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/DataExtractor.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/M68k.def"
34 #include "llvm/BinaryFormat/ELFRelocs/x86_64.def"
42 #include "llvm/BinaryFormat/ELFRelocs/i386.def"
49 #include "llvm/BinaryFormat/ELFRelocs/Mips.def"
56 #include "llvm/BinaryFormat/ELFRelocs/AArch64.def"
63 #include "llvm/BinaryFormat/ELFRelocs/ARM.def"
68 case ELF::EM_ARC_COMPACT
:
69 case ELF::EM_ARC_COMPACT2
:
71 #include "llvm/BinaryFormat/ELFRelocs/ARC.def"
78 #include "llvm/BinaryFormat/ELFRelocs/AVR.def"
85 #include "llvm/BinaryFormat/ELFRelocs/Hexagon.def"
92 #include "llvm/BinaryFormat/ELFRelocs/Lanai.def"
99 #include "llvm/BinaryFormat/ELFRelocs/PowerPC.def"
106 #include "llvm/BinaryFormat/ELFRelocs/PowerPC64.def"
113 #include "llvm/BinaryFormat/ELFRelocs/RISCV.def"
120 #include "llvm/BinaryFormat/ELFRelocs/SystemZ.def"
126 case ELF::EM_SPARC32PLUS
:
127 case ELF::EM_SPARCV9
:
129 #include "llvm/BinaryFormat/ELFRelocs/Sparc.def"
136 #include "llvm/BinaryFormat/ELFRelocs/AMDGPU.def"
143 #include "llvm/BinaryFormat/ELFRelocs/BPF.def"
150 #include "llvm/BinaryFormat/ELFRelocs/MSP430.def"
157 #include "llvm/BinaryFormat/ELFRelocs/VE.def"
164 #include "llvm/BinaryFormat/ELFRelocs/CSKY.def"
177 uint32_t llvm::object::getELFRelativeRelocationType(uint32_t Machine
) {
180 return ELF::R_X86_64_RELATIVE
;
183 return ELF::R_386_RELATIVE
;
186 case ELF::EM_AARCH64
:
187 return ELF::R_AARCH64_RELATIVE
;
189 return ELF::R_ARM_RELATIVE
;
190 case ELF::EM_ARC_COMPACT
:
191 case ELF::EM_ARC_COMPACT2
:
192 return ELF::R_ARC_RELATIVE
;
195 case ELF::EM_HEXAGON
:
196 return ELF::R_HEX_RELATIVE
;
202 return ELF::R_PPC64_RELATIVE
;
204 return ELF::R_RISCV_RELATIVE
;
206 return ELF::R_390_RELATIVE
;
208 case ELF::EM_SPARC32PLUS
:
209 case ELF::EM_SPARCV9
:
210 return ELF::R_SPARC_RELATIVE
;
212 return ELF::R_CKCORE_RELATIVE
;
223 StringRef
llvm::object::getELFSectionTypeName(uint32_t Machine
, unsigned Type
) {
227 STRINGIFY_ENUM_CASE(ELF
, SHT_ARM_EXIDX
);
228 STRINGIFY_ENUM_CASE(ELF
, SHT_ARM_PREEMPTMAP
);
229 STRINGIFY_ENUM_CASE(ELF
, SHT_ARM_ATTRIBUTES
);
230 STRINGIFY_ENUM_CASE(ELF
, SHT_ARM_DEBUGOVERLAY
);
231 STRINGIFY_ENUM_CASE(ELF
, SHT_ARM_OVERLAYSECTION
);
234 case ELF::EM_HEXAGON
:
235 switch (Type
) { STRINGIFY_ENUM_CASE(ELF
, SHT_HEX_ORDERED
); }
238 switch (Type
) { STRINGIFY_ENUM_CASE(ELF
, SHT_X86_64_UNWIND
); }
241 case ELF::EM_MIPS_RS3_LE
:
243 STRINGIFY_ENUM_CASE(ELF
, SHT_MIPS_REGINFO
);
244 STRINGIFY_ENUM_CASE(ELF
, SHT_MIPS_OPTIONS
);
245 STRINGIFY_ENUM_CASE(ELF
, SHT_MIPS_DWARF
);
246 STRINGIFY_ENUM_CASE(ELF
, SHT_MIPS_ABIFLAGS
);
250 switch (Type
) { STRINGIFY_ENUM_CASE(ELF
, SHT_RISCV_ATTRIBUTES
); }
257 STRINGIFY_ENUM_CASE(ELF
, SHT_NULL
);
258 STRINGIFY_ENUM_CASE(ELF
, SHT_PROGBITS
);
259 STRINGIFY_ENUM_CASE(ELF
, SHT_SYMTAB
);
260 STRINGIFY_ENUM_CASE(ELF
, SHT_STRTAB
);
261 STRINGIFY_ENUM_CASE(ELF
, SHT_RELA
);
262 STRINGIFY_ENUM_CASE(ELF
, SHT_HASH
);
263 STRINGIFY_ENUM_CASE(ELF
, SHT_DYNAMIC
);
264 STRINGIFY_ENUM_CASE(ELF
, SHT_NOTE
);
265 STRINGIFY_ENUM_CASE(ELF
, SHT_NOBITS
);
266 STRINGIFY_ENUM_CASE(ELF
, SHT_REL
);
267 STRINGIFY_ENUM_CASE(ELF
, SHT_SHLIB
);
268 STRINGIFY_ENUM_CASE(ELF
, SHT_DYNSYM
);
269 STRINGIFY_ENUM_CASE(ELF
, SHT_INIT_ARRAY
);
270 STRINGIFY_ENUM_CASE(ELF
, SHT_FINI_ARRAY
);
271 STRINGIFY_ENUM_CASE(ELF
, SHT_PREINIT_ARRAY
);
272 STRINGIFY_ENUM_CASE(ELF
, SHT_GROUP
);
273 STRINGIFY_ENUM_CASE(ELF
, SHT_SYMTAB_SHNDX
);
274 STRINGIFY_ENUM_CASE(ELF
, SHT_RELR
);
275 STRINGIFY_ENUM_CASE(ELF
, SHT_ANDROID_REL
);
276 STRINGIFY_ENUM_CASE(ELF
, SHT_ANDROID_RELA
);
277 STRINGIFY_ENUM_CASE(ELF
, SHT_ANDROID_RELR
);
278 STRINGIFY_ENUM_CASE(ELF
, SHT_LLVM_ODRTAB
);
279 STRINGIFY_ENUM_CASE(ELF
, SHT_LLVM_LINKER_OPTIONS
);
280 STRINGIFY_ENUM_CASE(ELF
, SHT_LLVM_CALL_GRAPH_PROFILE
);
281 STRINGIFY_ENUM_CASE(ELF
, SHT_LLVM_ADDRSIG
);
282 STRINGIFY_ENUM_CASE(ELF
, SHT_LLVM_DEPENDENT_LIBRARIES
);
283 STRINGIFY_ENUM_CASE(ELF
, SHT_LLVM_SYMPART
);
284 STRINGIFY_ENUM_CASE(ELF
, SHT_LLVM_PART_EHDR
);
285 STRINGIFY_ENUM_CASE(ELF
, SHT_LLVM_PART_PHDR
);
286 STRINGIFY_ENUM_CASE(ELF
, SHT_LLVM_BB_ADDR_MAP
);
287 STRINGIFY_ENUM_CASE(ELF
, SHT_GNU_ATTRIBUTES
);
288 STRINGIFY_ENUM_CASE(ELF
, SHT_GNU_HASH
);
289 STRINGIFY_ENUM_CASE(ELF
, SHT_GNU_verdef
);
290 STRINGIFY_ENUM_CASE(ELF
, SHT_GNU_verneed
);
291 STRINGIFY_ENUM_CASE(ELF
, SHT_GNU_versym
);
297 template <class ELFT
>
298 std::vector
<typename
ELFT::Rel
>
299 ELFFile
<ELFT
>::decode_relrs(Elf_Relr_Range relrs
) const {
300 // This function decodes the contents of an SHT_RELR packed relocation
303 // Proposal for adding SHT_RELR sections to generic-abi is here:
304 // https://groups.google.com/forum/#!topic/generic-abi/bX460iggiKg
306 // The encoded sequence of Elf64_Relr entries in a SHT_RELR section looks
307 // like [ AAAAAAAA BBBBBBB1 BBBBBBB1 ... AAAAAAAA BBBBBB1 ... ]
309 // i.e. start with an address, followed by any number of bitmaps. The address
310 // entry encodes 1 relocation. The subsequent bitmap entries encode up to 63
311 // relocations each, at subsequent offsets following the last address entry.
313 // The bitmap entries must have 1 in the least significant bit. The assumption
314 // here is that an address cannot have 1 in lsb. Odd addresses are not
317 // Excluding the least significant bit in the bitmap, each non-zero bit in
318 // the bitmap represents a relocation to be applied to a corresponding machine
319 // word that follows the base address word. The second least significant bit
320 // represents the machine word immediately following the initial address, and
321 // each bit that follows represents the next word, in linear order. As such,
322 // a single bitmap can encode up to 31 relocations in a 32-bit object, and
323 // 63 relocations in a 64-bit object.
325 // This encoding has a couple of interesting properties:
326 // 1. Looking at any entry, it is clear whether it's an address or a bitmap:
327 // even means address, odd means bitmap.
328 // 2. Just a simple list of addresses is a valid encoding.
332 Rel
.setType(getRelativeRelocationType(), false);
333 std::vector
<Elf_Rel
> Relocs
;
335 // Word type: uint32_t for Elf32, and uint64_t for Elf64.
336 typedef typename
ELFT::uint Word
;
338 // Word size in number of bytes.
339 const size_t WordSize
= sizeof(Word
);
341 // Number of bits used for the relocation offsets bitmap.
342 // These many relative relocations can be encoded in a single entry.
343 const size_t NBits
= 8*WordSize
- 1;
346 for (const Elf_Relr
&R
: relrs
) {
348 if ((Entry
&1) == 0) {
349 // Even entry: encodes the offset for next relocation.
350 Rel
.r_offset
= Entry
;
351 Relocs
.push_back(Rel
);
352 // Set base offset for subsequent bitmap entries.
353 Base
= Entry
+ WordSize
;
357 // Odd entry: encodes bitmap for relocations starting at base.
361 if ((Entry
&1) != 0) {
362 Rel
.r_offset
= Offset
;
363 Relocs
.push_back(Rel
);
368 // Advance base offset by NBits words.
369 Base
+= NBits
* WordSize
;
375 template <class ELFT
>
376 Expected
<std::vector
<typename
ELFT::Rela
>>
377 ELFFile
<ELFT
>::android_relas(const Elf_Shdr
&Sec
) const {
378 // This function reads relocations in Android's packed relocation format,
379 // which is based on SLEB128 and delta encoding.
380 Expected
<ArrayRef
<uint8_t>> ContentsOrErr
= getSectionContents(Sec
);
382 return ContentsOrErr
.takeError();
383 ArrayRef
<uint8_t> Content
= *ContentsOrErr
;
384 if (Content
.size() < 4 || Content
[0] != 'A' || Content
[1] != 'P' ||
385 Content
[2] != 'S' || Content
[3] != '2')
386 return createError("invalid packed relocation header");
387 DataExtractor
Data(Content
, isLE(), ELFT::Is64Bits
? 8 : 4);
388 DataExtractor::Cursor
Cur(/*Offset=*/4);
390 uint64_t NumRelocs
= Data
.getSLEB128(Cur
);
391 uint64_t Offset
= Data
.getSLEB128(Cur
);
395 return std::move(Cur
.takeError());
397 std::vector
<Elf_Rela
> Relocs
;
398 Relocs
.reserve(NumRelocs
);
400 uint64_t NumRelocsInGroup
= Data
.getSLEB128(Cur
);
402 return std::move(Cur
.takeError());
403 if (NumRelocsInGroup
> NumRelocs
)
404 return createError("relocation group unexpectedly large");
405 NumRelocs
-= NumRelocsInGroup
;
407 uint64_t GroupFlags
= Data
.getSLEB128(Cur
);
408 bool GroupedByInfo
= GroupFlags
& ELF::RELOCATION_GROUPED_BY_INFO_FLAG
;
409 bool GroupedByOffsetDelta
= GroupFlags
& ELF::RELOCATION_GROUPED_BY_OFFSET_DELTA_FLAG
;
410 bool GroupedByAddend
= GroupFlags
& ELF::RELOCATION_GROUPED_BY_ADDEND_FLAG
;
411 bool GroupHasAddend
= GroupFlags
& ELF::RELOCATION_GROUP_HAS_ADDEND_FLAG
;
413 uint64_t GroupOffsetDelta
;
414 if (GroupedByOffsetDelta
)
415 GroupOffsetDelta
= Data
.getSLEB128(Cur
);
419 GroupRInfo
= Data
.getSLEB128(Cur
);
421 if (GroupedByAddend
&& GroupHasAddend
)
422 Addend
+= Data
.getSLEB128(Cur
);
427 for (uint64_t I
= 0; Cur
&& I
!= NumRelocsInGroup
; ++I
) {
429 Offset
+= GroupedByOffsetDelta
? GroupOffsetDelta
: Data
.getSLEB128(Cur
);
431 R
.r_info
= GroupedByInfo
? GroupRInfo
: Data
.getSLEB128(Cur
);
432 if (GroupHasAddend
&& !GroupedByAddend
)
433 Addend
+= Data
.getSLEB128(Cur
);
438 return std::move(Cur
.takeError());
444 template <class ELFT
>
445 std::string ELFFile
<ELFT
>::getDynamicTagAsString(unsigned Arch
,
446 uint64_t Type
) const {
447 #define DYNAMIC_STRINGIFY_ENUM(tag, value) \
451 #define DYNAMIC_TAG(n, v)
453 case ELF::EM_AARCH64
:
455 #define AARCH64_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
456 #include "llvm/BinaryFormat/DynamicTags.def"
457 #undef AARCH64_DYNAMIC_TAG
461 case ELF::EM_HEXAGON
:
463 #define HEXAGON_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
464 #include "llvm/BinaryFormat/DynamicTags.def"
465 #undef HEXAGON_DYNAMIC_TAG
471 #define MIPS_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
472 #include "llvm/BinaryFormat/DynamicTags.def"
473 #undef MIPS_DYNAMIC_TAG
479 #define PPC_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
480 #include "llvm/BinaryFormat/DynamicTags.def"
481 #undef PPC_DYNAMIC_TAG
487 #define PPC64_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
488 #include "llvm/BinaryFormat/DynamicTags.def"
489 #undef PPC64_DYNAMIC_TAG
495 // Now handle all dynamic tags except the architecture specific ones
496 #define AARCH64_DYNAMIC_TAG(name, value)
497 #define MIPS_DYNAMIC_TAG(name, value)
498 #define HEXAGON_DYNAMIC_TAG(name, value)
499 #define PPC_DYNAMIC_TAG(name, value)
500 #define PPC64_DYNAMIC_TAG(name, value)
501 // Also ignore marker tags such as DT_HIOS (maps to DT_VERNEEDNUM), etc.
502 #define DYNAMIC_TAG_MARKER(name, value)
503 #define DYNAMIC_TAG(name, value) case value: return #name;
504 #include "llvm/BinaryFormat/DynamicTags.def"
506 #undef AARCH64_DYNAMIC_TAG
507 #undef MIPS_DYNAMIC_TAG
508 #undef HEXAGON_DYNAMIC_TAG
509 #undef PPC_DYNAMIC_TAG
510 #undef PPC64_DYNAMIC_TAG
511 #undef DYNAMIC_TAG_MARKER
512 #undef DYNAMIC_STRINGIFY_ENUM
514 return "<unknown:>0x" + utohexstr(Type
, true);
518 template <class ELFT
>
519 std::string ELFFile
<ELFT
>::getDynamicTagAsString(uint64_t Type
) const {
520 return getDynamicTagAsString(getHeader().e_machine
, Type
);
523 template <class ELFT
>
524 Expected
<typename
ELFT::DynRange
> ELFFile
<ELFT
>::dynamicEntries() const {
525 ArrayRef
<Elf_Dyn
> Dyn
;
527 auto ProgramHeadersOrError
= program_headers();
528 if (!ProgramHeadersOrError
)
529 return ProgramHeadersOrError
.takeError();
531 for (const Elf_Phdr
&Phdr
: *ProgramHeadersOrError
) {
532 if (Phdr
.p_type
== ELF::PT_DYNAMIC
) {
534 reinterpret_cast<const Elf_Dyn
*>(base() + Phdr
.p_offset
),
535 Phdr
.p_filesz
/ sizeof(Elf_Dyn
));
540 // If we can't find the dynamic section in the program headers, we just fall
541 // back on the sections.
543 auto SectionsOrError
= sections();
544 if (!SectionsOrError
)
545 return SectionsOrError
.takeError();
547 for (const Elf_Shdr
&Sec
: *SectionsOrError
) {
548 if (Sec
.sh_type
== ELF::SHT_DYNAMIC
) {
549 Expected
<ArrayRef
<Elf_Dyn
>> DynOrError
=
550 getSectionContentsAsArray
<Elf_Dyn
>(Sec
);
552 return DynOrError
.takeError();
559 return ArrayRef
<Elf_Dyn
>();
563 // TODO: this error is untested.
564 return createError("invalid empty dynamic section");
566 if (Dyn
.back().d_tag
!= ELF::DT_NULL
)
567 // TODO: this error is untested.
568 return createError("dynamic sections must be DT_NULL terminated");
573 template <class ELFT
>
574 Expected
<const uint8_t *>
575 ELFFile
<ELFT
>::toMappedAddr(uint64_t VAddr
, WarningHandler WarnHandler
) const {
576 auto ProgramHeadersOrError
= program_headers();
577 if (!ProgramHeadersOrError
)
578 return ProgramHeadersOrError
.takeError();
580 llvm::SmallVector
<Elf_Phdr
*, 4> LoadSegments
;
582 for (const Elf_Phdr
&Phdr
: *ProgramHeadersOrError
)
583 if (Phdr
.p_type
== ELF::PT_LOAD
)
584 LoadSegments
.push_back(const_cast<Elf_Phdr
*>(&Phdr
));
586 auto SortPred
= [](const Elf_Phdr_Impl
<ELFT
> *A
,
587 const Elf_Phdr_Impl
<ELFT
> *B
) {
588 return A
->p_vaddr
< B
->p_vaddr
;
590 if (!llvm::is_sorted(LoadSegments
, SortPred
)) {
592 WarnHandler("loadable segments are unsorted by virtual address"))
594 llvm::stable_sort(LoadSegments
, SortPred
);
597 const Elf_Phdr
*const *I
= llvm::upper_bound(
598 LoadSegments
, VAddr
, [](uint64_t VAddr
, const Elf_Phdr_Impl
<ELFT
> *Phdr
) {
599 return VAddr
< Phdr
->p_vaddr
;
602 if (I
== LoadSegments
.begin())
603 return createError("virtual address is not in any segment: 0x" +
604 Twine::utohexstr(VAddr
));
606 const Elf_Phdr
&Phdr
= **I
;
607 uint64_t Delta
= VAddr
- Phdr
.p_vaddr
;
608 if (Delta
>= Phdr
.p_filesz
)
609 return createError("virtual address is not in any segment: 0x" +
610 Twine::utohexstr(VAddr
));
612 uint64_t Offset
= Phdr
.p_offset
+ Delta
;
613 if (Offset
>= getBufSize())
614 return createError("can't map virtual address 0x" +
615 Twine::utohexstr(VAddr
) + " to the segment with index " +
616 Twine(&Phdr
- (*ProgramHeadersOrError
).data() + 1) +
617 ": the segment ends at 0x" +
618 Twine::utohexstr(Phdr
.p_offset
+ Phdr
.p_filesz
) +
619 ", which is greater than the file size (0x" +
620 Twine::utohexstr(getBufSize()) + ")");
622 return base() + Offset
;
625 template <class ELFT
>
626 Expected
<std::vector
<typename
ELFT::BBAddrMap
>>
627 ELFFile
<ELFT
>::decodeBBAddrMap(const Elf_Shdr
&Sec
) const {
628 Expected
<ArrayRef
<uint8_t>> ContentsOrErr
= getSectionContents(Sec
);
630 return ContentsOrErr
.takeError();
631 ArrayRef
<uint8_t> Content
= *ContentsOrErr
;
632 DataExtractor
Data(Content
, isLE(), ELFT::Is64Bits
? 8 : 4);
633 std::vector
<Elf_BBAddrMap
> FunctionEntries
;
635 DataExtractor::Cursor
Cur(0);
636 Error ULEBSizeErr
= Error::success();
638 // Helper to extract and decode the next ULEB128 value as uint32_t.
639 // Returns zero and sets ULEBSizeErr if the ULEB128 value exceeds the uint32_t
641 // Also returns zero if ULEBSizeErr is already in an error state.
642 auto ReadULEB128AsUInt32
= [&Data
, &Cur
, &ULEBSizeErr
]() -> uint32_t {
643 // Bail out and do not extract data if ULEBSizeErr is already set.
646 uint64_t Offset
= Cur
.tell();
647 uint64_t Value
= Data
.getULEB128(Cur
);
648 if (Value
> UINT32_MAX
) {
649 ULEBSizeErr
= createError(
650 "ULEB128 value at offset 0x" + Twine::utohexstr(Offset
) +
651 " exceeds UINT32_MAX (0x" + Twine::utohexstr(Value
) + ")");
654 return static_cast<uint32_t>(Value
);
657 while (!ULEBSizeErr
&& Cur
&& Cur
.tell() < Content
.size()) {
658 uintX_t Address
= static_cast<uintX_t
>(Data
.getAddress(Cur
));
659 uint32_t NumBlocks
= ReadULEB128AsUInt32();
660 std::vector
<typename
Elf_BBAddrMap::BBEntry
> BBEntries
;
661 for (uint32_t BlockID
= 0; !ULEBSizeErr
&& Cur
&& (BlockID
< NumBlocks
);
663 uint32_t Offset
= ReadULEB128AsUInt32();
664 uint32_t Size
= ReadULEB128AsUInt32();
665 uint32_t Metadata
= ReadULEB128AsUInt32();
666 BBEntries
.push_back({Offset
, Size
, Metadata
});
668 FunctionEntries
.push_back({Address
, BBEntries
});
670 // Either Cur is in the error state, or ULEBSizeError is set (not both), but
671 // we join the two errors here to be safe.
672 if (!Cur
|| ULEBSizeErr
)
673 return joinErrors(Cur
.takeError(), std::move(ULEBSizeErr
));
674 return FunctionEntries
;
677 template class llvm::object::ELFFile
<ELF32LE
>;
678 template class llvm::object::ELFFile
<ELF32BE
>;
679 template class llvm::object::ELFFile
<ELF64LE
>;
680 template class llvm::object::ELFFile
<ELF64BE
>;