1 //===-- ObjectFileMachO.cpp -------------------------------------*- C++ -*-===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 9 #include "llvm/ADT/StringRef.h" 10 11 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 12 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 13 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 14 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 15 #include "lldb/Core/Debugger.h" 16 #include "lldb/Core/FileSpecList.h" 17 #include "lldb/Core/Module.h" 18 #include "lldb/Core/ModuleSpec.h" 19 #include "lldb/Core/PluginManager.h" 20 #include "lldb/Core/Section.h" 21 #include "lldb/Core/StreamFile.h" 22 #include "lldb/Host/Host.h" 23 #include "lldb/Symbol/DWARFCallFrameInfo.h" 24 #include "lldb/Symbol/ObjectFile.h" 25 #include "lldb/Target/DynamicLoader.h" 26 #include "lldb/Target/MemoryRegionInfo.h" 27 #include "lldb/Target/Platform.h" 28 #include "lldb/Target/Process.h" 29 #include "lldb/Target/SectionLoadList.h" 30 #include "lldb/Target/Target.h" 31 #include "lldb/Target/Thread.h" 32 #include "lldb/Target/ThreadList.h" 33 #include "lldb/Utility/ArchSpec.h" 34 #include "lldb/Utility/DataBuffer.h" 35 #include "lldb/Utility/FileSpec.h" 36 #include "lldb/Utility/Log.h" 37 #include "lldb/Utility/RangeMap.h" 38 #include "lldb/Utility/RegisterValue.h" 39 #include "lldb/Utility/Status.h" 40 #include "lldb/Utility/StreamString.h" 41 #include "lldb/Utility/Timer.h" 42 #include "lldb/Utility/UUID.h" 43 44 #include "lldb/Host/SafeMachO.h" 45 46 #include "llvm/Support/MemoryBuffer.h" 47 48 #include "ObjectFileMachO.h" 49 50 #if defined(__APPLE__) && \ 51 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 52 // GetLLDBSharedCacheUUID() needs to call dlsym() 53 #include <dlfcn.h> 54 #endif 55 56 #ifndef __APPLE__ 57 #include "Utility/UuidCompatibility.h" 58 #else 59 #include <uuid/uuid.h> 60 #endif 61 62 #include <memory> 63 64 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 65 using namespace lldb; 66 using namespace lldb_private; 67 using namespace llvm::MachO; 68 69 // Some structure definitions needed for parsing the dyld shared cache files 70 // found on iOS devices. 71 72 struct lldb_copy_dyld_cache_header_v1 { 73 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 74 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 75 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 76 uint32_t imagesOffset; 77 uint32_t imagesCount; 78 uint64_t dyldBaseAddress; 79 uint64_t codeSignatureOffset; 80 uint64_t codeSignatureSize; 81 uint64_t slideInfoOffset; 82 uint64_t slideInfoSize; 83 uint64_t localSymbolsOffset; 84 uint64_t localSymbolsSize; 85 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 86 // and later 87 }; 88 89 struct lldb_copy_dyld_cache_mapping_info { 90 uint64_t address; 91 uint64_t size; 92 uint64_t fileOffset; 93 uint32_t maxProt; 94 uint32_t initProt; 95 }; 96 97 struct lldb_copy_dyld_cache_local_symbols_info { 98 uint32_t nlistOffset; 99 uint32_t nlistCount; 100 uint32_t stringsOffset; 101 uint32_t stringsSize; 102 uint32_t entriesOffset; 103 uint32_t entriesCount; 104 }; 105 struct lldb_copy_dyld_cache_local_symbols_entry { 106 uint32_t dylibOffset; 107 uint32_t nlistStartIndex; 108 uint32_t nlistCount; 109 }; 110 111 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 112 public: 113 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 114 const DataExtractor &data) 115 : RegisterContextDarwin_x86_64(thread, 0) { 116 SetRegisterDataFrom_LC_THREAD(data); 117 } 118 119 void InvalidateAllRegisters() override { 120 // Do nothing... registers are always valid... 121 } 122 123 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 124 lldb::offset_t offset = 0; 125 SetError(GPRRegSet, Read, -1); 126 SetError(FPURegSet, Read, -1); 127 SetError(EXCRegSet, Read, -1); 128 bool done = false; 129 130 while (!done) { 131 int flavor = data.GetU32(&offset); 132 if (flavor == 0) 133 done = true; 134 else { 135 uint32_t i; 136 uint32_t count = data.GetU32(&offset); 137 switch (flavor) { 138 case GPRRegSet: 139 for (i = 0; i < count; ++i) 140 (&gpr.rax)[i] = data.GetU64(&offset); 141 SetError(GPRRegSet, Read, 0); 142 done = true; 143 144 break; 145 case FPURegSet: 146 // TODO: fill in FPU regs.... 147 // SetError (FPURegSet, Read, -1); 148 done = true; 149 150 break; 151 case EXCRegSet: 152 exc.trapno = data.GetU32(&offset); 153 exc.err = data.GetU32(&offset); 154 exc.faultvaddr = data.GetU64(&offset); 155 SetError(EXCRegSet, Read, 0); 156 done = true; 157 break; 158 case 7: 159 case 8: 160 case 9: 161 // fancy flavors that encapsulate of the above flavors... 162 break; 163 164 default: 165 done = true; 166 break; 167 } 168 } 169 } 170 } 171 172 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 173 const char *alt_name, size_t reg_byte_size, 174 Stream &data) { 175 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 176 if (reg_info == NULL) 177 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 178 if (reg_info) { 179 lldb_private::RegisterValue reg_value; 180 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 181 if (reg_info->byte_size >= reg_byte_size) 182 data.Write(reg_value.GetBytes(), reg_byte_size); 183 else { 184 data.Write(reg_value.GetBytes(), reg_info->byte_size); 185 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 186 ++i) 187 data.PutChar(0); 188 } 189 return reg_byte_size; 190 } 191 } 192 // Just write zeros if all else fails 193 for (size_t i = 0; i < reg_byte_size; ++i) 194 data.PutChar(0); 195 return reg_byte_size; 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 WriteRegister(reg_ctx, "rax", NULL, 8, data); 206 WriteRegister(reg_ctx, "rbx", NULL, 8, data); 207 WriteRegister(reg_ctx, "rcx", NULL, 8, data); 208 WriteRegister(reg_ctx, "rdx", NULL, 8, data); 209 WriteRegister(reg_ctx, "rdi", NULL, 8, data); 210 WriteRegister(reg_ctx, "rsi", NULL, 8, data); 211 WriteRegister(reg_ctx, "rbp", NULL, 8, data); 212 WriteRegister(reg_ctx, "rsp", NULL, 8, data); 213 WriteRegister(reg_ctx, "r8", NULL, 8, data); 214 WriteRegister(reg_ctx, "r9", NULL, 8, data); 215 WriteRegister(reg_ctx, "r10", NULL, 8, data); 216 WriteRegister(reg_ctx, "r11", NULL, 8, data); 217 WriteRegister(reg_ctx, "r12", NULL, 8, data); 218 WriteRegister(reg_ctx, "r13", NULL, 8, data); 219 WriteRegister(reg_ctx, "r14", NULL, 8, data); 220 WriteRegister(reg_ctx, "r15", NULL, 8, data); 221 WriteRegister(reg_ctx, "rip", NULL, 8, data); 222 WriteRegister(reg_ctx, "rflags", NULL, 8, data); 223 WriteRegister(reg_ctx, "cs", NULL, 8, data); 224 WriteRegister(reg_ctx, "fs", NULL, 8, data); 225 WriteRegister(reg_ctx, "gs", NULL, 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 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 315 WriteRegister(reg_ctx, "err", NULL, 4, data); 316 WriteRegister(reg_ctx, "faultvaddr", NULL, 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 size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 404 const char *alt_name, size_t reg_byte_size, 405 Stream &data) { 406 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 407 if (reg_info == NULL) 408 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 409 if (reg_info) { 410 lldb_private::RegisterValue reg_value; 411 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 412 if (reg_info->byte_size >= reg_byte_size) 413 data.Write(reg_value.GetBytes(), reg_byte_size); 414 else { 415 data.Write(reg_value.GetBytes(), reg_info->byte_size); 416 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 417 ++i) 418 data.PutChar(0); 419 } 420 return reg_byte_size; 421 } 422 } 423 // Just write zeros if all else fails 424 for (size_t i = 0; i < reg_byte_size; ++i) 425 data.PutChar(0); 426 return reg_byte_size; 427 } 428 429 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 430 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 431 if (reg_ctx_sp) { 432 RegisterContext *reg_ctx = reg_ctx_sp.get(); 433 434 data.PutHex32(GPRRegSet); // Flavor 435 data.PutHex32(GPRWordCount); 436 WriteRegister(reg_ctx, "eax", NULL, 4, data); 437 WriteRegister(reg_ctx, "ebx", NULL, 4, data); 438 WriteRegister(reg_ctx, "ecx", NULL, 4, data); 439 WriteRegister(reg_ctx, "edx", NULL, 4, data); 440 WriteRegister(reg_ctx, "edi", NULL, 4, data); 441 WriteRegister(reg_ctx, "esi", NULL, 4, data); 442 WriteRegister(reg_ctx, "ebp", NULL, 4, data); 443 WriteRegister(reg_ctx, "esp", NULL, 4, data); 444 WriteRegister(reg_ctx, "ss", NULL, 4, data); 445 WriteRegister(reg_ctx, "eflags", NULL, 4, data); 446 WriteRegister(reg_ctx, "eip", NULL, 4, data); 447 WriteRegister(reg_ctx, "cs", NULL, 4, data); 448 WriteRegister(reg_ctx, "ds", NULL, 4, data); 449 WriteRegister(reg_ctx, "es", NULL, 4, data); 450 WriteRegister(reg_ctx, "fs", NULL, 4, data); 451 WriteRegister(reg_ctx, "gs", NULL, 4, data); 452 453 // Write out the EXC registers 454 data.PutHex32(EXCRegSet); 455 data.PutHex32(EXCWordCount); 456 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 457 WriteRegister(reg_ctx, "err", NULL, 4, data); 458 WriteRegister(reg_ctx, "faultvaddr", NULL, 4, data); 459 return true; 460 } 461 return false; 462 } 463 464 protected: 465 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 466 467 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 468 469 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 470 471 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 472 return 0; 473 } 474 475 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 476 return 0; 477 } 478 479 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 480 return 0; 481 } 482 }; 483 484 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 485 public: 486 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 487 const DataExtractor &data) 488 : RegisterContextDarwin_arm(thread, 0) { 489 SetRegisterDataFrom_LC_THREAD(data); 490 } 491 492 void InvalidateAllRegisters() override { 493 // Do nothing... registers are always valid... 494 } 495 496 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 497 lldb::offset_t offset = 0; 498 SetError(GPRRegSet, Read, -1); 499 SetError(FPURegSet, Read, -1); 500 SetError(EXCRegSet, Read, -1); 501 bool done = false; 502 503 while (!done) { 504 int flavor = data.GetU32(&offset); 505 uint32_t count = data.GetU32(&offset); 506 lldb::offset_t next_thread_state = offset + (count * 4); 507 switch (flavor) { 508 case GPRAltRegSet: 509 case GPRRegSet: 510 for (uint32_t i = 0; i < count; ++i) { 511 gpr.r[i] = data.GetU32(&offset); 512 } 513 514 // Note that gpr.cpsr is also copied by the above loop; this loop 515 // technically extends one element past the end of the gpr.r[] array. 516 517 SetError(GPRRegSet, Read, 0); 518 offset = next_thread_state; 519 break; 520 521 case FPURegSet: { 522 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 523 const int fpu_reg_buf_size = sizeof(fpu.floats); 524 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 525 fpu_reg_buf) == fpu_reg_buf_size) { 526 offset += fpu_reg_buf_size; 527 fpu.fpscr = data.GetU32(&offset); 528 SetError(FPURegSet, Read, 0); 529 } else { 530 done = true; 531 } 532 } 533 offset = next_thread_state; 534 break; 535 536 case EXCRegSet: 537 if (count == 3) { 538 exc.exception = data.GetU32(&offset); 539 exc.fsr = data.GetU32(&offset); 540 exc.far = data.GetU32(&offset); 541 SetError(EXCRegSet, Read, 0); 542 } 543 done = true; 544 offset = next_thread_state; 545 break; 546 547 // Unknown register set flavor, stop trying to parse. 548 default: 549 done = true; 550 } 551 } 552 } 553 554 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 555 const char *alt_name, size_t reg_byte_size, 556 Stream &data) { 557 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 558 if (reg_info == NULL) 559 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 560 if (reg_info) { 561 lldb_private::RegisterValue reg_value; 562 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 563 if (reg_info->byte_size >= reg_byte_size) 564 data.Write(reg_value.GetBytes(), reg_byte_size); 565 else { 566 data.Write(reg_value.GetBytes(), reg_info->byte_size); 567 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 568 ++i) 569 data.PutChar(0); 570 } 571 return reg_byte_size; 572 } 573 } 574 // Just write zeros if all else fails 575 for (size_t i = 0; i < reg_byte_size; ++i) 576 data.PutChar(0); 577 return reg_byte_size; 578 } 579 580 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 581 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 582 if (reg_ctx_sp) { 583 RegisterContext *reg_ctx = reg_ctx_sp.get(); 584 585 data.PutHex32(GPRRegSet); // Flavor 586 data.PutHex32(GPRWordCount); 587 WriteRegister(reg_ctx, "r0", NULL, 4, data); 588 WriteRegister(reg_ctx, "r1", NULL, 4, data); 589 WriteRegister(reg_ctx, "r2", NULL, 4, data); 590 WriteRegister(reg_ctx, "r3", NULL, 4, data); 591 WriteRegister(reg_ctx, "r4", NULL, 4, data); 592 WriteRegister(reg_ctx, "r5", NULL, 4, data); 593 WriteRegister(reg_ctx, "r6", NULL, 4, data); 594 WriteRegister(reg_ctx, "r7", NULL, 4, data); 595 WriteRegister(reg_ctx, "r8", NULL, 4, data); 596 WriteRegister(reg_ctx, "r9", NULL, 4, data); 597 WriteRegister(reg_ctx, "r10", NULL, 4, data); 598 WriteRegister(reg_ctx, "r11", NULL, 4, data); 599 WriteRegister(reg_ctx, "r12", NULL, 4, data); 600 WriteRegister(reg_ctx, "sp", NULL, 4, data); 601 WriteRegister(reg_ctx, "lr", NULL, 4, data); 602 WriteRegister(reg_ctx, "pc", NULL, 4, data); 603 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 604 605 // Write out the EXC registers 606 // data.PutHex32 (EXCRegSet); 607 // data.PutHex32 (EXCWordCount); 608 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 609 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 610 // WriteRegister (reg_ctx, "far", NULL, 4, data); 611 return true; 612 } 613 return false; 614 } 615 616 protected: 617 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 618 619 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 620 621 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 622 623 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 624 625 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 626 return 0; 627 } 628 629 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 630 return 0; 631 } 632 633 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 634 return 0; 635 } 636 637 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 638 return -1; 639 } 640 }; 641 642 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 643 public: 644 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 645 const DataExtractor &data) 646 : RegisterContextDarwin_arm64(thread, 0) { 647 SetRegisterDataFrom_LC_THREAD(data); 648 } 649 650 void InvalidateAllRegisters() override { 651 // Do nothing... registers are always valid... 652 } 653 654 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 655 lldb::offset_t offset = 0; 656 SetError(GPRRegSet, Read, -1); 657 SetError(FPURegSet, Read, -1); 658 SetError(EXCRegSet, Read, -1); 659 bool done = false; 660 while (!done) { 661 int flavor = data.GetU32(&offset); 662 uint32_t count = data.GetU32(&offset); 663 lldb::offset_t next_thread_state = offset + (count * 4); 664 switch (flavor) { 665 case GPRRegSet: 666 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 667 // 32-bit register) 668 if (count >= (33 * 2) + 1) { 669 for (uint32_t i = 0; i < 29; ++i) 670 gpr.x[i] = data.GetU64(&offset); 671 gpr.fp = data.GetU64(&offset); 672 gpr.lr = data.GetU64(&offset); 673 gpr.sp = data.GetU64(&offset); 674 gpr.pc = data.GetU64(&offset); 675 gpr.cpsr = data.GetU32(&offset); 676 SetError(GPRRegSet, Read, 0); 677 } 678 offset = next_thread_state; 679 break; 680 case FPURegSet: { 681 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 682 const int fpu_reg_buf_size = sizeof(fpu); 683 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 684 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 685 fpu_reg_buf) == fpu_reg_buf_size) { 686 SetError(FPURegSet, Read, 0); 687 } else { 688 done = true; 689 } 690 } 691 offset = next_thread_state; 692 break; 693 case EXCRegSet: 694 if (count == 4) { 695 exc.far = data.GetU64(&offset); 696 exc.esr = data.GetU32(&offset); 697 exc.exception = data.GetU32(&offset); 698 SetError(EXCRegSet, Read, 0); 699 } 700 offset = next_thread_state; 701 break; 702 default: 703 done = true; 704 break; 705 } 706 } 707 } 708 709 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 710 const char *alt_name, size_t reg_byte_size, 711 Stream &data) { 712 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 713 if (reg_info == NULL) 714 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 715 if (reg_info) { 716 lldb_private::RegisterValue reg_value; 717 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 718 if (reg_info->byte_size >= reg_byte_size) 719 data.Write(reg_value.GetBytes(), reg_byte_size); 720 else { 721 data.Write(reg_value.GetBytes(), reg_info->byte_size); 722 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 723 ++i) 724 data.PutChar(0); 725 } 726 return reg_byte_size; 727 } 728 } 729 // Just write zeros if all else fails 730 for (size_t i = 0; i < reg_byte_size; ++i) 731 data.PutChar(0); 732 return reg_byte_size; 733 } 734 735 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 736 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 737 if (reg_ctx_sp) { 738 RegisterContext *reg_ctx = reg_ctx_sp.get(); 739 740 data.PutHex32(GPRRegSet); // Flavor 741 data.PutHex32(GPRWordCount); 742 WriteRegister(reg_ctx, "x0", NULL, 8, data); 743 WriteRegister(reg_ctx, "x1", NULL, 8, data); 744 WriteRegister(reg_ctx, "x2", NULL, 8, data); 745 WriteRegister(reg_ctx, "x3", NULL, 8, data); 746 WriteRegister(reg_ctx, "x4", NULL, 8, data); 747 WriteRegister(reg_ctx, "x5", NULL, 8, data); 748 WriteRegister(reg_ctx, "x6", NULL, 8, data); 749 WriteRegister(reg_ctx, "x7", NULL, 8, data); 750 WriteRegister(reg_ctx, "x8", NULL, 8, data); 751 WriteRegister(reg_ctx, "x9", NULL, 8, data); 752 WriteRegister(reg_ctx, "x10", NULL, 8, data); 753 WriteRegister(reg_ctx, "x11", NULL, 8, data); 754 WriteRegister(reg_ctx, "x12", NULL, 8, data); 755 WriteRegister(reg_ctx, "x13", NULL, 8, data); 756 WriteRegister(reg_ctx, "x14", NULL, 8, data); 757 WriteRegister(reg_ctx, "x15", NULL, 8, data); 758 WriteRegister(reg_ctx, "x16", NULL, 8, data); 759 WriteRegister(reg_ctx, "x17", NULL, 8, data); 760 WriteRegister(reg_ctx, "x18", NULL, 8, data); 761 WriteRegister(reg_ctx, "x19", NULL, 8, data); 762 WriteRegister(reg_ctx, "x20", NULL, 8, data); 763 WriteRegister(reg_ctx, "x21", NULL, 8, data); 764 WriteRegister(reg_ctx, "x22", NULL, 8, data); 765 WriteRegister(reg_ctx, "x23", NULL, 8, data); 766 WriteRegister(reg_ctx, "x24", NULL, 8, data); 767 WriteRegister(reg_ctx, "x25", NULL, 8, data); 768 WriteRegister(reg_ctx, "x26", NULL, 8, data); 769 WriteRegister(reg_ctx, "x27", NULL, 8, data); 770 WriteRegister(reg_ctx, "x28", NULL, 8, data); 771 WriteRegister(reg_ctx, "fp", NULL, 8, data); 772 WriteRegister(reg_ctx, "lr", NULL, 8, data); 773 WriteRegister(reg_ctx, "sp", NULL, 8, data); 774 WriteRegister(reg_ctx, "pc", NULL, 8, data); 775 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 776 777 // Write out the EXC registers 778 // data.PutHex32 (EXCRegSet); 779 // data.PutHex32 (EXCWordCount); 780 // WriteRegister (reg_ctx, "far", NULL, 8, data); 781 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 782 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 783 return true; 784 } 785 return false; 786 } 787 788 protected: 789 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 790 791 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 792 793 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 794 795 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 796 797 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 798 return 0; 799 } 800 801 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 802 return 0; 803 } 804 805 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 806 return 0; 807 } 808 809 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 810 return -1; 811 } 812 }; 813 814 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 815 switch (magic) { 816 case MH_MAGIC: 817 case MH_CIGAM: 818 return sizeof(struct mach_header); 819 820 case MH_MAGIC_64: 821 case MH_CIGAM_64: 822 return sizeof(struct mach_header_64); 823 break; 824 825 default: 826 break; 827 } 828 return 0; 829 } 830 831 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 832 833 void ObjectFileMachO::Initialize() { 834 PluginManager::RegisterPlugin( 835 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 836 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 837 } 838 839 void ObjectFileMachO::Terminate() { 840 PluginManager::UnregisterPlugin(CreateInstance); 841 } 842 843 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 844 static ConstString g_name("mach-o"); 845 return g_name; 846 } 847 848 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 849 return "Mach-o object file reader (32 and 64 bit)"; 850 } 851 852 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 853 DataBufferSP &data_sp, 854 lldb::offset_t data_offset, 855 const FileSpec *file, 856 lldb::offset_t file_offset, 857 lldb::offset_t length) { 858 if (!data_sp) { 859 data_sp = MapFileData(*file, length, file_offset); 860 if (!data_sp) 861 return nullptr; 862 data_offset = 0; 863 } 864 865 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 866 return nullptr; 867 868 // Update the data to contain the entire file if it doesn't already 869 if (data_sp->GetByteSize() < length) { 870 data_sp = MapFileData(*file, length, file_offset); 871 if (!data_sp) 872 return nullptr; 873 data_offset = 0; 874 } 875 auto objfile_up = llvm::make_unique<ObjectFileMachO>( 876 module_sp, data_sp, data_offset, file, file_offset, length); 877 if (!objfile_up || !objfile_up->ParseHeader()) 878 return nullptr; 879 880 return objfile_up.release(); 881 } 882 883 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 884 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 885 const ProcessSP &process_sp, lldb::addr_t header_addr) { 886 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 887 std::unique_ptr<ObjectFile> objfile_up( 888 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 889 if (objfile_up.get() && objfile_up->ParseHeader()) 890 return objfile_up.release(); 891 } 892 return NULL; 893 } 894 895 size_t ObjectFileMachO::GetModuleSpecifications( 896 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 897 lldb::offset_t data_offset, lldb::offset_t file_offset, 898 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 899 const size_t initial_count = specs.GetSize(); 900 901 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 902 DataExtractor data; 903 data.SetData(data_sp); 904 llvm::MachO::mach_header header; 905 if (ParseHeader(data, &data_offset, header)) { 906 size_t header_and_load_cmds = 907 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 908 if (header_and_load_cmds >= data_sp->GetByteSize()) { 909 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 910 data.SetData(data_sp); 911 data_offset = MachHeaderSizeFromMagic(header.magic); 912 } 913 if (data_sp) { 914 ModuleSpec spec; 915 spec.GetFileSpec() = file; 916 spec.SetObjectOffset(file_offset); 917 spec.SetObjectSize(length); 918 919 spec.GetArchitecture() = GetArchitecture(header, data, data_offset); 920 if (spec.GetArchitecture().IsValid()) { 921 spec.GetUUID() = GetUUID(header, data, data_offset); 922 specs.Append(spec); 923 } 924 } 925 } 926 } 927 return specs.GetSize() - initial_count; 928 } 929 930 ConstString ObjectFileMachO::GetSegmentNameTEXT() { 931 static ConstString g_segment_name_TEXT("__TEXT"); 932 return g_segment_name_TEXT; 933 } 934 935 ConstString ObjectFileMachO::GetSegmentNameDATA() { 936 static ConstString g_segment_name_DATA("__DATA"); 937 return g_segment_name_DATA; 938 } 939 940 ConstString ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 941 static ConstString g_segment_name("__DATA_DIRTY"); 942 return g_segment_name; 943 } 944 945 ConstString ObjectFileMachO::GetSegmentNameDATA_CONST() { 946 static ConstString g_segment_name("__DATA_CONST"); 947 return g_segment_name; 948 } 949 950 ConstString ObjectFileMachO::GetSegmentNameOBJC() { 951 static ConstString g_segment_name_OBJC("__OBJC"); 952 return g_segment_name_OBJC; 953 } 954 955 ConstString ObjectFileMachO::GetSegmentNameLINKEDIT() { 956 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 957 return g_section_name_LINKEDIT; 958 } 959 960 ConstString ObjectFileMachO::GetSegmentNameDWARF() { 961 static ConstString g_section_name("__DWARF"); 962 return g_section_name; 963 } 964 965 ConstString ObjectFileMachO::GetSectionNameEHFrame() { 966 static ConstString g_section_name_eh_frame("__eh_frame"); 967 return g_section_name_eh_frame; 968 } 969 970 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 971 lldb::addr_t data_offset, 972 lldb::addr_t data_length) { 973 DataExtractor data; 974 data.SetData(data_sp, data_offset, data_length); 975 lldb::offset_t offset = 0; 976 uint32_t magic = data.GetU32(&offset); 977 return MachHeaderSizeFromMagic(magic) != 0; 978 } 979 980 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 981 DataBufferSP &data_sp, 982 lldb::offset_t data_offset, 983 const FileSpec *file, 984 lldb::offset_t file_offset, 985 lldb::offset_t length) 986 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 987 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 988 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 989 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 990 ::memset(&m_header, 0, sizeof(m_header)); 991 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 992 } 993 994 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 995 lldb::DataBufferSP &header_data_sp, 996 const lldb::ProcessSP &process_sp, 997 lldb::addr_t header_addr) 998 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 999 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 1000 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 1001 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 1002 ::memset(&m_header, 0, sizeof(m_header)); 1003 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1004 } 1005 1006 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 1007 lldb::offset_t *data_offset_ptr, 1008 llvm::MachO::mach_header &header) { 1009 data.SetByteOrder(endian::InlHostByteOrder()); 1010 // Leave magic in the original byte order 1011 header.magic = data.GetU32(data_offset_ptr); 1012 bool can_parse = false; 1013 bool is_64_bit = false; 1014 switch (header.magic) { 1015 case MH_MAGIC: 1016 data.SetByteOrder(endian::InlHostByteOrder()); 1017 data.SetAddressByteSize(4); 1018 can_parse = true; 1019 break; 1020 1021 case MH_MAGIC_64: 1022 data.SetByteOrder(endian::InlHostByteOrder()); 1023 data.SetAddressByteSize(8); 1024 can_parse = true; 1025 is_64_bit = true; 1026 break; 1027 1028 case MH_CIGAM: 1029 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1030 ? eByteOrderLittle 1031 : eByteOrderBig); 1032 data.SetAddressByteSize(4); 1033 can_parse = true; 1034 break; 1035 1036 case MH_CIGAM_64: 1037 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1038 ? eByteOrderLittle 1039 : eByteOrderBig); 1040 data.SetAddressByteSize(8); 1041 is_64_bit = true; 1042 can_parse = true; 1043 break; 1044 1045 default: 1046 break; 1047 } 1048 1049 if (can_parse) { 1050 data.GetU32(data_offset_ptr, &header.cputype, 6); 1051 if (is_64_bit) 1052 *data_offset_ptr += 4; 1053 return true; 1054 } else { 1055 memset(&header, 0, sizeof(header)); 1056 } 1057 return false; 1058 } 1059 1060 bool ObjectFileMachO::ParseHeader() { 1061 ModuleSP module_sp(GetModule()); 1062 if (module_sp) { 1063 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1064 bool can_parse = false; 1065 lldb::offset_t offset = 0; 1066 m_data.SetByteOrder(endian::InlHostByteOrder()); 1067 // Leave magic in the original byte order 1068 m_header.magic = m_data.GetU32(&offset); 1069 switch (m_header.magic) { 1070 case MH_MAGIC: 1071 m_data.SetByteOrder(endian::InlHostByteOrder()); 1072 m_data.SetAddressByteSize(4); 1073 can_parse = true; 1074 break; 1075 1076 case MH_MAGIC_64: 1077 m_data.SetByteOrder(endian::InlHostByteOrder()); 1078 m_data.SetAddressByteSize(8); 1079 can_parse = true; 1080 break; 1081 1082 case MH_CIGAM: 1083 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1084 ? eByteOrderLittle 1085 : eByteOrderBig); 1086 m_data.SetAddressByteSize(4); 1087 can_parse = true; 1088 break; 1089 1090 case MH_CIGAM_64: 1091 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1092 ? eByteOrderLittle 1093 : eByteOrderBig); 1094 m_data.SetAddressByteSize(8); 1095 can_parse = true; 1096 break; 1097 1098 default: 1099 break; 1100 } 1101 1102 if (can_parse) { 1103 m_data.GetU32(&offset, &m_header.cputype, 6); 1104 1105 if (ArchSpec mach_arch = GetArchitecture()) { 1106 // Check if the module has a required architecture 1107 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1108 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1109 return false; 1110 1111 if (SetModulesArchitecture(mach_arch)) { 1112 const size_t header_and_lc_size = 1113 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1114 if (m_data.GetByteSize() < header_and_lc_size) { 1115 DataBufferSP data_sp; 1116 ProcessSP process_sp(m_process_wp.lock()); 1117 if (process_sp) { 1118 data_sp = 1119 ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1120 } else { 1121 // Read in all only the load command data from the file on disk 1122 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset); 1123 if (data_sp->GetByteSize() != header_and_lc_size) 1124 return false; 1125 } 1126 if (data_sp) 1127 m_data.SetData(data_sp); 1128 } 1129 } 1130 return true; 1131 } 1132 } else { 1133 memset(&m_header, 0, sizeof(struct mach_header)); 1134 } 1135 } 1136 return false; 1137 } 1138 1139 ByteOrder ObjectFileMachO::GetByteOrder() const { 1140 return m_data.GetByteOrder(); 1141 } 1142 1143 bool ObjectFileMachO::IsExecutable() const { 1144 return m_header.filetype == MH_EXECUTE; 1145 } 1146 1147 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1148 return m_data.GetAddressByteSize(); 1149 } 1150 1151 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1152 Symtab *symtab = GetSymtab(); 1153 if (symtab) { 1154 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1155 if (symbol) { 1156 if (symbol->ValueIsAddress()) { 1157 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1158 if (section_sp) { 1159 const lldb::SectionType section_type = section_sp->GetType(); 1160 switch (section_type) { 1161 case eSectionTypeInvalid: 1162 return AddressClass::eUnknown; 1163 1164 case eSectionTypeCode: 1165 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1166 // For ARM we have a bit in the n_desc field of the symbol that 1167 // tells us ARM/Thumb which is bit 0x0008. 1168 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1169 return AddressClass::eCodeAlternateISA; 1170 } 1171 return AddressClass::eCode; 1172 1173 case eSectionTypeContainer: 1174 return AddressClass::eUnknown; 1175 1176 case eSectionTypeData: 1177 case eSectionTypeDataCString: 1178 case eSectionTypeDataCStringPointers: 1179 case eSectionTypeDataSymbolAddress: 1180 case eSectionTypeData4: 1181 case eSectionTypeData8: 1182 case eSectionTypeData16: 1183 case eSectionTypeDataPointers: 1184 case eSectionTypeZeroFill: 1185 case eSectionTypeDataObjCMessageRefs: 1186 case eSectionTypeDataObjCCFStrings: 1187 case eSectionTypeGoSymtab: 1188 return AddressClass::eData; 1189 1190 case eSectionTypeDebug: 1191 case eSectionTypeDWARFDebugAbbrev: 1192 case eSectionTypeDWARFDebugAbbrevDwo: 1193 case eSectionTypeDWARFDebugAddr: 1194 case eSectionTypeDWARFDebugAranges: 1195 case eSectionTypeDWARFDebugCuIndex: 1196 case eSectionTypeDWARFDebugFrame: 1197 case eSectionTypeDWARFDebugInfo: 1198 case eSectionTypeDWARFDebugInfoDwo: 1199 case eSectionTypeDWARFDebugLine: 1200 case eSectionTypeDWARFDebugLineStr: 1201 case eSectionTypeDWARFDebugLoc: 1202 case eSectionTypeDWARFDebugLocLists: 1203 case eSectionTypeDWARFDebugMacInfo: 1204 case eSectionTypeDWARFDebugMacro: 1205 case eSectionTypeDWARFDebugNames: 1206 case eSectionTypeDWARFDebugPubNames: 1207 case eSectionTypeDWARFDebugPubTypes: 1208 case eSectionTypeDWARFDebugRanges: 1209 case eSectionTypeDWARFDebugRngLists: 1210 case eSectionTypeDWARFDebugStr: 1211 case eSectionTypeDWARFDebugStrDwo: 1212 case eSectionTypeDWARFDebugStrOffsets: 1213 case eSectionTypeDWARFDebugStrOffsetsDwo: 1214 case eSectionTypeDWARFDebugTypes: 1215 case eSectionTypeDWARFAppleNames: 1216 case eSectionTypeDWARFAppleTypes: 1217 case eSectionTypeDWARFAppleNamespaces: 1218 case eSectionTypeDWARFAppleObjC: 1219 case eSectionTypeDWARFGNUDebugAltLink: 1220 return AddressClass::eDebug; 1221 1222 case eSectionTypeEHFrame: 1223 case eSectionTypeARMexidx: 1224 case eSectionTypeARMextab: 1225 case eSectionTypeCompactUnwind: 1226 return AddressClass::eRuntime; 1227 1228 case eSectionTypeAbsoluteAddress: 1229 case eSectionTypeELFSymbolTable: 1230 case eSectionTypeELFDynamicSymbols: 1231 case eSectionTypeELFRelocationEntries: 1232 case eSectionTypeELFDynamicLinkInfo: 1233 case eSectionTypeOther: 1234 return AddressClass::eUnknown; 1235 } 1236 } 1237 } 1238 1239 const SymbolType symbol_type = symbol->GetType(); 1240 switch (symbol_type) { 1241 case eSymbolTypeAny: 1242 return AddressClass::eUnknown; 1243 case eSymbolTypeAbsolute: 1244 return AddressClass::eUnknown; 1245 1246 case eSymbolTypeCode: 1247 case eSymbolTypeTrampoline: 1248 case eSymbolTypeResolver: 1249 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1250 // For ARM we have a bit in the n_desc field of the symbol that tells 1251 // us ARM/Thumb which is bit 0x0008. 1252 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1253 return AddressClass::eCodeAlternateISA; 1254 } 1255 return AddressClass::eCode; 1256 1257 case eSymbolTypeData: 1258 return AddressClass::eData; 1259 case eSymbolTypeRuntime: 1260 return AddressClass::eRuntime; 1261 case eSymbolTypeException: 1262 return AddressClass::eRuntime; 1263 case eSymbolTypeSourceFile: 1264 return AddressClass::eDebug; 1265 case eSymbolTypeHeaderFile: 1266 return AddressClass::eDebug; 1267 case eSymbolTypeObjectFile: 1268 return AddressClass::eDebug; 1269 case eSymbolTypeCommonBlock: 1270 return AddressClass::eDebug; 1271 case eSymbolTypeBlock: 1272 return AddressClass::eDebug; 1273 case eSymbolTypeLocal: 1274 return AddressClass::eData; 1275 case eSymbolTypeParam: 1276 return AddressClass::eData; 1277 case eSymbolTypeVariable: 1278 return AddressClass::eData; 1279 case eSymbolTypeVariableType: 1280 return AddressClass::eDebug; 1281 case eSymbolTypeLineEntry: 1282 return AddressClass::eDebug; 1283 case eSymbolTypeLineHeader: 1284 return AddressClass::eDebug; 1285 case eSymbolTypeScopeBegin: 1286 return AddressClass::eDebug; 1287 case eSymbolTypeScopeEnd: 1288 return AddressClass::eDebug; 1289 case eSymbolTypeAdditional: 1290 return AddressClass::eUnknown; 1291 case eSymbolTypeCompiler: 1292 return AddressClass::eDebug; 1293 case eSymbolTypeInstrumentation: 1294 return AddressClass::eDebug; 1295 case eSymbolTypeUndefined: 1296 return AddressClass::eUnknown; 1297 case eSymbolTypeObjCClass: 1298 return AddressClass::eRuntime; 1299 case eSymbolTypeObjCMetaClass: 1300 return AddressClass::eRuntime; 1301 case eSymbolTypeObjCIVar: 1302 return AddressClass::eRuntime; 1303 case eSymbolTypeReExported: 1304 return AddressClass::eRuntime; 1305 } 1306 } 1307 } 1308 return AddressClass::eUnknown; 1309 } 1310 1311 Symtab *ObjectFileMachO::GetSymtab() { 1312 ModuleSP module_sp(GetModule()); 1313 if (module_sp) { 1314 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1315 if (m_symtab_up == NULL) { 1316 m_symtab_up.reset(new Symtab(this)); 1317 std::lock_guard<std::recursive_mutex> symtab_guard( 1318 m_symtab_up->GetMutex()); 1319 ParseSymtab(); 1320 m_symtab_up->Finalize(); 1321 } 1322 } 1323 return m_symtab_up.get(); 1324 } 1325 1326 bool ObjectFileMachO::IsStripped() { 1327 if (m_dysymtab.cmd == 0) { 1328 ModuleSP module_sp(GetModule()); 1329 if (module_sp) { 1330 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1331 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1332 const lldb::offset_t load_cmd_offset = offset; 1333 1334 load_command lc; 1335 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1336 break; 1337 if (lc.cmd == LC_DYSYMTAB) { 1338 m_dysymtab.cmd = lc.cmd; 1339 m_dysymtab.cmdsize = lc.cmdsize; 1340 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1341 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1342 NULL) { 1343 // Clear m_dysymtab if we were unable to read all items from the 1344 // load command 1345 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1346 } 1347 } 1348 offset = load_cmd_offset + lc.cmdsize; 1349 } 1350 } 1351 } 1352 if (m_dysymtab.cmd) 1353 return m_dysymtab.nlocalsym <= 1; 1354 return false; 1355 } 1356 1357 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1358 EncryptedFileRanges result; 1359 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1360 1361 encryption_info_command encryption_cmd; 1362 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1363 const lldb::offset_t load_cmd_offset = offset; 1364 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1365 break; 1366 1367 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the 1368 // 3 fields we care about, so treat them the same. 1369 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1370 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1371 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1372 if (encryption_cmd.cryptid != 0) { 1373 EncryptedFileRanges::Entry entry; 1374 entry.SetRangeBase(encryption_cmd.cryptoff); 1375 entry.SetByteSize(encryption_cmd.cryptsize); 1376 result.Append(entry); 1377 } 1378 } 1379 } 1380 offset = load_cmd_offset + encryption_cmd.cmdsize; 1381 } 1382 1383 return result; 1384 } 1385 1386 void ObjectFileMachO::SanitizeSegmentCommand(segment_command_64 &seg_cmd, 1387 uint32_t cmd_idx) { 1388 if (m_length == 0 || seg_cmd.filesize == 0) 1389 return; 1390 1391 if (seg_cmd.fileoff > m_length) { 1392 // We have a load command that says it extends past the end of the file. 1393 // This is likely a corrupt file. We don't have any way to return an error 1394 // condition here (this method was likely invoked from something like 1395 // ObjectFile::GetSectionList()), so we just null out the section contents, 1396 // and dump a message to stdout. The most common case here is core file 1397 // debugging with a truncated file. 1398 const char *lc_segment_name = 1399 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1400 GetModule()->ReportWarning( 1401 "load command %u %s has a fileoff (0x%" PRIx64 1402 ") that extends beyond the end of the file (0x%" PRIx64 1403 "), ignoring this section", 1404 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1405 1406 seg_cmd.fileoff = 0; 1407 seg_cmd.filesize = 0; 1408 } 1409 1410 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1411 // We have a load command that says it extends past the end of the file. 1412 // This is likely a corrupt file. We don't have any way to return an error 1413 // condition here (this method was likely invoked from something like 1414 // ObjectFile::GetSectionList()), so we just null out the section contents, 1415 // and dump a message to stdout. The most common case here is core file 1416 // debugging with a truncated file. 1417 const char *lc_segment_name = 1418 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1419 GetModule()->ReportWarning( 1420 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1421 ") that extends beyond the end of the file (0x%" PRIx64 1422 "), the segment will be truncated to match", 1423 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1424 1425 // Truncate the length 1426 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1427 } 1428 } 1429 1430 static uint32_t GetSegmentPermissions(const segment_command_64 &seg_cmd) { 1431 uint32_t result = 0; 1432 if (seg_cmd.initprot & VM_PROT_READ) 1433 result |= ePermissionsReadable; 1434 if (seg_cmd.initprot & VM_PROT_WRITE) 1435 result |= ePermissionsWritable; 1436 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1437 result |= ePermissionsExecutable; 1438 return result; 1439 } 1440 1441 static lldb::SectionType GetSectionType(uint32_t flags, 1442 ConstString section_name) { 1443 1444 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1445 return eSectionTypeCode; 1446 1447 uint32_t mach_sect_type = flags & SECTION_TYPE; 1448 static ConstString g_sect_name_objc_data("__objc_data"); 1449 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1450 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1451 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1452 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1453 static ConstString g_sect_name_objc_const("__objc_const"); 1454 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1455 static ConstString g_sect_name_cfstring("__cfstring"); 1456 1457 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1458 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1459 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1460 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1461 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1462 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1463 static ConstString g_sect_name_dwarf_debug_loclists("__debug_loclists"); 1464 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1465 static ConstString g_sect_name_dwarf_debug_names("__debug_names"); 1466 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1467 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1468 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1469 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1470 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1471 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1472 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1473 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1474 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1475 static ConstString g_sect_name_eh_frame("__eh_frame"); 1476 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1477 static ConstString g_sect_name_text("__text"); 1478 static ConstString g_sect_name_data("__data"); 1479 static ConstString g_sect_name_go_symtab("__gosymtab"); 1480 1481 if (section_name == g_sect_name_dwarf_debug_abbrev) 1482 return eSectionTypeDWARFDebugAbbrev; 1483 if (section_name == g_sect_name_dwarf_debug_aranges) 1484 return eSectionTypeDWARFDebugAranges; 1485 if (section_name == g_sect_name_dwarf_debug_frame) 1486 return eSectionTypeDWARFDebugFrame; 1487 if (section_name == g_sect_name_dwarf_debug_info) 1488 return eSectionTypeDWARFDebugInfo; 1489 if (section_name == g_sect_name_dwarf_debug_line) 1490 return eSectionTypeDWARFDebugLine; 1491 if (section_name == g_sect_name_dwarf_debug_loc) 1492 return eSectionTypeDWARFDebugLoc; 1493 if (section_name == g_sect_name_dwarf_debug_loclists) 1494 return eSectionTypeDWARFDebugLocLists; 1495 if (section_name == g_sect_name_dwarf_debug_macinfo) 1496 return eSectionTypeDWARFDebugMacInfo; 1497 if (section_name == g_sect_name_dwarf_debug_names) 1498 return eSectionTypeDWARFDebugNames; 1499 if (section_name == g_sect_name_dwarf_debug_pubnames) 1500 return eSectionTypeDWARFDebugPubNames; 1501 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1502 return eSectionTypeDWARFDebugPubTypes; 1503 if (section_name == g_sect_name_dwarf_debug_ranges) 1504 return eSectionTypeDWARFDebugRanges; 1505 if (section_name == g_sect_name_dwarf_debug_str) 1506 return eSectionTypeDWARFDebugStr; 1507 if (section_name == g_sect_name_dwarf_debug_types) 1508 return eSectionTypeDWARFDebugTypes; 1509 if (section_name == g_sect_name_dwarf_apple_names) 1510 return eSectionTypeDWARFAppleNames; 1511 if (section_name == g_sect_name_dwarf_apple_types) 1512 return eSectionTypeDWARFAppleTypes; 1513 if (section_name == g_sect_name_dwarf_apple_namespaces) 1514 return eSectionTypeDWARFAppleNamespaces; 1515 if (section_name == g_sect_name_dwarf_apple_objc) 1516 return eSectionTypeDWARFAppleObjC; 1517 if (section_name == g_sect_name_objc_selrefs) 1518 return eSectionTypeDataCStringPointers; 1519 if (section_name == g_sect_name_objc_msgrefs) 1520 return eSectionTypeDataObjCMessageRefs; 1521 if (section_name == g_sect_name_eh_frame) 1522 return eSectionTypeEHFrame; 1523 if (section_name == g_sect_name_compact_unwind) 1524 return eSectionTypeCompactUnwind; 1525 if (section_name == g_sect_name_cfstring) 1526 return eSectionTypeDataObjCCFStrings; 1527 if (section_name == g_sect_name_go_symtab) 1528 return eSectionTypeGoSymtab; 1529 if (section_name == g_sect_name_objc_data || 1530 section_name == g_sect_name_objc_classrefs || 1531 section_name == g_sect_name_objc_superrefs || 1532 section_name == g_sect_name_objc_const || 1533 section_name == g_sect_name_objc_classlist) { 1534 return eSectionTypeDataPointers; 1535 } 1536 1537 switch (mach_sect_type) { 1538 // TODO: categorize sections by other flags for regular sections 1539 case S_REGULAR: 1540 if (section_name == g_sect_name_text) 1541 return eSectionTypeCode; 1542 if (section_name == g_sect_name_data) 1543 return eSectionTypeData; 1544 return eSectionTypeOther; 1545 case S_ZEROFILL: 1546 return eSectionTypeZeroFill; 1547 case S_CSTRING_LITERALS: // section with only literal C strings 1548 return eSectionTypeDataCString; 1549 case S_4BYTE_LITERALS: // section with only 4 byte literals 1550 return eSectionTypeData4; 1551 case S_8BYTE_LITERALS: // section with only 8 byte literals 1552 return eSectionTypeData8; 1553 case S_LITERAL_POINTERS: // section with only pointers to literals 1554 return eSectionTypeDataPointers; 1555 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1556 return eSectionTypeDataPointers; 1557 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1558 return eSectionTypeDataPointers; 1559 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1560 // the reserved2 field 1561 return eSectionTypeCode; 1562 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1563 // initialization 1564 return eSectionTypeDataPointers; 1565 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1566 // termination 1567 return eSectionTypeDataPointers; 1568 case S_COALESCED: 1569 return eSectionTypeOther; 1570 case S_GB_ZEROFILL: 1571 return eSectionTypeZeroFill; 1572 case S_INTERPOSING: // section with only pairs of function pointers for 1573 // interposing 1574 return eSectionTypeCode; 1575 case S_16BYTE_LITERALS: // section with only 16 byte literals 1576 return eSectionTypeData16; 1577 case S_DTRACE_DOF: 1578 return eSectionTypeDebug; 1579 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1580 return eSectionTypeDataPointers; 1581 default: 1582 return eSectionTypeOther; 1583 } 1584 } 1585 1586 struct ObjectFileMachO::SegmentParsingContext { 1587 const EncryptedFileRanges EncryptedRanges; 1588 lldb_private::SectionList &UnifiedList; 1589 uint32_t NextSegmentIdx = 0; 1590 uint32_t NextSectionIdx = 0; 1591 bool FileAddressesChanged = false; 1592 1593 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1594 lldb_private::SectionList &UnifiedList) 1595 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1596 }; 1597 1598 void ObjectFileMachO::ProcessSegmentCommand(const load_command &load_cmd_, 1599 lldb::offset_t offset, 1600 uint32_t cmd_idx, 1601 SegmentParsingContext &context) { 1602 segment_command_64 load_cmd; 1603 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1604 1605 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1606 return; 1607 1608 ModuleSP module_sp = GetModule(); 1609 const bool is_core = GetType() == eTypeCoreFile; 1610 const bool is_dsym = (m_header.filetype == MH_DSYM); 1611 bool add_section = true; 1612 bool add_to_unified = true; 1613 ConstString const_segname( 1614 load_cmd.segname, strnlen(load_cmd.segname, sizeof(load_cmd.segname))); 1615 1616 SectionSP unified_section_sp( 1617 context.UnifiedList.FindSectionByName(const_segname)); 1618 if (is_dsym && unified_section_sp) { 1619 if (const_segname == GetSegmentNameLINKEDIT()) { 1620 // We need to keep the __LINKEDIT segment private to this object file 1621 // only 1622 add_to_unified = false; 1623 } else { 1624 // This is the dSYM file and this section has already been created by the 1625 // object file, no need to create it. 1626 add_section = false; 1627 } 1628 } 1629 load_cmd.vmaddr = m_data.GetAddress(&offset); 1630 load_cmd.vmsize = m_data.GetAddress(&offset); 1631 load_cmd.fileoff = m_data.GetAddress(&offset); 1632 load_cmd.filesize = m_data.GetAddress(&offset); 1633 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1634 return; 1635 1636 SanitizeSegmentCommand(load_cmd, cmd_idx); 1637 1638 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1639 const bool segment_is_encrypted = 1640 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1641 1642 // Keep a list of mach segments around in case we need to get at data that 1643 // isn't stored in the abstracted Sections. 1644 m_mach_segments.push_back(load_cmd); 1645 1646 // Use a segment ID of the segment index shifted left by 8 so they never 1647 // conflict with any of the sections. 1648 SectionSP segment_sp; 1649 if (add_section && (const_segname || is_core)) { 1650 segment_sp = std::make_shared<Section>( 1651 module_sp, // Module to which this section belongs 1652 this, // Object file to which this sections belongs 1653 ++context.NextSegmentIdx 1654 << 8, // Section ID is the 1 based segment index 1655 // shifted right by 8 bits as not to collide with any of the 256 1656 // section IDs that are possible 1657 const_segname, // Name of this section 1658 eSectionTypeContainer, // This section is a container of other 1659 // sections. 1660 load_cmd.vmaddr, // File VM address == addresses as they are 1661 // found in the object file 1662 load_cmd.vmsize, // VM size in bytes of this section 1663 load_cmd.fileoff, // Offset to the data for this section in 1664 // the file 1665 load_cmd.filesize, // Size in bytes of this section as found 1666 // in the file 1667 0, // Segments have no alignment information 1668 load_cmd.flags); // Flags for this section 1669 1670 segment_sp->SetIsEncrypted(segment_is_encrypted); 1671 m_sections_up->AddSection(segment_sp); 1672 segment_sp->SetPermissions(segment_permissions); 1673 if (add_to_unified) 1674 context.UnifiedList.AddSection(segment_sp); 1675 } else if (unified_section_sp) { 1676 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1677 // Check to see if the module was read from memory? 1678 if (module_sp->GetObjectFile()->GetBaseAddress().IsValid()) { 1679 // We have a module that is in memory and needs to have its file 1680 // address adjusted. We need to do this because when we load a file 1681 // from memory, its addresses will be slid already, yet the addresses 1682 // in the new symbol file will still be unslid. Since everything is 1683 // stored as section offset, this shouldn't cause any problems. 1684 1685 // Make sure we've parsed the symbol table from the ObjectFile before 1686 // we go around changing its Sections. 1687 module_sp->GetObjectFile()->GetSymtab(); 1688 // eh_frame would present the same problems but we parse that on a per- 1689 // function basis as-needed so it's more difficult to remove its use of 1690 // the Sections. Realistically, the environments where this code path 1691 // will be taken will not have eh_frame sections. 1692 1693 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1694 1695 // Notify the module that the section addresses have been changed once 1696 // we're done so any file-address caches can be updated. 1697 context.FileAddressesChanged = true; 1698 } 1699 } 1700 m_sections_up->AddSection(unified_section_sp); 1701 } 1702 1703 struct section_64 sect64; 1704 ::memset(§64, 0, sizeof(sect64)); 1705 // Push a section into our mach sections for the section at index zero 1706 // (NO_SECT) if we don't have any mach sections yet... 1707 if (m_mach_sections.empty()) 1708 m_mach_sections.push_back(sect64); 1709 uint32_t segment_sect_idx; 1710 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1711 1712 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1713 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1714 ++segment_sect_idx) { 1715 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1716 sizeof(sect64.sectname)) == NULL) 1717 break; 1718 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1719 sizeof(sect64.segname)) == NULL) 1720 break; 1721 sect64.addr = m_data.GetAddress(&offset); 1722 sect64.size = m_data.GetAddress(&offset); 1723 1724 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1725 break; 1726 1727 // Keep a list of mach sections around in case we need to get at data that 1728 // isn't stored in the abstracted Sections. 1729 m_mach_sections.push_back(sect64); 1730 1731 if (add_section) { 1732 ConstString section_name( 1733 sect64.sectname, strnlen(sect64.sectname, sizeof(sect64.sectname))); 1734 if (!const_segname) { 1735 // We have a segment with no name so we need to conjure up segments 1736 // that correspond to the section's segname if there isn't already such 1737 // a section. If there is such a section, we resize the section so that 1738 // it spans all sections. We also mark these sections as fake so 1739 // address matches don't hit if they land in the gaps between the child 1740 // sections. 1741 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1742 sizeof(sect64.segname)); 1743 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1744 if (segment_sp.get()) { 1745 Section *segment = segment_sp.get(); 1746 // Grow the section size as needed. 1747 const lldb::addr_t sect64_min_addr = sect64.addr; 1748 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1749 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1750 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1751 const lldb::addr_t curr_seg_max_addr = 1752 curr_seg_min_addr + curr_seg_byte_size; 1753 if (sect64_min_addr >= curr_seg_min_addr) { 1754 const lldb::addr_t new_seg_byte_size = 1755 sect64_max_addr - curr_seg_min_addr; 1756 // Only grow the section size if needed 1757 if (new_seg_byte_size > curr_seg_byte_size) 1758 segment->SetByteSize(new_seg_byte_size); 1759 } else { 1760 // We need to change the base address of the segment and adjust the 1761 // child section offsets for all existing children. 1762 const lldb::addr_t slide_amount = 1763 sect64_min_addr - curr_seg_min_addr; 1764 segment->Slide(slide_amount, false); 1765 segment->GetChildren().Slide(-slide_amount, false); 1766 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1767 } 1768 1769 // Grow the section size as needed. 1770 if (sect64.offset) { 1771 const lldb::addr_t segment_min_file_offset = 1772 segment->GetFileOffset(); 1773 const lldb::addr_t segment_max_file_offset = 1774 segment_min_file_offset + segment->GetFileSize(); 1775 1776 const lldb::addr_t section_min_file_offset = sect64.offset; 1777 const lldb::addr_t section_max_file_offset = 1778 section_min_file_offset + sect64.size; 1779 const lldb::addr_t new_file_offset = 1780 std::min(section_min_file_offset, segment_min_file_offset); 1781 const lldb::addr_t new_file_size = 1782 std::max(section_max_file_offset, segment_max_file_offset) - 1783 new_file_offset; 1784 segment->SetFileOffset(new_file_offset); 1785 segment->SetFileSize(new_file_size); 1786 } 1787 } else { 1788 // Create a fake section for the section's named segment 1789 segment_sp = std::make_shared<Section>( 1790 segment_sp, // Parent section 1791 module_sp, // Module to which this section belongs 1792 this, // Object file to which this section belongs 1793 ++context.NextSegmentIdx 1794 << 8, // Section ID is the 1 based segment index 1795 // shifted right by 8 bits as not to 1796 // collide with any of the 256 section IDs 1797 // that are possible 1798 const_segname, // Name of this section 1799 eSectionTypeContainer, // This section is a container of 1800 // other sections. 1801 sect64.addr, // File VM address == addresses as they are 1802 // found in the object file 1803 sect64.size, // VM size in bytes of this section 1804 sect64.offset, // Offset to the data for this section in 1805 // the file 1806 sect64.offset ? sect64.size : 0, // Size in bytes of 1807 // this section as 1808 // found in the file 1809 sect64.align, 1810 load_cmd.flags); // Flags for this section 1811 segment_sp->SetIsFake(true); 1812 segment_sp->SetPermissions(segment_permissions); 1813 m_sections_up->AddSection(segment_sp); 1814 if (add_to_unified) 1815 context.UnifiedList.AddSection(segment_sp); 1816 segment_sp->SetIsEncrypted(segment_is_encrypted); 1817 } 1818 } 1819 assert(segment_sp.get()); 1820 1821 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1822 1823 SectionSP section_sp(new Section( 1824 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1825 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1826 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1827 sect64.flags)); 1828 // Set the section to be encrypted to match the segment 1829 1830 bool section_is_encrypted = false; 1831 if (!segment_is_encrypted && load_cmd.filesize != 0) 1832 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1833 sect64.offset) != NULL; 1834 1835 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1836 section_sp->SetPermissions(segment_permissions); 1837 segment_sp->GetChildren().AddSection(section_sp); 1838 1839 if (segment_sp->IsFake()) { 1840 segment_sp.reset(); 1841 const_segname.Clear(); 1842 } 1843 } 1844 } 1845 if (segment_sp && is_dsym) { 1846 if (first_segment_sectID <= context.NextSectionIdx) { 1847 lldb::user_id_t sect_uid; 1848 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1849 ++sect_uid) { 1850 SectionSP curr_section_sp( 1851 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1852 SectionSP next_section_sp; 1853 if (sect_uid + 1 <= context.NextSectionIdx) 1854 next_section_sp = 1855 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1856 1857 if (curr_section_sp.get()) { 1858 if (curr_section_sp->GetByteSize() == 0) { 1859 if (next_section_sp.get() != NULL) 1860 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1861 curr_section_sp->GetFileAddress()); 1862 else 1863 curr_section_sp->SetByteSize(load_cmd.vmsize); 1864 } 1865 } 1866 } 1867 } 1868 } 1869 } 1870 1871 void ObjectFileMachO::ProcessDysymtabCommand(const load_command &load_cmd, 1872 lldb::offset_t offset) { 1873 m_dysymtab.cmd = load_cmd.cmd; 1874 m_dysymtab.cmdsize = load_cmd.cmdsize; 1875 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1876 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1877 } 1878 1879 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1880 if (m_sections_up) 1881 return; 1882 1883 m_sections_up.reset(new SectionList()); 1884 1885 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1886 // bool dump_sections = false; 1887 ModuleSP module_sp(GetModule()); 1888 1889 offset = MachHeaderSizeFromMagic(m_header.magic); 1890 1891 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1892 struct load_command load_cmd; 1893 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1894 const lldb::offset_t load_cmd_offset = offset; 1895 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1896 break; 1897 1898 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1899 ProcessSegmentCommand(load_cmd, offset, i, context); 1900 else if (load_cmd.cmd == LC_DYSYMTAB) 1901 ProcessDysymtabCommand(load_cmd, offset); 1902 1903 offset = load_cmd_offset + load_cmd.cmdsize; 1904 } 1905 1906 if (context.FileAddressesChanged && module_sp) 1907 module_sp->SectionFileAddressesChanged(); 1908 } 1909 1910 class MachSymtabSectionInfo { 1911 public: 1912 MachSymtabSectionInfo(SectionList *section_list) 1913 : m_section_list(section_list), m_section_infos() { 1914 // Get the number of sections down to a depth of 1 to include all segments 1915 // and their sections, but no other sections that may be added for debug 1916 // map or 1917 m_section_infos.resize(section_list->GetNumSections(1)); 1918 } 1919 1920 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1921 if (n_sect == 0) 1922 return SectionSP(); 1923 if (n_sect < m_section_infos.size()) { 1924 if (!m_section_infos[n_sect].section_sp) { 1925 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1926 m_section_infos[n_sect].section_sp = section_sp; 1927 if (section_sp) { 1928 m_section_infos[n_sect].vm_range.SetBaseAddress( 1929 section_sp->GetFileAddress()); 1930 m_section_infos[n_sect].vm_range.SetByteSize( 1931 section_sp->GetByteSize()); 1932 } else { 1933 Host::SystemLog(Host::eSystemLogError, 1934 "error: unable to find section for section %u\n", 1935 n_sect); 1936 } 1937 } 1938 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1939 // Symbol is in section. 1940 return m_section_infos[n_sect].section_sp; 1941 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1942 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1943 file_addr) { 1944 // Symbol is in section with zero size, but has the same start address 1945 // as the section. This can happen with linker symbols (symbols that 1946 // start with the letter 'l' or 'L'. 1947 return m_section_infos[n_sect].section_sp; 1948 } 1949 } 1950 return m_section_list->FindSectionContainingFileAddress(file_addr); 1951 } 1952 1953 protected: 1954 struct SectionInfo { 1955 SectionInfo() : vm_range(), section_sp() {} 1956 1957 VMRange vm_range; 1958 SectionSP section_sp; 1959 }; 1960 SectionList *m_section_list; 1961 std::vector<SectionInfo> m_section_infos; 1962 }; 1963 1964 struct TrieEntry { 1965 TrieEntry() 1966 : name(), address(LLDB_INVALID_ADDRESS), flags(0), other(0), 1967 import_name() {} 1968 1969 void Clear() { 1970 name.Clear(); 1971 address = LLDB_INVALID_ADDRESS; 1972 flags = 0; 1973 other = 0; 1974 import_name.Clear(); 1975 } 1976 1977 void Dump() const { 1978 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1979 static_cast<unsigned long long>(address), 1980 static_cast<unsigned long long>(flags), 1981 static_cast<unsigned long long>(other), name.GetCString()); 1982 if (import_name) 1983 printf(" -> \"%s\"\n", import_name.GetCString()); 1984 else 1985 printf("\n"); 1986 } 1987 ConstString name; 1988 uint64_t address; 1989 uint64_t flags; 1990 uint64_t other; 1991 ConstString import_name; 1992 }; 1993 1994 struct TrieEntryWithOffset { 1995 lldb::offset_t nodeOffset; 1996 TrieEntry entry; 1997 1998 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1999 2000 void Dump(uint32_t idx) const { 2001 printf("[%3u] 0x%16.16llx: ", idx, 2002 static_cast<unsigned long long>(nodeOffset)); 2003 entry.Dump(); 2004 } 2005 2006 bool operator<(const TrieEntryWithOffset &other) const { 2007 return (nodeOffset < other.nodeOffset); 2008 } 2009 }; 2010 2011 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2012 const bool is_arm, 2013 std::vector<llvm::StringRef> &nameSlices, 2014 std::set<lldb::addr_t> &resolver_addresses, 2015 std::vector<TrieEntryWithOffset> &output) { 2016 if (!data.ValidOffset(offset)) 2017 return true; 2018 2019 const uint64_t terminalSize = data.GetULEB128(&offset); 2020 lldb::offset_t children_offset = offset + terminalSize; 2021 if (terminalSize != 0) { 2022 TrieEntryWithOffset e(offset); 2023 e.entry.flags = data.GetULEB128(&offset); 2024 const char *import_name = NULL; 2025 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2026 e.entry.address = 0; 2027 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2028 import_name = data.GetCStr(&offset); 2029 } else { 2030 e.entry.address = data.GetULEB128(&offset); 2031 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2032 e.entry.other = data.GetULEB128(&offset); 2033 uint64_t resolver_addr = e.entry.other; 2034 if (is_arm) 2035 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2036 resolver_addresses.insert(resolver_addr); 2037 } else 2038 e.entry.other = 0; 2039 } 2040 // Only add symbols that are reexport symbols with a valid import name 2041 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && 2042 import_name[0]) { 2043 std::string name; 2044 if (!nameSlices.empty()) { 2045 for (auto name_slice : nameSlices) 2046 name.append(name_slice.data(), name_slice.size()); 2047 } 2048 if (name.size() > 1) { 2049 // Skip the leading '_' 2050 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2051 } 2052 if (import_name) { 2053 // Skip the leading '_' 2054 e.entry.import_name.SetCString(import_name + 1); 2055 } 2056 output.push_back(e); 2057 } 2058 } 2059 2060 const uint8_t childrenCount = data.GetU8(&children_offset); 2061 for (uint8_t i = 0; i < childrenCount; ++i) { 2062 const char *cstr = data.GetCStr(&children_offset); 2063 if (cstr) 2064 nameSlices.push_back(llvm::StringRef(cstr)); 2065 else 2066 return false; // Corrupt data 2067 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2068 if (childNodeOffset) { 2069 if (!ParseTrieEntries(data, childNodeOffset, is_arm, nameSlices, 2070 resolver_addresses, output)) { 2071 return false; 2072 } 2073 } 2074 nameSlices.pop_back(); 2075 } 2076 return true; 2077 } 2078 2079 // Read the UUID out of a dyld_shared_cache file on-disk. 2080 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2081 const ByteOrder byte_order, 2082 const uint32_t addr_byte_size) { 2083 UUID dsc_uuid; 2084 DataBufferSP DscData = MapFileData( 2085 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2086 if (!DscData) 2087 return dsc_uuid; 2088 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2089 2090 char version_str[7]; 2091 lldb::offset_t offset = 0; 2092 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2093 version_str[6] = '\0'; 2094 if (strcmp(version_str, "dyld_v") == 0) { 2095 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2096 dsc_uuid = UUID::fromOptionalData( 2097 dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t)); 2098 } 2099 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2100 if (log && dsc_uuid.IsValid()) { 2101 log->Printf("Shared cache %s has UUID %s", dyld_shared_cache.GetPath().c_str(), 2102 dsc_uuid.GetAsString().c_str()); 2103 } 2104 return dsc_uuid; 2105 } 2106 2107 size_t ObjectFileMachO::ParseSymtab() { 2108 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2109 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2110 m_file.GetFilename().AsCString("")); 2111 ModuleSP module_sp(GetModule()); 2112 if (!module_sp) 2113 return 0; 2114 2115 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2116 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2117 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2118 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2119 FunctionStarts function_starts; 2120 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2121 uint32_t i; 2122 FileSpecList dylib_files; 2123 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2124 static const llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2125 static const llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2126 static const llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2127 2128 for (i = 0; i < m_header.ncmds; ++i) { 2129 const lldb::offset_t cmd_offset = offset; 2130 // Read in the load command and load command size 2131 struct load_command lc; 2132 if (m_data.GetU32(&offset, &lc, 2) == NULL) 2133 break; 2134 // Watch for the symbol table load command 2135 switch (lc.cmd) { 2136 case LC_SYMTAB: 2137 symtab_load_command.cmd = lc.cmd; 2138 symtab_load_command.cmdsize = lc.cmdsize; 2139 // Read in the rest of the symtab load command 2140 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2141 0) // fill in symoff, nsyms, stroff, strsize fields 2142 return 0; 2143 if (symtab_load_command.symoff == 0) { 2144 if (log) 2145 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2146 return 0; 2147 } 2148 2149 if (symtab_load_command.stroff == 0) { 2150 if (log) 2151 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2152 return 0; 2153 } 2154 2155 if (symtab_load_command.nsyms == 0) { 2156 if (log) 2157 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2158 return 0; 2159 } 2160 2161 if (symtab_load_command.strsize == 0) { 2162 if (log) 2163 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2164 return 0; 2165 } 2166 break; 2167 2168 case LC_DYLD_INFO: 2169 case LC_DYLD_INFO_ONLY: 2170 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2171 dyld_info.cmd = lc.cmd; 2172 dyld_info.cmdsize = lc.cmdsize; 2173 } else { 2174 memset(&dyld_info, 0, sizeof(dyld_info)); 2175 } 2176 break; 2177 2178 case LC_LOAD_DYLIB: 2179 case LC_LOAD_WEAK_DYLIB: 2180 case LC_REEXPORT_DYLIB: 2181 case LC_LOADFVMLIB: 2182 case LC_LOAD_UPWARD_DYLIB: { 2183 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2184 const char *path = m_data.PeekCStr(name_offset); 2185 if (path) { 2186 FileSpec file_spec(path); 2187 // Strip the path if there is @rpath, @executable, etc so we just use 2188 // the basename 2189 if (path[0] == '@') 2190 file_spec.GetDirectory().Clear(); 2191 2192 if (lc.cmd == LC_REEXPORT_DYLIB) { 2193 m_reexported_dylibs.AppendIfUnique(file_spec); 2194 } 2195 2196 dylib_files.Append(file_spec); 2197 } 2198 } break; 2199 2200 case LC_FUNCTION_STARTS: 2201 function_starts_load_command.cmd = lc.cmd; 2202 function_starts_load_command.cmdsize = lc.cmdsize; 2203 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2204 NULL) // fill in symoff, nsyms, stroff, strsize fields 2205 memset(&function_starts_load_command, 0, 2206 sizeof(function_starts_load_command)); 2207 break; 2208 2209 default: 2210 break; 2211 } 2212 offset = cmd_offset + lc.cmdsize; 2213 } 2214 2215 if (symtab_load_command.cmd) { 2216 Symtab *symtab = m_symtab_up.get(); 2217 SectionList *section_list = GetSectionList(); 2218 if (section_list == NULL) 2219 return 0; 2220 2221 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2222 const ByteOrder byte_order = m_data.GetByteOrder(); 2223 bool bit_width_32 = addr_byte_size == 4; 2224 const size_t nlist_byte_size = 2225 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2226 2227 DataExtractor nlist_data(NULL, 0, byte_order, addr_byte_size); 2228 DataExtractor strtab_data(NULL, 0, byte_order, addr_byte_size); 2229 DataExtractor function_starts_data(NULL, 0, byte_order, addr_byte_size); 2230 DataExtractor indirect_symbol_index_data(NULL, 0, byte_order, 2231 addr_byte_size); 2232 DataExtractor dyld_trie_data(NULL, 0, byte_order, addr_byte_size); 2233 2234 const addr_t nlist_data_byte_size = 2235 symtab_load_command.nsyms * nlist_byte_size; 2236 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2237 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2238 2239 ProcessSP process_sp(m_process_wp.lock()); 2240 Process *process = process_sp.get(); 2241 2242 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2243 2244 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2245 Target &target = process->GetTarget(); 2246 2247 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2248 2249 SectionSP linkedit_section_sp( 2250 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2251 // Reading mach file from memory in a process or core file... 2252 2253 if (linkedit_section_sp) { 2254 addr_t linkedit_load_addr = 2255 linkedit_section_sp->GetLoadBaseAddress(&target); 2256 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2257 // We might be trying to access the symbol table before the 2258 // __LINKEDIT's load address has been set in the target. We can't 2259 // fail to read the symbol table, so calculate the right address 2260 // manually 2261 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2262 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2263 } 2264 2265 const addr_t linkedit_file_offset = 2266 linkedit_section_sp->GetFileOffset(); 2267 const addr_t symoff_addr = linkedit_load_addr + 2268 symtab_load_command.symoff - 2269 linkedit_file_offset; 2270 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2271 linkedit_file_offset; 2272 2273 bool data_was_read = false; 2274 2275 #if defined(__APPLE__) && \ 2276 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2277 if (m_header.flags & 0x80000000u && 2278 process->GetAddressByteSize() == sizeof(void *)) { 2279 // This mach-o memory file is in the dyld shared cache. If this 2280 // program is not remote and this is iOS, then this process will 2281 // share the same shared cache as the process we are debugging and we 2282 // can read the entire __LINKEDIT from the address space in this 2283 // process. This is a needed optimization that is used for local iOS 2284 // debugging only since all shared libraries in the shared cache do 2285 // not have corresponding files that exist in the file system of the 2286 // device. They have been combined into a single file. This means we 2287 // always have to load these files from memory. All of the symbol and 2288 // string tables from all of the __LINKEDIT sections from the shared 2289 // libraries in the shared cache have been merged into a single large 2290 // symbol and string table. Reading all of this symbol and string 2291 // table data across can slow down debug launch times, so we optimize 2292 // this by reading the memory for the __LINKEDIT section from this 2293 // process. 2294 2295 UUID lldb_shared_cache; 2296 addr_t lldb_shared_cache_addr; 2297 GetLLDBSharedCacheUUID (lldb_shared_cache_addr, lldb_shared_cache); 2298 UUID process_shared_cache; 2299 addr_t process_shared_cache_addr; 2300 GetProcessSharedCacheUUID(process, process_shared_cache_addr, process_shared_cache); 2301 bool use_lldb_cache = true; 2302 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2303 (lldb_shared_cache != process_shared_cache 2304 || process_shared_cache_addr != lldb_shared_cache_addr)) { 2305 use_lldb_cache = false; 2306 } 2307 2308 PlatformSP platform_sp(target.GetPlatform()); 2309 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2310 data_was_read = true; 2311 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2312 eByteOrderLittle); 2313 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2314 eByteOrderLittle); 2315 if (function_starts_load_command.cmd) { 2316 const addr_t func_start_addr = 2317 linkedit_load_addr + function_starts_load_command.dataoff - 2318 linkedit_file_offset; 2319 function_starts_data.SetData( 2320 (void *)func_start_addr, 2321 function_starts_load_command.datasize, eByteOrderLittle); 2322 } 2323 } 2324 } 2325 #endif 2326 2327 if (!data_was_read) { 2328 // Always load dyld - the dynamic linker - from memory if we didn't 2329 // find a binary anywhere else. lldb will not register 2330 // dylib/framework/bundle loads/unloads if we don't have the dyld 2331 // symbols, we force dyld to load from memory despite the user's 2332 // target.memory-module-load-level setting. 2333 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2334 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2335 DataBufferSP nlist_data_sp( 2336 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2337 if (nlist_data_sp) 2338 nlist_data.SetData(nlist_data_sp, 0, 2339 nlist_data_sp->GetByteSize()); 2340 if (m_dysymtab.nindirectsyms != 0) { 2341 const addr_t indirect_syms_addr = linkedit_load_addr + 2342 m_dysymtab.indirectsymoff - 2343 linkedit_file_offset; 2344 DataBufferSP indirect_syms_data_sp( 2345 ReadMemory(process_sp, indirect_syms_addr, 2346 m_dysymtab.nindirectsyms * 4)); 2347 if (indirect_syms_data_sp) 2348 indirect_symbol_index_data.SetData( 2349 indirect_syms_data_sp, 0, 2350 indirect_syms_data_sp->GetByteSize()); 2351 // If this binary is outside the shared cache, 2352 // cache the string table. 2353 // Binaries in the shared cache all share a giant string table, and 2354 // we can't share the string tables across multiple ObjectFileMachO's, 2355 // so we'd end up re-reading this mega-strtab for every binary 2356 // in the shared cache - it would be a big perf problem. 2357 // For binaries outside the shared cache, it's faster to read the 2358 // entire strtab at once instead of piece-by-piece as we process 2359 // the nlist records. 2360 if ((m_header.flags & 0x80000000u) == 0) { 2361 DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, 2362 strtab_data_byte_size)); 2363 if (strtab_data_sp) { 2364 strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 2365 } 2366 } 2367 } 2368 } 2369 if (memory_module_load_level >= 2370 eMemoryModuleLoadLevelPartial) { 2371 if (function_starts_load_command.cmd) { 2372 const addr_t func_start_addr = 2373 linkedit_load_addr + function_starts_load_command.dataoff - 2374 linkedit_file_offset; 2375 DataBufferSP func_start_data_sp( 2376 ReadMemory(process_sp, func_start_addr, 2377 function_starts_load_command.datasize)); 2378 if (func_start_data_sp) 2379 function_starts_data.SetData(func_start_data_sp, 0, 2380 func_start_data_sp->GetByteSize()); 2381 } 2382 } 2383 } 2384 } 2385 } else { 2386 nlist_data.SetData(m_data, symtab_load_command.symoff, 2387 nlist_data_byte_size); 2388 strtab_data.SetData(m_data, symtab_load_command.stroff, 2389 strtab_data_byte_size); 2390 2391 if (dyld_info.export_size > 0) { 2392 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2393 dyld_info.export_size); 2394 } 2395 2396 if (m_dysymtab.nindirectsyms != 0) { 2397 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2398 m_dysymtab.nindirectsyms * 4); 2399 } 2400 if (function_starts_load_command.cmd) { 2401 function_starts_data.SetData(m_data, 2402 function_starts_load_command.dataoff, 2403 function_starts_load_command.datasize); 2404 } 2405 } 2406 2407 if (nlist_data.GetByteSize() == 0 && 2408 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2409 if (log) 2410 module_sp->LogMessage(log, "failed to read nlist data"); 2411 return 0; 2412 } 2413 2414 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2415 if (!have_strtab_data) { 2416 if (process) { 2417 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2418 if (log) 2419 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2420 return 0; 2421 } 2422 } else { 2423 if (log) 2424 module_sp->LogMessage(log, "failed to read strtab data"); 2425 return 0; 2426 } 2427 } 2428 2429 ConstString g_segment_name_TEXT = GetSegmentNameTEXT(); 2430 ConstString g_segment_name_DATA = GetSegmentNameDATA(); 2431 ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2432 ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2433 ConstString g_segment_name_OBJC = GetSegmentNameOBJC(); 2434 ConstString g_section_name_eh_frame = GetSectionNameEHFrame(); 2435 SectionSP text_section_sp( 2436 section_list->FindSectionByName(g_segment_name_TEXT)); 2437 SectionSP data_section_sp( 2438 section_list->FindSectionByName(g_segment_name_DATA)); 2439 SectionSP data_dirty_section_sp( 2440 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2441 SectionSP data_const_section_sp( 2442 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2443 SectionSP objc_section_sp( 2444 section_list->FindSectionByName(g_segment_name_OBJC)); 2445 SectionSP eh_frame_section_sp; 2446 if (text_section_sp.get()) 2447 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2448 g_section_name_eh_frame); 2449 else 2450 eh_frame_section_sp = 2451 section_list->FindSectionByName(g_section_name_eh_frame); 2452 2453 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2454 2455 // lldb works best if it knows the start address of all functions in a 2456 // module. Linker symbols or debug info are normally the best source of 2457 // information for start addr / size but they may be stripped in a released 2458 // binary. Two additional sources of information exist in Mach-O binaries: 2459 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2460 // function's start address in the 2461 // binary, relative to the text section. 2462 // eh_frame - the eh_frame FDEs have the start addr & size of 2463 // each function 2464 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2465 // all modern binaries. 2466 // Binaries built to run on older releases may need to use eh_frame 2467 // information. 2468 2469 if (text_section_sp && function_starts_data.GetByteSize()) { 2470 FunctionStarts::Entry function_start_entry; 2471 function_start_entry.data = false; 2472 lldb::offset_t function_start_offset = 0; 2473 function_start_entry.addr = text_section_sp->GetFileAddress(); 2474 uint64_t delta; 2475 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2476 0) { 2477 // Now append the current entry 2478 function_start_entry.addr += delta; 2479 function_starts.Append(function_start_entry); 2480 } 2481 } else { 2482 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2483 // load command claiming an eh_frame but it doesn't actually have the 2484 // eh_frame content. And if we have a dSYM, we don't need to do any of 2485 // this fill-in-the-missing-symbols works anyway - the debug info should 2486 // give us all the functions in the module. 2487 if (text_section_sp.get() && eh_frame_section_sp.get() && 2488 m_type != eTypeDebugInfo) { 2489 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2490 DWARFCallFrameInfo::EH); 2491 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2492 eh_frame.GetFunctionAddressAndSizeVector(functions); 2493 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2494 size_t count = functions.GetSize(); 2495 for (size_t i = 0; i < count; ++i) { 2496 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2497 functions.GetEntryAtIndex(i); 2498 if (func) { 2499 FunctionStarts::Entry function_start_entry; 2500 function_start_entry.addr = func->base - text_base_addr; 2501 function_starts.Append(function_start_entry); 2502 } 2503 } 2504 } 2505 } 2506 2507 const size_t function_starts_count = function_starts.GetSize(); 2508 2509 // For user process binaries (executables, dylibs, frameworks, bundles), if 2510 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2511 // going to assume the binary has been stripped. Don't allow assembly 2512 // language instruction emulation because we don't know proper function 2513 // start boundaries. 2514 // 2515 // For all other types of binaries (kernels, stand-alone bare board 2516 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2517 // sections - we should not make any assumptions about them based on that. 2518 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2519 m_allow_assembly_emulation_unwind_plans = false; 2520 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2521 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2522 2523 if (unwind_or_symbol_log) 2524 module_sp->LogMessage( 2525 unwind_or_symbol_log, 2526 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2527 } 2528 2529 const user_id_t TEXT_eh_frame_sectID = 2530 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2531 : static_cast<user_id_t>(NO_SECT); 2532 2533 lldb::offset_t nlist_data_offset = 0; 2534 2535 uint32_t N_SO_index = UINT32_MAX; 2536 2537 MachSymtabSectionInfo section_info(section_list); 2538 std::vector<uint32_t> N_FUN_indexes; 2539 std::vector<uint32_t> N_NSYM_indexes; 2540 std::vector<uint32_t> N_INCL_indexes; 2541 std::vector<uint32_t> N_BRAC_indexes; 2542 std::vector<uint32_t> N_COMM_indexes; 2543 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2544 typedef std::map<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2545 typedef std::map<const char *, uint32_t> ConstNameToSymbolIndexMap; 2546 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2547 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2548 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2549 // Any symbols that get merged into another will get an entry in this map 2550 // so we know 2551 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2552 uint32_t nlist_idx = 0; 2553 Symbol *symbol_ptr = NULL; 2554 2555 uint32_t sym_idx = 0; 2556 Symbol *sym = NULL; 2557 size_t num_syms = 0; 2558 std::string memory_symbol_name; 2559 uint32_t unmapped_local_symbols_found = 0; 2560 2561 std::vector<TrieEntryWithOffset> trie_entries; 2562 std::set<lldb::addr_t> resolver_addresses; 2563 2564 if (dyld_trie_data.GetByteSize() > 0) { 2565 std::vector<llvm::StringRef> nameSlices; 2566 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, 2567 resolver_addresses, trie_entries); 2568 2569 ConstString text_segment_name("__TEXT"); 2570 SectionSP text_segment_sp = 2571 GetSectionList()->FindSectionByName(text_segment_name); 2572 if (text_segment_sp) { 2573 const lldb::addr_t text_segment_file_addr = 2574 text_segment_sp->GetFileAddress(); 2575 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2576 for (auto &e : trie_entries) 2577 e.entry.address += text_segment_file_addr; 2578 } 2579 } 2580 } 2581 2582 typedef std::set<ConstString> IndirectSymbols; 2583 IndirectSymbols indirect_symbol_names; 2584 2585 #if defined(__APPLE__) && \ 2586 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2587 2588 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2589 // optimized by moving LOCAL symbols out of the memory mapped portion of 2590 // the DSC. The symbol information has all been retained, but it isn't 2591 // available in the normal nlist data. However, there *are* duplicate 2592 // entries of *some* 2593 // LOCAL symbols in the normal nlist data. To handle this situation 2594 // correctly, we must first attempt 2595 // to parse any DSC unmapped symbol information. If we find any, we set a 2596 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2597 2598 if (m_header.flags & 0x80000000u) { 2599 // Before we can start mapping the DSC, we need to make certain the 2600 // target process is actually using the cache we can find. 2601 2602 // Next we need to determine the correct path for the dyld shared cache. 2603 2604 ArchSpec header_arch; 2605 GetArchitecture(header_arch); 2606 char dsc_path[PATH_MAX]; 2607 char dsc_path_development[PATH_MAX]; 2608 2609 snprintf( 2610 dsc_path, sizeof(dsc_path), "%s%s%s", 2611 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2612 */ 2613 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2614 header_arch.GetArchitectureName()); 2615 2616 snprintf( 2617 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2618 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2619 */ 2620 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2621 header_arch.GetArchitectureName(), ".development"); 2622 2623 FileSpec dsc_nondevelopment_filespec(dsc_path, false); 2624 FileSpec dsc_development_filespec(dsc_path_development, false); 2625 FileSpec dsc_filespec; 2626 2627 UUID dsc_uuid; 2628 UUID process_shared_cache_uuid; 2629 addr_t process_shared_cache_base_addr; 2630 2631 if (process) { 2632 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, process_shared_cache_uuid); 2633 } 2634 2635 // First see if we can find an exact match for the inferior process 2636 // shared cache UUID in the development or non-development shared caches 2637 // on disk. 2638 if (process_shared_cache_uuid.IsValid()) { 2639 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2640 UUID dsc_development_uuid = GetSharedCacheUUID( 2641 dsc_development_filespec, byte_order, addr_byte_size); 2642 if (dsc_development_uuid.IsValid() && 2643 dsc_development_uuid == process_shared_cache_uuid) { 2644 dsc_filespec = dsc_development_filespec; 2645 dsc_uuid = dsc_development_uuid; 2646 } 2647 } 2648 if (!dsc_uuid.IsValid() && 2649 FileSystem::Instance().Exists(dsc_nondevelopment_filespec)) { 2650 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2651 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2652 if (dsc_nondevelopment_uuid.IsValid() && 2653 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2654 dsc_filespec = dsc_nondevelopment_filespec; 2655 dsc_uuid = dsc_nondevelopment_uuid; 2656 } 2657 } 2658 } 2659 2660 // Failing a UUID match, prefer the development dyld_shared cache if both 2661 // are present. 2662 if (!FileSystem::Instance().Exists(dsc_filespec)) { 2663 if (FileSystem::Instance().Exists(dsc_development_filespec)) { 2664 dsc_filespec = dsc_development_filespec; 2665 } else { 2666 dsc_filespec = dsc_nondevelopment_filespec; 2667 } 2668 } 2669 2670 /* The dyld_cache_header has a pointer to the 2671 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2672 The dyld_cache_local_symbols_info structure gives us three things: 2673 1. The start and count of the nlist records in the dyld_shared_cache 2674 file 2675 2. The start and size of the strings for these nlist records 2676 3. The start and count of dyld_cache_local_symbols_entry entries 2677 2678 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2679 dyld shared cache. 2680 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2681 the dyld shared cache. 2682 The dyld_cache_local_symbols_entry also lists the start of this 2683 dylib/framework's nlist records 2684 and the count of how many nlist records there are for this 2685 dylib/framework. 2686 */ 2687 2688 // Process the dyld shared cache header to find the unmapped symbols 2689 2690 DataBufferSP dsc_data_sp = MapFileData( 2691 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2692 if (!dsc_uuid.IsValid()) { 2693 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2694 } 2695 if (dsc_data_sp) { 2696 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2697 2698 bool uuid_match = true; 2699 if (dsc_uuid.IsValid() && process) { 2700 if (process_shared_cache_uuid.IsValid() && 2701 dsc_uuid != process_shared_cache_uuid) { 2702 // The on-disk dyld_shared_cache file is not the same as the one in 2703 // this process' memory, don't use it. 2704 uuid_match = false; 2705 ModuleSP module_sp(GetModule()); 2706 if (module_sp) 2707 module_sp->ReportWarning("process shared cache does not match " 2708 "on-disk dyld_shared_cache file, some " 2709 "symbol names will be missing."); 2710 } 2711 } 2712 2713 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2714 2715 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2716 2717 // If the mappingOffset points to a location inside the header, we've 2718 // opened an old dyld shared cache, and should not proceed further. 2719 if (uuid_match && 2720 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2721 2722 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2723 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2724 mappingOffset); 2725 2726 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2727 byte_order, addr_byte_size); 2728 offset = 0; 2729 2730 // The File addresses (from the in-memory Mach-O load commands) for 2731 // the shared libraries in the shared library cache need to be 2732 // adjusted by an offset to match up with the dylibOffset identifying 2733 // field in the dyld_cache_local_symbol_entry's. This offset is 2734 // recorded in mapping_offset_value. 2735 const uint64_t mapping_offset_value = 2736 dsc_mapping_info_data.GetU64(&offset); 2737 2738 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, 2739 localSymbolsOffset); 2740 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2741 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2742 2743 if (localSymbolsOffset && localSymbolsSize) { 2744 // Map the local symbols 2745 DataBufferSP dsc_local_symbols_data_sp = 2746 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2747 2748 if (dsc_local_symbols_data_sp) { 2749 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2750 byte_order, addr_byte_size); 2751 2752 offset = 0; 2753 2754 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2755 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2756 UndefinedNameToDescMap undefined_name_to_desc; 2757 SymbolIndexToName reexport_shlib_needs_fixup; 2758 2759 // Read the local_symbols_infos struct in one shot 2760 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2761 dsc_local_symbols_data.GetU32(&offset, 2762 &local_symbols_info.nlistOffset, 6); 2763 2764 SectionSP text_section_sp( 2765 section_list->FindSectionByName(GetSegmentNameTEXT())); 2766 2767 uint32_t header_file_offset = 2768 (text_section_sp->GetFileAddress() - mapping_offset_value); 2769 2770 offset = local_symbols_info.entriesOffset; 2771 for (uint32_t entry_index = 0; 2772 entry_index < local_symbols_info.entriesCount; 2773 entry_index++) { 2774 struct lldb_copy_dyld_cache_local_symbols_entry 2775 local_symbols_entry; 2776 local_symbols_entry.dylibOffset = 2777 dsc_local_symbols_data.GetU32(&offset); 2778 local_symbols_entry.nlistStartIndex = 2779 dsc_local_symbols_data.GetU32(&offset); 2780 local_symbols_entry.nlistCount = 2781 dsc_local_symbols_data.GetU32(&offset); 2782 2783 if (header_file_offset == local_symbols_entry.dylibOffset) { 2784 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2785 2786 // The normal nlist code cannot correctly size the Symbols 2787 // array, we need to allocate it here. 2788 sym = symtab->Resize( 2789 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2790 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2791 num_syms = symtab->GetNumSymbols(); 2792 2793 nlist_data_offset = 2794 local_symbols_info.nlistOffset + 2795 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2796 uint32_t string_table_offset = 2797 local_symbols_info.stringsOffset; 2798 2799 for (uint32_t nlist_index = 0; 2800 nlist_index < local_symbols_entry.nlistCount; 2801 nlist_index++) { 2802 ///////////////////////////// 2803 { 2804 struct nlist_64 nlist; 2805 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize( 2806 nlist_data_offset, nlist_byte_size)) 2807 break; 2808 2809 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked( 2810 &nlist_data_offset); 2811 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked( 2812 &nlist_data_offset); 2813 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked( 2814 &nlist_data_offset); 2815 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked( 2816 &nlist_data_offset); 2817 nlist.n_value = 2818 dsc_local_symbols_data.GetAddress_unchecked( 2819 &nlist_data_offset); 2820 2821 SymbolType type = eSymbolTypeInvalid; 2822 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2823 string_table_offset + nlist.n_strx); 2824 2825 if (symbol_name == NULL) { 2826 // No symbol should be NULL, even the symbols with no 2827 // string values should have an offset zero which 2828 // points to an empty C-string 2829 Host::SystemLog( 2830 Host::eSystemLogError, 2831 "error: DSC unmapped local symbol[%u] has invalid " 2832 "string table offset 0x%x in %s, ignoring symbol\n", 2833 entry_index, nlist.n_strx, 2834 module_sp->GetFileSpec().GetPath().c_str()); 2835 continue; 2836 } 2837 if (symbol_name[0] == '\0') 2838 symbol_name = NULL; 2839 2840 const char *symbol_name_non_abi_mangled = NULL; 2841 2842 SectionSP symbol_section; 2843 uint32_t symbol_byte_size = 0; 2844 bool add_nlist = true; 2845 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2846 bool demangled_is_synthesized = false; 2847 bool is_gsym = false; 2848 bool set_value = true; 2849 2850 assert(sym_idx < num_syms); 2851 2852 sym[sym_idx].SetDebug(is_debug); 2853 2854 if (is_debug) { 2855 switch (nlist.n_type) { 2856 case N_GSYM: 2857 // global symbol: name,,NO_SECT,type,0 2858 // Sometimes the N_GSYM value contains the address. 2859 2860 // FIXME: In the .o files, we have a GSYM and a debug 2861 // symbol for all the ObjC data. They 2862 // have the same address, but we want to ensure that 2863 // we always find only the real symbol, 'cause we 2864 // don't currently correctly attribute the 2865 // GSYM one to the ObjCClass/Ivar/MetaClass 2866 // symbol type. This is a temporary hack to make 2867 // sure the ObjectiveC symbols get treated correctly. 2868 // To do this right, we should coalesce all the GSYM 2869 // & global symbols that have the same address. 2870 2871 is_gsym = true; 2872 sym[sym_idx].SetExternal(true); 2873 2874 if (symbol_name && symbol_name[0] == '_' && 2875 symbol_name[1] == 'O') { 2876 llvm::StringRef symbol_name_ref(symbol_name); 2877 if (symbol_name_ref.startswith( 2878 g_objc_v2_prefix_class)) { 2879 symbol_name_non_abi_mangled = symbol_name + 1; 2880 symbol_name = 2881 symbol_name + g_objc_v2_prefix_class.size(); 2882 type = eSymbolTypeObjCClass; 2883 demangled_is_synthesized = true; 2884 2885 } else if (symbol_name_ref.startswith( 2886 g_objc_v2_prefix_metaclass)) { 2887 symbol_name_non_abi_mangled = symbol_name + 1; 2888 symbol_name = symbol_name + 2889 g_objc_v2_prefix_metaclass.size(); 2890 type = eSymbolTypeObjCMetaClass; 2891 demangled_is_synthesized = true; 2892 } else if (symbol_name_ref.startswith( 2893 g_objc_v2_prefix_ivar)) { 2894 symbol_name_non_abi_mangled = symbol_name + 1; 2895 symbol_name = 2896 symbol_name + g_objc_v2_prefix_ivar.size(); 2897 type = eSymbolTypeObjCIVar; 2898 demangled_is_synthesized = true; 2899 } 2900 } else { 2901 if (nlist.n_value != 0) 2902 symbol_section = section_info.GetSection( 2903 nlist.n_sect, nlist.n_value); 2904 type = eSymbolTypeData; 2905 } 2906 break; 2907 2908 case N_FNAME: 2909 // procedure name (f77 kludge): name,,NO_SECT,0,0 2910 type = eSymbolTypeCompiler; 2911 break; 2912 2913 case N_FUN: 2914 // procedure: name,,n_sect,linenumber,address 2915 if (symbol_name) { 2916 type = eSymbolTypeCode; 2917 symbol_section = section_info.GetSection( 2918 nlist.n_sect, nlist.n_value); 2919 2920 N_FUN_addr_to_sym_idx.insert( 2921 std::make_pair(nlist.n_value, sym_idx)); 2922 // We use the current number of symbols in the 2923 // symbol table in lieu of using nlist_idx in case 2924 // we ever start trimming entries out 2925 N_FUN_indexes.push_back(sym_idx); 2926 } else { 2927 type = eSymbolTypeCompiler; 2928 2929 if (!N_FUN_indexes.empty()) { 2930 // Copy the size of the function into the 2931 // original 2932 // STAB entry so we don't have 2933 // to hunt for it later 2934 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2935 ->SetByteSize(nlist.n_value); 2936 N_FUN_indexes.pop_back(); 2937 // We don't really need the end function STAB as 2938 // it contains the size which we already placed 2939 // with the original symbol, so don't add it if 2940 // we want a minimal symbol table 2941 add_nlist = false; 2942 } 2943 } 2944 break; 2945 2946 case N_STSYM: 2947 // static symbol: name,,n_sect,type,address 2948 N_STSYM_addr_to_sym_idx.insert( 2949 std::make_pair(nlist.n_value, sym_idx)); 2950 symbol_section = section_info.GetSection( 2951 nlist.n_sect, nlist.n_value); 2952 if (symbol_name && symbol_name[0]) { 2953 type = ObjectFile::GetSymbolTypeFromName( 2954 symbol_name + 1, eSymbolTypeData); 2955 } 2956 break; 2957 2958 case N_LCSYM: 2959 // .lcomm symbol: name,,n_sect,type,address 2960 symbol_section = section_info.GetSection( 2961 nlist.n_sect, nlist.n_value); 2962 type = eSymbolTypeCommonBlock; 2963 break; 2964 2965 case N_BNSYM: 2966 // We use the current number of symbols in the symbol 2967 // table in lieu of using nlist_idx in case we ever 2968 // start trimming entries out Skip these if we want 2969 // minimal symbol tables 2970 add_nlist = false; 2971 break; 2972 2973 case N_ENSYM: 2974 // Set the size of the N_BNSYM to the terminating 2975 // index of this N_ENSYM so that we can always skip 2976 // the entire symbol if we need to navigate more 2977 // quickly at the source level when parsing STABS 2978 // Skip these if we want minimal symbol tables 2979 add_nlist = false; 2980 break; 2981 2982 case N_OPT: 2983 // emitted with gcc2_compiled and in gcc source 2984 type = eSymbolTypeCompiler; 2985 break; 2986 2987 case N_RSYM: 2988 // register sym: name,,NO_SECT,type,register 2989 type = eSymbolTypeVariable; 2990 break; 2991 2992 case N_SLINE: 2993 // src line: 0,,n_sect,linenumber,address 2994 symbol_section = section_info.GetSection( 2995 nlist.n_sect, nlist.n_value); 2996 type = eSymbolTypeLineEntry; 2997 break; 2998 2999 case N_SSYM: 3000 // structure elt: name,,NO_SECT,type,struct_offset 3001 type = eSymbolTypeVariableType; 3002 break; 3003 3004 case N_SO: 3005 // source file name 3006 type = eSymbolTypeSourceFile; 3007 if (symbol_name == NULL) { 3008 add_nlist = false; 3009 if (N_SO_index != UINT32_MAX) { 3010 // Set the size of the N_SO to the terminating 3011 // index of this N_SO so that we can always skip 3012 // the entire N_SO if we need to navigate more 3013 // quickly at the source level when parsing STABS 3014 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3015 symbol_ptr->SetByteSize(sym_idx); 3016 symbol_ptr->SetSizeIsSibling(true); 3017 } 3018 N_NSYM_indexes.clear(); 3019 N_INCL_indexes.clear(); 3020 N_BRAC_indexes.clear(); 3021 N_COMM_indexes.clear(); 3022 N_FUN_indexes.clear(); 3023 N_SO_index = UINT32_MAX; 3024 } else { 3025 // We use the current number of symbols in the 3026 // symbol table in lieu of using nlist_idx in case 3027 // we ever start trimming entries out 3028 const bool N_SO_has_full_path = 3029 symbol_name[0] == '/'; 3030 if (N_SO_has_full_path) { 3031 if ((N_SO_index == sym_idx - 1) && 3032 ((sym_idx - 1) < num_syms)) { 3033 // We have two consecutive N_SO entries where 3034 // the first contains a directory and the 3035 // second contains a full path. 3036 sym[sym_idx - 1].GetMangled().SetValue( 3037 ConstString(symbol_name), false); 3038 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3039 add_nlist = false; 3040 } else { 3041 // This is the first entry in a N_SO that 3042 // contains a directory or 3043 // a full path to the source file 3044 N_SO_index = sym_idx; 3045 } 3046 } else if ((N_SO_index == sym_idx - 1) && 3047 ((sym_idx - 1) < num_syms)) { 3048 // This is usually the second N_SO entry that 3049 // contains just the filename, so here we combine 3050 // it with the first one if we are minimizing the 3051 // symbol table 3052 const char *so_path = 3053 sym[sym_idx - 1] 3054 .GetMangled() 3055 .GetDemangledName( 3056 lldb::eLanguageTypeUnknown) 3057 .AsCString(); 3058 if (so_path && so_path[0]) { 3059 std::string full_so_path(so_path); 3060 const size_t double_slash_pos = 3061 full_so_path.find("//"); 3062 if (double_slash_pos != std::string::npos) { 3063 // The linker has been generating bad N_SO 3064 // entries with doubled up paths 3065 // in the format "%s%s" where the first 3066 // string in the DW_AT_comp_dir, and the 3067 // second is the directory for the source 3068 // file so you end up with a path that looks 3069 // like "/tmp/src//tmp/src/" 3070 FileSpec so_dir(so_path, false); 3071 if (!FileSystem::Instance().Exists(so_dir)) { 3072 so_dir.SetFile( 3073 &full_so_path[double_slash_pos + 1], 3074 false); 3075 if (FileSystem::Instance().Exists(so_dir)) { 3076 // Trim off the incorrect path 3077 full_so_path.erase(0, 3078 double_slash_pos + 1); 3079 } 3080 } 3081 } 3082 if (*full_so_path.rbegin() != '/') 3083 full_so_path += '/'; 3084 full_so_path += symbol_name; 3085 sym[sym_idx - 1].GetMangled().SetValue( 3086 ConstString(full_so_path.c_str()), false); 3087 add_nlist = false; 3088 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3089 } 3090 } else { 3091 // This could be a relative path to a N_SO 3092 N_SO_index = sym_idx; 3093 } 3094 } 3095 break; 3096 3097 case N_OSO: 3098 // object file name: name,,0,0,st_mtime 3099 type = eSymbolTypeObjectFile; 3100 break; 3101 3102 case N_LSYM: 3103 // local sym: name,,NO_SECT,type,offset 3104 type = eSymbolTypeLocal; 3105 break; 3106 3107 //---------------------------------------------------------------------- 3108 // INCL scopes 3109 //---------------------------------------------------------------------- 3110 case N_BINCL: 3111 // include file beginning: name,,NO_SECT,0,sum We use 3112 // the current number of symbols in the symbol table 3113 // in lieu of using nlist_idx in case we ever start 3114 // trimming entries out 3115 N_INCL_indexes.push_back(sym_idx); 3116 type = eSymbolTypeScopeBegin; 3117 break; 3118 3119 case N_EINCL: 3120 // include file end: name,,NO_SECT,0,0 3121 // Set the size of the N_BINCL to the terminating 3122 // index of this N_EINCL so that we can always skip 3123 // the entire symbol if we need to navigate more 3124 // quickly at the source level when parsing STABS 3125 if (!N_INCL_indexes.empty()) { 3126 symbol_ptr = 3127 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3128 symbol_ptr->SetByteSize(sym_idx + 1); 3129 symbol_ptr->SetSizeIsSibling(true); 3130 N_INCL_indexes.pop_back(); 3131 } 3132 type = eSymbolTypeScopeEnd; 3133 break; 3134 3135 case N_SOL: 3136 // #included file name: name,,n_sect,0,address 3137 type = eSymbolTypeHeaderFile; 3138 3139 // We currently don't use the header files on darwin 3140 add_nlist = false; 3141 break; 3142 3143 case N_PARAMS: 3144 // compiler parameters: name,,NO_SECT,0,0 3145 type = eSymbolTypeCompiler; 3146 break; 3147 3148 case N_VERSION: 3149 // compiler version: name,,NO_SECT,0,0 3150 type = eSymbolTypeCompiler; 3151 break; 3152 3153 case N_OLEVEL: 3154 // compiler -O level: name,,NO_SECT,0,0 3155 type = eSymbolTypeCompiler; 3156 break; 3157 3158 case N_PSYM: 3159 // parameter: name,,NO_SECT,type,offset 3160 type = eSymbolTypeVariable; 3161 break; 3162 3163 case N_ENTRY: 3164 // alternate entry: name,,n_sect,linenumber,address 3165 symbol_section = section_info.GetSection( 3166 nlist.n_sect, nlist.n_value); 3167 type = eSymbolTypeLineEntry; 3168 break; 3169 3170 //---------------------------------------------------------------------- 3171 // Left and Right Braces 3172 //---------------------------------------------------------------------- 3173 case N_LBRAC: 3174 // left bracket: 0,,NO_SECT,nesting level,address We 3175 // use the current number of symbols in the symbol 3176 // table in lieu of using nlist_idx in case we ever 3177 // start trimming entries out 3178 symbol_section = section_info.GetSection( 3179 nlist.n_sect, nlist.n_value); 3180 N_BRAC_indexes.push_back(sym_idx); 3181 type = eSymbolTypeScopeBegin; 3182 break; 3183 3184 case N_RBRAC: 3185 // right bracket: 0,,NO_SECT,nesting level,address 3186 // Set the size of the N_LBRAC to the terminating 3187 // index of this N_RBRAC so that we can always skip 3188 // the entire symbol if we need to navigate more 3189 // quickly at the source level when parsing STABS 3190 symbol_section = section_info.GetSection( 3191 nlist.n_sect, nlist.n_value); 3192 if (!N_BRAC_indexes.empty()) { 3193 symbol_ptr = 3194 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3195 symbol_ptr->SetByteSize(sym_idx + 1); 3196 symbol_ptr->SetSizeIsSibling(true); 3197 N_BRAC_indexes.pop_back(); 3198 } 3199 type = eSymbolTypeScopeEnd; 3200 break; 3201 3202 case N_EXCL: 3203 // deleted include file: name,,NO_SECT,0,sum 3204 type = eSymbolTypeHeaderFile; 3205 break; 3206 3207 //---------------------------------------------------------------------- 3208 // COMM scopes 3209 //---------------------------------------------------------------------- 3210 case N_BCOMM: 3211 // begin common: name,,NO_SECT,0,0 3212 // We use the current number of symbols in the symbol 3213 // table in lieu of using nlist_idx in case we ever 3214 // start trimming entries out 3215 type = eSymbolTypeScopeBegin; 3216 N_COMM_indexes.push_back(sym_idx); 3217 break; 3218 3219 case N_ECOML: 3220 // end common (local name): 0,,n_sect,0,address 3221 symbol_section = section_info.GetSection( 3222 nlist.n_sect, nlist.n_value); 3223 // Fall through 3224 3225 case N_ECOMM: 3226 // end common: name,,n_sect,0,0 3227 // Set the size of the N_BCOMM to the terminating 3228 // index of this N_ECOMM/N_ECOML so that we can 3229 // always skip the entire symbol if we need to 3230 // navigate more quickly at the source level when 3231 // parsing STABS 3232 if (!N_COMM_indexes.empty()) { 3233 symbol_ptr = 3234 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3235 symbol_ptr->SetByteSize(sym_idx + 1); 3236 symbol_ptr->SetSizeIsSibling(true); 3237 N_COMM_indexes.pop_back(); 3238 } 3239 type = eSymbolTypeScopeEnd; 3240 break; 3241 3242 case N_LENG: 3243 // second stab entry with length information 3244 type = eSymbolTypeAdditional; 3245 break; 3246 3247 default: 3248 break; 3249 } 3250 } else { 3251 // uint8_t n_pext = N_PEXT & nlist.n_type; 3252 uint8_t n_type = N_TYPE & nlist.n_type; 3253 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3254 3255 switch (n_type) { 3256 case N_INDR: { 3257 const char *reexport_name_cstr = 3258 strtab_data.PeekCStr(nlist.n_value); 3259 if (reexport_name_cstr && reexport_name_cstr[0]) { 3260 type = eSymbolTypeReExported; 3261 ConstString reexport_name( 3262 reexport_name_cstr + 3263 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3264 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3265 set_value = false; 3266 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3267 indirect_symbol_names.insert( 3268 ConstString(symbol_name + 3269 ((symbol_name[0] == '_') ? 1 : 0))); 3270 } else 3271 type = eSymbolTypeUndefined; 3272 } break; 3273 3274 case N_UNDF: 3275 if (symbol_name && symbol_name[0]) { 3276 ConstString undefined_name( 3277 symbol_name + 3278 ((symbol_name[0] == '_') ? 1 : 0)); 3279 undefined_name_to_desc[undefined_name] = 3280 nlist.n_desc; 3281 } 3282 // Fall through 3283 case N_PBUD: 3284 type = eSymbolTypeUndefined; 3285 break; 3286 3287 case N_ABS: 3288 type = eSymbolTypeAbsolute; 3289 break; 3290 3291 case N_SECT: { 3292 symbol_section = section_info.GetSection( 3293 nlist.n_sect, nlist.n_value); 3294 3295 if (symbol_section == NULL) { 3296 // TODO: warn about this? 3297 add_nlist = false; 3298 break; 3299 } 3300 3301 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3302 type = eSymbolTypeException; 3303 } else { 3304 uint32_t section_type = 3305 symbol_section->Get() & SECTION_TYPE; 3306 3307 switch (section_type) { 3308 case S_CSTRING_LITERALS: 3309 type = eSymbolTypeData; 3310 break; // section with only literal C strings 3311 case S_4BYTE_LITERALS: 3312 type = eSymbolTypeData; 3313 break; // section with only 4 byte literals 3314 case S_8BYTE_LITERALS: 3315 type = eSymbolTypeData; 3316 break; // section with only 8 byte literals 3317 case S_LITERAL_POINTERS: 3318 type = eSymbolTypeTrampoline; 3319 break; // section with only pointers to literals 3320 case S_NON_LAZY_SYMBOL_POINTERS: 3321 type = eSymbolTypeTrampoline; 3322 break; // section with only non-lazy symbol 3323 // pointers 3324 case S_LAZY_SYMBOL_POINTERS: 3325 type = eSymbolTypeTrampoline; 3326 break; // section with only lazy symbol pointers 3327 case S_SYMBOL_STUBS: 3328 type = eSymbolTypeTrampoline; 3329 break; // section with only symbol stubs, byte 3330 // size of stub in the reserved2 field 3331 case S_MOD_INIT_FUNC_POINTERS: 3332 type = eSymbolTypeCode; 3333 break; // section with only function pointers for 3334 // initialization 3335 case S_MOD_TERM_FUNC_POINTERS: 3336 type = eSymbolTypeCode; 3337 break; // section with only function pointers for 3338 // termination 3339 case S_INTERPOSING: 3340 type = eSymbolTypeTrampoline; 3341 break; // section with only pairs of function 3342 // pointers for interposing 3343 case S_16BYTE_LITERALS: 3344 type = eSymbolTypeData; 3345 break; // section with only 16 byte literals 3346 case S_DTRACE_DOF: 3347 type = eSymbolTypeInstrumentation; 3348 break; 3349 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3350 type = eSymbolTypeTrampoline; 3351 break; 3352 default: 3353 switch (symbol_section->GetType()) { 3354 case lldb::eSectionTypeCode: 3355 type = eSymbolTypeCode; 3356 break; 3357 case eSectionTypeData: 3358 case eSectionTypeDataCString: // Inlined C string 3359 // data 3360 case eSectionTypeDataCStringPointers: // Pointers 3361 // to C 3362 // string 3363 // data 3364 case eSectionTypeDataSymbolAddress: // Address of 3365 // a symbol in 3366 // the symbol 3367 // table 3368 case eSectionTypeData4: 3369 case eSectionTypeData8: 3370 case eSectionTypeData16: 3371 type = eSymbolTypeData; 3372 break; 3373 default: 3374 break; 3375 } 3376 break; 3377 } 3378 3379 if (type == eSymbolTypeInvalid) { 3380 const char *symbol_sect_name = 3381 symbol_section->GetName().AsCString(); 3382 if (symbol_section->IsDescendant( 3383 text_section_sp.get())) { 3384 if (symbol_section->IsClear( 3385 S_ATTR_PURE_INSTRUCTIONS | 3386 S_ATTR_SELF_MODIFYING_CODE | 3387 S_ATTR_SOME_INSTRUCTIONS)) 3388 type = eSymbolTypeData; 3389 else 3390 type = eSymbolTypeCode; 3391 } else if (symbol_section->IsDescendant( 3392 data_section_sp.get()) || 3393 symbol_section->IsDescendant( 3394 data_dirty_section_sp.get()) || 3395 symbol_section->IsDescendant( 3396 data_const_section_sp.get())) { 3397 if (symbol_sect_name && 3398 ::strstr(symbol_sect_name, "__objc") == 3399 symbol_sect_name) { 3400 type = eSymbolTypeRuntime; 3401 3402 if (symbol_name) { 3403 llvm::StringRef symbol_name_ref( 3404 symbol_name); 3405 if (symbol_name_ref.startswith("_OBJC_")) { 3406 static const llvm::StringRef 3407 g_objc_v2_prefix_class( 3408 "_OBJC_CLASS_$_"); 3409 static const llvm::StringRef 3410 g_objc_v2_prefix_metaclass( 3411 "_OBJC_METACLASS_$_"); 3412 static const llvm::StringRef 3413 g_objc_v2_prefix_ivar( 3414 "_OBJC_IVAR_$_"); 3415 if (symbol_name_ref.startswith( 3416 g_objc_v2_prefix_class)) { 3417 symbol_name_non_abi_mangled = 3418 symbol_name + 1; 3419 symbol_name = 3420 symbol_name + 3421 g_objc_v2_prefix_class.size(); 3422 type = eSymbolTypeObjCClass; 3423 demangled_is_synthesized = true; 3424 } else if ( 3425 symbol_name_ref.startswith( 3426 g_objc_v2_prefix_metaclass)) { 3427 symbol_name_non_abi_mangled = 3428 symbol_name + 1; 3429 symbol_name = 3430 symbol_name + 3431 g_objc_v2_prefix_metaclass.size(); 3432 type = eSymbolTypeObjCMetaClass; 3433 demangled_is_synthesized = true; 3434 } else if (symbol_name_ref.startswith( 3435 g_objc_v2_prefix_ivar)) { 3436 symbol_name_non_abi_mangled = 3437 symbol_name + 1; 3438 symbol_name = 3439 symbol_name + 3440 g_objc_v2_prefix_ivar.size(); 3441 type = eSymbolTypeObjCIVar; 3442 demangled_is_synthesized = true; 3443 } 3444 } 3445 } 3446 } else if (symbol_sect_name && 3447 ::strstr(symbol_sect_name, 3448 "__gcc_except_tab") == 3449 symbol_sect_name) { 3450 type = eSymbolTypeException; 3451 } else { 3452 type = eSymbolTypeData; 3453 } 3454 } else if (symbol_sect_name && 3455 ::strstr(symbol_sect_name, 3456 "__IMPORT") == 3457 symbol_sect_name) { 3458 type = eSymbolTypeTrampoline; 3459 } else if (symbol_section->IsDescendant( 3460 objc_section_sp.get())) { 3461 type = eSymbolTypeRuntime; 3462 if (symbol_name && symbol_name[0] == '.') { 3463 llvm::StringRef symbol_name_ref(symbol_name); 3464 static const llvm::StringRef 3465 g_objc_v1_prefix_class( 3466 ".objc_class_name_"); 3467 if (symbol_name_ref.startswith( 3468 g_objc_v1_prefix_class)) { 3469 symbol_name_non_abi_mangled = symbol_name; 3470 symbol_name = symbol_name + 3471 g_objc_v1_prefix_class.size(); 3472 type = eSymbolTypeObjCClass; 3473 demangled_is_synthesized = true; 3474 } 3475 } 3476 } 3477 } 3478 } 3479 } break; 3480 } 3481 } 3482 3483 if (add_nlist) { 3484 uint64_t symbol_value = nlist.n_value; 3485 if (symbol_name_non_abi_mangled) { 3486 sym[sym_idx].GetMangled().SetMangledName( 3487 ConstString(symbol_name_non_abi_mangled)); 3488 sym[sym_idx].GetMangled().SetDemangledName( 3489 ConstString(symbol_name)); 3490 } else { 3491 bool symbol_name_is_mangled = false; 3492 3493 if (symbol_name && symbol_name[0] == '_') { 3494 symbol_name_is_mangled = symbol_name[1] == '_'; 3495 symbol_name++; // Skip the leading underscore 3496 } 3497 3498 if (symbol_name) { 3499 ConstString const_symbol_name(symbol_name); 3500 sym[sym_idx].GetMangled().SetValue( 3501 const_symbol_name, symbol_name_is_mangled); 3502 if (is_gsym && is_debug) { 3503 const char *gsym_name = 3504 sym[sym_idx] 3505 .GetMangled() 3506 .GetName(lldb::eLanguageTypeUnknown, 3507 Mangled::ePreferMangled) 3508 .GetCString(); 3509 if (gsym_name) 3510 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3511 } 3512 } 3513 } 3514 if (symbol_section) { 3515 const addr_t section_file_addr = 3516 symbol_section->GetFileAddress(); 3517 if (symbol_byte_size == 0 && 3518 function_starts_count > 0) { 3519 addr_t symbol_lookup_file_addr = nlist.n_value; 3520 // Do an exact address match for non-ARM addresses, 3521 // else get the closest since the symbol might be a 3522 // thumb symbol which has an address with bit zero 3523 // set 3524 FunctionStarts::Entry *func_start_entry = 3525 function_starts.FindEntry( 3526 symbol_lookup_file_addr, !is_arm); 3527 if (is_arm && func_start_entry) { 3528 // Verify that the function start address is the 3529 // symbol address (ARM) or the symbol address + 1 3530 // (thumb) 3531 if (func_start_entry->addr != 3532 symbol_lookup_file_addr && 3533 func_start_entry->addr != 3534 (symbol_lookup_file_addr + 1)) { 3535 // Not the right entry, NULL it out... 3536 func_start_entry = NULL; 3537 } 3538 } 3539 if (func_start_entry) { 3540 func_start_entry->data = true; 3541 3542 addr_t symbol_file_addr = func_start_entry->addr; 3543 uint32_t symbol_flags = 0; 3544 if (is_arm) { 3545 if (symbol_file_addr & 1) 3546 symbol_flags = 3547 MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3548 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3549 } 3550 3551 const FunctionStarts::Entry 3552 *next_func_start_entry = 3553 function_starts.FindNextEntry( 3554 func_start_entry); 3555 const addr_t section_end_file_addr = 3556 section_file_addr + 3557 symbol_section->GetByteSize(); 3558 if (next_func_start_entry) { 3559 addr_t next_symbol_file_addr = 3560 next_func_start_entry->addr; 3561 // Be sure the clear the Thumb address bit when 3562 // we calculate the size from the current and 3563 // next address 3564 if (is_arm) 3565 next_symbol_file_addr &= 3566 THUMB_ADDRESS_BIT_MASK; 3567 symbol_byte_size = std::min<lldb::addr_t>( 3568 next_symbol_file_addr - symbol_file_addr, 3569 section_end_file_addr - symbol_file_addr); 3570 } else { 3571 symbol_byte_size = 3572 section_end_file_addr - symbol_file_addr; 3573 } 3574 } 3575 } 3576 symbol_value -= section_file_addr; 3577 } 3578 3579 if (is_debug == false) { 3580 if (type == eSymbolTypeCode) { 3581 // See if we can find a N_FUN entry for any code 3582 // symbols. If we do find a match, and the name 3583 // matches, then we can merge the two into just the 3584 // function symbol to avoid duplicate entries in 3585 // the symbol table 3586 std::pair<ValueToSymbolIndexMap::const_iterator, 3587 ValueToSymbolIndexMap::const_iterator> 3588 range; 3589 range = N_FUN_addr_to_sym_idx.equal_range( 3590 nlist.n_value); 3591 if (range.first != range.second) { 3592 bool found_it = false; 3593 for (ValueToSymbolIndexMap::const_iterator pos = 3594 range.first; 3595 pos != range.second; ++pos) { 3596 if (sym[sym_idx].GetMangled().GetName( 3597 lldb::eLanguageTypeUnknown, 3598 Mangled::ePreferMangled) == 3599 sym[pos->second].GetMangled().GetName( 3600 lldb::eLanguageTypeUnknown, 3601 Mangled::ePreferMangled)) { 3602 m_nlist_idx_to_sym_idx[nlist_idx] = 3603 pos->second; 3604 // We just need the flags from the linker 3605 // symbol, so put these flags 3606 // into the N_FUN flags to avoid duplicate 3607 // symbols in the symbol table 3608 sym[pos->second].SetExternal( 3609 sym[sym_idx].IsExternal()); 3610 sym[pos->second].SetFlags(nlist.n_type << 16 | 3611 nlist.n_desc); 3612 if (resolver_addresses.find(nlist.n_value) != 3613 resolver_addresses.end()) 3614 sym[pos->second].SetType( 3615 eSymbolTypeResolver); 3616 sym[sym_idx].Clear(); 3617 found_it = true; 3618 break; 3619 } 3620 } 3621 if (found_it) 3622 continue; 3623 } else { 3624 if (resolver_addresses.find(nlist.n_value) != 3625 resolver_addresses.end()) 3626 type = eSymbolTypeResolver; 3627 } 3628 } else if (type == eSymbolTypeData || 3629 type == eSymbolTypeObjCClass || 3630 type == eSymbolTypeObjCMetaClass || 3631 type == eSymbolTypeObjCIVar) { 3632 // See if we can find a N_STSYM entry for any data 3633 // symbols. If we do find a match, and the name 3634 // matches, then we can merge the two into just the 3635 // Static symbol to avoid duplicate entries in the 3636 // symbol table 3637 std::pair<ValueToSymbolIndexMap::const_iterator, 3638 ValueToSymbolIndexMap::const_iterator> 3639 range; 3640 range = N_STSYM_addr_to_sym_idx.equal_range( 3641 nlist.n_value); 3642 if (range.first != range.second) { 3643 bool found_it = false; 3644 for (ValueToSymbolIndexMap::const_iterator pos = 3645 range.first; 3646 pos != range.second; ++pos) { 3647 if (sym[sym_idx].GetMangled().GetName( 3648 lldb::eLanguageTypeUnknown, 3649 Mangled::ePreferMangled) == 3650 sym[pos->second].GetMangled().GetName( 3651 lldb::eLanguageTypeUnknown, 3652 Mangled::ePreferMangled)) { 3653 m_nlist_idx_to_sym_idx[nlist_idx] = 3654 pos->second; 3655 // We just need the flags from the linker 3656 // symbol, so put these flags 3657 // into the N_STSYM flags to avoid duplicate 3658 // symbols in the symbol table 3659 sym[pos->second].SetExternal( 3660 sym[sym_idx].IsExternal()); 3661 sym[pos->second].SetFlags(nlist.n_type << 16 | 3662 nlist.n_desc); 3663 sym[sym_idx].Clear(); 3664 found_it = true; 3665 break; 3666 } 3667 } 3668 if (found_it) 3669 continue; 3670 } else { 3671 const char *gsym_name = 3672 sym[sym_idx] 3673 .GetMangled() 3674 .GetName(lldb::eLanguageTypeUnknown, 3675 Mangled::ePreferMangled) 3676 .GetCString(); 3677 if (gsym_name) { 3678 // Combine N_GSYM stab entries with the non 3679 // stab symbol 3680 ConstNameToSymbolIndexMap::const_iterator pos = 3681 N_GSYM_name_to_sym_idx.find(gsym_name); 3682 if (pos != N_GSYM_name_to_sym_idx.end()) { 3683 const uint32_t GSYM_sym_idx = pos->second; 3684 m_nlist_idx_to_sym_idx[nlist_idx] = 3685 GSYM_sym_idx; 3686 // Copy the address, because often the N_GSYM 3687 // address has an invalid address of zero 3688 // when the global is a common symbol 3689 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3690 symbol_section); 3691 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3692 symbol_value); 3693 // We just need the flags from the linker 3694 // symbol, so put these flags 3695 // into the N_GSYM flags to avoid duplicate 3696 // symbols in the symbol table 3697 sym[GSYM_sym_idx].SetFlags( 3698 nlist.n_type << 16 | nlist.n_desc); 3699 sym[sym_idx].Clear(); 3700 continue; 3701 } 3702 } 3703 } 3704 } 3705 } 3706 3707 sym[sym_idx].SetID(nlist_idx); 3708 sym[sym_idx].SetType(type); 3709 if (set_value) { 3710 sym[sym_idx].GetAddressRef().SetSection( 3711 symbol_section); 3712 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3713 } 3714 sym[sym_idx].SetFlags(nlist.n_type << 16 | 3715 nlist.n_desc); 3716 3717 if (symbol_byte_size > 0) 3718 sym[sym_idx].SetByteSize(symbol_byte_size); 3719 3720 if (demangled_is_synthesized) 3721 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3722 ++sym_idx; 3723 } else { 3724 sym[sym_idx].Clear(); 3725 } 3726 } 3727 ///////////////////////////// 3728 } 3729 break; // No more entries to consider 3730 } 3731 } 3732 3733 for (const auto &pos : reexport_shlib_needs_fixup) { 3734 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3735 if (undef_pos != undefined_name_to_desc.end()) { 3736 const uint8_t dylib_ordinal = 3737 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3738 if (dylib_ordinal > 0 && 3739 dylib_ordinal < dylib_files.GetSize()) 3740 sym[pos.first].SetReExportedSymbolSharedLibrary( 3741 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3742 } 3743 } 3744 } 3745 } 3746 } 3747 } 3748 } 3749 3750 // Must reset this in case it was mutated above! 3751 nlist_data_offset = 0; 3752 #endif 3753 3754 if (nlist_data.GetByteSize() > 0) { 3755 3756 // If the sym array was not created while parsing the DSC unmapped 3757 // symbols, create it now. 3758 if (sym == NULL) { 3759 sym = symtab->Resize(symtab_load_command.nsyms + 3760 m_dysymtab.nindirectsyms); 3761 num_syms = symtab->GetNumSymbols(); 3762 } 3763 3764 if (unmapped_local_symbols_found) { 3765 assert(m_dysymtab.ilocalsym == 0); 3766 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3767 nlist_idx = m_dysymtab.nlocalsym; 3768 } else { 3769 nlist_idx = 0; 3770 } 3771 3772 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3773 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3774 UndefinedNameToDescMap undefined_name_to_desc; 3775 SymbolIndexToName reexport_shlib_needs_fixup; 3776 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 3777 struct nlist_64 nlist; 3778 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, 3779 nlist_byte_size)) 3780 break; 3781 3782 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3783 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 3784 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 3785 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 3786 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 3787 3788 SymbolType type = eSymbolTypeInvalid; 3789 const char *symbol_name = NULL; 3790 3791 if (have_strtab_data) { 3792 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3793 3794 if (symbol_name == NULL) { 3795 // No symbol should be NULL, even the symbols with no string values 3796 // should have an offset zero which points to an empty C-string 3797 Host::SystemLog(Host::eSystemLogError, 3798 "error: symbol[%u] has invalid string table offset " 3799 "0x%x in %s, ignoring symbol\n", 3800 nlist_idx, nlist.n_strx, 3801 module_sp->GetFileSpec().GetPath().c_str()); 3802 continue; 3803 } 3804 if (symbol_name[0] == '\0') 3805 symbol_name = NULL; 3806 } else { 3807 const addr_t str_addr = strtab_addr + nlist.n_strx; 3808 Status str_error; 3809 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3810 str_error)) 3811 symbol_name = memory_symbol_name.c_str(); 3812 } 3813 const char *symbol_name_non_abi_mangled = NULL; 3814 3815 SectionSP symbol_section; 3816 lldb::addr_t symbol_byte_size = 0; 3817 bool add_nlist = true; 3818 bool is_gsym = false; 3819 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3820 bool demangled_is_synthesized = false; 3821 bool set_value = true; 3822 assert(sym_idx < num_syms); 3823 3824 sym[sym_idx].SetDebug(is_debug); 3825 3826 if (is_debug) { 3827 switch (nlist.n_type) { 3828 case N_GSYM: 3829 // global symbol: name,,NO_SECT,type,0 3830 // Sometimes the N_GSYM value contains the address. 3831 3832 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3833 // the ObjC data. They 3834 // have the same address, but we want to ensure that we always find 3835 // only the real symbol, 'cause we don't currently correctly 3836 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3837 // type. This is a temporary hack to make sure the ObjectiveC 3838 // symbols get treated correctly. To do this right, we should 3839 // coalesce all the GSYM & global symbols that have the same 3840 // address. 3841 is_gsym = true; 3842 sym[sym_idx].SetExternal(true); 3843 3844 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3845 llvm::StringRef symbol_name_ref(symbol_name); 3846 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3847 symbol_name_non_abi_mangled = symbol_name + 1; 3848 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3849 type = eSymbolTypeObjCClass; 3850 demangled_is_synthesized = true; 3851 3852 } else if (symbol_name_ref.startswith( 3853 g_objc_v2_prefix_metaclass)) { 3854 symbol_name_non_abi_mangled = symbol_name + 1; 3855 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3856 type = eSymbolTypeObjCMetaClass; 3857 demangled_is_synthesized = true; 3858 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3859 symbol_name_non_abi_mangled = symbol_name + 1; 3860 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3861 type = eSymbolTypeObjCIVar; 3862 demangled_is_synthesized = true; 3863 } 3864 } else { 3865 if (nlist.n_value != 0) 3866 symbol_section = 3867 section_info.GetSection(nlist.n_sect, nlist.n_value); 3868 type = eSymbolTypeData; 3869 } 3870 break; 3871 3872 case N_FNAME: 3873 // procedure name (f77 kludge): name,,NO_SECT,0,0 3874 type = eSymbolTypeCompiler; 3875 break; 3876 3877 case N_FUN: 3878 // procedure: name,,n_sect,linenumber,address 3879 if (symbol_name) { 3880 type = eSymbolTypeCode; 3881 symbol_section = 3882 section_info.GetSection(nlist.n_sect, nlist.n_value); 3883 3884 N_FUN_addr_to_sym_idx.insert( 3885 std::make_pair(nlist.n_value, sym_idx)); 3886 // We use the current number of symbols in the symbol table in 3887 // lieu of using nlist_idx in case we ever start trimming entries 3888 // out 3889 N_FUN_indexes.push_back(sym_idx); 3890 } else { 3891 type = eSymbolTypeCompiler; 3892 3893 if (!N_FUN_indexes.empty()) { 3894 // Copy the size of the function into the original STAB entry 3895 // so we don't have to hunt for it later 3896 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3897 ->SetByteSize(nlist.n_value); 3898 N_FUN_indexes.pop_back(); 3899 // We don't really need the end function STAB as it contains 3900 // the size which we already placed with the original symbol, 3901 // so don't add it if we want a minimal symbol table 3902 add_nlist = false; 3903 } 3904 } 3905 break; 3906 3907 case N_STSYM: 3908 // static symbol: name,,n_sect,type,address 3909 N_STSYM_addr_to_sym_idx.insert( 3910 std::make_pair(nlist.n_value, sym_idx)); 3911 symbol_section = 3912 section_info.GetSection(nlist.n_sect, nlist.n_value); 3913 if (symbol_name && symbol_name[0]) { 3914 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3915 eSymbolTypeData); 3916 } 3917 break; 3918 3919 case N_LCSYM: 3920 // .lcomm symbol: name,,n_sect,type,address 3921 symbol_section = 3922 section_info.GetSection(nlist.n_sect, nlist.n_value); 3923 type = eSymbolTypeCommonBlock; 3924 break; 3925 3926 case N_BNSYM: 3927 // We use the current number of symbols in the symbol table in lieu 3928 // of using nlist_idx in case we ever start trimming entries out 3929 // Skip these if we want minimal symbol tables 3930 add_nlist = false; 3931 break; 3932 3933 case N_ENSYM: 3934 // Set the size of the N_BNSYM to the terminating index of this 3935 // N_ENSYM so that we can always skip the entire symbol if we need 3936 // to navigate more quickly at the source level when parsing STABS 3937 // Skip these if we want minimal symbol tables 3938 add_nlist = false; 3939 break; 3940 3941 case N_OPT: 3942 // emitted with gcc2_compiled and in gcc source 3943 type = eSymbolTypeCompiler; 3944 break; 3945 3946 case N_RSYM: 3947 // register sym: name,,NO_SECT,type,register 3948 type = eSymbolTypeVariable; 3949 break; 3950 3951 case N_SLINE: 3952 // src line: 0,,n_sect,linenumber,address 3953 symbol_section = 3954 section_info.GetSection(nlist.n_sect, nlist.n_value); 3955 type = eSymbolTypeLineEntry; 3956 break; 3957 3958 case N_SSYM: 3959 // structure elt: name,,NO_SECT,type,struct_offset 3960 type = eSymbolTypeVariableType; 3961 break; 3962 3963 case N_SO: 3964 // source file name 3965 type = eSymbolTypeSourceFile; 3966 if (symbol_name == NULL) { 3967 add_nlist = false; 3968 if (N_SO_index != UINT32_MAX) { 3969 // Set the size of the N_SO to the terminating index of this 3970 // N_SO so that we can always skip the entire N_SO if we need 3971 // to navigate more quickly at the source level when parsing 3972 // STABS 3973 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3974 symbol_ptr->SetByteSize(sym_idx); 3975 symbol_ptr->SetSizeIsSibling(true); 3976 } 3977 N_NSYM_indexes.clear(); 3978 N_INCL_indexes.clear(); 3979 N_BRAC_indexes.clear(); 3980 N_COMM_indexes.clear(); 3981 N_FUN_indexes.clear(); 3982 N_SO_index = UINT32_MAX; 3983 } else { 3984 // We use the current number of symbols in the symbol table in 3985 // lieu of using nlist_idx in case we ever start trimming entries 3986 // out 3987 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3988 if (N_SO_has_full_path) { 3989 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3990 // We have two consecutive N_SO entries where the first 3991 // contains a directory and the second contains a full path. 3992 sym[sym_idx - 1].GetMangled().SetValue( 3993 ConstString(symbol_name), false); 3994 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3995 add_nlist = false; 3996 } else { 3997 // This is the first entry in a N_SO that contains a 3998 // directory or a full path to the source file 3999 N_SO_index = sym_idx; 4000 } 4001 } else if ((N_SO_index == sym_idx - 1) && 4002 ((sym_idx - 1) < num_syms)) { 4003 // This is usually the second N_SO entry that contains just the 4004 // filename, so here we combine it with the first one if we are 4005 // minimizing the symbol table 4006 const char *so_path = 4007 sym[sym_idx - 1] 4008 .GetMangled() 4009 .GetDemangledName(lldb::eLanguageTypeUnknown) 4010 .AsCString(); 4011 if (so_path && so_path[0]) { 4012 std::string full_so_path(so_path); 4013 const size_t double_slash_pos = full_so_path.find("//"); 4014 if (double_slash_pos != std::string::npos) { 4015 // The linker has been generating bad N_SO entries with 4016 // doubled up paths in the format "%s%s" where the first 4017 // string in the DW_AT_comp_dir, and the second is the 4018 // directory for the source file so you end up with a path 4019 // that looks like "/tmp/src//tmp/src/" 4020 FileSpec so_dir(so_path); 4021 if (!FileSystem::Instance().Exists(so_dir)) { 4022 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 4023 FileSpec::Style::native); 4024 if (FileSystem::Instance().Exists(so_dir)) { 4025 // Trim off the incorrect path 4026 full_so_path.erase(0, double_slash_pos + 1); 4027 } 4028 } 4029 } 4030 if (*full_so_path.rbegin() != '/') 4031 full_so_path += '/'; 4032 full_so_path += symbol_name; 4033 sym[sym_idx - 1].GetMangled().SetValue( 4034 ConstString(full_so_path.c_str()), false); 4035 add_nlist = false; 4036 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4037 } 4038 } else { 4039 // This could be a relative path to a N_SO 4040 N_SO_index = sym_idx; 4041 } 4042 } 4043 break; 4044 4045 case N_OSO: 4046 // object file name: name,,0,0,st_mtime 4047 type = eSymbolTypeObjectFile; 4048 break; 4049 4050 case N_LSYM: 4051 // local sym: name,,NO_SECT,type,offset 4052 type = eSymbolTypeLocal; 4053 break; 4054 4055 //---------------------------------------------------------------------- 4056 // INCL scopes 4057 //---------------------------------------------------------------------- 4058 case N_BINCL: 4059 // include file beginning: name,,NO_SECT,0,sum We use the current 4060 // number of symbols in the symbol table in lieu of using nlist_idx 4061 // in case we ever start trimming entries out 4062 N_INCL_indexes.push_back(sym_idx); 4063 type = eSymbolTypeScopeBegin; 4064 break; 4065 4066 case N_EINCL: 4067 // include file end: name,,NO_SECT,0,0 4068 // Set the size of the N_BINCL to the terminating index of this 4069 // N_EINCL so that we can always skip the entire symbol if we need 4070 // to navigate more quickly at the source level when parsing STABS 4071 if (!N_INCL_indexes.empty()) { 4072 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4073 symbol_ptr->SetByteSize(sym_idx + 1); 4074 symbol_ptr->SetSizeIsSibling(true); 4075 N_INCL_indexes.pop_back(); 4076 } 4077 type = eSymbolTypeScopeEnd; 4078 break; 4079 4080 case N_SOL: 4081 // #included file name: name,,n_sect,0,address 4082 type = eSymbolTypeHeaderFile; 4083 4084 // We currently don't use the header files on darwin 4085 add_nlist = false; 4086 break; 4087 4088 case N_PARAMS: 4089 // compiler parameters: name,,NO_SECT,0,0 4090 type = eSymbolTypeCompiler; 4091 break; 4092 4093 case N_VERSION: 4094 // compiler version: name,,NO_SECT,0,0 4095 type = eSymbolTypeCompiler; 4096 break; 4097 4098 case N_OLEVEL: 4099 // compiler -O level: name,,NO_SECT,0,0 4100 type = eSymbolTypeCompiler; 4101 break; 4102 4103 case N_PSYM: 4104 // parameter: name,,NO_SECT,type,offset 4105 type = eSymbolTypeVariable; 4106 break; 4107 4108 case N_ENTRY: 4109 // alternate entry: name,,n_sect,linenumber,address 4110 symbol_section = 4111 section_info.GetSection(nlist.n_sect, nlist.n_value); 4112 type = eSymbolTypeLineEntry; 4113 break; 4114 4115 //---------------------------------------------------------------------- 4116 // Left and Right Braces 4117 //---------------------------------------------------------------------- 4118 case N_LBRAC: 4119 // left bracket: 0,,NO_SECT,nesting level,address We use the 4120 // current number of symbols in the symbol table in lieu of using 4121 // nlist_idx in case we ever start trimming entries out 4122 symbol_section = 4123 section_info.GetSection(nlist.n_sect, nlist.n_value); 4124 N_BRAC_indexes.push_back(sym_idx); 4125 type = eSymbolTypeScopeBegin; 4126 break; 4127 4128 case N_RBRAC: 4129 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4130 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4131 // can always skip the entire symbol if we need to navigate more 4132 // quickly at the source level when parsing STABS 4133 symbol_section = 4134 section_info.GetSection(nlist.n_sect, nlist.n_value); 4135 if (!N_BRAC_indexes.empty()) { 4136 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4137 symbol_ptr->SetByteSize(sym_idx + 1); 4138 symbol_ptr->SetSizeIsSibling(true); 4139 N_BRAC_indexes.pop_back(); 4140 } 4141 type = eSymbolTypeScopeEnd; 4142 break; 4143 4144 case N_EXCL: 4145 // deleted include file: name,,NO_SECT,0,sum 4146 type = eSymbolTypeHeaderFile; 4147 break; 4148 4149 //---------------------------------------------------------------------- 4150 // COMM scopes 4151 //---------------------------------------------------------------------- 4152 case N_BCOMM: 4153 // begin common: name,,NO_SECT,0,0 4154 // We use the current number of symbols in the symbol table in lieu 4155 // of using nlist_idx in case we ever start trimming entries out 4156 type = eSymbolTypeScopeBegin; 4157 N_COMM_indexes.push_back(sym_idx); 4158 break; 4159 4160 case N_ECOML: 4161 // end common (local name): 0,,n_sect,0,address 4162 symbol_section = 4163 section_info.GetSection(nlist.n_sect, nlist.n_value); 4164 LLVM_FALLTHROUGH; 4165 4166 case N_ECOMM: 4167 // end common: name,,n_sect,0,0 4168 // Set the size of the N_BCOMM to the terminating index of this 4169 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4170 // we need to navigate more quickly at the source level when 4171 // parsing STABS 4172 if (!N_COMM_indexes.empty()) { 4173 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4174 symbol_ptr->SetByteSize(sym_idx + 1); 4175 symbol_ptr->SetSizeIsSibling(true); 4176 N_COMM_indexes.pop_back(); 4177 } 4178 type = eSymbolTypeScopeEnd; 4179 break; 4180 4181 case N_LENG: 4182 // second stab entry with length information 4183 type = eSymbolTypeAdditional; 4184 break; 4185 4186 default: 4187 break; 4188 } 4189 } else { 4190 // uint8_t n_pext = N_PEXT & nlist.n_type; 4191 uint8_t n_type = N_TYPE & nlist.n_type; 4192 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4193 4194 switch (n_type) { 4195 case N_INDR: { 4196 const char *reexport_name_cstr = 4197 strtab_data.PeekCStr(nlist.n_value); 4198 if (reexport_name_cstr && reexport_name_cstr[0]) { 4199 type = eSymbolTypeReExported; 4200 ConstString reexport_name( 4201 reexport_name_cstr + 4202 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4203 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4204 set_value = false; 4205 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4206 indirect_symbol_names.insert( 4207 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4208 } else 4209 type = eSymbolTypeUndefined; 4210 } break; 4211 4212 case N_UNDF: 4213 if (symbol_name && symbol_name[0]) { 4214 ConstString undefined_name(symbol_name + 4215 ((symbol_name[0] == '_') ? 1 : 0)); 4216 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4217 } 4218 LLVM_FALLTHROUGH; 4219 4220 case N_PBUD: 4221 type = eSymbolTypeUndefined; 4222 break; 4223 4224 case N_ABS: 4225 type = eSymbolTypeAbsolute; 4226 break; 4227 4228 case N_SECT: { 4229 symbol_section = 4230 section_info.GetSection(nlist.n_sect, nlist.n_value); 4231 4232 if (!symbol_section) { 4233 // TODO: warn about this? 4234 add_nlist = false; 4235 break; 4236 } 4237 4238 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4239 type = eSymbolTypeException; 4240 } else { 4241 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4242 4243 switch (section_type) { 4244 case S_CSTRING_LITERALS: 4245 type = eSymbolTypeData; 4246 break; // section with only literal C strings 4247 case S_4BYTE_LITERALS: 4248 type = eSymbolTypeData; 4249 break; // section with only 4 byte literals 4250 case S_8BYTE_LITERALS: 4251 type = eSymbolTypeData; 4252 break; // section with only 8 byte literals 4253 case S_LITERAL_POINTERS: 4254 type = eSymbolTypeTrampoline; 4255 break; // section with only pointers to literals 4256 case S_NON_LAZY_SYMBOL_POINTERS: 4257 type = eSymbolTypeTrampoline; 4258 break; // section with only non-lazy symbol pointers 4259 case S_LAZY_SYMBOL_POINTERS: 4260 type = eSymbolTypeTrampoline; 4261 break; // section with only lazy symbol pointers 4262 case S_SYMBOL_STUBS: 4263 type = eSymbolTypeTrampoline; 4264 break; // section with only symbol stubs, byte size of stub in 4265 // the reserved2 field 4266 case S_MOD_INIT_FUNC_POINTERS: 4267 type = eSymbolTypeCode; 4268 break; // section with only function pointers for initialization 4269 case S_MOD_TERM_FUNC_POINTERS: 4270 type = eSymbolTypeCode; 4271 break; // section with only function pointers for termination 4272 case S_INTERPOSING: 4273 type = eSymbolTypeTrampoline; 4274 break; // section with only pairs of function pointers for 4275 // interposing 4276 case S_16BYTE_LITERALS: 4277 type = eSymbolTypeData; 4278 break; // section with only 16 byte literals 4279 case S_DTRACE_DOF: 4280 type = eSymbolTypeInstrumentation; 4281 break; 4282 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4283 type = eSymbolTypeTrampoline; 4284 break; 4285 default: 4286 switch (symbol_section->GetType()) { 4287 case lldb::eSectionTypeCode: 4288 type = eSymbolTypeCode; 4289 break; 4290 case eSectionTypeData: 4291 case eSectionTypeDataCString: // Inlined C string data 4292 case eSectionTypeDataCStringPointers: // Pointers to C string 4293 // data 4294 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4295 // the symbol table 4296 case eSectionTypeData4: 4297 case eSectionTypeData8: 4298 case eSectionTypeData16: 4299 type = eSymbolTypeData; 4300 break; 4301 default: 4302 break; 4303 } 4304 break; 4305 } 4306 4307 if (type == eSymbolTypeInvalid) { 4308 const char *symbol_sect_name = 4309 symbol_section->GetName().AsCString(); 4310 if (symbol_section->IsDescendant(text_section_sp.get())) { 4311 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4312 S_ATTR_SELF_MODIFYING_CODE | 4313 S_ATTR_SOME_INSTRUCTIONS)) 4314 type = eSymbolTypeData; 4315 else 4316 type = eSymbolTypeCode; 4317 } else if (symbol_section->IsDescendant( 4318 data_section_sp.get()) || 4319 symbol_section->IsDescendant( 4320 data_dirty_section_sp.get()) || 4321 symbol_section->IsDescendant( 4322 data_const_section_sp.get())) { 4323 if (symbol_sect_name && 4324 ::strstr(symbol_sect_name, "__objc") == 4325 symbol_sect_name) { 4326 type = eSymbolTypeRuntime; 4327 4328 if (symbol_name) { 4329 llvm::StringRef symbol_name_ref(symbol_name); 4330 if (symbol_name_ref.startswith("_OBJC_")) { 4331 static const llvm::StringRef g_objc_v2_prefix_class( 4332 "_OBJC_CLASS_$_"); 4333 static const llvm::StringRef g_objc_v2_prefix_metaclass( 4334 "_OBJC_METACLASS_$_"); 4335 static const llvm::StringRef g_objc_v2_prefix_ivar( 4336 "_OBJC_IVAR_$_"); 4337 if (symbol_name_ref.startswith( 4338 g_objc_v2_prefix_class)) { 4339 symbol_name_non_abi_mangled = symbol_name + 1; 4340 symbol_name = 4341 symbol_name + g_objc_v2_prefix_class.size(); 4342 type = eSymbolTypeObjCClass; 4343 demangled_is_synthesized = true; 4344 } else if (symbol_name_ref.startswith( 4345 g_objc_v2_prefix_metaclass)) { 4346 symbol_name_non_abi_mangled = symbol_name + 1; 4347 symbol_name = 4348 symbol_name + g_objc_v2_prefix_metaclass.size(); 4349 type = eSymbolTypeObjCMetaClass; 4350 demangled_is_synthesized = true; 4351 } else if (symbol_name_ref.startswith( 4352 g_objc_v2_prefix_ivar)) { 4353 symbol_name_non_abi_mangled = symbol_name + 1; 4354 symbol_name = 4355 symbol_name + g_objc_v2_prefix_ivar.size(); 4356 type = eSymbolTypeObjCIVar; 4357 demangled_is_synthesized = true; 4358 } 4359 } 4360 } 4361 } else if (symbol_sect_name && 4362 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4363 symbol_sect_name) { 4364 type = eSymbolTypeException; 4365 } else { 4366 type = eSymbolTypeData; 4367 } 4368 } else if (symbol_sect_name && 4369 ::strstr(symbol_sect_name, "__IMPORT") == 4370 symbol_sect_name) { 4371 type = eSymbolTypeTrampoline; 4372 } else if (symbol_section->IsDescendant( 4373 objc_section_sp.get())) { 4374 type = eSymbolTypeRuntime; 4375 if (symbol_name && symbol_name[0] == '.') { 4376 llvm::StringRef symbol_name_ref(symbol_name); 4377 static const llvm::StringRef g_objc_v1_prefix_class( 4378 ".objc_class_name_"); 4379 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4380 symbol_name_non_abi_mangled = symbol_name; 4381 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4382 type = eSymbolTypeObjCClass; 4383 demangled_is_synthesized = true; 4384 } 4385 } 4386 } 4387 } 4388 } 4389 } break; 4390 } 4391 } 4392 4393 if (add_nlist) { 4394 uint64_t symbol_value = nlist.n_value; 4395 4396 if (symbol_name_non_abi_mangled) { 4397 sym[sym_idx].GetMangled().SetMangledName( 4398 ConstString(symbol_name_non_abi_mangled)); 4399 sym[sym_idx].GetMangled().SetDemangledName( 4400 ConstString(symbol_name)); 4401 } else { 4402 bool symbol_name_is_mangled = false; 4403 4404 if (symbol_name && symbol_name[0] == '_') { 4405 symbol_name_is_mangled = symbol_name[1] == '_'; 4406 symbol_name++; // Skip the leading underscore 4407 } 4408 4409 if (symbol_name) { 4410 ConstString const_symbol_name(symbol_name); 4411 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4412 symbol_name_is_mangled); 4413 } 4414 } 4415 4416 if (is_gsym) { 4417 const char *gsym_name = sym[sym_idx] 4418 .GetMangled() 4419 .GetName(lldb::eLanguageTypeUnknown, 4420 Mangled::ePreferMangled) 4421 .GetCString(); 4422 if (gsym_name) 4423 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4424 } 4425 4426 if (symbol_section) { 4427 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4428 if (symbol_byte_size == 0 && function_starts_count > 0) { 4429 addr_t symbol_lookup_file_addr = nlist.n_value; 4430 // Do an exact address match for non-ARM addresses, else get the 4431 // closest since the symbol might be a thumb symbol which has an 4432 // address with bit zero set 4433 FunctionStarts::Entry *func_start_entry = 4434 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4435 if (is_arm && func_start_entry) { 4436 // Verify that the function start address is the symbol address 4437 // (ARM) or the symbol address + 1 (thumb) 4438 if (func_start_entry->addr != symbol_lookup_file_addr && 4439 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4440 // Not the right entry, NULL it out... 4441 func_start_entry = NULL; 4442 } 4443 } 4444 if (func_start_entry) { 4445 func_start_entry->data = true; 4446 4447 addr_t symbol_file_addr = func_start_entry->addr; 4448 if (is_arm) 4449 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4450 4451 const FunctionStarts::Entry *next_func_start_entry = 4452 function_starts.FindNextEntry(func_start_entry); 4453 const addr_t section_end_file_addr = 4454 section_file_addr + symbol_section->GetByteSize(); 4455 if (next_func_start_entry) { 4456 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4457 // Be sure the clear the Thumb address bit when we calculate 4458 // the size from the current and next address 4459 if (is_arm) 4460 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4461 symbol_byte_size = std::min<lldb::addr_t>( 4462 next_symbol_file_addr - symbol_file_addr, 4463 section_end_file_addr - symbol_file_addr); 4464 } else { 4465 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4466 } 4467 } 4468 } 4469 symbol_value -= section_file_addr; 4470 } 4471 4472 if (!is_debug) { 4473 if (type == eSymbolTypeCode) { 4474 // See if we can find a N_FUN entry for any code symbols. If we 4475 // do find a match, and the name matches, then we can merge the 4476 // two into just the function symbol to avoid duplicate entries 4477 // in the symbol table 4478 std::pair<ValueToSymbolIndexMap::const_iterator, 4479 ValueToSymbolIndexMap::const_iterator> 4480 range; 4481 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4482 if (range.first != range.second) { 4483 bool found_it = false; 4484 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4485 pos != range.second; ++pos) { 4486 if (sym[sym_idx].GetMangled().GetName( 4487 lldb::eLanguageTypeUnknown, 4488 Mangled::ePreferMangled) == 4489 sym[pos->second].GetMangled().GetName( 4490 lldb::eLanguageTypeUnknown, 4491 Mangled::ePreferMangled)) { 4492 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4493 // We just need the flags from the linker symbol, so put 4494 // these flags into the N_FUN flags to avoid duplicate 4495 // symbols in the symbol table 4496 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4497 sym[pos->second].SetFlags(nlist.n_type << 16 | 4498 nlist.n_desc); 4499 if (resolver_addresses.find(nlist.n_value) != 4500 resolver_addresses.end()) 4501 sym[pos->second].SetType(eSymbolTypeResolver); 4502 sym[sym_idx].Clear(); 4503 found_it = true; 4504 break; 4505 } 4506 } 4507 if (found_it) 4508 continue; 4509 } else { 4510 if (resolver_addresses.find(nlist.n_value) != 4511 resolver_addresses.end()) 4512 type = eSymbolTypeResolver; 4513 } 4514 } else if (type == eSymbolTypeData || 4515 type == eSymbolTypeObjCClass || 4516 type == eSymbolTypeObjCMetaClass || 4517 type == eSymbolTypeObjCIVar) { 4518 // See if we can find a N_STSYM entry for any data symbols. If we 4519 // do find a match, and the name matches, then we can merge the 4520 // two into just the Static symbol to avoid duplicate entries in 4521 // the symbol table 4522 std::pair<ValueToSymbolIndexMap::const_iterator, 4523 ValueToSymbolIndexMap::const_iterator> 4524 range; 4525 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4526 if (range.first != range.second) { 4527 bool found_it = false; 4528 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4529 pos != range.second; ++pos) { 4530 if (sym[sym_idx].GetMangled().GetName( 4531 lldb::eLanguageTypeUnknown, 4532 Mangled::ePreferMangled) == 4533 sym[pos->second].GetMangled().GetName( 4534 lldb::eLanguageTypeUnknown, 4535 Mangled::ePreferMangled)) { 4536 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4537 // We just need the flags from the linker symbol, so put 4538 // these flags into the N_STSYM flags to avoid duplicate 4539 // symbols in the symbol table 4540 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4541 sym[pos->second].SetFlags(nlist.n_type << 16 | 4542 nlist.n_desc); 4543 sym[sym_idx].Clear(); 4544 found_it = true; 4545 break; 4546 } 4547 } 4548 if (found_it) 4549 continue; 4550 } else { 4551 // Combine N_GSYM stab entries with the non stab symbol 4552 const char *gsym_name = sym[sym_idx] 4553 .GetMangled() 4554 .GetName(lldb::eLanguageTypeUnknown, 4555 Mangled::ePreferMangled) 4556 .GetCString(); 4557 if (gsym_name) { 4558 ConstNameToSymbolIndexMap::const_iterator pos = 4559 N_GSYM_name_to_sym_idx.find(gsym_name); 4560 if (pos != N_GSYM_name_to_sym_idx.end()) { 4561 const uint32_t GSYM_sym_idx = pos->second; 4562 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4563 // Copy the address, because often the N_GSYM address has 4564 // an invalid address of zero when the global is a common 4565 // symbol 4566 sym[GSYM_sym_idx].GetAddressRef().SetSection( 4567 symbol_section); 4568 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4569 // We just need the flags from the linker symbol, so put 4570 // these flags into the N_GSYM flags to avoid duplicate 4571 // symbols in the symbol table 4572 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 4573 nlist.n_desc); 4574 sym[sym_idx].Clear(); 4575 continue; 4576 } 4577 } 4578 } 4579 } 4580 } 4581 4582 sym[sym_idx].SetID(nlist_idx); 4583 sym[sym_idx].SetType(type); 4584 if (set_value) { 4585 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4586 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4587 } 4588 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4589 4590 if (symbol_byte_size > 0) 4591 sym[sym_idx].SetByteSize(symbol_byte_size); 4592 4593 if (demangled_is_synthesized) 4594 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4595 4596 ++sym_idx; 4597 } else { 4598 sym[sym_idx].Clear(); 4599 } 4600 } 4601 4602 for (const auto &pos : reexport_shlib_needs_fixup) { 4603 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4604 if (undef_pos != undefined_name_to_desc.end()) { 4605 const uint8_t dylib_ordinal = 4606 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4607 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4608 sym[pos.first].SetReExportedSymbolSharedLibrary( 4609 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4610 } 4611 } 4612 } 4613 4614 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4615 4616 if (function_starts_count > 0) { 4617 uint32_t num_synthetic_function_symbols = 0; 4618 for (i = 0; i < function_starts_count; ++i) { 4619 if (!function_starts.GetEntryRef(i).data) 4620 ++num_synthetic_function_symbols; 4621 } 4622 4623 if (num_synthetic_function_symbols > 0) { 4624 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4625 num_syms = sym_idx + num_synthetic_function_symbols; 4626 sym = symtab->Resize(num_syms); 4627 } 4628 for (i = 0; i < function_starts_count; ++i) { 4629 const FunctionStarts::Entry *func_start_entry = 4630 function_starts.GetEntryAtIndex(i); 4631 if (!func_start_entry->data) { 4632 addr_t symbol_file_addr = func_start_entry->addr; 4633 uint32_t symbol_flags = 0; 4634 if (is_arm) { 4635 if (symbol_file_addr & 1) 4636 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4637 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4638 } 4639 Address symbol_addr; 4640 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4641 SectionSP symbol_section(symbol_addr.GetSection()); 4642 uint32_t symbol_byte_size = 0; 4643 if (symbol_section) { 4644 const addr_t section_file_addr = 4645 symbol_section->GetFileAddress(); 4646 const FunctionStarts::Entry *next_func_start_entry = 4647 function_starts.FindNextEntry(func_start_entry); 4648 const addr_t section_end_file_addr = 4649 section_file_addr + symbol_section->GetByteSize(); 4650 if (next_func_start_entry) { 4651 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4652 if (is_arm) 4653 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4654 symbol_byte_size = std::min<lldb::addr_t>( 4655 next_symbol_file_addr - symbol_file_addr, 4656 section_end_file_addr - symbol_file_addr); 4657 } else { 4658 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4659 } 4660 sym[sym_idx].SetID(synthetic_sym_id++); 4661 sym[sym_idx].GetMangled().SetDemangledName( 4662 GetNextSyntheticSymbolName()); 4663 sym[sym_idx].SetType(eSymbolTypeCode); 4664 sym[sym_idx].SetIsSynthetic(true); 4665 sym[sym_idx].GetAddressRef() = symbol_addr; 4666 if (symbol_flags) 4667 sym[sym_idx].SetFlags(symbol_flags); 4668 if (symbol_byte_size) 4669 sym[sym_idx].SetByteSize(symbol_byte_size); 4670 ++sym_idx; 4671 } 4672 } 4673 } 4674 } 4675 } 4676 } 4677 4678 // Trim our symbols down to just what we ended up with after removing any 4679 // symbols. 4680 if (sym_idx < num_syms) { 4681 num_syms = sym_idx; 4682 sym = symtab->Resize(num_syms); 4683 } 4684 4685 // Now synthesize indirect symbols 4686 if (m_dysymtab.nindirectsyms != 0) { 4687 if (indirect_symbol_index_data.GetByteSize()) { 4688 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4689 m_nlist_idx_to_sym_idx.end(); 4690 4691 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4692 ++sect_idx) { 4693 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4694 S_SYMBOL_STUBS) { 4695 uint32_t symbol_stub_byte_size = 4696 m_mach_sections[sect_idx].reserved2; 4697 if (symbol_stub_byte_size == 0) 4698 continue; 4699 4700 const uint32_t num_symbol_stubs = 4701 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4702 4703 if (num_symbol_stubs == 0) 4704 continue; 4705 4706 const uint32_t symbol_stub_index_offset = 4707 m_mach_sections[sect_idx].reserved1; 4708 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; 4709 ++stub_idx) { 4710 const uint32_t symbol_stub_index = 4711 symbol_stub_index_offset + stub_idx; 4712 const lldb::addr_t symbol_stub_addr = 4713 m_mach_sections[sect_idx].addr + 4714 (stub_idx * symbol_stub_byte_size); 4715 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4716 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4717 symbol_stub_offset, 4)) { 4718 const uint32_t stub_sym_id = 4719 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4720 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4721 continue; 4722 4723 NListIndexToSymbolIndexMap::const_iterator index_pos = 4724 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4725 Symbol *stub_symbol = NULL; 4726 if (index_pos != end_index_pos) { 4727 // We have a remapping from the original nlist index to a 4728 // current symbol index, so just look this up by index 4729 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4730 } else { 4731 // We need to lookup a symbol using the original nlist symbol 4732 // index since this index is coming from the S_SYMBOL_STUBS 4733 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4734 } 4735 4736 if (stub_symbol) { 4737 Address so_addr(symbol_stub_addr, section_list); 4738 4739 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4740 // Change the external symbol into a trampoline that makes 4741 // sense These symbols were N_UNDF N_EXT, and are useless 4742 // to us, so we can re-use them so we don't have to make up 4743 // a synthetic symbol for no good reason. 4744 if (resolver_addresses.find(symbol_stub_addr) == 4745 resolver_addresses.end()) 4746 stub_symbol->SetType(eSymbolTypeTrampoline); 4747 else 4748 stub_symbol->SetType(eSymbolTypeResolver); 4749 stub_symbol->SetExternal(false); 4750 stub_symbol->GetAddressRef() = so_addr; 4751 stub_symbol->SetByteSize(symbol_stub_byte_size); 4752 } else { 4753 // Make a synthetic symbol to describe the trampoline stub 4754 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4755 if (sym_idx >= num_syms) { 4756 sym = symtab->Resize(++num_syms); 4757 stub_symbol = NULL; // this pointer no longer valid 4758 } 4759 sym[sym_idx].SetID(synthetic_sym_id++); 4760 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4761 if (resolver_addresses.find(symbol_stub_addr) == 4762 resolver_addresses.end()) 4763 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4764 else 4765 sym[sym_idx].SetType(eSymbolTypeResolver); 4766 sym[sym_idx].SetIsSynthetic(true); 4767 sym[sym_idx].GetAddressRef() = so_addr; 4768 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4769 ++sym_idx; 4770 } 4771 } else { 4772 if (log) 4773 log->Warning("symbol stub referencing symbol table symbol " 4774 "%u that isn't in our minimal symbol table, " 4775 "fix this!!!", 4776 stub_sym_id); 4777 } 4778 } 4779 } 4780 } 4781 } 4782 } 4783 } 4784 4785 if (!trie_entries.empty()) { 4786 for (const auto &e : trie_entries) { 4787 if (e.entry.import_name) { 4788 // Only add indirect symbols from the Trie entries if we didn't have 4789 // a N_INDR nlist entry for this already 4790 if (indirect_symbol_names.find(e.entry.name) == 4791 indirect_symbol_names.end()) { 4792 // Make a synthetic symbol to describe re-exported symbol. 4793 if (sym_idx >= num_syms) 4794 sym = symtab->Resize(++num_syms); 4795 sym[sym_idx].SetID(synthetic_sym_id++); 4796 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4797 sym[sym_idx].SetType(eSymbolTypeReExported); 4798 sym[sym_idx].SetIsSynthetic(true); 4799 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4800 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4801 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4802 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4803 } 4804 ++sym_idx; 4805 } 4806 } 4807 } 4808 } 4809 4810 // StreamFile s(stdout, false); 4811 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4812 // symtab->Dump(&s, NULL, eSortOrderNone); 4813 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4814 // sizes 4815 symtab->CalculateSymbolSizes(); 4816 4817 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4818 // symtab->Dump(&s, NULL, eSortOrderNone); 4819 4820 return symtab->GetNumSymbols(); 4821 } 4822 return 0; 4823 } 4824 4825 void ObjectFileMachO::Dump(Stream *s) { 4826 ModuleSP module_sp(GetModule()); 4827 if (module_sp) { 4828 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4829 s->Printf("%p: ", static_cast<void *>(this)); 4830 s->Indent(); 4831 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4832 s->PutCString("ObjectFileMachO64"); 4833 else 4834 s->PutCString("ObjectFileMachO32"); 4835 4836 ArchSpec header_arch = GetArchitecture(); 4837 4838 *s << ", file = '" << m_file 4839 << "', triple = " << header_arch.GetTriple().getTriple() << "\n"; 4840 4841 SectionList *sections = GetSectionList(); 4842 if (sections) 4843 sections->Dump(s, NULL, true, UINT32_MAX); 4844 4845 if (m_symtab_up) 4846 m_symtab_up->Dump(s, NULL, eSortOrderNone); 4847 } 4848 } 4849 4850 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4851 const lldb_private::DataExtractor &data, 4852 lldb::offset_t lc_offset) { 4853 uint32_t i; 4854 struct uuid_command load_cmd; 4855 4856 lldb::offset_t offset = lc_offset; 4857 for (i = 0; i < header.ncmds; ++i) { 4858 const lldb::offset_t cmd_offset = offset; 4859 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4860 break; 4861 4862 if (load_cmd.cmd == LC_UUID) { 4863 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4864 4865 if (uuid_bytes) { 4866 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4867 // We pretend these object files have no UUID to prevent crashing. 4868 4869 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4870 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4871 0xbb, 0x14, 0xf0, 0x0d}; 4872 4873 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4874 return UUID(); 4875 4876 return UUID::fromOptionalData(uuid_bytes, 16); 4877 } 4878 return UUID(); 4879 } 4880 offset = cmd_offset + load_cmd.cmdsize; 4881 } 4882 return UUID(); 4883 } 4884 4885 static llvm::StringRef GetOSName(uint32_t cmd) { 4886 switch (cmd) { 4887 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4888 return llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4889 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4890 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4891 case llvm::MachO::LC_VERSION_MIN_TVOS: 4892 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4893 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4894 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4895 default: 4896 llvm_unreachable("unexpected LC_VERSION load command"); 4897 } 4898 } 4899 4900 namespace { 4901 struct OSEnv { 4902 llvm::StringRef os_type; 4903 llvm::StringRef environment; 4904 OSEnv(uint32_t cmd) { 4905 switch (cmd) { 4906 case PLATFORM_MACOS: 4907 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX); 4908 return; 4909 case PLATFORM_IOS: 4910 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4911 return; 4912 case PLATFORM_TVOS: 4913 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4914 return; 4915 case PLATFORM_WATCHOS: 4916 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4917 return; 4918 // NEED_BRIDGEOS_TRIPLE case PLATFORM_BRIDGEOS: 4919 // NEED_BRIDGEOS_TRIPLE os_type = llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS); 4920 // NEED_BRIDGEOS_TRIPLE return; 4921 #if defined (PLATFORM_IOSSIMULATOR) && defined (PLATFORM_TVOSSIMULATOR) && defined (PLATFORM_WATCHOSSIMULATOR) 4922 case PLATFORM_IOSSIMULATOR: 4923 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS); 4924 environment = 4925 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4926 return; 4927 case PLATFORM_TVOSSIMULATOR: 4928 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS); 4929 environment = 4930 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4931 return; 4932 case PLATFORM_WATCHOSSIMULATOR: 4933 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS); 4934 environment = 4935 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator); 4936 return; 4937 #endif 4938 default: { 4939 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | 4940 LIBLLDB_LOG_PROCESS)); 4941 if (log) 4942 log->Printf("unsupported platform in LC_BUILD_VERSION"); 4943 } 4944 } 4945 } 4946 }; 4947 4948 struct MinOS { 4949 uint32_t major_version, minor_version, patch_version; 4950 MinOS(uint32_t version) 4951 : major_version(version >> 16), 4952 minor_version((version >> 8) & 0xffu), 4953 patch_version(version & 0xffu) {} 4954 }; 4955 } // namespace 4956 4957 ArchSpec 4958 ObjectFileMachO::GetArchitecture(const llvm::MachO::mach_header &header, 4959 const lldb_private::DataExtractor &data, 4960 lldb::offset_t lc_offset) { 4961 ArchSpec arch; 4962 arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4963 4964 if (arch.IsValid()) { 4965 llvm::Triple &triple = arch.GetTriple(); 4966 4967 // Set OS to an unspecified unknown or a "*" so it can match any OS 4968 triple.setOS(llvm::Triple::UnknownOS); 4969 triple.setOSName(llvm::StringRef()); 4970 4971 if (header.filetype == MH_PRELOAD) { 4972 if (header.cputype == CPU_TYPE_ARM) { 4973 // If this is a 32-bit arm binary, and it's a standalone binary, force 4974 // the Vendor to Apple so we don't accidentally pick up the generic 4975 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4976 // frame pointer register; most other armv7 ABIs use a combination of 4977 // r7 and r11. 4978 triple.setVendor(llvm::Triple::Apple); 4979 } else { 4980 // Set vendor to an unspecified unknown or a "*" so it can match any 4981 // vendor This is required for correct behavior of EFI debugging on 4982 // x86_64 4983 triple.setVendor(llvm::Triple::UnknownVendor); 4984 triple.setVendorName(llvm::StringRef()); 4985 } 4986 return arch; 4987 } else { 4988 struct load_command load_cmd; 4989 llvm::SmallString<16> os_name; 4990 llvm::raw_svector_ostream os(os_name); 4991 4992 // See if there is an LC_VERSION_MIN_* load command that can give 4993 // us the OS type. 4994 lldb::offset_t offset = lc_offset; 4995 for (uint32_t i = 0; i < header.ncmds; ++i) { 4996 const lldb::offset_t cmd_offset = offset; 4997 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4998 break; 4999 5000 struct version_min_command version_min; 5001 switch (load_cmd.cmd) { 5002 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 5003 case llvm::MachO::LC_VERSION_MIN_MACOSX: 5004 case llvm::MachO::LC_VERSION_MIN_TVOS: 5005 case llvm::MachO::LC_VERSION_MIN_WATCHOS: { 5006 if (load_cmd.cmdsize != sizeof(version_min)) 5007 break; 5008 if (data.ExtractBytes(cmd_offset, sizeof(version_min), 5009 data.GetByteOrder(), &version_min) == 0) 5010 break; 5011 MinOS min_os(version_min.version); 5012 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.' 5013 << min_os.minor_version << '.' << min_os.patch_version; 5014 triple.setOSName(os.str()); 5015 return arch; 5016 } 5017 default: 5018 break; 5019 } 5020 5021 offset = cmd_offset + load_cmd.cmdsize; 5022 } 5023 5024 // See if there is an LC_BUILD_VERSION load command that can give 5025 // us the OS type. 5026 5027 offset = lc_offset; 5028 for (uint32_t i = 0; i < header.ncmds; ++i) { 5029 const lldb::offset_t cmd_offset = offset; 5030 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 5031 break; 5032 do { 5033 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { 5034 struct build_version_command build_version; 5035 if (load_cmd.cmdsize < sizeof(build_version)) { 5036 // Malformed load command. 5037 break; 5038 } 5039 if (data.ExtractBytes(cmd_offset, sizeof(build_version), 5040 data.GetByteOrder(), &build_version) == 0) 5041 break; 5042 MinOS min_os(build_version.minos); 5043 OSEnv os_env(build_version.platform); 5044 if (os_env.os_type.empty()) 5045 break; 5046 os << os_env.os_type << min_os.major_version << '.' 5047 << min_os.minor_version << '.' << min_os.patch_version; 5048 triple.setOSName(os.str()); 5049 if (!os_env.environment.empty()) 5050 triple.setEnvironmentName(os_env.environment); 5051 return arch; 5052 } 5053 } while (0); 5054 offset = cmd_offset + load_cmd.cmdsize; 5055 } 5056 5057 if (header.filetype != MH_KEXT_BUNDLE) { 5058 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 5059 // so lets not say our Vendor is Apple, leave it as an unspecified 5060 // unknown 5061 triple.setVendor(llvm::Triple::UnknownVendor); 5062 triple.setVendorName(llvm::StringRef()); 5063 } 5064 } 5065 } 5066 return arch; 5067 } 5068 5069 UUID ObjectFileMachO::GetUUID() { 5070 ModuleSP module_sp(GetModule()); 5071 if (module_sp) { 5072 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5073 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5074 return GetUUID(m_header, m_data, offset); 5075 } 5076 return UUID(); 5077 } 5078 5079 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5080 uint32_t count = 0; 5081 ModuleSP module_sp(GetModule()); 5082 if (module_sp) { 5083 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5084 struct load_command load_cmd; 5085 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5086 std::vector<std::string> rpath_paths; 5087 std::vector<std::string> rpath_relative_paths; 5088 std::vector<std::string> at_exec_relative_paths; 5089 uint32_t i; 5090 for (i = 0; i < m_header.ncmds; ++i) { 5091 const uint32_t cmd_offset = offset; 5092 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5093 break; 5094 5095 switch (load_cmd.cmd) { 5096 case LC_RPATH: 5097 case LC_LOAD_DYLIB: 5098 case LC_LOAD_WEAK_DYLIB: 5099 case LC_REEXPORT_DYLIB: 5100 case LC_LOAD_DYLINKER: 5101 case LC_LOADFVMLIB: 5102 case LC_LOAD_UPWARD_DYLIB: { 5103 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5104 const char *path = m_data.PeekCStr(name_offset); 5105 if (path) { 5106 if (load_cmd.cmd == LC_RPATH) 5107 rpath_paths.push_back(path); 5108 else { 5109 if (path[0] == '@') { 5110 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5111 rpath_relative_paths.push_back(path + strlen("@rpath")); 5112 else if (strncmp(path, "@executable_path", 5113 strlen("@executable_path")) == 0) 5114 at_exec_relative_paths.push_back(path 5115 + strlen("@executable_path")); 5116 } else { 5117 FileSpec file_spec(path); 5118 if (files.AppendIfUnique(file_spec)) 5119 count++; 5120 } 5121 } 5122 } 5123 } break; 5124 5125 default: 5126 break; 5127 } 5128 offset = cmd_offset + load_cmd.cmdsize; 5129 } 5130 5131 FileSpec this_file_spec(m_file); 5132 FileSystem::Instance().Resolve(this_file_spec); 5133 5134 if (!rpath_paths.empty()) { 5135 // Fixup all LC_RPATH values to be absolute paths 5136 std::string loader_path("@loader_path"); 5137 std::string executable_path("@executable_path"); 5138 for (auto &rpath : rpath_paths) { 5139 if (rpath.find(loader_path) == 0) { 5140 rpath.erase(0, loader_path.size()); 5141 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5142 } else if (rpath.find(executable_path) == 0) { 5143 rpath.erase(0, executable_path.size()); 5144 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5145 } 5146 } 5147 5148 for (const auto &rpath_relative_path : rpath_relative_paths) { 5149 for (const auto &rpath : rpath_paths) { 5150 std::string path = rpath; 5151 path += rpath_relative_path; 5152 // It is OK to resolve this path because we must find a file on disk 5153 // for us to accept it anyway if it is rpath relative. 5154 FileSpec file_spec(path); 5155 FileSystem::Instance().Resolve(file_spec); 5156 if (FileSystem::Instance().Exists(file_spec) && 5157 files.AppendIfUnique(file_spec)) { 5158 count++; 5159 break; 5160 } 5161 } 5162 } 5163 } 5164 5165 // We may have @executable_paths but no RPATHS. Figure those out here. 5166 // Only do this if this object file is the executable. We have no way to 5167 // get back to the actual executable otherwise, so we won't get the right 5168 // path. 5169 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5170 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5171 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5172 FileSpec file_spec = 5173 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5174 if (FileSystem::Instance().Exists(file_spec) && 5175 files.AppendIfUnique(file_spec)) 5176 count++; 5177 } 5178 } 5179 } 5180 return count; 5181 } 5182 5183 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5184 // If the object file is not an executable it can't hold the entry point. 5185 // m_entry_point_address is initialized to an invalid address, so we can just 5186 // return that. If m_entry_point_address is valid it means we've found it 5187 // already, so return the cached value. 5188 5189 if (!IsExecutable() || m_entry_point_address.IsValid()) 5190 return m_entry_point_address; 5191 5192 // Otherwise, look for the UnixThread or Thread command. The data for the 5193 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5194 // 5195 // uint32_t flavor - this is the flavor argument you would pass to 5196 // thread_get_state 5197 // uint32_t count - this is the count of longs in the thread state data 5198 // struct XXX_thread_state state - this is the structure from 5199 // <machine/thread_status.h> corresponding to the flavor. 5200 // <repeat this trio> 5201 // 5202 // So we just keep reading the various register flavors till we find the GPR 5203 // one, then read the PC out of there. 5204 // FIXME: We will need to have a "RegisterContext data provider" class at some 5205 // point that can get all the registers 5206 // out of data in this form & attach them to a given thread. That should 5207 // underlie the MacOS X User process plugin, and we'll also need it for the 5208 // MacOS X Core File process plugin. When we have that we can also use it 5209 // here. 5210 // 5211 // For now we hard-code the offsets and flavors we need: 5212 // 5213 // 5214 5215 ModuleSP module_sp(GetModule()); 5216 if (module_sp) { 5217 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5218 struct load_command load_cmd; 5219 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5220 uint32_t i; 5221 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5222 bool done = false; 5223 5224 for (i = 0; i < m_header.ncmds; ++i) { 5225 const lldb::offset_t cmd_offset = offset; 5226 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5227 break; 5228 5229 switch (load_cmd.cmd) { 5230 case LC_UNIXTHREAD: 5231 case LC_THREAD: { 5232 while (offset < cmd_offset + load_cmd.cmdsize) { 5233 uint32_t flavor = m_data.GetU32(&offset); 5234 uint32_t count = m_data.GetU32(&offset); 5235 if (count == 0) { 5236 // We've gotten off somehow, log and exit; 5237 return m_entry_point_address; 5238 } 5239 5240 switch (m_header.cputype) { 5241 case llvm::MachO::CPU_TYPE_ARM: 5242 if (flavor == 1 || 5243 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 from 5244 // mach/arm/thread_status.h 5245 { 5246 offset += 60; // This is the offset of pc in the GPR thread state 5247 // data structure. 5248 start_address = m_data.GetU32(&offset); 5249 done = true; 5250 } 5251 break; 5252 case llvm::MachO::CPU_TYPE_ARM64: 5253 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5254 { 5255 offset += 256; // This is the offset of pc in the GPR thread state 5256 // data structure. 5257 start_address = m_data.GetU64(&offset); 5258 done = true; 5259 } 5260 break; 5261 case llvm::MachO::CPU_TYPE_I386: 5262 if (flavor == 5263 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5264 { 5265 offset += 40; // This is the offset of eip in the GPR thread state 5266 // data structure. 5267 start_address = m_data.GetU32(&offset); 5268 done = true; 5269 } 5270 break; 5271 case llvm::MachO::CPU_TYPE_X86_64: 5272 if (flavor == 5273 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5274 { 5275 offset += 16 * 8; // This is the offset of rip in the GPR thread 5276 // state data structure. 5277 start_address = m_data.GetU64(&offset); 5278 done = true; 5279 } 5280 break; 5281 default: 5282 return m_entry_point_address; 5283 } 5284 // Haven't found the GPR flavor yet, skip over the data for this 5285 // flavor: 5286 if (done) 5287 break; 5288 offset += count * 4; 5289 } 5290 } break; 5291 case LC_MAIN: { 5292 ConstString text_segment_name("__TEXT"); 5293 uint64_t entryoffset = m_data.GetU64(&offset); 5294 SectionSP text_segment_sp = 5295 GetSectionList()->FindSectionByName(text_segment_name); 5296 if (text_segment_sp) { 5297 done = true; 5298 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5299 } 5300 } break; 5301 5302 default: 5303 break; 5304 } 5305 if (done) 5306 break; 5307 5308 // Go to the next load command: 5309 offset = cmd_offset + load_cmd.cmdsize; 5310 } 5311 5312 if (start_address != LLDB_INVALID_ADDRESS) { 5313 // We got the start address from the load commands, so now resolve that 5314 // address in the sections of this ObjectFile: 5315 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5316 start_address, GetSectionList())) { 5317 m_entry_point_address.Clear(); 5318 } 5319 } else { 5320 // We couldn't read the UnixThread load command - maybe it wasn't there. 5321 // As a fallback look for the "start" symbol in the main executable. 5322 5323 ModuleSP module_sp(GetModule()); 5324 5325 if (module_sp) { 5326 SymbolContextList contexts; 5327 SymbolContext context; 5328 if (module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5329 eSymbolTypeCode, contexts)) { 5330 if (contexts.GetContextAtIndex(0, context)) 5331 m_entry_point_address = context.symbol->GetAddress(); 5332 } 5333 } 5334 } 5335 } 5336 5337 return m_entry_point_address; 5338 } 5339 5340 lldb_private::Address ObjectFileMachO::GetBaseAddress() { 5341 lldb_private::Address header_addr; 5342 SectionList *section_list = GetSectionList(); 5343 if (section_list) { 5344 SectionSP text_segment_sp( 5345 section_list->FindSectionByName(GetSegmentNameTEXT())); 5346 if (text_segment_sp) { 5347 header_addr.SetSection(text_segment_sp); 5348 header_addr.SetOffset(0); 5349 } 5350 } 5351 return header_addr; 5352 } 5353 5354 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5355 ModuleSP module_sp(GetModule()); 5356 if (module_sp) { 5357 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5358 if (!m_thread_context_offsets_valid) { 5359 m_thread_context_offsets_valid = true; 5360 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5361 FileRangeArray::Entry file_range; 5362 thread_command thread_cmd; 5363 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5364 const uint32_t cmd_offset = offset; 5365 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 5366 break; 5367 5368 if (thread_cmd.cmd == LC_THREAD) { 5369 file_range.SetRangeBase(offset); 5370 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5371 m_thread_context_offsets.Append(file_range); 5372 } 5373 offset = cmd_offset + thread_cmd.cmdsize; 5374 } 5375 } 5376 } 5377 return m_thread_context_offsets.GetSize(); 5378 } 5379 5380 std::string ObjectFileMachO::GetIdentifierString() { 5381 std::string result; 5382 ModuleSP module_sp(GetModule()); 5383 if (module_sp) { 5384 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5385 5386 // First, look over the load commands for an LC_NOTE load command with 5387 // data_owner string "kern ver str" & use that if found. 5388 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5389 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5390 const uint32_t cmd_offset = offset; 5391 load_command lc; 5392 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5393 break; 5394 if (lc.cmd == LC_NOTE) 5395 { 5396 char data_owner[17]; 5397 m_data.CopyData (offset, 16, data_owner); 5398 data_owner[16] = '\0'; 5399 offset += 16; 5400 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5401 uint64_t size = m_data.GetU64_unchecked (&offset); 5402 5403 // "kern ver str" has a uint32_t version and then a nul terminated 5404 // c-string. 5405 if (strcmp ("kern ver str", data_owner) == 0) 5406 { 5407 offset = fileoff; 5408 uint32_t version; 5409 if (m_data.GetU32 (&offset, &version, 1) != nullptr) 5410 { 5411 if (version == 1) 5412 { 5413 uint32_t strsize = size - sizeof (uint32_t); 5414 char *buf = (char*) malloc (strsize); 5415 if (buf) 5416 { 5417 m_data.CopyData (offset, strsize, buf); 5418 buf[strsize - 1] = '\0'; 5419 result = buf; 5420 if (buf) 5421 free (buf); 5422 return result; 5423 } 5424 } 5425 } 5426 } 5427 } 5428 offset = cmd_offset + lc.cmdsize; 5429 } 5430 5431 // Second, make a pass over the load commands looking for an obsolete 5432 // LC_IDENT load command. 5433 offset = MachHeaderSizeFromMagic(m_header.magic); 5434 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5435 const uint32_t cmd_offset = offset; 5436 struct ident_command ident_command; 5437 if (m_data.GetU32(&offset, &ident_command, 2) == NULL) 5438 break; 5439 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5440 char *buf = (char *) malloc (ident_command.cmdsize); 5441 if (buf != nullptr 5442 && m_data.CopyData (offset, ident_command.cmdsize, buf) == ident_command.cmdsize) { 5443 buf[ident_command.cmdsize - 1] = '\0'; 5444 result = buf; 5445 } 5446 if (buf) 5447 free (buf); 5448 } 5449 offset = cmd_offset + ident_command.cmdsize; 5450 } 5451 5452 } 5453 return result; 5454 } 5455 5456 bool ObjectFileMachO::GetCorefileMainBinaryInfo (addr_t &address, UUID &uuid) { 5457 address = LLDB_INVALID_ADDRESS; 5458 uuid.Clear(); 5459 ModuleSP module_sp(GetModule()); 5460 if (module_sp) { 5461 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5462 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5463 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5464 const uint32_t cmd_offset = offset; 5465 load_command lc; 5466 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5467 break; 5468 if (lc.cmd == LC_NOTE) 5469 { 5470 char data_owner[17]; 5471 memset (data_owner, 0, sizeof (data_owner)); 5472 m_data.CopyData (offset, 16, data_owner); 5473 offset += 16; 5474 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5475 uint64_t size = m_data.GetU64_unchecked (&offset); 5476 5477 // "main bin spec" (main binary specification) data payload is 5478 // formatted: 5479 // uint32_t version [currently 1] 5480 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user process] 5481 // uint64_t address [ UINT64_MAX if address not specified ] 5482 // uuid_t uuid [ all zero's if uuid not specified ] 5483 // uint32_t log2_pagesize [ process page size in log base 2, e.g. 4k pages are 12. 0 for unspecified ] 5484 5485 if (strcmp ("main bin spec", data_owner) == 0 && size >= 32) 5486 { 5487 offset = fileoff; 5488 uint32_t version; 5489 if (m_data.GetU32 (&offset, &version, 1) != nullptr && version == 1) 5490 { 5491 uint32_t type = 0; 5492 uuid_t raw_uuid; 5493 memset (raw_uuid, 0, sizeof (uuid_t)); 5494 5495 if (m_data.GetU32(&offset, &type, 1) && 5496 m_data.GetU64(&offset, &address, 1) && 5497 m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) { 5498 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t)); 5499 return true; 5500 } 5501 } 5502 } 5503 } 5504 offset = cmd_offset + lc.cmdsize; 5505 } 5506 } 5507 return false; 5508 } 5509 5510 lldb::RegisterContextSP 5511 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5512 lldb_private::Thread &thread) { 5513 lldb::RegisterContextSP reg_ctx_sp; 5514 5515 ModuleSP module_sp(GetModule()); 5516 if (module_sp) { 5517 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5518 if (!m_thread_context_offsets_valid) 5519 GetNumThreadContexts(); 5520 5521 const FileRangeArray::Entry *thread_context_file_range = 5522 m_thread_context_offsets.GetEntryAtIndex(idx); 5523 if (thread_context_file_range) { 5524 5525 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5526 thread_context_file_range->GetByteSize()); 5527 5528 switch (m_header.cputype) { 5529 case llvm::MachO::CPU_TYPE_ARM64: 5530 reg_ctx_sp = 5531 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data); 5532 break; 5533 5534 case llvm::MachO::CPU_TYPE_ARM: 5535 reg_ctx_sp = 5536 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data); 5537 break; 5538 5539 case llvm::MachO::CPU_TYPE_I386: 5540 reg_ctx_sp = 5541 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data); 5542 break; 5543 5544 case llvm::MachO::CPU_TYPE_X86_64: 5545 reg_ctx_sp = 5546 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data); 5547 break; 5548 } 5549 } 5550 } 5551 return reg_ctx_sp; 5552 } 5553 5554 ObjectFile::Type ObjectFileMachO::CalculateType() { 5555 switch (m_header.filetype) { 5556 case MH_OBJECT: // 0x1u 5557 if (GetAddressByteSize() == 4) { 5558 // 32 bit kexts are just object files, but they do have a valid 5559 // UUID load command. 5560 if (GetUUID()) { 5561 // this checking for the UUID load command is not enough we could 5562 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5563 // this is required of kexts 5564 if (m_strata == eStrataInvalid) 5565 m_strata = eStrataKernel; 5566 return eTypeSharedLibrary; 5567 } 5568 } 5569 return eTypeObjectFile; 5570 5571 case MH_EXECUTE: 5572 return eTypeExecutable; // 0x2u 5573 case MH_FVMLIB: 5574 return eTypeSharedLibrary; // 0x3u 5575 case MH_CORE: 5576 return eTypeCoreFile; // 0x4u 5577 case MH_PRELOAD: 5578 return eTypeSharedLibrary; // 0x5u 5579 case MH_DYLIB: 5580 return eTypeSharedLibrary; // 0x6u 5581 case MH_DYLINKER: 5582 return eTypeDynamicLinker; // 0x7u 5583 case MH_BUNDLE: 5584 return eTypeSharedLibrary; // 0x8u 5585 case MH_DYLIB_STUB: 5586 return eTypeStubLibrary; // 0x9u 5587 case MH_DSYM: 5588 return eTypeDebugInfo; // 0xAu 5589 case MH_KEXT_BUNDLE: 5590 return eTypeSharedLibrary; // 0xBu 5591 default: 5592 break; 5593 } 5594 return eTypeUnknown; 5595 } 5596 5597 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5598 switch (m_header.filetype) { 5599 case MH_OBJECT: // 0x1u 5600 { 5601 // 32 bit kexts are just object files, but they do have a valid 5602 // UUID load command. 5603 if (GetUUID()) { 5604 // this checking for the UUID load command is not enough we could 5605 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5606 // this is required of kexts 5607 if (m_type == eTypeInvalid) 5608 m_type = eTypeSharedLibrary; 5609 5610 return eStrataKernel; 5611 } 5612 } 5613 return eStrataUnknown; 5614 5615 case MH_EXECUTE: // 0x2u 5616 // Check for the MH_DYLDLINK bit in the flags 5617 if (m_header.flags & MH_DYLDLINK) { 5618 return eStrataUser; 5619 } else { 5620 SectionList *section_list = GetSectionList(); 5621 if (section_list) { 5622 static ConstString g_kld_section_name("__KLD"); 5623 if (section_list->FindSectionByName(g_kld_section_name)) 5624 return eStrataKernel; 5625 } 5626 } 5627 return eStrataRawImage; 5628 5629 case MH_FVMLIB: 5630 return eStrataUser; // 0x3u 5631 case MH_CORE: 5632 return eStrataUnknown; // 0x4u 5633 case MH_PRELOAD: 5634 return eStrataRawImage; // 0x5u 5635 case MH_DYLIB: 5636 return eStrataUser; // 0x6u 5637 case MH_DYLINKER: 5638 return eStrataUser; // 0x7u 5639 case MH_BUNDLE: 5640 return eStrataUser; // 0x8u 5641 case MH_DYLIB_STUB: 5642 return eStrataUser; // 0x9u 5643 case MH_DSYM: 5644 return eStrataUnknown; // 0xAu 5645 case MH_KEXT_BUNDLE: 5646 return eStrataKernel; // 0xBu 5647 default: 5648 break; 5649 } 5650 return eStrataUnknown; 5651 } 5652 5653 llvm::VersionTuple ObjectFileMachO::GetVersion() { 5654 ModuleSP module_sp(GetModule()); 5655 if (module_sp) { 5656 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5657 struct dylib_command load_cmd; 5658 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5659 uint32_t version_cmd = 0; 5660 uint64_t version = 0; 5661 uint32_t i; 5662 for (i = 0; i < m_header.ncmds; ++i) { 5663 const lldb::offset_t cmd_offset = offset; 5664 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5665 break; 5666 5667 if (load_cmd.cmd == LC_ID_DYLIB) { 5668 if (version_cmd == 0) { 5669 version_cmd = load_cmd.cmd; 5670 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5671 break; 5672 version = load_cmd.dylib.current_version; 5673 } 5674 break; // Break for now unless there is another more complete version 5675 // number load command in the future. 5676 } 5677 offset = cmd_offset + load_cmd.cmdsize; 5678 } 5679 5680 if (version_cmd == LC_ID_DYLIB) { 5681 unsigned major = (version & 0xFFFF0000ull) >> 16; 5682 unsigned minor = (version & 0x0000FF00ull) >> 8; 5683 unsigned subminor = (version & 0x000000FFull); 5684 return llvm::VersionTuple(major, minor, subminor); 5685 } 5686 } 5687 return llvm::VersionTuple(); 5688 } 5689 5690 ArchSpec ObjectFileMachO::GetArchitecture() { 5691 ModuleSP module_sp(GetModule()); 5692 ArchSpec arch; 5693 if (module_sp) { 5694 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5695 5696 return GetArchitecture(m_header, m_data, 5697 MachHeaderSizeFromMagic(m_header.magic)); 5698 } 5699 return arch; 5700 } 5701 5702 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, addr_t &base_addr, UUID &uuid) { 5703 uuid.Clear(); 5704 base_addr = LLDB_INVALID_ADDRESS; 5705 if (process && process->GetDynamicLoader()) { 5706 DynamicLoader *dl = process->GetDynamicLoader(); 5707 LazyBool using_shared_cache; 5708 LazyBool private_shared_cache; 5709 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5710 private_shared_cache); 5711 } 5712 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5713 if (log) 5714 log->Printf("inferior process shared cache has a UUID of %s, base address 0x%" PRIx64 , uuid.GetAsString().c_str(), base_addr); 5715 } 5716 5717 // From dyld SPI header dyld_process_info.h 5718 typedef void *dyld_process_info; 5719 struct lldb_copy__dyld_process_cache_info { 5720 uuid_t cacheUUID; // UUID of cache used by process 5721 uint64_t cacheBaseAddress; // load address of dyld shared cache 5722 bool noCache; // process is running without a dyld cache 5723 bool privateCache; // process is using a private copy of its dyld cache 5724 }; 5725 5726 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with llvm 5727 // enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile errors. 5728 // So we need to use the actual underlying types of task_t and kern_return_t 5729 // below. 5730 extern "C" unsigned int /*task_t*/ mach_task_self(); 5731 5732 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5733 uuid.Clear(); 5734 base_addr = LLDB_INVALID_ADDRESS; 5735 5736 #if defined(__APPLE__) && \ 5737 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5738 uint8_t *(*dyld_get_all_image_infos)(void); 5739 dyld_get_all_image_infos = 5740 (uint8_t * (*)())dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5741 if (dyld_get_all_image_infos) { 5742 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5743 if (dyld_all_image_infos_address) { 5744 uint32_t *version = (uint32_t *) 5745 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5746 if (*version >= 13) { 5747 uuid_t *sharedCacheUUID_address = 0; 5748 int wordsize = sizeof(uint8_t *); 5749 if (wordsize == 8) { 5750 sharedCacheUUID_address = 5751 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5752 160); // sharedCacheUUID <mach-o/dyld_images.h> 5753 if (*version >= 15) 5754 base_addr = *(uint64_t *) ((uint8_t *) dyld_all_image_infos_address 5755 + 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5756 } else { 5757 sharedCacheUUID_address = 5758 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5759 84); // sharedCacheUUID <mach-o/dyld_images.h> 5760 if (*version >= 15) { 5761 base_addr = 0; 5762 base_addr = *(uint32_t *) ((uint8_t *) dyld_all_image_infos_address 5763 + 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5764 } 5765 } 5766 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t)); 5767 } 5768 } 5769 } else { 5770 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5771 dyld_process_info (*dyld_process_info_create)(unsigned int /* task_t */ task, uint64_t timestamp, unsigned int /*kern_return_t*/ *kernelError); 5772 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5773 void (*dyld_process_info_release)(dyld_process_info info); 5774 5775 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, unsigned int /*kern_return_t*/ *)) 5776 dlsym (RTLD_DEFAULT, "_dyld_process_info_create"); 5777 dyld_process_info_get_cache = (void (*)(void *, void *)) 5778 dlsym (RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5779 dyld_process_info_release = (void (*)(void *)) 5780 dlsym (RTLD_DEFAULT, "_dyld_process_info_release"); 5781 5782 if (dyld_process_info_create && dyld_process_info_get_cache) { 5783 unsigned int /*kern_return_t */ kern_ret; 5784 dyld_process_info process_info = dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5785 if (process_info) { 5786 struct lldb_copy__dyld_process_cache_info sc_info; 5787 memset (&sc_info, 0, sizeof (struct lldb_copy__dyld_process_cache_info)); 5788 dyld_process_info_get_cache (process_info, &sc_info); 5789 if (sc_info.cacheBaseAddress != 0) { 5790 base_addr = sc_info.cacheBaseAddress; 5791 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t)); 5792 } 5793 dyld_process_info_release (process_info); 5794 } 5795 } 5796 } 5797 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5798 if (log && uuid.IsValid()) 5799 log->Printf("lldb's in-memory shared cache has a UUID of %s base address of 0x%" PRIx64, uuid.GetAsString().c_str(), base_addr); 5800 #endif 5801 } 5802 5803 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { 5804 if (!m_min_os_version) { 5805 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5806 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5807 const lldb::offset_t load_cmd_offset = offset; 5808 5809 version_min_command lc; 5810 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5811 break; 5812 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5813 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5814 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5815 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5816 if (m_data.GetU32(&offset, &lc.version, 5817 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5818 const uint32_t xxxx = lc.version >> 16; 5819 const uint32_t yy = (lc.version >> 8) & 0xffu; 5820 const uint32_t zz = lc.version & 0xffu; 5821 if (xxxx) { 5822 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5823 break; 5824 } 5825 } 5826 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5827 // struct build_version_command { 5828 // uint32_t cmd; /* LC_BUILD_VERSION */ 5829 // uint32_t cmdsize; /* sizeof(struct build_version_command) plus */ 5830 // /* ntools * sizeof(struct build_tool_version) */ 5831 // uint32_t platform; /* platform */ 5832 // uint32_t minos; /* X.Y.Z is encoded in nibbles xxxx.yy.zz */ 5833 // uint32_t sdk; /* X.Y.Z is encoded in nibbles xxxx.yy.zz */ 5834 // uint32_t ntools; /* number of tool entries following this */ 5835 // }; 5836 5837 offset += 4; // skip platform 5838 uint32_t minos = m_data.GetU32(&offset); 5839 5840 const uint32_t xxxx = minos >> 16; 5841 const uint32_t yy = (minos >> 8) & 0xffu; 5842 const uint32_t zz = minos & 0xffu; 5843 if (xxxx) { 5844 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz); 5845 break; 5846 } 5847 } 5848 5849 offset = load_cmd_offset + lc.cmdsize; 5850 } 5851 5852 if (!m_min_os_version) { 5853 // Set version to an empty value so we don't keep trying to 5854 m_min_os_version = llvm::VersionTuple(); 5855 } 5856 } 5857 5858 return *m_min_os_version; 5859 } 5860 5861 uint32_t ObjectFileMachO::GetSDKVersion(uint32_t *versions, 5862 uint32_t num_versions) { 5863 if (m_sdk_versions.empty()) { 5864 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5865 bool success = false; 5866 for (uint32_t i = 0; !success && i < m_header.ncmds; ++i) { 5867 const lldb::offset_t load_cmd_offset = offset; 5868 5869 version_min_command lc; 5870 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5871 break; 5872 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5873 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5874 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5875 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5876 if (m_data.GetU32(&offset, &lc.version, 5877 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5878 const uint32_t xxxx = lc.sdk >> 16; 5879 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5880 const uint32_t zz = lc.sdk & 0xffu; 5881 if (xxxx) { 5882 m_sdk_versions.push_back(xxxx); 5883 m_sdk_versions.push_back(yy); 5884 m_sdk_versions.push_back(zz); 5885 success = true; 5886 } else { 5887 GetModule()->ReportWarning( 5888 "minimum OS version load command with invalid (0) version found."); 5889 } 5890 } 5891 } 5892 offset = load_cmd_offset + lc.cmdsize; 5893 } 5894 5895 if (!success) { 5896 offset = MachHeaderSizeFromMagic(m_header.magic); 5897 for (uint32_t i = 0; !success && i < m_header.ncmds; ++i) { 5898 const lldb::offset_t load_cmd_offset = offset; 5899 5900 version_min_command lc; 5901 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5902 break; 5903 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { 5904 // struct build_version_command { 5905 // uint32_t cmd; /* LC_BUILD_VERSION */ 5906 // uint32_t cmdsize; /* sizeof(struct 5907 // build_version_command) plus */ 5908 // /* ntools * sizeof(struct 5909 // build_tool_version) */ 5910 // uint32_t platform; /* platform */ 5911 // uint32_t minos; /* X.Y.Z is encoded in nibbles 5912 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded 5913 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number 5914 // of tool entries following this */ 5915 // }; 5916 5917 offset += 4; // skip platform 5918 uint32_t minos = m_data.GetU32(&offset); 5919 5920 const uint32_t xxxx = minos >> 16; 5921 const uint32_t yy = (minos >> 8) & 0xffu; 5922 const uint32_t zz = minos & 0xffu; 5923 if (xxxx) { 5924 m_sdk_versions.push_back(xxxx); 5925 m_sdk_versions.push_back(yy); 5926 m_sdk_versions.push_back(zz); 5927 success = true; 5928 } 5929 } 5930 offset = load_cmd_offset + lc.cmdsize; 5931 } 5932 } 5933 5934 if (!success) { 5935 // Push an invalid value so we don't try to find 5936 // the version # again on the next call to this 5937 // method. 5938 m_sdk_versions.push_back(UINT32_MAX); 5939 } 5940 } 5941 5942 // Legitimate version numbers will have 3 entries pushed 5943 // on to m_sdk_versions. If we only have one value, it's 5944 // the sentinel value indicating that this object file 5945 // does not have a valid minimum os version #. 5946 if (m_sdk_versions.size() > 1) { 5947 if (versions != NULL && num_versions > 0) { 5948 for (size_t i = 0; i < num_versions; ++i) { 5949 if (i < m_sdk_versions.size()) 5950 versions[i] = m_sdk_versions[i]; 5951 else 5952 versions[i] = 0; 5953 } 5954 } 5955 return m_sdk_versions.size(); 5956 } 5957 // Call the superclasses version that will empty out the data 5958 return ObjectFile::GetSDKVersion(versions, num_versions); 5959 } 5960 5961 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5962 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5963 } 5964 5965 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5966 return m_allow_assembly_emulation_unwind_plans; 5967 } 5968 5969 //------------------------------------------------------------------ 5970 // PluginInterface protocol 5971 //------------------------------------------------------------------ 5972 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5973 return GetPluginNameStatic(); 5974 } 5975 5976 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5977 5978 Section *ObjectFileMachO::GetMachHeaderSection() { 5979 // Find the first address of the mach header which is the first non-zero file 5980 // sized section whose file offset is zero. This is the base file address of 5981 // the mach-o file which can be subtracted from the vmaddr of the other 5982 // segments found in memory and added to the load address 5983 ModuleSP module_sp = GetModule(); 5984 if (!module_sp) 5985 return nullptr; 5986 SectionList *section_list = GetSectionList(); 5987 if (!section_list) 5988 return nullptr; 5989 const size_t num_sections = section_list->GetSize(); 5990 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5991 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5992 if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) 5993 return section; 5994 } 5995 return nullptr; 5996 } 5997 5998 bool ObjectFileMachO::SectionIsLoadable(const Section *section) { 5999 if (!section) 6000 return false; 6001 const bool is_dsym = (m_header.filetype == MH_DSYM); 6002 if (section->GetFileSize() == 0 && !is_dsym) 6003 return false; 6004 if (section->IsThreadSpecific()) 6005 return false; 6006 if (GetModule().get() != section->GetModule().get()) 6007 return false; 6008 // Be careful with __LINKEDIT and __DWARF segments 6009 if (section->GetName() == GetSegmentNameLINKEDIT() || 6010 section->GetName() == GetSegmentNameDWARF()) { 6011 // Only map __LINKEDIT and __DWARF if we have an in memory image and 6012 // this isn't a kernel binary like a kext or mach_kernel. 6013 const bool is_memory_image = (bool)m_process_wp.lock(); 6014 const Strata strata = GetStrata(); 6015 if (is_memory_image == false || strata == eStrataKernel) 6016 return false; 6017 } 6018 return true; 6019 } 6020 6021 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 6022 lldb::addr_t header_load_address, const Section *header_section, 6023 const Section *section) { 6024 ModuleSP module_sp = GetModule(); 6025 if (module_sp && header_section && section && 6026 header_load_address != LLDB_INVALID_ADDRESS) { 6027 lldb::addr_t file_addr = header_section->GetFileAddress(); 6028 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) 6029 return section->GetFileAddress() - file_addr + header_load_address; 6030 } 6031 return LLDB_INVALID_ADDRESS; 6032 } 6033 6034 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 6035 bool value_is_offset) { 6036 ModuleSP module_sp = GetModule(); 6037 if (module_sp) { 6038 size_t num_loaded_sections = 0; 6039 SectionList *section_list = GetSectionList(); 6040 if (section_list) { 6041 const size_t num_sections = section_list->GetSize(); 6042 6043 if (value_is_offset) { 6044 // "value" is an offset to apply to each top level segment 6045 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6046 // Iterate through the object file sections to find all of the 6047 // sections that size on disk (to avoid __PAGEZERO) and load them 6048 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6049 if (SectionIsLoadable(section_sp.get())) 6050 if (target.GetSectionLoadList().SetSectionLoadAddress( 6051 section_sp, section_sp->GetFileAddress() + value)) 6052 ++num_loaded_sections; 6053 } 6054 } else { 6055 // "value" is the new base address of the mach_header, adjust each 6056 // section accordingly 6057 6058 Section *mach_header_section = GetMachHeaderSection(); 6059 if (mach_header_section) { 6060 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 6061 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 6062 6063 lldb::addr_t section_load_addr = 6064 CalculateSectionLoadAddressForMemoryImage( 6065 value, mach_header_section, section_sp.get()); 6066 if (section_load_addr != LLDB_INVALID_ADDRESS) { 6067 if (target.GetSectionLoadList().SetSectionLoadAddress( 6068 section_sp, section_load_addr)) 6069 ++num_loaded_sections; 6070 } 6071 } 6072 } 6073 } 6074 } 6075 return num_loaded_sections > 0; 6076 } 6077 return false; 6078 } 6079 6080 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 6081 const FileSpec &outfile, Status &error) { 6082 if (process_sp) { 6083 Target &target = process_sp->GetTarget(); 6084 const ArchSpec target_arch = target.GetArchitecture(); 6085 const llvm::Triple &target_triple = target_arch.GetTriple(); 6086 if (target_triple.getVendor() == llvm::Triple::Apple && 6087 (target_triple.getOS() == llvm::Triple::MacOSX || 6088 target_triple.getOS() == llvm::Triple::IOS || 6089 target_triple.getOS() == llvm::Triple::WatchOS || 6090 target_triple.getOS() == llvm::Triple::TvOS)) { 6091 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) { 6092 bool make_core = false; 6093 switch (target_arch.GetMachine()) { 6094 case llvm::Triple::aarch64: 6095 case llvm::Triple::arm: 6096 case llvm::Triple::thumb: 6097 case llvm::Triple::x86: 6098 case llvm::Triple::x86_64: 6099 make_core = true; 6100 break; 6101 default: 6102 error.SetErrorStringWithFormat("unsupported core architecture: %s", 6103 target_triple.str().c_str()); 6104 break; 6105 } 6106 6107 if (make_core) { 6108 std::vector<segment_command_64> segment_load_commands; 6109 // uint32_t range_info_idx = 0; 6110 MemoryRegionInfo range_info; 6111 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6112 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6113 const ByteOrder byte_order = target_arch.GetByteOrder(); 6114 if (range_error.Success()) { 6115 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6116 const addr_t addr = range_info.GetRange().GetRangeBase(); 6117 const addr_t size = range_info.GetRange().GetByteSize(); 6118 6119 if (size == 0) 6120 break; 6121 6122 // Calculate correct protections 6123 uint32_t prot = 0; 6124 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6125 prot |= VM_PROT_READ; 6126 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6127 prot |= VM_PROT_WRITE; 6128 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6129 prot |= VM_PROT_EXECUTE; 6130 6131 // printf ("[%3u] [0x%16.16" PRIx64 " - 6132 // 0x%16.16" PRIx64 ") %c%c%c\n", 6133 // range_info_idx, 6134 // addr, 6135 // size, 6136 // (prot & VM_PROT_READ ) ? 'r' : 6137 // '-', 6138 // (prot & VM_PROT_WRITE ) ? 'w' : 6139 // '-', 6140 // (prot & VM_PROT_EXECUTE) ? 'x' : 6141 // '-'); 6142 6143 if (prot != 0) { 6144 uint32_t cmd_type = LC_SEGMENT_64; 6145 uint32_t segment_size = sizeof(segment_command_64); 6146 if (addr_byte_size == 4) { 6147 cmd_type = LC_SEGMENT; 6148 segment_size = sizeof(segment_command); 6149 } 6150 segment_command_64 segment = { 6151 cmd_type, // uint32_t cmd; 6152 segment_size, // uint32_t cmdsize; 6153 {0}, // char segname[16]; 6154 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6155 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6156 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6157 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6158 prot, // uint32_t maxprot; 6159 prot, // uint32_t initprot; 6160 0, // uint32_t nsects; 6161 0}; // uint32_t flags; 6162 segment_load_commands.push_back(segment); 6163 } else { 6164 // No protections and a size of 1 used to be returned from old 6165 // debugservers when we asked about a region that was past the 6166 // last memory region and it indicates the end... 6167 if (size == 1) 6168 break; 6169 } 6170 6171 range_error = process_sp->GetMemoryRegionInfo( 6172 range_info.GetRange().GetRangeEnd(), range_info); 6173 if (range_error.Fail()) 6174 break; 6175 } 6176 6177 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6178 6179 mach_header_64 mach_header; 6180 if (addr_byte_size == 8) { 6181 mach_header.magic = MH_MAGIC_64; 6182 } else { 6183 mach_header.magic = MH_MAGIC; 6184 } 6185 mach_header.cputype = target_arch.GetMachOCPUType(); 6186 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6187 mach_header.filetype = MH_CORE; 6188 mach_header.ncmds = segment_load_commands.size(); 6189 mach_header.flags = 0; 6190 mach_header.reserved = 0; 6191 ThreadList &thread_list = process_sp->GetThreadList(); 6192 const uint32_t num_threads = thread_list.GetSize(); 6193 6194 // Make an array of LC_THREAD data items. Each one contains the 6195 // contents of the LC_THREAD load command. The data doesn't contain 6196 // the load command + load command size, we will add the load command 6197 // and load command size as we emit the data. 6198 std::vector<StreamString> LC_THREAD_datas(num_threads); 6199 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6200 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6201 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6202 LC_THREAD_data.SetByteOrder(byte_order); 6203 } 6204 for (uint32_t thread_idx = 0; thread_idx < num_threads; 6205 ++thread_idx) { 6206 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6207 if (thread_sp) { 6208 switch (mach_header.cputype) { 6209 case llvm::MachO::CPU_TYPE_ARM64: 6210 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6211 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6212 break; 6213 6214 case llvm::MachO::CPU_TYPE_ARM: 6215 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6216 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6217 break; 6218 6219 case llvm::MachO::CPU_TYPE_I386: 6220 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6221 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6222 break; 6223 6224 case llvm::MachO::CPU_TYPE_X86_64: 6225 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6226 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6227 break; 6228 } 6229 } 6230 } 6231 6232 // The size of the load command is the size of the segments... 6233 if (addr_byte_size == 8) { 6234 mach_header.sizeofcmds = segment_load_commands.size() * 6235 sizeof(struct segment_command_64); 6236 } else { 6237 mach_header.sizeofcmds = 6238 segment_load_commands.size() * sizeof(struct segment_command); 6239 } 6240 6241 // and the size of all LC_THREAD load command 6242 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6243 ++mach_header.ncmds; 6244 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6245 } 6246 6247 printf("mach_header: 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x " 6248 "0x%8.8x 0x%8.8x\n", 6249 mach_header.magic, mach_header.cputype, mach_header.cpusubtype, 6250 mach_header.filetype, mach_header.ncmds, 6251 mach_header.sizeofcmds, mach_header.flags, 6252 mach_header.reserved); 6253 6254 // Write the mach header 6255 buffer.PutHex32(mach_header.magic); 6256 buffer.PutHex32(mach_header.cputype); 6257 buffer.PutHex32(mach_header.cpusubtype); 6258 buffer.PutHex32(mach_header.filetype); 6259 buffer.PutHex32(mach_header.ncmds); 6260 buffer.PutHex32(mach_header.sizeofcmds); 6261 buffer.PutHex32(mach_header.flags); 6262 if (addr_byte_size == 8) { 6263 buffer.PutHex32(mach_header.reserved); 6264 } 6265 6266 // Skip the mach header and all load commands and align to the next 6267 // 0x1000 byte boundary 6268 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6269 if (file_offset & 0x00000fff) { 6270 file_offset += 0x00001000ull; 6271 file_offset &= (~0x00001000ull + 1); 6272 } 6273 6274 for (auto &segment : segment_load_commands) { 6275 segment.fileoff = file_offset; 6276 file_offset += segment.filesize; 6277 } 6278 6279 // Write out all of the LC_THREAD load commands 6280 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6281 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6282 buffer.PutHex32(LC_THREAD); 6283 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6284 buffer.Write(LC_THREAD_data.GetString().data(), 6285 LC_THREAD_data_size); 6286 } 6287 6288 // Write out all of the segment load commands 6289 for (const auto &segment : segment_load_commands) { 6290 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6291 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6292 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6293 segment.cmd, segment.cmdsize, segment.vmaddr, 6294 segment.vmaddr + segment.vmsize, segment.fileoff, 6295 segment.filesize, segment.maxprot, segment.initprot, 6296 segment.nsects, segment.flags); 6297 6298 buffer.PutHex32(segment.cmd); 6299 buffer.PutHex32(segment.cmdsize); 6300 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6301 if (addr_byte_size == 8) { 6302 buffer.PutHex64(segment.vmaddr); 6303 buffer.PutHex64(segment.vmsize); 6304 buffer.PutHex64(segment.fileoff); 6305 buffer.PutHex64(segment.filesize); 6306 } else { 6307 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6308 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6309 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6310 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6311 } 6312 buffer.PutHex32(segment.maxprot); 6313 buffer.PutHex32(segment.initprot); 6314 buffer.PutHex32(segment.nsects); 6315 buffer.PutHex32(segment.flags); 6316 } 6317 6318 File core_file; 6319 std::string core_file_path(outfile.GetPath()); 6320 error = FileSystem::Instance().Open(core_file, outfile, 6321 File::eOpenOptionWrite | 6322 File::eOpenOptionTruncate | 6323 File::eOpenOptionCanCreate); 6324 if (error.Success()) { 6325 // Read 1 page at a time 6326 uint8_t bytes[0x1000]; 6327 // Write the mach header and load commands out to the core file 6328 size_t bytes_written = buffer.GetString().size(); 6329 error = core_file.Write(buffer.GetString().data(), bytes_written); 6330 if (error.Success()) { 6331 // Now write the file data for all memory segments in the process 6332 for (const auto &segment : segment_load_commands) { 6333 if (core_file.SeekFromStart(segment.fileoff) == -1) { 6334 error.SetErrorStringWithFormat( 6335 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6336 segment.fileoff, core_file_path.c_str()); 6337 break; 6338 } 6339 6340 printf("Saving %" PRId64 6341 " bytes of data for memory region at 0x%" PRIx64 "\n", 6342 segment.vmsize, segment.vmaddr); 6343 addr_t bytes_left = segment.vmsize; 6344 addr_t addr = segment.vmaddr; 6345 Status memory_read_error; 6346 while (bytes_left > 0 && error.Success()) { 6347 const size_t bytes_to_read = 6348 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6349 const size_t bytes_read = process_sp->ReadMemory( 6350 addr, bytes, bytes_to_read, memory_read_error); 6351 if (bytes_read == bytes_to_read) { 6352 size_t bytes_written = bytes_read; 6353 error = core_file.Write(bytes, bytes_written); 6354 bytes_left -= bytes_read; 6355 addr += bytes_read; 6356 } else { 6357 // Some pages within regions are not readable, those should 6358 // be zero filled 6359 memset(bytes, 0, bytes_to_read); 6360 size_t bytes_written = bytes_to_read; 6361 error = core_file.Write(bytes, bytes_written); 6362 bytes_left -= bytes_to_read; 6363 addr += bytes_to_read; 6364 } 6365 } 6366 } 6367 } 6368 } 6369 } else { 6370 error.SetErrorString( 6371 "process doesn't support getting memory region info"); 6372 } 6373 } 6374 return true; // This is the right plug to handle saving core files for 6375 // this process 6376 } 6377 } 6378 return false; 6379 } 6380