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
13using namespace llvm;
14using 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
22StringRef 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  default:
149    break;
150  }
151  return "Unknown";
152}
153
154#undef ELF_RELOC
155
156uint32_t llvm::object::getELFRelativeRelocationType(uint32_t Machine) {
157  switch (Machine) {
158  case ELF::EM_X86_64:
159    return ELF::R_X86_64_RELATIVE;
160  case ELF::EM_386:
161  case ELF::EM_IAMCU:
162    return ELF::R_386_RELATIVE;
163  case ELF::EM_MIPS:
164    break;
165  case ELF::EM_AARCH64:
166    return ELF::R_AARCH64_RELATIVE;
167  case ELF::EM_ARM:
168    return ELF::R_ARM_RELATIVE;
169  case ELF::EM_ARC_COMPACT:
170  case ELF::EM_ARC_COMPACT2:
171    return ELF::R_ARC_RELATIVE;
172  case ELF::EM_AVR:
173    break;
174  case ELF::EM_HEXAGON:
175    return ELF::R_HEX_RELATIVE;
176  case ELF::EM_LANAI:
177    break;
178  case ELF::EM_PPC:
179    break;
180  case ELF::EM_PPC64:
181    return ELF::R_PPC64_RELATIVE;
182  case ELF::EM_RISCV:
183    return ELF::R_RISCV_RELATIVE;
184  case ELF::EM_S390:
185    return ELF::R_390_RELATIVE;
186  case ELF::EM_SPARC:
187  case ELF::EM_SPARC32PLUS:
188  case ELF::EM_SPARCV9:
189    return ELF::R_SPARC_RELATIVE;
190  case ELF::EM_AMDGPU:
191    break;
192  case ELF::EM_BPF:
193    break;
194  default:
195    break;
196  }
197  return 0;
198}
199
200StringRef llvm::object::getELFSectionTypeName(uint32_t Machine, unsigned Type) {
201  switch (Machine) {
202  case ELF::EM_ARM:
203    switch (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);
209    }
210    break;
211  case ELF::EM_HEXAGON:
212    switch (Type) { STRINGIFY_ENUM_CASE(ELF, SHT_HEX_ORDERED); }
213    break;
214  case ELF::EM_X86_64:
215    switch (Type) { STRINGIFY_ENUM_CASE(ELF, SHT_X86_64_UNWIND); }
216    break;
217  case ELF::EM_MIPS:
218  case ELF::EM_MIPS_RS3_LE:
219    switch (Type) {
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);
224    }
225    break;
226  default:
227    break;
228  }
229
230  switch (Type) {
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_LLVM_PART_EHDR);
259    STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_PART_PHDR);
260    STRINGIFY_ENUM_CASE(ELF, SHT_GNU_ATTRIBUTES);
261    STRINGIFY_ENUM_CASE(ELF, SHT_GNU_HASH);
262    STRINGIFY_ENUM_CASE(ELF, SHT_GNU_verdef);
263    STRINGIFY_ENUM_CASE(ELF, SHT_GNU_verneed);
264    STRINGIFY_ENUM_CASE(ELF, SHT_GNU_versym);
265  default:
266    return "Unknown";
267  }
268}
269
270template <class ELFT>
271Expected<std::vector<typename ELFT::Rela>>
272ELFFile<ELFT>::decode_relrs(Elf_Relr_Range relrs) const {
273  // This function decodes the contents of an SHT_RELR packed relocation
274  // section.
275  //
276  // Proposal for adding SHT_RELR sections to generic-abi is here:
277  //   https://groups.google.com/forum/#!topic/generic-abi/bX460iggiKg
278  //
279  // The encoded sequence of Elf64_Relr entries in a SHT_RELR section looks
280  // like [ AAAAAAAA BBBBBBB1 BBBBBBB1 ... AAAAAAAA BBBBBB1 ... ]
281  //
282  // i.e. start with an address, followed by any number of bitmaps. The address
283  // entry encodes 1 relocation. The subsequent bitmap entries encode up to 63
284  // relocations each, at subsequent offsets following the last address entry.
285  //
286  // The bitmap entries must have 1 in the least significant bit. The assumption
287  // here is that an address cannot have 1 in lsb. Odd addresses are not
288  // supported.
289  //
290  // Excluding the least significant bit in the bitmap, each non-zero bit in
291  // the bitmap represents a relocation to be applied to a corresponding machine
292  // word that follows the base address word. The second least significant bit
293  // represents the machine word immediately following the initial address, and
294  // each bit that follows represents the next word, in linear order. As such,
295  // a single bitmap can encode up to 31 relocations in a 32-bit object, and
296  // 63 relocations in a 64-bit object.
297  //
298  // This encoding has a couple of interesting properties:
299  // 1. Looking at any entry, it is clear whether it's an address or a bitmap:
300  //    even means address, odd means bitmap.
301  // 2. Just a simple list of addresses is a valid encoding.
302
303  Elf_Rela Rela;
304  Rela.r_info = 0;
305  Rela.r_addend = 0;
306  Rela.setType(getRelativeRelocationType(), false);
307  std::vector<Elf_Rela> Relocs;
308
309  // Word type: uint32_t for Elf32, and uint64_t for Elf64.
310  typedef typename ELFT::uint Word;
311
312  // Word size in number of bytes.
313  const size_t WordSize = sizeof(Word);
314
315  // Number of bits used for the relocation offsets bitmap.
316  // These many relative relocations can be encoded in a single entry.
317  const size_t NBits = 8*WordSize - 1;
318
319  Word Base = 0;
320  for (const Elf_Relr &R : relrs) {
321    Word Entry = R;
322    if ((Entry&1) == 0) {
323      // Even entry: encodes the offset for next relocation.
324      Rela.r_offset = Entry;
325      Relocs.push_back(Rela);
326      // Set base offset for subsequent bitmap entries.
327      Base = Entry + WordSize;
328      continue;
329    }
330
331    // Odd entry: encodes bitmap for relocations starting at base.
332    Word Offset = Base;
333    while (Entry != 0) {
334      Entry >>= 1;
335      if ((Entry&1) != 0) {
336        Rela.r_offset = Offset;
337        Relocs.push_back(Rela);
338      }
339      Offset += WordSize;
340    }
341
342    // Advance base offset by NBits words.
343    Base += NBits * WordSize;
344  }
345
346  return Relocs;
347}
348
349template <class ELFT>
350Expected<std::vector<typename ELFT::Rela>>
351ELFFile<ELFT>::android_relas(const Elf_Shdr *Sec) const {
352  // This function reads relocations in Android's packed relocation format,
353  // which is based on SLEB128 and delta encoding.
354  Expected<ArrayRef<uint8_t>> ContentsOrErr = getSectionContents(Sec);
355  if (!ContentsOrErr)
356    return ContentsOrErr.takeError();
357  const uint8_t *Cur = ContentsOrErr->begin();
358  const uint8_t *End = ContentsOrErr->end();
359  if (ContentsOrErr->size() < 4 || Cur[0] != 'A' || Cur[1] != 'P' ||
360      Cur[2] != 'S' || Cur[3] != '2')
361    return createError("invalid packed relocation header");
362  Cur += 4;
363
364  const char *ErrStr = nullptr;
365  auto ReadSLEB = [&]() -> int64_t {
366    if (ErrStr)
367      return 0;
368    unsigned Len;
369    int64_t Result = decodeSLEB128(Cur, &Len, End, &ErrStr);
370    Cur += Len;
371    return Result;
372  };
373
374  uint64_t NumRelocs = ReadSLEB();
375  uint64_t Offset = ReadSLEB();
376  uint64_t Addend = 0;
377
378  if (ErrStr)
379    return createError(ErrStr);
380
381  std::vector<Elf_Rela> Relocs;
382  Relocs.reserve(NumRelocs);
383  while (NumRelocs) {
384    uint64_t NumRelocsInGroup = ReadSLEB();
385    if (NumRelocsInGroup > NumRelocs)
386      return createError("relocation group unexpectedly large");
387    NumRelocs -= NumRelocsInGroup;
388
389    uint64_t GroupFlags = ReadSLEB();
390    bool GroupedByInfo = GroupFlags & ELF::RELOCATION_GROUPED_BY_INFO_FLAG;
391    bool GroupedByOffsetDelta = GroupFlags & ELF::RELOCATION_GROUPED_BY_OFFSET_DELTA_FLAG;
392    bool GroupedByAddend = GroupFlags & ELF::RELOCATION_GROUPED_BY_ADDEND_FLAG;
393    bool GroupHasAddend = GroupFlags & ELF::RELOCATION_GROUP_HAS_ADDEND_FLAG;
394
395    uint64_t GroupOffsetDelta;
396    if (GroupedByOffsetDelta)
397      GroupOffsetDelta = ReadSLEB();
398
399    uint64_t GroupRInfo;
400    if (GroupedByInfo)
401      GroupRInfo = ReadSLEB();
402
403    if (GroupedByAddend && GroupHasAddend)
404      Addend += ReadSLEB();
405
406    if (!GroupHasAddend)
407      Addend = 0;
408
409    for (uint64_t I = 0; I != NumRelocsInGroup; ++I) {
410      Elf_Rela R;
411      Offset += GroupedByOffsetDelta ? GroupOffsetDelta : ReadSLEB();
412      R.r_offset = Offset;
413      R.r_info = GroupedByInfo ? GroupRInfo : ReadSLEB();
414      if (GroupHasAddend && !GroupedByAddend)
415        Addend += ReadSLEB();
416      R.r_addend = Addend;
417      Relocs.push_back(R);
418
419      if (ErrStr)
420        return createError(ErrStr);
421    }
422
423    if (ErrStr)
424      return createError(ErrStr);
425  }
426
427  return Relocs;
428}
429
430template <class ELFT>
431std::string ELFFile<ELFT>::getDynamicTagAsString(unsigned Arch,
432                                                 uint64_t Type) const {
433#define DYNAMIC_STRINGIFY_ENUM(tag, value)                                     \
434  case value:                                                                  \
435    return #tag;
436
437#define DYNAMIC_TAG(n, v)
438  switch (Arch) {
439  case ELF::EM_AARCH64:
440    switch (Type) {
441#define AARCH64_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
442#include "llvm/BinaryFormat/DynamicTags.def"
443#undef AARCH64_DYNAMIC_TAG
444    }
445    break;
446
447  case ELF::EM_HEXAGON:
448    switch (Type) {
449#define HEXAGON_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
450#include "llvm/BinaryFormat/DynamicTags.def"
451#undef HEXAGON_DYNAMIC_TAG
452    }
453    break;
454
455  case ELF::EM_MIPS:
456    switch (Type) {
457#define MIPS_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
458#include "llvm/BinaryFormat/DynamicTags.def"
459#undef MIPS_DYNAMIC_TAG
460    }
461    break;
462
463  case ELF::EM_PPC64:
464    switch (Type) {
465#define PPC64_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
466#include "llvm/BinaryFormat/DynamicTags.def"
467#undef PPC64_DYNAMIC_TAG
468    }
469    break;
470  }
471#undef DYNAMIC_TAG
472  switch (Type) {
473// Now handle all dynamic tags except the architecture specific ones
474#define AARCH64_DYNAMIC_TAG(name, value)
475#define MIPS_DYNAMIC_TAG(name, value)
476#define HEXAGON_DYNAMIC_TAG(name, value)
477#define PPC64_DYNAMIC_TAG(name, value)
478// Also ignore marker tags such as DT_HIOS (maps to DT_VERNEEDNUM), etc.
479#define DYNAMIC_TAG_MARKER(name, value)
480#define DYNAMIC_TAG(name, value) case value: return #name;
481#include "llvm/BinaryFormat/DynamicTags.def"
482#undef DYNAMIC_TAG
483#undef AARCH64_DYNAMIC_TAG
484#undef MIPS_DYNAMIC_TAG
485#undef HEXAGON_DYNAMIC_TAG
486#undef PPC64_DYNAMIC_TAG
487#undef DYNAMIC_TAG_MARKER
488#undef DYNAMIC_STRINGIFY_ENUM
489  default:
490    return "<unknown:>0x" + utohexstr(Type, true);
491  }
492}
493
494template <class ELFT>
495std::string ELFFile<ELFT>::getDynamicTagAsString(uint64_t Type) const {
496  return getDynamicTagAsString(getHeader()->e_machine, Type);
497}
498
499template <class ELFT>
500Expected<typename ELFT::DynRange> ELFFile<ELFT>::dynamicEntries() const {
501  ArrayRef<Elf_Dyn> Dyn;
502  size_t DynSecSize = 0;
503
504  auto ProgramHeadersOrError = program_headers();
505  if (!ProgramHeadersOrError)
506    return ProgramHeadersOrError.takeError();
507
508  for (const Elf_Phdr &Phdr : *ProgramHeadersOrError) {
509    if (Phdr.p_type == ELF::PT_DYNAMIC) {
510      Dyn = makeArrayRef(
511          reinterpret_cast<const Elf_Dyn *>(base() + Phdr.p_offset),
512          Phdr.p_filesz / sizeof(Elf_Dyn));
513      DynSecSize = Phdr.p_filesz;
514      break;
515    }
516  }
517
518  // If we can't find the dynamic section in the program headers, we just fall
519  // back on the sections.
520  if (Dyn.empty()) {
521    auto SectionsOrError = sections();
522    if (!SectionsOrError)
523      return SectionsOrError.takeError();
524
525    for (const Elf_Shdr &Sec : *SectionsOrError) {
526      if (Sec.sh_type == ELF::SHT_DYNAMIC) {
527        Expected<ArrayRef<Elf_Dyn>> DynOrError =
528            getSectionContentsAsArray<Elf_Dyn>(&Sec);
529        if (!DynOrError)
530          return DynOrError.takeError();
531        Dyn = *DynOrError;
532        DynSecSize = Sec.sh_size;
533        break;
534      }
535    }
536
537    if (!Dyn.data())
538      return ArrayRef<Elf_Dyn>();
539  }
540
541  if (Dyn.empty())
542    // TODO: this error is untested.
543    return createError("invalid empty dynamic section");
544
545  if (DynSecSize % sizeof(Elf_Dyn) != 0)
546    // TODO: this error is untested.
547    return createError("malformed dynamic section");
548
549  if (Dyn.back().d_tag != ELF::DT_NULL)
550    // TODO: this error is untested.
551    return createError("dynamic sections must be DT_NULL terminated");
552
553  return Dyn;
554}
555
556template <class ELFT>
557Expected<const uint8_t *> ELFFile<ELFT>::toMappedAddr(uint64_t VAddr) const {
558  auto ProgramHeadersOrError = program_headers();
559  if (!ProgramHeadersOrError)
560    return ProgramHeadersOrError.takeError();
561
562  llvm::SmallVector<Elf_Phdr *, 4> LoadSegments;
563
564  for (const Elf_Phdr &Phdr : *ProgramHeadersOrError)
565    if (Phdr.p_type == ELF::PT_LOAD)
566      LoadSegments.push_back(const_cast<Elf_Phdr *>(&Phdr));
567
568  const Elf_Phdr *const *I =
569      std::upper_bound(LoadSegments.begin(), LoadSegments.end(), VAddr,
570                       [](uint64_t VAddr, const Elf_Phdr_Impl<ELFT> *Phdr) {
571                         return VAddr < Phdr->p_vaddr;
572                       });
573
574  if (I == LoadSegments.begin())
575    return createError("virtual address is not in any segment: 0x" +
576                       Twine::utohexstr(VAddr));
577  --I;
578  const Elf_Phdr &Phdr = **I;
579  uint64_t Delta = VAddr - Phdr.p_vaddr;
580  if (Delta >= Phdr.p_filesz)
581    return createError("virtual address is not in any segment: 0x" +
582                       Twine::utohexstr(VAddr));
583  return base() + Phdr.p_offset + Delta;
584}
585
586template class llvm::object::ELFFile<ELF32LE>;
587template class llvm::object::ELFFile<ELF32BE>;
588template class llvm::object::ELFFile<ELF64LE>;
589template class llvm::object::ELFFile<ELF64BE>;
590