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