1 //===- ELF.cpp - ELF object file implementation ---------------------------===//
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
7 //===----------------------------------------------------------------------===//
9 #include "llvm/Object/ELF.h"
10 #include "llvm/BinaryFormat/ELF.h"
11 #include "llvm/Support/LEB128.h"
14 using namespace object;
16 #define STRINGIFY_ENUM_CASE(ns, name) \
20 #define ELF_RELOC(name, value) STRINGIFY_ENUM_CASE(ELF, name)
22 StringRef llvm::object::getELFRelocationTypeName(uint32_t Machine,
27 #include "llvm/BinaryFormat/ELFRelocs/x86_64.def"
35 #include "llvm/BinaryFormat/ELFRelocs/i386.def"
42 #include "llvm/BinaryFormat/ELFRelocs/Mips.def"
49 #include "llvm/BinaryFormat/ELFRelocs/AArch64.def"
56 #include "llvm/BinaryFormat/ELFRelocs/ARM.def"
61 case ELF::EM_ARC_COMPACT:
62 case ELF::EM_ARC_COMPACT2:
64 #include "llvm/BinaryFormat/ELFRelocs/ARC.def"
71 #include "llvm/BinaryFormat/ELFRelocs/AVR.def"
78 #include "llvm/BinaryFormat/ELFRelocs/Hexagon.def"
85 #include "llvm/BinaryFormat/ELFRelocs/Lanai.def"
92 #include "llvm/BinaryFormat/ELFRelocs/PowerPC.def"
99 #include "llvm/BinaryFormat/ELFRelocs/PowerPC64.def"
106 #include "llvm/BinaryFormat/ELFRelocs/RISCV.def"
113 #include "llvm/BinaryFormat/ELFRelocs/SystemZ.def"
119 case ELF::EM_SPARC32PLUS:
120 case ELF::EM_SPARCV9:
122 #include "llvm/BinaryFormat/ELFRelocs/Sparc.def"
129 #include "llvm/BinaryFormat/ELFRelocs/AMDGPU.def"
136 #include "llvm/BinaryFormat/ELFRelocs/BPF.def"
143 #include "llvm/BinaryFormat/ELFRelocs/MSP430.def"
156 uint32_t llvm::object::getELFRelativeRelocationType(uint32_t Machine) {
159 return ELF::R_X86_64_RELATIVE;
162 return ELF::R_386_RELATIVE;
165 case ELF::EM_AARCH64:
166 return ELF::R_AARCH64_RELATIVE;
168 return ELF::R_ARM_RELATIVE;
169 case ELF::EM_ARC_COMPACT:
170 case ELF::EM_ARC_COMPACT2:
171 return ELF::R_ARC_RELATIVE;
174 case ELF::EM_HEXAGON:
175 return ELF::R_HEX_RELATIVE;
181 return ELF::R_PPC64_RELATIVE;
183 return ELF::R_RISCV_RELATIVE;
185 return ELF::R_390_RELATIVE;
187 case ELF::EM_SPARC32PLUS:
188 case ELF::EM_SPARCV9:
189 return ELF::R_SPARC_RELATIVE;
200 StringRef llvm::object::getELFSectionTypeName(uint32_t Machine, unsigned Type) {
204 STRINGIFY_ENUM_CASE(ELF, SHT_ARM_EXIDX);
205 STRINGIFY_ENUM_CASE(ELF, SHT_ARM_PREEMPTMAP);
206 STRINGIFY_ENUM_CASE(ELF, SHT_ARM_ATTRIBUTES);
207 STRINGIFY_ENUM_CASE(ELF, SHT_ARM_DEBUGOVERLAY);
208 STRINGIFY_ENUM_CASE(ELF, SHT_ARM_OVERLAYSECTION);
211 case ELF::EM_HEXAGON:
212 switch (Type) { STRINGIFY_ENUM_CASE(ELF, SHT_HEX_ORDERED); }
215 switch (Type) { STRINGIFY_ENUM_CASE(ELF, SHT_X86_64_UNWIND); }
218 case ELF::EM_MIPS_RS3_LE:
220 STRINGIFY_ENUM_CASE(ELF, SHT_MIPS_REGINFO);
221 STRINGIFY_ENUM_CASE(ELF, SHT_MIPS_OPTIONS);
222 STRINGIFY_ENUM_CASE(ELF, SHT_MIPS_DWARF);
223 STRINGIFY_ENUM_CASE(ELF, SHT_MIPS_ABIFLAGS);
231 STRINGIFY_ENUM_CASE(ELF, SHT_NULL);
232 STRINGIFY_ENUM_CASE(ELF, SHT_PROGBITS);
233 STRINGIFY_ENUM_CASE(ELF, SHT_SYMTAB);
234 STRINGIFY_ENUM_CASE(ELF, SHT_STRTAB);
235 STRINGIFY_ENUM_CASE(ELF, SHT_RELA);
236 STRINGIFY_ENUM_CASE(ELF, SHT_HASH);
237 STRINGIFY_ENUM_CASE(ELF, SHT_DYNAMIC);
238 STRINGIFY_ENUM_CASE(ELF, SHT_NOTE);
239 STRINGIFY_ENUM_CASE(ELF, SHT_NOBITS);
240 STRINGIFY_ENUM_CASE(ELF, SHT_REL);
241 STRINGIFY_ENUM_CASE(ELF, SHT_SHLIB);
242 STRINGIFY_ENUM_CASE(ELF, SHT_DYNSYM);
243 STRINGIFY_ENUM_CASE(ELF, SHT_INIT_ARRAY);
244 STRINGIFY_ENUM_CASE(ELF, SHT_FINI_ARRAY);
245 STRINGIFY_ENUM_CASE(ELF, SHT_PREINIT_ARRAY);
246 STRINGIFY_ENUM_CASE(ELF, SHT_GROUP);
247 STRINGIFY_ENUM_CASE(ELF, SHT_SYMTAB_SHNDX);
248 STRINGIFY_ENUM_CASE(ELF, SHT_RELR);
249 STRINGIFY_ENUM_CASE(ELF, SHT_ANDROID_REL);
250 STRINGIFY_ENUM_CASE(ELF, SHT_ANDROID_RELA);
251 STRINGIFY_ENUM_CASE(ELF, SHT_ANDROID_RELR);
252 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_ODRTAB);
253 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_LINKER_OPTIONS);
254 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_CALL_GRAPH_PROFILE);
255 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_ADDRSIG);
256 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_DEPENDENT_LIBRARIES);
257 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_SYMPART);
258 STRINGIFY_ENUM_CASE(ELF, SHT_GNU_ATTRIBUTES);
259 STRINGIFY_ENUM_CASE(ELF, SHT_GNU_HASH);
260 STRINGIFY_ENUM_CASE(ELF, SHT_GNU_verdef);
261 STRINGIFY_ENUM_CASE(ELF, SHT_GNU_verneed);
262 STRINGIFY_ENUM_CASE(ELF, SHT_GNU_versym);
268 template <class ELFT>
269 Expected<std::vector<typename ELFT::Rela>>
270 ELFFile<ELFT>::decode_relrs(Elf_Relr_Range relrs) const {
271 // This function decodes the contents of an SHT_RELR packed relocation
274 // Proposal for adding SHT_RELR sections to generic-abi is here:
275 // https://groups.google.com/forum/#!topic/generic-abi/bX460iggiKg
277 // The encoded sequence of Elf64_Relr entries in a SHT_RELR section looks
278 // like [ AAAAAAAA BBBBBBB1 BBBBBBB1 ... AAAAAAAA BBBBBB1 ... ]
280 // i.e. start with an address, followed by any number of bitmaps. The address
281 // entry encodes 1 relocation. The subsequent bitmap entries encode up to 63
282 // relocations each, at subsequent offsets following the last address entry.
284 // The bitmap entries must have 1 in the least significant bit. The assumption
285 // here is that an address cannot have 1 in lsb. Odd addresses are not
288 // Excluding the least significant bit in the bitmap, each non-zero bit in
289 // the bitmap represents a relocation to be applied to a corresponding machine
290 // word that follows the base address word. The second least significant bit
291 // represents the machine word immediately following the initial address, and
292 // each bit that follows represents the next word, in linear order. As such,
293 // a single bitmap can encode up to 31 relocations in a 32-bit object, and
294 // 63 relocations in a 64-bit object.
296 // This encoding has a couple of interesting properties:
297 // 1. Looking at any entry, it is clear whether it's an address or a bitmap:
298 // even means address, odd means bitmap.
299 // 2. Just a simple list of addresses is a valid encoding.
304 Rela.setType(getRelativeRelocationType(), false);
305 std::vector<Elf_Rela> Relocs;
307 // Word type: uint32_t for Elf32, and uint64_t for Elf64.
308 typedef typename ELFT::uint Word;
310 // Word size in number of bytes.
311 const size_t WordSize = sizeof(Word);
313 // Number of bits used for the relocation offsets bitmap.
314 // These many relative relocations can be encoded in a single entry.
315 const size_t NBits = 8*WordSize - 1;
318 for (const Elf_Relr &R : relrs) {
320 if ((Entry&1) == 0) {
321 // Even entry: encodes the offset for next relocation.
322 Rela.r_offset = Entry;
323 Relocs.push_back(Rela);
324 // Set base offset for subsequent bitmap entries.
325 Base = Entry + WordSize;
329 // Odd entry: encodes bitmap for relocations starting at base.
333 if ((Entry&1) != 0) {
334 Rela.r_offset = Offset;
335 Relocs.push_back(Rela);
340 // Advance base offset by NBits words.
341 Base += NBits * WordSize;
347 template <class ELFT>
348 Expected<std::vector<typename ELFT::Rela>>
349 ELFFile<ELFT>::android_relas(const Elf_Shdr *Sec) const {
350 // This function reads relocations in Android's packed relocation format,
351 // which is based on SLEB128 and delta encoding.
352 Expected<ArrayRef<uint8_t>> ContentsOrErr = getSectionContents(Sec);
354 return ContentsOrErr.takeError();
355 const uint8_t *Cur = ContentsOrErr->begin();
356 const uint8_t *End = ContentsOrErr->end();
357 if (ContentsOrErr->size() < 4 || Cur[0] != 'A' || Cur[1] != 'P' ||
358 Cur[2] != 'S' || Cur[3] != '2')
359 return createError("invalid packed relocation header");
362 const char *ErrStr = nullptr;
363 auto ReadSLEB = [&]() -> int64_t {
367 int64_t Result = decodeSLEB128(Cur, &Len, End, &ErrStr);
372 uint64_t NumRelocs = ReadSLEB();
373 uint64_t Offset = ReadSLEB();
377 return createError(ErrStr);
379 std::vector<Elf_Rela> Relocs;
380 Relocs.reserve(NumRelocs);
382 uint64_t NumRelocsInGroup = ReadSLEB();
383 if (NumRelocsInGroup > NumRelocs)
384 return createError("relocation group unexpectedly large");
385 NumRelocs -= NumRelocsInGroup;
387 uint64_t GroupFlags = ReadSLEB();
388 bool GroupedByInfo = GroupFlags & ELF::RELOCATION_GROUPED_BY_INFO_FLAG;
389 bool GroupedByOffsetDelta = GroupFlags & ELF::RELOCATION_GROUPED_BY_OFFSET_DELTA_FLAG;
390 bool GroupedByAddend = GroupFlags & ELF::RELOCATION_GROUPED_BY_ADDEND_FLAG;
391 bool GroupHasAddend = GroupFlags & ELF::RELOCATION_GROUP_HAS_ADDEND_FLAG;
393 uint64_t GroupOffsetDelta;
394 if (GroupedByOffsetDelta)
395 GroupOffsetDelta = ReadSLEB();
399 GroupRInfo = ReadSLEB();
401 if (GroupedByAddend && GroupHasAddend)
402 Addend += ReadSLEB();
407 for (uint64_t I = 0; I != NumRelocsInGroup; ++I) {
409 Offset += GroupedByOffsetDelta ? GroupOffsetDelta : ReadSLEB();
411 R.r_info = GroupedByInfo ? GroupRInfo : ReadSLEB();
412 if (GroupHasAddend && !GroupedByAddend)
413 Addend += ReadSLEB();
418 return createError(ErrStr);
422 return createError(ErrStr);
428 template <class ELFT>
429 std::string ELFFile<ELFT>::getDynamicTagAsString(unsigned Arch,
430 uint64_t Type) const {
431 #define DYNAMIC_STRINGIFY_ENUM(tag, value) \
435 #define DYNAMIC_TAG(n, v)
437 case ELF::EM_AARCH64:
439 #define AARCH64_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
440 #include "llvm/BinaryFormat/DynamicTags.def"
441 #undef AARCH64_DYNAMIC_TAG
445 case ELF::EM_HEXAGON:
447 #define HEXAGON_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
448 #include "llvm/BinaryFormat/DynamicTags.def"
449 #undef HEXAGON_DYNAMIC_TAG
455 #define MIPS_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
456 #include "llvm/BinaryFormat/DynamicTags.def"
457 #undef MIPS_DYNAMIC_TAG
463 #define PPC64_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
464 #include "llvm/BinaryFormat/DynamicTags.def"
465 #undef PPC64_DYNAMIC_TAG
471 // Now handle all dynamic tags except the architecture specific ones
472 #define AARCH64_DYNAMIC_TAG(name, value)
473 #define MIPS_DYNAMIC_TAG(name, value)
474 #define HEXAGON_DYNAMIC_TAG(name, value)
475 #define PPC64_DYNAMIC_TAG(name, value)
476 // Also ignore marker tags such as DT_HIOS (maps to DT_VERNEEDNUM), etc.
477 #define DYNAMIC_TAG_MARKER(name, value)
478 #define DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
479 #include "llvm/BinaryFormat/DynamicTags.def"
481 #undef AARCH64_DYNAMIC_TAG
482 #undef MIPS_DYNAMIC_TAG
483 #undef HEXAGON_DYNAMIC_TAG
484 #undef PPC64_DYNAMIC_TAG
485 #undef DYNAMIC_TAG_MARKER
486 #undef DYNAMIC_STRINGIFY_ENUM
488 return "<unknown:>0x" + utohexstr(Type, true);
492 template <class ELFT>
493 std::string ELFFile<ELFT>::getDynamicTagAsString(uint64_t Type) const {
494 return getDynamicTagAsString(getHeader()->e_machine, Type);
497 template <class ELFT>
498 Expected<typename ELFT::DynRange> ELFFile<ELFT>::dynamicEntries() const {
499 ArrayRef<Elf_Dyn> Dyn;
500 size_t DynSecSize = 0;
502 auto ProgramHeadersOrError = program_headers();
503 if (!ProgramHeadersOrError)
504 return ProgramHeadersOrError.takeError();
506 for (const Elf_Phdr &Phdr : *ProgramHeadersOrError) {
507 if (Phdr.p_type == ELF::PT_DYNAMIC) {
509 reinterpret_cast<const Elf_Dyn *>(base() + Phdr.p_offset),
510 Phdr.p_filesz / sizeof(Elf_Dyn));
511 DynSecSize = Phdr.p_filesz;
516 // If we can't find the dynamic section in the program headers, we just fall
517 // back on the sections.
519 auto SectionsOrError = sections();
520 if (!SectionsOrError)
521 return SectionsOrError.takeError();
523 for (const Elf_Shdr &Sec : *SectionsOrError) {
524 if (Sec.sh_type == ELF::SHT_DYNAMIC) {
525 Expected<ArrayRef<Elf_Dyn>> DynOrError =
526 getSectionContentsAsArray<Elf_Dyn>(&Sec);
528 return DynOrError.takeError();
530 DynSecSize = Sec.sh_size;
536 return ArrayRef<Elf_Dyn>();
540 // TODO: this error is untested.
541 return createError("invalid empty dynamic section");
543 if (DynSecSize % sizeof(Elf_Dyn) != 0)
544 // TODO: this error is untested.
545 return createError("malformed dynamic section");
547 if (Dyn.back().d_tag != ELF::DT_NULL)
548 // TODO: this error is untested.
549 return createError("dynamic sections must be DT_NULL terminated");
554 template <class ELFT>
555 Expected<const uint8_t *> ELFFile<ELFT>::toMappedAddr(uint64_t VAddr) const {
556 auto ProgramHeadersOrError = program_headers();
557 if (!ProgramHeadersOrError)
558 return ProgramHeadersOrError.takeError();
560 llvm::SmallVector<Elf_Phdr *, 4> LoadSegments;
562 for (const Elf_Phdr &Phdr : *ProgramHeadersOrError)
563 if (Phdr.p_type == ELF::PT_LOAD)
564 LoadSegments.push_back(const_cast<Elf_Phdr *>(&Phdr));
566 const Elf_Phdr *const *I =
567 std::upper_bound(LoadSegments.begin(), LoadSegments.end(), VAddr,
568 [](uint64_t VAddr, const Elf_Phdr_Impl<ELFT> *Phdr) {
569 return VAddr < Phdr->p_vaddr;
572 if (I == LoadSegments.begin())
573 return createError("virtual address is not in any segment: 0x" +
574 Twine::utohexstr(VAddr));
576 const Elf_Phdr &Phdr = **I;
577 uint64_t Delta = VAddr - Phdr.p_vaddr;
578 if (Delta >= Phdr.p_filesz)
579 return createError("virtual address is not in any segment: 0x" +
580 Twine::utohexstr(VAddr));
581 return base() + Phdr.p_offset + Delta;
584 template class llvm::object::ELFFile<ELF32LE>;
585 template class llvm::object::ELFFile<ELF32BE>;
586 template class llvm::object::ELFFile<ELF64LE>;
587 template class llvm::object::ELFFile<ELF64BE>;