LLVM  10.0.0svn
ELF.cpp
Go to the documentation of this file.
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 
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;
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 
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:
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 
201  switch (Machine) {
202  case ELF::EM_ARM:
203  switch (Type) {
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) {
224  }
225  break;
226  default:
227  break;
228  }
229 
230  switch (Type) {
263  default:
264  return "Unknown";
265  }
266 }
267 
268 template <class ELFT>
271  // This function decodes the contents of an SHT_RELR packed relocation
272  // section.
273  //
274  // Proposal for adding SHT_RELR sections to generic-abi is here:
275  // https://groups.google.com/forum/#!topic/generic-abi/bX460iggiKg
276  //
277  // The encoded sequence of Elf64_Relr entries in a SHT_RELR section looks
278  // like [ AAAAAAAA BBBBBBB1 BBBBBBB1 ... AAAAAAAA BBBBBB1 ... ]
279  //
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.
283  //
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
286  // supported.
287  //
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.
295  //
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.
300 
301  Elf_Rela Rela;
302  Rela.r_info = 0;
303  Rela.r_addend = 0;
304  Rela.setType(getRelativeRelocationType(), false);
305  std::vector<Elf_Rela> Relocs;
306 
307  // Word type: uint32_t for Elf32, and uint64_t for Elf64.
308  typedef typename ELFT::uint Word;
309 
310  // Word size in number of bytes.
311  const size_t WordSize = sizeof(Word);
312 
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;
316 
317  Word Base = 0;
318  for (const Elf_Relr &R : relrs) {
319  Word Entry = R;
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;
326  continue;
327  }
328 
329  // Odd entry: encodes bitmap for relocations starting at base.
330  Word Offset = Base;
331  while (Entry != 0) {
332  Entry >>= 1;
333  if ((Entry&1) != 0) {
334  Rela.r_offset = Offset;
335  Relocs.push_back(Rela);
336  }
337  Offset += WordSize;
338  }
339 
340  // Advance base offset by NBits words.
341  Base += NBits * WordSize;
342  }
343 
344  return Relocs;
345 }
346 
347 template <class ELFT>
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);
353  if (!ContentsOrErr)
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");
360  Cur += 4;
361 
362  const char *ErrStr = nullptr;
363  auto ReadSLEB = [&]() -> int64_t {
364  if (ErrStr)
365  return 0;
366  unsigned Len;
367  int64_t Result = decodeSLEB128(Cur, &Len, End, &ErrStr);
368  Cur += Len;
369  return Result;
370  };
371 
372  uint64_t NumRelocs = ReadSLEB();
373  uint64_t Offset = ReadSLEB();
374  uint64_t Addend = 0;
375 
376  if (ErrStr)
377  return createError(ErrStr);
378 
379  std::vector<Elf_Rela> Relocs;
380  Relocs.reserve(NumRelocs);
381  while (NumRelocs) {
382  uint64_t NumRelocsInGroup = ReadSLEB();
383  if (NumRelocsInGroup > NumRelocs)
384  return createError("relocation group unexpectedly large");
385  NumRelocs -= NumRelocsInGroup;
386 
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;
392 
393  uint64_t GroupOffsetDelta;
394  if (GroupedByOffsetDelta)
395  GroupOffsetDelta = ReadSLEB();
396 
397  uint64_t GroupRInfo;
398  if (GroupedByInfo)
399  GroupRInfo = ReadSLEB();
400 
401  if (GroupedByAddend && GroupHasAddend)
402  Addend += ReadSLEB();
403 
404  if (!GroupHasAddend)
405  Addend = 0;
406 
407  for (uint64_t I = 0; I != NumRelocsInGroup; ++I) {
408  Elf_Rela R;
409  Offset += GroupedByOffsetDelta ? GroupOffsetDelta : ReadSLEB();
410  R.r_offset = Offset;
411  R.r_info = GroupedByInfo ? GroupRInfo : ReadSLEB();
412  if (GroupHasAddend && !GroupedByAddend)
413  Addend += ReadSLEB();
414  R.r_addend = Addend;
415  Relocs.push_back(R);
416 
417  if (ErrStr)
418  return createError(ErrStr);
419  }
420 
421  if (ErrStr)
422  return createError(ErrStr);
423  }
424 
425  return Relocs;
426 }
427 
428 template <class ELFT>
429 std::string ELFFile<ELFT>::getDynamicTagAsString(unsigned Arch,
430  uint64_t Type) const {
431 #define DYNAMIC_STRINGIFY_ENUM(tag, value) \
432  case value: \
433  return #tag;
434 
435 #define DYNAMIC_TAG(n, v)
436  switch (Arch) {
437  case ELF::EM_AARCH64:
438  switch (Type) {
439 #define AARCH64_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
440 #include "llvm/BinaryFormat/DynamicTags.def"
441 #undef AARCH64_DYNAMIC_TAG
442  }
443  break;
444 
445  case ELF::EM_HEXAGON:
446  switch (Type) {
447 #define HEXAGON_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
448 #include "llvm/BinaryFormat/DynamicTags.def"
449 #undef HEXAGON_DYNAMIC_TAG
450  }
451  break;
452 
453  case ELF::EM_MIPS:
454  switch (Type) {
455 #define MIPS_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
456 #include "llvm/BinaryFormat/DynamicTags.def"
457 #undef MIPS_DYNAMIC_TAG
458  }
459  break;
460 
461  case ELF::EM_PPC64:
462  switch (Type) {
463 #define PPC64_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
464 #include "llvm/BinaryFormat/DynamicTags.def"
465 #undef PPC64_DYNAMIC_TAG
466  }
467  break;
468  }
469 #undef DYNAMIC_TAG
470  switch (Type) {
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"
480 #undef DYNAMIC_TAG
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
487  default:
488  return "<unknown:>0x" + utohexstr(Type, true);
489  }
490 }
491 
492 template <class ELFT>
493 std::string ELFFile<ELFT>::getDynamicTagAsString(uint64_t Type) const {
494  return getDynamicTagAsString(getHeader()->e_machine, Type);
495 }
496 
497 template <class ELFT>
499  ArrayRef<Elf_Dyn> Dyn;
500  size_t DynSecSize = 0;
501 
502  auto ProgramHeadersOrError = program_headers();
503  if (!ProgramHeadersOrError)
504  return ProgramHeadersOrError.takeError();
505 
506  for (const Elf_Phdr &Phdr : *ProgramHeadersOrError) {
507  if (Phdr.p_type == ELF::PT_DYNAMIC) {
508  Dyn = makeArrayRef(
509  reinterpret_cast<const Elf_Dyn *>(base() + Phdr.p_offset),
510  Phdr.p_filesz / sizeof(Elf_Dyn));
511  DynSecSize = Phdr.p_filesz;
512  break;
513  }
514  }
515 
516  // If we can't find the dynamic section in the program headers, we just fall
517  // back on the sections.
518  if (Dyn.empty()) {
519  auto SectionsOrError = sections();
520  if (!SectionsOrError)
521  return SectionsOrError.takeError();
522 
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);
527  if (!DynOrError)
528  return DynOrError.takeError();
529  Dyn = *DynOrError;
530  DynSecSize = Sec.sh_size;
531  break;
532  }
533  }
534 
535  if (!Dyn.data())
536  return ArrayRef<Elf_Dyn>();
537  }
538 
539  if (Dyn.empty())
540  // TODO: this error is untested.
541  return createError("invalid empty dynamic section");
542 
543  if (DynSecSize % sizeof(Elf_Dyn) != 0)
544  // TODO: this error is untested.
545  return createError("malformed dynamic section");
546 
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");
550 
551  return Dyn;
552 }
553 
554 template <class ELFT>
556  auto ProgramHeadersOrError = program_headers();
557  if (!ProgramHeadersOrError)
558  return ProgramHeadersOrError.takeError();
559 
561 
562  for (const Elf_Phdr &Phdr : *ProgramHeadersOrError)
563  if (Phdr.p_type == ELF::PT_LOAD)
564  LoadSegments.push_back(const_cast<Elf_Phdr *>(&Phdr));
565 
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;
570  });
571 
572  if (I == LoadSegments.begin())
573  return createError("virtual address is not in any segment: 0x" +
574  Twine::utohexstr(VAddr));
575  --I;
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;
582 }
583 
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>;
typename ELFT::Dyn Elf_Dyn
Definition: ELF.h:75
const T & back() const
back - Get the last element.
Definition: ArrayRef.h:157
This class represents lattice values for constants.
Definition: AllocatorList.h:23
StringRef getELFSectionTypeName(uint32_t Machine, uint32_t Type)
Expected< std::vector< Elf_Rela > > decode_relrs(Elf_Relr_Range relrs) const
Definition: ELF.cpp:270
Error takeError()
Take ownership of the stored error.
Definition: Error.h:552
Expected< const uint8_t * > toMappedAddr(uint64_t VAddr) const
Definition: ELF.cpp:555
ArrayRef< T > makeArrayRef(const T &OneElt)
Construct an ArrayRef from a single element.
Definition: ArrayRef.h:450
int64_t decodeSLEB128(const uint8_t *p, unsigned *n=nullptr, const uint8_t *end=nullptr, const char **error=nullptr)
Utility function to decode a SLEB128 value.
Definition: LEB128.h:161
support::ulittle32_t Word
Definition: IRSymtab.h:50
typename ELFT::Phdr Elf_Phdr
Definition: ELF.h:76
Tagged union holding either a T or a Error.
Definition: CachePruning.h:22
uint32_t getELFRelativeRelocationType(uint32_t Machine)
Definition: ELF.cpp:156
static Error createError(const Twine &Err)
Definition: ELF.h:47
typename ELFT::Relr Elf_Relr
Definition: ELF.h:79
ArrayRef - Represent a constant reference to an array (0 or more elements consecutively in memory)...
Definition: APInt.h:32
COFF::MachineTypes Machine
Definition: COFFYAML.cpp:365
The instances of the Type class are immutable: once they are created, they are never changed...
Definition: Type.h:45
std::string getDynamicTagAsString(unsigned Arch, uint64_t Type) const
Definition: ELF.cpp:429
typename ELFT::RelrRange Elf_Relr_Range
Definition: ELF.h:95
typename ELFT::Rela Elf_Rela
Definition: ELF.h:78
const T * data() const
Definition: ArrayRef.h:145
StringRef getELFRelocationTypeName(uint32_t Machine, uint32_t Type)
Definition: ELF.cpp:22
This is a &#39;vector&#39; (really, a variable-sized array), optimized for the case when the array is small...
Definition: SmallVector.h:837
#define STRINGIFY_ENUM_CASE(ns, name)
Definition: ELF.cpp:16
static Twine utohexstr(const uint64_t &Val)
Definition: Twine.h:387
Expected< std::vector< Elf_Rela > > android_relas(const Elf_Shdr *Sec) const
Definition: ELF.cpp:349
#define I(x, y, z)
Definition: MD5.cpp:58
Expected< Elf_Dyn_Range > dynamicEntries() const
Definition: ELF.cpp:498
StringRef - Represent a constant reference to a string, i.e.
Definition: StringRef.h:48
typename ELFT::Shdr Elf_Shdr
Definition: ELF.h:73
bool empty() const
empty - Check if the array is empty.
Definition: ArrayRef.h:143
auto upper_bound(R &&Range, T &&Value) -> decltype(adl_begin(Range))
Provide wrappers to std::upper_bound which take ranges instead of having to pass begin/end explicitly...
Definition: STLExtras.h:1303
std::string utohexstr(uint64_t X, bool LowerCase=false)
Definition: StringExtras.h:124