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_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
);
270 template <class ELFT
>
271 Expected
<std::vector
<typename
ELFT::Rela
>>
272 ELFFile
<ELFT
>::decode_relrs(Elf_Relr_Range relrs
) const {
273 // This function decodes the contents of an SHT_RELR packed relocation
276 // Proposal for adding SHT_RELR sections to generic-abi is here:
277 // https://groups.google.com/forum/#!topic/generic-abi/bX460iggiKg
279 // The encoded sequence of Elf64_Relr entries in a SHT_RELR section looks
280 // like [ AAAAAAAA BBBBBBB1 BBBBBBB1 ... AAAAAAAA BBBBBB1 ... ]
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.
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
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.
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.
306 Rela
.setType(getRelativeRelocationType(), false);
307 std::vector
<Elf_Rela
> Relocs
;
309 // Word type: uint32_t for Elf32, and uint64_t for Elf64.
310 typedef typename
ELFT::uint Word
;
312 // Word size in number of bytes.
313 const size_t WordSize
= sizeof(Word
);
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;
320 for (const Elf_Relr
&R
: relrs
) {
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
;
331 // Odd entry: encodes bitmap for relocations starting at base.
335 if ((Entry
&1) != 0) {
336 Rela
.r_offset
= Offset
;
337 Relocs
.push_back(Rela
);
342 // Advance base offset by NBits words.
343 Base
+= NBits
* WordSize
;
349 template <class ELFT
>
350 Expected
<std::vector
<typename
ELFT::Rela
>>
351 ELFFile
<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
);
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");
364 const char *ErrStr
= nullptr;
365 auto ReadSLEB
= [&]() -> int64_t {
369 int64_t Result
= decodeSLEB128(Cur
, &Len
, End
, &ErrStr
);
374 uint64_t NumRelocs
= ReadSLEB();
375 uint64_t Offset
= ReadSLEB();
379 return createError(ErrStr
);
381 std::vector
<Elf_Rela
> Relocs
;
382 Relocs
.reserve(NumRelocs
);
384 uint64_t NumRelocsInGroup
= ReadSLEB();
385 if (NumRelocsInGroup
> NumRelocs
)
386 return createError("relocation group unexpectedly large");
387 NumRelocs
-= NumRelocsInGroup
;
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
;
395 uint64_t GroupOffsetDelta
;
396 if (GroupedByOffsetDelta
)
397 GroupOffsetDelta
= ReadSLEB();
401 GroupRInfo
= ReadSLEB();
403 if (GroupedByAddend
&& GroupHasAddend
)
404 Addend
+= ReadSLEB();
409 for (uint64_t I
= 0; I
!= NumRelocsInGroup
; ++I
) {
411 Offset
+= GroupedByOffsetDelta
? GroupOffsetDelta
: ReadSLEB();
413 R
.r_info
= GroupedByInfo
? GroupRInfo
: ReadSLEB();
414 if (GroupHasAddend
&& !GroupedByAddend
)
415 Addend
+= ReadSLEB();
420 return createError(ErrStr
);
424 return createError(ErrStr
);
430 template <class ELFT
>
431 std::string ELFFile
<ELFT
>::getDynamicTagAsString(unsigned Arch
,
432 uint64_t Type
) const {
433 #define DYNAMIC_STRINGIFY_ENUM(tag, value) \
437 #define DYNAMIC_TAG(n, v)
439 case ELF::EM_AARCH64
:
441 #define AARCH64_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
442 #include "llvm/BinaryFormat/DynamicTags.def"
443 #undef AARCH64_DYNAMIC_TAG
447 case ELF::EM_HEXAGON
:
449 #define HEXAGON_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
450 #include "llvm/BinaryFormat/DynamicTags.def"
451 #undef HEXAGON_DYNAMIC_TAG
457 #define MIPS_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
458 #include "llvm/BinaryFormat/DynamicTags.def"
459 #undef MIPS_DYNAMIC_TAG
465 #define PPC64_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
466 #include "llvm/BinaryFormat/DynamicTags.def"
467 #undef PPC64_DYNAMIC_TAG
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) DYNAMIC_STRINGIFY_ENUM(name, value)
481 #include "llvm/BinaryFormat/DynamicTags.def"
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
490 return "<unknown:>0x" + utohexstr(Type
, true);
494 template <class ELFT
>
495 std::string ELFFile
<ELFT
>::getDynamicTagAsString(uint64_t Type
) const {
496 return getDynamicTagAsString(getHeader()->e_machine
, Type
);
499 template <class ELFT
>
500 Expected
<typename
ELFT::DynRange
> ELFFile
<ELFT
>::dynamicEntries() const {
501 ArrayRef
<Elf_Dyn
> Dyn
;
502 size_t DynSecSize
= 0;
504 auto ProgramHeadersOrError
= program_headers();
505 if (!ProgramHeadersOrError
)
506 return ProgramHeadersOrError
.takeError();
508 for (const Elf_Phdr
&Phdr
: *ProgramHeadersOrError
) {
509 if (Phdr
.p_type
== ELF::PT_DYNAMIC
) {
511 reinterpret_cast<const Elf_Dyn
*>(base() + Phdr
.p_offset
),
512 Phdr
.p_filesz
/ sizeof(Elf_Dyn
));
513 DynSecSize
= Phdr
.p_filesz
;
518 // If we can't find the dynamic section in the program headers, we just fall
519 // back on the sections.
521 auto SectionsOrError
= sections();
522 if (!SectionsOrError
)
523 return SectionsOrError
.takeError();
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
);
530 return DynOrError
.takeError();
532 DynSecSize
= Sec
.sh_size
;
538 return ArrayRef
<Elf_Dyn
>();
542 // TODO: this error is untested.
543 return createError("invalid empty dynamic section");
545 if (DynSecSize
% sizeof(Elf_Dyn
) != 0)
546 // TODO: this error is untested.
547 return createError("malformed dynamic section");
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");
556 template <class ELFT
>
557 Expected
<const uint8_t *> ELFFile
<ELFT
>::toMappedAddr(uint64_t VAddr
) const {
558 auto ProgramHeadersOrError
= program_headers();
559 if (!ProgramHeadersOrError
)
560 return ProgramHeadersOrError
.takeError();
562 llvm::SmallVector
<Elf_Phdr
*, 4> LoadSegments
;
564 for (const Elf_Phdr
&Phdr
: *ProgramHeadersOrError
)
565 if (Phdr
.p_type
== ELF::PT_LOAD
)
566 LoadSegments
.push_back(const_cast<Elf_Phdr
*>(&Phdr
));
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
;
574 if (I
== LoadSegments
.begin())
575 return createError("virtual address is not in any segment: 0x" +
576 Twine::utohexstr(VAddr
));
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
;
586 template class llvm::object::ELFFile
<ELF32LE
>;
587 template class llvm::object::ELFFile
<ELF32BE
>;
588 template class llvm::object::ELFFile
<ELF64LE
>;
589 template class llvm::object::ELFFile
<ELF64BE
>;