1 //===- ELF.cpp - ELF object file implementation ---------------------------===// 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/Object/ELF.h" 10 #include "llvm/BinaryFormat/ELF.h" 11 #include "llvm/Support/LEB128.h" 12 13 using namespace llvm; 14 using namespace object; 15 16 #define STRINGIFY_ENUM_CASE(ns, name) \ 17 case ns::name: \ 18 return #name; 19 20 #define ELF_RELOC(name, value) STRINGIFY_ENUM_CASE(ELF, name) 21 22 StringRef llvm::object::getELFRelocationTypeName(uint32_t Machine, 23 uint32_t Type) { 24 switch (Machine) { 25 case ELF::EM_X86_64: 26 switch (Type) { 27 #include "llvm/BinaryFormat/ELFRelocs/x86_64.def" 28 default: 29 break; 30 } 31 break; 32 case ELF::EM_386: 33 case ELF::EM_IAMCU: 34 switch (Type) { 35 #include "llvm/BinaryFormat/ELFRelocs/i386.def" 36 default: 37 break; 38 } 39 break; 40 case ELF::EM_MIPS: 41 switch (Type) { 42 #include "llvm/BinaryFormat/ELFRelocs/Mips.def" 43 default: 44 break; 45 } 46 break; 47 case ELF::EM_AARCH64: 48 switch (Type) { 49 #include "llvm/BinaryFormat/ELFRelocs/AArch64.def" 50 default: 51 break; 52 } 53 break; 54 case ELF::EM_ARM: 55 switch (Type) { 56 #include "llvm/BinaryFormat/ELFRelocs/ARM.def" 57 default: 58 break; 59 } 60 break; 61 case ELF::EM_ARC_COMPACT: 62 case ELF::EM_ARC_COMPACT2: 63 switch (Type) { 64 #include "llvm/BinaryFormat/ELFRelocs/ARC.def" 65 default: 66 break; 67 } 68 break; 69 case ELF::EM_AVR: 70 switch (Type) { 71 #include "llvm/BinaryFormat/ELFRelocs/AVR.def" 72 default: 73 break; 74 } 75 break; 76 case ELF::EM_HEXAGON: 77 switch (Type) { 78 #include "llvm/BinaryFormat/ELFRelocs/Hexagon.def" 79 default: 80 break; 81 } 82 break; 83 case ELF::EM_LANAI: 84 switch (Type) { 85 #include "llvm/BinaryFormat/ELFRelocs/Lanai.def" 86 default: 87 break; 88 } 89 break; 90 case ELF::EM_PPC: 91 switch (Type) { 92 #include "llvm/BinaryFormat/ELFRelocs/PowerPC.def" 93 default: 94 break; 95 } 96 break; 97 case ELF::EM_PPC64: 98 switch (Type) { 99 #include "llvm/BinaryFormat/ELFRelocs/PowerPC64.def" 100 default: 101 break; 102 } 103 break; 104 case ELF::EM_RISCV: 105 switch (Type) { 106 #include "llvm/BinaryFormat/ELFRelocs/RISCV.def" 107 default: 108 break; 109 } 110 break; 111 case ELF::EM_S390: 112 switch (Type) { 113 #include "llvm/BinaryFormat/ELFRelocs/SystemZ.def" 114 default: 115 break; 116 } 117 break; 118 case ELF::EM_SPARC: 119 case ELF::EM_SPARC32PLUS: 120 case ELF::EM_SPARCV9: 121 switch (Type) { 122 #include "llvm/BinaryFormat/ELFRelocs/Sparc.def" 123 default: 124 break; 125 } 126 break; 127 case ELF::EM_AMDGPU: 128 switch (Type) { 129 #include "llvm/BinaryFormat/ELFRelocs/AMDGPU.def" 130 default: 131 break; 132 } 133 break; 134 case ELF::EM_BPF: 135 switch (Type) { 136 #include "llvm/BinaryFormat/ELFRelocs/BPF.def" 137 default: 138 break; 139 } 140 break; 141 case ELF::EM_MSP430: 142 switch (Type) { 143 #include "llvm/BinaryFormat/ELFRelocs/MSP430.def" 144 default: 145 break; 146 } 147 break; 148 case ELF::EM_VE: 149 switch (Type) { 150 #include "llvm/BinaryFormat/ELFRelocs/VE.def" 151 default: 152 break; 153 } 154 break; 155 default: 156 break; 157 } 158 return "Unknown"; 159 } 160 161 #undef ELF_RELOC 162 163 uint32_t llvm::object::getELFRelativeRelocationType(uint32_t Machine) { 164 switch (Machine) { 165 case ELF::EM_X86_64: 166 return ELF::R_X86_64_RELATIVE; 167 case ELF::EM_386: 168 case ELF::EM_IAMCU: 169 return ELF::R_386_RELATIVE; 170 case ELF::EM_MIPS: 171 break; 172 case ELF::EM_AARCH64: 173 return ELF::R_AARCH64_RELATIVE; 174 case ELF::EM_ARM: 175 return ELF::R_ARM_RELATIVE; 176 case ELF::EM_ARC_COMPACT: 177 case ELF::EM_ARC_COMPACT2: 178 return ELF::R_ARC_RELATIVE; 179 case ELF::EM_AVR: 180 break; 181 case ELF::EM_HEXAGON: 182 return ELF::R_HEX_RELATIVE; 183 case ELF::EM_LANAI: 184 break; 185 case ELF::EM_PPC: 186 break; 187 case ELF::EM_PPC64: 188 return ELF::R_PPC64_RELATIVE; 189 case ELF::EM_RISCV: 190 return ELF::R_RISCV_RELATIVE; 191 case ELF::EM_S390: 192 return ELF::R_390_RELATIVE; 193 case ELF::EM_SPARC: 194 case ELF::EM_SPARC32PLUS: 195 case ELF::EM_SPARCV9: 196 return ELF::R_SPARC_RELATIVE; 197 case ELF::EM_AMDGPU: 198 break; 199 case ELF::EM_BPF: 200 break; 201 default: 202 break; 203 } 204 return 0; 205 } 206 207 StringRef llvm::object::getELFSectionTypeName(uint32_t Machine, unsigned Type) { 208 switch (Machine) { 209 case ELF::EM_ARM: 210 switch (Type) { 211 STRINGIFY_ENUM_CASE(ELF, SHT_ARM_EXIDX); 212 STRINGIFY_ENUM_CASE(ELF, SHT_ARM_PREEMPTMAP); 213 STRINGIFY_ENUM_CASE(ELF, SHT_ARM_ATTRIBUTES); 214 STRINGIFY_ENUM_CASE(ELF, SHT_ARM_DEBUGOVERLAY); 215 STRINGIFY_ENUM_CASE(ELF, SHT_ARM_OVERLAYSECTION); 216 } 217 break; 218 case ELF::EM_HEXAGON: 219 switch (Type) { STRINGIFY_ENUM_CASE(ELF, SHT_HEX_ORDERED); } 220 break; 221 case ELF::EM_X86_64: 222 switch (Type) { STRINGIFY_ENUM_CASE(ELF, SHT_X86_64_UNWIND); } 223 break; 224 case ELF::EM_MIPS: 225 case ELF::EM_MIPS_RS3_LE: 226 switch (Type) { 227 STRINGIFY_ENUM_CASE(ELF, SHT_MIPS_REGINFO); 228 STRINGIFY_ENUM_CASE(ELF, SHT_MIPS_OPTIONS); 229 STRINGIFY_ENUM_CASE(ELF, SHT_MIPS_DWARF); 230 STRINGIFY_ENUM_CASE(ELF, SHT_MIPS_ABIFLAGS); 231 } 232 break; 233 case ELF::EM_RISCV: 234 switch (Type) { STRINGIFY_ENUM_CASE(ELF, SHT_RISCV_ATTRIBUTES); } 235 break; 236 default: 237 break; 238 } 239 240 switch (Type) { 241 STRINGIFY_ENUM_CASE(ELF, SHT_NULL); 242 STRINGIFY_ENUM_CASE(ELF, SHT_PROGBITS); 243 STRINGIFY_ENUM_CASE(ELF, SHT_SYMTAB); 244 STRINGIFY_ENUM_CASE(ELF, SHT_STRTAB); 245 STRINGIFY_ENUM_CASE(ELF, SHT_RELA); 246 STRINGIFY_ENUM_CASE(ELF, SHT_HASH); 247 STRINGIFY_ENUM_CASE(ELF, SHT_DYNAMIC); 248 STRINGIFY_ENUM_CASE(ELF, SHT_NOTE); 249 STRINGIFY_ENUM_CASE(ELF, SHT_NOBITS); 250 STRINGIFY_ENUM_CASE(ELF, SHT_REL); 251 STRINGIFY_ENUM_CASE(ELF, SHT_SHLIB); 252 STRINGIFY_ENUM_CASE(ELF, SHT_DYNSYM); 253 STRINGIFY_ENUM_CASE(ELF, SHT_INIT_ARRAY); 254 STRINGIFY_ENUM_CASE(ELF, SHT_FINI_ARRAY); 255 STRINGIFY_ENUM_CASE(ELF, SHT_PREINIT_ARRAY); 256 STRINGIFY_ENUM_CASE(ELF, SHT_GROUP); 257 STRINGIFY_ENUM_CASE(ELF, SHT_SYMTAB_SHNDX); 258 STRINGIFY_ENUM_CASE(ELF, SHT_RELR); 259 STRINGIFY_ENUM_CASE(ELF, SHT_ANDROID_REL); 260 STRINGIFY_ENUM_CASE(ELF, SHT_ANDROID_RELA); 261 STRINGIFY_ENUM_CASE(ELF, SHT_ANDROID_RELR); 262 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_ODRTAB); 263 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_LINKER_OPTIONS); 264 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_CALL_GRAPH_PROFILE); 265 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_ADDRSIG); 266 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_DEPENDENT_LIBRARIES); 267 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_SYMPART); 268 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_PART_EHDR); 269 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_PART_PHDR); 270 STRINGIFY_ENUM_CASE(ELF, SHT_GNU_ATTRIBUTES); 271 STRINGIFY_ENUM_CASE(ELF, SHT_GNU_HASH); 272 STRINGIFY_ENUM_CASE(ELF, SHT_GNU_verdef); 273 STRINGIFY_ENUM_CASE(ELF, SHT_GNU_verneed); 274 STRINGIFY_ENUM_CASE(ELF, SHT_GNU_versym); 275 default: 276 return "Unknown"; 277 } 278 } 279 280 template <class ELFT> 281 std::vector<typename ELFT::Rel> 282 ELFFile<ELFT>::decode_relrs(Elf_Relr_Range relrs) const { 283 // This function decodes the contents of an SHT_RELR packed relocation 284 // section. 285 // 286 // Proposal for adding SHT_RELR sections to generic-abi is here: 287 // https://groups.google.com/forum/#!topic/generic-abi/bX460iggiKg 288 // 289 // The encoded sequence of Elf64_Relr entries in a SHT_RELR section looks 290 // like [ AAAAAAAA BBBBBBB1 BBBBBBB1 ... AAAAAAAA BBBBBB1 ... ] 291 // 292 // i.e. start with an address, followed by any number of bitmaps. The address 293 // entry encodes 1 relocation. The subsequent bitmap entries encode up to 63 294 // relocations each, at subsequent offsets following the last address entry. 295 // 296 // The bitmap entries must have 1 in the least significant bit. The assumption 297 // here is that an address cannot have 1 in lsb. Odd addresses are not 298 // supported. 299 // 300 // Excluding the least significant bit in the bitmap, each non-zero bit in 301 // the bitmap represents a relocation to be applied to a corresponding machine 302 // word that follows the base address word. The second least significant bit 303 // represents the machine word immediately following the initial address, and 304 // each bit that follows represents the next word, in linear order. As such, 305 // a single bitmap can encode up to 31 relocations in a 32-bit object, and 306 // 63 relocations in a 64-bit object. 307 // 308 // This encoding has a couple of interesting properties: 309 // 1. Looking at any entry, it is clear whether it's an address or a bitmap: 310 // even means address, odd means bitmap. 311 // 2. Just a simple list of addresses is a valid encoding. 312 313 Elf_Rel Rel; 314 Rel.r_info = 0; 315 Rel.setType(getRelativeRelocationType(), false); 316 std::vector<Elf_Rel> Relocs; 317 318 // Word type: uint32_t for Elf32, and uint64_t for Elf64. 319 typedef typename ELFT::uint Word; 320 321 // Word size in number of bytes. 322 const size_t WordSize = sizeof(Word); 323 324 // Number of bits used for the relocation offsets bitmap. 325 // These many relative relocations can be encoded in a single entry. 326 const size_t NBits = 8*WordSize - 1; 327 328 Word Base = 0; 329 for (const Elf_Relr &R : relrs) { 330 Word Entry = R; 331 if ((Entry&1) == 0) { 332 // Even entry: encodes the offset for next relocation. 333 Rel.r_offset = Entry; 334 Relocs.push_back(Rel); 335 // Set base offset for subsequent bitmap entries. 336 Base = Entry + WordSize; 337 continue; 338 } 339 340 // Odd entry: encodes bitmap for relocations starting at base. 341 Word Offset = Base; 342 while (Entry != 0) { 343 Entry >>= 1; 344 if ((Entry&1) != 0) { 345 Rel.r_offset = Offset; 346 Relocs.push_back(Rel); 347 } 348 Offset += WordSize; 349 } 350 351 // Advance base offset by NBits words. 352 Base += NBits * WordSize; 353 } 354 355 return Relocs; 356 } 357 358 template <class ELFT> 359 Expected<std::vector<typename ELFT::Rela>> 360 ELFFile<ELFT>::android_relas(const Elf_Shdr *Sec) const { 361 // This function reads relocations in Android's packed relocation format, 362 // which is based on SLEB128 and delta encoding. 363 Expected<ArrayRef<uint8_t>> ContentsOrErr = getSectionContents(Sec); 364 if (!ContentsOrErr) 365 return ContentsOrErr.takeError(); 366 const uint8_t *Cur = ContentsOrErr->begin(); 367 const uint8_t *End = ContentsOrErr->end(); 368 if (ContentsOrErr->size() < 4 || Cur[0] != 'A' || Cur[1] != 'P' || 369 Cur[2] != 'S' || Cur[3] != '2') 370 return createError("invalid packed relocation header"); 371 Cur += 4; 372 373 const char *ErrStr = nullptr; 374 auto ReadSLEB = [&]() -> int64_t { 375 if (ErrStr) 376 return 0; 377 unsigned Len; 378 int64_t Result = decodeSLEB128(Cur, &Len, End, &ErrStr); 379 Cur += Len; 380 return Result; 381 }; 382 383 uint64_t NumRelocs = ReadSLEB(); 384 uint64_t Offset = ReadSLEB(); 385 uint64_t Addend = 0; 386 387 if (ErrStr) 388 return createError(ErrStr); 389 390 std::vector<Elf_Rela> Relocs; 391 Relocs.reserve(NumRelocs); 392 while (NumRelocs) { 393 uint64_t NumRelocsInGroup = ReadSLEB(); 394 if (NumRelocsInGroup > NumRelocs) 395 return createError("relocation group unexpectedly large"); 396 NumRelocs -= NumRelocsInGroup; 397 398 uint64_t GroupFlags = ReadSLEB(); 399 bool GroupedByInfo = GroupFlags & ELF::RELOCATION_GROUPED_BY_INFO_FLAG; 400 bool GroupedByOffsetDelta = GroupFlags & ELF::RELOCATION_GROUPED_BY_OFFSET_DELTA_FLAG; 401 bool GroupedByAddend = GroupFlags & ELF::RELOCATION_GROUPED_BY_ADDEND_FLAG; 402 bool GroupHasAddend = GroupFlags & ELF::RELOCATION_GROUP_HAS_ADDEND_FLAG; 403 404 uint64_t GroupOffsetDelta; 405 if (GroupedByOffsetDelta) 406 GroupOffsetDelta = ReadSLEB(); 407 408 uint64_t GroupRInfo; 409 if (GroupedByInfo) 410 GroupRInfo = ReadSLEB(); 411 412 if (GroupedByAddend && GroupHasAddend) 413 Addend += ReadSLEB(); 414 415 if (!GroupHasAddend) 416 Addend = 0; 417 418 for (uint64_t I = 0; I != NumRelocsInGroup; ++I) { 419 Elf_Rela R; 420 Offset += GroupedByOffsetDelta ? GroupOffsetDelta : ReadSLEB(); 421 R.r_offset = Offset; 422 R.r_info = GroupedByInfo ? GroupRInfo : ReadSLEB(); 423 if (GroupHasAddend && !GroupedByAddend) 424 Addend += ReadSLEB(); 425 R.r_addend = Addend; 426 Relocs.push_back(R); 427 428 if (ErrStr) 429 return createError(ErrStr); 430 } 431 432 if (ErrStr) 433 return createError(ErrStr); 434 } 435 436 return Relocs; 437 } 438 439 template <class ELFT> 440 std::string ELFFile<ELFT>::getDynamicTagAsString(unsigned Arch, 441 uint64_t Type) const { 442 #define DYNAMIC_STRINGIFY_ENUM(tag, value) \ 443 case value: \ 444 return #tag; 445 446 #define DYNAMIC_TAG(n, v) 447 switch (Arch) { 448 case ELF::EM_AARCH64: 449 switch (Type) { 450 #define AARCH64_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value) 451 #include "llvm/BinaryFormat/DynamicTags.def" 452 #undef AARCH64_DYNAMIC_TAG 453 } 454 break; 455 456 case ELF::EM_HEXAGON: 457 switch (Type) { 458 #define HEXAGON_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value) 459 #include "llvm/BinaryFormat/DynamicTags.def" 460 #undef HEXAGON_DYNAMIC_TAG 461 } 462 break; 463 464 case ELF::EM_MIPS: 465 switch (Type) { 466 #define MIPS_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value) 467 #include "llvm/BinaryFormat/DynamicTags.def" 468 #undef MIPS_DYNAMIC_TAG 469 } 470 break; 471 472 case ELF::EM_PPC64: 473 switch (Type) { 474 #define PPC64_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value) 475 #include "llvm/BinaryFormat/DynamicTags.def" 476 #undef PPC64_DYNAMIC_TAG 477 } 478 break; 479 } 480 #undef DYNAMIC_TAG 481 switch (Type) { 482 // Now handle all dynamic tags except the architecture specific ones 483 #define AARCH64_DYNAMIC_TAG(name, value) 484 #define MIPS_DYNAMIC_TAG(name, value) 485 #define HEXAGON_DYNAMIC_TAG(name, value) 486 #define PPC64_DYNAMIC_TAG(name, value) 487 // Also ignore marker tags such as DT_HIOS (maps to DT_VERNEEDNUM), etc. 488 #define DYNAMIC_TAG_MARKER(name, value) 489 #define DYNAMIC_TAG(name, value) case value: return #name; 490 #include "llvm/BinaryFormat/DynamicTags.def" 491 #undef DYNAMIC_TAG 492 #undef AARCH64_DYNAMIC_TAG 493 #undef MIPS_DYNAMIC_TAG 494 #undef HEXAGON_DYNAMIC_TAG 495 #undef PPC64_DYNAMIC_TAG 496 #undef DYNAMIC_TAG_MARKER 497 #undef DYNAMIC_STRINGIFY_ENUM 498 default: 499 return "<unknown:>0x" + utohexstr(Type, true); 500 } 501 } 502 503 template <class ELFT> 504 std::string ELFFile<ELFT>::getDynamicTagAsString(uint64_t Type) const { 505 return getDynamicTagAsString(getHeader()->e_machine, Type); 506 } 507 508 template <class ELFT> 509 Expected<typename ELFT::DynRange> ELFFile<ELFT>::dynamicEntries() const { 510 ArrayRef<Elf_Dyn> Dyn; 511 512 auto ProgramHeadersOrError = program_headers(); 513 if (!ProgramHeadersOrError) 514 return ProgramHeadersOrError.takeError(); 515 516 for (const Elf_Phdr &Phdr : *ProgramHeadersOrError) { 517 if (Phdr.p_type == ELF::PT_DYNAMIC) { 518 Dyn = makeArrayRef( 519 reinterpret_cast<const Elf_Dyn *>(base() + Phdr.p_offset), 520 Phdr.p_filesz / sizeof(Elf_Dyn)); 521 break; 522 } 523 } 524 525 // If we can't find the dynamic section in the program headers, we just fall 526 // back on the sections. 527 if (Dyn.empty()) { 528 auto SectionsOrError = sections(); 529 if (!SectionsOrError) 530 return SectionsOrError.takeError(); 531 532 for (const Elf_Shdr &Sec : *SectionsOrError) { 533 if (Sec.sh_type == ELF::SHT_DYNAMIC) { 534 Expected<ArrayRef<Elf_Dyn>> DynOrError = 535 getSectionContentsAsArray<Elf_Dyn>(&Sec); 536 if (!DynOrError) 537 return DynOrError.takeError(); 538 Dyn = *DynOrError; 539 break; 540 } 541 } 542 543 if (!Dyn.data()) 544 return ArrayRef<Elf_Dyn>(); 545 } 546 547 if (Dyn.empty()) 548 // TODO: this error is untested. 549 return createError("invalid empty dynamic section"); 550 551 if (Dyn.back().d_tag != ELF::DT_NULL) 552 // TODO: this error is untested. 553 return createError("dynamic sections must be DT_NULL terminated"); 554 555 return Dyn; 556 } 557 558 template <class ELFT> 559 Expected<const uint8_t *> ELFFile<ELFT>::toMappedAddr(uint64_t VAddr) const { 560 auto ProgramHeadersOrError = program_headers(); 561 if (!ProgramHeadersOrError) 562 return ProgramHeadersOrError.takeError(); 563 564 llvm::SmallVector<Elf_Phdr *, 4> LoadSegments; 565 566 for (const Elf_Phdr &Phdr : *ProgramHeadersOrError) 567 if (Phdr.p_type == ELF::PT_LOAD) 568 LoadSegments.push_back(const_cast<Elf_Phdr *>(&Phdr)); 569 570 const Elf_Phdr *const *I = 571 std::upper_bound(LoadSegments.begin(), LoadSegments.end(), VAddr, 572 [](uint64_t VAddr, const Elf_Phdr_Impl<ELFT> *Phdr) { 573 return VAddr < Phdr->p_vaddr; 574 }); 575 576 if (I == LoadSegments.begin()) 577 return createError("virtual address is not in any segment: 0x" + 578 Twine::utohexstr(VAddr)); 579 --I; 580 const Elf_Phdr &Phdr = **I; 581 uint64_t Delta = VAddr - Phdr.p_vaddr; 582 if (Delta >= Phdr.p_filesz) 583 return createError("virtual address is not in any segment: 0x" + 584 Twine::utohexstr(VAddr)); 585 586 uint64_t Offset = Phdr.p_offset + Delta; 587 if (Offset >= getBufSize()) 588 return createError("can't map virtual address 0x" + 589 Twine::utohexstr(VAddr) + " to the segment with index " + 590 Twine(&Phdr - (*ProgramHeadersOrError).data() + 1) + 591 ": the segment ends at 0x" + 592 Twine::utohexstr(Phdr.p_offset + Phdr.p_filesz) + 593 ", which is greater than the file size (0x" + 594 Twine::utohexstr(getBufSize()) + ")"); 595 596 return base() + Offset; 597 } 598 599 template class llvm::object::ELFFile<ELF32LE>; 600 template class llvm::object::ELFFile<ELF32BE>; 601 template class llvm::object::ELFFile<ELF64LE>; 602 template class llvm::object::ELFFile<ELF64BE>; 603