTERS: 4141 type = eSymbolTypeTrampoline; 4142 break; // section with only non-lazy symbol pointers 4143 case S_LAZY_SYMBOL_POINTERS: 4144 type = eSymbolTypeTrampoline; 4145 break; // section with only lazy symbol pointers 4146 case S_SYMBOL_STUBS: 4147 type = eSymbolTypeTrampoline; 4148 break; // section with only symbol stubs, byte size of stub in 4149 // the reserved2 field 4150 case S_MOD_INIT_FUNC_POINTERS: 4151 type = eSymbolTypeCode; 4152 break; // section with only function pointers for initialization 4153 case S_MOD_TERM_FUNC_POINTERS: 4154 type = eSymbolTypeCode; 4155 break; // section with only function pointers for termination 4156 case S_INTERPOSING: 4157 type = eSymbolTypeTrampoline; 4158 break; // section with only pairs of function pointers for 4159 // interposing 4160 case S_16BYTE_LITERALS: 4161 type = eSymbolTypeData; 4162 break; // section with only 16 byte literals 4163 case S_DTRACE_DOF: 4164 type = eSymbolTypeInstrumentation; 4165 break; 4166 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4167 type = eSymbolTypeTrampoline; 4168 break; 4169 default: 4170 switch (symbol_section->GetType()) { 4171 case lldb::eSectionTypeCode: 4172 type = eSymbolTypeCode; 4173 break; 4174 case eSectionTypeData: 4175 case eSectionTypeDataCString: // Inlined C string data 4176 case eSectionTypeDataCStringPointers: // Pointers to C string 4177 // data 4178 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4179 // the symbol table 4180 case eSectionTypeData4: 4181 case eSectionTypeData8: 4182 case eSectionTypeData16: 4183 type = eSymbolTypeData; 4184 break; 4185 default: 4186 break; 4187 } 4188 break; 4189 } 4190 4191 if (type == eSymbolTypeInvalid) { 4192 const char *symbol_sect_name = 4193 symbol_section->GetName().AsCString(); 4194 if (symbol_section->IsDescendant(text_section_sp.get())) { 4195 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4196 S_ATTR_SELF_MODIFYING_CODE | 4197 S_ATTR_SOME_INSTRUCTIONS)) 4198 type = eSymbolTypeData; 4199 else 4200 type = eSymbolTypeCode; 4201 } else if (symbol_section->IsDescendant(data_section_sp.get()) || 4202 symbol_section->IsDescendant( 4203 data_dirty_section_sp.get()) || 4204 symbol_section->IsDescendant( 4205 data_const_section_sp.get())) { 4206 if (symbol_sect_name && 4207 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) { 4208 type = eSymbolTypeRuntime; 4209 4210 if (symbol_name) { 4211 llvm::StringRef symbol_name_ref(symbol_name); 4212 if (symbol_name_ref.startswith("_OBJC_")) { 4213 llvm::StringRef g_objc_v2_prefix_class( 4214 "_OBJC_CLASS_$_"); 4215 llvm::StringRef g_objc_v2_prefix_metaclass( 4216 "_OBJC_METACLASS_$_"); 4217 llvm::StringRef g_objc_v2_prefix_ivar( 4218 "_OBJC_IVAR_$_"); 4219 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 4220 symbol_name_non_abi_mangled = symbol_name + 1; 4221 symbol_name = 4222 symbol_name + g_objc_v2_prefix_class.size(); 4223 type = eSymbolTypeObjCClass; 4224 demangled_is_synthesized = true; 4225 } else if (symbol_name_ref.startswith( 4226 g_objc_v2_prefix_metaclass)) { 4227 symbol_name_non_abi_mangled = symbol_name + 1; 4228 symbol_name = 4229 symbol_name + g_objc_v2_prefix_metaclass.size(); 4230 type = eSymbolTypeObjCMetaClass; 4231 demangled_is_synthesized = true; 4232 } else if (symbol_name_ref.startswith( 4233 g_objc_v2_prefix_ivar)) { 4234 symbol_name_non_abi_mangled = symbol_name + 1; 4235 symbol_name = 4236 symbol_name + g_objc_v2_prefix_ivar.size(); 4237 type = eSymbolTypeObjCIVar; 4238 demangled_is_synthesized = true; 4239 } 4240 } 4241 } 4242 } else if (symbol_sect_name && 4243 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4244 symbol_sect_name) { 4245 type = eSymbolTypeException; 4246 } else { 4247 type = eSymbolTypeData; 4248 } 4249 } else if (symbol_sect_name && 4250 ::strstr(symbol_sect_name, "__IMPORT") == 4251 symbol_sect_name) { 4252 type = eSymbolTypeTrampoline; 4253 } else if (symbol_section->IsDescendant(objc_section_sp.get())) { 4254 type = eSymbolTypeRuntime; 4255 if (symbol_name && symbol_name[0] == '.') { 4256 llvm::StringRef symbol_name_ref(symbol_name); 4257 llvm::StringRef g_objc_v1_prefix_class( 4258 ".objc_class_name_"); 4259 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4260 symbol_name_non_abi_mangled = symbol_name; 4261 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4262 type = eSymbolTypeObjCClass; 4263 demangled_is_synthesized = true; 4264 } 4265 } 4266 } 4267 } 4268 } 4269 } break; 4270 } 4271 } 4272 4273 if (!add_nlist) { 4274 sym[sym_idx].Clear(); 4275 return true; 4276 } 4277 4278 uint64_t symbol_value = nlist.n_value; 4279 4280 if (symbol_name_non_abi_mangled) { 4281 sym[sym_idx].GetMangled().SetMangledName( 4282 ConstString(symbol_name_non_abi_mangled)); 4283 sym[sym_idx].GetMangled().SetDemangledName(ConstString(symbol_name)); 4284 } else { 4285 bool symbol_name_is_mangled = false; 4286 4287 if (symbol_name && symbol_name[0] == '_') { 4288 symbol_name_is_mangled = symbol_name[1] == '_'; 4289 symbol_name++; // Skip the leading underscore 4290 } 4291 4292 if (symbol_name) { 4293 ConstString const_symbol_name(symbol_name); 4294 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4295 symbol_name_is_mangled); 4296 } 4297 } 4298 4299 if (is_gsym) { 4300 const char *gsym_name = 4301 sym[sym_idx] 4302 .GetMangled() 4303 .GetName(lldb::eLanguageTypeUnknown, Mangled::ePreferMangled) 4304 .GetCString(); 4305 if (gsym_name) 4306 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4307 } 4308 4309 if (symbol_section) { 4310 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4311 if (symbol_byte_size == 0 && function_starts_count > 0) { 4312 addr_t symbol_lookup_file_addr = nlist.n_value; 4313 // Do an exact address match for non-ARM addresses, else get the 4314 // closest since the symbol might be a thumb symbol which has an 4315 // address with bit zero set. 4316 FunctionStarts::Entry *func_start_entry = 4317 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4318 if (is_arm && func_start_entry) { 4319 // Verify that the function start address is the symbol address 4320 // (ARM) or the symbol address + 1 (thumb). 4321 if (func_start_entry->addr != symbol_lookup_file_addr && 4322 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4323 // Not the right entry, NULL it out... 4324 func_start_entry = nullptr; 4325 } 4326 } 4327 if (func_start_entry) { 4328 func_start_entry->data = true; 4329 4330 addr_t symbol_file_addr = func_start_entry->addr; 4331 if (is_arm) 4332 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4333 4334 const FunctionStarts::Entry *next_func_start_entry = 4335 function_starts.FindNextEntry(func_start_entry); 4336 const addr_t section_end_file_addr = 4337 section_file_addr + symbol_section->GetByteSize(); 4338 if (next_func_start_entry) { 4339 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4340 // Be sure the clear the Thumb address bit when we calculate the 4341 // size from the current and next address 4342 if (is_arm) 4343 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4344 symbol_byte_size = std::min<lldb::addr_t>( 4345 next_symbol_file_addr - symbol_file_addr, 4346 section_end_file_addr - symbol_file_addr); 4347 } else { 4348 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4349 } 4350 } 4351 } 4352 symbol_value -= section_file_addr; 4353 } 4354 4355 if (!is_debug) { 4356 if (type == eSymbolTypeCode) { 4357 // See if we can find a N_FUN entry for any code symbols. If we do 4358 // find a match, and the name matches, then we can merge the two into 4359 // just the function symbol to avoid duplicate entries in the symbol 4360 // table. 4361 std::pair<ValueToSymbolIndexMap::const_iterator, 4362 ValueToSymbolIndexMap::const_iterator> 4363 range; 4364 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4365 if (range.first != range.second) { 4366 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4367 pos != range.second; ++pos) { 4368 if (sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, 4369 Mangled::ePreferMangled) == 4370 sym[pos->second].GetMangled().GetName( 4371 lldb::eLanguageTypeUnknown, Mangled::ePreferMangled)) { 4372 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4373 // We just need the flags from the linker symbol, so put these 4374 // flags into the N_FUN flags to avoid duplicate symbols in the 4375 // symbol table. 4376 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4377 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4378 if (resolver_addresses.find(nlist.n_value) != 4379 resolver_addresses.end()) 4380 sym[pos->second].SetType(eSymbolTypeResolver); 4381 sym[sym_idx].Clear(); 4382 return true; 4383 } 4384 } 4385 } else { 4386 if (resolver_addresses.find(nlist.n_value) != 4387 resolver_addresses.end()) 4388 type = eSymbolTypeResolver; 4389 } 4390 } else if (type == eSymbolTypeData || type == eSymbolTypeObjCClass || 4391 type == eSymbolTypeObjCMetaClass || 4392 type == eSymbolTypeObjCIVar) { 4393 // See if we can find a N_STSYM entry for any data symbols. If we do 4394 // find a match, and the name matches, then we can merge the two into 4395 // just the Static symbol to avoid duplicate entries in the symbol 4396 // table. 4397 std::pair<ValueToSymbolIndexMap::const_iterator, 4398 ValueToSymbolIndexMap::const_iterator> 4399 range; 4400 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4401 if (range.first != range.second) { 4402 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4403 pos != range.second; ++pos) { 4404 if (sym[sym_idx].GetMangled().GetName(lldb::eLanguageTypeUnknown, 4405 Mangled::ePreferMangled) == 4406 sym[pos->second].GetMangled().GetName( 4407 lldb::eLanguageTypeUnknown, Mangled::ePreferMangled)) { 4408 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4409 // We just need the flags from the linker symbol, so put these 4410 // flags into the N_STSYM flags to avoid duplicate symbols in 4411 // the symbol table. 4412 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4413 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4414 sym[sym_idx].Clear(); 4415 return true; 4416 } 4417 } 4418 } else { 4419 // Combine N_GSYM stab entries with the non stab symbol. 4420 const char *gsym_name = sym[sym_idx] 4421 .GetMangled() 4422 .GetName(lldb::eLanguageTypeUnknown, 4423 Mangled::ePreferMangled) 4424 .GetCString(); 4425 if (gsym_name) { 4426 ConstNameToSymbolIndexMap::const_iterator pos = 4427 N_GSYM_name_to_sym_idx.find(gsym_name); 4428 if (pos != N_GSYM_name_to_sym_idx.end()) { 4429 const uint32_t GSYM_sym_idx = pos->second; 4430 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4431 // Copy the address, because often the N_GSYM address has an 4432 // invalid address of zero when the global is a common symbol. 4433 sym[GSYM_sym_idx].GetAddressRef().SetSection(symbol_section); 4434 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4435 // We just need the flags from the linker symbol, so put these 4436 // flags into the N_GSYM flags to avoid duplicate symbols in 4437 // the symbol table. 4438 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4439 sym[sym_idx].Clear(); 4440 return true; 4441 } 4442 } 4443 } 4444 } 4445 } 4446 4447 sym[sym_idx].SetID(nlist_idx); 4448 sym[sym_idx].SetType(type); 4449 if (set_value) { 4450 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4451 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4452 } 4453 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4454 if (nlist.n_desc & N_WEAK_REF) 4455 sym[sym_idx].SetIsWeak(true); 4456 4457 if (symbol_byte_size > 0) 4458 sym[sym_idx].SetByteSize(symbol_byte_size); 4459 4460 if (demangled_is_synthesized) 4461 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4462 4463 ++sym_idx; 4464 return true; 4465 }; 4466 4467 // First parse all the nlists but don't process them yet. See the next 4468 // comment for an explanation why. 4469 std::vector<struct nlist_64> nlists; 4470 nlists.reserve(symtab_load_command.nsyms); 4471 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 4472 if (auto nlist = 4473 ParseNList(nlist_data, nlist_data_offset, nlist_byte_size)) 4474 nlists.push_back(*nlist); 4475 else 4476 break; 4477 } 4478 4479 // Now parse all the debug symbols. This is needed to merge non-debug 4480 // symbols in the next step. Non-debug symbols are always coalesced into 4481 // the debug symbol. Doing this in one step would mean that some symbols 4482 // won't be merged. 4483 nlist_idx = 0; 4484 for (auto &nlist : nlists) { 4485 if (!ParseSymbolLambda(nlist, nlist_idx++, DebugSymbols)) 4486 break; 4487 } 4488 4489 // Finally parse all the non debug symbols. 4490 nlist_idx = 0; 4491 for (auto &nlist : nlists) { 4492 if (!ParseSymbolLambda(nlist, nlist_idx++, NonDebugSymbols)) 4493 break; 4494 } 4495 4496 for (const auto &pos : reexport_shlib_needs_fixup) { 4497 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4498 if (undef_pos != undefined_name_to_desc.end()) { 4499 const uint8_t dylib_ordinal = 4500 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4501 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4502 sym[pos.first].SetReExportedSymbolSharedLibrary( 4503 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4504 } 4505 } 4506 } 4507 4508 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4509 4510 if (function_starts_count > 0) { 4511 uint32_t num_synthetic_function_symbols = 0; 4512 for (i = 0; i < function_starts_count; ++i) { 4513 if (!function_starts.GetEntryRef(i).data) 4514 ++num_synthetic_function_symbols; 4515 } 4516 4517 if (num_synthetic_function_symbols > 0) { 4518 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4519 num_syms = sym_idx + num_synthetic_function_symbols; 4520 sym = symtab->Resize(num_syms); 4521 } 4522 for (i = 0; i < function_starts_count; ++i) { 4523 const FunctionStarts::Entry *func_start_entry = 4524 function_starts.GetEntryAtIndex(i); 4525 if (!func_start_entry->data) { 4526 addr_t symbol_file_addr = func_start_entry->addr; 4527 uint32_t symbol_flags = 0; 4528 if (is_arm) { 4529 if (symbol_file_addr & 1) 4530 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4531 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4532 } 4533 Address symbol_addr; 4534 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4535 SectionSP symbol_section(symbol_addr.GetSection()); 4536 uint32_t symbol_byte_size = 0; 4537 if (symbol_section) { 4538 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4539 const FunctionStarts::Entry *next_func_start_entry = 4540 function_starts.FindNextEntry(func_start_entry); 4541 const addr_t section_end_file_addr = 4542 section_file_addr + symbol_section->GetByteSize(); 4543 if (next_func_start_entry) { 4544 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4545 if (is_arm) 4546 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4547 symbol_byte_size = std::min<lldb::addr_t>( 4548 next_symbol_file_addr - symbol_file_addr, 4549 section_end_file_addr - symbol_file_addr); 4550 } else { 4551 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4552 } 4553 sym[sym_idx].SetID(synthetic_sym_id++); 4554 sym[sym_idx].GetMangled().SetDemangledName( 4555 GetNextSyntheticSymbolName()); 4556 sym[sym_idx].SetType(eSymbolTypeCode); 4557 sym[sym_idx].SetIsSynthetic(true); 4558 sym[sym_idx].GetAddressRef() = symbol_addr; 4559 if (symbol_flags) 4560 sym[sym_idx].SetFlags(symbol_flags); 4561 if (symbol_byte_size) 4562 sym[sym_idx].SetByteSize(symbol_byte_size); 4563 ++sym_idx; 4564 } 4565 } 4566 } 4567 } 4568 } 4569 } 4570 4571 // Trim our symbols down to just what we ended up with after removing any 4572 // symbols. 4573 if (sym_idx < num_syms) { 4574 num_syms = sym_idx; 4575 sym = symtab->Resize(num_syms); 4576 } 4577 4578 // Now synthesize indirect symbols 4579 if (m_dysymtab.nindirectsyms != 0) { 4580 if (indirect_symbol_index_data.GetByteSize()) { 4581 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4582 m_nlist_idx_to_sym_idx.end(); 4583 4584 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4585 ++sect_idx) { 4586 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4587 S_SYMBOL_STUBS) { 4588 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4589 if (symbol_stub_byte_size == 0) 4590 continue; 4591 4592 const uint32_t num_symbol_stubs = 4593 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4594 4595 if (num_symbol_stubs == 0) 4596 continue; 4597 4598 const uint32_t symbol_stub_index_offset = 4599 m_mach_sections[sect_idx].reserved1; 4600 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) { 4601 const uint32_t symbol_stub_index = 4602 symbol_stub_index_offset + stub_idx; 4603 const lldb::addr_t symbol_stub_addr = 4604 m_mach_sections[sect_idx].addr + 4605 (stub_idx * symbol_stub_byte_size); 4606 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4607 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4608 symbol_stub_offset, 4)) { 4609 const uint32_t stub_sym_id = 4610 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4611 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4612 continue; 4613 4614 NListIndexToSymbolIndexMap::const_iterator index_pos = 4615 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4616 Symbol *stub_symbol = nullptr; 4617 if (index_pos != end_index_pos) { 4618 // We have a remapping from the original nlist index to a 4619 // current symbol index, so just look this up by index 4620 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4621 } else { 4622 // We need to lookup a symbol using the original nlist symbol 4623 // index since this index is coming from the S_SYMBOL_STUBS 4624 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4625 } 4626 4627 if (stub_symbol) { 4628 Address so_addr(symbol_stub_addr, section_list); 4629 4630 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4631 // Change the external symbol into a trampoline that makes 4632 // sense These symbols were N_UNDF N_EXT, and are useless 4633 // to us, so we can re-use them so we don't have to make up 4634 // a synthetic symbol for no good reason. 4635 if (resolver_addresses.find(symbol_stub_addr) == 4636 resolver_addresses.end()) 4637 stub_symbol->SetType(eSymbolTypeTrampoline); 4638 else 4639 stub_symbol->SetType(eSymbolTypeResolver); 4640 stub_symbol->SetExternal(false); 4641 stub_symbol->GetAddressRef() = so_addr; 4642 stub_symbol->SetByteSize(symbol_stub_byte_size); 4643 } else { 4644 // Make a synthetic symbol to describe the trampoline stub 4645 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4646 if (sym_idx >= num_syms) { 4647 sym = symtab->Resize(++num_syms); 4648 stub_symbol = nullptr; // this pointer no longer valid 4649 } 4650 sym[sym_idx].SetID(synthetic_sym_id++); 4651 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4652 if (resolver_addresses.find(symbol_stub_addr) == 4653 resolver_addresses.end()) 4654 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4655 else 4656 sym[sym_idx].SetType(eSymbolTypeResolver); 4657 sym[sym_idx].SetIsSynthetic(true); 4658 sym[sym_idx].GetAddressRef() = so_addr; 4659 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4660 ++sym_idx; 4661 } 4662 } else { 4663 if (log) 4664 log->Warning("symbol stub referencing symbol table symbol " 4665 "%u that isn't in our minimal symbol table, " 4666 "fix this!!!", 4667 stub_sym_id); 4668 } 4669 } 4670 } 4671 } 4672 } 4673 } 4674 } 4675 4676 if (!trie_entries.empty()) { 4677 for (const auto &e : trie_entries) { 4678 if (e.entry.import_name) { 4679 // Only add indirect symbols from the Trie entries if we didn't have 4680 // a N_INDR nlist entry for this already 4681 if (indirect_symbol_names.find(e.entry.name) == 4682 indirect_symbol_names.end()) { 4683 // Make a synthetic symbol to describe re-exported symbol. 4684 if (sym_idx >= num_syms) 4685 sym = symtab->Resize(++num_syms); 4686 sym[sym_idx].SetID(synthetic_sym_id++); 4687 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4688 sym[sym_idx].SetType(eSymbolTypeReExported); 4689 sym[sym_idx].SetIsSynthetic(true); 4690 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4691 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4692 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4693 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4694 } 4695 ++sym_idx; 4696 } 4697 } 4698 } 4699 } 4700 4701 // StreamFile s(stdout, false); 4702 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4703 // symtab->Dump(&s, NULL, eSortOrderNone); 4704 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4705 // sizes 4706 symtab->CalculateSymbolSizes(); 4707 4708 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4709 // symtab->Dump(&s, NULL, eSortOrderNone); 4710 4711 return symtab->GetNumSymbols(); 4712 } 4713 4714 void ObjectFileMachO::Dump(Stream *s) { 4715 ModuleSP module_sp(GetModule()); 4716 if (module_sp) { 4717 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4718 s->Printf("%p: ", static_cast<void *>(this)); 4719 s->Indent(); 4720 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4721 s->PutCString("ObjectFileMachO64"); 4722 else 4723 s->PutCString("ObjectFileMachO32"); 4724 4725 *s << ", file = '" << m_file; 4726 ModuleSpecList all_specs; 4727 ModuleSpec base_spec; 4728 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), 4729 base_spec, all_specs); 4730 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 4731 *s << "', triple"; 4732 if (e) 4733 s->Printf("[%d]", i); 4734 *s << " = "; 4735 *s << all_specs.GetModuleSpecRefAtIndex(i) 4736 .GetArchitecture() 4737 .GetTriple() 4738 .getTriple(); 4739 } 4740 *s << "\n"; 4741 SectionList *sections = GetSectionList(); 4742 if (sections) 4743 sections->Dump(s, nullptr, true, UINT32_MAX); 4744 4745 if (m_symtab_up) 4746 m_symtab_up->Dump(s, nullptr, eSortOrderNone); 4747 } 4748 } 4749 4750 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4751 const lldb_private::DataExtractor &data, 4752 lldb::offset_t lc_offset) { 4753 uint32_t i; 4754 struct uuid_command load_cmd; 4755 4756 lldb::offset_t offset = lc_offset; 4757 for (i = 0; i < header.ncmds; ++i) { 4758 const lldb::offset_t cmd_offset = offset; 4759 if (data.GetU32(&offset, &load_cmd, 2) == nullptr) 4760 break; 4761 4762 if (load_cmd.cmd == LC_UUID) { 4763 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4764 4765 if (uuid_bytes) { 4766 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4767 // We pretend these object files have no UUID to prevent crashing. 4768 4769 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4770 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4771 0xbb, 0x14, 0xf0, 0x0d}; 4772 4773 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4774 return UUID(); 4775 4776 return UUID::fromOptionalData(uuid_bytes, 16); 4777 } 4778 return UUID(); 4779 } 4780 offset = cmd_offset + load_cmd.cmdsize; 4781 } 4782 return UUID(); 4783 } 4784 4785 static llvm::StringRef GetOSName(uint32_t cmd) { 4786 switch (cmd) { 4787 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4788 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4789 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4790 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4791 case llvm::MachO::LC_VERSION_MIN_TVOS: 4792 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4793 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4794 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4795 default: 4796 llvm_unreachable("unexpected LC_VERSION load command"); 4797 } 4798 } 4799 4800 namespace { 4801 struct OSEnv { 4802 llvm::StringRef os_type; 4803 llvm::StringRef environment; 4804 OSEnv(uint32_t cmd) { 4805 switch (cmd) { 4806 case llvm::MachO::PLATFORM_MACOS: 4807 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4808 return; 4809 case llvm::MachO::PLATFORM_IOS: 4810 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4811 return; 4812 case llvm::MachO::PLATFORM_TVOS: 4813 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4814 return; 4815 case llvm::MachO::PLATFORM_WATCHOS: 4816 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4817 return; 4818 // NEED_BRIDGEOS_TRIPLE case llvm::MachO::PLATFORM_BRIDGEOS: 4819 // NEED_BRIDGEOS_TRIPLE os_type = 4820 // llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4821 // NEED_BRIDGEOS_TRIPLE return; 4822 case llvm::MachO::PLATFORM_MACCATALYST: 4823 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4824 environment = llvm::Triple::getEnvironmentTypeName(llvm::Triple::MacABI); 4825 return; 4826 case llvm::MachO::PLATFORM_IOSSIMULATOR: 4827 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4828 environment = 4829 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4830 return; 4831 case llvm::MachO::PLATFORM_TVOSSIMULATOR: 4832 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4833 environment = 4834 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4835 return; 4836 case llvm::MachO::PLATFORM_WATCHOSSIMULATOR: 4837 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4838 environment = 4839 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4840 return; 4841 default: { 4842 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 4843 LIBLLDB_LOG_PROCESS)); 4844 LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION"); 4845 } 4846 } 4847 } 4848 }; 4849 4850 struct MinOS { 4851 uint32_t major_version, minor_version, patch_version; 4852 MinOS(uint32_t version) 4853 : major_version(version >> 16), minor_version((version >> 8) & 0xffu), 4854 patch_version(version & 0xffu) {} 4855 }; 4856 } // namespace 4857 4858 void ObjectFileMachO::GetAllArchSpecs(const llvm::MachO::mach_header &header, 4859 const lldb_private::DataExtractor &data, 4860 lldb::offset_t lc_offset, 4861 ModuleSpec &base_spec, 4862 lldb_private::ModuleSpecList &all_specs) { 4863 auto &base_arch = base_spec.GetArchitecture(); 4864 base_arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4865 if (!base_arch.IsValid()) 4866 return; 4867 4868 bool found_any = false; 4869 auto add_triple = [&](const llvm::Triple &triple) { 4870 auto spec = base_spec; 4871 spec.GetArchitecture().GetTriple() = triple; 4872 if (spec.GetArchitecture().IsValid()) { 4873 spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset); 4874 all_specs.Append(spec); 4875 found_any = true; 4876 } 4877 }; 4878 4879 // Set OS to an unspecified unknown or a "*" so it can match any OS 4880 llvm::Triple base_triple = base_arch.GetTriple(); 4881 base_triple.setOS(llvm::Triple::UnknownOS); 4882 base_triple.setOSName(llvm::StringRef()); 4883 4884 if (header.filetype == MH_PRELOAD) { 4885 if (header.cputype == CPU_TYPE_ARM) { 4886 // If this is a 32-bit arm binary, and it's a standalone binary, force 4887 // the Vendor to Apple so we don't accidentally pick up the generic 4888 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4889 // frame pointer register; most other armv7 ABIs use a combination of 4890 // r7 and r11. 4891 base_triple.setVendor(llvm::Triple::Apple); 4892 } else { 4893 // Set vendor to an unspecified unknown or a "*" so it can match any 4894 // vendor This is required for correct behavior of EFI debugging on 4895 // x86_64 4896 base_triple.setVendor(llvm::Triple::UnknownVendor); 4897 base_triple.setVendorName(llvm::StringRef()); 4898 } 4899 return add_triple(base_triple); 4900 } 4901 4902 struct load_command load_cmd; 4903 4904 // See if there is an LC_VERSION_MIN_* load command that can give 4905 // us the OS type. 4906 lldb::offset_t offset = lc_offset; 4907 for (uint32_t i = 0; i < header.ncmds; ++i) { 4908 const lldb::offset_t cmd_offset = offset; 4909 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4910 break; 4911 4912 struct version_min_command version_min; 4913 switch (load_cmd.cmd) { 4914 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4915 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4916 case llvm::MachO::LC_VERSION_MIN_TVOS: 4917 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 4918 if (load_cmd.cmdsize != sizeof(version_min)) 4919 break; 4920 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 4921 data.GetByteOrder(), &version_min) == 0) 4922 break; 4923 MinOS min_os(version_min.version); 4924 llvm::SmallString<32> os_name; 4925 llvm::raw_svector_ostream os(os_name); 4926 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 4927 << min_os.minor_version << '.' << min_os.patch_version; 4928 4929 auto triple = base_triple; 4930 triple.setOSName(os.str()); 4931 os_name.clear(); 4932 add_triple(triple); 4933 break; 4934 } 4935 default: 4936 break; 4937 } 4938 4939 offset = cmd_offset + load_cmd.cmdsize; 4940 } 4941 4942 // See if there are LC_BUILD_VERSION load commands that can give 4943 // us the OS type. 4944 offset = lc_offset; 4945 for (uint32_t i = 0; i < header.ncmds; ++i) { 4946 const lldb::offset_t cmd_offset = offset; 4947 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4948 break; 4949 4950 do { 4951 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 4952 struct build_version_command build_version; 4953 if (load_cmd.cmdsize < sizeof(build_version)) { 4954 // Malformed load command. 4955 break; 4956 } 4957 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 4958 data.GetByteOrder(), &build_version) == 0) 4959 break; 4960 MinOS min_os(build_version.minos); 4961 OSEnv os_env(build_version.platform); 4962 llvm::SmallString<16> os_name; 4963 llvm::raw_svector_ostream os(os_name); 4964 os << os_env.os_type << min_os.major_version << '.' 4965 << min_os.minor_version << '.' << min_os.patch_version; 4966 auto triple = base_triple; 4967 triple.setOSName(os.str()); 4968 os_name.clear(); 4969 if (!os_env.environment.empty()) 4970 triple.setEnvironmentName(os_env.environment); 4971 add_triple(triple); 4972 } 4973 } while (0); 4974 offset = cmd_offset + load_cmd.cmdsize; 4975 } 4976 4977 if (!found_any) { 4978 if (header.filetype == MH_KEXT_BUNDLE) { 4979 base_triple.setVendor(llvm::Triple::Apple); 4980 add_triple(base_triple); 4981 } else { 4982 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 4983 // so lets not say our Vendor is Apple, leave it as an unspecified 4984 // unknown. 4985 base_triple.setVendor(llvm::Triple::UnknownVendor); 4986 base_triple.setVendorName(llvm::StringRef()); 4987 add_triple(base_triple); 4988 } 4989 } 4990 } 4991 4992 ArchSpec ObjectFileMachO::GetArchitecture( 4993 ModuleSP module_sp, const llvm::MachO::mach_header &header, 4994 const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) { 4995 ModuleSpecList all_specs; 4996 ModuleSpec base_spec; 4997 GetAllArchSpecs(header, data, MachHeaderSizeFromMagic(header.magic), 4998 base_spec, all_specs); 4999 5000 // If the object file offers multiple alternative load commands, 5001 // pick the one that matches the module. 5002 if (module_sp) { 5003 const ArchSpec &module_arch = module_sp->GetArchitecture(); 5004 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { 5005 ArchSpec mach_arch = 5006 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); 5007 if (module_arch.IsCompatibleMatch(mach_arch)) 5008 return mach_arch; 5009 } 5010 } 5011 5012 // Return the first arch we found. 5013 if (all_specs.GetSize() == 0) 5014 return {}; 5015 return all_specs.GetModuleSpecRefAtIndex(0).GetArchitecture(); 5016 } 5017 5018 UUID ObjectFileMachO::GetUUID() { 5019 ModuleSP module_sp(GetModule()); 5020 if (module_sp) { 5021 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5022 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5023 return GetUUID(m_header, m_data, offset); 5024 } 5025 return UUID(); 5026 } 5027 5028 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5029 uint32_t count = 0; 5030 ModuleSP module_sp(GetModule()); 5031 if (module_sp) { 5032 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5033 struct load_command load_cmd; 5034 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5035 std::vector<std::string> rpath_paths; 5036 std::vector<std::string> rpath_relative_paths; 5037 std::vector<std::string> at_exec_relative_paths; 5038 uint32_t i; 5039 for (i = 0; i < m_header.ncmds; ++i) { 5040 const uint32_t cmd_offset = offset; 5041 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5042 break; 5043 5044 switch (load_cmd.cmd) { 5045 case LC_RPATH: 5046 case LC_LOAD_DYLIB: 5047 case LC_LOAD_WEAK_DYLIB: 5048 case LC_REEXPORT_DYLIB: 5049 case LC_LOAD_DYLINKER: 5050 case LC_LOADFVMLIB: 5051 case LC_LOAD_UPWARD_DYLIB: { 5052 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5053 const char *path = m_data.PeekCStr(name_offset); 5054 if (path) { 5055 if (load_cmd.cmd == LC_RPATH) 5056 rpath_paths.push_back(path); 5057 else { 5058 if (path[0] == '@') { 5059 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5060 rpath_relative_paths.push_back(path + strlen("@rpath")); 5061 else if (strncmp(path, "@executable_path", 5062 strlen("@executable_path")) == 0) 5063 at_exec_relative_paths.push_back(path + 5064 strlen("@executable_path")); 5065 } else { 5066 FileSpec file_spec(path); 5067 if (files.AppendIfUnique(file_spec)) 5068 count++; 5069 } 5070 } 5071 } 5072 } break; 5073 5074 default: 5075 break; 5076 } 5077 offset = cmd_offset + load_cmd.cmdsize; 5078 } 5079 5080 FileSpec this_file_spec(m_file); 5081 FileSystem::Instance().Resolve(this_file_spec); 5082 5083 if (!rpath_paths.empty()) { 5084 // Fixup all LC_RPATH values to be absolute paths 5085 std::string loader_path("@loader_path"); 5086 std::string executable_path("@executable_path"); 5087 for (auto &rpath : rpath_paths) { 5088 if (rpath.find(loader_path) == 0) { 5089 rpath.erase(0, loader_path.size()); 5090 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5091 } else if (rpath.find(executable_path) == 0) { 5092 rpath.erase(0, executable_path.size()); 5093 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5094 } 5095 } 5096 5097 for (const auto &rpath_relative_path : rpath_relative_paths) { 5098 for (const auto &rpath : rpath_paths) { 5099 std::string path = rpath; 5100 path += rpath_relative_path; 5101 // It is OK to resolve this path because we must find a file on disk 5102 // for us to accept it anyway if it is rpath relative. 5103 FileSpec file_spec(path); 5104 FileSystem::Instance().Resolve(file_spec); 5105 if (FileSystem::Instance().Exists(file_spec) && 5106 files.AppendIfUnique(file_spec)) { 5107 count++; 5108 break; 5109 } 5110 } 5111 } 5112 } 5113 5114 // We may have @executable_paths but no RPATHS. Figure those out here. 5115 // Only do this if this object file is the executable. We have no way to 5116 // get back to the actual executable otherwise, so we won't get the right 5117 // path. 5118 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5119 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5120 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5121 FileSpec file_spec = 5122 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5123 if (FileSystem::Instance().Exists(file_spec) && 5124 files.AppendIfUnique(file_spec)) 5125 count++; 5126 } 5127 } 5128 } 5129 return count; 5130 } 5131 5132 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5133 // If the object file is not an executable it can't hold the entry point. 5134 // m_entry_point_address is initialized to an invalid address, so we can just 5135 // return that. If m_entry_point_address is valid it means we've found it 5136 // already, so return the cached value. 5137 5138 if ((!IsExecutable() && !IsDynamicLoader()) || 5139 m_entry_point_address.IsValid()) { 5140 return m_entry_point_address; 5141 } 5142 5143 // Otherwise, look for the UnixThread or Thread command. The data for the 5144 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5145 // 5146 // uint32_t flavor - this is the flavor argument you would pass to 5147 // thread_get_state 5148 // uint32_t count - this is the count of longs in the thread state data 5149 // struct XXX_thread_state state - this is the structure from 5150 // <machine/thread_status.h> corresponding to the flavor. 5151 // <repeat this trio> 5152 // 5153 // So we just keep reading the various register flavors till we find the GPR 5154 // one, then read the PC out of there. 5155 // FIXME: We will need to have a "RegisterContext data provider" class at some 5156 // point that can get all the registers 5157 // out of data in this form & attach them to a given thread. That should 5158 // underlie the MacOS X User process plugin, and we'll also need it for the 5159 // MacOS X Core File process plugin. When we have that we can also use it 5160 // here. 5161 // 5162 // For now we hard-code the offsets and flavors we need: 5163 // 5164 // 5165 5166 ModuleSP module_sp(GetModule()); 5167 if (module_sp) { 5168 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5169 struct load_command load_cmd; 5170 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5171 uint32_t i; 5172 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5173 bool done = false; 5174 5175 for (i = 0; i < m_header.ncmds; ++i) { 5176 const lldb::offset_t cmd_offset = offset; 5177 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5178 break; 5179 5180 switch (load_cmd.cmd) { 5181 case LC_UNIXTHREAD: 5182 case LC_THREAD: { 5183 while (offset < cmd_offset + load_cmd.cmdsize) { 5184 uint32_t flavor = m_data.GetU32(&offset); 5185 uint32_t count = m_data.GetU32(&offset); 5186 if (count == 0) { 5187 // We've gotten off somehow, log and exit; 5188 return m_entry_point_address; 5189 } 5190 5191 switch (m_header.cputype) { 5192 case llvm::MachO::CPU_TYPE_ARM: 5193 if (flavor == 1 || 5194 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 5195 // from mach/arm/thread_status.h 5196 { 5197 offset += 60; // This is the offset of pc in the GPR thread state 5198 // data structure. 5199 start_address = m_data.GetU32(&offset); 5200 done = true; 5201 } 5202 break; 5203 case llvm::MachO::CPU_TYPE_ARM64: 5204 case llvm::MachO::CPU_TYPE_ARM64_32: 5205 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5206 { 5207 offset += 256; // This is the offset of pc in the GPR thread state 5208 // data structure. 5209 start_address = m_data.GetU64(&offset); 5210 done = true; 5211 } 5212 break; 5213 case llvm::MachO::CPU_TYPE_I386: 5214 if (flavor == 5215 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5216 { 5217 offset += 40; // This is the offset of eip in the GPR thread state 5218 // data structure. 5219 start_address = m_data.GetU32(&offset); 5220 done = true; 5221 } 5222 break; 5223 case llvm::MachO::CPU_TYPE_X86_64: 5224 if (flavor == 5225 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5226 { 5227 offset += 16 * 8; // This is the offset of rip in the GPR thread 5228 // state data structure. 5229 start_address = m_data.GetU64(&offset); 5230 done = true; 5231 } 5232 break; 5233 default: 5234 return m_entry_point_address; 5235 } 5236 // Haven't found the GPR flavor yet, skip over the data for this 5237 // flavor: 5238 if (done) 5239 break; 5240 offset += count * 4; 5241 } 5242 } break; 5243 case LC_MAIN: { 5244 ConstString text_segment_name("__TEXT"); 5245 uint64_t entryoffset = m_data.GetU64(&offset); 5246 SectionSP text_segment_sp = 5247 GetSectionList()->FindSectionByName(text_segment_name); 5248 if (text_segment_sp) { 5249 done = true; 5250 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5251 } 5252 } break; 5253 5254 default: 5255 break; 5256 } 5257 if (done) 5258 break; 5259 5260 // Go to the next load command: 5261 offset = cmd_offset + load_cmd.cmdsize; 5262 } 5263 5264 if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) { 5265 if (GetSymtab()) { 5266 Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType( 5267 ConstString("_dyld_start"), SymbolType::eSymbolTypeCode, 5268 Symtab::eDebugAny, Symtab::eVisibilityAny); 5269 if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) { 5270 start_address = dyld_start_sym->GetAddress().GetFileAddress(); 5271 } 5272 } 5273 } 5274 5275 if (start_address != LLDB_INVALID_ADDRESS) { 5276 // We got the start address from the load commands, so now resolve that 5277 // address in the sections of this ObjectFile: 5278 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5279 start_address, GetSectionList())) { 5280 m_entry_point_address.Clear(); 5281 } 5282 } else { 5283 // We couldn't read the UnixThread load command - maybe it wasn't there. 5284 // As a fallback look for the "start" symbol in the main executable. 5285 5286 ModuleSP module_sp(GetModule()); 5287 5288 if (module_sp) { 5289 SymbolContextList contexts; 5290 SymbolContext context; 5291 module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5292 eSymbolTypeCode, contexts); 5293 if (contexts.GetSize()) { 5294 if (contexts.GetContextAtIndex(0, context)) 5295 m_entry_point_address = context.symbol->GetAddress(); 5296 } 5297 } 5298 } 5299 } 5300 5301 return m_entry_point_address; 5302 } 5303 5304 lldb_private::Address ObjectFileMachO::GetBaseAddress() { 5305 lldb_private::Address header_addr; 5306 SectionList *section_list = GetSectionList(); 5307 if (section_list) { 5308 SectionSP text_segment_sp( 5309 section_list->FindSectionByName(GetSegmentNameTEXT())); 5310 if (text_segment_sp) { 5311 header_addr.SetSection(text_segment_sp); 5312 header_addr.SetOffset(0); 5313 } 5314 } 5315 return header_addr; 5316 } 5317 5318 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5319 ModuleSP module_sp(GetModule()); 5320 if (module_sp) { 5321 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5322 if (!m_thread_context_offsets_valid) { 5323 m_thread_context_offsets_valid = true; 5324 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5325 FileRangeArray::Entry file_range; 5326 thread_command thread_cmd; 5327 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5328 const uint32_t cmd_offset = offset; 5329 if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr) 5330 break; 5331 5332 if (thread_cmd.cmd == LC_THREAD) { 5333 file_range.SetRangeBase(offset); 5334 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5335 m_thread_context_offsets.Append(file_range); 5336 } 5337 offset = cmd_offset + thread_cmd.cmdsize; 5338 } 5339 } 5340 } 5341 return m_thread_context_offsets.GetSize(); 5342 } 5343 5344 std::string ObjectFileMachO::GetIdentifierString() { 5345 std::string result; 5346 ModuleSP module_sp(GetModule()); 5347 if (module_sp) { 5348 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5349 5350 // First, look over the load commands for an LC_NOTE load command with 5351 // data_owner string "kern ver str" & use that if found. 5352 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5353 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5354 const uint32_t cmd_offset = offset; 5355 load_command lc; 5356 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5357 break; 5358 if (lc.cmd == LC_NOTE) { 5359 char data_owner[17]; 5360 m_data.CopyData(offset, 16, data_owner); 5361 data_owner[16] = '\0'; 5362 offset += 16; 5363 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5364 uint64_t size = m_data.GetU64_unchecked(&offset); 5365 5366 // "kern ver str" has a uint32_t version and then a nul terminated 5367 // c-string. 5368 if (strcmp("kern ver str", data_owner) == 0) { 5369 offset = fileoff; 5370 uint32_t version; 5371 if (m_data.GetU32(&offset, &version, 1) != nullptr) { 5372 if (version == 1) { 5373 uint32_t strsize = size - sizeof(uint32_t); 5374 char *buf = (char *)malloc(strsize); 5375 if (buf) { 5376 m_data.CopyData(offset, strsize, buf); 5377 buf[strsize - 1] = '\0'; 5378 result = buf; 5379 if (buf) 5380 free(buf); 5381 return result; 5382 } 5383 } 5384 } 5385 } 5386 } 5387 offset = cmd_offset + lc.cmdsize; 5388 } 5389 5390 // Second, make a pass over the load commands looking for an obsolete 5391 // LC_IDENT load command. 5392 offset = MachHeaderSizeFromMagic(m_header.magic); 5393 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5394 const uint32_t cmd_offset = offset; 5395 struct ident_command ident_command; 5396 if (m_data.GetU32(&offset, &ident_command, 2) == nullptr) 5397 break; 5398 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5399 char *buf = (char *)malloc(ident_command.cmdsize); 5400 if (buf != nullptr && m_data.CopyData(offset, ident_command.cmdsize, 5401 buf) == ident_command.cmdsize) { 5402 buf[ident_command.cmdsize - 1] = '\0'; 5403 result = buf; 5404 } 5405 if (buf) 5406 free(buf); 5407 } 5408 offset = cmd_offset + ident_command.cmdsize; 5409 } 5410 } 5411 return result; 5412 } 5413 5414 bool ObjectFileMachO::GetCorefileMainBinaryInfo(addr_t &address, UUID &uuid) { 5415 address = LLDB_INVALID_ADDRESS; 5416 uuid.Clear(); 5417 ModuleSP module_sp(GetModule()); 5418 if (module_sp) { 5419 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5420 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5421 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5422 const uint32_t cmd_offset = offset; 5423 load_command lc; 5424 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5425 break; 5426 if (lc.cmd == LC_NOTE) { 5427 char data_owner[17]; 5428 memset(data_owner, 0, sizeof(data_owner)); 5429 m_data.CopyData(offset, 16, data_owner); 5430 offset += 16; 5431 uint64_t fileoff = m_data.GetU64_unchecked(&offset); 5432 uint64_t size = m_data.GetU64_unchecked(&offset); 5433 5434 // "main bin spec" (main binary specification) data payload is 5435 // formatted: 5436 // uint32_t version [currently 1] 5437 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user 5438 // process] uint64_t address [ UINT64_MAX if address not 5439 // specified ] uuid_t uuid [ all zero's if uuid not 5440 // specified ] uint32_t log2_pagesize [ process page size in log base 5441 // 2, e.g. 4k pages are 12. 0 for unspecified ] 5442 5443 if (strcmp("main bin spec", data_owner) == 0 && size >= 32) { 5444 offset = fileoff; 5445 uint32_t version; 5446 if (m_data.GetU32(&offset, &version, 1) != nullptr && version == 1) { 5447 uint32_t type = 0; 5448 uuid_t raw_uuid; 5449 memset(raw_uuid, 0, sizeof(uuid_t)); 5450 5451 if (m_data.GetU32(&offset, &type, 1) && 5452 m_data.GetU64(&offset, &address, 1) && 5453 m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5454 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5455 return true; 5456 } 5457 } 5458 } 5459 } 5460 offset = cmd_offset + lc.cmdsize; 5461 } 5462 } 5463 return false; 5464 } 5465 5466 lldb::RegisterContextSP 5467 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5468 lldb_private::Thread &thread) { 5469 lldb::RegisterContextSP reg_ctx_sp; 5470 5471 ModuleSP module_sp(GetModule()); 5472 if (module_sp) { 5473 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5474 if (!m_thread_context_offsets_valid) 5475 GetNumThreadContexts(); 5476 5477 const FileRangeArray::Entry *thread_context_file_range = 5478 m_thread_context_offsets.GetEntryAtIndex(idx); 5479 if (thread_context_file_range) { 5480 5481 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5482 thread_context_file_range->GetByteSize()); 5483 5484 switch (m_header.cputype) { 5485 case llvm::MachO::CPU_TYPE_ARM64: 5486 case llvm::MachO::CPU_TYPE_ARM64_32: 5487 reg_ctx_sp = 5488 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data); 5489 break; 5490 5491 case llvm::MachO::CPU_TYPE_ARM: 5492 reg_ctx_sp = 5493 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data); 5494 break; 5495 5496 case llvm::MachO::CPU_TYPE_I386: 5497 reg_ctx_sp = 5498 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data); 5499 break; 5500 5501 case llvm::MachO::CPU_TYPE_X86_64: 5502 reg_ctx_sp = 5503 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data); 5504 break; 5505 } 5506 } 5507 } 5508 return reg_ctx_sp; 5509 } 5510 5511 ObjectFile::Type ObjectFileMachO::CalculateType() { 5512 switch (m_header.filetype) { 5513 case MH_OBJECT: // 0x1u 5514 if (GetAddressByteSize() == 4) { 5515 // 32 bit kexts are just object files, but they do have a valid 5516 // UUID load command. 5517 if (GetUUID()) { 5518 // this checking for the UUID load command is not enough we could 5519 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5520 // this is required of kexts 5521 if (m_strata == eStrataInvalid) 5522 m_strata = eStrataKernel; 5523 return eTypeSharedLibrary; 5524 } 5525 } 5526 return eTypeObjectFile; 5527 5528 case MH_EXECUTE: 5529 return eTypeExecutable; // 0x2u 5530 case MH_FVMLIB: 5531 return eTypeSharedLibrary; // 0x3u 5532 case MH_CORE: 5533 return eTypeCoreFile; // 0x4u 5534 case MH_PRELOAD: 5535 return eTypeSharedLibrary; // 0x5u 5536 case MH_DYLIB: 5537 return eTypeSharedLibrary; // 0x6u 5538 case MH_DYLINKER: 5539 return eTypeDynamicLinker; // 0x7u 5540 case MH_BUNDLE: 5541 return eTypeSharedLibrary; // 0x8u 5542 case MH_DYLIB_STUB: 5543 return eTypeStubLibrary; // 0x9u 5544 case MH_DSYM: 5545 return eTypeDebugInfo; // 0xAu 5546 case MH_KEXT_BUNDLE: 5547 return eTypeSharedLibrary; // 0xBu 5548 default: 5549 break; 5550 } 5551 return eTypeUnknown; 5552 } 5553 5554 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5555 switch (m_header.filetype) { 5556 case MH_OBJECT: // 0x1u 5557 { 5558 // 32 bit kexts are just object files, but they do have a valid 5559 // UUID load command. 5560 if (GetUUID()) { 5561 // this checking for the UUID load command is not enough we could 5562 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5563 // this is required of kexts 5564 if (m_type == eTypeInvalid) 5565 m_type = eTypeSharedLibrary; 5566 5567 return eStrataKernel; 5568 } 5569 } 5570 return eStrataUnknown; 5571 5572 case MH_EXECUTE: // 0x2u 5573 // Check for the MH_DYLDLINK bit in the flags 5574 if (m_header.flags & MH_DYLDLINK) { 5575 return eStrataUser; 5576 } else { 5577 SectionList *section_list = GetSectionList(); 5578 if (section_list) { 5579 static ConstString g_kld_section_name("__KLD"); 5580 if (section_list->FindSectionByName(g_kld_section_name)) 5581 return eStrataKernel; 5582 } 5583 } 5584 return eStrataRawImage; 5585 5586 case MH_FVMLIB: 5587 return eStrataUser; // 0x3u 5588 case MH_CORE: 5589 return eStrataUnknown; // 0x4u 5590 case MH_PRELOAD: 5591 return eStrataRawImage; // 0x5u 5592 case MH_DYLIB: 5593 return eStrataUser; // 0x6u 5594 case MH_DYLINKER: 5595 return eStrataUser; // 0x7u 5596 case MH_BUNDLE: 5597 return eStrataUser; // 0x8u 5598 case MH_DYLIB_STUB: 5599 return eStrataUser; // 0x9u 5600 case MH_DSYM: 5601 return eStrataUnknown; // 0xAu 5602 case MH_KEXT_BUNDLE: 5603 return eStrataKernel; // 0xBu 5604 default: 5605 break; 5606 } 5607 return eStrataUnknown; 5608 } 5609 5610 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5611 ModuleSP module_sp(GetModule()); 5612 if (module_sp) { 5613 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5614 struct dylib_command load_cmd; 5615 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5616 uint32_t version_cmd = 0; 5617 uint64_t version = 0; 5618 uint32_t i; 5619 for (i = 0; i < m_header.ncmds; ++i) { 5620 const lldb::offset_t cmd_offset = offset; 5621 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr) 5622 break; 5623 5624 if (load_cmd.cmd == LC_ID_DYLIB) { 5625 if (version_cmd == 0) { 5626 version_cmd = load_cmd.cmd; 5627 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr) 5628 break; 5629 version = load_cmd.dylib.current_version; 5630 } 5631 break; // Break for now unless there is another more complete version 5632 // number load command in the future. 5633 } 5634 offset = cmd_offset + load_cmd.cmdsize; 5635 } 5636 5637 if (version_cmd == LC_ID_DYLIB) { 5638 unsigned major = (version & 0xFFFF0000ull) >> 16; 5639 unsigned minor = (version & 0x0000FF00ull) >> 8; 5640 unsigned subminor = (version & 0x000000FFull); 5641 return llvm::VersionTuple(major, minor, subminor); 5642 } 5643 } 5644 return llvm::VersionTuple(); 5645 } 5646 5647 ArchSpec ObjectFileMachO::GetArchitecture() { 5648 ModuleSP module_sp(GetModule()); 5649 ArchSpec arch; 5650 if (module_sp) { 5651 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5652 5653 return GetArchitecture(module_sp, m_header, m_data, 5654 MachHeaderSizeFromMagic(m_header.magic)); 5655 } 5656 return arch; 5657 } 5658 5659 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, 5660 addr_t &base_addr, UUID &uuid) { 5661 uuid.Clear(); 5662 base_addr = LLDB_INVALID_ADDRESS; 5663 if (process && process->GetDynamicLoader()) { 5664 DynamicLoader *dl = process->GetDynamicLoader(); 5665 LazyBool using_shared_cache; 5666 LazyBool private_shared_cache; 5667 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5668 private_shared_cache); 5669 } 5670 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5671 LIBLLDB_LOG_PROCESS)); 5672 LLDB_LOGF( 5673 log, 5674 "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64, 5675 uuid.GetAsString().c_str(), base_addr); 5676 } 5677 5678 // From dyld SPI header dyld_process_info.h 5679 typedef void *dyld_process_info; 5680 struct lldb_copy__dyld_process_cache_info { 5681 uuid_t cacheUUID; // UUID of cache used by process 5682 uint64_t cacheBaseAddress; // load address of dyld shared cache 5683 bool noCache; // process is running without a dyld cache 5684 bool privateCache; // process is using a private copy of its dyld cache 5685 }; 5686 5687 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with 5688 // llvm enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile 5689 // errors. So we need to use the actual underlying types of task_t and 5690 // kern_return_t below. 5691 extern "C" unsigned int /*task_t*/ mach_task_self(); 5692 5693 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5694 uuid.Clear(); 5695 base_addr = LLDB_INVALID_ADDRESS; 5696 5697 #if defined(__APPLE__) && \ 5698 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5699 uint8_t *(*dyld_get_all_image_infos)(void); 5700 dyld_get_all_image_infos = 5701 (uint8_t * (*)()) dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5702 if (dyld_get_all_image_infos) { 5703 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5704 if (dyld_all_image_infos_address) { 5705 uint32_t *version = (uint32_t *) 5706 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5707 if (*version >= 13) { 5708 uuid_t *sharedCacheUUID_address = 0; 5709 int wordsize = sizeof(uint8_t *); 5710 if (wordsize == 8) { 5711 sharedCacheUUID_address = 5712 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5713 160); // sharedCacheUUID <mach-o/dyld_images.h> 5714 if (*version >= 15) 5715 base_addr = 5716 *(uint64_t 5717 *)((uint8_t *)dyld_all_image_infos_address + 5718 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5719 } else { 5720 sharedCacheUUID_address = 5721 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5722 84); // sharedCacheUUID <mach-o/dyld_images.h> 5723 if (*version >= 15) { 5724 base_addr = 0; 5725 base_addr = 5726 *(uint32_t 5727 *)((uint8_t *)dyld_all_image_infos_address + 5728 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5729 } 5730 } 5731 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5732 } 5733 } 5734 } else { 5735 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5736 dyld_process_info (*dyld_process_info_create)( 5737 unsigned int /* task_t */ task, uint64_t timestamp, 5738 unsigned int /*kern_return_t*/ *kernelError); 5739 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5740 void (*dyld_process_info_release)(dyld_process_info info); 5741 5742 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, 5743 unsigned int /*kern_return_t*/ *)) 5744 dlsym(RTLD_DEFAULT, "_dyld_process_info_create"); 5745 dyld_process_info_get_cache = (void (*)(void *, void *))dlsym( 5746 RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5747 dyld_process_info_release = 5748 (void (*)(void *))dlsym(RTLD_DEFAULT, "_dyld_process_info_release"); 5749 5750 if (dyld_process_info_create && dyld_process_info_get_cache) { 5751 unsigned int /*kern_return_t */ kern_ret; 5752 dyld_process_info process_info = 5753 dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5754 if (process_info) { 5755 struct lldb_copy__dyld_process_cache_info sc_info; 5756 memset(&sc_info, 0, sizeof(struct lldb_copy__dyld_process_cache_info)); 5757 dyld_process_info_get_cache(process_info, &sc_info); 5758 if (sc_info.cacheBaseAddress != 0) { 5759 base_addr = sc_info.cacheBaseAddress; 5760 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 5761 } 5762 dyld_process_info_release(process_info); 5763 } 5764 } 5765 } 5766 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 5767 LIBLLDB_LOG_PROCESS)); 5768 if (log && uuid.IsValid()) 5769 LLDB_LOGF(log, 5770 "lldb's in-memory shared cache has a UUID of %s base address of " 5771 "0x%" PRIx64, 5772 uuid.GetAsString().c_str(), base_addr); 5773 #endif 5774 } 5775 5776 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5777 if (!m_min_os_version) { 5778 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5779 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5780 const lldb::offset_t load_cmd_offset = offset; 5781 5782 version_min_command lc; 5783 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5784 break; 5785 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5786 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5787 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5788 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5789 if (m_data.GetU32(&offset, &lc.version, 5790 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5791 const uint32_t xxxx = lc.version >> 16; 5792 const uint32_t yy = (lc.version >> 8) & 0xffu; 5793 const uint32_t zz = lc.version & 0xffu; 5794 if (xxxx) { 5795 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5796 break; 5797 } 5798 } 5799 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5800 // struct build_version_command { 5801 // uint32_t cmd; /* LC_BUILD_VERSION */ 5802 // uint32_t cmdsize; /* sizeof(struct 5803 // build_version_command) plus */ 5804 // /* ntools * sizeof(struct 5805 // build_tool_version) */ 5806 // uint32_t platform; /* platform */ 5807 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5808 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded in 5809 // nibbles xxxx.yy.zz */ uint32_t ntools; /* number of 5810 // tool entries following this */ 5811 // }; 5812 5813 offset += 4; // skip platform 5814 uint32_t minos = m_data.GetU32(&offset); 5815 5816 const uint32_t xxxx = minos >> 16; 5817 const uint32_t yy = (minos >> 8) & 0xffu; 5818 const uint32_t zz = minos & 0xffu; 5819 if (xxxx) { 5820 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5821 break; 5822 } 5823 } 5824 5825 offset = load_cmd_offset + lc.cmdsize; 5826 } 5827 5828 if (!m_min_os_version) { 5829 // Set version to an empty value so we don't keep trying to 5830 m_min_os_version = llvm::VersionTuple(); 5831 } 5832 } 5833 5834 return *m_min_os_version; 5835 } 5836 5837 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { 5838 if (!m_sdk_versions.hasValue()) { 5839 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5840 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5841 const lldb::offset_t load_cmd_offset = offset; 5842 5843 version_min_command lc; 5844 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5845 break; 5846 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5847 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5848 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5849 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5850 if (m_data.GetU32(&offset, &lc.version, 5851 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5852 const uint32_t xxxx = lc.sdk >> 16; 5853 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5854 const uint32_t zz = lc.sdk & 0xffu; 5855 if (xxxx) { 5856 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 5857 break; 5858 } else { 5859 GetModule()->ReportWarning("minimum OS version load command with " 5860 "invalid (0) version found."); 5861 } 5862 } 5863 } 5864 offset = load_cmd_offset + lc.cmdsize; 5865 } 5866 5867 if (!m_sdk_versions.hasValue()) { 5868 offset = MachHeaderSizeFromMagic(m_header.magic); 5869 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5870 const lldb::offset_t load_cmd_offset = offset; 5871 5872 version_min_command lc; 5873 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr) 5874 break; 5875 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5876 // struct build_version_command { 5877 // uint32_t cmd; /* LC_BUILD_VERSION */ 5878 // uint32_t cmdsize; /* sizeof(struct 5879 // build_version_command) plus */ 5880 // /* ntools * sizeof(struct 5881 // build_tool_version) */ 5882 // uint32_t platform; /* platform */ 5883 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5884 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded 5885 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number 5886 // of tool entries following this */ 5887 // }; 5888 5889 offset += 4; // skip platform 5890 uint32_t minos = m_data.GetU32(&offset); 5891 5892 const uint32_t xxxx = minos >> 16; 5893 const uint32_t yy = (minos >> 8) & 0xffu; 5894 const uint32_t zz = minos & 0xffu; 5895 if (xxxx) { 5896 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz); 5897 break; 5898 } 5899 } 5900 offset = load_cmd_offset + lc.cmdsize; 5901 } 5902 } 5903 5904 if (!m_sdk_versions.hasValue()) 5905 m_sdk_versions = llvm::VersionTuple(); 5906 } 5907 5908 return m_sdk_versions.getValue(); 5909 } 5910 5911 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5912 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5913 } 5914 5915 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5916 return m_allow_assembly_emulation_unwind_plans; 5917 } 5918 5919 // PluginInterface protocol 5920 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5921 return GetPluginNameStatic(); 5922 } 5923 5924 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5925 5926 Section *ObjectFileMachO::GetMachHeaderSection() { 5927 // Find the first address of the mach header which is the first non-zero file 5928 // sized section whose file offset is zero. This is the base file address of 5929 // the mach-o file which can be subtracted from the vmaddr of the other 5930 // segments found in memory and added to the load address 5931 ModuleSP module_sp = GetModule(); 5932 if (!module_sp) 5933 return nullptr; 5934 SectionList *section_list = GetSectionList(); 5935 if (!section_list) 5936 return nullptr; 5937 const size_t num_sections = section_list->GetSize(); 5938 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5939 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5940 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 5941 return section; 5942 } 5943 return nullptr; 5944 } 5945 5946 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 5947 if (!section) 5948 return false; 5949 const bool is_dsym = (m_header.filetype == MH_DSYM); 5950 if (section->GetFileSize() == 0 && !is_dsym) 5951 return false; 5952 if (section->IsThreadSpecific()) 5953 return false; 5954 if (GetModule().get() != section->GetModule().get()) 5955 return false; 5956 // Be careful with __LINKEDIT and __DWARF segments 5957 if (section->GetName() == GetSegmentNameLINKEDIT() || 5958 section->GetName() == GetSegmentNameDWARF()) { 5959 // Only map __LINKEDIT and __DWARF if we have an in memory image and 5960 // this isn't a kernel binary like a kext or mach_kernel. 5961 const bool is_memory_image = (bool)m_process_wp.lock(); 5962 const Strata strata = GetStrata(); 5963 if (is_memory_image == false || strata == eStrataKernel) 5964 return false; 5965 } 5966 return true; 5967 } 5968 5969 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 5970 lldb::addr_t header_load_address, const Section *header_section, 5971 const Section *section) { 5972 ModuleSP module_sp = GetModule(); 5973 if (module_sp && header_section && section && 5974 header_load_address != LLDB_INVALID_ADDRESS) { 5975 lldb::addr_t file_addr = header_section->GetFileAddress(); 5976 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 5977 return section->GetFileAddress() - file_addr + header_load_address; 5978 } 5979 return LLDB_INVALID_ADDRESS; 5980 } 5981 5982 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 5983 bool value_is_offset) { 5984 ModuleSP module_sp = GetModule(); 5985 if (!module_sp) 5986 return false; 5987 5988 SectionList *section_list = GetSectionList(); 5989 if (!section_list) 5990 return false; 5991 5992 size_t num_loaded_sections = 0; 5993 const size_t num_sections = section_list->GetSize(); 5994 5995 if (value_is_offset) { 5996 // "value" is an offset to apply to each top level segment 5997 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5998 // Iterate through the object file sections to find all of the 5999 // sections that size on disk (to avoid __PAGEZERO) and load them 6000 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6001 if (SectionIsLoadable(section_sp.get())) 6002 if (target.GetSectionLoadList().SetSectionLoadAddress( 6003 section_sp, section_sp->GetFileAddress() + value)) 6004 ++num_loaded_sections; 6005 } 6006 } else { 6007 // "value" is the new base address of the mach_header, adjust each 6008 // section accordingly 6009 6010 Section *mach_header_section = GetMachHeaderSection(); 6011 if (mach_header_section) { 6012 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6013 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6014 6015 lldb::addr_t section_load_addr = 6016 CalculateSectionLoadAddressForMemoryImage( 6017 value, mach_header_section, section_sp.get()); 6018 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6019 if (target.GetSectionLoadList().SetSectionLoadAddress( 6020 section_sp, section_load_addr)) 6021 ++num_loaded_sections; 6022 } 6023 } 6024 } 6025 } 6026 return num_loaded_sections > 0; 6027 } 6028 6029 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6030 const FileSpec &outfile, Status &error) { 6031 if (!process_sp) 6032 return false; 6033 6034 Target &target = process_sp->GetTarget(); 6035 const ArchSpec target_arch = target.GetArchitecture(); 6036 const llvm::Triple &target_triple = target_arch.GetTriple(); 6037 if (target_triple.getVendor() == llvm::Triple::Apple && 6038 (target_triple.getOS() == llvm::Triple::MacOSX || 6039 target_triple.getOS() == llvm::Triple::IOS || 6040 target_triple.getOS() == llvm::Triple::WatchOS || 6041 target_triple.getOS() == llvm::Triple::TvOS)) { 6042 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) 6043 // { 6044 bool make_core = false; 6045 switch (target_arch.GetMachine()) { 6046 case llvm::Triple::aarch64: 6047 case llvm::Triple::aarch64_32: 6048 case llvm::Triple::arm: 6049 case llvm::Triple::thumb: 6050 case llvm::Triple::x86: 6051 case llvm::Triple::x86_64: 6052 make_core = true; 6053 break; 6054 default: 6055 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6056 target_triple.str().c_str()); 6057 break; 6058 } 6059 6060 if (make_core) { 6061 std::vector<segment_command_64> segment_load_commands; 6062 // uint32_t range_info_idx = 0; 6063 MemoryRegionInfo range_info; 6064 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6065 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6066 const ByteOrder byte_order = target_arch.GetByteOrder(); 6067 if (range_error.Success()) { 6068 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6069 const addr_t addr = range_info.GetRange().GetRangeBase(); 6070 const addr_t size = range_info.GetRange().GetByteSize(); 6071 6072 if (size == 0) 6073 break; 6074 6075 // Calculate correct protections 6076 uint32_t prot = 0; 6077 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6078 prot |= VM_PROT_READ; 6079 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6080 prot |= VM_PROT_WRITE; 6081 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6082 prot |= VM_PROT_EXECUTE; 6083 6084 if (prot != 0) { 6085 uint32_t cmd_type = LC_SEGMENT_64; 6086 uint32_t segment_size = sizeof(segment_command_64); 6087 if (addr_byte_size == 4) { 6088 cmd_type = LC_SEGMENT; 6089 segment_size = sizeof(segment_command); 6090 } 6091 segment_command_64 segment = { 6092 cmd_type, // uint32_t cmd; 6093 segment_size, // uint32_t cmdsize; 6094 {0}, // char segname[16]; 6095 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6096 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6097 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6098 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6099 prot, // uint32_t maxprot; 6100 prot, // uint32_t initprot; 6101 0, // uint32_t nsects; 6102 0}; // uint32_t flags; 6103 segment_load_commands.push_back(segment); 6104 } else { 6105 // No protections and a size of 1 used to be returned from old 6106 // debugservers when we asked about a region that was past the 6107 // last memory region and it indicates the end... 6108 if (size == 1) 6109 break; 6110 } 6111 6112 range_error = process_sp->GetMemoryRegionInfo( 6113 range_info.GetRange().GetRangeEnd(), range_info); 6114 if (range_error.Fail()) 6115 break; 6116 } 6117 6118 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6119 6120 mach_header_64 mach_header; 6121 if (addr_byte_size == 8) { 6122 mach_header.magic = MH_MAGIC_64; 6123 } else { 6124 mach_header.magic = MH_MAGIC; 6125 } 6126 mach_header.cputype = target_arch.GetMachOCPUType(); 6127 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6128 mach_header.filetype = MH_CORE; 6129 mach_header.ncmds = segment_load_commands.size(); 6130 mach_header.flags = 0; 6131 mach_header.reserved = 0; 6132 ThreadList &thread_list = process_sp->GetThreadList(); 6133 const uint32_t num_threads = thread_list.GetSize(); 6134 6135 // Make an array of LC_THREAD data items. Each one contains the 6136 // contents of the LC_THREAD load command. The data doesn't contain 6137 // the load command + load command size, we will add the load command 6138 // and load command size as we emit the data. 6139 std::vector<StreamString> LC_THREAD_datas(num_threads); 6140 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6141 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6142 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6143 LC_THREAD_data.SetByteOrder(byte_order); 6144 } 6145 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) { 6146 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6147 if (thread_sp) { 6148 switch (mach_header.cputype) { 6149 case llvm::MachO::CPU_TYPE_ARM64: 6150 case llvm::MachO::CPU_TYPE_ARM64_32: 6151 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6152 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6153 break; 6154 6155 case llvm::MachO::CPU_TYPE_ARM: 6156 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6157 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6158 break; 6159 6160 case llvm::MachO::CPU_TYPE_I386: 6161 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6162 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6163 break; 6164 6165 case llvm::MachO::CPU_TYPE_X86_64: 6166 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6167 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6168 break; 6169 } 6170 } 6171 } 6172 6173 // The size of the load command is the size of the segments... 6174 if (addr_byte_size == 8) { 6175 mach_header.sizeofcmds = 6176 segment_load_commands.size() * sizeof(struct segment_command_64); 6177 } else { 6178 mach_header.sizeofcmds = 6179 segment_load_commands.size() * sizeof(struct segment_command); 6180 } 6181 6182 // and the size of all LC_THREAD load command 6183 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6184 ++mach_header.ncmds; 6185 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6186 } 6187 6188 // Write the mach header 6189 buffer.PutHex32(mach_header.magic); 6190 buffer.PutHex32(mach_header.cputype); 6191 buffer.PutHex32(mach_header.cpusubtype); 6192 buffer.PutHex32(mach_header.filetype); 6193 buffer.PutHex32(mach_header.ncmds); 6194 buffer.PutHex32(mach_header.sizeofcmds); 6195 buffer.PutHex32(mach_header.flags); 6196 if (addr_byte_size == 8) { 6197 buffer.PutHex32(mach_header.reserved); 6198 } 6199 6200 // Skip the mach header and all load commands and align to the next 6201 // 0x1000 byte boundary 6202 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6203 if (file_offset & 0x00000fff) { 6204 file_offset += 0x00001000ull; 6205 file_offset &= (~0x00001000ull + 1); 6206 } 6207 6208 for (auto &segment : segment_load_commands) { 6209 segment.fileoff = file_offset; 6210 file_offset += segment.filesize; 6211 } 6212 6213 // Write out all of the LC_THREAD load commands 6214 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6215 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6216 buffer.PutHex32(LC_THREAD); 6217 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6218 buffer.Write(LC_THREAD_data.GetString().data(), LC_THREAD_data_size); 6219 } 6220 6221 // Write out all of the segment load commands 6222 for (const auto &segment : segment_load_commands) { 6223 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6224 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6225 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6226 segment.cmd, segment.cmdsize, segment.vmaddr, 6227 segment.vmaddr + segment.vmsize, segment.fileoff, 6228 segment.filesize, segment.maxprot, segment.initprot, 6229 segment.nsects, segment.flags); 6230 6231 buffer.PutHex32(segment.cmd); 6232 buffer.PutHex32(segment.cmdsize); 6233 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6234 if (addr_byte_size == 8) { 6235 buffer.PutHex64(segment.vmaddr); 6236 buffer.PutHex64(segment.vmsize); 6237 buffer.PutHex64(segment.fileoff); 6238 buffer.PutHex64(segment.filesize); 6239 } else { 6240 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6241 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6242 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6243 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6244 } 6245 buffer.PutHex32(segment.maxprot); 6246 buffer.PutHex32(segment.initprot); 6247 buffer.PutHex32(segment.nsects); 6248 buffer.PutHex32(segment.flags); 6249 } 6250 6251 std::string core_file_path(outfile.GetPath()); 6252 auto core_file = FileSystem::Instance().Open( 6253 outfile, File::eOpenOptionWrite | File::eOpenOptionTruncate | 6254 File::eOpenOptionCanCreate); 6255 if (!core_file) { 6256 error = core_file.takeError(); 6257 } else { 6258 // Read 1 page at a time 6259 uint8_t bytes[0x1000]; 6260 // Write the mach header and load commands out to the core file 6261 size_t bytes_written = buffer.GetString().size(); 6262 error = 6263 core_file.get()->Write(buffer.GetString().data(), bytes_written); 6264 if (error.Success()) { 6265 // Now write the file data for all memory segments in the process 6266 for (const auto &segment : segment_load_commands) { 6267 if (core_file.get()->SeekFromStart(segment.fileoff) == -1) { 6268 error.SetErrorStringWithFormat( 6269 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6270 segment.fileoff, core_file_path.c_str()); 6271 break; 6272 } 6273 6274 printf("Saving %" PRId64 6275 " bytes of data for memory region at 0x%" PRIx64 "\n", 6276 segment.vmsize, segment.vmaddr); 6277 addr_t bytes_left = segment.vmsize; 6278 addr_t addr = segment.vmaddr; 6279 Status memory_read_error; 6280 while (bytes_left > 0 && error.Success()) { 6281 const size_t bytes_to_read = 6282 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6283 6284 // In a savecore setting, we don't really care about caching, 6285 // as the data is dumped and very likely never read again, 6286 // so we call ReadMemoryFromInferior to bypass it. 6287 const size_t bytes_read = process_sp->ReadMemoryFromInferior( 6288 addr, bytes, bytes_to_read, memory_read_error); 6289 6290 if (bytes_read == bytes_to_read) { 6291 size_t bytes_written = bytes_read; 6292 error = core_file.get()->Write(bytes, bytes_written); 6293 bytes_left -= bytes_read; 6294 addr += bytes_read; 6295 } else { 6296 // Some pages within regions are not readable, those should 6297 // be zero filled 6298 memset(bytes, 0, bytes_to_read); 6299 size_t bytes_written = bytes_to_read; 6300 error = core_file.get()->Write(bytes, bytes_written); 6301 bytes_left -= bytes_to_read; 6302 addr += bytes_to_read; 6303 } 6304 } 6305 } 6306 } 6307 } 6308 } else { 6309 error.SetErrorString( 6310 "process doesn't support getting memory region info"); 6311 } 6312 } 6313 return true; // This is the right plug to handle saving core files for 6314 // this process 6315 } 6316 return false; 6317 } 6318